From eb8172f3dd626345f562246b21b31830fd88d577 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 13:51:46 +0100 Subject: [PATCH 01/79] feat(write-skill): preserve context-dependent outcomes --- skills/write-skill/SKILL.md | 34 +++++++++++---------- skills/write-skill/agents/openai.yaml | 2 +- skills/write-skill/references/CRAFT.md | 36 +++++++++++------------ skills/write-skill/references/SECTIONS.md | 9 +++--- skills/write-skill/references/TESTING.md | 18 +++++++----- 5 files changed, 53 insertions(+), 46 deletions(-) diff --git a/skills/write-skill/SKILL.md b/skills/write-skill/SKILL.md index f6f1c34..e665315 100644 --- a/skills/write-skill/SKILL.md +++ b/skills/write-skill/SKILL.md @@ -1,6 +1,6 @@ --- name: write-skill -description: Creates and updates predictable agent skills through confirmed behavioural contracts, evidence-backed concepts, and forward testing. Use when authoring or revising a skill. +description: Creates and updates compact agent skills through confirmed behavioural contracts, evidence-backed choices, and fresh-agent evaluation. Use when authoring or revising a skill. metadata: invocation: user disable-model-invocation: true @@ -8,46 +8,50 @@ disable-model-invocation: true # Write Skill -**Minimalist instruction** turns confirmed behaviours into the smallest skill bundle that reliably teaches them. +Creates or updates the smallest skill bundle that reliably teaches a confirmed process while leaving project-dependent results to runtime evidence and user decisions. ## Process ### 1. Establish the behavioural contract -Inspect the request, complete target bundle, discoverable callers, and host conventions. Invoke `$elicit` and use the **main success scenario** to confirm the skill's purpose, trigger, primary behaviour, required inputs, observable result, and resource needs. Apply **YAGNI** to speculative branches: retain an exception only when representative evidence, the primary behaviour, or a necessary safety or permission boundary requires it. Existing and new skills reach one explicit behavioural contract with a dominant thread and only its material exceptions. +Inspect the request, complete target bundle, direct callers and callees, and host conventions. Invoke `$elicit` to confirm the skill's purpose, trigger, required inputs, common successful path, observable result, and resource needs. Retain an exception only when representative evidence, the primary behaviour, or a necessary safety, permission, or prerequisite boundary requires it. Existing and new skills reach one explicit behavioural contract with a dominant path and only its material exceptions. ### 2. Compare governing methodologies -Use the confirmed behavioural contract as the fixed scope and decision authority for both research passes; evidence selects how to teach the confirmed behaviour. Invoke `$research` to compare credible governing methodologies, then **design it twice**: continue until at least two credible options emerge or the evidence reaches saturation. Present the supported options, behavioural consequences, and recommendation for the user to choose. Offer only evidence-supported alternatives; when one methodology survives, compare it with a methodology-free process. The user's selection or evidence-backed absence is explicit. +Use the confirmed behavioural contract as the fixed scope and decision authority for both research passes; evidence selects how to teach the confirmed behaviour. Invoke `$research` to compare credible governing methodologies and a methodology-free process until further evidence is unlikely to change the decision set. Analyse the report against the contract. Call a recommendation strong only when one option materially fits better than every credible alternative; otherwise state that no strong recommendation exists. Present the supported options, behavioural consequences, trade-offs, and recommendation strength for the user to choose. The user's selection, including no methodology, is explicit. ### 3. Select supporting concepts -Invoke `$research` to find established principles, theories, methods, or techniques that reinforce distinct concerns without competing with the selected governor. Explain each candidate's intended behavioural effect and let the user decide; let the evidence determine the count. Every retained concept earns a distinct role. +Invoke `$research` to find established principles, theories, methods, or techniques that reinforce distinct concerns without competing with the selected methodology or replacing an evidence-backed absence. Analyse whether each name changes behaviour, overlaps another concept, or is more precise in plain language. Apply the same recommendation-strength rule, explain each candidate's behavioural job, and let the user decide. Zero concepts is valid; every retained concept earns a distinct role. ### 4. Confirm the design -Present one complete synthesis of the behavioural contract, selected concepts, structure, resources, main success scenario, material exceptions, and observable success conditions. Obtain explicit confirmation before following the remaining process. +Present one complete synthesis of the behavioural contract, selected methodology or absence, supporting concepts, structure, resources, common successful path, material exceptions, constraint levels, and observable success conditions. Obtain explicit confirmation before following the remaining process. ### 5. Write the bundle -Create or update through one path. Treat the confirmed design as closed: encode its main success scenario and retained material exceptions without adding new behaviour during drafting. Give every `SKILL.md` frontmatter, one H1, a concise introduction, and exactly one `## Process`; place a selected governing methodology in bold where it fits naturally in the introduction. Bold each supporting concept at its first behaviour-governing use. Consult [Skill Sections](references/SECTIONS.md) when invocation metadata, optional sections, or resource placement needs detail. The bundle expresses the confirmed design. +Create or update through one path. Treat the confirmed design as closed: encode its common successful path and retained material exceptions without adding behaviour during drafting. Give every `SKILL.md` frontmatter, one H1, a concise introduction, and exactly one `## Process`; when the user selected a governing methodology, place it in bold where it fits naturally in the introduction. Bold each supporting concept at its first behaviour-governing use. Consult [Skill Sections](references/SECTIONS.md) for invocation metadata, optional sections, and resource placement. The bundle expresses the confirmed design in the fewest instructions that reliably change behaviour. -### 6. Separate the concerns +### 6. Place runtime information -Apply **separation of concerns** so each step or subsection carries one coherent behavioural idea and observable outcome. Split independently actionable instructions, concepts, or completion criteria; keep sentences together when they jointly govern the same action. Consult [Skill Craft](references/CRAFT.md) when the split, vocabulary, emphasis, or disclosure boundary is unclear. Every section remains focused and substantial. +Apply **progressive disclosure** to the Agent Skills loading model: keep discovery conditions in metadata, instructions required on every invocation in `SKILL.md`, and conditional detail in task-needed resources. Link every resource beside the condition that loads it. The common path remains complete without loading an irrelevant branch. -### 7. Compress the language +### 7. Calibrate the instructions -Apply minimalist instruction and **DRY** until every remaining word changes behaviour, preserves a condition, or improves navigation. Replace explanations with canonical leading words when the agent already knows the concept and keep each meaning in one authoritative location. Use **ironic process theory** as a salience check: state the positive target behaviour and pair an essential prohibition with the safe action that satisfies it. Consult [Skill Craft](references/CRAFT.md) when negative framing or semantic duplication remains unclear. Every remaining instruction is behaviourally necessary, authoritative, and positively framed. +Apply **degrees of freedom** to each instruction and evaluation criterion. Keep project-dependent reasoning and results flexible; make required sequences, safety, permissions, fixed formats, schemas, scripts, and other fragile deterministic behaviour correspondingly exact. Give each step one coherent action and observable outcome, keep each rule in one authoritative location, state the desired action first, and pair an essential prohibition with the permitted route. Consult [Skill Craft](references/CRAFT.md) when recommendation strength, wording, structure, authority, or composition remains unclear. Every constraint is no stronger than the confirmed behaviour requires. ### 8. Validate the mechanics Run [scripts/validate-skill.js](scripts/validate-skill.js), inspect every bundled script, and execute each within a disposable filesystem using inert fixtures, isolated credentials, and an environment incapable of external mutation. When that boundary is unavailable, leave the script unexecuted and report the limitation. The mechanical contract passes within the safe execution boundary. -### 9. Forward-test the behaviour +### 9. Verify direct composition -Give a fresh agent only the finished bundle and a realistic main-success invocation, then compare its observable process and result with the confirmed contract. Add the smallest scenario for each retained material exception. Consult [Forward Testing](references/TESTING.md) when scenario selection, isolation, or pass evidence needs detail. Repair, recompress, revalidate, and retest until the skill reliably invokes the intended behaviour. +Trace every direct caller and callee across trigger, supplied inputs and authority, promised result, responsibility owner, and stopping boundary. Remove overlap, omissions, contradictions, and invented obligations. Exercise the least restrictive connected path that would expose a genuine incompatibility. Direct compositions preserve the confirmed contract. -### 10. Report the result +### 10. Evaluate the behaviour -Return the changed files, research evidence, mechanical results, forward-test scenarios and outcomes, unexecuted scripts, and remaining uncertainty. The user receives the finished bundle and evidence that its contract holds. +Give a fresh agent only the finished bundle, a realistic request, and task-local evidence. Compare its observable process and result with the confirmed contract. Judge variable results by their correctness and suitability for the supplied evidence; inspect exact content only where the contract fixes it. Evaluate the common successful path and the smallest scenario for each retained material exception. Consult [Fresh-Agent Behavioural Evaluation](references/TESTING.md) for isolation, evidence, and repair. Repair the smallest steering cause, recompress, revalidate, and reevaluate until the contract holds. + +### 11. Report the result + +Return the changed files, research evidence, mechanical and composition results, behavioural-evaluation scenarios and outcomes, unexecuted scripts, and remaining uncertainty. The user receives the finished bundle and evidence that its contract holds. diff --git a/skills/write-skill/agents/openai.yaml b/skills/write-skill/agents/openai.yaml index 2223e0f..bc4b0a4 100644 --- a/skills/write-skill/agents/openai.yaml +++ b/skills/write-skill/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Write Skill' - short_description: 'Create and update predictable agent skills' + short_description: 'Create and update compact, evidence-tested skills' policy: allow_implicit_invocation: false diff --git a/skills/write-skill/references/CRAFT.md b/skills/write-skill/references/CRAFT.md index d55e740..c718f44 100644 --- a/skills/write-skill/references/CRAFT.md +++ b/skills/write-skill/references/CRAFT.md @@ -1,45 +1,43 @@ # Skill Craft -Use this reference when a methodology, supporting concept, section boundary, leading word, or disclosure decision remains unclear. +Use this reference when recommendation strength, supporting concepts, wording, structure, authority, or direct composition remains unclear. -## Compare methodologies +## Judge recommendations -Research methods against the confirmed behaviour rather than selecting one for familiarity. Use **design it twice** to compare credible candidates through: +Research candidates against the confirmed behaviour rather than selecting one for familiarity. Compare each credible option, including no methodology, through: - the process each method would impose; - the behaviours it strengthens or suppresses; -- its fit across the main success scenario and retained material exceptions; +- its fit across the common successful path and retained material exceptions; - the local adaptations it would require; and - the evidence supporting its published meaning. -Continue until multiple credible choices emerge or further research is unlikely to change the set. Recommend the strongest fit and expose its trade-offs. When only one survives, compare it with a methodology-free process. +Stop when further evidence is unlikely to change the decision set. Call a recommendation strong only when one option materially fits the contract better than every credible alternative. Otherwise state that no strong recommendation exists. Present the evidence, adaptations, consequences, and uncertainty without manufacturing a ranking; the user decides. ## Select supporting concepts -A supporting concept earns inclusion only when it governs a concern the selected methodology leaves unresolved. Name its behavioural job in one sentence and test whether removing it changes the instructions. Use the governor and plain language alone when they already determine the behaviour; retain as many concepts as distinct concerns require. +A supporting concept earns inclusion only when its established meaning governs a distinct unresolved concern. Name its behavioural job in one sentence, test whether deleting its name changes execution, and compare it with plain language and overlapping concepts. Retain as many as distinct concerns require, including zero, and keep a short local mapping where the canonical name is broader than the required behaviour. -## Choose leading words +## Write the smallest effective bundle -Prefer the canonical name of a recognised method, principle, theory, or technique already present in the agent's knowledge. A leading word earns its place when it replaces explanation and sharpens a decision, action, or stopping condition. Remove or replace a term that forward testing shows to be decorative. +Assume the agent already knows general software practice. Retain a word only when it changes behaviour, preserves a condition, or improves navigation. Prefer a canonical term when it replaces explanation and sharpens a decision; remove it when the local rule must restate the entire useful meaning. -Bold the governing methodology naturally within the introduction when one was selected. Bold a supporting concept where it first governs behaviour, then use plain text unless renewed emphasis changes the instruction. +Bold a selected governing methodology naturally within the introduction. Bold a supporting concept where it first governs behaviour, then use plain text unless renewed emphasis changes execution. -## Separate concerns +## Shape coherent instructions Give a step or subsection one coherent behavioural idea and one observable outcome. Split it when any sentence could be acted on, tested, reordered, or completed independently. Keep supporting detail together when separating it would make the action harder to understand or create headings without meaningful content. -## Disclose depth +Match precision to the confirmed behaviour. Leave valid project-dependent choices open. Specify exact sequences, permissions, formats, assets, scripts, and other fragile or deterministic behaviour only to the degree the contract fixes them. -Keep knowledge inline when every invocation needs it to act correctly. Move a retained exception to a reference when its detail would obscure the common path. Write every pointer as the precise loading condition. Pull must-have material back inline when forward tests show that a sharper pointer still misses the behaviour. +## Preserve one authority -## Remove no-ops +Keep each rule and definition in one authoritative location. Allow deliberate summaries, emphasis, and caller-facing pointers when they navigate to that authority without becoming competing versions. Treat the environment as authoritative for readily discoverable commands, configuration, and structure; document only what inspection cannot reveal safely. -Test each sentence in isolation: would deleting it change agent behaviour, preserve a necessary condition, or impair navigation? Delete the whole sentence when the answer is no. Prefer a stronger canonical term over several weak adjectives, and prefer one checkable bound over exhortations to be careful or thorough. +## State the target behaviour -## Remove duplication +State the desired action or result first. Retain a prohibition when the excluded behaviour defines safety, permission, scope, or a deterministic contract, and pair it with the permitted route. Prefer a checkable bound over vague exhortations. -Apply **DRY** to meaning rather than tokens. Give each behaviour, rule, and definition one authoritative expression. Repeating a canonical term can focus attention; repeating its explanation inflates prominence and creates competing authorities. +## Trace direct composition -## State positive behaviour - -Use ironic process theory as a salience check during final compression. Describe the action the agent should perform, replace avoidable negative framing with that target, and reserve prohibitions for essential safety boundaries that cannot be expressed positively. Pair each retained prohibition with the safe action that satisfies it. The instruction keeps the intended behaviour most salient. +For each direct caller and callee, trace the trigger, inputs, authority, result, responsibility owner, and stopping boundary. Compare the real obligations rather than repeated wording. Remove gaps, contradictions, duplicated responsibility, and behaviour neither contract owns. Use the least restrictive check that still exposes an incompatible connection. diff --git a/skills/write-skill/references/SECTIONS.md b/skills/write-skill/references/SECTIONS.md index 572f708..bd91528 100644 --- a/skills/write-skill/references/SECTIONS.md +++ b/skills/write-skill/references/SECTIONS.md @@ -22,7 +22,7 @@ Add `agents/openai.yaml` with a human-readable `interface.display_name`, a 25– ## Introduction -The fixed spine and emphasis rules in the main workflow are authoritative. Use the introduction to explain only the selected methodology's context-specific adaptation. +The fixed spine and emphasis rules in the main workflow are authoritative. State the skill's purpose concisely. When the user selected a governing methodology, explain only its context-specific adaptation; a methodology-free skill needs no substitute label. ## Process details @@ -40,11 +40,12 @@ Rename or combine optional sections when that improves the confirmed behaviour. ## Bundled resources -Use **progressive disclosure** as an information hierarchy: +Use progressive disclosure through the Agent Skills loading hierarchy: -- Keep the common execution path in `SKILL.md`. +- Put discovery conditions in frontmatter metadata. +- Keep instructions required on every invocation in `SKILL.md`. - Put conditional or extensive runtime guidance in `references/`. - Put files consumed or copied into generated output in `assets/`. - Put deterministic, repeated, or fragile operations in `scripts/`. -Add a resource when a retained material exception or other conditional depth improves execution. Link every resource directly from `SKILL.md` beside a precise condition that tells the agent when it may help. Keep references one level deep and each meaning in one authoritative location. References contain runtime guidance rather than general concept explanations. +Add a resource only when retained conditional depth improves execution. Link every resource directly from `SKILL.md` beside the condition that loads it. Keep references one level deep and each meaning in one authoritative location. References contain runtime guidance rather than general concept explanations. diff --git a/skills/write-skill/references/TESTING.md b/skills/write-skill/references/TESTING.md index 263937e..123370d 100644 --- a/skills/write-skill/references/TESTING.md +++ b/skills/write-skill/references/TESTING.md @@ -1,26 +1,30 @@ -# Forward Testing +# Fresh-Agent Behavioural Evaluation -Use this reference when scenario selection, context isolation, or observable pass evidence needs more detail than the main success scenario. +Use this reference when scenario selection, context isolation, result variability, or observable pass evidence needs more detail than the common successful path. ## Preserve the evaluation boundary -Give the fresh agent the finished bundle and a realistic user request. Keep the intended answer, design rationale, suspected failure, and prior test output outside the evaluation context. Use an inert workspace or read-only artifacts; permit safe local writes only when the scenario requires them. A test demonstrates transferable steering when the finished bundle and task-local evidence supply the result. +Give the fresh agent the finished bundle, a realistic user request, and only task-local evidence. Keep the intended answer, design rationale, suspected failure, prior output, and hidden hints outside the evaluation context. Use an inert workspace or read-only artefacts; permit safe local writes only when the scenario requires them. An evaluation demonstrates transferable steering when the bundle and supplied evidence account for the process and result. ## Select scenarios -Run the main success scenario for every created or rewritten skill. Add the smallest scenario for a retained material exception when it distinguishes: +Run the common successful path for every created or rewritten skill. Add the smallest scenario for a retained material exception when it distinguishes: - an evidenced invocation that changes the required process or result; - a necessary safety, permission, or prerequisite boundary; - an optional reference needed by the retained exception; - a fragile script or deterministic output contract; or -- wording whose effect depends on a leading word. +- wording whose effect depends on a selected concept or precise instruction. -Prefer one scenario that distinguishes several competing behaviours when its failure remains diagnosable. The scenario set contains only confirmed behaviour or material risk rather than hypothetical combinations. +Prefer one scenario that distinguishes several competing behaviours when failure remains diagnosable. The scenario set contains only confirmed behaviour or material risk rather than quotas or hypothetical combinations. ## Define evidence -Translate the confirmed contract into observable pass conditions before reading the result. Inspect the agent's actions, resource reads, decisions, output, and postconditions. Pass only when the skill invokes the confirmed process, loads relevant context, follows every required branch, and derives its result from the evaluation boundary. +Translate the confirmed contract into observable pass conditions before reading the result. Inspect the agent's actions, resource reads, decisions, boundaries, result, and postconditions. + +For variable work, judge whether the agent followed the confirmed process, used the correct authorities and resources, respected user decisions and stopping boundaries, and produced a correct result suitable for the supplied evidence. Do not require one hidden solution, route, or phrasing. + +For fixed assets, schemas, scripts, formats, or deterministic behaviour, inspect the exact properties promised by the contract. Pass only when the degree of scrutiny matches the claim the skill makes. ## Repair the smallest cause From a485d74eff978a53367c5d3da5a554247566b31a Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 13:53:01 +0100 Subject: [PATCH 02/79] docs: add lifecycle skills implementation plan --- SKILLS_PLAN.md | 425 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 425 insertions(+) create mode 100644 SKILLS_PLAN.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md new file mode 100644 index 0000000..4bf818c --- /dev/null +++ b/SKILLS_PLAN.md @@ -0,0 +1,425 @@ +# Propulsion skills plan + +## Status + +This is the authoritative implementation brief for the Propulsion lifecycle skill set. It fixes the intended outcomes, composition, boundaries and acceptance conditions without prescribing the finished wording or detailed process of each skill. + +The existing suite may be changed or replaced where necessary. Skill authoring remains responsible for researching credible governing methodologies, allowing a methodology-free design when none fits, testing the common successful path and handling only material exceptions. + +## Objective + +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, backlog refinement, implementation, pull-request publication and review, debugging, incident response and retirement. + +Propulsion is not a workflow platform. It does not carry a product automatically through its lifecycle, maintain its own work state or require a setup framework. A user may enter at any point where the necessary inputs already exist and leave when that session's outcome is complete. + +## Design principles + +1. **One bounded session outcome.** Each lifecycle skill represents a natural working session and ends with one independently useful outcome. +2. **Narrow progressively.** Product intent narrows into a feature specification, solution design, implementation-ready tickets and, finally, one implemented body of work. +3. **Enter proportionately.** Substantial work uses the complete applicable chain. Small, understood work may begin directly with implementation. +4. **Compose within the requested outcome.** A skill may invoke another skill when its outcome is required to complete the requested session. It must not silently advance into a later lifecycle outcome. +5. **Keep authorities separate.** Product documents own intent, code and tests own implemented behaviour, Git owns change history and the task-management tool owns work state. +6. **Document what cannot be inferred safely.** Durable documents preserve intent, language, constraints and rationale without copying observable code, configuration, CI or tracker state. +7. **Use native project tools.** Propulsion does not introduce a manifest, tracker adapter, lifecycle state model or completion-horizon schema. +8. **Treat agreement as authority.** Invoking a skill authorises the mutations intrinsic to its agreed outcome. Questions resolve the outcome; they are not followed by a redundant application gate. +9. **Make quality evidence cumulative.** Implementation satisfies project checks and adds proportionate evidence without weakening existing constraints or installing unrelated infrastructure merely to make a check available. +10. **Keep the public surface clear.** Lifecycle skills are the primary entry points. Reusable utilities remain independently useful and may be composed where their explicit trigger applies. + +## Lifecycle coverage + +| Area | Purpose | +| --- | --- | +| Establish | Create the durable product and technical foundations for a new or existing system. | +| Define | Turn one product capability or enhancement into an approved statement of intended behaviour. | +| Plan | Select a buildable solution and decompose it into implementation-ready work. | +| Refine | Keep incoming and existing work current, prioritised, ready and feasible for the next delivery period. | +| Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | +| Debug and Respond | Diagnose and repair defects, stabilise active incidents and learn from them afterwards. | +| Retire | Plan the safe removal of an obsolete capability and route its execution through normal delivery. | + +These areas are entry points, not mandatory phase gates. Release, deployment and continuous production monitoring are outside Propulsion's core lifecycle. + +## Session and authority model + +Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture a missing feature specification or solution design. + +Reusable utilities are user-invokable and may be model-invoked when their documented trigger is present. A caller remains responsible for verifying the utility's result and completing its own outcome. + +Invoking a skill grants authority for its expected repository and external mutations once the skill has reached shared understanding. Material product, architecture, priority or scope questions remain with the user, but an agreed decision does not require a second permission prompt before it is applied. Work outside the invoked outcome remains unauthorised. + +Every skill: + +1. Inspects the request, repository guidance, existing artefacts and available tools before asking questions. +2. Reuses current authoritative inputs rather than recreating earlier lifecycle work. +3. Invokes required and triggered supporting skills. +4. Persists its outcome in the owning document or system. +5. Verifies the resulting artefact or external state. +6. Stops at its stated boundary and identifies the next available session without starting it automatically. + +## Documentation and tool authorities + +### Living project foundations + +| Authority | Owns | +| --- | --- | +| `PRODUCT.md` | Product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capability catalogue. | +| `CONTEXT.md` | Ubiquitous language, domain concepts, business rules and conceptual boundaries. | +| `ARCHITECTURE.md` | Enduring architecture drivers, system context, responsibilities, technology strategy, data ownership, integrations and cross-cutting technical constraints. | + +`CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-context` inline. + +Architecture documentation records boundaries, constraints and rationale that cannot be inferred reliably. It does not mirror classes, files, framework wiring, executable configuration or other implementation detail. Focused ADRs are created only when a consequential or surprising decision needs durable rationale beyond `ARCHITECTURE.md`. + +### Feature records + +A substantial feature uses the project convention where one exists, otherwise: + +```text +docs/features// +├── specification.md +└── solution-design.md +``` + +The feature specification owns externally observable intent and constraints. The solution design owns the selected system-specific implementation. A useful test is whether a statement would remain true if the feature were rebuilt using a different architecture: if so, it normally belongs in the specification. + +Small, understood changes do not require feature documents merely to enter implementation. + +### Bounded supporting records + +Research reports, architecture reviews, incident reviews and retirement plans are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. + +### Task-management tool + +The project's task-management tool must be named in `AGENTS.md`. A tracker-backed skill that cannot find this preference asks the user which tool the project uses, invokes `maintain-agents` to record it and then resumes its original outcome. + +The skill uses the selected tool's native issues, relationships, statuses, priorities, sprints, cycles, milestones or queues. If the named tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. + +## Lifecycle skill catalogue + +### Establish + +#### `define-product` + +- **Outcome:** A user-confirmed product foundation broad enough to guide later feature definition without becoming a roadmap or delivery plan. +- **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, representative repository evidence and applicable external evidence. +- **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. +- **Composition:** Requires `elicit` and `maintain-context`; conditionally invokes `research` and `prototype`. +- **Stops:** After product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capabilities are documented. It does not define technical architecture, detailed features, priorities or tickets. + +#### `define-architecture` + +- **Outcome:** A user-confirmed technical foundation capable of constraining later feature solutions without duplicating the implementation. +- **Inputs:** `PRODUCT.md`, `CONTEXT.md`, existing system evidence and material technical, organisational or regulatory constraints. +- **Output:** `ARCHITECTURE.md` and any warranted focused ADRs. +- **Composition:** Requires `elicit` and `modular-design`; conditionally invokes `research`, `prototype` and `maintain-context`. +- **Stops:** After the system context, principal boundaries, ownership, technology strategy, data, integrations and cross-cutting quality approach are established. It does not scaffold the system or create implementation work. + +### Define + +#### `specify-feature` + +- **Outcome:** One approved feature or enhancement expressed as externally observable intent. +- **Inputs:** A capability from `PRODUCT.md`, an already-elicited conversation or another agreed feature idea, plus `CONTEXT.md`, relevant current behaviour and supporting evidence. +- **Output:** `docs/features//specification.md`, or the project's equivalent. +- **Composition:** Requires `elicit` and `maintain-context`; conditionally invokes `research` and `prototype`. +- **Stops:** After the problem, actors, behaviour, rules, states, scenarios, acceptance conditions, constraints and exclusions are unambiguous. It does not select the implementation or create tickets. + +### Plan + +#### `design-feature` + +- **Outcome:** An approved, buildable technical solution for one feature specification. +- **Inputs:** The approved feature specification, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. +- **Output:** `docs/features//solution-design.md` and any warranted ADRs. +- **Composition:** Requires `modular-design`; conditionally invokes `elicit`, `research`, `prototype` and `maintain-context`. +- **Stops:** After affected boundaries, interfaces, data, integrations, security, migration, operational effects and verification seams are resolved. It does not decompose or implement the work. + +#### `create-tickets` + +- **Outcome:** An approved feature is represented by implementation-ready, dependency-aware work in the project's task-management tool. +- **Inputs:** The approved feature specification and solution design, plus relevant project guidance and current tracker state. +- **Output:** Native tracker items linked to their authoritative feature documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. +- **Composition:** Conditionally invokes `elicit` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. +- **Stops:** After the created items and relationships are read back and verified. It does not select an iteration or begin implementation. + +### Refine + +#### `triage-work` + +- **Outcome:** Each incoming request in scope is closed with an evidence-backed reason or routed to the appropriate lifecycle area. +- **Inputs:** The selected incoming items, their source evidence, current product and implementation authorities and the configured task-management tool. +- **Output:** Verified native dispositions, classifications, links and next routes. +- **Composition:** Conditionally invokes `elicit`, `research`, `maintain-context` and `maintain-agents`. +- **Stops:** After disposition. It does not perform feature specification, debugging, implementation, retirement planning or incident response. + +#### `refine-backlog` + +- **Outcome:** The selected backlog is current, prioritised and honest about scope, readiness, dependencies and blockers. +- **Inputs:** A bounded backlog view, its linked authorities, current project direction and native tracker relationships. +- **Output:** Verified updates, splits, closures, dependencies, priorities and readiness represented in the task-management tool. +- **Composition:** Conditionally invokes `elicit` for unresolved readiness, scope or priority decisions and `maintain-agents` when the task-management preference is missing. +- **Stops:** After backlog health and ordering are current. It does not commit the project to a sprint, cycle, milestone or implementation sequence. + +#### `plan-iteration` + +- **Outcome:** A feasible body of ready, unblocked work is selected for the project's next delivery period. +- **Inputs:** The refined backlog, current priorities, dependencies, available capacity and the project's native planning mechanism. +- **Output:** A verified sprint, cycle, milestone or ordered next-work queue in the task-management tool. +- **Composition:** Conditionally invokes `elicit` for selection trade-offs and `maintain-agents` when the task-management preference is missing. +- **Stops:** After the selected body of work is recorded. It does not implement any item. + +#### `review-architecture` + +- **Outcome:** A bounded, evidence-backed assessment identifies a small number of high-value architecture or maintainability improvements. +- **Inputs:** An explicitly scoped codebase area, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. +- **Output:** An independently useful architecture review report with prioritised findings, evidence, benefits and trade-offs. +- **Composition:** Requires `modular-design`; conditionally invokes `research`. +- **Stops:** After reporting. It does not refactor the code or create speculative tickets automatically. + +### Deliver + +#### `implement` + +- **Outcome:** One ticket or small agreed body of work exists as a minimal, verified local change. +- **Inputs:** The stated ticket or confirmed work, linked product and technical authorities, repository guidance and current code. +- **Output:** Working code, tests and other in-scope artefacts with reproducible verification and no unresolved required review finding. +- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit`, `research`, `prototype`, `modular-design`, `tdd` and `maintain-context`. +- **Stops:** After the agreed behaviour is implemented and verified. It does not commit or publish the change. + +#### `commit` + +- **Outcome:** Eligible reviewed work is recorded in one or more coherent Conventional Commits. +- **Inputs:** The current Git state, repository guidance, recent history and any user-supplied scope. +- **Output:** Verified atomic commits that preserve unrelated staged, unstaged and untracked work. +- **Composition:** No required utility. +- **Stops:** After reporting created commits and remaining work. It does not push or create a pull request. + +#### `pull-request` + +- **Outcome:** The current branch and its complete committed work are represented by one accurate pull request. +- **Inputs:** The current branch, its base, full branch history, repository guidance and any existing pull request. +- **Output:** A pushed branch and created or updated pull request whose title and body describe the whole branch. +- **Composition:** Conditionally invokes `commit` when eligible uncommitted work remains. +- **Stops:** When the published head, base, content and review state are verified. It does not merge, release or deploy the work. + +#### `review-pull-request` + +- **Outcome:** Someone else's pinned pull request receives an independent, evidence-backed assessment. +- **Inputs:** The complete pull-request change, originating specification or work item, repository standards and relevant product and technical authorities. +- **Output:** Prioritised findings covering behavioural conformance, regressions, minimality, conventions, maintainability, architecture and verification evidence. +- **Composition:** Requires `verify-change` and `code-review`; conditionally invokes `modular-design` and `research`. +- **Stops:** After reporting findings. It does not change the contributor's branch or publish comments, approvals or requested changes unless the user's invocation includes that outcome. + +### Debug and Respond + +#### `debug` + +- **Outcome:** A defect has an evidence-backed root cause and, when authorised by the requested outcome, a minimal verified repair. +- **Inputs:** Expected and observed behaviour, the failing signal, relevant environment evidence and current implementation authorities. +- **Output:** A reproducible diagnosis and, for repair work, regression protection plus a reviewed local change. +- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `prototype`, `modular-design`, `tdd` and `maintain-context`. +- **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. + +#### `respond-to-incident` + +- **Outcome:** An active production incident is understood sufficiently to stabilise service safely and verify the resulting service state. +- **Inputs:** The incident signal, production evidence, affected systems, available operational controls and current communications. +- **Output:** A verified mitigation or recovery, concise timeline, current impact and explicit unresolved risks or follow-up needs in the owning incident system. +- **Composition:** Conditionally invokes `debug`, `verify-change` and `research` when they are necessary to stabilise the incident. +- **Stops:** When service is stable or the exact external blocker is established. It does not perform the later learning review. + +#### `review-incident` + +- **Outcome:** A stabilised incident produces a blameless, evidence-based explanation and owned follow-up work. +- **Inputs:** The incident timeline, technical and operational evidence, mitigations, communications and current product and architecture context. +- **Output:** A durable incident review and verified follow-up items routed into the normal lifecycle. +- **Composition:** Conditionally invokes `research`, `maintain-context` and `maintain-agents` when tracker-backed follow-up work lacks a configured task-management preference. +- **Stops:** After learning and follow-up ownership are recorded. It does not implement the resulting changes. + +### Retire + +#### `plan-retirement` + +- **Outcome:** An obsolete capability has an approved plan for safe retirement. +- **Inputs:** The capability, product intent, users, dependencies, interfaces, data, integrations, support obligations and current implementation evidence. +- **Output:** A retirement plan covering deprecation, migration, communication, removal, archival, access, recovery and observable completion conditions. +- **Composition:** Requires `elicit`; conditionally invokes `research` and `maintain-context`. +- **Stops:** After the retirement outcome and obligations are approved. Execution returns to `create-tickets`, `implement`, `commit` and `pull-request` in later sessions. + +## Reusable utility catalogue + +| Utility | Independently useful outcome | +| --- | --- | +| `elicit` | Resolves dependency-ordered user decisions one at a time until shared understanding is confirmed. | +| `research` | Answers a material question using high-trust evidence and creates a cited report when the conclusion must survive the session. | +| `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | +| `maintain-context` | Keeps `CONTEXT.md` aligned with confirmed ubiquitous language, domain rules and conceptual boundaries. | +| `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | +| `tdd` | Drives red-green-refactor through a stable observable seam and runnable test suite. | +| `verify-change` | Selects and executes the applicable project and risk-triggered quality harnesses and reports reproducible evidence and limitations. | +| `code-review` | Independently assesses a scoped change for required behaviour, regressions, minimality, conventions, maintainability, code smells and architecture. | +| `maintain-agents` | Records confirmed, non-inferable repository-wide agent guidance while keeping permanent context lean. | + +`write-skill` remains a separate suite-development tool. It is not presented as part of the software-delivery lifecycle. + +## Quality model + +### Research prerequisite + +Before finalising `implement`, `tdd`, `verify-change` or `code-review`, conduct focused research using primary technical sources, established standards and credible field practices. Persist one cited report that establishes: + +- a small universal quality baseline +- selection triggers for additional harnesses +- what each harness can and cannot demonstrate +- ways metrics and tests can be gamed +- how executable evidence and engineering judgement complement each other +- how to report incomplete confidence when required infrastructure is absent + +The research evaluates practices used successfully in real projects without turning every available harness into mandatory ceremony. + +### Verification portfolio + +`verify-change` discovers the repository's existing commands, tests, CI-equivalent checks and QA procedures, then selects the applicable portfolio for the changed behaviour and risk. The portfolio may include: + +- unit, acceptance, integration, contract and system tests +- executable business examples such as Gherkin where they improve precision +- static types, linting, formatting and framework diagnostics +- coverage analysis and mutation testing +- security, accessibility, performance, compatibility and resilience checks +- reproducible manual or specialist QA procedures + +The repository's existing constraints remain authoritative. A skill may strengthen evidence for the current change but must not weaken or skip required checks silently. + +When an existing project cannot support meaningful verification, the skill reports the limitation and recommends explicit planned work. It does not install a project-wide test framework or quality tool merely to satisfy its own process. New projects establish required infrastructure through their architecture and initial delivery tickets. + +### Maintainability judgement + +Passing checks does not prove that a change is well designed. `code-review` and `modular-design` also require the implementation to: + +- contain only the behaviour and code needed for the agreed outcome +- follow the language, framework and repository conventions +- reuse established helpers and abstractions where they fit +- avoid duplication, speculative generality and recognised code smells +- preserve or improve information hiding and module boundaries +- remain understandable and economical to change + +Review findings are corrected through the relevant implementation and verification loop until no required finding remains unresolved. + +## Representative flows + +### Establish a product + +```text +define-product + → later define-architecture +``` + +Product and architecture planning are separate sessions and stop after their respective documents. + +### Deliver a substantial feature + +```text +specify-feature + → later design-feature + → later create-tickets + → implement each selected ticket in a separate session + → commit whenever a coherent unit is ready + → pull-request when the branch is ready for human review +``` + +### Deliver a small understood change + +```text +elicit when needed + → implement + → commit + → pull-request +``` + +### Refine team work + +```text +triage-work + → refine-backlog + → plan-iteration when the project uses a delivery period +``` + +The skills may be invoked independently. Triage may route a valid defect directly to `debug` or a new capability to `specify-feature`. + +### Debug a known defect + +```text +debug + → commit + → pull-request +``` + +### Respond to an incident + +```text +respond-to-incident + → later review-incident + → follow-up work enters Define, Plan, Refine or Deliver +``` + +### Retire a capability + +```text +plan-retirement + → later create-tickets + → normal implementation and publication sessions +``` + +### Review another contributor's work + +```text +review-pull-request +``` + +The author's normal flow does not invoke `review-pull-request`; `implement` already includes independent change review. + +## Implementation order + +1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. +2. Align the shared utilities: `elicit`, `research`, `prototype`, `maintain-context`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. +3. Build the progressive foundation path: `define-product`, `define-architecture`, `specify-feature`, `design-feature` and `create-tickets`. +4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. +5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. +6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. +7. Remove `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. + +Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. + +## Acceptance conditions + +The revised suite is ready when: + +- every common lifecycle task has one obvious user-facing entry point +- every lifecycle skill produces one bounded outcome and stops at its documented boundary +- substantial work narrows cleanly from product foundation through specification, design, tickets and per-item implementation +- small, clear work can enter implementation without ceremonial documents or tickets +- required and conditional composition is explicit and does not duplicate supporting skill instructions +- task-management skills use the project tool named in `AGENTS.md` and its native concepts without Propulsion configuration +- documentation preserves enduring intent without copying observable implementation or live tool state +- implementation always completes applicable verification and independent review +- absent quality infrastructure is reported honestly and never causes unrelated tooling to be installed implicitly +- external mutations remain within the outcome authorised by the invoked skill and are read back after application +- release, deployment and continuous monitoring remain outside the suite +- each skill passes a fresh-agent common-path test and every material permission, missing-input and stopping boundary identified during authoring +- representative cross-session flows pass without transcript dependency or an orchestrating lifecycle skill +- `bun run checks` and `git diff --check` pass + +## Explicit non-goals + +Propulsion will not: + +- provide a lifecycle navigator or end-to-end delivery orchestrator +- require a Propulsion manifest, setup skill, state model, completion horizon or tracker adapter +- impose Scrum, sprints or another project-management method +- create separate skills for individual test types, metrics or quality commands +- require discovery, architecture or feature documents for small work that is already sufficiently understood +- duplicate code, CI, tracker or production state in prose documents +- release, deploy or continuously monitor software +- merge a pull request merely because it has been created or reviewed +- turn incident response and incident learning into one session +- perform retirement implementation or production decommissioning inside `plan-retirement` From bdcc9f2a34f78b595b3a1f15b9a787d73e489584 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 17:54:19 +0100 Subject: [PATCH 03/79] feat(elicit): expand domain-neutral elicitation --- SKILLS_PLAN.md | 2 +- skills/elicit/SKILL.md | 32 ++++++++++++++------------------ skills/elicit/agents/openai.yaml | 2 +- 3 files changed, 16 insertions(+), 20 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 4bf818c..66a9f3f 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -250,7 +250,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | Utility | Independently useful outcome | | --- | --- | -| `elicit` | Resolves dependency-ordered user decisions one at a time until shared understanding is confirmed. | +| `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | | `research` | Answers a material question using high-trust evidence and creates a cited report when the conclusion must survive the session. | | `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | | `maintain-context` | Keeps `CONTEXT.md` aligned with confirmed ubiquitous language, domain rules and conceptual boundaries. | diff --git a/skills/elicit/SKILL.md b/skills/elicit/SKILL.md index 515dcaa..2e41b1c 100644 --- a/skills/elicit/SKILL.md +++ b/skills/elicit/SKILL.md @@ -1,6 +1,6 @@ --- name: elicit -description: Elicits user-confirmed decisions one at a time until shared understanding is complete. Use when requirements, constraints, trade-offs, boundaries, or intent need the user's direction before acting. +description: Elicits material user-held information and user-authoritative decisions one question at a time. Use when shared understanding requires the user's knowledge, preferences, constraints, or direction. metadata: invocation: model disable-model-invocation: false @@ -8,34 +8,30 @@ disable-model-invocation: false # Elicit -**Goal-oriented requirements engineering** refines a request into user-confirmed decisions until every branch reaches shared understanding. +Uses an internal **decision tree** to resolve material information and decisions, reaching **theoretical saturation** when further questioning cannot change the shared understanding. ## Process -### 1. Establish the root goal +### 1. Establish the basis -Inspect the request, conversation, and accessible task-scoped material before questioning. Derive every available fact and identify the user's intended outcome as the root goal. Keep fact-finding read-only; when it requires a state-changing operation, make authorisation the active decision. Represent every unavailable fact as a branch to resolve with the user. The root goal and factual basis are explicit. +Inspect the request, conversation, caller context, and accessible task-scoped evidence before questioning. Derive available facts and identify the user's intended outcome. Keep inspection read-only; when necessary fact discovery would change state, ask exactly one question to obtain explicit authority before performing it. The outcome and factual basis are explicit. -### 2. Refine the decision tree +### 2. Maintain the decision tree -Build and continually update an internal, dependency-ordered decision tree. Use lightweight goal refinement to expand the entire request across outcome, scope, terminology, inputs, outputs, prerequisites, dependencies, constraints, flows, exceptions, permissions, risks, trade-offs, and success conditions. Explore alternatives, scenarios, obstacles, and conflicts; preserve compatible decisions when the tree changes. Select the highest-impact unresolved decision whose dependencies are resolved. One active decision is explicit. +Build and continually recompute an internal, dependency-ordered decision tree of unresolved user-held information and user-authoritative decisions. Open a branch only when its answer could materially change the outcome, boundaries, requirements, constraints, permissions, trade-offs, risks, or success conditions. Preserve compatible answers as branches open, close, become irrelevant, or reopen. Select the highest-impact unresolved branch whose prerequisites are resolved. One active branch is explicit. -### 3. Ask one issue +### 3. Ask one question -Use **Issue-Based Information Systems** to frame the active decision as one issue with a recommended position, viable alternatives, and the decisive arguments and trade-offs. Offer only positions that resolve the issue in the current tree. Derive the recommendation from the confirmed goal, evidence, conventions, consequences, and prior decisions. Ask exactly one question per turn, ask it once, and leave room for the user's own answer. Treat the user as the sole decision authority. The active decision has one explicit response. +Ask exactly one explicit question per turn in natural conversational language. Explain why an information question matters. For a decision, present only viable positions and their decisive trade-offs, recommending one when the outcome, evidence, applicable conventions, consequences, and prior answers provide a defensible basis. Leave room for the user's own answer and treat the user as the sole authority over their information and decisions. The active branch has one explicit response. -### 4. Scaffold an answer +### 4. Adapt and validate -When the user cannot answer, apply **contingent scaffolding** to the same issue through plain-language restatement, clearer alternatives and trade-offs, examples, or scenarios. When needed, decompose it into the highest-impact prerequisite decisions, resolve them one at a time, and recombine their answers. Continue adapting the issue until the user resolves it. The active decision has a user-confirmed answer. +Test the response internally against the outcome, prior answers, assumptions, implications, scenarios, edge cases, obstacles, conflicts, and consequences. Advance the tree when the response is clear and consistent. When a material ambiguity or conflict remains, adapt the active question through plain-language restatement, clearer choices and trade-offs, examples, scenarios, or decomposition into prerequisite questions resolved one at a time. Recombine their answers and reopen every affected branch. The active branch is resolved consistently. -### 5. Validate the answer +### 5. Reach theoretical saturation -Test each answer against the root goal, prior decisions, scenarios, counterexamples, edge cases, obstacles, conflicts, and consequences. Use **Socratic questioning** to probe the single highest-impact uncertainty in its assumptions, evidence, implications, or viewpoints. Keep an ambiguous or conflicting answer open, state discovered constraints and consequences as facts, and add every exposed decision to the tree. Preserve compatible answers and rebuild affected branches. The answer is consistent and every consequence is represented. +Repeat steps 2–4 until every material branch is resolved. Then make a complete materiality and consistency pass across the outcome, scenarios, obstacles, conflicts, consequences, and edge cases. Theoretical saturation is reached only when the pass adds or reopens no branch whose answer could materially change the shared understanding. The decision tree is saturated. -### 6. Reach theoretical saturation +### 6. Confirm shared understanding -Continue steps 1–5 until **theoretical saturation**: every branch has a confirmed answer, all dependencies and answers are consistent, and a final goal-refinement, scenario, obstacle, conflict, consequence, and edge-case pass produces no new branch. Shared understanding is ready for synthesis. - -### 7. Confirm shared understanding - -Present one concise, self-contained synthesis of the outcome, boundaries, decisions, constraints, and observable success conditions, then ask for explicit agreement. Treat a correction or rejection as new evidence, reopen every affected branch, and continue from step 2 until saturation returns before presenting the revised synthesis. Affirmative agreement completes elicitation. Return only the user-confirmed synthesis to the caller. +Present one concise, self-contained synthesis of the outcome, relevant facts, decisions, boundaries, constraints, and observable success conditions, then ask exactly one question for explicit agreement. Obtain a new response to that complete synthesis; only this response completes elicitation. Treat a correction or rejection as new evidence, reopen every affected branch, and continue from step 2 until saturation returns before presenting the revised synthesis. Return only the user-confirmed synthesis to the caller without performing downstream work. diff --git a/skills/elicit/agents/openai.yaml b/skills/elicit/agents/openai.yaml index dc7e24b..e687258 100644 --- a/skills/elicit/agents/openai.yaml +++ b/skills/elicit/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Elicit' - short_description: 'Reach shared understanding one decision at a time' + short_description: 'Resolve material information and decisions' policy: allow_implicit_invocation: true From dd74920883707dbb2af7f857554eda6da89d981a Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 17:54:38 +0100 Subject: [PATCH 04/79] docs: add skills implementation checklist --- SKILLS_PLAN.md | 33 +++++++++++++++++++++++++++++++++ 1 file changed, 33 insertions(+) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 66a9f3f..ebd6ed4 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -388,6 +388,39 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. 7. Remove `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. +### Implementation checklist + +- [x] Complete quality-harness research +- [x] Update `elicit` skill +- [ ] Update `research` skill +- [ ] Create `prototype` skill +- [ ] Update `maintain-context` skill +- [ ] Update `modular-design` skill +- [ ] Update `tdd` skill +- [ ] Create `verify-change` skill +- [ ] Update `code-review` skill +- [ ] Update `maintain-agents` skill +- [ ] Update `define-product` skill +- [ ] Create `define-architecture` skill +- [ ] Create `specify-feature` skill +- [ ] Create `design-feature` skill +- [ ] Create `create-tickets` skill +- [ ] Update `implement` skill +- [ ] Update `commit` skill +- [ ] Replace `pr` with `pull-request` +- [ ] Create `review-pull-request` skill +- [ ] Create `triage-work` skill +- [ ] Create `refine-backlog` skill +- [ ] Create `plan-iteration` skill +- [ ] Update `review-architecture` skill +- [ ] Update `debug` skill +- [ ] Create `respond-to-incident` skill +- [ ] Create `review-incident` skill +- [ ] Create `plan-retirement` skill +- [ ] Delete `elicit-with-context` skill +- [ ] Update the README +- [ ] Run the suite-wide acceptance pass + Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. ## Acceptance conditions From 76036e6b3036a324fd193906140552b79cce4370 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 19:34:14 +0100 Subject: [PATCH 05/79] feat(research): strengthen evidence reporting --- SKILLS_PLAN.md | 2 +- skills/research/SKILL.md | 20 ++++++++++--------- skills/research/agents/openai.yaml | 2 +- .../assets/research-report-template.md | 14 ++++++------- .../research/references/REPORT-LIFECYCLE.md | 6 +++++- 5 files changed, 25 insertions(+), 19 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index ebd6ed4..c62c3b1 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -392,7 +392,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill -- [ ] Update `research` skill +- [x] Update `research` skill - [ ] Create `prototype` skill - [ ] Update `maintain-context` skill - [ ] Update `modular-design` skill diff --git a/skills/research/SKILL.md b/skills/research/SKILL.md index c6c4cfc..181599a 100644 --- a/skills/research/SKILL.md +++ b/skills/research/SKILL.md @@ -1,6 +1,6 @@ --- name: research -description: Researches questions against high-trust primary sources and persists cited reports. Use when a durable evidence-backed answer is needed. +description: Researches material questions using critically appraised sources and persists verified cited reports. Use when a trustworthy answer must survive the session. metadata: invocation: model disable-model-invocation: false @@ -8,30 +8,32 @@ disable-model-invocation: false # Research -**Rapid evidence assessment** turns a scoped question into an auditable report grounded in high-trust primary sources. +Turns a scoped material question into a trustworthy report while a fresh agent contains the investigation context. ## Process ### 1. Define the research contract -Define the research question, intended use, scope, exclusions, currency needs, and source hierarchy. Inspect task-relevant repository context and `docs/research/` for related reports before searching. Resolve ambiguity that could materially change the investigation. The research contract and applicable prior evidence are explicit. +Define the research question, intended use, scope, exclusions, currency needs, and domain evidence constraints. Inspect task-relevant repository context and `docs/research/` for related reports before searching. Resolve ambiguity that could materially change the investigation. When the contract requires a formal research, appraisal, or reporting method, consult its current authoritative guidance, apply the controls the investigation can satisfy, and disclose every unmet prerequisite instead of claiming conformance. The research contract and applicable prior evidence are explicit. ### 2. Assign the investigation -Give a fresh agent the complete research contract, relevant repository context, primary-source standard, and output contract. One fresh agent owns source discovery, appraisal, synthesis, and report writing; the caller verifies the finished report. +Give one fresh agent the complete research contract, relevant repository context, evidence rules, and report contract. It owns discovery, appraisal, synthesis, and report writing; the caller later verifies and reads the finished report. The investigation context remains outside the caller. -### 3. Discover and appraise primary evidence +### 3. Discover and appraise the evidence -Discover the strongest applicable primary evidence, including official documentation, source code, standards, original publications, first-party APIs, and first-party data. Use secondary sources only as discovery leads, then apply **backward citation searching** to trace material claims to their originals. Critically appraise authority and access, validity, currency, applicability, completeness, and bias. Primary-source status sets the hierarchy; appraisal determines the trust warranted. The evidence set is relevant, current enough for the question, and traceable. +Identify the material claims the answer must establish and select the source type capable of establishing each one. Trace every material claim to the source that owns it, such as a standard, official documentation or source code, an original study or dataset, or an originator's work. Rigorous syntheses may support discovery and interpretation; unverified commentary may locate, interpret, or challenge original evidence but carries no independent evidential weight. + +Apply **critical appraisal** to determine what each source can establish and the trust it warrants. Assess authority and access to the fact, methodological or technical validity, currency, applicability, completeness, incentives or bias, and consistency with the underlying evidence. Treat an originator as authoritative for the method they define, not as proof that the method produces a claimed effect. When a material source participates in a citation graph, apply **citation searching** backward to cited originals and forward to later validation, correction, extension, or criticism; keep it supplementary to direct discovery. The evidence set is relevant, current enough for the question, and traceable. ### 4. Synthesize the findings -Compare independent evidence through **triangulation**, treating sources that repeat the same upstream claim as one evidence route. Test emerging conclusions through **falsification** by seeking contrary evidence and plausible alternatives. Distinguish direct evidence, inference, conflict, and unknowns; narrow or qualify conclusions when the evidence cannot support a stronger answer. Stop discovery when each material claim is supported or explicitly unresolved and further primary-source work is unlikely to change the answer. Every material finding is proportionate to the evidence. +Corroborate material conclusions through genuinely independent evidence routes, treating sources that repeat one upstream claim as one route. Seek contrary evidence, plausible alternatives, and conditions that would change the conclusion; distinguish missing evidence from evidence against. Separate evidence, inference, conflict, and unknowns, and qualify conclusions when the evidence cannot support a stronger answer. Stop discovery when each material claim is supported or explicitly unresolved and further trustworthy source work is unlikely to change the answer. Every material finding is proportionate to the evidence. ### 5. Write the research report -Persist the result at `docs/research/YYYYMMDD-{research-title}.md` using [the research report template](assets/research-report-template.md). Use a concise lowercase hyphenated title and claim-level links to primary evidence. Replace every placeholder, retain only applicable lifecycle fields, and complete every applicable section. Record the research date, material search locations or approaches, appraisal basis, and synthesis method for **auditability** without retaining the raw search trail. When related research already exists, apply [the research report lifecycle](references/REPORT-LIFECYCLE.md). The report is concise, auditable, and proportionate to its evidence. +Persist the result at `docs/research/YYYYMMDD-{research-title}.md` using [the research report template](assets/research-report-template.md). Use a concise lowercase hyphenated title, retain only applicable lineage fields, complete every stable section, and structure `Findings` for the subject. Give material claims direct links to the evidence that supports them. Record the research date, material search locations or approaches, evidence selection, appraisal and synthesis basis, important deviations, and verification constraints without retaining the raw search trail. When related research already exists, apply [the research report lifecycle](references/REPORT-LIFECYCLE.md). The report is concise, inspectable, and proportionate to its evidence. ### 6. Verify and hand off -Verify that each material claim is supported by its cited primary source, evidence routes are genuinely independent, every link and relative report path resolves, conflicts and uncertainty are visible, and the recorded method makes the investigation auditable. Return the report path, concise concrete findings, and unresolved limitations to the caller. The caller receives a validated durable result without the raw search context. +The caller verifies that every material claim is supported by the cited source, evidence routes are genuinely independent, contrary evidence and uncertainty are visible, every link and relative report path resolves, and the recorded method makes the investigation inspectable. Return the report path, concise concrete findings, and unresolved limitations. The report owns the evidence; the caller owns every resulting decision or downstream mutation. diff --git a/skills/research/agents/openai.yaml b/skills/research/agents/openai.yaml index 0f9f9ed..f9f37c9 100644 --- a/skills/research/agents/openai.yaml +++ b/skills/research/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Research' - short_description: 'Research primary sources into cited reports' + short_description: 'Research evidence into verified cited reports' policy: allow_implicit_invocation: true diff --git a/skills/research/assets/research-report-template.md b/skills/research/assets/research-report-template.md index 8e382bd..c5fee5d 100644 --- a/skills/research/assets/research-report-template.md +++ b/skills/research/assets/research-report-template.md @@ -3,7 +3,7 @@ title: '{research title}' createdAt: YYYY-MM-DD updatedAt: YYYY-MM-DD status: current -# Retain each applicable relationship and remove this guidance. +# Retain each applicable lineage field and remove this guidance. # supersedes: './YYYYMMDD-research-title.md' # supersededBy: './YYYYMMDD-research-title.md' --- @@ -26,11 +26,11 @@ status: current ### {Finding} -{Material claims with direct links to supporting primary sources. Identify inferences explicitly.} +{Material claims with direct links to supporting evidence. Identify inferences explicitly.} -## Conflicts +## Conflicts and contrary evidence -{Conflicting evidence and its effect on the conclusion, or "None found."} +{Conflicting or contrary evidence and its effect on the conclusion, or "None found."} ## Limitations @@ -38,8 +38,8 @@ status: current ## Method -{The research date, material search locations, terms or approaches, appraisal basis, triangulation and falsification method, and auditability constraints without the raw search trail.} +{The research date, material search locations or approaches, evidence selection, appraisal and synthesis basis, important deviations, and verification constraints without the raw search trail.} -## Primary sources +## Sources -- [{Source title}]({URL or repository-relative path}) — {publisher or owner, version or publication date, accessed YYYY-MM-DD, and relevance} +- [{Source title}]({URL or repository-relative path}) — {publisher or owner, version or publication date, accessed YYYY-MM-DD, and evidential role} diff --git a/skills/research/references/REPORT-LIFECYCLE.md b/skills/research/references/REPORT-LIFECYCLE.md index c01349d..5942d9f 100644 --- a/skills/research/references/REPORT-LIFECYCLE.md +++ b/skills/research/references/REPORT-LIFECYCLE.md @@ -8,7 +8,11 @@ Update a report in place only to repair wording, formatting, or a link to the sa ## Create a substantive refresh -New evidence, changed scope, or a materially changed finding creates a new dated snapshot with a relative `supersedes` link. Mark the previous snapshot `superseded`, add its relative `supersededBy` link, and preserve its historical findings. The report history distinguishes correction from substantive refresh. +New evidence, changed scope, or a materially changed finding creates a new dated snapshot with a relative `supersedes` link. Mark the previous snapshot `superseded`, add its reciprocal relative `supersededBy` link, and preserve its historical findings. Verify both links. The report history distinguishes correction from substantive refresh. + +## Keep related reports current + +When related reports answer independently useful questions, leave both `current` and add no lineage fields. Describe a useful relationship in the report body or sources instead of creating a false supersession history. ## Resolve a dated-path collision From 557920f5c630c1b4d2851acdc167b3e046316249 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 20:36:10 +0100 Subject: [PATCH 06/79] feat(prototype): add disposable visual experiments --- SKILLS_PLAN.md | 2 +- skills/prototype/SKILL.md | 41 +++++++++++++++++++++++++++++ skills/prototype/agents/openai.yaml | 5 ++++ 3 files changed, 47 insertions(+), 1 deletion(-) create mode 100644 skills/prototype/SKILL.md create mode 100644 skills/prototype/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index c62c3b1..2abcfb7 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -393,7 +393,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill - [x] Update `research` skill -- [ ] Create `prototype` skill +- [x] Create `prototype` skill - [ ] Update `maintain-context` skill - [ ] Update `modular-design` skill - [ ] Update `tdd` skill diff --git a/skills/prototype/SKILL.md b/skills/prototype/SKILL.md new file mode 100644 index 0000000..f9456db --- /dev/null +++ b/skills/prototype/SKILL.md @@ -0,0 +1,41 @@ +--- +name: prototype +description: Creates disposable visual artefacts that resolve one bounded uncertainty. Use when seeing, comparing, or interacting would answer a question better than discussion alone. +metadata: + invocation: model +disable-model-invocation: false +--- + +# Prototype + +Creates a temporary visual experiment that makes one uncertainty inspectable, returns what it established, and leaves no prototype code, files, records, or external effects behind. + +## Process + +### 1. Bound the question + +Inspect the request, caller outcome, repository guidance, relevant product and implementation evidence, intended audience, and safely available data. State one uncertainty and the observations that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the comparison criteria and keep the decision with the user. Invoke `$elicit` when material framing information or a user-authoritative choice remains unresolved. When a visible or explorable artefact cannot discriminate the question, return that boundary to the caller without prototyping. The question, audience, decision authority, and discriminators are explicit. + +### 2. Select the representation + +Apply **prototypes as filters and manifestations**: identify the qualities the artefact must expose, what it will omit, and whether an omission could distort the answer; then choose the simplest trustworthy material, resolution, scope, and project context. Use a static artefact when appearance or explanation is sufficient and interaction when behaviour, sequence, state, or experience matters. Build one focused artefact for feasibility or explanation. For comparative design, place two or three materially different alternatives in the same artefact under the same content, data, and constraints; vary structure, hierarchy, or primary interaction rather than decoration alone. The selected form exposes the decisive qualities without implying completeness. + +### 3. Establish the disposable boundary + +Inspect the current project state and identify every file, route, component, process, or data source the prototype will temporarily affect. Use authentic application context and safe read-only data when they materially affect the answer; otherwise work in an isolated temporary location. Temporary adjustments to an existing surface are permitted only when their exact removal can be verified without disturbing pre-existing work. Stub external writes and irreversible actions, and clearly distinguish simulated behaviour from implemented capability. The prototype has an isolated, reversible boundary with no real external mutation. + +### 4. Build the artefact + +Implement only the fidelity and behaviour needed to expose the discriminators. Reuse project-native layout, components, styling, and realistic content density when they affect interpretation, but do not add production hardening, reusable abstractions, tests, or unrelated infrastructure. When state or sequence matters, include guided representative and adverse scenarios and allow direct exploration when action order could reveal unexpected behaviour. Keep all prototype material clearly identifiable as temporary. One faithful disposable artefact is ready to inspect. + +### 5. Render and present it + +Render the artefact in its intended medium and inspect its visual output and relevant interactions. Verify that the question, alternatives or scenarios, simulated boundaries, and decisive observations are understandable to the intended audience; correct anything that could make the comparison misleading. Expose a live view or rendered capture directly in the conversation before cleanup and give the user the simplest controls needed to compare or explore it; a source listing or path that will disappear before it can be viewed is insufficient. When the environment cannot render and expose the result, remove the prototype and report the exact missing capability. The user evaluates the rendered artefact rather than its source. + +### 6. Iterate to a conclusion + +Revise the same artefact while comparison or exploration continues to produce useful evidence. Let the agreed observations settle an objective question and obtain the user's choice for subjective product or design direction. Stop when the uncertainty is resolved or when further useful iterations cannot distinguish the options. The experiment has a supported conclusion or an exact remaining uncertainty and missing discriminator. + +### 7. Remove and return + +After the rendered result has been exposed and the experiment has concluded, remove the artefact and revert only the prototype-specific project adjustments, then verify that no prototype material remains and all pre-existing work is intact. Do not promote, commit, document, deploy, or otherwise retain any prototype code or artefact. Return the concise conclusion to the current conversation or caller; when inconclusive, return the remaining uncertainty, evidence gathered, and missing discriminator instead. The caller receives the result with no durable prototype residue. diff --git a/skills/prototype/agents/openai.yaml b/skills/prototype/agents/openai.yaml new file mode 100644 index 0000000..71e000c --- /dev/null +++ b/skills/prototype/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Prototype' + short_description: 'Resolve uncertainty with disposable visual experiments' +policy: + allow_implicit_invocation: true From f96225a2f2770cad3fdb6d11a1d80e356dc88eb3 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 21:09:20 +0100 Subject: [PATCH 07/79] feat(maintain-context): maintain ubiquitous language inline --- SKILLS_PLAN.md | 6 +-- skills/maintain-context/SKILL.md | 40 ++++++++----------- skills/maintain-context/agents/openai.yaml | 2 +- .../assets/context-template.md | 5 +-- 4 files changed, 21 insertions(+), 32 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 2abcfb7..d44ff7c 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -63,7 +63,7 @@ Every skill: | Authority | Owns | | --- | --- | | `PRODUCT.md` | Product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capability catalogue. | -| `CONTEXT.md` | Ubiquitous language, domain concepts, business rules and conceptual boundaries. | +| `CONTEXT.md` | Canonical project terms and their concise project-specific meanings. | | `ARCHITECTURE.md` | Enduring architecture drivers, system context, responsibilities, technology strategy, data ownership, integrations and cross-cutting technical constraints. | `CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-context` inline. @@ -253,7 +253,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | | `research` | Answers a material question using high-trust evidence and creates a cited report when the conclusion must survive the session. | | `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | -| `maintain-context` | Keeps `CONTEXT.md` aligned with confirmed ubiquitous language, domain rules and conceptual boundaries. | +| `maintain-context` | Keeps `CONTEXT.md` aligned inline with confirmed ubiquitous language. | | `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | | `tdd` | Drives red-green-refactor through a stable observable seam and runnable test suite. | | `verify-change` | Selects and executes the applicable project and risk-triggered quality harnesses and reports reproducible evidence and limitations. | @@ -394,7 +394,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `elicit` skill - [x] Update `research` skill - [x] Create `prototype` skill -- [ ] Update `maintain-context` skill +- [x] Update `maintain-context` skill - [ ] Update `modular-design` skill - [ ] Update `tdd` skill - [ ] Create `verify-change` skill diff --git a/skills/maintain-context/SKILL.md b/skills/maintain-context/SKILL.md index c783dec..24fa12c 100644 --- a/skills/maintain-context/SKILL.md +++ b/skills/maintain-context/SKILL.md @@ -1,6 +1,6 @@ --- name: maintain-context -description: Actively maintains project language and architecture decisions. Use when domain terms or consequential codebase decisions emerge or change. +description: Continually maintains confirmed project terminology and rare ADRs. Use when project language is introduced, changed, misused, or contradicted, or an accepted decision needs durable rationale. metadata: invocation: model disable-model-invocation: false @@ -8,43 +8,35 @@ disable-model-invocation: false # Maintain Context -**Ubiquitous Language** is an active discipline: challenge and refine project terms during ordinary discussion, then write each resolution into one root `CONTEXT.md` before the conversation moves on. Architecture decision records preserve only rare consequential choices. +**Ubiquitous Language** keeps project-specific terminology consistent across conversation, documentation, and code. Maintain one root `CONTEXT.md` as a concise glossary during ordinary work; preserve only rare qualifying architecture decisions separately. ## Process -### 1. Maintain the language inline +### 1. Notice project language -Apply this loop to each material domain term while the discussion is taking place: +Act whenever a material project term is introduced, changed, misused, or contradicted during the caller's work. Compare its meanings in the active discussion and root `CONTEXT.md`, then inspect only the relevant code or authoritative documents needed to expose a conflict. State the conflicting meanings and their sources plainly. The term and any disagreement are explicit. -- Compare it with the single root `CONTEXT.md` and inspect only the relevant code. -- Challenge glossary misuse immediately and quote the conflicting meanings. -- Sharpen vague or overloaded language by proposing one precise canonical term. -- Test the proposed meaning with concrete scenarios and edge cases that expose its boundaries. -- Use **Model-Driven Design** to surface disagreement between language and implementation. Treat code as evidence of current behaviour and the user's confirmed answer as intent. -- When the user resolves the term, update `CONTEXT.md` before continuing the discussion. Keep a genuine uncertainty explicit and leave its glossary entry unresolved. +### 2. Resolve the meaning -Create the root file lazily from the [context template](assets/context-template.md) when the first term resolves. The durable language stays current with the conversation rather than accumulating for handoff. +Challenge an inconsistent, vague, or overloaded meaning immediately. Propose one precise canonical term and test its project-specific meaning with concrete examples or boundary cases suited to the disagreement. Treat every source as evidence rather than automatic authority and obtain confirmation from the user or the caller's already-confirmed outcome. When the meaning remains unresolved, keep the conflict explicit and leave `CONTEXT.md`, including any existing entry, unchanged. One meaning is confirmed or its uncertainty remains visible. -### 2. Keep the glossary rigorous +### 3. Update the glossary inline -Keep `CONTEXT.md` as a glossary of project-specific domain language: define each term's meaning in one or two sentences, leaving implementation and specifications to their authoritative sources. Give each meaning one authoritative entry, and add `_Avoid_` only for aliases or ambiguous alternatives that actually occur. +As soon as a meaning is confirmed, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md) when the first term resolves. Give each canonical term one authoritative entry and define its project-specific meaning in one or two concise sentences, including only what is needed to explain the term. Group entries under descriptive headings when natural clusters emerge; otherwise keep one flat language list. -Use **Conceptual Contours** to group related terms under descriptive subheadings when meaningful domain clusters emerge; keep one flat language list when the terms form a cohesive area. If context-dependent meanings conflict with the single-context structure, surface that ambiguity instead of inventing another context file. +Change no code, specification, product, architecture, or other project artifact. Return every remaining conflict to the caller or artifact owner. The glossary reflects the confirmed language and other authorities retain their ownership. -### 3. Offer ADRs sparingly +### 4. Offer a rare ADR -Use **Architecture Decision Records** only for an accepted codebase decision that passes all three gates: +Consider an architecture decision record only for an already accepted codebase decision that passes every gate: - changing it later has meaningful cost; - a future reader would find it surprising without context; and -- viable alternatives created a genuine trade-off. +- viable alternatives created a genuine trade-off; and +- neither `ARCHITECTURE.md` nor a feature record already explains it adequately. -Offer an ADR when all three pass and let the user decide whether to record it. A decision that misses any gate remains routine and produces no ADR. +When every gate passes, offer the ADR and let the user decide whether to record it. A decision that misses any gate remains routine and produces no ADR. The exceptional record has explicit user agreement. -### 4. Record a qualifying decision +### 5. Record and verify the result -After the user accepts the offer, create `docs/adr/` lazily and write the next record from the [ADR template](assets/adr-template.md). Derive the next four-digit sequence from filenames alone and name it `NNNN-decision-shaped-slug.md`. State the decision first, then only the context and significant ramifications needed to explain it; use exactly `Decision`, `Context`, and `Ramifications` as content sections, in that order. Link supporting material from the record. The ADR is brief, sequentially numbered, and readable from its filename. - -### 5. Verify and hand off - -Re-read each changed artifact against the resolved language, accepted decisions, and relevant code evidence. Report the files changed and the exact uncertainty behind any unresolved contradiction. +When the user accepts the offer, create `docs/adr/` lazily and write the next record from the [ADR template](assets/adr-template.md). Derive the next four-digit sequence from filenames alone and name it `NNNN-decision-shaped-slug.md`. Use exactly `Decision`, `Context`, and `Ramifications` as content sections, in that order, and link only the supporting material needed to understand the choice. In every path, re-read each changed artifact against the confirmed language, accepted decision, and relevant evidence. Report changed files and each unresolved conflict with its owning artifact. The glossary and any qualifying ADR are concise, verified, and ready for the caller to continue. diff --git a/skills/maintain-context/agents/openai.yaml b/skills/maintain-context/agents/openai.yaml index e459f7c..8ff35b5 100644 --- a/skills/maintain-context/agents/openai.yaml +++ b/skills/maintain-context/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Maintain Context' - short_description: 'Maintain project language and decisions' + short_description: 'Maintain project language and rare ADRs' policy: allow_implicit_invocation: true diff --git a/skills/maintain-context/assets/context-template.md b/skills/maintain-context/assets/context-template.md index 486cc18..3f5bcd3 100644 --- a/skills/maintain-context/assets/context-template.md +++ b/skills/maintain-context/assets/context-template.md @@ -1,10 +1,7 @@ # {Project Name} Context -{Describe the project domain and why this language exists in one or two sentences.} - ## Language ### {Optional descriptive concept group} -**{Canonical term}**: {Define what the term is in one or two sentences.}\ -_Avoid_: {List observed aliases or ambiguous alternatives; omit when none exist.} +**{Canonical term}**: {Define the term's project-specific meaning in one or two concise sentences.} From 8df74dcf86cb1024013e28237d0244e9e117c66c Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 14 Aug 2026 21:17:17 +0100 Subject: [PATCH 08/79] docs(context): remove duplicated skill definitions --- CONTEXT.md | 134 ++++------------------------------------------------- 1 file changed, 10 insertions(+), 124 deletions(-) diff --git a/CONTEXT.md b/CONTEXT.md index fa2742b..d72d9c6 100644 --- a/CONTEXT.md +++ b/CONTEXT.md @@ -1,135 +1,21 @@ -# Propulsion - -Propulsion is an agentic coding workflow composed of skills that steer a coding agent through repeatable engineering work. +# Propulsion Context ## Language -**Predictability**: The degree to which a skill makes the agent follow the same process on every run, without requiring the same output.\ -_Avoid_: Consistency, output determinism - -**User-invoked skill**: A skill the user selects explicitly. This is the default skill type in Propulsion.\ -_Avoid_: Manual skill, command - -**Model-invoked skill**: A skill the agent may select autonomously or invoke from another skill. Use this exception only when autonomous discovery would naturally help during ordinary coding work often enough to earn its permanent context cost.\ -_Avoid_: Automatic skill - -**Invocation policy**: Client-specific metadata within a skill bundle that controls whether an agent may select that skill implicitly. The same intent may require different policy fields in different clients.\ -_Avoid_: Invocation flag, frontmatter setting - -**Skill-only distribution**: Distributing Propulsion directly as Agent Skills, using the skills installer for remote sources and filesystem links for local sources. Propulsion does not require client-specific plugin packaging.\ -_Avoid_: Plugin distribution - -**Elicitation**: Requirements elicitation adapted to establish discoverable facts, resolve a dependency-ordered decision tree with the user, and confirm shared understanding before downstream action.\ -_Avoid_: Interrogation, clarification - -**Theoretical saturation**: The point at which further elicitation within the agreed scope reveals no new material decisions, allowing the shared understanding to be presented for confirmation.\ -_Avoid_: Question limit, exhaustive questioning - -**Main success scenario**: The representative invocation path that delivers a skill's intended outcome and remains dominant during design, refinement, and forward testing.\ -_Avoid_: Every possible path, exhaustive scenario set - -**Material exception**: A non-common path that earns explicit skill behaviour because representative evidence, the main success scenario, or a necessary safety or permission boundary requires it. Speculative variation remains agent judgement.\ -_Avoid_: Edge case, hypothetical branch - -### Skill anatomy - -**Authoring workflow**: A skill that guides an agent through creating or updating another skill with an explicit process. It makes execution predictable without prescribing the authored skill's outcome.\ -_Avoid_: Design guide, skill reference - -**Composable skill**: An independently useful skill that may be invoked alone or coordinated by another skill without requiring the rest of a prescribed workflow.\ -_Avoid_: Workflow stage, mandatory step - -**Router skill**: A lightweight skill that invokes other skills to produce a combined outcome. It contains only coordination unique to that outcome; called skills remain authoritative and the router neither repeats nor overrides their context.\ -_Avoid_: Pipeline skill - -**Implement skill**: The user-invoked workflow that coordinates a confirmed implementation through TDD and independent code review. It technically adjudicates review findings, remediates required corrections, accepts optional improvements only when proportionate, and elicits user intent when a finding would change behaviour, contracts, architecture, or scope. - -**TDD skill**: The model-invoked workflow that applies red-green-refactor when an existing runnable test suite can exercise the behaviour change through a stable public seam. It uses Test Desiderata to favour valuable tests that respond to behaviour without coupling to code structure. - -**Code-review skill**: The model-invoked workflow that assesses a scoped code change independently for requirements and code health, then reports evidence-validated findings including code smells and refactor opportunities without changing the code.\ -_Avoid_: Review skill - -**Review-architecture skill**: The user-invoked workflow that analyses a project's architecture and produces an HTML report of high-value, context-aware redesign opportunities without changing the implementation.\ -_Avoid_: Improve-architecture skill - -**Architecture review report**: A single-file interactive artifact named `docs/architecture/YYYYMMDD-{scope}-architecture-review.html` that guides the user through a small set of prioritised recommendations in concise plain language. It uses visualisation and progressive disclosure to explain affected architecture, expected improvements, evidence, and trade-offs without presenting a wall of technical detail. Verified CDN dependencies may supply scripts, styles, fonts, and diagram libraries.\ -_Avoid_: Static architecture audit - -**Architecture module**: A cohesive capability with a small explicit contract and a hidden implementation. Other modules depend on the contract rather than its internal classes, adapters, or framework wiring.\ -_Avoid_: Directory, namespace - -**Deep module**: An architecture module whose small, stable interface hides substantial cohesive implementation. The implementation may be decomposed into focused internal actions for reuse and maintainability without exposing that decomposition to consumers.\ -_Avoid_: Large class, shallow module - -**Debug skill**: The model-invoked workflow that reproduces a code issue, establishes its root cause, applies the smallest correction, and verifies the result. An explicit diagnosis-only request stops before mutation.\ -_Avoid_: Diagnose skill - -**Maintain-agents skill**: The user-invoked workflow that creates or aggressively compresses the root `AGENTS.md` into project-wide runtime guidance and one canonical completion check. It removes narrower workflows from permanent context and reports their appropriate destinations. - -**Define-product skill**: The user-invoked workflow that inspects existing product knowledge, composes contextual elicitation and conditional research, and maintains a root `PRODUCT.md` plus canonical language in `CONTEXT.md`. Concise strategic framing leads into a journey-organised catalogue of high-level feature descriptions without becoming a delivery plan. - -**Primary source**: Original high-trust evidence such as official documentation, source code, standards, publications, first-party APIs, or first-party data. Secondary sources may aid discovery but findings trace their claims back to primary evidence.\ -_Avoid_: Trusted write-up - -**Research report**: A cited Markdown snapshot named `docs/research/YYYYMMDD-{research-title}.md` that answers a research question from primary evidence and records its scope, findings, and unresolved limitations. Substantive re-research creates a linked superseding snapshot; minor corrections update the existing report and its metadata.\ -_Avoid_: Research answer - -**Research skill**: The model-invoked rapid evidence assessment workflow that gives a fresh agent ownership of primary-source discovery, appraisal, synthesis, and report writing. The caller verifies the cited research report and receives its concise findings; other skills invoke it only when the evidence warrants that durable record.\ -_Avoid_: Web search - -**Description**: A concise statement of what a skill does and the conditions under which it should be invoked.\ -_Avoid_: Summary, tagline - -**Skill name**: A short command that states the skill's action and fits naturally into a user instruction. Prefer one imperative verb, then a short imperative phrase, with established nouns reserved for operations they already name clearly.\ -_Avoid_: Title, label - -**Branch**: A distinct route through a skill for a particular use case or condition. Branches share the skill's common process without duplicating it.\ -_Avoid_: Separate workflow, mode - -**Process section**: The required `## Process` section that contains a skill's instructions. It uses numbered subheadings only when order matters and descriptive subheadings or direct prose otherwise.\ -_Avoid_: Steps section, instructions section - -**Leading word**: A recognised term from an established method, principle, theory, or technique, specific enough to invoke the agent's existing knowledge without further explanation. A skill explains only its context-specific adaptation or constraints.\ -_Avoid_: Coined term, theme, slogan - -**Governing methodology**: An established methodology selected through research to determine a skill's process when one credibly fits. Research may conclude that none is suitable; a selected methodology appears by canonical name in the skill without source attribution.\ -_Avoid_: Core concept, main theme - -**Supporting concept**: An established principle, theory, or technique that reinforces the governing methodology for a distinct concern without competing with it.\ -_Avoid_: Secondary concept - -**Skills plan**: The self-contained high-level handoff for Propulsion's fixed v1 skill suite. It records suite principles, the skill catalogue and standard briefs, composition and invocation, implementation order, and acceptance criteria without carrying source citations, discarded scope, or finished skill instructions.\ -_Avoid_: Skill specification, backlog - -**Lossless compression**: Reducing a skill to the fewest words and structures that preserve its behaviour, conditions, constraints, and technical meaning.\ -_Avoid_: Trimming, shortening, minimalism - -**Degrees of freedom**: The amount of judgement a skill leaves to the agent. Match it to the work's fragility so the process is predictable without predetermining valid outcomes.\ -_Avoid_: Flexibility, strictness - -**Ironic process theory**: The tendency for a negated concept to become more salient. Skills state the positive target behaviour and pair an essential safety boundary with the safe action that satisfies it.\ -_Avoid_: Prohibition-only rule, negative prompting +### Skill model -**Prerequisite**: A condition that must be true before a skill can begin. Its failure stops the skill or routes the work elsewhere.\ -_Avoid_: Setup step, pre-flight check +**Predictability**: The degree to which a skill makes the agent follow the same process on every run, without requiring the same output. -**Step**: A numbered subheading used when actions within a process or branch must occur in order. It isolates one coherent behavioural concern, describes the action, and ends in an observable postcondition.\ -_Avoid_: Instruction, rule +**User-invoked skill**: A skill the user selects explicitly. This is the default skill type in Propulsion. -**Postcondition**: An observable state that marks a step complete without requiring a separate completion section.\ -_Avoid_: Completion criterion, completion gate +**Model-invoked skill**: A skill the agent may select autonomously or invoke from another skill. Use this exception only when autonomous discovery would naturally help during ordinary coding work often enough to earn its permanent context cost. -**Rule**: A cross-cutting invariant that constrains multiple instructions or the finished output.\ -_Avoid_: Step, reminder +**Invocation policy**: Client-specific metadata within a skill bundle that controls whether an agent may select that skill implicitly. The same intent may require different policy fields in different clients. -**Handoff**: A transfer or route that becomes available after the skill's process is complete.\ -_Avoid_: Next step, final step +**Skill-only distribution**: Distributing Propulsion directly as Agent Skills, using the skills installer for remote sources and filesystem links for local sources. Propulsion does not require client-specific plugin packaging. -**Reference**: Conditional or extensive runtime guidance linked once beside a precise loading condition. The agent loads it only when that branch or decision needs the additional depth.\ -_Avoid_: Background, resource +### Composition -**Acceptance testing**: A semantic validation that traces every elicited invocation through a finished skill's branches, resource pointers, and observable postconditions.\ -_Avoid_: Checklist, structural validation +**Composable skill**: An independently useful skill that may be invoked alone or coordinated by another skill without requiring the rest of a prescribed workflow. -**Characterization testing**: Capturing an existing skill's observable invocation and process before improvement so intentional changes remain distinct from regressions.\ -_Avoid_: Snapshot, preservation rule +**Router skill**: A lightweight skill that invokes other skills to produce a combined outcome. It contains only coordination unique to that outcome; called skills remain authoritative and the router neither repeats nor overrides their context. From bc37d332c9daf63d9ee781270813a370fa1ccb25 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 09:13:20 +0100 Subject: [PATCH 09/79] feat(modular-design): teach structural maintainability --- SKILLS_PLAN.md | 2 +- skills/code-review/SKILL.md | 2 +- skills/debug/SKILL.md | 2 +- skills/implement/SKILL.md | 2 +- skills/modular-design/SKILL.md | 46 ++++++------------- skills/modular-design/agents/openai.yaml | 2 +- .../modular-design/references/TECHNIQUES.md | 20 +++----- .../references/architecture-analysis.md | 2 +- skills/tdd/SKILL.md | 2 +- 9 files changed, 28 insertions(+), 52 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index d44ff7c..b7604b0 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -395,7 +395,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `research` skill - [x] Create `prototype` skill - [x] Update `maintain-context` skill -- [ ] Update `modular-design` skill +- [x] Update `modular-design` skill - [ ] Update `tdd` skill - [ ] Create `verify-change` skill - [ ] Update `code-review` skill diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 083740b..511d95a 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -28,7 +28,7 @@ The Spec packet applies **bidirectional requirements traceability**: trace every The Standards packet applies repository standards first, then residual **Google code-review criteria** across whole-change understanding, correctness and concurrency risks, test presence and validity, comments, and affected documentation. Include the complete [Fowler code-smell work aid](references/CODE-SMELLS.md). Add **Test Desiderata** when tests change; the relevant **ISO/IEC 25010:2023** characteristic when the repository adopts it or the change exposes a concrete residual product-quality concern; an applicable **SEI CERT** rule when supported-language code exposes its construct; and the relevant frozen **OWASP ASVS 5.0.0** requirement when Web code crosses that security boundary. Load only the implicated part of a conditional benchmark. -When the fixed change alters modular architecture, invoke `$modular-design` and include the applicable standard in the Standards packet. +When the fixed change presents a material structural-maintainability decision, invoke `$modular-design` and include the applicable standard in the Standards packet. Within Standards, repository rules and demonstrably configured tooling govern the concerns they cover. General work aids fill uncovered diagnostic roles and yield to an explicit repository choice. A smell or benchmark cue begins an investigation; it becomes a finding only when the scoped code supplies exact evidence and a concrete consequence. diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 0e125e7..2a707af 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -22,7 +22,7 @@ Gather evidence and state falsifiable causal hypotheses in evidence-supported or ### 3. Test one repair -When the user requested diagnosis only, stop before mutation and follow the Handoff. Invoke `$modular-design` when the supported repair changes modular architecture. Invoke `$tdd` when its runnable-suite and stable-seam prerequisite applies; preserve the established reproduction as its Red signal and let it own the minimal Green repair and regression protection. When TDD is not applicable, apply one smallest change that would repair the leading cause if the hypothesis is correct. Keep the attempt within the authorised mutation boundary and leave unrelated cleanup or refactoring outside it. One evidence-led repair is ready for a causal verdict. +When the user requested diagnosis only, stop before mutation and follow the Handoff. Invoke `$modular-design` when the supported repair presents a material structural-maintainability decision, and carry its constraints into `$tdd` when the runnable-suite and stable-seam prerequisite applies; preserve the established reproduction as TDD's Red signal and let it own the minimal Green repair and regression protection. When TDD is not applicable, apply one smallest change that would repair the leading cause if the hypothesis is correct. Keep the attempt within the authorised mutation boundary and leave unrelated cleanup or refactoring outside it. One evidence-led repair is ready for a causal verdict. ### 4. Accept or revert the repair diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 32b451b..0322ca3 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -12,7 +12,7 @@ A tailored **Software Formal Inspection** lifecycle routes confirmed work throug ## Process -Invoke `$tdd` to implement the confirmed work, then invoke `$code-review` on the result. Validate every finding against the confirmed request and current codebase, then disposition it: +Inspect the confirmed work, its applicable authorities, and the relevant current code. Invoke `$modular-design` when that evidence presents a material structural-maintainability decision and carry its constraints into `$tdd`, then invoke `$code-review` independently on the result. Validate every finding against the confirmed request and current codebase, then disposition it: - A required correction is a verified violation of confirmed behaviour, the specification, a repository rule, or a necessary safety boundary, or another defect that leaves the change untrustworthy. - A proportional improvement is valid but non-blocking; accept it only when its concrete benefit justifies the added change and complexity. diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index 4620ee4..44901fd 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -1,6 +1,6 @@ --- name: modular-design -description: Defines an evidence-backed modular architecture standard. Use when designing or assessing modules, interfaces, dependencies, ownership, or seams. +description: Applies an evidence-backed structural-maintainability standard. Use when designing or assessing ownership, boundaries, contracts, dependencies, seams, or change propagation. metadata: invocation: model disable-model-invocation: false @@ -8,48 +8,32 @@ disable-model-invocation: false # Modular Design -**Information hiding** makes code safer to change by assigning cohesive knowledge and change-prone decisions to modules whose callers learn only a small, stable contract. +Supplies structural constraints inside a caller's workflow by locating each material decision with a coherent owner, making the smallest complete change, and keeping caller knowledge proportionate to the capability provided. ## Process -Apply this standard inside the caller's workflow. Let the caller own repository inspection, quality priorities, design comparison, implementation, verification, and artifacts; this skill supplies architecture knowledge without starting a separate process or producing its own output. Explicit project constraints, domain language, and architecture decisions govern where they conflict with the baseline. +### 1. Fix the behaviour and evidence -## Standard +Apply this standard inside the caller's workflow. Use the scoped behaviour, project guidance, architecture decisions, domain language, and relevant implementation evidence; explicit project authorities govern when they conflict with this baseline. Let the caller own inspection, design comparison, implementation, verification, and artifacts. Leave correctness, testing, security, performance, formatting, and general style to their applicable authorities rather than selecting their methods or outputs here. The behaviour, evidence, and material structural decision are explicit. -### Name the architecture precisely +### 2. Assign the affected knowledge -- A **module** is a cohesive capability with an interface and implementation, regardless of whether code expresses it as a function, object, package, process, or tier-spanning slice. -- An **interface** is everything callers must know to use the module correctly, including behaviour, data shapes, invariants, ordering, errors, configuration, side effects, and material performance characteristics. It is broader than a language `interface` declaration. -- An **implementation** is the hidden representation, policy, algorithm, sequencing, framework detail, and collaboration that fulfils the interface. -- A **seam** is a controlled place where behaviour can be observed or substituted without editing the calling location. -- An **adapter** translates between a module's contract and a technology, protocol, framework, or external system. +Identify the representation, invariant, policy, protocol, sequencing, or framework decision the behaviour introduces or changes. Keep it with its natural existing owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each material rule once, and let code that changes for the same evidenced reason change together. The affected knowledge has one coherent owner rather than being repeated across callers. -Use this vocabulary for reasoning while preserving established project and framework names in code and reports. +### 3. Make the smallest coherent change -### Hide owned knowledge +Deliver the complete scoped behaviour without speculative abstraction or unrelated cleanup. Permit a bounded enabling refactor only when the behaviour cannot fit cleanly without it; change no more surrounding structure than the new behaviour needs. Keep a stable local concrete dependency direct unless meaningful variation, isolation, observation, migration, or policy ownership justifies a boundary. -Decompose around difficult, consequential, or change-prone knowledge. Give one module ownership of each representation, invariant, policy, protocol, sequencing rule, or framework decision that other modules should not repeat. A change to hidden knowledge should remain behind its interface unless the promised behaviour changes. +Realize the design through established project and framework idioms. Objects may own identity, state, invariants, and cohesive behaviour; functions, closures, structural types, language modules, components, and framework primitives are equally valid when they preserve the same ownership and contract. Introduce a language interface only when callers need a stable contract distinct from its realization, and use inheritance only for genuine substitutability or a required framework extension. The result is the smallest idiomatic change that completely delivers the behaviour. -### Prefer deep cohesive modules +### 4. Keep the contract deep -Apply **deep modules**: make the caller-visible interface markedly simpler than the cohesive capability it exposes. Remove or absorb shallow wrappers that repeat another interface, scatter one decision across callers, or add navigation without hiding knowledge. +Treat a module as a cohesive capability, whether code expresses it as a function, object, package, process, or tier-spanning slice. Its contract is everything callers must know, including behaviour, data, invariants, ordering, errors, configuration, side effects, and material performance characteristics. -Use **cohesion and coupling** qualitatively. Keep knowledge that changes for the same reason together; separate unrelated actors, models, or policies. Reduce cross-module knowledge, coordination, cycles, and change propagation while retaining necessary collaboration. Do not optimize file size, class count, method count, or mechanical coupling scores as substitutes for architectural evidence. +Apply **deep modules**: make the caller-visible contract markedly simpler than the cohesive capability and knowledge hidden behind it. Absorb or remove shallow forwarding layers that merely repeat another contract or redistribute navigation; retain a layer when it owns translation or another evidenced boundary responsibility. The contract repays its caller burden without combining unrelated knowledge to appear deeper. -### Choose an idiomatic realization +### 5. Check locality and applicable techniques -Prefer object-oriented realization where the language and framework make it natural: objects own identity, state, invariants, and cohesive behaviour; purposeful action or use-case entry points hide a complete operation; collaborators are composed; and nominal interfaces express meaningful variation or ownership boundaries. +Trace plausible project-evidenced changes through the owner, contract, callers, and affected code. Reconsider the boundary when one owned decision would still require scattered changes, when several elements must always change together, or when only an unrelated part of the proposed owner would change. Do not substitute file size, class count, method count, dependency scores, or speculative agent-context savings for this evidence. -Treat functions, closures, structural types, and language modules as equivalent realizations when they provide the same ownership, contract, and hiding. In frontend frameworks, keep components and framework-specific state or effect primitives focused on presentation and interaction, and place durable policy behind framework-neutral modules when that separation is cohesive. Framework-owned code may use framework types at its own edge. - -### Load only applicable techniques - -Read [Modular Design Techniques](references/TECHNIQUES.md) when evidence presents a volatile mechanism, application-to-technology boundary, competing domain model, need for controlled observation or substitution, entangled deterministic policy and effects, or an architectural promise that needs repeatable protection. Use only the technique whose stated condition is present. - -## Rules - -- Keep a stable local concrete dependency direct when no meaningful knowledge, variation, isolation, observation, or migration need justifies another abstraction. -- Introduce a language interface only when callers need a stable contract distinct from a realization; an interface that mirrors one concrete type without hiding knowledge is ceremony. -- Let a cohesive module contain several internal actions. A class or function with one entry point is valuable only when it hides a complete capability rather than forwards the call. -- Prefer composition in object-oriented code; use inheritance for a genuine substitutable type or required framework extension contract. -- Optimize the knowledge callers require, not repository fragmentation or speculative token savings. Reduced agent context is an inference to verify, not proof of correctness. +Decide from the current evidence whether it presents a policy-to-mechanism dependency, an application-to-technology conversation, competing domain models, a need for controlled substitution or observation, or deterministic decisions entangled with effects. When none is present, keep [Modular Design Techniques](references/TECHNIQUES.md) unloaded and continue with the direct process; do not open it merely to confirm that a condition is absent. Otherwise read the reference and apply only the technique whose stated condition is present. The resulting structure keeps the owned knowledge and its likely changes local without introducing an unevidenced abstraction. diff --git a/skills/modular-design/agents/openai.yaml b/skills/modular-design/agents/openai.yaml index 5214400..8e43c61 100644 --- a/skills/modular-design/agents/openai.yaml +++ b/skills/modular-design/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Modular Design' - short_description: 'Apply an evidence-backed modular design standard' + short_description: 'Apply structural maintainability decisions' policy: allow_implicit_invocation: true diff --git a/skills/modular-design/references/TECHNIQUES.md b/skills/modular-design/references/TECHNIQUES.md index 5c27e2b..96b351f 100644 --- a/skills/modular-design/references/TECHNIQUES.md +++ b/skills/modular-design/references/TECHNIQUES.md @@ -1,14 +1,14 @@ # Modular Design Techniques -Load this reference only when repository evidence satisfies a technique's condition. Information hiding and the standard in `SKILL.md` remain authoritative; a pattern earns its place by solving the evidenced boundary problem. +Load this reference only when repository evidence satisfies a technique's condition. The direct process in `SKILL.md` remains authoritative; use one mechanism to satisfy overlapping conditions when possible rather than adding duplicate layers. -## Protect policy from a volatile mechanism +## Direct dependencies towards policy Apply **dependency inversion** when stable domain or application policy directly knows a materially volatile framework, device, vendor, persistence, transport, or delivery mechanism. Define the smallest policy-owned contract in the policy's language and make the mechanism satisfy it. Keep a stable local concrete dependency direct when inversion would only create an interface-per-class and wiring. ## Isolate an application conversation -Apply **ports and adapters** when one application capability must support multiple technologies, replacement, isolated execution, or a material external boundary. Let a port describe the purpose of the conversation and let adapters translate UI, persistence, network, vendor, or test technology. Keep framework types at framework-owned edges and avoid wrapping every framework call. +Apply **ports and adapters** when an application-purpose conversation must remain independent of multiple technologies, replacement, isolated execution, or a material external boundary. Let the port state the application's need and let adapters translate UI, persistence, network, vendor, or test technology. Keep framework types at framework-owned edges and avoid wrapping every framework call. ## Separate domain models @@ -16,18 +16,10 @@ Apply **bounded contexts** when the same term or concept has competing meanings, Apply **conceptual contours** when domain language and observed axes of change reveal a more natural capability grain than technical layers or uniform class sizes. Keep cohesive entity behaviour and invariants together; use a standalone domain service only for a significant process that belongs to no entity or value object. -## Create purposeful variation or observation +## Create controlled variation or observation -Add a **seam** when testing, diagnosis, replacement, observation, or staged migration needs controlled variation. Use the smallest existing enabling point—such as a parameter, collaborator, provider, module export, or framework facility—before adding a nominal interface. Direct behaviour remains preferable when it is already local, stable, and observable. +Add a **seam** when substitution, observation, replacement, or staged migration needs behaviour to vary without editing the calling location. Supply the smallest enabling point that selects the behaviour, such as a parameter, collaborator, provider, module export, or framework facility. The caller owns the reason and verification; direct behaviour remains preferable when it is already local, stable, and observable. ## Separate decisions from effects -Apply **functional core, imperative shell** when substantial deterministic policy is entangled with I/O, time, mutable state, or framework lifecycle behaviour. Express the policy as transformations of explicit values and keep effects in a thin shell. Preserve stateful objects when identity, lifecycle, effect ordering, or invariant ownership is intrinsic; do not turn the shell into the unencapsulated application. - -## Preserve an architectural promise - -Define an **architecture fitness function** when a material contract, dependency rule, quality threshold, or runtime characteristic has a faithful objective signal. State the property, signal, expected result, and execution point. Prefer contract tests, dependency rules, adapter conformance, cycle checks, or measured runtime thresholds; retain qualitative review when a metric would be a misleading proxy. - -## Realize a capability entry point - -Use an action object, use-case class, function, command, or framework-native entry point when it exposes one meaningful actor goal and hides cohesive sequencing, policy, invariants, failure handling, or transaction behaviour. Reuse the entry point when several delivery mechanisms need the same capability. Absorb or remove it when it merely forwards to another module or becomes generic for hypothetical reuse. +Apply **functional core, imperative shell** when substantial deterministic decisions are entangled with I/O, time, mutable state, or framework lifecycle behaviour. Express the decisions as transformations of explicit values and keep effects at a narrow edge. Preserve stateful ownership when identity, lifecycle, effect ordering, or invariant enforcement is intrinsic; do not leave an unencapsulated application in the shell. diff --git a/skills/review-architecture/references/architecture-analysis.md b/skills/review-architecture/references/architecture-analysis.md index fb21338..2b02717 100644 --- a/skills/review-architecture/references/architecture-analysis.md +++ b/skills/review-architecture/references/architecture-analysis.md @@ -1,6 +1,6 @@ # Architecture Analysis -Load this reference while selecting and ranking redesigns. `$modular-design` owns architecture vocabulary, information hiding, deep modules, realization, and conditional techniques; this reference owns review-specific evidence and judgment. +Load this reference while selecting and ranking redesigns. `$modular-design` owns knowledge and boundary decisions, contracts, deep modules, locality, realization, and conditional techniques; this reference owns review-specific evidence and judgment. ## Quality Priority diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 3a365b8..4feace0 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -18,7 +18,7 @@ TDD applies when an existing runnable test suite can exercise the requested beha ### 1. Establish the baseline -Read repository instructions, identify the relevant test command, and run the existing suite. Use **tracer bullets** to select the smallest end-to-end behaviour that advances the request, then choose the narrowest stable public seam that can observe the slice without exposing hidden structure. Invoke `$modular-design` when the slice changes modular architecture. Identify a **test oracle**—a requirement, worked example, invariant, contract, trusted reference, accepted prior behaviour, or explicit domain decision—capable of distinguishing the expected outcome from the implementation. Consult [Test Quality](references/TEST-QUALITY.md) when the seam, oracle, or proposed assertion could couple to representation. The baseline, behaviour, seam, oracle, and applicable modular constraints are explicit. +Read repository instructions, identify the relevant test command, and run the existing suite. Use **tracer bullets** to select the smallest end-to-end behaviour that advances the request. Apply caller-supplied modular constraints, or invoke `$modular-design` when the slice presents a new material structural-maintainability decision, then choose the narrowest stable public seam that observes the slice without exposing hidden structure. Identify a **test oracle**—a requirement, worked example, invariant, contract, trusted reference, accepted prior behaviour, or explicit domain decision—capable of distinguishing the expected outcome from the implementation. Consult [Test Quality](references/TEST-QUALITY.md) when the seam, oracle, or proposed assertion could couple to representation. The baseline, behaviour, seam, oracle, and applicable modular constraints are explicit. ### 2. Red From 5aad60e845f9aa90591206317497193523afb3d1 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 11:31:04 +0100 Subject: [PATCH 10/79] feat(tdd): enforce durable behavioural tests --- SKILLS_PLAN.md | 4 ++-- skills/debug/SKILL.md | 2 +- skills/tdd/SKILL.md | 14 +++++++------- skills/tdd/references/TEST-DOUBLES.md | 2 +- skills/tdd/references/TEST-QUALITY.md | 17 +++++++++-------- 5 files changed, 20 insertions(+), 19 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index b7604b0..7ae5db4 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -255,7 +255,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | | `maintain-context` | Keeps `CONTEXT.md` aligned inline with confirmed ubiquitous language. | | `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | -| `tdd` | Drives red-green-refactor through a stable observable seam and runnable test suite. | +| `tdd` | Drives red-green-refactor through a stable observable boundary and runnable test suite. | | `verify-change` | Selects and executes the applicable project and risk-triggered quality harnesses and reports reproducible evidence and limitations. | | `code-review` | Independently assesses a scoped change for required behaviour, regressions, minimality, conventions, maintainability, code smells and architecture. | | `maintain-agents` | Records confirmed, non-inferable repository-wide agent guidance while keeping permanent context lean. | @@ -396,7 +396,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `prototype` skill - [x] Update `maintain-context` skill - [x] Update `modular-design` skill -- [ ] Update `tdd` skill +- [x] Update `tdd` skill - [ ] Create `verify-change` skill - [ ] Update `code-review` skill - [ ] Update `maintain-agents` skill diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 2a707af..7da24b3 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -22,7 +22,7 @@ Gather evidence and state falsifiable causal hypotheses in evidence-supported or ### 3. Test one repair -When the user requested diagnosis only, stop before mutation and follow the Handoff. Invoke `$modular-design` when the supported repair presents a material structural-maintainability decision, and carry its constraints into `$tdd` when the runnable-suite and stable-seam prerequisite applies; preserve the established reproduction as TDD's Red signal and let it own the minimal Green repair and regression protection. When TDD is not applicable, apply one smallest change that would repair the leading cause if the hypothesis is correct. Keep the attempt within the authorised mutation boundary and leave unrelated cleanup or refactoring outside it. One evidence-led repair is ready for a causal verdict. +When the user requested diagnosis only, stop before mutation and follow the Handoff. Invoke `$modular-design` when the supported repair presents a material structural-maintainability decision, and carry its constraints into `$tdd` when a runnable suite can exercise the repair through a stable observable boundary; preserve the established reproduction as TDD's Red signal and let it own the minimal Green repair and regression protection. When TDD is not applicable, apply one smallest change that would repair the leading cause if the hypothesis is correct. Keep the attempt within the authorised mutation boundary and leave unrelated cleanup or refactoring outside it. One evidence-led repair is ready for a causal verdict. ### 4. Accept or revert the repair diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 4feace0..f11293f 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -1,6 +1,6 @@ --- name: tdd -description: Implements observable features and bug fixes through red-green-refactor. Use when an existing runnable test suite can exercise the change at a stable public seam. +description: Implements observable features and bug fixes through red-green-refactor with durable tests. Use when an existing runnable test suite can exercise the change at a stable observable boundary. metadata: invocation: model disable-model-invocation: false @@ -8,21 +8,21 @@ disable-model-invocation: false # Test-Driven Development -**Classicist TDD** builds one observable behaviour at a time through red-green-refactor, testing the narrowest stable public seam with real internal collaborators and doubling only uncontrollable boundaries. +Builds one observable behaviour at a time through red-green-refactor. Retained tests detect promised behaviour changes while surviving changes to hidden structure. ## Prerequisite -TDD applies when an existing runnable test suite can exercise the requested behaviour through a stable public seam. Otherwise return control with the missing condition; the caller owns any decision to create a test harness or reshape a public contract. Apply TDD to observable behaviour, including configuration with observable effects; leave documentation, configuration-only maintenance, and behaviour-preserving refactors with the caller. +TDD applies when an existing runnable test suite can exercise the requested behaviour through a stable observable boundary. Otherwise return control with the exact missing condition; the caller owns any decision to create a test harness or reshape a public contract. Apply TDD to observable behaviour, including configuration with observable effects; leave documentation, configuration-only maintenance, and behaviour-preserving refactors with the caller. ## Process ### 1. Establish the baseline -Read repository instructions, identify the relevant test command, and run the existing suite. Use **tracer bullets** to select the smallest end-to-end behaviour that advances the request. Apply caller-supplied modular constraints, or invoke `$modular-design` when the slice presents a new material structural-maintainability decision, then choose the narrowest stable public seam that observes the slice without exposing hidden structure. Identify a **test oracle**—a requirement, worked example, invariant, contract, trusted reference, accepted prior behaviour, or explicit domain decision—capable of distinguishing the expected outcome from the implementation. Consult [Test Quality](references/TEST-QUALITY.md) when the seam, oracle, or proposed assertion could couple to representation. The baseline, behaviour, seam, oracle, and applicable modular constraints are explicit. +Read repository instructions, identify the relevant test command, and run the existing relevant suite. Record baseline failures and continue only when the focused Red and Green signals remain distinguishable; otherwise return the blocker. Use **tracer bullets** to select the smallest end-to-end behaviour that produces one requested observable result. Apply caller-supplied modular constraints, or invoke `$modular-design` when the slice presents a new material structural-maintainability decision. Choose the narrowest boundary that exposes the result, hides structure allowed to change, and credibly exercises production behaviour. Identify a **test oracle**—a requirement, worked example, invariant, contract, trusted reference, accepted prior behaviour, or explicit domain decision—capable of disagreeing with the implementation; when none exists, return the missing behavioural decision. Consult [Test Quality](references/TEST-QUALITY.md) when the boundary, oracle, or proposed assertion needs deeper guidance. The baseline, behaviour, boundary, oracle, and applicable modular constraints are explicit. ### 2. Red -Use **Arrange-Act-Assert** to add one focused test. Apply the **Test Desiderata**, especially behavioural sensitivity, structure insensitivity, specificity, determinism, readability, and production prediction. Keep internal collaborators real. When an uncontrollable boundary must be controlled or observed, choose the least powerful **Test Double** that supplies the required evidence; consult [Test Doubles](references/TEST-DOUBLES.md) before introducing a double or interaction assertion. +Add one focused test: arrange only the necessary state, act once through the selected boundary, assert the complete promised outcome, and clean up resources the test acquires. Name one plausible promised-behaviour defect that must make the test fail and one hidden structural change that must leave it unchanged. Keep internal collaborators real. When an uncontrollable boundary must be controlled or observed, choose the least powerful **Test Double** that supplies the required evidence; consult [Test Doubles](references/TEST-DOUBLES.md) before introducing a double or interaction assertion. For a bug, reproduce the incorrect behaviour; adopt an already-failing regression test only when it independently specifies the desired behaviour. Run the focused test and confirm that it fails for the expected behavioural reason. When it fails because of the test or environment, remain in Red: correct an in-scope defect or report the blocker, then rerun until the intended failure is observed. Meaningful red evidence exists before Green begins. @@ -32,8 +32,8 @@ Implement only enough production code to satisfy the behaviour, then run the foc ### 4. Refactor -Improve the test and production code while keeping behaviour fixed. Preserve the test across changes to algorithms, collaborators, storage, rendering, or other hidden structure; when structure alone breaks it, move the observation back to the public outcome. Run the focused tests after each material change until the design is clear and green. The cycle ends without a refactor regression. +Improve the production design behind the selected boundary while keeping behaviour fixed. Preserve the test across changes to algorithms, collaborators, storage, rendering, or other hidden structure; when structure alone breaks it, move the observation back to the promised outcome. Clarify test code only without weakening its oracle or quality counterfactuals. Run the focused tests after each material change until the design is clear and green. The cycle ends without a refactor regression. ### 5. Complete the cycles -Repeat Red, Green, and Refactor for each remaining behaviour, then run the complete relevant suite. Report the behaviours delivered, red and green evidence, refactors, commands, results, and unresolved baseline failures. The requested behaviour and retained tests are verified. +Repeat Red, Green, and Refactor for each remaining behaviour, then run the complete relevant suite. Report the behaviours delivered, Red and Green evidence, refactors, commands, results, unresolved baseline failures, and other limitations. The requested behaviour and durable retained tests are verified; later verification and review remain with the caller. diff --git a/skills/tdd/references/TEST-DOUBLES.md b/skills/tdd/references/TEST-DOUBLES.md index ded54f2..3c90b73 100644 --- a/skills/tdd/references/TEST-DOUBLES.md +++ b/skills/tdd/references/TEST-DOUBLES.md @@ -14,7 +14,7 @@ Start with the real collaborator, then introduce only the capability the test ne | Spy | Record an otherwise invisible boundary effect | Inspect only promised boundary facts | | Mock | Specify a required external interaction protocol | Verify only contractually material calls | -Prefer state verification: act through the public seam, then inspect its result or a recorded boundary effect. Use interaction verification when the interaction is itself observable behaviour, such as one idempotency-keyed payment request or committing only after a durable write. +Prefer state verification: act through the selected observable boundary, then inspect its result or a recorded boundary effect. Use interaction verification when the interaction is itself observable behaviour, such as one idempotency-keyed payment request or committing only after a durable write. ## Keep the contract visible diff --git a/skills/tdd/references/TEST-QUALITY.md b/skills/tdd/references/TEST-QUALITY.md index 1a53249..7657103 100644 --- a/skills/tdd/references/TEST-QUALITY.md +++ b/skills/tdd/references/TEST-QUALITY.md @@ -1,8 +1,8 @@ # Test Quality -Use this guide when choosing a seam or oracle, or when an assertion may couple the test to representation rather than promised behaviour. +Use this guide when choosing an observable boundary or oracle, or when an assertion may couple the test to representation rather than promised behaviour. -## Choose the test seam +## Choose the observable boundary Test the narrowest boundary that satisfies all three conditions: @@ -10,7 +10,7 @@ Test the narrowest boundary that satisfies all three conditions: - **Stable:** it hides algorithms, collaborator graphs, storage layouts, rendering wrappers, generated selectors, and other decisions that may change while behaviour remains fixed. - **Predictive:** exercising it provides credible evidence that the behaviour will work in production. -Use a coarser companion test only when the narrow seam cannot predict a material integration outcome. A substitutable private hook is not a suitable assertion boundary merely because it is convenient to replace. +Use a coarser companion test only when the narrow boundary cannot predict a material integration outcome. A substitutable private hook is not a suitable assertion boundary merely because it is convenient to replace. ## Construct the oracle @@ -36,17 +36,18 @@ When no credible oracle exists, return the missing behavioural decision instead ## Structure one behaviour -Use Arrange-Act-Assert: +Keep the test's phases explicit: 1. **Arrange** only the state and collaborators needed for the behaviour. -2. **Act** once through the selected public seam. +2. **Act** once through the selected observable boundary. 3. **Assert** the complete promised outcome and material unchanged state. +4. **Clean up** resources or external state acquired by the test. -One behaviour may require several cohesive assertions. Split a test when it contains multiple independent Acts or when its name cannot state one behavioural rule. Clean up resources or external state acquired by the test. +One behaviour may require several cohesive assertions. Split a test when it contains multiple independent Acts or when its name cannot state one behavioural rule. -## Apply the Test Desiderata +## Challenge the test's quality -Retain tests that are isolated, composable, deterministic, fast, writable, readable, behavioural, structure-insensitive, automated, specific, predictive, and inspiring. Treat the properties as trade-offs: a slower test can earn its cost through prediction, while a faster test still needs credible production evidence. +Before retaining a test, name one plausible promised-behaviour defect that it must detect and one hidden structural change that it must survive. Require deterministic, readable assertions and credible evidence for production behaviour. Treat speed, isolation, and fidelity as trade-offs: a slower test can earn its cost through prediction, while a faster test still needs credible production evidence. Use two counterfactuals: From 49dee507238a9ffd4221b75f423e671739ac7dfa Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 15:04:23 +0100 Subject: [PATCH 11/79] feat(verify-change): add quality harness verification --- SKILLS_PLAN.md | 2 +- skills/verify-change/SKILL.md | 66 ++++++++++++++++++++ skills/verify-change/agents/openai.yaml | 5 ++ skills/verify-change/references/HARNESSES.md | 43 +++++++++++++ 4 files changed, 115 insertions(+), 1 deletion(-) create mode 100644 skills/verify-change/SKILL.md create mode 100644 skills/verify-change/agents/openai.yaml create mode 100644 skills/verify-change/references/HARNESSES.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 7ae5db4..cf73e0f 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -397,7 +397,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `maintain-context` skill - [x] Update `modular-design` skill - [x] Update `tdd` skill -- [ ] Create `verify-change` skill +- [x] Create `verify-change` skill - [ ] Update `code-review` skill - [ ] Update `maintain-agents` skill - [ ] Update `define-product` skill diff --git a/skills/verify-change/SKILL.md b/skills/verify-change/SKILL.md new file mode 100644 index 0000000..6fbe2b4 --- /dev/null +++ b/skills/verify-change/SKILL.md @@ -0,0 +1,66 @@ +--- +name: verify-change +description: Verifies fixed software changes with applicable quality harnesses and reproducible evidence. Use for a diff, revision, pull request, repair, or completed implementation. +metadata: + invocation: model +disable-model-invocation: false +--- + +# Verify Change + +Verifies a fixed software change through the smallest faithful portfolio of repository-required and risk-triggered quality evidence without repairing or certifying the implementation. + +## Process + +### 1. Fix the verification basis + +Read repository instructions and resolve the exact change set from caller context or the user's stated diff, revision, branch comparison, pull request, repair, or other fixed scope. Capture its revisions when applicable, changed and untracked paths, current repository status, intended behaviour, governing specifications and decisions, repository rules, architecture promises, known risks, and relevant pre-existing results. Inspect the patch, affected implementation, tests, quality configuration, and enough surrounding evidence to understand what changed without performing an independent code review. Ask when the scope or expected behaviour is materially ambiguous; when it cannot be fixed, return the exact blocker without running checks. The change, authorities, initial state, and verification target are explicit. + +### 2. Build the evidence map + +Apply **bidirectional verification traceability** for evidence selection and accounting: trace each applicable behaviour, repository rule, architecture promise, and material risk forward to the changed scope and evidence capable of assessing it; trace each material changed path, selected harness, result, limitation, and verdict statement back to its governing claim or rule. Do not use this map to replace the independent conformance and maintainability judgement owned by code review. + +Discover existing repository commands, tests, CI-equivalent gates, and reproducible QA procedures. Every applicable repository-required gate remains selected. Add a risk-triggered harness only when it tests a named claim or exposes a distinct material failure mode; multiple checks resting on the same signal do not create independent assurance. When a specialist risk, unfamiliar harness, or evidential boundary makes selection unclear, consult [Harness Selection](references/HARNESSES.md). Do not add tests, author an ad hoc executable check, generated-input campaign, benchmark, or oracle, install tools, create project-wide infrastructure, or alter CI to fill a gap, even when the addition would be transient. A read-only inspection or probe may characterise a directly observable artefact or diagnose existing evidence, but it does not become a missing behavioural harness. Record the gap for the verdict and recommend explicit planned work. Each material claim has a selected existing evidence route or a stated limitation. + +### 3. Bound the execution + +For each selected command or procedure, establish the revision or diff, environment, configuration, fixtures or data, seed, material tool versions, expected outputs, and local or external effects. Prefer a check, dry-run, read-only, or disposable mode. Ordinary caches, build artefacts, temporary files, disposable services, and isolated local test data are permitted only when their boundary is understood; clean up only effects created and owned by this run. Never remove or overwrite pre-existing user work. + +Do not run an operation that can alter source, tests, snapshots, baselines, suppressions, dependencies, durable project data, or an external system unless the user has separately authorised that exact verification effect and it does not remediate the change. Use a non-mutating mode when available; otherwise record the check as not run and its consequence. The execution plan is reproducible and stays within available authority. + +### 4. Execute and appraise the evidence + +Run the portfolio in widening order where project dependencies permit: focused reproduction or behavioural evidence; configured static, formatting, type, framework, and architecture checks; related component, integration, contract, or system evidence; broader regression and build gates; then activated specialist procedures. A faster or narrower check does not substitute for a slower faithful one. Existing CI or supplied evidence counts only when its provenance ties it to the frozen change and exposes enough of the execution boundary and result to appraise; otherwise run the applicable local equivalent or record the limitation. + +For each executable behavioural check, identify its **test oracle**: the requirement, invariant, worked example, versioned contract, trusted reference, accepted prior behaviour, explicit domain decision, or other authority capable of distinguishing acceptable from incorrect behaviour. Treat implementation-derived expectations, regenerated snapshots, proxy metrics, and execution without a discriminating outcome as weak or absent oracles. Ask what plausible counterexample would refute the claim and what defect in the harness, configuration, inputs, exclusions, or oracle could make a passing result misleading. For any metric, state the claim it informs and how the observed value bears on that claim; never let a proxy erase a failed gate or unsupported behaviour. + +Record one evidence entry per claim with: + +- the exact claim; +- the harness, its repository or procedure provenance, why it applies, and whether it is required or risk-triggered; +- the execution boundary; +- `pass`, `fail`, `flaky/inconclusive`, or `not run`, with salient output, exit status, and artefact location where applicable; +- the inputs, paths, states, rules, environments, or acceptance criteria actually exercised; +- the limitation and residual risk; and +- the in-scope resolution taken, or the planned work or user decision needed for stronger evidence. + +Preserve every retry, seed, failing counterexample, and contradictory observation. A flaky or inconclusive result never becomes a pass through repetition. A required gate that fails remains a failure even when evidence proves it predates the change; state that provenance and whether it masks change-specific signals. The executed portfolio has inspectable claim-level results rather than an undifferentiated green summary. + +### 5. Reinspect scope and harness integrity + +Compare the final repository state and diff with the frozen basis. Account for changes to source, tests, snapshots, dependencies, generated artefacts, quality configuration, baselines, ignore lists, suppressions, and scripts. Distinguish declared temporary effects from material changes, but do not revert an unexpected change when ownership is uncertain. When any in-scope source or measurement-path change occurred after the scope was fixed, identify the paths, mark affected evidence stale, and stop without silently retargeting the run. The final evidence refers to the frozen change and an intact measurement path, or its invalidation is explicit. + +### 6. Determine the verification verdict + +Use exactly one verification-specific verdict: + +- `passed within stated scope` when every applicable required gate passed, each material claim has faithful evidence, no material result is unreliable or stale, and all remaining limitations are stated; +- `partially verified` when useful evidence passed but a named material claim or risk lacks a faithful harness or reliable result; +- `unverified` when the requested behaviour could not be exercised, an applicable required gate did not run, or stale evidence leaves no adequate basis for the requested scope; or +- `failed` when an applicable required gate or acceptance criterion failed and remains unresolved. + +A failed required gate takes precedence over incomplete evidence. Do not compute a confidence percentage or describe static inspection, compilation, coverage, or any other single route as proof beyond the claim it exercised. The verdict characterises verification evidence only; it does not certify the complete implementation or replace independent code review. + +### 7. Report and stop + +Return the exact frozen scope, authorities, material risks, selected and omitted harnesses, execution boundaries, commands or procedures, claim-level evidence entries, before-and-after repository state, verdict, limitations, residual risks, and any recommended planned work. Name every check that was not executed and why. Do not repair a failure, add a missing harness, weaken a gate, commit, publish, release, deploy, or continuously monitor. The caller receives reproducible evidence and one bounded verdict for its own remediation, review, or lifecycle decision. diff --git a/skills/verify-change/agents/openai.yaml b/skills/verify-change/agents/openai.yaml new file mode 100644 index 0000000..57b4514 --- /dev/null +++ b/skills/verify-change/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Verify Change' + short_description: 'Run proportionate quality harnesses for a fixed change' +policy: + allow_implicit_invocation: true diff --git a/skills/verify-change/references/HARNESSES.md b/skills/verify-change/references/HARNESSES.md new file mode 100644 index 0000000..2439c1a --- /dev/null +++ b/skills/verify-change/references/HARNESSES.md @@ -0,0 +1,43 @@ +# Harness Selection + +Load this reference when a material risk, unfamiliar quality harness, or unclear evidential boundary makes the smallest faithful portfolio difficult to select. Repository rules, specifications, architecture decisions, configured checks, and accepted thresholds remain authoritative. + +## Select by claim and failure mode + +Start with the claim, not a preferred tool. Choose an existing harness only when its observation and oracle can distinguish a failure of that claim within a known environment. Retain every applicable repository-required gate. Add another existing route only when it exposes a distinct material failure mode or reduces a named oracle, environment, or state-space weakness enough to justify its feedback cost. Do not invent a transient test, fuzzer, benchmark, generator, or reference implementation to close a project evidence gap; return that need as planned implementation work. + +Run cheap diagnostic checks early when dependencies permit, but never substitute speed for fidelity. Prefer deterministic, local, narrowly diagnostic, machine-readable procedures for correction feedback. Preserve slower integration, specialist, or manual procedures when they alone exercise a material boundary. A required unreliable check remains unreliable evidence rather than a pass. + +## General-purpose harnesses + +| Harness family | Select when | Evidence supplied | Principal boundary and integrity check | +| --- | --- | --- | --- | +| Focused unit or component test | A stable observable boundary can exercise the changed behaviour economically. | The oracle accepted the exercised nominal, boundary, failure, or off-nominal examples. | It does not establish collaborations or environments outside the boundary. Reject tautological assertions, implementation-derived expectations, and hidden-structure coupling. | +| Executable acceptance or business example | A specification, acceptance criterion, or worked example is authoritative and precision benefits from execution. | The system produced the specified externally observable outcome for the examples. | Examples are not exhaustive. Keep business language tied to real production behaviour rather than step definitions that merely restate code. | +| Integration, contract, or system test | The change crosses a process, service, database, framework, protocol, hardware, or user-facing boundary. | Selected collaborators, interfaces, side effects, or end-to-end paths worked in the exercised configuration. | Contract tests do not prove provider internals or complete collaboration; isolated integration does not prove the deployed system. Record fakes, versions, topology, and omitted real boundaries. | +| Static type, lint, format, framework, or architecture check | The repository configures it, or a named structural or language rule governs the change. | The analysed source satisfied the configured rules in the inspected scope. | It proves only encoded and enabled rules. Inspect baselines, exclusions, suppressions, generated files, database freshness, and changed configuration. | +| Build or package check | Build inputs, dependencies, generated artefacts, packaging, or runtime compilation changed, or the repository requires it. | The declared inputs produced the expected artefact in the recorded environment. | One successful local build does not prove portability or reproducibility. Record dependency resolution, generated changes, environment, and artefact identity. | +| Broader regression suite | Changed behaviour or dependencies could affect established nearby or system behaviour, or the repository requires the suite. | Previously encoded expectations still passed across the suite's actual coverage. | Passing regressions inherit every oracle, exclusion, environment, and flakiness weakness in the suite. Distinguish focused new-behaviour evidence from pass-to-pass regression evidence. | +| Coverage analysis | A coverage gap could reveal changed code or a material path never exercised. | Which configured code, branches, or paths executed under the selected suite. | Coverage is not correctness or test effectiveness. Do not target a percentage without repository authority or treat execution as a discriminating oracle. | +| Changed-code mutation analysis | The sensitivity of changed or high-consequence tests is materially uncertain and existing infrastructure can run it economically. | Selected seeded faults were detected or survived within the mutation scope. | Survivors may be equivalent, irrelevant, unreachable, or genuine gaps. Record operator set, exclusions, timeouts, scope, and survivor appraisal; do not mandate whole-suite mutation. | +| Reproducible manual or specialist procedure | Automation cannot faithfully exercise a visual, physical, usability, domain, or contextual claim. | A named actor followed recorded steps against stated inputs and observed retained outcomes. | One observation may depend on competence and environment. Preserve steps, versions, data, screenshots or logs, expected result, actual result, and specialist limitations. | + +## Risk-triggered specialist harnesses + +| Material risk cue | Candidate existing evidence | What a clean result cannot establish | +| --- | --- | --- | +| Trust boundary, authentication or authorisation, sensitive data, cryptography, untrusted input, command or query construction, file handling, dependency or security configuration | Threat or attack-surface analysis, configured secret/dependency/SAST checks, security-focused tests, DAST, fuzzing, penetration procedure, qualified security review | One scanner or finite attack set cannot establish absence of vulnerabilities. Confirm scope, rule/database version, environment, permissions, suppressions, and untested threats. | +| User interface, semantic structure, keyboard or assistive-technology interaction, contrast, focus, motion, or accessibility requirement | Configured automated accessibility check plus interaction, browser, assistive-technology, or knowledgeable human evaluation as the claim requires | Automated rules cover only machine-detectable conditions; one browser, device, or evaluator does not establish conformance across the target population. | +| Hot path, algorithm, query, serialization, start-up, resource constraint, volume, latency, or throughput requirement | Focused benchmark, profile, load, stress, or soak procedure against an authoritative budget and representative workload | A result does not generalise beyond workload, environment, warm-up, sample size, variance, versions, and competing load. Compare a relevant baseline; do not infer improvement from one noisy run. | +| Supported runtime, browser, device, operating system, database, architecture, locale, time zone, or dependency matrix | Repository matrix, compatibility suite, representative build/run, or targeted manual procedure | Tested combinations do not cover omitted versions or interactions. Use the authoritative support matrix and record exact combinations. | +| Retry, timeout, failure recovery, partial outage, queue, distributed boundary, or degraded dependency | Fault injection, failure-path integration test, resilience procedure, recovery drill, or state-transition assertion | Synthetic faults may not match operational timing or correlated failures. Record injected conditions, recovery oracle, state cleanup, and unexercised modes. | +| Shared mutable state, parallel work, lock, transaction, async ordering, idempotency, or race-sensitive path | Deterministic schedule test when available, repeated stress with preserved seed and trace, invariant check, or focused concurrency procedure | Stress cannot prove absence of races; bounded schedule exploration cannot prove states beyond its model and bounds. Preserve the failing schedule or trace. | +| Schema, migration, backfill, serialization, data transformation, retention, destructive operation, or rollback promise | Representative production-shaped fixtures, forward and reverse migration where required, invariant/reconciliation queries, backup/restore or recovery procedure | Synthetic or sampled data may omit real distributions and invalid legacy states. Record scale, constraints, before/after invariants, failure recovery, and environment fidelity. | +| Environment-sensitive build, generated output, compiler/toolchain change, signed or distributed artefact | Clean build, repeated build under relevant declared variance, artefact comparison, provenance or signature check | Same-machine repetition does not establish reproducibility elsewhere. Record all material inputs, toolchain, environment variation, and comparison method. | +| High-consequence protocol, state machine, concurrency algorithm, or invariant already supported by project modelling infrastructure | Model checking, proof checking, symbolic execution, or other bounded formal analysis | The result applies to the stated model, assumptions, abstraction, and explored bounds; it does not prove model-to-code or environment correspondence without separate traceability. | + +## Guard against self-confirming evidence + +Inspect every changed test, snapshot, baseline, suppression, quality rule, ignore list, generator, fixture, benchmark, and harness script that participates in the verdict. Confirm that its change has specification or repository authority and that its oracle can disagree with the implementation. Treat a weakened or unexplained measurement path as a limitation even when its command is green, and leave independent acceptability judgement to code review. + +For generated or sampled inputs, record the generator domain, distribution, seed, shrink or reduction result, and failing counterexample where applicable. For metrics, record the governing claim, measurement scope, interpretation, variance or threshold authority, and blind spots. Never combine unrelated evidence into a composite score that can hide a required failure or unsupported material risk. From 74db7253d3699c2a543e4d6776dc8c99dddc59af Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 20:54:41 +0100 Subject: [PATCH 12/79] feat(code-review): add independent diagnostic review --- SKILLS_PLAN.md | 2 +- skills/code-review/SKILL.md | 60 +++++++++++--------- skills/code-review/agents/openai.yaml | 2 +- skills/code-review/references/CODE-SMELLS.md | 2 +- 4 files changed, 37 insertions(+), 29 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index cf73e0f..9cbd9b0 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -398,7 +398,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `modular-design` skill - [x] Update `tdd` skill - [x] Create `verify-change` skill -- [ ] Update `code-review` skill +- [x] Update `code-review` skill - [ ] Update `maintain-agents` skill - [ ] Update `define-product` skill - [ ] Create `define-architecture` skill diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 511d95a..468042a 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,6 +1,6 @@ --- name: code-review -description: Reviews scoped code changes against a specification and applicable standards. Use when assessing a diff, branch, pull request, or completed implementation. +description: Independently diagnoses a fixed code change against intended behaviour and engineering standards. Use when reviewing code directly or during implementation, debugging, or pull-request review. metadata: invocation: model disable-model-invocation: false @@ -8,56 +8,64 @@ disable-model-invocation: false # Code Review -**Tailored software formal inspection** prepares fixed evidence packets for independent Standards and Spec inspectors, then presents their diagnostic findings without changing the reviewed work. +Independently diagnoses a fixed code change through isolated Spec and Standards reviews, returning evidence for the caller without verifying, approving, or changing the work. ## Process -### 1. Fix the inspection scope +### 1. Freeze the change -Use the caller-supplied scope, whether uncommitted work, a revision range, a branch comparison, a pull request, or another exact change set. Resolve every revision, capture the patch and changed-path list once through read-only inspection, and include the complete contents of in-scope untracked files. Confirm that the captured change set is non-empty. Ask the user when the scope is missing or ambiguous; report the exact blocker and stop when it is invalid or empty. The inspection has one fixed work product. +Use a caller-supplied frozen basis when available; otherwise resolve the exact uncommitted diff, revision range, branch comparison, pull-request change, repair, or other scoped work product. Read repository instructions, resolve every revision, and capture the patch, changed paths, complete in-scope changed and untracked files, and repository state once. Ask when scope is materially ambiguous. When it is invalid or empty, return the exact blocker without reviewing. The review has one explicit, non-empty work product. -### 2. Resolve the inspection sources +### 2. Resolve the authorities and context -Find the specification from caller context, supplied paths or tickets, issue references and change history, then relevant repository documentation. When none is found, ask the user; omit the Spec inspection only after the user confirms that no specification exists. Independently identify applicable repository instructions, architecture decisions, coding standards, language policies, configured checks, and local conventions. Read the changed files in full, relevant tests, and enough surrounding code to judge the patch. The specification and Standards authorities are explicit. +Identify intended behaviour from the caller's fixed authority, such as a specification, ticket, acceptance criteria, confirmed request, or another project source. Extract the authoritative statement without carrying surrounding conversation into inspection. When no intended-behaviour authority can be found, ask the user to confirm a Standards-only review unless the caller already supplied that confirmation. -### 3. Prepare the work aids +Independently resolve applicable repository instructions, architecture decisions, language and framework policies, configured checks, and established local conventions. Read every changed file in full, relevant tests, affected contracts and consumers, and enough surrounding implementation to understand each change. Reuse relevant caller-supplied `$modular-design` constraints, or invoke `$modular-design` when ownership, boundaries, dependencies, contracts, or change propagation present a material structural concern. -Create one self-contained packet per applicable axis with the fixed patch, changed paths, relevant source context, authority sources, priority definitions, output schema, and read-only verification boundary. Exclude conversation history and the other inspector's materials. +Apply the universal Standards criteria directly: correctness and regression risk, scoped minimality, repository conventions, changed-test validity, maintainability, recognised smells, and architecture. Activate security, performance, accessibility, compatibility, resilience, concurrency, migration, or another specialist concern only when task evidence or project authority exposes it. Use ISO/IEC 25010 only when an authority adopts it. For an implicated supported-language or Web-security construct, consult the current official CERT rule or stable versioned OWASP ASVS requirement narrowly and include its exact applicability and exceptions. When a concrete maintainability shape still needs recognised diagnostic vocabulary, consult [Code Smells](references/CODE-SMELLS.md); a smell prompts investigation and is never finding authority by itself. The two axes have explicit, task-relevant evidence without a universal external checklist. -The Spec packet applies **bidirectional requirements traceability**: trace every applicable requirement into the changed implementation and relevant tests, and every introduced behaviour back to specification authority. It investigates missing, partial, incorrect, conflicting, and unrequested behaviour and relevant unhandled cases. +### 3. Prepare the inspection packets -The Standards packet applies repository standards first, then residual **Google code-review criteria** across whole-change understanding, correctness and concurrency risks, test presence and validity, comments, and affected documentation. Include the complete [Fowler code-smell work aid](references/CODE-SMELLS.md). Add **Test Desiderata** when tests change; the relevant **ISO/IEC 25010:2023** characteristic when the repository adopts it or the change exposes a concrete residual product-quality concern; an applicable **SEI CERT** rule when supported-language code exposes its construct; and the relevant frozen **OWASP ASVS 5.0.0** requirement when Web code crosses that security boundary. Load only the implicated part of a conditional benchmark. +Create one self-contained, operational packet per applicable axis. Both packets receive the frozen scope, complete changed files and necessary surrounding source, their applicable authorities and criteria, the priority definitions and output schema below, and the read-only boundary. Exclude conversation history, undocumented implementation rationale, `$verify-change` results or conclusions, and the other reviewer's materials. -When the fixed change presents a material structural-maintainability decision, invoke `$modular-design` and include the applicable standard in the Standards packet. +The Spec packet directs its reviewer to: -Within Standards, repository rules and demonstrably configured tooling govern the concerns they cover. General work aids fill uncovered diagnostic roles and yield to an explicit repository choice. A smell or benchmark cue begins an investigation; it becomes a finding only when the scoped code supplies exact evidence and a concrete consequence. +1. account for every applicable statement of intended behaviour in the changed implementation and relevant tests; +2. account for every introduced or altered behaviour against an identified authority; and +3. follow affected contracts, states, data shapes, side effects, and consumers far enough to expose missing, partial, conflicting, excess, and regressed behaviour. -### 4. Assign the inspections +The Standards packet directs its reviewer to: -Give each packet to a separate fresh agent and run the Standards and Spec inspections in parallel when both apply. Each inspector owns candidate discovery, **falsification**, authority and code-evidence validation, consequence analysis, and priority validation for its axis. It may run a targeted check only when the command and execution boundary demonstrate that it cannot mutate the checkout, repository state, external systems, or durable project data; otherwise it records the limitation. Each inspector returns only findings that survive its validation. +1. understand every changed line and the necessary whole-file and system context; +2. follow affected contracts, states, data shapes, side effects, callers, and dependencies far enough to expose correctness and regression risks; +3. apply the universal and triggered criteria resolved in step 2; and +4. when tests changed, ask whether they detect a promised-behaviour defect, avoid coupling to hidden structure, remain deterministic and readable, and credibly predict the promised result. -Use **risk-based prioritisation** within each axis: `critical` for immediate data loss, security compromise, or production failure; `high` for incorrect requirements or major behaviour, security, reliability, or maintenance risk; `medium` for a concrete defect or significant code, design, or test weakness; and `low` for a local but worthwhile issue. +Each packet tells the reviewer how to inspect its axis rather than supplying labels alone. -### 5. Present the inspection report +### 4. Run the independent reviews -Check that each assigned packet produced the required output fields, returning an incomplete report to its originating inspector for completion from the same packet. Present the Standards and Spec outputs separately without substantive re-review, merging, deduplication, or cross-axis reranking. Preserve each inspector's findings and ordering. The caller receives the two independent inspection results. +Give each applicable packet to a separate fresh agent, in parallel when possible and otherwise sequentially. Do not substitute the coordinating agent; when fresh-agent delegation is unavailable, mark the affected axis not performed and state the exact limitation. -## Rules +Each reviewer owns candidate discovery, code and authority validation, consequence analysis, and priority validation for its axis. Before retaining a candidate, actively try to disprove it through contradicting authority, an already-handled path, a repository-sanctioned exception, surrounding code, and the strongest benign interpretation. Run a targeted probe only when its command and boundary demonstrate that it cannot mutate the checkout, repository state, durable data, or an external system; otherwise omit it and record the limitation. Report a pre-existing issue only when the fixed change introduces or worsens it, makes it newly consequential, or cannot conform because of it. Return only findings that survive these checks. -- Keep the inspection read-only and return evidence for the caller's implementation process. -- Report only issues introduced by or materially relevant to the fixed change. -- Prefer specification, repository, and code evidence over general guidance or personal preference. -- Hold structural, test, security, and product-quality findings to the same evidence, consequence, and priority standard as behavioural defects. +Use consequence-based priority within each axis: `critical` for immediate data loss, security compromise, or production failure; `high` for incorrect required behaviour or major security, reliability, or maintainability risk; `medium` for a concrete defect or significant code, design, or test weakness; and `low` for a local but worthwhile issue. Priority communicates impact and order, not remediation authority. + +### 5. Preserve the independent results + +Check that each reviewer returned the required fields, returning an incomplete result to that same reviewer for structural completion from the original packet. Reinspect the scoped revisions, diff, paths, and in-scope untracked content. When the frozen work product drifted, identify the changed paths, mark affected findings and clean results stale, and stop without silently retargeting the review. + +Present Standards and Spec separately without substantive re-review, merging, deduplication, suppression, or cross-axis reranking. Preserve each reviewer's findings and ordering. The caller receives independent diagnostic evidence for its own adjudication. ## Handoff -State the exact scope, specification source or user-confirmed absence, Standards sources, and any check that could not run. Return `## Standards` and `## Spec`; use `No findings.` for a clean axis and state when the Spec inspection was omitted. Format each finding as: +State the exact frozen scope, intended-behaviour authority or user-confirmed absence, Standards authorities, any omitted or unperformed axis, every probe that could not run, and any scope drift. Return `## Standards` and `## Spec`; use `No findings.` for a clean axis and state when Spec was omitted or either axis was not performed. Format each finding as: ```markdown ### [priority] Concise finding -- Evidence: exact code `path:line`, applicable authority, and observed fact -- Consequence: concrete behaviour or code-health impact +- Evidence: exact code `path:line`, applicable authority or criterion, and observed fact +- Consequence: concrete behavioural or code-health impact ``` -End with `## Summary` and the finding count for each axis. When neither axis contains a material finding, say the fixed change is clean plainly. +End with `## Summary` and the finding count for each completed axis. Do not repair the code, adjudicate findings, issue a verification or approval verdict, publish review comments, or alter pull-request state. diff --git a/skills/code-review/agents/openai.yaml b/skills/code-review/agents/openai.yaml index a243ba1..4f4a4fa 100644 --- a/skills/code-review/agents/openai.yaml +++ b/skills/code-review/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Code Review' - short_description: 'Review code against standards and specification' + short_description: 'Independently diagnose a fixed code change' policy: allow_implicit_invocation: true diff --git a/skills/code-review/references/CODE-SMELLS.md b/skills/code-review/references/CODE-SMELLS.md index 38e2b5d..3185738 100644 --- a/skills/code-review/references/CODE-SMELLS.md +++ b/skills/code-review/references/CODE-SMELLS.md @@ -1,6 +1,6 @@ # Fowler Code Smells -Use this complete second-edition catalogue as a diagnostic work aid for the Standards inspection. Each cue identifies a code shape to investigate, not a finding by itself. Test the relevant benign interpretation against the scoped code, repository authorities, and concrete consequence. +Load this complete second-edition catalogue only when a concrete maintainability shape needs recognised diagnostic vocabulary. Each cue identifies a code shape to investigate, not a finding or authority by itself. Test the relevant benign interpretation against the scoped code, repository authorities, and concrete consequence. 1. **Mysterious Name** — An identifier does not communicate its role, meaning, or unit in the surrounding domain. A repository-defined term or tightly conventional local name may already be precise. 2. **Duplicated Code** — Materially similar behaviour appears in multiple locations that may need to evolve together. Superficially similar code may represent different responsibilities or change for different reasons. From d7976a2642dfef8c53a3625198860a0e0ffdb1e3 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 21:33:44 +0100 Subject: [PATCH 13/79] feat(maintain-agents): enforce lean repository guidance --- SKILLS_PLAN.md | 2 +- skills/maintain-agents/SKILL.md | 22 ++++++++++------------ skills/maintain-agents/agents/openai.yaml | 2 +- 3 files changed, 12 insertions(+), 14 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9cbd9b0..bf1ed44 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -399,7 +399,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `tdd` skill - [x] Create `verify-change` skill - [x] Update `code-review` skill -- [ ] Update `maintain-agents` skill +- [x] Update `maintain-agents` skill - [ ] Update `define-product` skill - [ ] Create `define-architecture` skill - [ ] Create `specify-feature` skill diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index 038ca5f..badc1da 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -1,6 +1,6 @@ --- name: maintain-agents -description: Creates and compresses lean root AGENTS.md files. Use when initializing or improving repository-wide agent guidance. +description: Maintains lean root AGENTS.md files containing confirmed repository-wide guidance. Use when creating, updating, or compressing project-wide agent instructions. metadata: invocation: user disable-model-invocation: true @@ -8,26 +8,24 @@ disable-model-invocation: true # Maintain AGENTS.md -**Progressive Disclosure** keeps root `AGENTS.md` guidance limited to behaviour every repository task needs while narrower instructions remain discoverable on demand. +**Progressive Disclosure** keeps root `AGENTS.md` guidance limited to behaviour every repository task needs while narrower instructions remain discoverable only when relevant. ## Process -### 1. Inspect the instruction surface +### 1. Establish the instruction surface -Locate the repository root. Read the root `AGENTS.md`, applicable instruction layers, and enough manifests, task runners, CI, and contributor documentation to recover each rule's intent and identify project checks. The target, instruction chain, rules, and check entry points are explicit. +Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Inspect repository evidence only far enough to recover each rule's intent, determine whether a candidate is inferable, and verify referenced commands. Invoke `$elicit` when authority, scope, or conflicting guidance requires a material user decision. The authorised candidates and active constraints are explicit. ### 2. Allocate the guidance -Classify every existing and proposed instruction by runtime scope. Retain only concise behaviour governing the whole repository, plus the required correction and completion instructions. Report useful narrower guidance with its smallest discoverable owner: an invocable skill for reusable workflows, a scoped instruction file for directory rules, or executable enforcement for mechanical constraints. Leave destinations unchanged. Discard stale guidance, rationale, boilerplate, repository description, personal preferences, and task-, component-, or workflow-specific instructions. Every retained rule earns its permanent context cost. +Keep a rule at root only when it is confirmed, cannot be inferred reliably from repository evidence, and governs the repository as a whole. Ensure the root file itself states the immediate correction loop: when a user correction establishes a reusable repository-wide rule, ask whether to add it to `AGENTS.md`. Apply the same retention gate to checks, tools, preferences, and overrides; their existence does not authorise a new root rule. -### 3. Establish the required guidance +Remove stale, inferable, explanatory, or narrower guidance. Report useful narrower guidance with its smallest discoverable owner—an invocable skill for reusable workflows, a scoped instruction file for directory rules, or executable enforcement for mechanical constraints—and leave that destination unchanged. Every candidate is retained, displaced, or removed for an explicit reason. -Ensure the file states: `Ask immediately whether to add a reusable repository-wide rule to AGENTS.md when a user correction establishes it.` When project checks exist, add one instruction to run them after implementation and before handoff. Prefer one canonical aggregate command covering the configured suites; otherwise list every applicable individual command. Invoke `$elicit` when candidates materially differ or a command has unusual external effects. Omit the instruction only when no project checks exist. The correction loop and completion commands are explicit. +### 3. Write the root instructions -### 4. Write the root instructions +Create or rewrite only the root `AGENTS.md`. Use direct imperative lines and headings only when they navigate multiple instruction groups. Retain only words that require behaviour or preserve a necessary condition. State each repository policy once: merge instructions only when the same policy change would require both to change, and keep independently changeable rules separate. The root file is lean and unambiguous. -Create or rewrite only the root `AGENTS.md`. Use direct imperative lines. Remove headings unless they navigate multiple instruction groups. Apply **Minimalist Instruction** until every word changes behaviour or preserves a necessary condition. Apply **DRY** to meaning: when changing one rule requires changing multiple instructions, merge them into one authoritative expression. No semantic duplicates remain. +### 4. Verify and hand off -### 5. Verify and hand off - -Re-read every line for repository-wide scope, behavioural value, and semantic duplication. Verify each command exists; run the checks after implementation and before handoff when safe and applicable. Return the changed file, check results or limitations, displaced guidance with destinations, and unresolved conflicts. The user receives a lean verified root file and a visible account of displaced guidance. +Re-read every line for confirmed authority, non-inferability, repository-wide scope, behavioural value, conflicts, and semantic duplication. Verify each referenced command and reported destination. Inspect a command's definition before running it; run applicable checks required by active confirmed guidance only when their effects are safe and authorised, otherwise report the limitation. Return the changed root file, verification evidence or limitations, displaced guidance with destinations, and unresolved conflicts. The user receives a lean verified root file without unrequested changes to narrower owners. diff --git a/skills/maintain-agents/agents/openai.yaml b/skills/maintain-agents/agents/openai.yaml index 4755f87..6601cff 100644 --- a/skills/maintain-agents/agents/openai.yaml +++ b/skills/maintain-agents/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Maintain AGENTS.md' - short_description: 'Create or compress repository-wide agent guidance' + short_description: 'Maintain confirmed repository-wide agent guidance' policy: allow_implicit_invocation: false From fe9f91268da74322aa36c51c2ce936695fed6b91 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 21:35:52 +0100 Subject: [PATCH 14/79] Reset implementation checklist --- SKILLS_PLAN.md | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index bf1ed44..a188ba2 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -391,15 +391,15 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ### Implementation checklist - [x] Complete quality-harness research -- [x] Update `elicit` skill -- [x] Update `research` skill -- [x] Create `prototype` skill -- [x] Update `maintain-context` skill -- [x] Update `modular-design` skill -- [x] Update `tdd` skill -- [x] Create `verify-change` skill -- [x] Update `code-review` skill -- [x] Update `maintain-agents` skill +- [ ] Update `elicit` skill +- [ ] Update `research` skill +- [ ] Create `prototype` skill +- [ ] Update `maintain-context` skill +- [ ] Update `modular-design` skill +- [ ] Update `tdd` skill +- [ ] Create `verify-change` skill +- [ ] Update `code-review` skill +- [ ] Update `maintain-agents` skill - [ ] Update `define-product` skill - [ ] Create `define-architecture` skill - [ ] Create `specify-feature` skill From 1d8610f6102d3cc5346ee8883b8e82bb08f401fc Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 21:47:57 +0100 Subject: [PATCH 15/79] feat(elicit): tighten shared understanding --- SKILLS_PLAN.md | 2 +- skills/elicit/SKILL.md | 16 ++++++++-------- 2 files changed, 9 insertions(+), 9 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index a188ba2..43c7a26 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -391,7 +391,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ### Implementation checklist - [x] Complete quality-harness research -- [ ] Update `elicit` skill +- [x] Update `elicit` skill - [ ] Update `research` skill - [ ] Create `prototype` skill - [ ] Update `maintain-context` skill diff --git a/skills/elicit/SKILL.md b/skills/elicit/SKILL.md index 2e41b1c..1ec0564 100644 --- a/skills/elicit/SKILL.md +++ b/skills/elicit/SKILL.md @@ -1,6 +1,6 @@ --- name: elicit -description: Elicits material user-held information and user-authoritative decisions one question at a time. Use when shared understanding requires the user's knowledge, preferences, constraints, or direction. +description: Elicits material user-held information and user-authoritative decisions one question at a time until confirmed. Use when shared understanding depends on the user, not safely discoverable facts. metadata: invocation: model disable-model-invocation: false @@ -8,30 +8,30 @@ disable-model-invocation: false # Elicit -Uses an internal **decision tree** to resolve material information and decisions, reaching **theoretical saturation** when further questioning cannot change the shared understanding. +Uses an internal **decision tree** to resolve only the material branches of shared understanding, then applies adapted **theoretical saturation** and obtains explicit confirmation of one complete synthesis. ## Process ### 1. Establish the basis -Inspect the request, conversation, caller context, and accessible task-scoped evidence before questioning. Derive available facts and identify the user's intended outcome. Keep inspection read-only; when necessary fact discovery would change state, ask exactly one question to obtain explicit authority before performing it. The outcome and factual basis are explicit. +Inspect the request, conversation, caller context, and accessible task-scoped evidence before questioning. Derive discoverable facts and identify the user's intended outcome. Keep inspection read-only. When a required fact could be discovered only through a state change, make whether the caller should perform that lookup the active authority decision and leave the lookup to the caller. The outcome and factual basis are explicit. ### 2. Maintain the decision tree -Build and continually recompute an internal, dependency-ordered decision tree of unresolved user-held information and user-authoritative decisions. Open a branch only when its answer could materially change the outcome, boundaries, requirements, constraints, permissions, trade-offs, risks, or success conditions. Preserve compatible answers as branches open, close, become irrelevant, or reopen. Select the highest-impact unresolved branch whose prerequisites are resolved. One active branch is explicit. +Build and continually recompute an internal, dependency-ordered decision tree covering only material outcome, scope, terminology, inputs, outputs, prerequisites, dependencies, constraints, flows, exceptions, permissions, risks, trade-offs, and success conditions. Open a branch only when runtime evidence or an answer exposes it and resolving it could materially change the shared understanding. Preserve compatible answers as branches open, close, become irrelevant, or reopen. Select the highest-impact unresolved branch whose prerequisites are resolved. Keep the tree internal; one active issue is explicit. ### 3. Ask one question -Ask exactly one explicit question per turn in natural conversational language. Explain why an information question matters. For a decision, present only viable positions and their decisive trade-offs, recommending one when the outcome, evidence, applicable conventions, consequences, and prior answers provide a defensible basis. Leave room for the user's own answer and treat the user as the sole authority over their information and decisions. The active branch has one explicit response. +Ask exactly one explicit question about the active issue, once per turn, in natural conversational language. Explain why an information question matters. For a decision, present only viable positions and their decisive trade-offs, recommending one only when the outcome, evidence, applicable conventions, consequences, and prior answers provide a defensible basis. Leave room for the user's own answer and treat the user as the sole authority over their information and decisions. The active issue has one explicit response. ### 4. Adapt and validate -Test the response internally against the outcome, prior answers, assumptions, implications, scenarios, edge cases, obstacles, conflicts, and consequences. Advance the tree when the response is clear and consistent. When a material ambiguity or conflict remains, adapt the active question through plain-language restatement, clearer choices and trade-offs, examples, scenarios, or decomposition into prerequisite questions resolved one at a time. Recombine their answers and reopen every affected branch. The active branch is resolved consistently. +Test each response internally against the root goal, previous decisions, scenarios, counterexamples, conflicts, and consequences. Keep material ambiguity open. When the response is unclear or conflicting, or the user cannot answer, scaffold the same issue through plain-language restatement, clearer choices and trade-offs, examples, scenarios, or prerequisite questions resolved one at a time. Do not offer an escape hatch or decide for the user. Recombine the answers and reopen every affected branch. The active issue is resolved consistently. ### 5. Reach theoretical saturation -Repeat steps 2–4 until every material branch is resolved. Then make a complete materiality and consistency pass across the outcome, scenarios, obstacles, conflicts, consequences, and edge cases. Theoretical saturation is reached only when the pass adds or reopens no branch whose answer could materially change the shared understanding. The decision tree is saturated. +Repeat steps 2–4 until every material branch is resolved. Then make an adapted theoretical-saturation pass across the tree, scenarios, counterexamples, conflicts, and consequences. Saturation is reached only when the pass adds or reopens no branch whose answer could materially change the shared understanding. The decision tree is saturated. ### 6. Confirm shared understanding -Present one concise, self-contained synthesis of the outcome, relevant facts, decisions, boundaries, constraints, and observable success conditions, then ask exactly one question for explicit agreement. Obtain a new response to that complete synthesis; only this response completes elicitation. Treat a correction or rejection as new evidence, reopen every affected branch, and continue from step 2 until saturation returns before presenting the revised synthesis. Return only the user-confirmed synthesis to the caller without performing downstream work. +Present one concise, self-contained synthesis of the outcome, relevant facts, decisions, scope, constraints, and observable success conditions, then ask exactly one question for explicit agreement. Only a new explicit user response confirming the complete synthesis completes elicitation; selecting an earlier option does not. Treat a correction or rejection as new evidence, reopen every affected branch, and continue from step 2 until saturation returns before presenting the revised synthesis. Return only the confirmed synthesis; the caller retains mutation, persistence, verification, downstream action, and stopping ownership. From 290823bc119b4c21e3b9394516da1b996a3fdfd5 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 21:54:24 +0100 Subject: [PATCH 16/79] feat(research): tighten evidence handoff --- SKILLS_PLAN.md | 2 +- skills/research/SKILL.md | 14 +++++++------- skills/research/agents/openai.yaml | 2 +- 3 files changed, 9 insertions(+), 9 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 43c7a26..f1fc9b3 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -392,7 +392,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill -- [ ] Update `research` skill +- [x] Update `research` skill - [ ] Create `prototype` skill - [ ] Update `maintain-context` skill - [ ] Update `modular-design` skill diff --git a/skills/research/SKILL.md b/skills/research/SKILL.md index 181599a..ccb847e 100644 --- a/skills/research/SKILL.md +++ b/skills/research/SKILL.md @@ -1,6 +1,6 @@ --- name: research -description: Researches material questions using critically appraised sources and persists verified cited reports. Use when a trustworthy answer must survive the session. +description: Researches one material question through critically appraised evidence and persists a verified cited report. Use when the answer must survive the current conversation. metadata: invocation: model disable-model-invocation: false @@ -8,27 +8,27 @@ disable-model-invocation: false # Research -Turns a scoped material question into a trustworthy report while a fresh agent contains the investigation context. +Answers one material question with critically appraised evidence and persists the result as a verified cited report. ## Process ### 1. Define the research contract -Define the research question, intended use, scope, exclusions, currency needs, and domain evidence constraints. Inspect task-relevant repository context and `docs/research/` for related reports before searching. Resolve ambiguity that could materially change the investigation. When the contract requires a formal research, appraisal, or reporting method, consult its current authoritative guidance, apply the controls the investigation can satisfy, and disclose every unmet prerequisite instead of claiming conformance. The research contract and applicable prior evidence are explicit. +Define the research question, intended downstream use, scope, currency needs, and evidence constraints. Inspect task-local and project evidence, including related reports in `docs/research/`, before discovery. Resolve available facts and expose any material missing scope or evidence constraint rather than silently widening the investigation. The fixed research contract and applicable prior evidence are explicit. ### 2. Assign the investigation -Give one fresh agent the complete research contract, relevant repository context, evidence rules, and report contract. It owns discovery, appraisal, synthesis, and report writing; the caller later verifies and reads the finished report. The investigation context remains outside the caller. +Give one fresh research agent the fixed contract, relevant runtime evidence, evidence rules, and report contract without a desired conclusion. It owns discovery, critical appraisal, synthesis, and report writing as one evidence chain. The caller later reads and verifies the finished report. ### 3. Discover and appraise the evidence -Identify the material claims the answer must establish and select the source type capable of establishing each one. Trace every material claim to the source that owns it, such as a standard, official documentation or source code, an original study or dataset, or an originator's work. Rigorous syntheses may support discovery and interpretation; unverified commentary may locate, interpret, or challenge original evidence but carries no independent evidential weight. +Identify the material claims the answer must establish and select source types capable of establishing each one. Prefer primary and high-authority sources; use secondary sources for discovery or when they are the appropriate evidence for the claim. Trace every material claim to citations the caller can verify. Apply **critical appraisal** to determine what each source can establish and the trust it warrants. Assess authority and access to the fact, methodological or technical validity, currency, applicability, completeness, incentives or bias, and consistency with the underlying evidence. Treat an originator as authoritative for the method they define, not as proof that the method produces a claimed effect. When a material source participates in a citation graph, apply **citation searching** backward to cited originals and forward to later validation, correction, extension, or criticism; keep it supplementary to direct discovery. The evidence set is relevant, current enough for the question, and traceable. ### 4. Synthesize the findings -Corroborate material conclusions through genuinely independent evidence routes, treating sources that repeat one upstream claim as one route. Seek contrary evidence, plausible alternatives, and conditions that would change the conclusion; distinguish missing evidence from evidence against. Separate evidence, inference, conflict, and unknowns, and qualify conclusions when the evidence cannot support a stronger answer. Stop discovery when each material claim is supported or explicitly unresolved and further trustworthy source work is unlikely to change the answer. Every material finding is proportionate to the evidence. +Corroborate material conclusions through genuinely independent evidence routes, treating sources that repeat one upstream claim as one route. Seek contrary evidence, plausible alternatives, and conditions that would change the conclusion; distinguish missing evidence from evidence against. Separate supported evidence, inference, conflicts, contrary evidence, and unknowns. When credible evidence is insufficient, give an honest limited or unresolved conclusion. Stop discovery when each material claim is supported or explicitly unresolved and further trustworthy source work is unlikely to change the answer. Every material finding is proportionate to the evidence. ### 5. Write the research report @@ -36,4 +36,4 @@ Persist the result at `docs/research/YYYYMMDD-{research-title}.md` using [the re ### 6. Verify and hand off -The caller verifies that every material claim is supported by the cited source, evidence routes are genuinely independent, contrary evidence and uncertainty are visible, every link and relative report path resolves, and the recorded method makes the investigation inspectable. Return the report path, concise concrete findings, and unresolved limitations. The report owns the evidence; the caller owns every resulting decision or downstream mutation. +The caller reads the report and verifies that every material claim is supported by the cited source, evidence routes are genuinely independent, conflicts, contrary evidence, and uncertainty are visible, every link and relative report path resolves, and the recorded method makes the investigation inspectable. Return the report path, concise conclusion, conflicts and limitations, and verification performed. Stop there: the report owns the evidence, while the caller owns every resulting product, methodology, architecture, implementation, or other downstream decision and mutation. diff --git a/skills/research/agents/openai.yaml b/skills/research/agents/openai.yaml index f9f37c9..57e8b6f 100644 --- a/skills/research/agents/openai.yaml +++ b/skills/research/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Research' - short_description: 'Research evidence into verified cited reports' + short_description: 'Research one question into a verified cited report' policy: allow_implicit_invocation: true From a1959bdd5be1c6197ac82921866bafab545e9c8e Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:03:56 +0100 Subject: [PATCH 17/79] feat(prototype): enforce disposable visual experiments --- SKILLS_PLAN.md | 2 +- skills/prototype/SKILL.md | 24 ++++++++++++------------ 2 files changed, 13 insertions(+), 13 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index f1fc9b3..b65533d 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -393,7 +393,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill - [x] Update `research` skill -- [ ] Create `prototype` skill +- [x] Create `prototype` skill - [ ] Update `maintain-context` skill - [ ] Update `modular-design` skill - [ ] Update `tdd` skill diff --git a/skills/prototype/SKILL.md b/skills/prototype/SKILL.md index f9456db..1a6ed84 100644 --- a/skills/prototype/SKILL.md +++ b/skills/prototype/SKILL.md @@ -1,6 +1,6 @@ --- name: prototype -description: Creates disposable visual artefacts that resolve one bounded uncertainty. Use when seeing, comparing, or interacting would answer a question better than discussion alone. +description: Creates disposable visual experiments that answer one bounded question. Use when seeing, comparing, or interacting with an artefact would resolve uncertainty better than discussion. metadata: invocation: model disable-model-invocation: false @@ -8,34 +8,34 @@ disable-model-invocation: false # Prototype -Creates a temporary visual experiment that makes one uncertainty inspectable, returns what it established, and leaves no prototype code, files, records, or external effects behind. +Creates the least elaborate temporary visual experiment that can answer one bounded question, then returns the evidence and removes all prototype material. ## Process -### 1. Bound the question +### 1. Bound the experiment -Inspect the request, caller outcome, repository guidance, relevant product and implementation evidence, intended audience, and safely available data. State one uncertainty and the observations that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the comparison criteria and keep the decision with the user. Invoke `$elicit` when material framing information or a user-authoritative choice remains unresolved. When a visible or explorable artefact cannot discriminate the question, return that boundary to the caller without prototyping. The question, audience, decision authority, and discriminators are explicit. +Inspect the request, caller outcome and authority, repository guidance, relevant authentic runtime context, intended evaluators, and safely available data. State one question and the observable reaction, comparison, or result that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the decision criteria and keep the choice with the user. Invoke `$elicit` only when user-held information is needed to bound this visual question; return broader discovery to the caller. When a visual artefact cannot discriminate the question, stop and return that boundary without prototyping. The question, audience, criteria, decision authority, and discriminators are explicit. -### 2. Select the representation +### 2. Confirm presentation and disposal -Apply **prototypes as filters and manifestations**: identify the qualities the artefact must expose, what it will omit, and whether an omission could distort the answer; then choose the simplest trustworthy material, resolution, scope, and project context. Use a static artefact when appearance or explanation is sufficient and interaction when behaviour, sequence, state, or experience matters. Build one focused artefact for feasibility or explanation. For comparative design, place two or three materially different alternatives in the same artefact under the same content, data, and constraints; vary structure, hierarchy, or primary interaction rather than decoration alone. The selected form exposes the decisive qualities without implying completeness. +Inspect the current project state, available rendering or viewing capability, and a safe reversible workspace. Confirm a real presentation path that can expose a live view or rendered capture directly to the user, and record the pre-existing state needed to verify cleanup. If the result cannot be rendered and presented, stop before building, clean any preliminary attempt, and report the missing capability. Identify every file, route, component, process, dependency, fixture, branch, data source, and external effect the experiment could touch. The experiment has a presentation path and an exact disposable boundary before construction begins. -### 3. Establish the disposable boundary +### 3. Choose the representation -Inspect the current project state and identify every file, route, component, process, or data source the prototype will temporarily affect. Use authentic application context and safe read-only data when they materially affect the answer; otherwise work in an isolated temporary location. Temporary adjustments to an existing surface are permitted only when their exact removal can be verified without disturbing pre-existing work. Stub external writes and irreversible actions, and clearly distinguish simulated behaviour from implemented capability. The prototype has an isolated, reversible boundary with no real external mutation. +Choose the simplest trustworthy sketch, static mock-up, interactive stub, or other visual manifestation that exposes the discriminators. Identify the qualities it must show, what it may omit, and whether an omission could distort the answer. Use interaction only when behaviour, sequence, state, or experience is material. When comparing alternatives, show the materially different options under the same content, data, and constraints. The selected form exposes the decisive qualities without implying production completeness. ### 4. Build the artefact -Implement only the fidelity and behaviour needed to expose the discriminators. Reuse project-native layout, components, styling, and realistic content density when they affect interpretation, but do not add production hardening, reusable abstractions, tests, or unrelated infrastructure. When state or sequence matters, include guided representative and adverse scenarios and allow direct exploration when action order could reveal unexpected behaviour. Keep all prototype material clearly identifiable as temporary. One faithful disposable artefact is ready to inspect. +Build only the fidelity and behaviour needed to expose the discriminators. Use authentic project layout, components, styling, content, or safe read-only data only when they materially affect interpretation and remain behind the reversible boundary; otherwise use isolated temporary material. Keep live systems read-only, stub external writes and irreversible actions, do not expose credentials, and distinguish simulated behaviour from implemented capability. Do not widen the experiment into production architecture, reusable abstractions, tests, hardening, or unrelated polish. One faithful disposable artefact is ready to render. ### 5. Render and present it -Render the artefact in its intended medium and inspect its visual output and relevant interactions. Verify that the question, alternatives or scenarios, simulated boundaries, and decisive observations are understandable to the intended audience; correct anything that could make the comparison misleading. Expose a live view or rendered capture directly in the conversation before cleanup and give the user the simplest controls needed to compare or explore it; a source listing or path that will disappear before it can be viewed is insufficient. When the environment cannot render and expose the result, remove the prototype and report the exact missing capability. The user evaluates the rendered artefact rather than its source. +Render the artefact in its intended medium and inspect its visual output and relevant interactions. Verify that the question, alternatives or scenarios, simulated boundaries, omissions, and decisive observations are understandable to the intended audience; correct anything that could make the experiment misleading. Present the live view or rendered capture directly to the user before cleanup, keep it available while evaluation requires it, and provide only the controls needed to compare or explore it. Source inspection or a path that will disappear before viewing is not presentation. The user has an opportunity to inspect the visual result. ### 6. Iterate to a conclusion -Revise the same artefact while comparison or exploration continues to produce useful evidence. Let the agreed observations settle an objective question and obtain the user's choice for subjective product or design direction. Stop when the uncertainty is resolved or when further useful iterations cannot distinguish the options. The experiment has a supported conclusion or an exact remaining uncertainty and missing discriminator. +Revise the same artefact only when another pass could change the answer to the bounded question. Let the agreed observations settle an objective question and obtain the user's choice for a subjective one. Stop when the question is answered or further useful iteration cannot discriminate it. If the user instead requests a durable or production outcome, stop the prototype and return that changed outcome to the appropriate caller or downstream skill rather than retaining or converting the experiment. The experiment has a supported conclusion or an exact remaining uncertainty and missing discriminator. ### 7. Remove and return -After the rendered result has been exposed and the experiment has concluded, remove the artefact and revert only the prototype-specific project adjustments, then verify that no prototype material remains and all pre-existing work is intact. Do not promote, commit, document, deploy, or otherwise retain any prototype code or artefact. Return the concise conclusion to the current conversation or caller; when inconclusive, return the remaining uncertainty, evidence gathered, and missing discriminator instead. The caller receives the result with no durable prototype residue. +After the user has had the opportunity to inspect the result and the experiment has concluded, remove every prototype artefact, dependency, fixture, branch, process, and temporary change. Revert only prototype-specific adjustments, then compare with the recorded starting state and verify that all pre-existing work remains intact. Preserve no prototype report, code, data, branch, or capture by default. Return the answer or explicitly inconclusive result in the current session or caller handoff with the evidence, limitations, remaining uncertainty, and missing discriminator where applicable. The caller retains every product, specification, architecture, implementation, and durable-update decision; it receives the result with no prototype residue. From ed796af57692b40a7599a07631825f2ef31b9f8b Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:13:28 +0100 Subject: [PATCH 18/79] feat(maintain-context): constrain durable project language --- SKILLS_PLAN.md | 2 +- skills/maintain-context/SKILL.md | 24 ++++++++++--------- .../maintain-context/assets/adr-template.md | 2 +- 3 files changed, 15 insertions(+), 13 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index b65533d..bb3a394 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -394,7 +394,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `elicit` skill - [x] Update `research` skill - [x] Create `prototype` skill -- [ ] Update `maintain-context` skill +- [x] Update `maintain-context` skill - [ ] Update `modular-design` skill - [ ] Update `tdd` skill - [ ] Create `verify-change` skill diff --git a/skills/maintain-context/SKILL.md b/skills/maintain-context/SKILL.md index 24fa12c..b726c30 100644 --- a/skills/maintain-context/SKILL.md +++ b/skills/maintain-context/SKILL.md @@ -8,35 +8,37 @@ disable-model-invocation: false # Maintain Context -**Ubiquitous Language** keeps project-specific terminology consistent across conversation, documentation, and code. Maintain one root `CONTEXT.md` as a concise glossary during ordinary work; preserve only rare qualifying architecture decisions separately. +Adapted **Ubiquitous Language** keeps confirmed project-specific terminology consistent. Maintain one concise root `CONTEXT.md` glossary inline during ordinary work and, exceptionally, preserve the rationale for a qualifying accepted decision in an ADR. ## Process -### 1. Notice project language +### 1. Inspect the language -Act whenever a material project term is introduced, changed, misused, or contradicted during the caller's work. Compare its meanings in the active discussion and root `CONTEXT.md`, then inspect only the relevant code or authoritative documents needed to expose a conflict. State the conflicting meanings and their sources plainly. The term and any disagreement are explicit. +When project language is introduced, changed, misused, or contradicted, inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the code or product evidence needed to understand that language. Challenge inconsistent uses of established terms. The candidate term, current meaning, and any contradiction are explicit without loading unrelated project context. ### 2. Resolve the meaning -Challenge an inconsistent, vague, or overloaded meaning immediately. Propose one precise canonical term and test its project-specific meaning with concrete examples or boundary cases suited to the disagreement. Treat every source as evidence rather than automatic authority and obtain confirmation from the user or the caller's already-confirmed outcome. When the meaning remains unresolved, keep the conflict explicit and leave `CONTEXT.md`, including any existing entry, unchanged. One meaning is confirmed or its uncertainty remains visible. +Accept terminology already established by authoritative project evidence or explicitly agreed with the user. A caller's agreed in-scope update supplies authority to persist that meaning without another permission prompt. When a material meaning, conflict, rename, or deprecation remains unresolved, return it to an active `$elicit` caller or invoke `$elicit` when none is active, and leave durable context, including the existing entry, unchanged until agreement. One meaning is confirmed or the unresolved decision returns to the caller. ### 3. Update the glossary inline -As soon as a meaning is confirmed, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md) when the first term resolves. Give each canonical term one authoritative entry and define its project-specific meaning in one or two concise sentences, including only what is needed to explain the term. Group entries under descriptive headings when natural clusters emerge; otherwise keep one flat language list. +As confirmed language arises, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and the document's local format; edit only affected entries and prefer changing an existing entry to adding a duplicate. Reconcile a confirmed change across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence, and remove obsolete wording only when its replacement is confirmed and losing it will not obscure that evidence. -Change no code, specification, product, architecture, or other project artifact. Return every remaining conflict to the caller or artifact owner. The glossary reflects the confirmed language and other authorities retain their ownership. +Give each canonical term one short plain-language definition, normally one or two sentences, using project-native wording. Include only durable project-specific terms, distinctions, roles, boundaries, or aliases that future agents need to interpret the project consistently; group them naturally when useful. `CONTEXT.md` explains what the language means. It does not store skill procedures, generic guidance, task status, roadmaps, volatile implementation facts, architecture inventories, schemas or endpoints, acceptance criteria, or anything reliably discoverable from code. Keep product intent in `PRODUCT.md`, system structure in `ARCHITECTURE.md`, feature behaviour in its specification, and implementation detail in code. The glossary is current without duplicating another authority. -### 4. Offer a rare ADR +### 4. Gate a rare ADR -Consider an architecture decision record only for an already accepted codebase decision that passes every gate: +Consider an ADR only for an accepted technical or architectural decision that passes every gate: - changing it later has meaningful cost; - a future reader would find it surprising without context; and - viable alternatives created a genuine trade-off; and -- neither `ARCHITECTURE.md` nor a feature record already explains it adequately. +- no existing durable authority is the better owner, including `PRODUCT.md`, `ARCHITECTURE.md`, a feature document, or a dedicated governance record. -When every gate passes, offer the ADR and let the user decide whether to record it. A decision that misses any gate remains routine and produces no ADR. The exceptional record has explicit user agreement. +When every gate passes, obtain the user's explicit agreement to preserve the rationale durably. Explicit agreement recorded in the relevant conversation or caller handoff satisfies this gate without another prompt. A decision that misses a gate, lacks that agreement, or is routine or discoverable produces no ADR. The exceptional record has a confirmed owner and purpose. ### 5. Record and verify the result -When the user accepts the offer, create `docs/adr/` lazily and write the next record from the [ADR template](assets/adr-template.md). Derive the next four-digit sequence from filenames alone and name it `NNNN-decision-shaped-slug.md`. Use exactly `Decision`, `Context`, and `Ramifications` as content sections, in that order, and link only the supporting material needed to understand the choice. In every path, re-read each changed artifact against the confirmed language, accepted decision, and relevant evidence. Report changed files and each unresolved conflict with its owning artifact. The glossary and any qualifying ADR are concise, verified, and ready for the caller to continue. +When the user agrees, follow the project's existing ADR convention; when none exists, create `docs/architecture/decisions/` lazily. Write the next record from the [ADR template](assets/adr-template.md) and name it `NNNN-short-title.md`. Use `0001` when no sibling filename matches `NNNN-*.md`; otherwise add one to the maximum four-digit `NNNN` among matching siblings. Keep exactly three substantive sections in order: `Decision`, `Context`, and `Ramifications`. Record the accepted choice, why it was needed, only the rejected or constrained alternatives necessary to explain its trade-offs, and its material consequences. + +Re-read each changed context artifact against the confirmed meaning, accepted decision, and relevant evidence. Return the changed files or why no durable update was warranted, plus unresolved terminology for the caller to resolve. Change no caller-owned product, architecture, feature, code, external state, or verification outcome. The caller receives a concise context update and retains its surrounding outcome. diff --git a/skills/maintain-context/assets/adr-template.md b/skills/maintain-context/assets/adr-template.md index fbf0562..a9f181a 100644 --- a/skills/maintain-context/assets/adr-template.md +++ b/skills/maintain-context/assets/adr-template.md @@ -6,7 +6,7 @@ ## Context -{Explain the problem, forces, and serious alternatives needed to understand the decision.} +{Explain why the decision was needed and only the rejected or constrained alternatives necessary to understand its trade-offs.} ## Ramifications From 2b9349e828e200e8a5035c12fd8d5de038b6e2b6 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:20:26 +0100 Subject: [PATCH 19/79] feat(modular-design): clarify passive structural constraints --- SKILLS_PLAN.md | 2 +- skills/modular-design/SKILL.md | 16 ++++++++++------ 2 files changed, 11 insertions(+), 7 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index bb3a394..01e370b 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -395,7 +395,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `research` skill - [x] Create `prototype` skill - [x] Update `maintain-context` skill -- [ ] Update `modular-design` skill +- [x] Update `modular-design` skill - [ ] Update `tdd` skill - [ ] Create `verify-change` skill - [ ] Update `code-review` skill diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index 44901fd..aa9dcc6 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -8,21 +8,21 @@ disable-model-invocation: false # Modular Design -Supplies structural constraints inside a caller's workflow by locating each material decision with a coherent owner, making the smallest complete change, and keeping caller knowledge proportionate to the capability provided. +Supplies structural findings and constraints inside a caller's workflow by locating each material decision with a coherent owner, making the smallest complete change, and keeping caller knowledge proportionate to the capability provided. ## Process ### 1. Fix the behaviour and evidence -Apply this standard inside the caller's workflow. Use the scoped behaviour, project guidance, architecture decisions, domain language, and relevant implementation evidence; explicit project authorities govern when they conflict with this baseline. Let the caller own inspection, design comparison, implementation, verification, and artifacts. Leave correctness, testing, security, performance, formatting, and general style to their applicable authorities rather than selecting their methods or outputs here. The behaviour, evidence, and material structural decision are explicit. +Start from confirmed scoped behaviour and runtime-available caller and project evidence: project guidance, architecture decisions, domain language, relevant source, contracts, and tests. Explicit project authorities and established language and framework conventions govern when they conflict with this baseline. Keep the assessment structural: ownership, information hiding, cohesion, coupling, contracts, dependencies, seams, invariants, policies, protocols, sequencing, side effects, and change propagation. Leave product authority, behavioural correctness, security, performance, accessibility, formatting, testing, and general style to their applicable authorities except where they materially shape a contract or boundary. The behaviour, evidence, and material structural decision are explicit. ### 2. Assign the affected knowledge -Identify the representation, invariant, policy, protocol, sequencing, or framework decision the behaviour introduces or changes. Keep it with its natural existing owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each material rule once, and let code that changes for the same evidenced reason change together. The affected knowledge has one coherent owner rather than being repeated across callers. +Identify the representation, contract, invariant, policy, protocol, sequencing, side effect, or framework decision the behaviour introduces or changes. Keep it with its natural existing owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each material rule once, and let code that changes for the same project-evidenced reason change together. The affected knowledge has one coherent owner rather than being repeated across callers. ### 3. Make the smallest coherent change -Deliver the complete scoped behaviour without speculative abstraction or unrelated cleanup. Permit a bounded enabling refactor only when the behaviour cannot fit cleanly without it; change no more surrounding structure than the new behaviour needs. Keep a stable local concrete dependency direct unless meaningful variation, isolation, observation, migration, or policy ownership justifies a boundary. +Deliver the complete scoped behaviour without speculative abstraction or unrelated cleanup. Permit a bounded enabling refactor only when the behaviour cannot fit coherently without it; change no more surrounding structure than the new behaviour needs. Keep a stable local concrete dependency direct unless meaningful variation, isolation, observation, migration, a technology boundary, or policy-mechanism separation justifies a seam. Realize the design through established project and framework idioms. Objects may own identity, state, invariants, and cohesive behaviour; functions, closures, structural types, language modules, components, and framework primitives are equally valid when they preserve the same ownership and contract. Introduce a language interface only when callers need a stable contract distinct from its realization, and use inheritance only for genuine substitutability or a required framework extension. The result is the smallest idiomatic change that completely delivers the behaviour. @@ -34,6 +34,10 @@ Apply **deep modules**: make the caller-visible contract markedly simpler than t ### 5. Check locality and applicable techniques -Trace plausible project-evidenced changes through the owner, contract, callers, and affected code. Reconsider the boundary when one owned decision would still require scattered changes, when several elements must always change together, or when only an unrelated part of the proposed owner would change. Do not substitute file size, class count, method count, dependency scores, or speculative agent-context savings for this evidence. +Trace plausible current-project changes through the owner, contract, consumers, and affected code. Reconsider the boundary when one owned decision would still require scattered changes, when several elements must always change together but are split, or when the proposed owner changes mostly for unrelated reasons. Do not substitute file size, class count, method count, dependency scores, abstract metrics, or speculative agent-context savings for this evidence. -Decide from the current evidence whether it presents a policy-to-mechanism dependency, an application-to-technology conversation, competing domain models, a need for controlled substitution or observation, or deterministic decisions entangled with effects. When none is present, keep [Modular Design Techniques](references/TECHNIQUES.md) unloaded and continue with the direct process; do not open it merely to confirm that a condition is absent. Otherwise read the reference and apply only the technique whose stated condition is present. The resulting structure keeps the owned knowledge and its likely changes local without introducing an unevidenced abstraction. +Load [Modular Design Techniques](references/TECHNIQUES.md) only when current evidence already presents one of its conditions: policy directly depends on a volatile mechanism; an application conversation needs independence from technologies, replacement, isolated execution, or a material external boundary; domain meanings or change contours compete; controlled substitution, observation, or migration is needed; or deterministic decisions are entangled with effects. Do not open the reference to discover whether a condition exists. Apply only the fitting technique. The resulting structure keeps the owned knowledge and its likely changes local without introducing an unevidenced abstraction. + +## Handoff + +Return the structural findings or constraints, their runtime evidence, the responsible owner and contract, the change-locality trace, any triggered technique, and unresolved structural uncertainty. The caller retains elicitation, requirements, architecture and review artifacts, implementation, tests, verification, external mutation, and the decision to continue or stop its workflow. From 53cc1e8c457bc402d80e5e1a59bfaf84d9a75e52 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:28:06 +0100 Subject: [PATCH 20/79] feat(tdd): enforce durable behavioural cycles --- SKILLS_PLAN.md | 2 +- skills/tdd/SKILL.md | 10 +++++----- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 01e370b..7ae5db4 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -396,7 +396,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `prototype` skill - [x] Update `maintain-context` skill - [x] Update `modular-design` skill -- [ ] Update `tdd` skill +- [x] Update `tdd` skill - [ ] Create `verify-change` skill - [ ] Update `code-review` skill - [ ] Update `maintain-agents` skill diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index f11293f..81a6cf6 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -12,23 +12,23 @@ Builds one observable behaviour at a time through red-green-refactor. Retained t ## Prerequisite -TDD applies when an existing runnable test suite can exercise the requested behaviour through a stable observable boundary. Otherwise return control with the exact missing condition; the caller owns any decision to create a test harness or reshape a public contract. Apply TDD to observable behaviour, including configuration with observable effects; leave documentation, configuration-only maintenance, and behaviour-preserving refactors with the caller. +TDD applies only when an existing runnable test suite can credibly exercise the requested behaviour through a stable observable boundary. Otherwise return control with the exact missing condition. Do not install or invent a test framework or add a harness to satisfy this prerequisite; new-project test infrastructure belongs in architecture and initial delivery work. Apply TDD to observable behaviour, including configuration with observable effects; leave documentation, configuration-only maintenance without an exercisable effect, and behaviour-preserving refactors with the caller. ## Process ### 1. Establish the baseline -Read repository instructions, identify the relevant test command, and run the existing relevant suite. Record baseline failures and continue only when the focused Red and Green signals remain distinguishable; otherwise return the blocker. Use **tracer bullets** to select the smallest end-to-end behaviour that produces one requested observable result. Apply caller-supplied modular constraints, or invoke `$modular-design` when the slice presents a new material structural-maintainability decision. Choose the narrowest boundary that exposes the result, hides structure allowed to change, and credibly exercises production behaviour. Identify a **test oracle**—a requirement, worked example, invariant, contract, trusted reference, accepted prior behaviour, or explicit domain decision—capable of disagreeing with the implementation; when none exists, return the missing behavioural decision. Consult [Test Quality](references/TEST-QUALITY.md) when the boundary, oracle, or proposed assertion needs deeper guidance. The baseline, behaviour, boundary, oracle, and applicable modular constraints are explicit. +Read repository instructions, caller-confirmed behaviour and acceptance evidence, applicable modular constraints, relevant production code, and current tests. Identify the test command and run the smallest relevant focused baseline. Record baseline failures without attributing a pre-existing failure to the change; continue only when the focused Red and Green signals remain distinguishable, otherwise return the blocker. Use **tracer bullets** to select the smallest end-to-end behaviour that produces one requested observable result. Apply caller-supplied modular constraints, or invoke `$modular-design` only when the slice presents a new material structural-maintainability decision. Choose the narrowest boundary that exposes the result, hides structure allowed to change, and credibly exercises production behaviour. Identify a **test oracle**—a requirement, accepted example, published protocol, independent invariant, trusted reference or separate model, accepted prior behaviour, or explicit domain decision—capable of disagreeing with the implementation; when none exists, return the missing behavioural decision. Consult [Test Quality](references/TEST-QUALITY.md) when the boundary, oracle, or proposed assertion needs deeper guidance. The baseline, behaviour, boundary, oracle, and applicable modular constraints are explicit. ### 2. Red -Add one focused test: arrange only the necessary state, act once through the selected boundary, assert the complete promised outcome, and clean up resources the test acquires. Name one plausible promised-behaviour defect that must make the test fail and one hidden structural change that must leave it unchanged. Keep internal collaborators real. When an uncontrollable boundary must be controlled or observed, choose the least powerful **Test Double** that supplies the required evidence; consult [Test Doubles](references/TEST-DOUBLES.md) before introducing a double or interaction assertion. +Add one focused test: arrange only the necessary state, act once through the selected boundary, assert the complete promised outcome and any material unchanged state, and clean up resources the test acquires. Name one plausible promised-behaviour defect that must make the test fail and one hidden structural change that must leave it unchanged. Keep internal collaborators real. When an uncontrollable boundary must be controlled or observed, choose the least powerful **Test Double** that supplies the required evidence; consult [Test Doubles](references/TEST-DOUBLES.md) before introducing a double or interaction assertion. For a bug, reproduce the incorrect behaviour; adopt an already-failing regression test only when it independently specifies the desired behaviour. Run the focused test and confirm that it fails for the expected behavioural reason. When it fails because of the test or environment, remain in Red: correct an in-scope defect or report the blocker, then rerun until the intended failure is observed. Meaningful red evidence exists before Green begins. ### 3. Green -Implement only enough production code to satisfy the behaviour, then run the focused test and relevant nearby tests. The new behaviour passes without speculative production code or hidden baseline failures. +Implement only enough production code to satisfy the behaviour using project and framework conventions, then run the focused test and relevant nearby tests. The new behaviour passes without speculative production code or hidden baseline failures. ### 4. Refactor @@ -36,4 +36,4 @@ Improve the production design behind the selected boundary while keeping behavio ### 5. Complete the cycles -Repeat Red, Green, and Refactor for each remaining behaviour, then run the complete relevant suite. Report the behaviours delivered, Red and Green evidence, refactors, commands, results, unresolved baseline failures, and other limitations. The requested behaviour and durable retained tests are verified; later verification and review remain with the caller. +Repeat Red, Green, and Refactor for each remaining confirmed behaviour, then run the complete relevant suite. Report the behaviours delivered, exact Red and Green evidence, refactors, commands and results, unresolved baseline failures, and other limitations. The requested behaviour and durable retained tests are implemented; the caller owns the overall work outcome, broader quality portfolio, later `$verify-change`, independent `$code-review`, commit or artefact decisions, and whether its workflow can stop. From 1c878ae971e56448acd652098ac4462cf8219b17 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:37:02 +0100 Subject: [PATCH 21/79] feat(verify-change): distinguish stale evidence --- SKILLS_PLAN.md | 2 +- skills/verify-change/SKILL.md | 5 +++-- 2 files changed, 4 insertions(+), 3 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 7ae5db4..cf73e0f 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -397,7 +397,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `maintain-context` skill - [x] Update `modular-design` skill - [x] Update `tdd` skill -- [ ] Create `verify-change` skill +- [x] Create `verify-change` skill - [ ] Update `code-review` skill - [ ] Update `maintain-agents` skill - [ ] Update `define-product` skill diff --git a/skills/verify-change/SKILL.md b/skills/verify-change/SKILL.md index 6fbe2b4..24d7be6 100644 --- a/skills/verify-change/SKILL.md +++ b/skills/verify-change/SKILL.md @@ -39,7 +39,8 @@ Record one evidence entry per claim with: - the exact claim; - the harness, its repository or procedure provenance, why it applies, and whether it is required or risk-triggered; - the execution boundary; -- `pass`, `fail`, `flaky/inconclusive`, or `not run`, with salient output, exit status, and artefact location where applicable; +- the observed result: `pass`, `fail`, `flaky-inconclusive`, or `not run`, with salient output, exit status, and artefact location where applicable; +- evidence integrity: `current`, or `stale` when post-basis drift applies; - the inputs, paths, states, rules, environments, or acceptance criteria actually exercised; - the limitation and residual risk; and - the in-scope resolution taken, or the planned work or user decision needed for stronger evidence. @@ -48,7 +49,7 @@ Preserve every retry, seed, failing counterexample, and contradictory observatio ### 5. Reinspect scope and harness integrity -Compare the final repository state and diff with the frozen basis. Account for changes to source, tests, snapshots, dependencies, generated artefacts, quality configuration, baselines, ignore lists, suppressions, and scripts. Distinguish declared temporary effects from material changes, but do not revert an unexpected change when ownership is uncertain. When any in-scope source or measurement-path change occurred after the scope was fixed, identify the paths, mark affected evidence stale, and stop without silently retargeting the run. The final evidence refers to the frozen change and an intact measurement path, or its invalidation is explicit. +Compare the final repository state and diff with the frozen basis. Account for changes to source, tests, snapshots, dependencies, generated artefacts, quality configuration, baselines, ignore lists, suppressions, and scripts. Distinguish declared temporary effects from material changes, but do not revert an unexpected change when ownership is uncertain. When any in-scope source or measurement-path change occurred after the scope was fixed, identify the paths, retain each observed result, set its evidence integrity to `stale`, treat it as unable to satisfy the affected claim, and stop without silently retargeting the run. The final evidence refers to the frozen change and an intact measurement path, or its invalidation is explicit. ### 6. Determine the verification verdict From 2108b57e964ca98f2129397400fd712be9eaaee1 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:48:26 +0100 Subject: [PATCH 22/79] feat(code-review): formalize independent inspection --- SKILLS_PLAN.md | 2 +- skills/code-review/SKILL.md | 12 ++++++------ 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index cf73e0f..9cbd9b0 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -398,7 +398,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `modular-design` skill - [x] Update `tdd` skill - [x] Create `verify-change` skill -- [ ] Update `code-review` skill +- [x] Update `code-review` skill - [ ] Update `maintain-agents` skill - [ ] Update `define-product` skill - [ ] Create `define-architecture` skill diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 468042a..b78992b 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -8,7 +8,7 @@ disable-model-invocation: false # Code Review -Independently diagnoses a fixed code change through isolated Spec and Standards reviews, returning evidence for the caller without verifying, approving, or changing the work. +A tailored **Software Formal Inspection** independently diagnoses a fixed code change through isolated Spec and Standards reviews, returning evidence for the caller without verifying, approving, or changing the work. ## Process @@ -22,7 +22,7 @@ Identify intended behaviour from the caller's fixed authority, such as a specifi Independently resolve applicable repository instructions, architecture decisions, language and framework policies, configured checks, and established local conventions. Read every changed file in full, relevant tests, affected contracts and consumers, and enough surrounding implementation to understand each change. Reuse relevant caller-supplied `$modular-design` constraints, or invoke `$modular-design` when ownership, boundaries, dependencies, contracts, or change propagation present a material structural concern. -Apply the universal Standards criteria directly: correctness and regression risk, scoped minimality, repository conventions, changed-test validity, maintainability, recognised smells, and architecture. Activate security, performance, accessibility, compatibility, resilience, concurrency, migration, or another specialist concern only when task evidence or project authority exposes it. Use ISO/IEC 25010 only when an authority adopts it. For an implicated supported-language or Web-security construct, consult the current official CERT rule or stable versioned OWASP ASVS requirement narrowly and include its exact applicability and exceptions. When a concrete maintainability shape still needs recognised diagnostic vocabulary, consult [Code Smells](references/CODE-SMELLS.md); a smell prompts investigation and is never finding authority by itself. The two axes have explicit, task-relevant evidence without a universal external checklist. +Apply the universal Standards criteria directly: correctness and regression risk; scoped minimality; language, framework, and repository conventions, including established helpers; changed-test validity and durability; maintainability and economy; recognised smells; and modular architecture. Activate security, performance, accessibility, compatibility, resilience, concurrency, migration, or another specialist concern only when task evidence or project authority exposes it. Use ISO/IEC 25010 only when an authority adopts it. For an implicated supported-language or Web-security construct, consult the current official CERT rule or stable versioned OWASP ASVS requirement narrowly and include its exact applicability and exceptions. When a concrete maintainability shape still needs recognised diagnostic vocabulary, consult [Code Smells](references/CODE-SMELLS.md); a smell prompts investigation and is never finding authority by itself. The two axes have explicit, task-relevant evidence without a universal external checklist. ### 3. Prepare the inspection packets @@ -39,7 +39,7 @@ The Standards packet directs its reviewer to: 1. understand every changed line and the necessary whole-file and system context; 2. follow affected contracts, states, data shapes, side effects, callers, and dependencies far enough to expose correctness and regression risks; 3. apply the universal and triggered criteria resolved in step 2; and -4. when tests changed, ask whether they detect a promised-behaviour defect, avoid coupling to hidden structure, remain deterministic and readable, and credibly predict the promised result. +4. when tests changed, ask whether they detect a promised-behaviour defect, survive behaviour-preserving changes to hidden structure, remain deterministic and readable, and credibly predict the promised result. Each packet tells the reviewer how to inspect its axis rather than supplying labels alone. @@ -53,13 +53,13 @@ Use consequence-based priority within each axis: `critical` for immediate data l ### 5. Preserve the independent results -Check that each reviewer returned the required fields, returning an incomplete result to that same reviewer for structural completion from the original packet. Reinspect the scoped revisions, diff, paths, and in-scope untracked content. When the frozen work product drifted, identify the changed paths, mark affected findings and clean results stale, and stop without silently retargeting the review. +Validate only that each reviewer returned the required fields, returning an incomplete result to that same reviewer for structural completion from the original packet. Reinspect the scoped revisions, diff, paths, and in-scope untracked content. When the frozen work product drifted, preserve each affected axis's frozen result, prefix it with `Status: stale` and the changed paths and basis mismatch, and stop without silently retargeting the review. Present Standards and Spec separately without substantive re-review, merging, deduplication, suppression, or cross-axis reranking. Preserve each reviewer's findings and ordering. The caller receives independent diagnostic evidence for its own adjudication. ## Handoff -State the exact frozen scope, intended-behaviour authority or user-confirmed absence, Standards authorities, any omitted or unperformed axis, every probe that could not run, and any scope drift. Return `## Standards` and `## Spec`; use `No findings.` for a clean axis and state when Spec was omitted or either axis was not performed. Format each finding as: +State the exact frozen scope, intended-behaviour authority or user-confirmed absence, Standards authorities, any omitted or unperformed axis, every probe run or omitted, and any scope drift. Return `## Standards` and `## Spec`; use `No findings.` only for a current clean axis, never for a stale clean result without its `Status: stale` prefix, and state when Spec was omitted or either axis was not performed. Format each finding as: ```markdown ### [priority] Concise finding @@ -68,4 +68,4 @@ State the exact frozen scope, intended-behaviour authority or user-confirmed abs - Consequence: concrete behavioural or code-health impact ``` -End with `## Summary` and the finding count for each completed axis. Do not repair the code, adjudicate findings, issue a verification or approval verdict, publish review comments, or alter pull-request state. +End with `## Summary` and the finding count for each completed axis, labelling an affected frozen count `stale`. The caller owns adjudication, remediation, re-review, verification, and any publication or pull-request decision. Do not repair the code, adjudicate findings, issue a verification or approval verdict, publish review comments, or alter pull-request state. From ce1cf4d68372d70715b286dc2a1bdaeb30cb6e10 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 22:56:32 +0100 Subject: [PATCH 23/79] feat(maintain-agents): keep repository guidance lean --- SKILLS_PLAN.md | 2 +- skills/maintain-agents/SKILL.md | 18 ++++++++++-------- skills/maintain-agents/agents/openai.yaml | 2 +- 3 files changed, 12 insertions(+), 10 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9cbd9b0..bf1ed44 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -399,7 +399,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `tdd` skill - [x] Create `verify-change` skill - [x] Update `code-review` skill -- [ ] Update `maintain-agents` skill +- [x] Update `maintain-agents` skill - [ ] Update `define-product` skill - [ ] Create `define-architecture` skill - [ ] Create `specify-feature` skill diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index badc1da..cb6567e 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -2,30 +2,32 @@ name: maintain-agents description: Maintains lean root AGENTS.md files containing confirmed repository-wide guidance. Use when creating, updating, or compressing project-wide agent instructions. metadata: - invocation: user -disable-model-invocation: true + invocation: model +disable-model-invocation: false --- # Maintain AGENTS.md -**Progressive Disclosure** keeps root `AGENTS.md` guidance limited to behaviour every repository task needs while narrower instructions remain discoverable only when relevant. +**Progressive Disclosure** keeps root `AGENTS.md` guidance limited to confirmed, non-inferable behaviour that governs the repository as a whole, while narrower guidance stays with its discoverable owner. Adapt **Minimalist Instruction** and **DRY** to preserve each behavioural policy once in the fewest words that retain its conditions. ## Process ### 1. Establish the instruction surface -Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Inspect repository evidence only far enough to recover each rule's intent, determine whether a candidate is inferable, and verify referenced commands. Invoke `$elicit` when authority, scope, or conflicting guidance requires a material user decision. The authorised candidates and active constraints are explicit. +Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Treat a caller handoff carrying an agreed preference as authority; do not ask for duplicate permission to record it. Inspect repository evidence only far enough to recover intent, scope, inferability, conflicts, and referenced commands. Invoke `$elicit` only when material authority, scope, or conflict remains unresolved. The authorised candidates and active constraints are explicit. ### 2. Allocate the guidance -Keep a rule at root only when it is confirmed, cannot be inferred reliably from repository evidence, and governs the repository as a whole. Ensure the root file itself states the immediate correction loop: when a user correction establishes a reusable repository-wide rule, ask whether to add it to `AGENTS.md`. Apply the same retention gate to checks, tools, preferences, and overrides; their existence does not authorise a new root rule. +Keep a rule at root only when it is confirmed, cannot be inferred reliably from repository evidence, governs the repository as a whole, and materially changes agent behaviour. Apply every gate equally to checks, tools, preferences, and overrides. A paired-repository relationship or the user's exact task-management tool choice can qualify; mere tool or script existence, framework convention, code inventory, task status, product language, workflow procedure, or directory-only rule cannot. -Remove stale, inferable, explanatory, or narrower guidance. Report useful narrower guidance with its smallest discoverable owner—an invocable skill for reusable workflows, a scoped instruction file for directory rules, or executable enforcement for mechanical constraints—and leave that destination unchanged. Every candidate is retained, displaced, or removed for an explicit reason. +Ensure the root file itself contains one immediate correction loop: when a user correction establishes a reusable repository-wide rule, ask whether to add it to `AGENTS.md`. Preserve project-native wording when an equivalent confirmed rule already exists; do not add a duplicate. + +Remove stale, inferable, explanatory, or narrower guidance. Report useful narrower guidance with its smallest discoverable proposed owner: an invocable skill for a reusable workflow, a scoped instruction file for a directory, executable enforcement for a mechanical rule, or the applicable product, context, or architecture authority. Leave those owners unchanged. Every candidate is retained, displaced, or removed for an explicit reason. ### 3. Write the root instructions -Create or rewrite only the root `AGENTS.md`. Use direct imperative lines and headings only when they navigate multiple instruction groups. Retain only words that require behaviour or preserve a necessary condition. State each repository policy once: merge instructions only when the same policy change would require both to change, and keep independently changeable rules separate. The root file is lean and unambiguous. +Create or update only the root `AGENTS.md`, preserving unrelated user work. If no candidate survives allocation or the current wording is already equivalent, make no change. Use direct imperative lines and headings only when they navigate multiple instruction groups. Retain only words that require behaviour or preserve a necessary condition. State each repository policy once: merge instructions only when the same policy change requires both statements to change, and keep independently changeable rules separate. Do not add a manifest, lifecycle state, task-tool adapter, exhaustive command list, architecture or code documentation, or skill procedure. The root file is lean and unambiguous. ### 4. Verify and hand off -Re-read every line for confirmed authority, non-inferability, repository-wide scope, behavioural value, conflicts, and semantic duplication. Verify each referenced command and reported destination. Inspect a command's definition before running it; run applicable checks required by active confirmed guidance only when their effects are safe and authorised, otherwise report the limitation. Return the changed root file, verification evidence or limitations, displaced guidance with destinations, and unresolved conflicts. The user receives a lean verified root file without unrequested changes to narrower owners. +Re-read every line against authority, non-inferability, repository-wide scope, behavioural value, conflicts, and semantic duplication. Verify each proposed owner and every referenced command definition before deciding whether to run it. Run applicable checks required by active guidance only when their effects, including aggregate command writes, are safe and authorised; otherwise report the limitation without weakening or replacing the check. Return the changed root file or no-op, verification evidence or limitations, displaced guidance with proposed destinations, and unresolved conflicts. Invocation authorises the agreed root edit without another application prompt. The user receives a lean verified root file without unrequested changes to narrower owners. diff --git a/skills/maintain-agents/agents/openai.yaml b/skills/maintain-agents/agents/openai.yaml index 6601cff..25e0133 100644 --- a/skills/maintain-agents/agents/openai.yaml +++ b/skills/maintain-agents/agents/openai.yaml @@ -2,4 +2,4 @@ interface: display_name: 'Maintain AGENTS.md' short_description: 'Maintain confirmed repository-wide agent guidance' policy: - allow_implicit_invocation: false + allow_implicit_invocation: true From b2a356325c1240769966b9f873e388a6d75ea2f5 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 23:10:50 +0100 Subject: [PATCH 24/79] feat(define-product): establish breadth-first product foundations --- SKILLS_PLAN.md | 2 +- skills/define-product/SKILL.md | 32 ++++++++++------- skills/define-product/agents/openai.yaml | 2 +- .../define-product/assets/product-template.md | 30 ++++++++++------ skills/define-product/references/DISCOVERY.md | 36 +++++++++++-------- 5 files changed, 62 insertions(+), 40 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index bf1ed44..a111015 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -400,7 +400,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `verify-change` skill - [x] Update `code-review` skill - [x] Update `maintain-agents` skill -- [ ] Update `define-product` skill +- [x] Update `define-product` skill - [ ] Create `define-architecture` skill - [ ] Create `specify-feature` skill - [ ] Create `design-feature` skill diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 32d42b0..4d5f96b 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -1,6 +1,6 @@ --- name: define-product -description: Discovers and maintains a durable product definition centred on high-level feature descriptions. Use when externalising a new product idea or refining an existing product. +description: Establishes a durable whole-product foundation for later feature definition. Use when externalising a new product idea or redefining an existing product. metadata: invocation: user disable-model-invocation: true @@ -8,34 +8,40 @@ disable-model-invocation: true # Define Product -A product definition externalises product knowledge into a durable reference for later planning and development. Concise strategic framing leads into a complete catalogue of high-level features without becoming a delivery plan. +Establishes one user-confirmed product foundation using an adapted **Product Vision Board** for strategic completeness and **User Story Mapping** for a mile-wide, inch-deep view of the whole product. The result guides later feature definition without becoming a feature specification, roadmap, or technical design. ## Process -### 1. Inspect existing knowledge +### 1. Inspect the product evidence -Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, applicable research and decisions, and the smallest representative repository evidence. Treat code and tests as evidence of current behaviour and the user as the authority on intent; surface contradictions between them. Derive available facts before questioning so the user supplies decisions and knowledge the existing material cannot establish. The known product and unresolved discovery surface are explicit. +Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, applicable research and decisions, and, for an existing system, the smallest representative public contracts, entry points, data boundaries, code, tests, and configuration. Treat the repository as evidence of current behaviour and the user as authority on intended product direction. Surface contradictions, distinguish observed behaviour from stated direction, and treat missing evidence as unknown rather than absence. Do not turn the inspection into an implementation inventory. When an existing product is difficult to establish, load [Discovery Techniques](references/DISCOVERY.md). The known product evidence, its provenance, and the unresolved product decisions are explicit. -### 2. Frame the product +### 2. Establish the product frame -Invoke `$elicit-with-context` and use the **Product Vision Board** dimensions to confirm the executive summary, vision, intended users and needs, value proposition, market position, business model, goals, success signals, pricing, boundaries, and non-goals. Invoke `$research` when external evidence could materially inform a decision about competitors, market conditions, pricing, regulation, or another product claim; keep its report authoritative and link applicable findings. When a framing dimension stalls, load only the relevant section of [Discovery Techniques](references/DISCOVERY.md). The concise product frame is user-confirmed and externally supported where material. +Invoke `$elicit` for material user-held knowledge and decisions throughout steps 2–5. Reuse the confirmed synthesis from a prior free-form `$elicit` session as resolved branches and ask only about gaps exposed by the product evidence. Establish concise executive orientation: purpose and vision; intended users, actors, and their needs and desired outcomes; value and meaningful alternatives or market position when applicable; product and business outcomes with observable success signals; business model or pricing only when material; and durable boundaries and non-goals. Derive discoverable facts before asking. The concise product frame is complete enough to orient the whole-product pass. + +Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. + +Invoke `$research` when a material external claim could change the product decision and its answer must survive this session; keep the report authoritative and link its decision-relevant conclusion rather than copying it wholesale. Use an incidental lookup when the answer need not persist. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would answer one bounded visual question better than discussion. Research and prototype results return evidence to this product session; neither owns `PRODUCT.md` or continuation. + +Keep `PRODUCT.md` unchanged until the complete whole-product synthesis is explicitly confirmed. Before then, supporting utilities may perform only the writes authorised by their own contracts. ### 3. Map the whole product -Use **User Story Mapping's big-picture techniques** without adopting its backlog or delivery workflow. Map the product mile-wide and inch-deep: inventory the known user-facing areas and candidate features before exploring any one feature in depth. Arrange the areas as a narrative backbone following the natural user journey; place a genuinely cross-cutting feature in the smallest coherent user-facing area rather than inventing a false sequence. Include observed features, confirmed direction, and product ideas from the inspected material and the user. Load the feature-mapping guidance in [Discovery Techniques](references/DISCOVERY.md) when the product surface is difficult to expose. The whole feature surface is visible at low resolution. +Map the product breadth-first before exploring any one capability. Walk each intended user from entry through recurring value, recovery, and exit or completion where applicable; use those journeys to expose the natural user-facing product areas and their high-level capabilities. Add commercial, account, trust, support, and genuinely cross-cutting areas only where the product needs them. Use the big-picture guidance in [Discovery Techniques](references/DISCOVERY.md) when the product surface is difficult to expose. The whole product is visible at low resolution without stories, screens, releases, or priorities. -### 4. Explore each feature +### 4. Clarify the capability catalogue -Work through the mapped features with `$elicit-with-context`, asking only for unresolved knowledge. For each feature, confirm its user value, high-level behaviour, meaningful boundaries, and one status: `Current` for observed product behaviour, `Confirmed direction` for intended behaviour the user has decided, or `Idea` for direction retained without commitment. Stop at the information needed for a concise feature description; leave prioritisation, sequencing, release slicing, estimates, tickets, architecture, implementation, user-story decomposition, and acceptance criteria to downstream work. Each feature is ready to become a self-contained mini-brief. +Clarify every mapped capability only to mini-brief depth: who benefits and the value they receive, the externally observable high-level behaviour, and meaningful boundaries that prevent a later feature session from assuming adjacent behaviour. For an existing product, distinguish `Current` observed behaviour, `Confirmed direction`, and uncommitted `Idea` where the distinction prevents conflation; these describe product truth, not delivery progress or priority. Do not manufacture current claims for a blank product. Leave detailed rules, states, scenarios, acceptance conditions, technical solution, architecture, sequencing, releases, estimates, and tickets to their owning later sessions. Each capability is independently understandable without becoming a feature specification. -### 5. Walk the catalogue +### 5. Confirm the whole-product synthesis -Narrate the complete product journey area by area with the user. Correct missing, duplicate, misplaced, or contradictory features and resolve every exposed question through `$elicit-with-context`; represent genuine uncertainty through `Idea` and contextual product language rather than an unanswered-questions inventory. Obtain final confirmation of the strategic frame and complete feature catalogue. The product definition is coherent and ready to persist. +Walk every journey and capability across the relevant user perspectives. Resolve material omissions, duplicates, false groupings, broken transitions, contradictions, and capability descriptions that conceal more than one product behaviour. Have `$elicit` present one complete synthesis of the product frame, journeys, and capability catalogue and obtain the user's explicit confirmation; if the user rejects or corrects it, reopen the affected product work and reconfirm the complete revised synthesis. The confirmed foundation is coherent and ready to persist. ### 6. Write the product definition -After final confirmation, create or update the single root `PRODUCT.md` from the [Product Definition Template](assets/product-template.md). Write the executive summary beneath the title, then the strategic sections, followed by feature areas as `##` headings and individual features as `###` headings. Give every feature its status and concise prose covering user value, high-level behaviour, and meaningful boundaries. Present user-confirmed decisions as ordinary product prose, cite useful repository evidence for current claims, and link external claims to their research reports. Preserve canonical language from `CONTEXT.md` and one authoritative meaning for each statement. The durable document makes the product and its features easy to understand and use in later work. +After explicit confirmation, create or update the single root `PRODUCT.md` using the [Product Foundation Template](assets/product-template.md) where the project has no established convention; the invocation already supplies authority for this agreed write. Preserve unrelated confirmed content and local structure when it differs only stylistically. Record the concise frame and explicit key journeys, then use natural user-facing areas as `##` headings and capabilities as `###` headings without a redundant umbrella. Preserve confirmed status distinctions only where material. Link external claims to durable research and keep canonical term meanings in `CONTEXT.md`. `PRODUCT.md` is the sole product authority, not a backlog, status tracker, assumptions ledger, or implementation mirror. ### 7. Verify the definition -Verify useful `Current` claims against repository evidence, intended direction against user confirmation, `Idea` statuses against the confirmed catalogue, and external claims against linked research. Check that every mapped feature has one mini-brief, the feature order tells a coherent product story, and excluded delivery detail has stayed downstream. Reconcile the finished document with `CONTEXT.md` and applicable decisions, then return changed files, supporting research, and any limitations in repository verification. The user receives a complete product foundation ready to inform feature planning. +Verify current claims against inspected evidence, direction and ideas against user confirmation, and material external claims against linked research. Check coverage from frame to journeys to every capability mini-brief, coherence across user perspectives, consistency with `CONTEXT.md`, preservation of unrelated confirmed content, and absence of architecture, detailed feature behaviour, priorities, roadmap, releases, tickets, or implementation detail. Verify every `$maintain-context` result and report changed files, evidence, and limitations. Stop after `PRODUCT.md` and applicable ubiquitous language are updated and verified; later architecture and feature definition remain separate sessions. diff --git a/skills/define-product/agents/openai.yaml b/skills/define-product/agents/openai.yaml index b14341f..967400c 100644 --- a/skills/define-product/agents/openai.yaml +++ b/skills/define-product/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Define Product' - short_description: 'Discover and document a product direction' + short_description: 'Establish a durable whole-product foundation' policy: allow_implicit_invocation: false diff --git a/skills/define-product/assets/product-template.md b/skills/define-product/assets/product-template.md index f6ff06f..8f09af1 100644 --- a/skills/define-product/assets/product-template.md +++ b/skills/define-product/assets/product-template.md @@ -1,35 +1,45 @@ # {Product name} -{Concise executive summary stating who the product serves, the outcome it enables, and what distinguishes its approach. Include overall product maturity only when it materially improves orientation.} +{Concise executive orientation stating who the product serves, the outcome it enables, and what distinguishes its approach. Include overall maturity only when it materially improves orientation.} -## Vision +## Purpose and vision -{The product's enduring purpose and positive change, independent of a particular implementation.} +{The product's purpose, enduring vision, and positive change, independent of a particular implementation.} ## Intended users and needs -- **{User or customer}**: {Prioritised need or desired outcome.} +- **{User or customer}**: {Needs and desired outcomes.} ## Value proposition and market position {Why intended users would choose this product over meaningful alternatives, manual work, or doing nothing. Describe supported differentiation and link material external claims to the applicable research report.} -## Business model, goals, and success signals +## Outcomes and success signals -{Describe the business model and applicable pricing decisions, then the product or business outcomes that justify investment and the observable signals of progress.} +{Describe the product and business outcomes that justify investment and the observable signals of progress. Include the business model and pricing only when material.} ## Product boundaries and non-goals - {A durable boundary or non-goal and the focus it preserves.} - +## Key user journeys -## {Feature area} +### {Journey name} + +**Actors:** {Intended users or other consequential actors.} + +{Concise journey from entry through recurring value, recovery, and exit or completion where applicable. Reference the user-facing areas and capabilities involved without adding screen flows or acceptance criteria.} + + + +## {Capability area} {Optional one-sentence orientation when the area's purpose is not clear from its name.} -### {Feature name} +### {Capability name} + + _Status: {Current | Confirmed direction | Idea}._ -{Concise prose describing the feature's user value, high-level behaviour, and meaningful boundaries. Cite repository evidence only when it usefully supports a Current claim; link external claims to their authoritative research report.} +{Concise prose stating who benefits and why, externally observable high-level behaviour, and meaningful boundaries. Leave detailed rules, states, scenarios, acceptance conditions, architecture, implementation, priorities, roadmaps, backlog items, sequencing, releases, estimates, and tickets to their owning later work. Link material external claims to their authoritative research report.} diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index dce55d9..0a0ab33 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -1,6 +1,6 @@ # Discovery Techniques -Load only the section needed by the active discovery branch. `$elicit-with-context` remains authoritative for questioning, confirmation, and canonical language; `$research` remains authoritative for external evidence. +Load only the section needed by the active discovery branch. `$elicit` owns material questioning and final synthesis confirmation; `$maintain-context` owns glossary persistence; `$research` owns durable external evidence; `$prototype` owns a bounded disposable visual experiment. `define-product` retains the product outcome and stopping boundary. ## Establish an existing product @@ -12,12 +12,12 @@ Use the **Product Vision Board** as a completeness check for the concise opening 1. What positive change and enduring purpose define the vision? 2. Which users, customers, and influential actors matter, and what outcomes do they need? -3. Why would they choose this product over alternatives, manual work, or doing nothing? -4. What market position or differentiation makes that choice plausible? -5. Which business model, revenue, costs, channels, goals, success signals, and pricing decisions matter now? +3. Why would they choose this product over meaningful alternatives, manual work, or doing nothing? +4. What market position or differentiation matters, when applicable? +5. Which product or business outcomes and observable success signals matter, and are business model or pricing decisions material now? 6. Which boundaries and non-goals keep the product coherent? -Keep this frame proportionate to its supporting role. Move into whole-product feature discovery once these dimensions are clear. +Keep this frame proportionate to its supporting role. Move into whole-product capability discovery once these dimensions are clear. ## Map the big picture @@ -25,27 +25,33 @@ Use the **mile-wide, inch-deep** pass before local detail: 1. Name the product's main users and their entry points. 2. Narrate how each user moves from first contact through recurring value and eventual exit or completion. -3. Record the user-facing product areas along that journey. -4. Beneath each area, inventory current features, confirmed direction, and ideas at one-line resolution. -5. Add commercial, account, trust, support, and other cross-cutting features that the main journey did not expose. +3. Include recurring value, recovery, and exit or completion where they apply. +4. Record the user-facing product areas along those journeys. +5. Beneath each area, inventory capabilities at one-line resolution. +6. For an existing product, distinguish observed current behaviour, confirmed direction, and uncommitted ideas where material. +7. Add commercial, account, trust, support, and other cross-cutting capabilities that the main journeys did not expose. -This pass creates a feature surface, not cards, stories, priorities, screens, releases, or architecture. +This pass creates a capability surface, not cards, stories, priorities, screens, releases, or architecture. -## Explore a feature +## Clarify a capability Resolve only the information required by the mini-brief: - **User value**: who benefits and what becomes possible or easier? - **High-level behaviour**: what does the product do from the user's perspective? -- **Status**: is it `Current`, `Confirmed direction`, or `Idea`? -- **Meaningful boundaries**: what nearby behaviour could a later planner reasonably but incorrectly assume belongs to it? +- **Meaningful boundaries**: what nearby behaviour could a later feature session reasonably but incorrectly assume belongs to it? +- **Status when material**: for an existing product, is it observed `Current` behaviour, `Confirmed direction`, or an uncommitted `Idea`? -Use examples or scenarios when behaviour remains ambiguous. Stop when a downstream planning session can understand the feature's product intent without receiving its implementation or delivery design. +Use a concise example only when high-level behaviour remains ambiguous. Stop when a later feature-definition session can understand the capability's product intent without receiving detailed rules, states, scenarios, acceptance conditions, implementation, or delivery design. ## Walk the complete catalogue -Read the mapped product back as one narrative. Change perspective across intended users and check entry, recurring-use, recovery, commercial, trust, support, and exit paths where applicable. Look for missing transitions, duplicated features, false journey positions, contradictions, and feature descriptions that hide more than one distinct product behaviour. Resolve each finding during elicitation, then repeat the walk until it exposes nothing new. +Read the mapped product back as one narrative. Change perspective across intended users and check entry, recurring-use, recovery, commercial, trust, support, and exit paths where applicable. Look for missing transitions, duplicated capabilities, false groupings, contradictions, and mini-briefs that hide more than one distinct product behaviour. Return each material finding to elicitation, then repeat the walk until it exposes nothing new. ## Establish external position -Compare the product with alternatives users employ today, including manual work and doing nothing. Invoke `$research` when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision. Keep researched claims in the research report and bring only the decision-relevant conclusion and link into `PRODUCT.md`. +Compare the product with meaningful alternatives users employ today, including manual work and doing nothing. Invoke `$research` only when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision and the answer must survive the session. Keep researched claims in the report and bring only the decision-relevant conclusion and link into `PRODUCT.md`; use an incidental lookup when no durable evidence record is needed. + +## Resolve a visual product question + +Invoke `$prototype` only when one bounded uncertainty is better resolved by seeing, comparing, or interacting with a visual artefact than by discussion. Give it the question, audience, authentic constraints, and product decision it informs. Use its returned evidence in elicitation after the disposable artefact is removed; do not preserve prototype material in the product foundation. From a68a3d1b50cdd571cf063a0696ffbbab7b0ed564 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 15 Aug 2026 23:27:45 +0100 Subject: [PATCH 25/79] feat(define-architecture): add technical foundation workflow --- SKILLS_PLAN.md | 2 +- skills/define-architecture/SKILL.md | 67 +++++++++++++++++++ skills/define-architecture/agents/openai.yaml | 5 ++ .../assets/architecture-template.md | 39 +++++++++++ 4 files changed, 112 insertions(+), 1 deletion(-) create mode 100644 skills/define-architecture/SKILL.md create mode 100644 skills/define-architecture/agents/openai.yaml create mode 100644 skills/define-architecture/assets/architecture-template.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index a111015..f6f3b85 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -401,7 +401,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `code-review` skill - [x] Update `maintain-agents` skill - [x] Update `define-product` skill -- [ ] Create `define-architecture` skill +- [x] Create `define-architecture` skill - [ ] Create `specify-feature` skill - [ ] Create `design-feature` skill - [ ] Create `create-tickets` skill diff --git a/skills/define-architecture/SKILL.md b/skills/define-architecture/SKILL.md new file mode 100644 index 0000000..6a2b497 --- /dev/null +++ b/skills/define-architecture/SKILL.md @@ -0,0 +1,67 @@ +--- +name: define-architecture +description: Establishes a durable technical foundation for a new or existing system. Use when the product foundation is confirmed and later feature solutions need system-wide architecture constraints. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Define Architecture + +Establishes one user-confirmed, system-specific technical foundation breadth-first, then tests it against representative change before recording it. Selective **arc42 architecture-description concepts** keep drivers, constraints, context, responsibility views, cross-cutting concerns, runtime, deployment, and risks inspectable without adopting its full template. The result constrains later feature solutions without duplicating current implementation or designing an individual feature. + +## Prerequisites + +Require a confirmed root `PRODUCT.md` and relevant root `CONTEXT.md`. Inspect both before continuing. When either is missing, or the product foundation is too incomplete or contradictory to establish its technical drivers, report the exact prerequisite and stop; leave product definition to a later `$define-product` invocation rather than manufacturing it here. + +## Process + +### 1. Establish the evidence boundary + +Inspect the request, `PRODUCT.md`, `CONTEXT.md`, existing `ARCHITECTURE.md` and ADRs, repository guidance, and material technical, organisational, regulatory, and operational constraints. For an existing system, inspect the smallest representative runtime and deployment evidence needed to understand principal entry points, responsibilities, data stores and flows, integrations, contracts, and trust boundaries. Treat code, configuration, contracts, tests, and deployment evidence as current implementation evidence, not intended architecture by default. Preserve confirmed unrelated content and local documentation conventions. The evidence is separated into confirmed intent, observed current architecture, and unresolved gaps before questioning begins. + +Invoke `$elicit` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Invoke `$prototype` only when one bounded visual or interactive uncertainty would be resolved better by the least elaborate disposable artefact than by discussion. Supporting utilities own only their contract-specific mutations and return their results to this session. + +Keep `ARCHITECTURE.md` unchanged until the complete synthesis is explicitly confirmed. + +### 2. Establish drivers and context + +Establish the architecture goals, product and organisational drivers, highest-priority material quality needs, fixed constraints, external actors and systems, scope, and trust boundaries. Express quality needs with enough observable context to compare consequences without manufacturing precision. Distinguish an existing system's observed current state, confirmed target direction, and unresolved gap wherever conflating them could mislead later design. The forces that the technical foundation must balance are explicit before a solution structure is proposed. + +### 3. Establish responsibilities and boundaries + +Map the principal system capabilities or responsibility boundaries, their coherent owners, information hidden behind each boundary, dependency and communication direction, and major contracts future feature designs must respect. Invoke `$modular-design` with the confirmed behaviour, drivers, constraints, and current evidence; use its structural findings as passive constraints while retaining ownership of architecture alternatives, user decisions, and this document. Prefer the smallest justified boundaries and deep, stable contracts that localise plausible change. Compare viable alternatives and their consequences for each material structural trade-off, then leave the choice to the user. The proposed foundation has coherent ownership and explicit dependency rules without mirroring the source tree. + +### 4. Complete the technical foundation + +Establish only the system-wide decisions and constraints that are material across the following concerns: + +- technology strategy and justified constraints; +- data domains, ownership, lifecycle, consistency, privacy, and retention; +- integrations, external dependencies, protocols, and failure boundaries; +- security, reliability and resilience, observability, performance and scalability, accessibility and compatibility, operability, and testing or verification strategy; and +- deployment and runtime topology at an enduring conceptual level. + +Omit a concern when it is not material. Record what future designs must preserve, what cannot yet be reliably inferred, and why a constraint exists. Leave endpoints, schemas, classes, files, framework wiring, executable configuration, library inventories, ephemeral deployment details, ticket or roadmap state, and feature-specific solutions to their owning evidence and later sessions. The proposed foundation is proportionate and contains no speculative implementation inventory. + +Invoke `$maintain-context` inline when consequential terminology is confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update. Treat `ARCHITECTURE.md` as the main living synthesis. Identify and test an ADR candidate only when `$maintain-context`'s four gates appear satisfied, then include its rationale and the user's explicit durable-record agreement in the final synthesis. Do not invoke `$maintain-context` to record an ADR before that whole synthesis is confirmed, and reject routine, discoverable, or section-level choices as candidates. + +### 5. Walk representative change + +Walk representative current and likely capability changes from product intent through actors, boundaries, contracts, data, integrations, cross-cutting constraints, and runtime consequences. Use these walks to expose contradictions, missing ownership, unstable dependency direction, or a foundation that would force scattered change. Reopen only the evidenced architecture branch needed to resolve a failure, compare the viable consequences, and avoid elaborating the feature itself. The foundation can constrain plausible later solution designs without pretending to settle them. + +### 6. Confirm the complete synthesis + +Have `$elicit` present one concise, complete synthesis covering drivers and context, responsibilities and dependency direction, technology and data strategy, integrations and material cross-cutting constraints, current-versus-target distinctions, trade-offs, evolution constraints, unresolved material risks, and every proposed ADR rationale. Obtain explicit user confirmation of the whole synthesis. Create no ADR record from any candidate until the complete synthesis, or a revised synthesis after rejection, is confirmed. If the user rejects or corrects it, leave `ARCHITECTURE.md` unchanged, reopen only affected evidence-backed branches, repeat the relevant change walks, and reconfirm the complete revised synthesis. The confirmed synthesis supplies authority for its agreed document writes without a second permission prompt. + +### 7. Record the architecture + +Create or update the single root `ARCHITECTURE.md` using the [Architecture Foundation Template](assets/architecture-template.md) when the project has no stronger convention. Preserve unrelated confirmed content and local structure when it differs only stylistically. Omit empty or immaterial sections rather than filling boilerplate. Use prose, tables, or repository-native diagrams only when they materially clarify relationships. Keep canonical term meanings in `CONTEXT.md`, external evidence in research reports, and exceptional rationale in qualifying ADRs instead of duplicating them. + +After writing `ARCHITECTURE.md`, invoke `$maintain-context` for every confirmed qualifying ADR and for confirmed terminology not already applied. The prior architecture confirmation supplies explicit agreement only for ADR rationales included in that synthesis. + +### 8. Verify and stop + +Verify every recorded statement against the confirmed synthesis and its authority: product intent against `PRODUCT.md`; terms against `CONTEXT.md`; current claims against inspected system evidence; target direction and trade-offs against user confirmation; and external claims against cited research. Check that responsibilities, dependency directions, data ownership, integration failure boundaries, and material quality constraints agree with one another and that the representative changes still fit. Confirm preservation of unrelated content and absence of implementation inventory, feature design, scaffolding, priorities, tickets, releases, or deployment execution. + +Report changed files, evidence and limitations, current-versus-target gaps, unresolved material risks, and the next available later session without starting it. Stop after the foundation and warranted context records are written and verified; do not scaffold code, select an individual feature solution, create work items, implement, release, or deploy. diff --git a/skills/define-architecture/agents/openai.yaml b/skills/define-architecture/agents/openai.yaml new file mode 100644 index 0000000..7067e7a --- /dev/null +++ b/skills/define-architecture/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Define Architecture' + short_description: 'Establish an enduring technical foundation' +policy: + allow_implicit_invocation: false diff --git a/skills/define-architecture/assets/architecture-template.md b/skills/define-architecture/assets/architecture-template.md new file mode 100644 index 0000000..24c3d37 --- /dev/null +++ b/skills/define-architecture/assets/architecture-template.md @@ -0,0 +1,39 @@ +# Architecture + + + +## Orientation + + + +## System context + + + +## Responsibilities and boundaries + + + +## Technology strategy + + + +## Data + + + +## Integrations + + + +## Cross-cutting constraints + + + +## Runtime and deployment + + + +## Evolution, trade-offs, and risks + + From 7dbd3782bc4ba1dc03c90eace4c73bce2594d4ab Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 07:51:27 +0100 Subject: [PATCH 26/79] feat(specify-feature): add observable feature definition --- SKILLS_PLAN.md | 2 +- skills/specify-feature/SKILL.md | 59 +++++++++++++++++++ skills/specify-feature/agents/openai.yaml | 5 ++ .../assets/specification-template.md | 53 +++++++++++++++++ 4 files changed, 118 insertions(+), 1 deletion(-) create mode 100644 skills/specify-feature/SKILL.md create mode 100644 skills/specify-feature/agents/openai.yaml create mode 100644 skills/specify-feature/assets/specification-template.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index f6f3b85..0e8c2ff 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -402,7 +402,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `maintain-agents` skill - [x] Update `define-product` skill - [x] Create `define-architecture` skill -- [ ] Create `specify-feature` skill +- [x] Create `specify-feature` skill - [ ] Create `design-feature` skill - [ ] Create `create-tickets` skill - [ ] Update `implement` skill diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md new file mode 100644 index 0000000..cec6f00 --- /dev/null +++ b/skills/specify-feature/SKILL.md @@ -0,0 +1,59 @@ +--- +name: specify-feature +description: Defines one substantial feature or enhancement as approved, observable product intent. Use when a product capability or agreed feature idea needs a durable specification before solution design. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Specify Feature + +Defines one substantial feature or enhancement without selecting its implementation. **Goal-oriented use cases** connect actors and beneficiaries to complete observable journeys; selective **Specification by Example** turns ambiguous rules and acceptance boundaries into discriminating examples rather than an implementation test inventory. + +## Process + +### 1. Inspect the feature evidence + +Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, the existing feature specification and paired solution design when present, and the project's feature-record convention. Inspect only the relevant current public contracts, documentation, tests, UI, research, and implementation evidence needed to understand observable behaviour. Treat `PRODUCT.md` as authority for the product foundation, the user as authority for the intended change, and the current system as evidence of present behaviour rather than intended future behaviour. Distinguish current, confirmed future, and unresolved claims; surface contradictions and unknowns without silently resolving them. + +Accept one capability from `PRODUCT.md`, a confirmed synthesis returned by a prior `$elicit` session, or another agreed feature idea whose product context is sufficient. Do not manufacture a missing product foundation when the requested specification depends on one; report the exact prerequisite and stop. A direct agreed idea may proceed without a formal catalogue when its actors, value, and product boundary are sufficiently grounded. The source, authority, current evidence, convention, and unresolved feature decisions are explicit. + +### 2. Bound one proportionate feature + +Invoke `$elicit` throughout steps 2–5 for material user-held knowledge and decisions. Reuse every compatible branch from a prior confirmed synthesis and ask only about gaps exposed by inspection. Bound one coherent feature or enhancement, its problem or opportunity, beneficiaries, desired outcome, relationship to the product capability, and exclusions that separate adjacent behaviour. When inspection shows that the change is small, already understood, and needs no durable intent record, explain why a specification would add ceremony, name `$implement` as the proportionate next session, and stop without writing or invoking it. Invocation alone does not bypass this gate; proceed only when the user has separately established that a durable feature specification is required. The session owns exactly one warranted feature and a clear reason for specifying it. + +Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. + +Invoke `$research` when a material external, domain, regulatory, or standards claim could change the specification and its evidence must survive this session. Link its durable conclusion instead of copying the report. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would resolve one bounded visual or interactive uncertainty better than discussion. Both return evidence to this session and own neither the specification nor continuation. + +Keep the specification unchanged until the complete synthesis is explicitly confirmed. Supporting utilities may perform only the mutations authorised by their own contracts before then. + +### 3. Establish observable behaviour + +Establish only the applicable feature-level detail: actors and their goals; scope and boundaries; externally observable behaviour and business or domain rules; material states, transitions, and lifecycle; main, adverse, recovery, and edge scenarios; and external product, regulatory, or compatibility constraints. Walk each actor from entry through goal, consequential interactions with other actors or systems, failure or refusal, recovery, and completion where applicable. Distinguish current behaviour from intended change wherever conflating them could misstate the feature. + +Use concrete examples when an abstract rule, boundary, state transition, or scenario remains open to materially different interpretations. Include an example only when it discriminates valid from invalid behaviour or exposes a missing branch; do not expand the specification into exhaustive cases. Resolve contradictions and material ambiguity as they emerge. The feature's complete observable behaviour is understandable without assuming a technical solution. + +### 4. Establish acceptance and traceability + +State acceptance conditions as unambiguous outcomes and evidence observable by an actor, consumer, or published contract. Cover the material behaviour, rules, state transitions, adverse paths, exclusions, and external constraints established for this feature. Use exact protocol or schema detail only when it is itself a published external contract; otherwise leave endpoints, components, services, tables, internal events, data models, algorithms, frameworks, libraries, deployment, and verification implementation to solution design. + +Apply the rebuild test to every proposed statement: it normally belongs in the specification only when it should remain true if the feature were rebuilt using a different architecture. Retain a system-specific statement only when it is an externally binding product, regulatory, or compatibility boundary. Link applicable product capability, research, context terms, current evidence, and stable material dependencies or assumptions without duplicating their authorities. The acceptance boundary is observable, testable, traceable, and independent of implementation. + +### 5. Challenge the complete feature + +Walk the feature across every material actor goal, interaction, state, rule, main path, adverse path, recovery, edge condition, acceptance outcome, and exclusion. Test counterexamples at the feature boundary and look for hidden adjacent features, conflicting rules, impossible transitions, undefined ownership, unsupported current claims, and acceptance conditions that inspect private internals or prescribe a solution. Reopen only evidence-backed branches that could materially change the specification. Stop elaborating when another case would not change externally observable intent. The proposed feature is coherent and complete at feature depth without becoming solution design or an implementation backlog. + +### 6. Confirm one full synthesis + +Have `$elicit` present one concise, self-contained synthesis of the problem or opportunity, desired outcome, actors and goals, scope and exclusions, observable behaviour and rules, states and transitions, material scenarios, acceptance conditions, external constraints, traceability, and unresolved limitations. Obtain explicit confirmation of the whole synthesis. Selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the specification unchanged, reopen every affected evidence-backed branch, repeat the applicable walks and counterexamples, and confirm the complete revised synthesis. The confirmed synthesis supplies authority for the agreed specification write without another application prompt. + +### 7. Write the specification + +Follow the project's established feature-record convention. Otherwise choose a stable lowercase feature slug and create `docs/features//specification.md` using the [Feature Specification Template](assets/specification-template.md). When refining an existing specification, preserve unrelated confirmed content and its local structure when differences are stylistic. Include only material sections and omit empty boilerplate. Keep canonical meanings in `CONTEXT.md`, external evidence in research reports, and implementation choices in a distinct `solution-design.md` when one exists. The durable specification records exactly the confirmed observable intent. + +### 8. Verify and stop + +Verify every statement against its authority: product foundation against `PRODUCT.md`; intended change against the confirmed synthesis; terminology against `CONTEXT.md`; current behaviour against inspected evidence; and material external claims against linked research. Rewalk traceability, actors, rules, states, scenarios, acceptance conditions, exclusions, and constraints for coherence and missing branches. Apply the rebuild test and confirm that acceptance remains observable without CSS classes, DOM or private internals, architecture, endpoints unless externally contractual, detailed data models, algorithms, frameworks, deployment, ticket breakdown, estimates, priorities, roadmaps, or releases. Verify the path, links, preservation of unrelated content, separation from any solution design, and every `$maintain-context` result. + +Report changed files, supporting evidence, verification, and unresolved limitations. Stop after the approved specification and applicable ubiquitous language are written and verified; do not select a solution, create tickets, scaffold, implement, publish, prioritise, or continue into another lifecycle outcome. diff --git a/skills/specify-feature/agents/openai.yaml b/skills/specify-feature/agents/openai.yaml new file mode 100644 index 0000000..3303890 --- /dev/null +++ b/skills/specify-feature/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Specify Feature' + short_description: 'Define one feature as observable product intent' +policy: + allow_implicit_invocation: false diff --git a/skills/specify-feature/assets/specification-template.md b/skills/specify-feature/assets/specification-template.md new file mode 100644 index 0000000..efd7ef8 --- /dev/null +++ b/skills/specify-feature/assets/specification-template.md @@ -0,0 +1,53 @@ +# {Feature name} + + + +## Summary + + + +## Actors and goals + +- **{Actor or beneficiary}:** {Goal or outcome sought from this feature.} + +## Scope and boundaries + +### Included + +- {Observable behaviour within this feature.} + +### Excluded + +- {Adjacent behaviour deliberately outside this feature.} + +## Observable behaviour and rules + + + +## States and transitions + + + +## Scenarios + +### {Main, adverse, recovery, or edge scenario} + + + +## Acceptance conditions + +- {An unambiguous, testable outcome observable by an actor, consumer, or published external contract.} + +## External constraints + + + +## Dependencies and assumptions + + + +## Evidence and traceability + +- **Product foundation:** {Link to the applicable PRODUCT.md capability or state the agreed direct feature source.} +- **Current behaviour:** {Link to relevant public contracts, docs, tests, UI evidence, or other inspected evidence when material.} +- **Research:** {Link material durable external evidence when applicable.} From 0d48dd8e951c5b11edc934f7bdef029af6eac0ef Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 08:14:21 +0100 Subject: [PATCH 27/79] feat(design-feature): add buildable solution design --- SKILLS_PLAN.md | 2 +- skills/design-feature/SKILL.md | 97 +++++++++++++++++++ skills/design-feature/agents/openai.yaml | 5 + .../assets/solution-design-template.md | 69 +++++++++++++ 4 files changed, 172 insertions(+), 1 deletion(-) create mode 100644 skills/design-feature/SKILL.md create mode 100644 skills/design-feature/agents/openai.yaml create mode 100644 skills/design-feature/assets/solution-design-template.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 0e8c2ff..5dfcf80 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -403,7 +403,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `define-product` skill - [x] Create `define-architecture` skill - [x] Create `specify-feature` skill -- [ ] Create `design-feature` skill +- [x] Create `design-feature` skill - [ ] Create `create-tickets` skill - [ ] Update `implement` skill - [ ] Update `commit` skill diff --git a/skills/design-feature/SKILL.md b/skills/design-feature/SKILL.md new file mode 100644 index 0000000..d9632a1 --- /dev/null +++ b/skills/design-feature/SKILL.md @@ -0,0 +1,97 @@ +--- +name: design-feature +description: Designs an approved feature specification into a confirmed, buildable, system-specific technical solution. Use when feature behaviour is settled and needs solution design before ticket creation. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Design Feature + +Designs one approved feature specification into one buildable technical solution without redefining the feature or implementing it. A force- and risk-led process uses selective **arc42 viewpoints** to expose the feature's context, boundaries, runtime, deployment, cross-cutting concerns, decisions, and risks; **bidirectional requirements-to-design traceability** and **scenario walkthroughs** keep the result complete and necessary. + +## Prerequisites + +Require the approved feature `specification.md`, root `ARCHITECTURE.md`, and relevant root `CONTEXT.md` and ADRs. Inspect them before continuing. When an authority is missing, inadequate, or contradictory in a way that prevents a complete solution, report the exact prerequisite or gap and stop. Leave missing feature intent to a later `$specify-feature` session and missing system-wide constraints to a later `$define-architecture` session; do not invoke either or manufacture its outcome. + +Refine an existing paired solution design when requested, preserving unrelated confirmed content and local conventions. + +## Process + +### 1. Freeze authorities and evidence + +Inspect the request, prerequisite authorities, existing paired solution design, repository guidance, and the project's feature-record convention. Inspect only the relevant code, tests, contracts, schemas, configuration, runtime and deployment evidence, similar implemented patterns, and material external technical or regulatory evidence needed to design this feature. Treat the specification as authority for what and why, `ARCHITECTURE.md` and applicable ADRs as authority for system-wide constraints, `CONTEXT.md` as authority for project language, and current implementation as evidence of current constraints and idioms rather than automatic intended design. The user remains authority for material system-specific trade-offs, feature scope, and any proposed departure from the architecture. + +Separate confirmed future behaviour, architecture constraints, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide decision is absent or the feature would require changing the architecture, identify the exact architecture gap and stop rather than embedding a new foundation in this design. + +Invoke `$elicit` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Invoke `$prototype` only when one bounded visual, interactive, or technically observable uncertainty can be discriminated by its disposable artefact. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. + +Invoke `$maintain-context` inline whenever consequential project terminology is independently confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update; this skill retains the solution-design outcome and verification. Apply a glossary change before whole-design confirmation only when its meaning is confirmed independently of the proposed solution. Keep terminology that depends on an unconfirmed design choice unchanged until that choice is confirmed. + +The solution-design authority, runtime evidence, unresolved gaps, and permitted decision routes are explicit before solution selection begins. + +### 2. Map behaviour to design forces + +Map every material behaviour, rule, state, scenario, external constraint, and acceptance condition in the specification to affected system responsibilities and a credible verification seam. Identify the feature's entry points, trust boundaries, owned knowledge, data and integration flows, current-versus-target differences, and material quality, migration, operational, compatibility, and evolution forces. Trace each proposed new component, contract, data shape, migration, or operational obligation back to a specification behaviour or architecture need; remove anything with no such origin. + +Invoke `$modular-design` with the confirmed behaviour, architecture constraints, project language, and current evidence. Use its structural findings as passive constraints on ownership, information hiding, cohesion, contracts, dependency direction, seams, and change locality while retaining responsibility for alternatives, user decisions, the complete solution, verification, and stopping. + +The affected architecture, design forces, risks, and two-way traceability are complete enough to compare coherent solutions without selecting implementation detail prematurely. + +### 3. Compare and select the solution + +Compare viable system-specific alternatives for every material choice. Prefer the smallest coherent complete solution that follows established project and framework idioms, keeps knowledge with its natural owner, offers deep stable contracts, localises plausible change, and avoids speculative abstraction. Consider the consequences for the relevant forces, failure modes, migration, operations, verification, and likely evolution. Do not let current code win by default when it conflicts with the confirmed target, and do not introduce a new seam or component without evidenced variation, ownership, boundary, migration, or observation value. + +Resolve material trade-offs with the user through `$elicit` when invoked. Select one internally coherent solution and record each material selected option and its consequences. Retain rejected alternatives only when they are needed to understand a consequential trade-off, risk, or qualifying ADR; avoid alternative essays. + +The selected solution is justified by runtime constraints and confirmed decisions rather than familiarity or template completeness. + +### 4. Complete the technical design + +Develop only the applicable design depth needed for safe ticket decomposition: + +- map acceptance and behaviour to affected responsibilities, boundaries, ownership, information hiding, dependency direction, and enabling refactors; +- define internal and external interfaces, contracts, events, protocols, data shapes, invariants, ordering, errors, and compatibility at the precision consumers need; +- resolve data ownership, lifecycle, consistency, concurrency, transactions, privacy, retention, migration, backfill, rollback, and recovery; +- resolve integration protocols, authentication, authorization, idempotency, retries, timeouts, partial failure, reconciliation, and ownership of failure; +- resolve material security and threat controls, performance and capacity, resilience, observability, operability, support, and technical accessibility or compatibility concerns; +- define runtime and deployment effects plus rollout or transition constraints only where safe implementation depends on them; and +- identify the project-required and risk-triggered verification seams, evidence, environments, fixtures, or infrastructure needed to establish the designed behaviour. + +Omit immaterial concerns. Keep externally contractual detail already owned by the specification exact; otherwise select internal detail only to the level that constrains implementation meaningfully. Remain specific enough to decompose later without turning the design into code, a file or class inventory, framework wiring, executable configuration, or an exhaustive task list. + +The solution is coherent across structure, data, integrations, cross-cutting concerns, transition, operations, and verification. + +### 5. Walk implementation scenarios + +Walk the solution end to end through the main behaviour, material adverse and recovery paths, likely evolution, and migration and rollback where applicable. Follow each scenario across responsibilities, contracts, data, integrations, trust boundaries, failure ownership, observability, and verification seams. Challenge concurrency, partial execution, retries, incompatible versions, failed migration or backfill, restoration, and operational diagnosis when the runtime evidence makes them material. + +Use these walks to expose missing ownership, contradictions, duplicated rules, unstable dependencies, untraceable design elements, acceptance conditions without verification seams, or a solution that cannot be divided safely. Identify implementation-sized seams, genuine dependencies, enabling refactors, and sequencing constraints only far enough to prove the design is decomposable; do not create tickets, ticket IDs, estimates, priorities, or a delivery plan. Reopen only the affected design choice and repeat the applicable walk after any change. + +The selected solution survives its material operating, change, failure, transition, and verification scenarios and has credible decomposition seams. + +### 6. Confirm the complete design + +Present one concise, self-contained synthesis covering authorities, drivers, selected solution, affected responsibilities and dependency direction, interfaces and contracts, data, integrations, material cross-cutting concerns, migration and compatibility, operations, verification, implementation dependencies, risks, trade-offs, and both traceability directions. Include every proposed terminology update and ADR rationale. + +Identify an ADR candidate only when `$maintain-context`'s four gates appear to pass and the user agrees that the durable rationale is warranted. Most feature choices belong only in the solution design. Obtain a new explicit response confirming the whole synthesis; selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the solution design and ADRs unchanged, leave terminology proposed only within the rejected synthesis unapplied, reopen every affected evidence-backed choice, repeat the applicable traces and scenarios, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid after rejection. + +The confirmed synthesis supplies authority for the agreed writes without another application prompt. + +### 7. Write the solution design and context + +Follow the project's established feature-record convention. Otherwise write beside the specification at `docs/features//solution-design.md` using the [Solution Design Template](assets/solution-design-template.md). Preserve unrelated confirmed content and local structure when refining an existing design. Include only material sections and omit empty boilerplate. Link authorities and durable evidence rather than duplicating them. + +Record the selected how and its consequences at design precision. Use a small example only when indispensable to make a contract unambiguous. Do not rewrite the specification's what or why, duplicate live implementation state, prescribe file-by-file or method-by-method work, or include tickets, code scaffolding, estimates, priorities, releases, or deployment execution. + +After writing the solution design, invoke `$maintain-context` for every confirmed qualifying ADR and reconcile confirmed terminology not already applied inline. The prior whole-design confirmation supplies authority for design-dependent terminology and explicit durable-rationale agreement only for ADRs included in that synthesis. It alone owns `CONTEXT.md` and ADR files; this skill retains the solution-design outcome and verification. + +The approved solution and only its warranted durable context are recorded in their owning artifacts. + +### 8. Verify and stop + +Verify every material specification behaviour and acceptance condition has a design responsibility and verification seam, and every introduced responsibility, component, contract, data shape, migration, integration, operational obligation, and verification mechanism traces back to a specification or architecture need. Rewalk the main, adverse, evolution, migration, and rollback scenarios that apply. Check responsibility ownership, contract compatibility, dependency direction, data lifecycle, integration failure handling, cross-cutting constraints, operational diagnosability, safe transition, and implementation decomposability for internal coherence. + +Verify statements against their authorities, links and paired path, preservation of unrelated content, and every `$maintain-context` result. Confirm the design has not changed product behaviour, smuggled in a system-wide architecture decision, duplicated live state, created tickets, or become code or file-level implementation instructions. + +Report changed files, authorities and supporting evidence, selected solution and material trade-offs, verification performed, and unresolved limitations. Stop after the confirmed solution design and applicable context records are written and verified; do not create tickets, estimate, prioritise, scaffold, implement, commit, publish, release, or deploy. diff --git a/skills/design-feature/agents/openai.yaml b/skills/design-feature/agents/openai.yaml new file mode 100644 index 0000000..783b7b2 --- /dev/null +++ b/skills/design-feature/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Design Feature' + short_description: 'Design a buildable feature solution' +policy: + allow_implicit_invocation: false diff --git a/skills/design-feature/assets/solution-design-template.md b/skills/design-feature/assets/solution-design-template.md new file mode 100644 index 0000000..af59bcc --- /dev/null +++ b/skills/design-feature/assets/solution-design-template.md @@ -0,0 +1,69 @@ +# {Feature name} solution design + + + +## Summary and authorities + + + +## Design drivers + + + +## Selected solution + + + +## Affected responsibilities and boundaries + + + +## Interfaces and contracts + + + +## Data + + + +## Integrations + + + +## Cross-cutting concerns + + + +## Migration and compatibility + + + +## Operations + + + +## Verification strategy and seams + + + +## Implementation and dependency notes + + + +## Risks and trade-offs + + + +## Traceability + +### Specification to design + +| Specification behaviour or acceptance condition | Design responsibility or contract | Verification seam or evidence | +| --- | --- | --- | +| {Use a stable specification label or concise exact reference.} | {Owning design element.} | {Observable verification boundary.} | + +### Design to authority + +| Introduced design element or obligation | Originating specification or architecture need | +| --- | --- | +| {Responsibility, component, contract, data shape, migration, integration, operational change, or verification mechanism.} | {Linked or exact authority reference.} | From 349dc9c03489419e9eab098f256541cf485ecbfc Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:15:34 +0100 Subject: [PATCH 28/79] feat(implement): enforce reviewed verified candidates --- SKILLS_PLAN.md | 2 +- skills/implement/SKILL.md | 36 +++++++++++++++++++++++++---- skills/implement/agents/openai.yaml | 2 +- 3 files changed, 34 insertions(+), 6 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 5dfcf80..5575aac 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -405,7 +405,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `specify-feature` skill - [x] Create `design-feature` skill - [ ] Create `create-tickets` skill -- [ ] Update `implement` skill +- [x] Update `implement` skill - [ ] Update `commit` skill - [ ] Replace `pr` with `pull-request` - [ ] Create `review-pull-request` skill diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 0322ca3..cb36fc2 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -1,6 +1,6 @@ --- name: implement -description: Routes confirmed implementation work through TDD and code review until complete. Use when the user asks to implement a clear software request. +description: Implements one ticket or small agreed software change with proportionate quality evidence. Use when the user asks to build confirmed work locally. metadata: invocation: user disable-model-invocation: true @@ -8,15 +8,43 @@ disable-model-invocation: true # Implement -A tailored **Software Formal Inspection** lifecycle routes confirmed work through `$tdd`, independent `$code-review`, technical disposition, and verified rework. +A tailored **Software Formal Inspection** lifecycle implements one confirmed body of work, independently reviews a fixed candidate, and verifies that same candidate. TDD supplies the implementation method only when its prerequisites hold. ## Process -Inspect the confirmed work, its applicable authorities, and the relevant current code. Invoke `$modular-design` when that evidence presents a material structural-maintainability decision and carry its constraints into `$tdd`, then invoke `$code-review` independently on the result. Validate every finding against the confirmed request and current codebase, then disposition it: +### 1. Fix the outcome + +Inspect the stated ticket or agreed work, repository instructions, linked product and technical authorities, current code and tests, and the complete repository state. A clear small request is sufficient authority; do not manufacture a ticket, specification, or other ceremony. Invoke `$elicit` one question at a time only when user-held ambiguity could materially change behaviour, contracts, architecture, scope, or acceptance. State the observable outcome, exclusions, applicable structural constraints, material risks, and evidence expected before editing. The implementation basis is explicit and bounded. + +### 2. Build the smallest complete change + +Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge question whose evidence must survive the conversation, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-context` when independently confirmed durable terminology arises or a rare qualifying decision needs durable rationale; each utility returns control to this outcome. + +Invoke `$tdd` only when an existing runnable test suite can credibly exercise the requested observable behaviour at a stable boundary with an independent oracle. When evidence already shows any prerequisite is absent, keep control here rather than invoking TDD merely to receive that result. Let TDD own the Red-Green-Refactor cycles and returned code and tests when it applies. Otherwise implement through the smallest conventional path and use the strongest available project-native feedback; update existing tests only when they can meaningfully express changed behaviour. Do not install or invent a test or quality framework to satisfy this process. When meaningful infrastructure is absent, preserve the limitation for verification and recommend separately planned work rather than adding unrelated harness infrastructure. + +Inspect the resulting diff and every in-scope untracked file. Account for each production addition against the agreed outcome or a necessary supporting change, and identify every change to tests, snapshots, baselines, suppressions, dependencies, quality configuration, scripts, or generated artefacts as part of the measurement path. The candidate is minimal, coherent, and ready for independent inspection. + +### 3. Review a fixed candidate + +Freeze the complete candidate, including its exact diff or revisions, changed and untracked paths, repository state, intended-behaviour authority, applicable structural constraints, and named risks. Invoke `$code-review` independently on that basis. If the basis drifts or either applicable review axis is stale or unperformed, re-freeze and repeat the review when the scope and ownership remain clear; otherwise stop with the exact blocker. Do not treat implementation self-inspection, tests, or verification as a substitute for the independent review. + +Validate every Standards and Spec finding against the confirmed request, its authorities, and current code, then give it one technical disposition: - A required correction is a verified violation of confirmed behaviour, the specification, a repository rule, or a necessary safety boundary, or another defect that leaves the change untrustworthy. - A proportional improvement is valid but non-blocking; accept it only when its concrete benefit justifies the added change and complexity. - A rejected finding is falsified, unsupported, superseded, or outside the confirmed scope. - A user decision is required when accepting or rejecting the finding would change behaviour, contracts, architecture, or scope, or when its evidence cannot be obtained safely. -Treat priority as evidence of impact and ordering, not remediation authority. Route each accepted change through `$tdd`, then repeat `$code-review` after any correction. Complete when the requested work is verified, every finding has a technical disposition, and no required finding remains unresolved. +For a user-decision finding, invoke `$elicit` and stop mutation, further adjudication, and verification until its confirmed synthesis returns. Resume from the fixed outcome; when the decision changes the candidate or any review authority, freeze the resulting basis and repeat independent review. + +Treat priority as evidence of impact and ordering, not remediation authority. Apply every accepted correction through `$tdd` when its prerequisite applies and otherwise through the conventional path. After any candidate change, freeze the complete new candidate and repeat independent review. Continue until every finding has a supported disposition, no required finding remains unresolved, and the complete current candidate has been independently reviewed. The reviewed candidate is fixed and current. + +### 4. Verify the reviewed candidate + +Invoke `$verify-change` on the exact reviewed candidate, supplying its authorities, named risks, review dispositions, relevant implementation evidence, and identified measurement-path changes. Let it select and run every applicable repository-required and risk-triggered existing harness and return claim-level evidence, reproducible commands, limitations, residual risks, and one honest verdict. Metrics support only their named claims; they never prove quality alone. + +When verification exposes an in-scope failure, repair it through the applicable TDD or conventional path, then repeat independent review and verification on the new fixed candidate. When verification evidence is stale because source or measurement-path state changed, establish ownership and scope, then re-freeze, re-review, and re-verify; do not silently retarget old evidence. Preserve a pre-existing or out-of-scope required-gate failure as a blocker rather than broadening the work. When missing infrastructure leaves the result partially verified or unverified and no in-scope repair can supply faithful evidence, report that limitation and explicit planned work without manufacturing a harness or claiming successful completion. + +### 5. Hand off the local result + +Report the agreed behaviour delivered, changed artefacts, review scope and every finding disposition, verification scope and verdict, exact commands and material results, limitations, residual risks, and any blocker or planned follow-up. Claim successful completion only when the behaviour exists locally, applicable tests and artefacts are present, the complete current candidate has a current independent review with no unresolved required finding, and verification passed within its stated scope. A partial, unverified, or failed verdict is an explicit incomplete or blocked handoff. Stop without committing, pushing, opening a pull request, releasing, deploying, changing tracker state, or authoring unrelated product, specification, or design documents. diff --git a/skills/implement/agents/openai.yaml b/skills/implement/agents/openai.yaml index 0e09bd3..100f67c 100644 --- a/skills/implement/agents/openai.yaml +++ b/skills/implement/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Implement' - short_description: 'Implement through TDD and review' + short_description: 'Implement and verify one bounded change' policy: allow_implicit_invocation: false From d23c2cbda817801bd5b9390b7426e93f3944875e Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:23:08 +0100 Subject: [PATCH 29/79] feat(create-tickets): add vertical tracker decomposition --- SKILLS_PLAN.md | 2 +- skills/create-tickets/SKILL.md | 75 ++++++++++++++++++++++++ skills/create-tickets/agents/openai.yaml | 6 ++ 3 files changed, 82 insertions(+), 1 deletion(-) create mode 100644 skills/create-tickets/SKILL.md create mode 100644 skills/create-tickets/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 5575aac..2f68c25 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -404,7 +404,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `define-architecture` skill - [x] Create `specify-feature` skill - [x] Create `design-feature` skill -- [ ] Create `create-tickets` skill +- [x] Create `create-tickets` skill - [x] Update `implement` skill - [ ] Update `commit` skill - [ ] Replace `pr` with `pull-request` diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md new file mode 100644 index 0000000..e54e730 --- /dev/null +++ b/skills/create-tickets/SKILL.md @@ -0,0 +1,75 @@ +--- +name: create-tickets +description: Creates verified native tracker items from an approved feature specification and solution design. Use when a feature needs implementation-ready, dependency-aware tickets. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Create Tickets + +Turns one approved feature into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules and examples into observable acceptance evidence. + +## Prerequisites + +Require the approved feature specification and its approved solution design. Inspect both before continuing. When either document is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent to a later `$specify-feature` session and missing technical resolution to a later `$design-feature` session; do not invoke either or manufacture its outcome. + +Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user directly which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool or required account is unavailable or not writable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. + +## Process + +### 1. Fix the authorities and native surface + +Inspect the request, repository guidance, approved specification and solution design, their linked product, context, architecture, ADR, and research authorities, and only the implementation evidence needed to understand existing work. Treat the specification as authority for observable intent, the solution design as authority for the selected implementation and genuine technical prerequisites, and the task-management tool as authority for live work state. Discover obvious facts before asking the user. + +Inspect the configured tool's available schema and capabilities before planning any mutation: native item types, bodies, links, parent or grouping concepts, dependency direction, duplicate or related relationships, and read-back operations. Use only concepts the tool actually supports. Keep the tracker read-only through proposal confirmation. + +The source authorities, tracker scope, native capabilities, and access boundary are explicit. + +### 2. Reconcile existing work + +Read the relevant open and closed tracker state before decomposing. Search by authoritative document links or identifiers, feature and outcome language, affected project concepts, and related or parent items. Classify candidates as exact matches, partial overlaps, related work, or unrelated work. Reuse an exact match, propose a bounded reconciliation for a partial overlap, preserve merely related work, and avoid creating a second representation of the same outcome. + +Record the classification and evidence internally. Do not copy live tracker state into repository documentation. The later proposal can distinguish create, update, relationship-only, and no-op reuse actions without inventing a parallel backlog. + +The decomposition starts from a current, duplicate-aware view of native work. + +### 3. Form vertical outcomes + +Build an internal two-way coverage map from every material specification behaviour, rule, state, scenario, constraint, and acceptance condition, plus every necessary solution responsibility, transition, and verification seam, to proposed tracker outcomes. Each outcome must trace back to an approved source, and every approved requirement necessary to deliver the feature must have an owner. + +Form the smallest set of coherent vertical outcomes. Each item spans every affected layer needed to produce one observable result for an actor, consumer, published contract, migration, or operational boundary. Include a shared enabling change in the earliest outcome that needs it unless the approved solution establishes a separately observable and independently verifiable prerequisite. Do not split work by frontend, backend, database, service, class, file, architecture layer, test type, or another implementation inventory merely because those parts exist. One item is correct when the feature is already one bounded independently implementable outcome. + +For each item, define a concise native title, one coherent outcome, enough bounded specification and solution context for a separate implementation session, links to both authoritative feature documents, observable acceptance evidence, and applicable relationships. Derive acceptance evidence from approved rules and discriminating examples; use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. + +Check each item directly for independent value, bounded size, and testability. Minimise coupling between items while representing unavoidable prerequisites honestly. Infer a blocker only when the approved solution or required state transition establishes that one outcome cannot begin or complete before another; do not convert user-journey order, convenient sequencing, or tracker display order into dependency. Interpret the configured tool's relationship direction explicitly and reject cycles. + +When a question exposes missing feature intent or technical design, stop with the owning source gap. Invoke `$elicit` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. + +The draft items are vertically coherent, source-complete, independently actionable where possible, and linked only by genuine acyclic prerequisites. + +### 4. Challenge and propose the set + +Trace every proposed item back to its approved need and every approved need forward to exactly the necessary items. Remove unapproved scope, duplicated outcomes, speculative abstractions, horizontal work, incidental dependencies, and detail that belongs only in code. Confirm that an implementation agent could complete each item in one bounded session without rediscovering feature intent or selecting a missing technical design. + +Repeat the live duplicate search using the final titles, outcomes, authority links, and concepts. Reconcile the result into one complete proposal containing every native create, update, relationship-only, and no-op reuse action; the exact item contents; relationship types and directions; and any native capability limitation or residual mismatch. Do not include estimates, priorities, sprint or cycle assignments, milestones, target dates, releases, implementation sequencing beyond genuine blockers, or speculative future work. + +Present the complete proposal for explicit user confirmation. Selecting an earlier decomposition option is not confirmation of the full set. If the user rejects or revises it, keep the tracker unchanged, reopen every affected coverage, slicing, relationship, and reconciliation decision, repeat the duplicate search, and present the complete revised proposal. The confirmed proposal supplies authority for exactly its agreed tracker mutations without another application question. + +The user and agent share one exact, duplicate-aware mutation plan. + +### 5. Apply the confirmed work + +Immediately before writing, refresh every matching candidate and affected relationship. If concurrent change materially alters the confirmed create, update, no-op, or dependency plan, make no tracker mutation and return to complete-proposal confirmation. Otherwise use the configured tool's native operations to create, update, and link exactly the agreed items. Preserve unrelated content and fields on reused items. Use native idempotency, duplicate, parent, related, and blocking concepts where available; never silently substitute a different meaning when a capability is absent. + +Create in dependency-safe order only where the tool requires identifiers before relationships. Agreement authorises the confirmed writes; do not ask whether to apply them again and do not extend the mutation to convenient adjacent work. + +The configured tracker contains exactly the agreed native representation or an explicitly surfaced tool limitation. + +### 6. Read back, verify, and stop + +Read every created or updated item and every affected relationship back from the tool. Compare titles, outcomes, bounded context, authoritative links, acceptance evidence, relationship types and directions, and create, update, or reuse disposition with the confirmed proposal. Correct an in-scope discrepancy when the agreed value and affected record are unambiguous; stop and report a mismatch when correction would require a new user decision, unsupported capability, or mutation of unrelated work. + +Repeat the relevant duplicate search and verify two-way source coverage, vertical coherence, independent actionability, observable completion, genuine acyclic blockers, and preservation of unrelated native state. Report exact item links, created, updated, related, and reused records, verification performed, unsupported capabilities, concurrent-state or duplicate uncertainty, and unresolved source limitations. + +Stop after the native items and relationships are read back and verified. Do not change product, feature, solution, architecture, or context documents; select or schedule an iteration; estimate or prioritise work; scaffold or implement code; commit or publish changes; release; or deploy. diff --git a/skills/create-tickets/agents/openai.yaml b/skills/create-tickets/agents/openai.yaml new file mode 100644 index 0000000..eb40b46 --- /dev/null +++ b/skills/create-tickets/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Create Tickets' + short_description: 'Create implementation-ready tracker work' + default_prompt: 'Use $create-tickets to turn this approved feature specification and solution design into implementation-ready tracker items.' +policy: + allow_implicit_invocation: false From f2f9f605e9e890ba1e7002e5834d1611a4b68382 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:25:16 +0100 Subject: [PATCH 30/79] feat(commit): preserve unrelated git state --- SKILLS_PLAN.md | 2 +- skills/commit/SKILL.md | 18 +++++++++++------- 2 files changed, 12 insertions(+), 8 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 2f68c25..a5ff35b 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -406,7 +406,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `design-feature` skill - [x] Create `create-tickets` skill - [x] Update `implement` skill -- [ ] Update `commit` skill +- [x] Update `commit` skill - [ ] Replace `pr` with `pull-request` - [ ] Create `review-pull-request` skill - [ ] Create `triage-work` skill diff --git a/skills/commit/SKILL.md b/skills/commit/SKILL.md index 8756033..bef73c3 100644 --- a/skills/commit/SKILL.md +++ b/skills/commit/SKILL.md @@ -1,6 +1,6 @@ --- name: commit -description: Creates coherent Conventional Commits from eligible changed work. Use to commit reviewed changes. +description: Records eligible reviewed local work in coherent Conventional Commits while preserving unrelated changes. Use when the user asks to commit completed work. metadata: invocation: user disable-model-invocation: true @@ -8,22 +8,26 @@ disable-model-invocation: true # Commit -**Conventional Commits** turns eligible changed work into coherent commits whose messages state each change's intent. +**Conventional Commits** records eligible local work in reviewable history without absorbing or disturbing unrelated changes. ## Process -### 1. Inspect the changed work +### 1. Establish the eligible scope -Inspect repository instructions, the current Git state, staged, unstaged, and untracked changes, and recent commit subjects. Apply any requested scope or message constraint. When no eligible change remains, report it and stop; otherwise the complete candidate work is explicit. +Inspect repository instructions, the current branch, recent commit subjects, and the exact staged, unstaged, and untracked state; inspect ignored files when they may explain a change or conceal a requested artefact. Read the diffs, candidate untracked content, and any suspicious new configuration, environment, or credential artefact, then apply any user-supplied scope or message constraint. Treat the request as authority to commit work whose ownership and intent are clear, but do not assume every change belongs to it. Preserve pre-existing and unrelated work, including user-staged content. Ask only when material ownership or scope remains ambiguous after inspection. + +Exclude work that is conflicted, clearly incomplete, unsafe to attribute, or contains a credential or other sensitive material, and report the exact blocker instead of changing source, tests, or documentation to make it eligible. Do not substitute implementation, review, or broad verification for this inspection. When nothing eligible remains, report the state and stop; otherwise every candidate change has an explicit owner and intent. ### 2. Group coherent changes -Partition the candidate work into **atomic commits** by coherent intent. Keep changes together when they serve the same purpose and leave unrelated or ambiguous work untouched. Each group has one explainable purpose. +Partition eligible work into **atomic commits** by coherent intent. Keep code, tests, documentation, configuration, and generated output together when they realise one purpose; split genuinely independent purposes only when each resulting commit remains understandable and complete. Use hunk-level grouping when necessary and safe. When unrelated content is already staged, isolate a whole-path group with a path-limited commit only after verifying its exact content and that the existing index will be preserved; leave a partial-path or otherwise unsafe group uncommitted. Each proposed group has one truthful outcome and exact paths or hunks. ### 3. Create the commits -For each group, stage its exact files or hunks, inspect the staged diff, and commit it with an accurate `type[(scope)][!]: description` message using the repository's Git setup. Ask before altering ambiguous user-staged work. On failure, preserve the resulting Git state and report the blocker. +For each group, stage only its intended paths or hunks, then inspect the complete staged diff and staged name/status before committing; for a safely isolated path-limited group, inspect the exact content that command will record and the preserved index instead. Follow the repository's supported Conventional Commit types, scopes, bodies, and trailers where established; otherwise use `type[(scope)][!]: description`. State the intent or outcome rather than a file list, represent breaking changes accurately, and honour a requested message or scope only when it truthfully describes the complete commit content. + +Create the commit with the repository's Git setup and allow its hooks to run. Do not bypass a hook, rewrite history, or use destructive cleanup. If staging, a hook, or the commit changes files or fails, inspect the resulting state, preserve it, and report the exact result instead of blindly recommitting or discarding anything. A commit exists only when Git reports a new object for the intended staged content. ### 4. Verify the result -Verify each created commit and inspect the remaining status. Return each hash and subject plus any work left uncommitted. +After each commit, verify its object, hash, subject, and changed content. After all groups, inspect the branch status, staged and unstaged diffs, and remaining untracked work. Return every created hash and subject, all remaining work, and any limitation; a created commit is not evidence of a clean tree. Stop without pushing, opening or updating a pull request, merging, releasing, deploying, or changing the implementation. From d7db26ca50e7d9f9447bc9873923dd9fe77ab453 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:30:01 +0100 Subject: [PATCH 31/79] feat(pull-request): publish complete branch outcomes --- SKILLS_PLAN.md | 2 +- skills/pr/SKILL.md | 29 ------------- skills/pull-request/SKILL.md | 43 +++++++++++++++++++ .../{pr => pull-request}/agents/openai.yaml | 2 +- 4 files changed, 45 insertions(+), 31 deletions(-) delete mode 100644 skills/pr/SKILL.md create mode 100644 skills/pull-request/SKILL.md rename skills/{pr => pull-request}/agents/openai.yaml (57%) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index a5ff35b..7c5826f 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -407,7 +407,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `create-tickets` skill - [x] Update `implement` skill - [x] Update `commit` skill -- [ ] Replace `pr` with `pull-request` +- [x] Replace `pr` with `pull-request` - [ ] Create `review-pull-request` skill - [ ] Create `triage-work` skill - [ ] Create `refine-backlog` skill diff --git a/skills/pr/SKILL.md b/skills/pr/SKILL.md deleted file mode 100644 index 475567d..0000000 --- a/skills/pr/SKILL.md +++ /dev/null @@ -1,29 +0,0 @@ ---- -name: pr -description: Publishes the current branch and creates or updates its pull request. Use when committed work is ready for review. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Pull Request - -Publish the current branch through the repository's available setup and represent its complete work in one pull request. - -## Process - -### 1. Inspect the repository setup - -Inspect repository instructions, Git status, the current branch, its remote and base, available publication tooling, and any existing pull request for the branch. Use the user-supplied base or the repository default. When no usable publication path exists, report the blocker; otherwise the publication context is explicit. - -### 2. Establish the grouped work - -Invoke `$commit` when eligible changed work remains, then inspect the commit history and **whole-branch change scope** against the base. When the branch has no publishable change, report it and stop; otherwise the complete pull-request scope is explicit. - -### 3. Write the pull request - -Write a `type[(scope)][!]: description` title that summarizes the complete grouped work. Apply **BLUF** by writing one succinct paragraph that begins directly with what the grouped work changes and why. Follow additional repository requirements only when they explicitly mandate them. - -### 4. Publish and verify - -Use the available repository mechanism to publish the current branch without rewriting remote history and create or update its one pull request. Make it ready for review unless the user requested a draft. Verify the head, base, title, body, and review state, then return the pull-request URL. diff --git a/skills/pull-request/SKILL.md b/skills/pull-request/SKILL.md new file mode 100644 index 0000000..9625b63 --- /dev/null +++ b/skills/pull-request/SKILL.md @@ -0,0 +1,43 @@ +--- +name: pull-request +description: Publishes the current branch as one accurate pull request. Use when the branch is ready for external review. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Pull Request + +Publish the current branch through the repository's available setup and represent its complete committed work in one verified pull request. Treat invocation, plus resolution of any material ambiguity, as authority for its intrinsic commit, push, and pull-request mutations; do not ask for a separate application confirmation. + +## Process + +### 1. Establish the publication target + +Inspect repository instructions and pull-request templates, Git status, the current branch and revision, remotes and upstreams, the repository's default or user-supplied base, remote branch state, available host tooling and authentication, and every pull request associated with the head branch. Resolve the exact head, base, remote, and existing pull request from this evidence. Ask only when branch ownership, the base, the remote, or a requested draft-state transition remains materially ambiguous. + +Use the repository's native publication path. When none is available or authenticated, report the exact access requirement and stop. Do not publish from the default or another inappropriate protected branch; when a matching pull request is closed, merged, or otherwise not safely editable, ask how the user wants to proceed rather than creating a duplicate. The intended publication target and the one pull request to create or update are explicit before mutation. + +### 2. Resolve local work + +Inspect the content and ownership of all staged, unstaged, and untracked changes. When eligible coherent work belongs to the requested branch outcome, invoke `$commit` with that scope, then re-inspect the branch, revision, and complete worktree. Leave ambiguous, unrelated, ineligible, or intentionally excluded work untouched and disclose it; a dirty tree does not prevent publication when the branch's committed result is complete. + +Compare the complete branch history and whole-branch diff against the chosen base, not only the latest commit or current transcript. When the base has no meaningful branch change, report that state and stop. The exact committed outcome to publish is fixed. + +### 3. Prepare the complete pull request + +Shape the title through **Conventional Commits** as `type[(scope)][!]: description`, truthfully summarizing the whole branch. Apply **BLUF** by starting the body directly with what the grouped change does and why. Then include concise material verification evidence and limitations, linked issue or specification context supported by the branch evidence, and breaking, migration, compatibility, or operational effects when present. + +Follow explicit repository templates and instructions where they require a different shape. Do not invent tests, issue-closing keywords, links, rollout claims, or verification. Prefer an outcome summary over a commit list or file inventory. The title and body account for all material branch work without overstating its evidence. + +### 4. Publish one pull request + +Push the current revision with the repository's normal mechanism, setting its upstream when needed. Do not rewrite history or force-push. When the remote cannot fast-forward, authentication fails, or publication otherwise fails, preserve the current state, report the exact blocker, and stop rather than substituting a destructive or different publication route. + +Create one pull request when none exists, or update the matching pull request's base, title, and body as needed for the confirmed outcome. Make a new pull request ready for review unless the user requested a draft or ready status would be dishonest. Preserve an existing pull request's draft or ready state unless the user requested or confirmed a transition, and preserve unrelated labels, reviewers, discussion, and review state. + +### 5. Verify the published result + +Read back the remote head revision and the pull request's identity, URL, head, base, title, complete body, and draft or ready state. Verify that the remote head equals the intended local revision and every field matches the confirmed publication. Correct an unambiguous in-scope discrepancy and read it back again; otherwise report the exact unsupported capability or external blocker. + +Return whether the pull request was created or updated, its URL and verified head/base/state, the verification evidence represented, and any excluded local work or limitation. Stop without merging, approving, reviewing the author's work, releasing, deploying, deleting a branch, changing tracker state beyond supported pull-request text, or beginning implementation. diff --git a/skills/pr/agents/openai.yaml b/skills/pull-request/agents/openai.yaml similarity index 57% rename from skills/pr/agents/openai.yaml rename to skills/pull-request/agents/openai.yaml index b844df6..c155d54 100644 --- a/skills/pr/agents/openai.yaml +++ b/skills/pull-request/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Pull Request' - short_description: 'Publish a concise pull request' + short_description: 'Publish one accurate branch pull request' policy: allow_implicit_invocation: false From 0a34937e09a234465cb5917f4b42139137a99600 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:37:34 +0100 Subject: [PATCH 32/79] feat(review-pull-request): add external change assessment --- SKILLS_PLAN.md | 2 +- skills/review-pull-request/SKILL.md | 61 +++++++++++++++++++ skills/review-pull-request/agents/openai.yaml | 5 ++ 3 files changed, 67 insertions(+), 1 deletion(-) create mode 100644 skills/review-pull-request/SKILL.md create mode 100644 skills/review-pull-request/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 7c5826f..96c69c5 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -408,7 +408,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `implement` skill - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` -- [ ] Create `review-pull-request` skill +- [x] Create `review-pull-request` skill - [ ] Create `triage-work` skill - [ ] Create `refine-backlog` skill - [ ] Create `plan-iteration` skill diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md new file mode 100644 index 0000000..bd8e06e --- /dev/null +++ b/skills/review-pull-request/SKILL.md @@ -0,0 +1,61 @@ +--- +name: review-pull-request +description: Independently reviews someone else's pinned pull request with code findings and verification evidence. Use when assessing an external contributor's pull request. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Review Pull Request + +Coordinates `$code-review`'s tailored **Software Formal Inspection** and `$verify-change`'s **bidirectional verification traceability** against one frozen pull-request revision, then reports the independent findings and verification evidence without changing the contributor's work. + +## Process + +### 1. Resolve the review target + +Inspect the request, repository guidance, available host tooling and authentication, pull-request metadata and discussion, repository state, and relevant product, architecture, context, ADR, specification, ticket, and acceptance authorities. Resolve one exact repository and pull-request identity. Confirm intent when the selected pull request belongs to the user or current author's branch; this skill normally reviews work submitted by someone else rather than duplicating `$implement`'s author-side review. + +Identify the intended-behaviour authority without inferring intent from the implementation. When none exists, follow `$code-review`'s Standards-only confirmation rule. The review target, authorship, authorities, and requested report-only or publication outcome are explicit. + +### 2. Freeze the complete candidate + +Resolve and record the pull request's exact base SHA and head SHA, full patch, commits and revisions, changed paths and complete changed files, relevant current code, discussion, and check evidence. Review the whole pull request rather than its latest commit, comment, or visible excerpt. + +Prefer host data or a checkout already representing the pinned candidate. When execution needs another filesystem view, use an isolated disposable worktree or temporary checkout only when its boundary and ownership are explicit. Do not switch, reset, clean, or overwrite the user's working tree. Stop with the exact access or isolation requirement when no safe complete view is available. The candidate is complete, immutable for the review, and safe to inspect. + +### 3. Run the independent code review + +Invoke `$modular-design` only when the complete candidate presents a material ownership, boundary, contract, dependency, seam, or change-propagation concern that needs focused structural constraints. Supply its result to `$code-review`; modular design owns the structural assessment while code review alone owns any retained review finding. Invoke `$research` only when the assessment depends on a material external-knowledge question whose evidence must survive the session and its report can be persisted without altering the contributor candidate. Supply that cited report as evidence rather than letting it decide a finding. When no safe durable-report boundary exists, state the limitation instead of changing contributor work. + +Invoke `$code-review` with the frozen base and head SHAs, complete candidate, intended-behaviour and Standards authorities, and a read-only boundary. Require its independent Standards and applicable Spec reviews. Preserve their findings, distinctions, priorities, evidence, ordering, omitted axes, and limitations without merging, suppressing, cross-axis reranking, adjudicating, or remediating them. + +Re-read the remote head after the review. If it differs from the pinned head, retain the result as stale for its original revision. Pin the new base and head and repeat the complete review only when the user's requested outcome still identifies the same pull request unambiguously; otherwise stop and ask for direction. The current code-review result refers to one explicit candidate. + +### 4. Verify the same candidate + +Invoke `$verify-change` on exactly the base and head SHAs reviewed in step 3. Supply the intended-behaviour and repository authorities, material risks exposed by the complete change, relevant existing CI evidence, and the isolated execution boundary. Let it select and appraise applicable existing required and risk-triggered harnesses and return claim-level evidence, exact commands or procedures, limitations, residual risk, and its verification-specific verdict. + +Use existing CI evidence only when its provenance ties it to the pinned head and its execution boundary and result are inspectable. When local verification cannot run safely or the repository lacks faithful infrastructure, retain the exact `not run`, partial, or unverified evidence; do not install tools, add tests or ad hoc harnesses, alter source, tests, snapshots, baselines, suppressions, dependencies, or checks, or treat static inspection as execution. Passing checks do not erase a code-review finding, and verification never certifies the pull request. + +Re-read the remote head after verification. If it changed, preserve both results as stale for their original revision and repeat steps 2–4 on the newly pinned revision only while the requested outcome remains clear. Stop with the stale results and exact blocker when repeated drift prevents one stable current assessment. The completed review and verification evidence refer to identical current SHAs or are explicitly stale. + +### 5. Report the assessment + +Report: + +- the repository, pull request, frozen base and head SHAs, full-change scope, and any drift; +- the intended-behaviour authority or confirmed Standards-only basis, plus applicable repository and technical authorities; +- the independent Standards and Spec findings in their returned sections and order, including `No findings.` only for a genuinely current completed axis; +- the verification verdict and claim-level evidence, exact commands or procedures, execution boundaries, and material results; +- existing pull-request check evidence whose provenance was accepted; +- every omitted, unsafe, unavailable, unperformed, or stale review or verification axis; and +- limitations, residual risks, and a concise finding-count and evidence summary. + +Do not add an approval, request-changes, merge, or other disposition merely from green checks or the absence of findings. Provide a disposition only when the user's requested outcome includes it and the complete current evidence supports it. + +### 6. Publish only an included outcome + +Stop after the chat report by default. Do not publish comments, approve, request changes, merge, close, edit pull-request metadata, change tracker state, modify the contributor branch, commit, push, release, or deploy. + +When the invocation explicitly includes publishing review feedback or a host review state, present the exact supported comments and disposition for confirmation as part of resolving that outcome. After agreement, publish only that content through the repository's native host tooling without another application gate, then read back the comments and review state. Never publish an unsupported conclusion or merge, fix, or otherwise advance the pull request. diff --git a/skills/review-pull-request/agents/openai.yaml b/skills/review-pull-request/agents/openai.yaml new file mode 100644 index 0000000..66d5bd3 --- /dev/null +++ b/skills/review-pull-request/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Review Pull Request' + short_description: 'Independently assess an external pull request' +policy: + allow_implicit_invocation: false From 8727898ba252e0b7471b527a9972c6809d3c6876 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 09:40:29 +0100 Subject: [PATCH 33/79] feat(triage-work): add evidence-led intake routing --- SKILLS_PLAN.md | 2 +- skills/triage-work/SKILL.md | 69 +++++++++++++++++++++++++++ skills/triage-work/agents/openai.yaml | 6 +++ 3 files changed, 76 insertions(+), 1 deletion(-) create mode 100644 skills/triage-work/SKILL.md create mode 100644 skills/triage-work/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 96c69c5..da10e06 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -409,7 +409,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` - [x] Create `review-pull-request` skill -- [ ] Create `triage-work` skill +- [x] Create `triage-work` skill - [ ] Create `refine-backlog` skill - [ ] Create `plan-iteration` skill - [ ] Update `review-architecture` skill diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md new file mode 100644 index 0000000..0e4f819 --- /dev/null +++ b/skills/triage-work/SKILL.md @@ -0,0 +1,69 @@ +--- +name: triage-work +description: Triages bounded incoming software requests into verified native tracker dispositions and lifecycle routes. Use for recurring intake review before backlog refinement. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Triage Work + +Determines what each selected incoming request is, records an evidence-led native disposition, and routes it to the right later lifecycle session without performing that work. + +## Prerequisites + +Require an explicitly bounded set of incoming items or queue. When the scope is unclear, resolve only the selection before inspecting or mutating items outside it. + +Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool, account, selected queue, source evidence, or required write operation is unavailable, stop the affected item with the exact access or connection needed rather than falling back to Markdown or another tracker. + +## Process + +### 1. Establish the intake surface + +Inspect the request, repository guidance, configured tool, selected items, and their current native state before mutation. Inspect the tool's available item types, statuses, resolutions, labels or classifications, comments, links, duplicate and related relationships, transfer or routing operations, and read-back capabilities. Use only concepts and meanings the current project and tool actually support. + +For each selected item, read its description, attachments, discussion, source report, prior dispositions, and linked work. Treat reporter wording as a claim rather than the classification. When the source is not yet represented in the tracker, include the smallest native intake record needed to preserve its evidence and route in the later proposal; do not turn it into an implementation ticket. The bounded items, available evidence, native capabilities, and access limits are explicit. + +### 2. Establish enough evidence to classify + +Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, `ARCHITECTURE.md`, ADRs, documentation, public contracts, tests, code, operational evidence, and external authorities only as far as each item needs. Distinguish intended behaviour, current observable behaviour, reported observation, and unsupported inference. Seek enough corroboration to decide a disposition, not a root cause or solution. A safe focused observation may establish that a defect report is credible; full reproduction, causal investigation, mitigation, repair, or feature discovery belongs to its later routed session. + +Search open and closed tracker work and relevant repository records using source identifiers, authoritative links, actors, outcomes, symptoms, affected concepts, and discriminating details. Classify a candidate internally as exact match, possible overlap, related work, or unrelated. Mark a duplicate only when the evidence establishes the same requested outcome or fault; preserve an uncertain resemblance as related or unresolved and state what would distinguish it. + +When a material external evidence question could change the disposition and its answer must survive this session, invoke `$research` with that bounded question, then read and verify the report's material claims and citations before using it. Invoke `$maintain-context` only when the evidence independently confirms consequential durable project language or qualifying rationale; a proposed classification or route alone does not establish either. Verify its returned context result before continuing. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. + +### 3. Choose one honest disposition per item + +Classify each item by the lifecycle outcome it actually needs: + +- route a new or changed externally observable capability to a later `$specify-feature` session; +- route incorrect existing behaviour to a later `$debug` session; +- route active production impact requiring stabilisation to a later `$respond-to-incident` session; +- route removal or deprecation intent to a later `$plan-retirement` session; +- link an exact existing representation and use the supported duplicate or equivalent native resolution; +- link actionable work whose intent and solution are already defined to its existing appropriate work without creating another specification, design, or ticket; or +- close answered, invalid, not-actionable, or out-of-scope requests only with a concise reason grounded in inspected evidence. + +When evidence cannot yet support one of those outcomes, retain an honest native needs-information or unresolved disposition when available and identify the minimum missing fact and who can supply it. Invoke `$elicit` only when that fact or a materially different classification or disposition depends on user-held knowledge or authority. Supply the evidence and viable consequences, use only its confirmed synthesis, and retain ownership of proposal, mutation, verification, and stopping. Do not ask the user for repository or tracker facts that inspection can establish. + +For each item, prepare a proposed native disposition containing its classification, concise evidence-led reason, exact duplicate or related links, supported state and field changes, precise next route and required input, and any native capability limitation. Do not add priority, readiness, estimates, iteration placement, ticket decomposition, root cause, solution, or speculative downstream work. Every selected item has one supported disposition or one explicit evidence blocker. + +### 4. Confirm the mutation plan + +Recheck each proposal against the source evidence, current authorities, possible duplicates, and the native meanings discovered in step 1. Present the complete material mutation plan: records to create, close, update, transfer, comment on, classify, or link; exact relationship meanings and targets; reasons; routes; no-op items; and limitations. Keep the tracker unchanged while resolving material decisions. + +Obtain explicit user confirmation of the complete plan. Selecting an earlier classification option is not confirmation of all mutations. If the user rejects or revises a proposal, reopen every affected evidence, classification, duplicate, and routing decision and present the complete revised plan. Confirmation supplies authority for exactly the agreed mutations, including uncontroversial native fields mechanically entailed by them, without another application prompt. + +### 5. Apply and verify the agreed dispositions + +Immediately before writing, re-read every affected item, relationship, and duplicate candidate. If drift changes an agreed disposition or target, leave that item unchanged and return its revised complete proposal for confirmation; continue independent batch items only when their plans remain safe. + +Apply only the agreed native mutations and relationships. Preserve unrelated content, discussion, fields, and metadata. When the tool lacks an agreed classification, resolution, link, or route, use another supported native operation only when it preserves the exact confirmed meaning; otherwise leave that part unapplied and report the capability limit rather than simulating it with a misleading field or relationship. + +Read back every affected record and relationship. Verify source preservation, classification, reason, state, fields, duplicate or related target and direction, next route, and absence of unintended changes against the confirmed plan. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop that item and report the mismatch. Each applied disposition is verified in its owning system. + +### 6. Report and stop at intake + +Report every selected item with its native link, applied or no-op disposition, evidence basis, duplicate or related work, next lifecycle route or missing input, read-back evidence, limitations, and unresolved uncertainty. Distinguish completed, blocked, and unchanged items so a blocker on one does not overstate or erase safe batch progress. + +Stop after the native dispositions and routes are verified. Do not invoke the routed lifecycle skill or perform feature elicitation or specification, solution design, ticket decomposition, backlog scope or readiness refinement, priority or iteration planning, root-cause debugging, repair, incident stabilisation, retirement planning, implementation, commit, publication, release, or deployment. Live work state remains in the configured tracker rather than a repository mirror. diff --git a/skills/triage-work/agents/openai.yaml b/skills/triage-work/agents/openai.yaml new file mode 100644 index 0000000..710eafa --- /dev/null +++ b/skills/triage-work/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Triage Work' + short_description: 'Triage incoming work into verified routes' + default_prompt: 'Use $triage-work to triage these incoming requests into verified tracker dispositions and lifecycle routes.' +policy: + allow_implicit_invocation: false From b2ac485297125467094f26a4ea853a72e0be5b80 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 10:08:27 +0100 Subject: [PATCH 34/79] feat(plan-iteration): add feasible native work selection --- SKILLS_PLAN.md | 2 +- skills/plan-iteration/SKILL.md | 65 ++++++++++++++++++++++++ skills/plan-iteration/agents/openai.yaml | 6 +++ 3 files changed, 72 insertions(+), 1 deletion(-) create mode 100644 skills/plan-iteration/SKILL.md create mode 100644 skills/plan-iteration/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index da10e06..9fda5c2 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -411,7 +411,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `review-pull-request` skill - [x] Create `triage-work` skill - [ ] Create `refine-backlog` skill -- [ ] Create `plan-iteration` skill +- [x] Create `plan-iteration` skill - [ ] Update `review-architecture` skill - [ ] Update `debug` skill - [ ] Create `respond-to-incident` skill diff --git a/skills/plan-iteration/SKILL.md b/skills/plan-iteration/SKILL.md new file mode 100644 index 0000000..9002954 --- /dev/null +++ b/skills/plan-iteration/SKILL.md @@ -0,0 +1,65 @@ +--- +name: plan-iteration +description: Selects and records a feasible body of ready, unblocked work in the project's native planning mechanism. Use when planning the next delivery period or ordered next-work queue from a refined backlog. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Plan Iteration + +Selects a coherent body of work that current evidence makes feasible for the project's next delivery period, then records and verifies that selection in the configured task-management tool. The result is a forecast of selected work, not a promise of completion or a release plan. + +## Prerequisites + +Require an explicitly bounded backlog whose candidate items have already been refined. An eligible item has observable bounded scope, a current priority, enough approved intent and technical direction for its next implementation step, and no unresolved blocker. Exclude an item that fails any condition, preserve the exact reason, and route it to a later `$refine-backlog` session; do not refine, estimate, split, or rescope it here. + +Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or introduce a Propulsion manifest, adapter, configuration, or state model. When the configured tool, account, bounded backlog, or required read or write operation is unavailable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. + +## Process + +### 1. Establish the native planning surface + +Inspect the request, repository guidance, bounded backlog, configured tool, and current planning state before proposing any mutation. Discover the tool's native sprint, cycle, milestone, or ordered-queue concepts; period identity and dates; membership and ordering; dependency direction; priority, readiness, sizing, capacity, and assignment fields; existing selections; and read-back operations. Use only concepts with the meanings established by the project and tool. + +Determine the next planning horizon from project evidence or user authority. A formal-period project uses its established or explicitly confirmed native period. A project without a formal period uses only its existing native ordered next-work mechanism; tool support alone does not establish that the project uses sprints, cycles, or milestones. When neither a native period nor an ordered next-work mechanism can represent the outcome, report the capability gap and agree only a representation the configured tool actually supports. Do not create a repository mirror. + +The planning horizon, native representation, existing membership, and capability limits are explicit. + +### 2. Establish honest capacity and eligibility + +Read current project-native evidence for the horizon: comparable historical throughput or sizing where it exists, current commitments, team availability, specialist constraints, dependency timing, operational load, and other work that consumes the same capacity. Never invent velocity, estimates, dates, availability, or precision, and do not treat points or item counts as interchangeable capacity unless the project's own evidence supports that use. + +When evidence cannot resolve a material capacity, horizon, or trade-off decision, invoke `$elicit` with the evidence and viable consequences. Let it ask one user-authoritative question at a time and use only its confirmed synthesis; retain ownership of selection, mutation, verification, and stopping. State the weakest honest capacity bound when precise capacity is unavailable rather than manufacturing a number. + +Refresh each candidate's observable scope, current priority, readiness for its next implementation step, dependencies, blockers, and current native state. Treat a relationship as blocking only when the dependency meaning and evidence show the item cannot begin or complete without it. Do not confuse preference, display order, or convenient sequence with a blocker. The eligible set, exclusions with reasons, and capacity assumptions and limitations are current and explicit. + +### 3. Form one feasible selection + +Consider eligible items in current priority order. Select a coherent body subject to the capacity evidence, whole-item fit, genuine dependencies, specialist availability, operational commitments, and material delivery risk. Include a prerequisite only when it is ready, necessary, and feasible in the same horizon; exclude work blocked by an external or out-of-horizon prerequisite. + +Do not fill capacity automatically, add lower-priority work merely to maximise utilisation, or promise that selected work will finish. Preserve a high-priority item that does not fit as an exclusion rather than splitting, rescoping, or lowering its readiness standard. Keep intentionally unused capacity visible when uncertainty, operational load, item shape, or a constrained capability makes that the honest plan. + +The proposed body is priority-led, dependency-valid, coherent, and feasible within the stated evidence without changing the backlog. + +### 4. Confirm the complete native plan + +Present one complete proposal containing the planning horizon, selected items, native membership and order, exclusions and reasons, included prerequisites, capacity evidence and assumptions, existing assignments or commitments being preserved, exact intended mutations, capability limitations, and residual feasibility risk. Keep the tracker unchanged while resolving material decisions. + +Obtain explicit user confirmation of the complete proposal. If the user rejects or revises it, reopen every affected capacity, eligibility, dependency, coherence, ordering, and representation decision and present the complete revised proposal. Confirmation supplies authority for exactly the agreed native mutations without another application question. + +The user and agent share one exact, feasible mutation plan. + +### 5. Refresh and apply the agreed selection + +Immediately before writing, re-read the target period or queue, every selected or displaced membership and order, relevant capacity state, priorities, readiness, blockers, and dependencies. When drift materially changes eligibility, feasibility, membership, order, or an agreed mutation, keep the affected state unchanged and return to complete-proposal confirmation. Preserve an independent agreed mutation only when the drift cannot affect its feasibility or meaning. + +Apply exactly the confirmed native assignments, membership, and order. Preserve unrelated period, queue, backlog, field, relationship, and assignment state. Use the tool's native operations and meanings; when an agreed capability is unsupported, leave that part unapplied and report the limitation rather than simulating it with a misleading field, relationship, or prose record. + +The configured tracker contains exactly the safely applicable agreed selection. + +### 6. Read back, verify, and stop + +Read back the target period or queue, selected membership and order, affected relationships and assignments, and every changed field. Compare them with the confirmed proposal and verify that selected items remain ready and unblocked, prerequisites and dependency directions are valid, capacity assumptions still hold within their stated precision, and unrelated native state was preserved. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop and report the mismatch or unsupported capability. + +Report the verified native period or queue, selected and excluded items with reasons, applied and no-op mutations, capacity evidence and limitations, read-back evidence, unsupported capabilities, drift, and residual feasibility risk. Stop after the selected body is recorded and verified. Do not refine or estimate backlog work, change specifications or designs, create tickets, implement, commit, publish, release, or deploy. diff --git a/skills/plan-iteration/agents/openai.yaml b/skills/plan-iteration/agents/openai.yaml new file mode 100644 index 0000000..d7b5000 --- /dev/null +++ b/skills/plan-iteration/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Plan Iteration' + short_description: 'Select a feasible body of ready work' + default_prompt: 'Use $plan-iteration to select and record a feasible body of ready work for the next delivery period.' +policy: + allow_implicit_invocation: false From fb4564780c22dc650d35c43a24a6390c095c3c95 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 10:09:28 +0100 Subject: [PATCH 35/79] feat(refine-backlog): add evidence-led backlog health --- SKILLS_PLAN.md | 2 +- skills/refine-backlog/SKILL.md | 79 ++++++++++++++++++++++++ skills/refine-backlog/agents/openai.yaml | 6 ++ 3 files changed, 86 insertions(+), 1 deletion(-) create mode 100644 skills/refine-backlog/SKILL.md create mode 100644 skills/refine-backlog/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9fda5c2..531490d 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -410,7 +410,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Replace `pr` with `pull-request` - [x] Create `review-pull-request` skill - [x] Create `triage-work` skill -- [ ] Create `refine-backlog` skill +- [x] Create `refine-backlog` skill - [x] Create `plan-iteration` skill - [ ] Update `review-architecture` skill - [ ] Update `debug` skill diff --git a/skills/refine-backlog/SKILL.md b/skills/refine-backlog/SKILL.md new file mode 100644 index 0000000..3abddfa --- /dev/null +++ b/skills/refine-backlog/SKILL.md @@ -0,0 +1,79 @@ +--- +name: refine-backlog +description: Refines a bounded software backlog into verified current, prioritised, and honestly ready native tracker work. Use for recurring backlog health and ordering sessions before iteration planning. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Refine Backlog + +Keeps a selected backlog aligned, coherent, and usable for later delivery decisions without committing work to an iteration or manufacturing missing product or technical definition. + +## Prerequisites + +Require an explicitly bounded native backlog view or item set. Resolve only the selection when scope is unclear; do not inspect or mutate work outside it. + +Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, treat their answer as authority to record that confirmed preference, invoke `$maintain-agents`, verify the root instruction, and resume without a separate permission question. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool, selected backlog, linked authority, or required read or write operation is unavailable, stop the affected work with the exact access, connection, or evidence needed rather than falling back to Markdown or another tracker. + +## Process + +### 1. Establish the refinement surface + +Inspect the request, repository guidance, configured tool, bounded backlog, and current project direction before mutation. Inspect the tool's native item types, fields, statuses, priorities or ordering, parent and split concepts, duplicate and related relationships, blocker direction, readiness signals, and read-back capabilities. Use only meanings supported by the current project and tool; do not invent a universal workflow or Definition of Ready. + +Read every selected item, its discussion and evidence, linked feature, product, solution, architecture, or other owning authorities, and its native relationships. Search relevant open and closed work for duplicates and overlap. Inspect only enough current product, documentation, code, tests, or operational evidence to judge whether the item remains current. Treat stale tracker text and reporter wording as claims rather than authority. + +The bounded items, available native operations, current authorities, and access limits are explicit. + +### 2. Assess every selected item + +Assess each item against the evidence applicable to its lifecycle position: + +- whether the work is still needed and aligned with current project direction; +- whether its scope expresses one coherent observable outcome; +- whether linked intent and technical design are sufficient for the next step it claims to be ready for; +- whether its acceptance evidence makes completion observable; +- whether its size and internal coherence support one bounded later session; +- whether it duplicates or overlaps existing work; +- whether dependencies and blockers are genuine, acyclic, and represented in the correct direction; +- how its priority compares with the other bounded work under the user's current product direction; and +- whether it is honestly ready for its next step using project-native evidence and fields, assessed separately from availability unless the project's native meaning explicitly includes blockers. + +Do not derive priority from age, reporter urgency, effort, tracker order, or agent preference. Priority is a relative product decision owned by the user or another explicit project authority. Do not invent missing feature intent or technical design, estimates, acceptance evidence, or readiness claims. Identify the exact source gap, keep the item honestly not ready using a supported native representation, and route resolution to a later `$specify-feature`, `$design-feature`, or other owning session without invoking it. Preserve the distinction between an item that is sufficiently refined but blocked and one whose definition is incomplete; do not lower readiness merely to duplicate a separately represented blocker unless that is the project's confirmed native meaning. + +Classify matching work as an exact duplicate, partial overlap, related work, or unrelated work from discriminating evidence. Split an existing overbroad implementation item only into coherent outcomes already authorised by its sources. When a feature-level placeholder links approved specification and solution work but has no implementation tickets, do not create children or otherwise perform its initial decomposition: leave that outcome to a later `$create-tickets` session. Renaming the proposed children does not make that decomposition refinement. + +Every item has an evidence-led health assessment, relative-priority basis or open decision, and honest next-step readiness. + +### 3. Form the complete proposal + +Prepare the smallest native change set that makes the bounded backlog current. It may clarify an item, split supported scope, reconcile or close obsolete, duplicate, invalid, or fully superseded work with evidence, correct relationships and blocker directions, set an agreed native priority or order, and represent honest readiness. Preserve a no-op when current state already expresses the supported result. + +Invoke `$elicit` only when a material scope, priority, or readiness decision depends on user-held knowledge or authority. Supply inspected evidence, viable options, and consequences; let it ask one question at a time and use only its confirmed synthesis. Retain responsibility for the complete proposal, mutation, verification, and stopping. + +Recheck duplicates and affected relationships, then present one complete proposal containing every create, update, split, reconciliation, closure, link, priority, readiness, and no-op action; exact relationship meanings and directions; evidence-led reasons; later routes; and native capability limitations. Keep the tracker unchanged while decisions remain open. Obtain explicit user confirmation of the whole set. Selecting one earlier option is not confirmation. If the user rejects or revises any part, reopen every affected assessment and relationship, then present the complete revised set. Confirmation authorises exactly the agreed native mutations without another application prompt. + +The user and agent share one exact, duplicate-aware refinement plan. + +### 4. Refresh and apply the agreed changes + +Immediately before writing, re-read every affected item, matching candidate, and relationship. When drift materially changes an agreed action, leave that item and its coupled mutations unchanged and return its complete revised proposal for confirmation. Continue only independent batch items whose agreed changes remain safe. + +Apply only the confirmed native mutations and preserve unrelated content, discussion, fields, and metadata. Create only relationships required to express an evidenced split, reconciliation, duplicate, overlap, or genuine blocker; native parent or split structure does not also justify an incidental related link between siblings. Use a different supported native operation only when it preserves the exact confirmed meaning; otherwise leave that part unapplied and report the capability limit. Do not turn a convenient ordering preference into a blocker or extend the mutation to adjacent backlog work. + +The configured tracker contains every still-valid agreed refinement or an explicit item-level blocker. + +### 5. Read back and verify the backlog + +Read every affected record and relationship back from the tool. Compare scope, observable outcome, authority links, acceptance evidence, state, priority or order, readiness, parent or split structure, duplicate disposition, and blocker type, target, and direction with the confirmed proposal. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop the affected item and report the mismatch. + +Repeat the relevant duplicate and overlap search. Verify that each selected item is current, supported at its claimed lifecycle position, coherent, relatively prioritised where authority permits, honest about readiness, and connected only by genuine blockers. Verify closure reasons and preservation of unrelated native state. A blocked item does not erase verified independent progress. + +The refined backlog and any partial completion are reproducibly verified in the owning tracker. + +### 6. Report and stop before planning + +Report each selected item with its native link, applied or no-op disposition, evidence basis, resulting priority and readiness, duplicates or overlap, dependencies and blockers, later owning route, read-back evidence, limitations, and unresolved decisions. Distinguish completed, blocked, drifted, and unchanged work. + +Stop after backlog health and relative ordering are current and verified. Do not select or assign a sprint, cycle, milestone, iteration, release, or delivery body; establish implementation sequence beyond genuine blockers; invoke feature specification, solution design, or ticket creation; implement, debug, commit, publish, release, or deploy. Live backlog state remains in the configured tracker rather than a repository mirror. diff --git a/skills/refine-backlog/agents/openai.yaml b/skills/refine-backlog/agents/openai.yaml new file mode 100644 index 0000000..91c58ca --- /dev/null +++ b/skills/refine-backlog/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Refine Backlog' + short_description: 'Keep backlog work current and honestly ready' + default_prompt: 'Use $refine-backlog to refine this bounded backlog into verified current, prioritised work.' +policy: + allow_implicit_invocation: false From 673a7847b607908b1b0965be35a0a2b5a27e9fe7 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 10:26:53 +0100 Subject: [PATCH 36/79] feat(debug): require causal reviewed repairs --- SKILLS_PLAN.md | 2 +- skills/debug/SKILL.md | 44 ++++++++++++++++++--------- skills/debug/agents/openai.yaml | 4 +-- skills/debug/references/TECHNIQUES.md | 4 +-- 4 files changed, 34 insertions(+), 20 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 531490d..f468aab 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -413,7 +413,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `refine-backlog` skill - [x] Create `plan-iteration` skill - [ ] Update `review-architecture` skill -- [ ] Update `debug` skill +- [x] Update `debug` skill - [ ] Create `respond-to-incident` skill - [ ] Create `review-incident` skill - [ ] Create `plan-retirement` skill diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 7da24b3..aaeb58b 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -1,37 +1,51 @@ --- name: debug -description: Reproduces, isolates, repairs, and verifies code issues. Use when debugging failures, regressions, runtime errors, or incorrect behaviour. +description: Diagnoses defects and conditionally makes verified repairs. Use for failures, regressions, runtime errors, performance faults, incorrect behaviour, or causal debugging called by incident response. metadata: - invocation: model -disable-model-invocation: false + invocation: user +disable-model-invocation: true --- # Debug -**Scientific debugging** turns a repeatable failing signal into a verified causal repair through falsifiable hypotheses, predicted observations, and discriminating experiments. +**Scientific debugging** establishes an evidence-backed causal explanation through falsifiable hypotheses, predicted observations, and discriminating experiments. It stops at diagnosis when requested and permits a repair only when repair is part of the agreed outcome. ## Process -### 1. Establish a repeatable signal +### 1. Fix the outcome and signal -Read repository instructions and establish the authorised scope, expected behaviour, observed behaviour, and exact verdict that distinguishes them. Run the smallest reliable reproduction and record its invocation or probe, input, environment, expected verdict, and observed verdict. A failing test, benchmark, trace, log pattern, captured artefact, or targeted external probe may supply the signal when a local reproduction cannot. Use [Debugging Techniques](references/TECHNIQUES.md) when the signal or next experiment is not obvious; for an intermittent or concurrent fault, component or production boundary, or performance regression, load [Nondeterministic Faults](references/NONDETERMINISTIC.md), [Boundary Evidence](references/BOUNDARY-EVIDENCE.md), or [Performance Faults](references/PERFORMANCE.md) respectively. Preserve pre-existing user work and identify the mutation boundary for later repair attempts. When no reliable signal can be established, leave the implementation unchanged and report the evidence, blocker, and next discriminating experiment. The original fault has a repeatable red signal or an explicit evidence boundary. +Read repository instructions, the requested outcome, expected and observed behaviour, environment evidence, current implementation and tests, and applicable product, feature, architecture, and operational authorities. Resolve whether the outcome is diagnosis only or diagnosis plus repair; ask when that distinction is materially ambiguous. Record the authorised local, external, and production boundaries before probing or editing. -### 2. Isolate the root cause +Define the smallest reliable signal whose verdict distinguishes expected from observed behaviour. Use a focused test, runtime error, trace or log, captured artefact, controlled benchmark, safe external observation, or production-boundary evidence according to the fault. Record its invocation or probe, input, environment, expected verdict, observed verdict, and fidelity limitations. Do not claim a local reproduction when only external or captured evidence exists. Run an external or production probe only when it is read-only and within available authority; obtain explicit authority before any probe that can mutate durable or external state, expose sensitive data, or materially affect service. Preserve the original unminimised signal for the final causal check. -Gather evidence and state falsifiable causal hypotheses in evidence-supported order. For the leading hypothesis, state the observation it predicts, then run the cheapest experiment that distinguishes it from the credible alternatives. Change one variable and record the hypothesis, prediction, experiment, and observation. Treat an observation that does not discriminate as an incomplete experiment and sharpen it before continuing. Repeat until the evidence supports one leading cause strongly enough to justify a minimal repair experiment. The cause, causal mechanism, and evidence against symptom-level alternatives are explicit. +When the signal or next experiment is unclear, load [Debugging Techniques](references/TECHNIQUES.md). For an intermittent, flaky, timing-sensitive, order-dependent, or concurrent fault, load [Nondeterministic Faults](references/NONDETERMINISTIC.md). For a component, process, service, environment, or production boundary, load [Boundary Evidence](references/BOUNDARY-EVIDENCE.md). For a latency, throughput, resource, query, or scale regression, load [Performance Faults](references/PERFORMANCE.md); also load the nondeterministic guide when its classifier is noisy. If no reliable signal can be established, leave implementation unchanged and stop with the exact evidence gap and next smallest discriminating experiment. The outcome, authority, and evidence boundary are explicit. -### 3. Test one repair +### 2. Establish the causal explanation -When the user requested diagnosis only, stop before mutation and follow the Handoff. Invoke `$modular-design` when the supported repair presents a material structural-maintainability decision, and carry its constraints into `$tdd` when a runnable suite can exercise the repair through a stable observable boundary; preserve the established reproduction as TDD's Red signal and let it own the minimal Green repair and regression protection. When TDD is not applicable, apply one smallest change that would repair the leading cause if the hypothesis is correct. Keep the attempt within the authorised mutation boundary and leave unrelated cleanup or refactoring outside it. One evidence-led repair is ready for a causal verdict. +State the plausible causal hypotheses in evidence-supported order. For the leading hypothesis, predict an observation that distinguishes it from credible alternatives, then run the smallest safe experiment capable of producing that observation. Change one variable at a time and record the hypothesis, prediction, experiment, observation, and whether the result supports, rejects, or fails to discriminate the hypothesis. Treat correlation, disappearance of a symptom, and a passing retry as evidence to explain rather than proof of cause. -### 4. Accept or revert the repair +Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge conclusion must survive the conversation and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-context` only after independently confirmed terminology arises or an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. -Run the original signal. When it changes as predicted, retain the repair provisionally and continue to verification. When it remains red or changes for a different reason, record the contradictory evidence and revert only the production, configuration, and throwaway-test changes introduced by that attempt; preserve the established reproduction and all pre-existing user work. Reconsider the hypotheses, experiment, system boundary, or architecture whenever the observations no longer support the causal model, then return to isolation without stacking another repair onto the failed one. The working tree contains either one supported repair or no residue from an unsuccessful attempt. +### 3. Test one causal repair -### 5. Verify and clean up +For a repair outcome, select the smallest change that addresses the supported cause. Invoke `$modular-design` only when the repair presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, and carry its constraints into the repair. Exclude unrelated cleanup, broad refactoring, speculative hardening, and other plausible defects. -Re-run the original unminimised reproduction, focused regression coverage, relevant nearby checks, and the repository-prescribed wider checks. Remove temporary instrumentation and throwaway harnesses, or retain them deliberately as documented diagnostics. Separate unrelated pre-existing failures from repair regressions. The original fault and causal account are confirmed by the evidence, the smallest supported repair remains, and relevant checks pass; otherwise the exact remaining failure and uncertainty are explicit. +Invoke `$tdd` only when an existing runnable suite can exercise the desired observable behaviour through a stable boundary with an independent oracle. An established reproduction may supply TDD's Red only when it independently specifies that desired behaviour and fails for the expected behavioural reason; a crash, log, trace, performance symptom, or implementation-derived expectation is not automatically a retained test or valid Red. Otherwise apply one conventional minimal repair and add or update regression protection only when existing project infrastructure can meaningfully express the changed behaviour. Do not install or invent a test or quality framework for this repair. + +Run the original signal against the attempt. Retain the repair provisionally only when the signal changes as the causal model predicted. When it does not, record the contradiction and revert only changes introduced and owned by that attempt when their exact ownership is known, preserving the original signal, diagnostic evidence, and all pre-existing user work. Do not use destructive reset or checkout, discard uncertain state, or stack another unsupported fix. Remove owned temporary instrumentation and throwaway harnesses before freezing a supported candidate; retain a diagnostic only when the requested scope and project authority make it durable. Return to the causal hypotheses with either one supported candidate or no repair residue. + +### 4. Review the fixed candidate + +Freeze the complete supported candidate: exact diff or revisions, changed and untracked paths, repository state, intended-behaviour authority, causal account, original signal, applicable structural constraints, and named risks. Invoke `$code-review` independently on that basis. If the candidate drifts or an applicable review axis is stale or unperformed, re-freeze and repeat the review when scope and ownership remain clear; otherwise stop with the exact blocker. + +Adjudicate every finding against the requested outcome, project authorities, causal evidence, and current code. Correct each supported required finding through the applicable TDD or conventional repair path; reject a finding only with concrete contrary evidence. When a finding requires a user-authoritative change to behaviour, contract, architecture, or scope, ask one material question and pause mutation until the answer is confirmed. After any candidate change, re-run the original signal, freeze the new candidate, and repeat independent review. Continue only when the complete current candidate has no unresolved required finding. + +### 5. Verify the reviewed repair + +Invoke `$verify-change` on the exact reviewed candidate. Supply the intended behaviour, causal claim, original unminimised signal, focused regression evidence, applicable nearby and repository requirements, named risks, review dispositions, and any changed tests, snapshots, baselines, suppressions, scripts, configuration, dependencies, or generated artefacts. A passing regression alone does not prove the root cause; require the original signal to change as predicted and explain the causal mechanism and evidence against material alternatives. + +When verification exposes an in-scope failure, return to the causal repair path, then independently re-review and freshly verify the new fixed candidate. When source or measurement-path drift makes evidence stale, establish ownership, re-freeze, re-review, and re-verify rather than retargeting old evidence. Preserve a required pre-existing or out-of-scope failure as a blocker instead of widening the repair. Claim a completed repair only when the current review has no unresolved required finding and verification returns `passed within stated scope`; treat `partially verified`, `unverified`, or `failed` as incomplete or blocked and report it honestly. ## Handoff -Report the expected and observed behaviour, original failing signal, hypotheses, predictions, experiments and observations, root cause and causal mechanism, reverted attempts, retained change, regression protection, verification commands and results, and any blocker or unresolved uncertainty. For diagnosis-only work, state plainly that no implementation was changed. +Report the requested outcome; expected and observed behaviour; authority and environment boundary; original signal; hypotheses, predictions, experiments, and contradictory evidence; root cause and causal mechanism or exact unresolved evidence boundary; reverted attempts; retained change and regression protection; review scope and finding dispositions; verification scope, commands, evidence, verdict, limitations, residual risks, and next discriminating experiment or planned work. For diagnosis-only work, state that no implementation was changed. Stop after the diagnosis or verified local repair without committing, pushing, opening a pull request, changing tracker state, owning incident mitigation or communications, performing an incident review, releasing, deploying, monitoring, or adding speculative documentation. diff --git a/skills/debug/agents/openai.yaml b/skills/debug/agents/openai.yaml index bac6c82..445ebaf 100644 --- a/skills/debug/agents/openai.yaml +++ b/skills/debug/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Debug' - short_description: 'Debug issues through verified repair' + short_description: 'Find root causes and verify repairs' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false diff --git a/skills/debug/references/TECHNIQUES.md b/skills/debug/references/TECHNIQUES.md index 9cd3d5c..60d0ad8 100644 --- a/skills/debug/references/TECHNIQUES.md +++ b/skills/debug/references/TECHNIQUES.md @@ -19,7 +19,7 @@ A passing retry does not turn an intermittent failure green. Use [Nondeterminist | Situation | Technique | Observable verdict and retained artefact | | --- | --- | --- | -| A stable test seam reaches the fault | Focused failing test or minimal harness | Assert the exact behaviour and preserve the smallest fixture. If retained as regression coverage, `$tdd` remains authoritative. | +| A stable test seam reaches the fault | Focused failing test or minimal harness | Assert the exact behaviour and preserve the smallest fixture. It is a candidate TDD Red only when it independently specifies the desired behaviour and satisfies every TDD prerequisite. | | The fault is at an HTTP boundary | HTTP request script | Assert the relevant status, body, and headers; retain a redacted request and response rather than relying only on process exit. | | The fault is a CLI contract | CLI invocation with fixture input | Assert exit status, stdout, and stderr as applicable; record flags, working directory, and relevant environment. | | Only a production request or event exposes the fault | Capture and replay | Use [Boundary Evidence](BOUNDARY-EVIDENCE.md) to capture the smallest authorised event and assess replay fidelity. | @@ -46,7 +46,7 @@ An experiment is complete when its observation confirms or rejects a stated hypo ## Preserve Useful Evidence -- Promote a minimised reproducer to regression protection at the strongest stable seam when appropriate; keep `$tdd` authoritative for the test and repair cycle. +- Promote a minimised reproducer to regression protection at the strongest stable seam only when it independently specifies desired observable behaviour. Invoke `$tdd` only when its complete prerequisite holds. - Remove temporary instrumentation and throwaway harnesses after use, or retain them deliberately as documented diagnostics. Tag temporary probes so cleanup is mechanically checkable. - Store only sanitised captures and fixtures that the repository is authorised to retain. Report external artefacts without copying sensitive data into the workspace. - Preserve the original reproduction across repair attempts. When a repair fails its prediction, retain the observation but revert the attempt before testing the next hypothesis. From 21a6a4b0f85e3b45d78239f7c521eb0f54363df4 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 11:51:05 +0100 Subject: [PATCH 37/79] feat(review-architecture): simplify evidence-led assessments --- SKILLS_PLAN.md | 2 +- skills/review-architecture/SKILL.md | 68 +++++++++++++++---- skills/review-architecture/agents/openai.yaml | 2 +- .../references/architecture-analysis.md | 46 ------------- .../references/report-design.md | 52 -------------- 5 files changed, 58 insertions(+), 112 deletions(-) delete mode 100644 skills/review-architecture/references/architecture-analysis.md delete mode 100644 skills/review-architecture/references/report-design.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index f468aab..498fef3 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -412,7 +412,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `triage-work` skill - [x] Create `refine-backlog` skill - [x] Create `plan-iteration` skill -- [ ] Update `review-architecture` skill +- [x] Update `review-architecture` skill - [x] Update `debug` skill - [ ] Create `respond-to-incident` skill - [ ] Create `review-incident` skill diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index 910f9bb..1f25c4b 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -1,6 +1,6 @@ --- name: review-architecture -description: Reviews a codebase or scope for high-value modular redesigns, persists a Markdown review, and opens a disposable visual HTML shortlist. Use when modular architecture needs assessment. +description: Assesses an explicitly bounded codebase area for a few high-value architecture improvements and reports the evidence without changing implementation. Use for architecture or maintainability reviews. metadata: invocation: user disable-model-invocation: true @@ -8,26 +8,70 @@ disable-model-invocation: true # Review Architecture -The **visual information-seeking mantra** turns an evidence-backed modular review into an impact-grouped overview for people and a durable implementation reference for agents. +Produces one canonical Markdown assessment and a user-friendly disposable HTML explanation of the highest-value structural improvements in an explicitly bounded codebase area. It preserves the reviewed implementation and stops before detailed redesign, tickets, or delivery work. ## Process -### 1. Establish the review +### 1. Bound and freeze the review -Use the user's explicit scope or the whole repository with the slug `full-codebase`. Inspect applicable project context, architecture decisions, source, tests, contracts, schemas, dependencies, runtime configuration, and documentation. Exclude generated output, vendored dependencies, caches, and binaries unless they participate in a material boundary. Invoke `$modular-design` as the architecture authority. The scope, exclusions, project constraints, and modular baseline are explicit. +Inspect the request, repository guidance, available tools, and current repository state. Establish the explicit codebase scope, exclusions, review purpose, applicable architecture or maintainability qualities, and any user-held priority that could change qualification or rank. Do not silently expand an ambiguous scope to the whole repository. Ask only for missing scope or priority that cannot be discovered and would materially change the result. -### 2. Select the redesigns +Capture the reviewed revision and starting worktree state. Read root `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and product constraints when present, then inspect only the scoped implementation, tests, contracts, schemas, configuration, history, and operational evidence needed to assess the area. Treat missing or conflicting authorities as limitations unless they prevent a responsible comparison, in which case expose the exact gap. Separate intended constraints, observed implementation, inferences, and unknowns. The review boundary, authorities, evidence baseline, and limitations are explicit before findings are formed. -Read [Architecture Analysis](references/architecture-analysis.md). Recover confirmed and inferred quality drivers, map the current capabilities and contracts, and trace material architecture pressure to precise repository locations. Apply **design it twice** to every serious candidate and load `$modular-design` techniques only when their conditions fit. Invoke `$research` only when a recommendation materially depends on an external claim requiring durable verification. Retain every redesign that clears the evidence threshold, assign stable two-digit IDs in ranked order, and group the set by explained `high`, `medium`, or `low` impact. Zero recommendations is valid. +### 2. Map structural evidence -### 3. Persist the review +Map capabilities, owned knowledge, contracts, consumers, dependencies, runtime boundaries, and verification seams instead of equating directories, classes, packages, services, or deployment units with modules. Invoke `$modular-design` with the fixed scope, authorities, and current evidence. Use its findings as passive structural authority for ownership, information hiding, cohesion, coupling, contracts, dependency direction, seams, and change locality; retain responsibility for investigation, candidate comparison, prioritisation, both reports, and stopping. -Write the canonical report to `docs/architecture/YYYYMMDD-{scope}-architecture-review.md`; preserve an existing path with `-2`, `-3`, and so on unless replacement is explicit. For each recommendation, make the issue, fix, benefit, affected architecture, current and target design, evidence, rejected alternative, costs, risks, dependencies, migration route, smallest useful slice, containment, and fitness checks independently understandable. Record coverage without a qualifying redesign and material limitations. Stop before implementation or a file-by-file plan. +Trace material pressure to precise current repository locations. Look for rules represented in multiple owners, unstable or leaking contracts, dependencies pointing against established responsibility, changes that repeatedly propagate across boundaries, and regression or operational evidence tied to structure. Use coherent repository history or metrics only to locate and challenge candidates, discounting bulk, generated, renamed, bot-authored, or obsolete changes; neither a metric nor a smell label proves a recommendation. Express a material quality or change pressure as a compact project-evidenced scenario when that makes its stimulus, affected capability, expected response, or observable consequence clearer. The evidence map distinguishes observations, supported inferences, and uncertainty. -### 4. Present the visual shortlist +Invoke `$research` only when a material external claim must survive the session. Read and verify its cited report before using the conclusion, and keep that report explicitly within the review scope. Do not import a general architecture method, standard, or universal quality model when project evidence does not make it authoritative. -Read [Report Design](references/report-design.md), then create a disposable `architecture-review-{timestamp}.html` in the operating system's temporary directory and open it for the user. Give it the same recommendation IDs, ordering, conclusions, and technical substance as the Markdown report. Show the complete set as `High impact`, `Medium impact`, and `Low impact`; make each collapsed card understandable through an aligned before-and-after visual and plain-language `Issue`, `Fix`, and `Benefit`, with technical depth available on demand. Add filters only when they materially improve navigation. +### 3. Compare and qualify candidates -### 5. Verify and hand off +For every serious candidate, apply **Design It Twice**: compare the current design with at least one materially different responsibility, boundary, contract, or dependency arrangement. Consider the status quo and another credible alternative when either could reasonably win. Do not count naming, file placement, interface syntax, or a straw design as an alternative, and do not elaborate any option beyond what the review decision requires. -Verify the Markdown paths, links, structure, and implementation sufficiency, then compare both artifacts for matching IDs and claims. Inspect the HTML at desktop and narrow widths, exercising disclosure, applicable filters, pointer and keyboard operation, focus, overflow, external dependencies, and print output. Correct material defects; when browser inspection is unavailable, mark visual acceptance incomplete. Return both report paths, scope and exclusions, recommendation IDs and count, invoked research, validation performed, and unresolved limitations. Preserve the reviewed implementation unchanged. +Retain a recommendation only when it has all of: + +- a material benefit to an explicit constraint or quality; +- exact current repository evidence and a credible causal explanation; +- a clear current and target responsibility, contract, or dependency difference; +- a bounded transition seam, smallest useful step, and containment or recovery route; +- observable verification or fitness evidence for the claimed benefit; and +- costs, trade-offs, risks, alternatives, and material uncertainty visible enough to support a decision. + +Reject local cosmetic cleanup, a smell without consequence, speculative abstraction, and a redesign whose expected benefit or transition cannot be evidenced. Zero recommendations is a valid successful result. + +Select only the small decision-useful shortlist with the greatest consequence or quality priority, evidence reach, confidence, expected benefit, and migration feasibility. Rank it qualitatively with an explained ordering; do not calculate composite scores or confidence percentages. Record reviewed areas and lower-value candidates that were not retained only to the degree needed to make coverage honest. The shortlist contains the highest-value supported recommendations, not an inventory of conceivable redesigns. + +### 4. Resolve the report outcome + +Present the material scope, limitations, concise outcome, and proposed ordered shortlist before persistence when a user-authoritative trade-off could still change them. Compare the live choices and let the user decide; do not convert an unavailable preference into an architectural conclusion. Once shared understanding is reached, the invocation authorises both report artifacts without a second application prompt. + +### 5. Write the canonical review + +Follow the project's existing architecture-review convention. Otherwise write `docs/architecture/YYYYMMDD-{scope-slug}-architecture-review.md`, preserving an unrelated existing file with a numeric suffix unless replacement was explicit. Use **plain language** for the outcome, finding titles, and short `Issue`, `Change`, and `Benefit` summaries while retaining precise project and architecture terms. + +Make the report independently useful with: + +1. scope, exclusions, reviewed revision, authorities, qualities, and concise outcome; +2. an ordered finding index, or an explicit zero-finding result; +3. for each numbered finding, precise evidence; current and target responsibilities and contracts; benefit; trade-offs, cost, risk, and uncertainty; the rejected alternative or status quo; transition seam, smallest useful step, coexistence or containment; and observable verification or fitness evidence; +4. reviewed coverage, candidates not retained when material, limitations, stale-evidence risk, and any invoked research. + +Use a small repository-native diagram only when it materially clarifies a complex ownership, dependency, runtime, or transition relationship. Keep the report at architecture-decision depth: it can support a later decision or body of work, but it is not a file-by-file implementation plan. + +### 6. Present the explanatory view + +Create a self-contained disposable HTML file in the operating system's temporary directory and open it for the user with an available browser tool. Treat Markdown as canonical and HTML as a linear explanatory view, not a second independently maintained report. + +Show the outcome and linked finding index first. Give each numbered finding one section whose always-visible content includes the same title and core `Issue`, `Change`, and `Benefit` claims as Markdown. Add a paired current-and-target visual only when it clarifies the material relationship; stack it without losing correspondence at narrow widths. Put ancillary evidence, alternatives, transition detail, and limitations in native `details` disclosures only when hiding them improves scanning; keep information required to understand or decide the recommendation visible. Use no filters, tabs, nested disclosure, dashboard cards, or JavaScript-dependent controls for the small shortlist. + +Keep the page readable as one column, semantic and keyboard-operable, usable without colour, and printable with all content available. Make printing open every native disclosure through a self-contained `beforeprint` behaviour or an equivalent that works while the on-screen disclosures start closed; do not assume print CSS can expose a closed `details` subtree. Escape repository-derived content for HTML, avoid unverified external dependencies, and include the canonical Markdown path, reviewed revision, scope, and limitations. A zero-finding view states the result directly and shows coverage and limitations without empty recommendation furniture. + +### 7. Verify and stop + +Re-read both artifacts against the frozen evidence. Verify every repository path and material claim, the Markdown structure and links, the minimal matching finding numbers, order, titles, and core claims across artifacts, and the separation between observation, inference, and uncertainty. Recheck the source revision and implementation paths; when material source drift occurred, reassess affected findings or mark the evidence stale rather than presenting it as current. + +Inspect the actual HTML in an available browser at ordinary and narrow widths. Exercise native disclosures by pointer and keyboard, check focus, reflow, overflow, diagrams, colour independence, and a print render with every disclosure open, and correct material defects. When browser inspection or printing is unavailable, perform structural and source checks and report the exact visual limitation. + +Confirm the reviewed implementation and its existing tests, configuration, architecture, context, product, and feature documents remain unchanged apart from concurrent pre-existing work. Return both report paths, scope and exclusions, finding count and order, invoked research, verification performed, limitations, and any stale evidence. Stop without refactoring, implementing, editing project authorities, creating tickets or specifications, scheduling work, committing, publishing, releasing, or deploying. diff --git a/skills/review-architecture/agents/openai.yaml b/skills/review-architecture/agents/openai.yaml index 9ea18e8..2ead7b8 100644 --- a/skills/review-architecture/agents/openai.yaml +++ b/skills/review-architecture/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Review Architecture' - short_description: 'Review architecture and explain redesigns' + short_description: 'Find high-value architecture improvements' policy: allow_implicit_invocation: false diff --git a/skills/review-architecture/references/architecture-analysis.md b/skills/review-architecture/references/architecture-analysis.md deleted file mode 100644 index 2b02717..0000000 --- a/skills/review-architecture/references/architecture-analysis.md +++ /dev/null @@ -1,46 +0,0 @@ -# Architecture Analysis - -Load this reference while selecting and ranking redesigns. `$modular-design` owns knowledge and boundary decisions, contracts, deep modules, locality, realization, and conditional techniques; this reference owns review-specific evidence and judgment. - -## Quality Priority - -Apply the first evidenced priority in this order: - -1. Safety, correctness, security, and data integrity constraints. -2. Explicit project quality drivers. -3. Qualities defined by `$modular-design`. -4. Testability and migration safety. -5. Operability, reliability, performance, scalability, and portability when material to the system. - -Recover drivers from product and domain context, architecture decisions, public promises, tests, operational configuration, incidents, recurring changes, and repository history when available. Express a material driver as a scenario with a stimulus, affected capability, expected response, and observable measure. Label inference and confidence; ask the user only when an unavailable priority would materially change qualification or rank. - -## Evidence and Comparison - -Map capabilities rather than assuming directories, classes, packages, services, or deployment units are architecture modules. For each material pressure, trace the owned knowledge, public contract, consumers, dependencies, runtime boundaries, verification seams, and repeated change propagation to repository-relative paths and precise locations. Separate observations, supported conclusions, and uncertainty. Metrics may locate candidates but cannot prove a redesign. - -For every serious candidate compare at least two materially different capability boundaries. State each contract, hidden knowledge, dependency direction, quality effects, framework and runtime fit, migration seam, first useful slice, containment, costs, and risks. A naming, file-placement, or interface-syntax variation is not a second design. Prefer the alternative that hides more consequential knowledge behind the simpler stable contract while satisfying the higher-priority evidence. - -## Qualification and Rank - -A recommendation qualifies only when it: - -- materially improves a priority quality; -- traces its problem and expected improvement to repository evidence; -- hides or realigns architectural knowledge rather than rearranging code locally; -- has a credible incremental route with visible dependencies and risks; and -- is supported strongly enough to recommend with material uncertainty exposed. - -Retain every qualifying redesign and none below the threshold. Record non-qualifying areas only as review coverage. Assign labels with a one-sentence evidence rationale: - -- **Impact:** `high` changes a constraint or explicit driver, or removes repeated high-reach pressure; `medium` materially improves a bounded capability; `low` produces a worthwhile but contained architectural gain. -- **Effort:** `high` crosses several boundaries or requires staged data, contract, or deployment work; `medium` needs coordinated changes; `low` is contained behind an existing seam. -- **Risk:** `high` threatens behaviour, data, security, public contracts, or runtime continuity; `medium` needs managed integration; `low` is isolated and readily reversible. -- **Confidence:** `high` follows direct repeated evidence and executable verification; `medium` combines credible evidence with limited inference; `low` depends materially on missing context. - -Do not calculate a composite score. Order impact groups `high`, `medium`, then `low`; rank within a group by quality priority, evidence reach, confidence, and migration feasibility. Assign `01`, `02`, and onward after ranking, and preserve those IDs across both reports regardless of filtering. - -## Migration and Fitness - -Define stages that keep the system operable and verifiable: prerequisites, smallest independently useful slice, old/new coexistence, data or contract transition, containment or rollback, and superseded-path removal. Replace a high-risk boundary incrementally; replace a safe local boundary atomically. Stop before a file-by-file implementation plan. - -Pair each claimed benefit with observable fitness evidence. State the signal, expected result, and where it should run, using existing contract tests, dependency checks, change-impact checks, adapter conformance, runtime thresholds, telemetry, or deployment signals when they prove the quality. Propose new machinery only when existing verification cannot. diff --git a/skills/review-architecture/references/report-design.md b/skills/review-architecture/references/report-design.md deleted file mode 100644 index dc07cd7..0000000 --- a/skills/review-architecture/references/report-design.md +++ /dev/null @@ -1,52 +0,0 @@ -# Architecture Review Report Design - -Load this reference after the recommendation set is complete. The Markdown report is the durable implementation record; the disposable HTML is its visual decision view. They adapt presentation to their readers without changing recommendation IDs, order, claims, or technical substance. - -## Markdown Record - -Use this reading order: - -1. Title, date, scope, exclusions, one-sentence outcome, and recommendation index. -2. `High impact`, `Medium impact`, and `Low impact` recommendation groups, omitting empty groups. -3. Review coverage and evidence limitations. -4. Method, invoked research reports, and validation status. - -Name each recommendation `{ID}. {action-led title}`. Lead with three plain-language fields: - -- **Issue:** the present architecture friction and consequence. -- **Fix:** the ownership, contract, or dependency change. -- **Benefit:** the concrete quality improvement. - -Then preserve the affected modules and contracts, current and target design, repository evidence with precise locations, before-and-after explanation, rejected alternative, framework and runtime fit, costs, risks, uncertainty, dependencies, migration and coexistence stages, containment or rollback, smallest useful slice, and fitness checks. A downstream agent must be able to receive the Markdown path plus an ID and understand the bounded change without reopening the review. - -## HTML Shortlist - -Apply overview first and details on demand. The initial viewport identifies the review and exposes the impact-grouped recommendation set without introductory prose. Each card shows its ID, title, impact, before-and-after visual, `Issue`, `Fix`, and `Benefit`. Put evidence, alternatives, effort, risk, confidence, migration, and fitness checks in native disclosure. Show a direct zero-result state when no redesign qualifies. - -Use filters only when the set is large enough that impact groups alone do not support comparison. Filtering changes visibility, never IDs, ranking, or report content; show the visible count and a clear reset. - -## Visual Language - -Apply **visual juxtaposition**: give every recommendation an aligned current/target pair that answers one question about changed ownership, hidden knowledge, dependency direction, runtime flow, or migration. Keep corresponding concepts in corresponding positions with consistent names, shapes, direction, and scale. At narrow widths, stack the pair while preserving that visual grammar. Simplify the visual rather than shrinking an unreadable whole-system map. - -Choose the smallest useful form: a boundary or dependency graph, quality-scenario flow, ownership sketch, cross-section, or staged migration. Mermaid, inline SVG, and semantic HTML/CSS are all valid. Mix techniques when the evidence benefits; avoid ornamental diagrams. Give every visual an accessible name and adjacent textual explanation. - -Use strong hierarchy, generous spacing, readable line lengths, restrained colour, and consistent cards. Paths and contracts may use monospace. Colour reinforces words and shapes rather than carrying meaning. Prefer semantic controls, visible focus, and native disclosure; interactions remain keyboard-operable. Provide responsive and print styles that preserve all content and expand technical details for printing. - -Scripts, styles, fonts, and diagram libraries may load from verified CDNs when useful. Keep the generated artifact in one HTML file apart from those dependencies, escape repository-derived text for its destination context, and verify every selected dependency at generation time. - -## Plain Language - -Use **plain language** for `Issue`, `Fix`, and `Benefit`: state the point first, use active verbs and familiar concrete words, and keep one idea in each field. Preserve established domain, project, framework, and architecture terms when they are more precise. Move substantiation into disclosure instead of weakening or repeating the summary. A collapsed card succeeds when it is concise and distinguishable from every other recommendation. - -## Visual QA - -Inspect the actual temporary file with network access: - -1. Confirm external dependencies and every before-and-after visual render. -2. Compare recommendation IDs, order, labels, and claims with the Markdown record. -3. Exercise disclosure and applicable filters by pointer and keyboard; verify focus, state, count, and reset. -4. Inspect desktop, narrow, zero-result, and print states for hierarchy, wrapping, overflow, clipping, and content loss. -5. Confirm the report remains understandable without colour and that evidence paths remain legible. - -Correct material defects and rerun affected checks. When browser inspection is unavailable, complete structural and source checks, then mark visual acceptance incomplete in both artifacts and the handoff. From 09448cbed2147b5b8a41bad52f872d31e19449ed Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:04:55 +0100 Subject: [PATCH 38/79] feat(respond-to-incident): add mitigation-first command --- SKILLS_PLAN.md | 2 +- skills/respond-to-incident/SKILL.md | 53 +++++++++++++++++++ skills/respond-to-incident/agents/openai.yaml | 6 +++ 3 files changed, 60 insertions(+), 1 deletion(-) create mode 100644 skills/respond-to-incident/SKILL.md create mode 100644 skills/respond-to-incident/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 498fef3..3e2e5e0 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -414,7 +414,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `plan-iteration` skill - [x] Update `review-architecture` skill - [x] Update `debug` skill -- [ ] Create `respond-to-incident` skill +- [x] Create `respond-to-incident` skill - [ ] Create `review-incident` skill - [ ] Create `plan-retirement` skill - [ ] Delete `elicit-with-context` skill diff --git a/skills/respond-to-incident/SKILL.md b/skills/respond-to-incident/SKILL.md new file mode 100644 index 0000000..e71a598 --- /dev/null +++ b/skills/respond-to-incident/SKILL.md @@ -0,0 +1,53 @@ +--- +name: respond-to-incident +description: Stabilises active production incidents through coordinated, evidence-led mitigation and verified service recovery. Use when current production impact requires an emergency response. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Respond to Incident + +An **adapted Incident Command System** scales clear command, operations, and communications ownership to the active incident so responders can reduce user impact safely and establish the resulting service state. + +## Process + +### 1. Establish the incident basis + +Inspect the request, repository guidance, current incident and communication records, operational authorities, runbooks, available tools and controls, relevant product and architecture constraints, and current production evidence. Resolve the incident identity; source-reported time; current impact and affected users, tenants, regions, systems, data, and dependencies; known recent changes; safety, privacy, security, compliance, and support constraints; current responders and owners; evidence sources and freshness; and accessible mutation boundary. Use the project's severity meaning only when it changes the response; do not invent a severity scheme or guess current production state. + +Define the signals that can distinguish continued impact, partial recovery, and credible stability before taking action. Prefer the owning incident, observability, and operational systems over copied state. When evidence is stale, contradictory, or unavailable, seek the smallest safe current observation; if no trustworthy signal or safe observation exists, keep production unchanged and establish the exact evidence or access blocker and next owner. The active impact, response authority, and evidence boundary are explicit. + +### 2. Establish proportionate command + +Map the project's current incident roles onto three responsibilities: incident command owns impact, priorities, mutation decisions, coordination, escalation, and the stopping decision; operations owns bounded technical investigation and controlled action; communications owns factual updates and the response timeline. One responder may retain every responsibility; delegate only when incident complexity or concurrent work requires it, and preserve existing project-native command and handoff authority. + +Adopt the owning incident record and established communication channel where available. Keep current impact, status, mitigation, command and action owners, next communication time, blockers, and unresolved risks current there. Maintain one concise timeline throughout: timestamp, observation or decision, action and actor, expected result, observed result, and reversal when applicable. When the owning record is inaccessible, retain a bounded session timeline and identify the exact record-update handoff rather than creating a competing repository record. The response has one visible command owner and one current factual account. + +### 3. Select the safest useful mitigation + +Prioritise reducing user impact over broad root-cause work. Start with applicable runbook or automated recovery actions and compare credible mitigation options by expected impact reduction, time, reversibility, blast radius, operational and data risk, dependencies, and signals capable of verifying direct effect and user or service health. Prefer the reversible, low-blast-radius option that is likely to stabilise service while preserving evidence. A successful command, quiet alert, or plausible diagnosis is not recovery evidence. + +Ask the incident commander or user to resolve a material trade-off when an option changes intended product behaviour, risks data loss or integrity, affects security or compliance, has a material irreversible effect, exceeds accessible authority, or requires a genuinely command-owned priority. Present the live evidence, viable options, consequences, exact target, intended effect, and available rollback or containment. Once the mitigation is agreed, the invocation supplies authority for its intrinsic mutations without a redundant application prompt; still satisfy required external confirmations and obtain any distinct operational authority the environment requires. One authorised mitigation and its expected evidence are explicit. + +### 4. Execute and appraise one controlled action + +Immediately before action, re-read the target and relevant service state, resolve its exact identity and scope, and record the expected result plus safe rollback or containment. Use least privilege and the narrowest native operational control. Preserve logs, traces, data, and unrelated user work; avoid destructive, broad, recursively targeted, glob-selected, or evidence-erasing operations. For a security or data-integrity incident, restrict exposure and preserve forensic evidence; defer destructive eradication, rebuild, or data repair until its targets, consequences, authority, and recovery path are exact. + +Record the action and actor, execute only the agreed operation, read back the mutated target, and immediately inspect both its direct effect and current user or service signals, including a credible counter-signal. When the action fails or worsens impact, perform only a pre-agreed safe rollback or containment, verify that result, preserve the contradictory evidence, and return to mitigation selection. Do not stack unsupported actions or interpret command success as service recovery. Each attempt leaves a timestamped, read-back result and either one supported current state or an explicit uncertainty. + +Invoke `$debug` only when a bounded causal diagnosis or local repair is necessary to stabilise this incident. Supply the exact signal, production boundary, time constraint, and stopping condition; retain ownership of impact, mitigation, production actions, communications, and recovery. Do not let diagnosis become open-ended or delay a safe mitigation. Invoke `$verify-change` for a frozen local code or configuration repair, or a fixed operational change that existing harnesses can appraise; its verdict supplements but never replaces current production recovery evidence. Invoke `$research` only for one material external question whose evidence must survive the session. + +Do not silently commit, push, publish, release, or deploy a repair. Use a project-native route only when the requested incident outcome and current operational authority explicitly include it. When stabilisation requires publication or deployment but no authorised project-native mechanism is available, stop that route with the exact handoff, owner, artefact, and service risk rather than improvising a Propulsion release process. + +### 5. Establish the service state + +After an effective action, observe the direct effect, user and service health, and counter-signals across the evidence window warranted by the system, runbook, traffic pattern, and incident risk. Do not invent an arbitrary duration. Check for partial or regional recovery, tenant or cohort skew, retries and hidden backlog, delayed processing, data damage, dependency degradation, recurrence, and changed alert or metric coverage when applicable. Distinguish mitigated impact, restored service, and remaining unknowns; silence alone does not establish stability. + +Keep the incident record and communications current as evidence changes. State observed facts, affected scope, action underway, user guidance, and next update without speculative root cause, blame, fabricated metrics, or unsupported recovery times. Service is stable only when trustworthy current signals and counter-signals support that conclusion within the stated scope and limitations. + +### 6. Conclude the active response + +Conclude with exactly one terminal state: verified mitigation or recovery with bounded ongoing observation and ownership, or an exact external blocker, escalation, or handoff. Read back the owning incident record where accessible and report current impact, stability evidence and its window, timeline, actions and reversals, communications, unresolved data or service risk, remaining observation and owner, access or authority limitations, and follow-up needs. Do not call the incident resolved merely because symptoms quieted or a mitigation executed. + +Stop before the later learning session. Do not write a postmortem or root-cause narrative, create speculative follow-up tickets, perform broad cleanup, plan a release, continuously monitor beyond bounded recovery evidence, retire a capability, or invoke `$review-incident`. The active incident response ends with a credible service state or precise handoff; later `$review-incident` owns blameless learning and follow-up routing. diff --git a/skills/respond-to-incident/agents/openai.yaml b/skills/respond-to-incident/agents/openai.yaml new file mode 100644 index 0000000..1ce243b --- /dev/null +++ b/skills/respond-to-incident/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Respond to Incident' + short_description: 'Stabilise an active production incident safely' + default_prompt: 'Use $respond-to-incident to coordinate a safe mitigation and verify the resulting production service state.' +policy: + allow_implicit_invocation: false From fc54d77753aad550e94ae6e38c08e4a85f918f89 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:16:00 +0100 Subject: [PATCH 39/79] feat(review-incident): add blameless evidence-led learning --- SKILLS_PLAN.md | 2 +- skills/review-incident/SKILL.md | 75 +++++++++++++++++++++++ skills/review-incident/agents/openai.yaml | 6 ++ 3 files changed, 82 insertions(+), 1 deletion(-) create mode 100644 skills/review-incident/SKILL.md create mode 100644 skills/review-incident/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 3e2e5e0..c889e6c 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -415,7 +415,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `review-architecture` skill - [x] Update `debug` skill - [x] Create `respond-to-incident` skill -- [ ] Create `review-incident` skill +- [x] Create `review-incident` skill - [ ] Create `plan-retirement` skill - [ ] Delete `elicit-with-context` skill - [ ] Update the README diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md new file mode 100644 index 0000000..3f76795 --- /dev/null +++ b/skills/review-incident/SKILL.md @@ -0,0 +1,75 @@ +--- +name: review-incident +description: Reviews stabilised production incidents into durable, blameless explanations and owned follow-up work. Use when active incident response has ended. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Review Incident + +Explains one stabilised incident from evidence and records the smallest useful set of owned learning outcomes without reopening active response or implementing the resulting work. + +## Prerequisites + +Require one fixed incident identity, its owning incident record, and current evidence that production is stable. Re-read the latest service state and its evidence window rather than trusting a resolved label or an earlier handoff. When impact is still active, stability is uncertain, or a bounded observation remains under incident command, stop the review and route the incident to a later `$respond-to-incident` session with the exact current signal, owner, and missing evidence. Do not operate production, change communications, or perform learning analysis as a substitute for stabilisation. + +Require enough incident evidence to distinguish what happened from what is merely believed: the timeline and its revisions, impact, detections, actions and reversals, mitigations, communications, current state, unresolved data or service risks, and relevant technical, operational, product, architecture, and context authorities. When a material gap prevents a responsible explanation, identify the evidence, source, and owner needed and stop without manufacturing a narrative. A gap that can remain explicit without distorting the review is an unknown, not a reason to invent certainty. + +## Process + +### 1. Establish the review and evidence boundary + +Inspect the request, repository guidance, owning incident system and project review convention, complete response handoff, current product and architecture context, available tools, evidence revisions, participants, owners, and access limits. Preserve the incident system as authority for live incident state and the project convention as authority for the durable review. Use a repository record only when durable learning belongs there; it must not become a mirror of volatile operational or tracker state. + +Set the incident scope, affected period, systems and populations, review participants, evidence cut-off, and the sensitivity of logs, communications, customer or personnel data, security details, and forensic material. Retain secrets, credentials, personal data, exploit detail, and other restricted evidence only in an appropriate access-controlled owner; cite a stable evidence identity or a safe summary elsewhere. When no available owner can preserve a required sensitive conclusion safely, report that exact persistence blocker instead of exposing or erasing it. + +Record each material statement as an observed fact with its source and time, an interpretation supported by named evidence, or an unknown. Preserve conflicting sources and superseded timeline entries with their provenance until the conflict is resolved; a later edit is evidence, not permission to rewrite history silently. The review has one bounded, current, and inspectable evidence base. + +### 2. Reconstruct the incident + +Build a timestamped sequence from initial conditions through impact, detection, escalation, response, mitigation, recovery, and the verified stable state. Account for who or what observed each event, the information available at that moment, decisions and actions, expected and observed results, communications, reversals, and revisions. Distinguish detection delay, response delay, recovery delay, and incomplete visibility when the evidence supports them. Do not infer an event time from record order or collapse contradictory clocks into false precision. + +Apply **local rationality** to explain why decisions and adaptations made sense to participants with the goals, pressures, signals, tools, permissions, and uncertainty available then. Compare **work-as-done** with documented or assumed work only when the difference reveals a material operating condition; neither deviation nor human error is a sufficient causal conclusion. Counter hindsight by asking what evidence would have made a different action reasonable at the time. State a counterfactual only when evidence supports both its changed condition and plausible consequence. + +Explain the interacting technical, process, organisational, tooling, environmental, and dependency conditions that shaped the outcome. Examine intended and actual defenses; what worked, failed, was missing, or created surprise; detection and escalation; communication flow; mitigation effectiveness and side effects; latent conditions; goal conflicts; previous relevant signals; and credible recurrence paths. Treat a trigger as distinct from the conditions that allowed impact and recovery. Preserve multiple contributing conditions and uncertainty rather than forcing a single root cause, stopping at individual action, or making responsibility blameless by making it vague. + +Invoke `$research` only when one material external conclusion must survive the review, then read and verify its cited report before relying on it. Research does not replace incident evidence or turn a plausible industry explanation into a fact about this incident. The reconstructed account explains the incident proportionately without blame or false certainty. + +### 3. Derive proportionate learning outcomes + +Identify only follow-up outcomes that directly address an evidenced contributing condition, recurrence path, defense gap, response weakness, or material learning unknown. For each candidate, state the incident evidence and rationale, the desired outcome rather than an unsupported implementation prescription, one accountable owner, the appropriate later lifecycle route, and observable evidence that will show completion and, where meaningful, effectiveness. Distinguish reducing recurrence likelihood, reducing impact, improving detection or response, and resolving an unknown. + +Route a behaviour or product change to a later `$specify-feature` session; a technical solution or decomposition to later `$design-feature` or `$create-tickets` sessions once their authorities exist; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal design path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology or rare accepted rationale to `$maintain-context`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. + +Challenge candidates against recurrence value, impact reduction, evidence strength, ownership, observability, cost, and overlap with current work. Reuse or link an exact existing outcome, preserve related work as related, merge duplicates, and remove generic reminders, `be more careful` actions, speculative backlog dumping, and action inflation. Zero follow-ups is valid when the evidence supports no useful new outcome. A critical learning action without an owner remains an explicit completion blocker rather than becoming anonymous backlog work. + +When agreed follow-ups need tracker mutations, require the task-management tool named in the root `AGENTS.md`. If absent, ask the user which tool the project uses, invoke `$maintain-agents` with that confirmed preference, verify the root instruction, and resume. Inspect relevant open and closed native tracker state, schema, relationships, and read-back operations before proposing changes. When the configured tool or account is unavailable or not writable, preserve the review and report the exact tracker blocker; do not fall back to Markdown, another tracker, or a repository backlog mirror. + +The candidate set is small, evidence-traceable, owned, observable, duplicate-aware, and routed without prescribing or performing later work. + +### 4. Confirm the complete review + +Choose the owning incident system or established project convention for the durable review. When neither exists and a repository record is appropriate and safe, propose `docs/incidents/YYYYMMDD-{incident-slug}-review.md`; preserve an unrelated existing path with a numeric suffix unless replacement is explicit. Keep a repository review durable and concise: incident identity and status, impact, evidence boundary and limitations, timestamped sequence, contributing conditions, defenses and response learning, follow-up outcomes and routes, and unresolved unknowns. Link to the owning incident and tracker records rather than copying their changing state. + +Present one complete synthesis for confirmation: the factual account, supported interpretations, contradictions and unknowns, contributing conditions and recurrence paths, defenses and response learning, proposed durable owner and exact write, sensitive-evidence treatment, and every follow-up create, update, link, reuse, owner, route, and effectiveness measure. Keep all durable systems unchanged while material review conclusions or ownership remain open. + +Obtain explicit user confirmation of the complete synthesis and mutation plan. If the user rejects or revises it, reopen every affected timeline, explanation, condition, action, owner, route, sensitivity, and persistence decision and present the complete revised synthesis. Confirmation authorises exactly the agreed review and tracker writes without another application prompt. Independently confirmed durable project terminology may invoke `$maintain-context` inline; terminology or rationale established by this review waits until the complete synthesis is confirmed. Verify its returned change before continuing. + +The user and agent share one exact, blameless explanation and owned persistence plan. + +### 5. Persist and verify the agreed outcome + +Immediately before mutation, re-read the current incident state, material evidence revisions, target review record, matching tracker items, owners, and affected relationships. When drift makes the incident active or stability uncertain, leave the review and tracker unchanged and route back to `$respond-to-incident`. When drift materially changes the explanation, follow-up set, owner, route, or target, leave coupled mutations unchanged and return to complete-review confirmation. Continue an independent agreed write only when the drift cannot change its meaning. + +Write the confirmed review in its owning system or project location and apply exactly the confirmed native tracker creates, updates, links, owners, and routes that remain valid. Preserve unrelated record content, access controls, discussion, fields, and relationships. Use only native concepts whose meanings match the agreed plan; leave an unsupported part unapplied and report the capability limit rather than simulating it with a misleading field or document. + +Read the durable review, each affected tracker item, owner, link, and relationship back from its owning system. Verify the incident identity and stable status, evidence labels and provenance, timeline and contradictions, impact, contributing conditions, defenses, mitigation effects, limitations, follow-up rationale, ownership, lifecycle route, observable completion or effectiveness evidence, relationship meaning and direction, sensitive-data boundary, and preservation of unrelated state. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise report the mismatch and partial completion without widening authority. + +The owning systems contain the confirmed review and every safely applicable follow-up, verified against current state. + +### 6. Report and stop after learning + +Return the durable review link or path, incident and evidence boundary, stabilisation evidence, concise explanation, contradictions and unknowns, follow-up links with owners and routes, no-op or zero-action result, read-back verification, partial completion, access or capability limitations, and any unresolved critical action. Distinguish observed fact, supported interpretation, and remaining uncertainty in the handoff. + +Stop without operating production, resuming incident command or communications, implementing fixes, changing code, specifying or designing features, decomposing general delivery work, refining priorities, selecting an iteration, committing, pushing, opening or reviewing a pull request, releasing, deploying, or continuously monitoring. Later lifecycle sessions own every routed outcome. diff --git a/skills/review-incident/agents/openai.yaml b/skills/review-incident/agents/openai.yaml new file mode 100644 index 0000000..f6ca86c --- /dev/null +++ b/skills/review-incident/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Review Incident' + short_description: 'Review a stable incident and own its lessons' + default_prompt: 'Use $review-incident to explain this stabilised incident and record owned, evidence-led follow-up work.' +policy: + allow_implicit_invocation: false From 9a85fb1fe12cd0737b9f20d537438cb2bc95012a Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:22:39 +0100 Subject: [PATCH 40/79] feat(plan-retirement): add risk-led capability exit plans --- SKILLS_PLAN.md | 2 +- skills/plan-retirement/SKILL.md | 64 +++++++++++++++++ skills/plan-retirement/agents/openai.yaml | 6 ++ .../assets/retirement-plan-template.md | 71 +++++++++++++++++++ 4 files changed, 142 insertions(+), 1 deletion(-) create mode 100644 skills/plan-retirement/SKILL.md create mode 100644 skills/plan-retirement/agents/openai.yaml create mode 100644 skills/plan-retirement/assets/retirement-plan-template.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index c889e6c..72ae9ed 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -416,7 +416,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `debug` skill - [x] Create `respond-to-incident` skill - [x] Create `review-incident` skill -- [ ] Create `plan-retirement` skill +- [x] Create `plan-retirement` skill - [ ] Delete `elicit-with-context` skill - [ ] Update the README - [ ] Run the suite-wide acceptance pass diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md new file mode 100644 index 0000000..0022cdf --- /dev/null +++ b/skills/plan-retirement/SKILL.md @@ -0,0 +1,64 @@ +--- +name: plan-retirement +description: Plans the safe retirement of one obsolete capability. Use when deprecation, migration, disablement, removal, archival, or recovery need an approved durable plan before delivery work. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Plan Retirement + +Turns evidence and user-authoritative decisions into one proportionate, reversible plan for retiring an obsolete capability. The plan reduces total system complexity without treating missing usage evidence as proof that no one depends on the capability or crossing into retirement execution. + +## Process + +### 1. Establish the retirement evidence + +Inspect the request, repository guidance, product intent, `CONTEXT.md`, `ARCHITECTURE.md`, applicable ADRs, prior decisions and plans, and the smallest sufficient current implementation and runtime evidence. Trace the capability through entry points; users, tenants, callers, and operators; internal and external dependencies; public and private interfaces, APIs, events, schemas, configuration and flags; data ownership, flows, access, retention, archival, and recovery; integrations and providers; permissions and trust boundaries; observability; runbooks, support and communications; documentation; and known replacements. + +Treat code, tests, contracts, configuration, telemetry, inventories, tracker records, and operational systems as evidence within their actual scope and freshness. Distinguish confirmed use, confirmed non-use, and unknown use. Absence of observed traffic, repository references, or known consumers is not proof of no dependence. Identify lifecycle authorities and contradictions without copying mutable operational or tracker state into the plan. The capability boundary, evidence provenance, affected parties, obligations, current unknowns, and accessible authority are explicit. + +### 2. Resolve retirement intent and obligations + +Invoke `$elicit` and retain responsibility for the plan, writes, verification, and stopping boundary. Give it the inspected evidence and use one dependency-ordered question at a time to resolve the obsolete capability and reason for retiring it; acceptable end state and non-goals; timing or obligation trade-offs; consumer and operator migration; replacement expectations; compatibility and coexistence; data retention, deletion, archival, access, and recovery; integration and provider obligations; communication and support; security, privacy, compliance, and contractual constraints; acceptable reversibility and risk; ownership; and observable completion. Reuse compatible confirmed answers and ask only about material gaps. + +Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation claim could change the plan and its evidence must survive the session. Read and verify its cited report before relying on the conclusion; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-context` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. + +Keep every retirement-plan artefact unchanged until the complete synthesis is explicitly confirmed. The intended retirement outcome, authoritative obligations, unresolved decisions, and conditions that must gate irreversible work are honest and explicit. + +### 3. Compare proportionate retirement strategies + +Compare the status quo, replacement or migration, disablement without removal, staged retirement, and immediate removal where each is credible. Assess intended benefit, affected parties, consumer and operational risk, evidence gaps, cost and complexity, residual obligations, reversibility, recovery, and the condition that ends any temporary compatibility path. Prefer the least irreversible strategy that can produce the required outcome and evidence. Do not introduce permanent coexistence, shims, flags, duplicate data, or compatibility scaffolding without an explicit exit condition. + +Apply supporting techniques only when their runtime trigger is present: + +- Use **API deprecation and sunset semantics** for an externally consumed contract, distinguishing notice of discouragement from the planned end of availability and retaining direct consumer evidence, migration guidance, compatibility obligations, and completion signals. +- Use **Parallel Change** when an incompatible interface or schema must coexist while consumers or data migrate; plan the expansion, observed migration, and contraction gates without assuming the final removal is safe. +- Use the **Strangler Fig pattern** only when a substantial capability is being replaced incrementally across a stable routing or responsibility seam; keep ownership, coexistence, and the removal condition explicit. +- Use **controlled stop and recovery** when use is dormant, uncertain, or operationally mediated; plan a reversible disablement, evidence window, counter-signals, tested recovery, abort authority, and reassessment before removal. + +Choose direct atomic removal only when current evidence demonstrates that the change is local and reversible and that no consumer, data, public contract, integration, support, security, compliance, or recovery obligation requires a staged transition. Otherwise select only the applicable stages from deprecate, migrate or coexist, disable, observe, remove, and archive or close. The selected strategy is justified by evidence and confirmed trade-offs rather than ceremony or a preferred migration pattern. + +### 4. Build the complete retirement plan + +Define the selected end state and explicit non-goals, affected parties, prerequisites, owners, contract and deprecation policy, migration and compatibility path, communication and support plan, data disposition and access rules, integration and provider handling, operational and monitoring effects, security and compliance checks, archival and recovery, unresolved risks and decisions, and observable completion conditions. + +For each applicable stage, state its bounded outcome, entry evidence and prerequisites, affected contracts or state, owner, verification and counter-signals, abort conditions and authority, rollback or recovery route, residual obligations, and evidence required to exit. Use confirmed dates, deadlines, notice periods, adoption claims, retention rules, and production mechanisms only when their authority exists; otherwise preserve the unknown and its resolution gate. Ensure the final removal and closure eliminate temporary paths and stale obligations rather than merely hiding the capability. + +Trace every stage and obligation back to retirement intent, a consumer or dependent, current evidence, an external authority, or an explicit user decision. Compare the selected plan with the credible alternatives and retain only the rejected choices necessary to understand a material consequence or reversal. The proposed plan is safe to decompose later without becoming a file-by-file implementation plan, task list, estimate, sprint, release, or deployment procedure. + +### 5. Confirm the complete synthesis + +Have `$elicit` present one concise, self-contained synthesis covering the evidence boundary and unknown use, retirement outcome and non-goals, affected parties and obligations, alternatives, selected strategy, stages and ownership, migration and compatibility, communications and support, data and integrations, operations, security and compliance, verification, abort and recovery, completion conditions, unresolved risks, context changes, durable destination, and exact proposed write. Obtain explicit confirmation of the whole synthesis; selecting an earlier option is not confirmation. + +If the user rejects or revises it, leave the retirement plan and design-dependent context unchanged, reopen every affected evidence, obligation, option, stage, owner, and completion branch, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid. The confirmed synthesis supplies authority for its agreed plan and context writes without another application prompt. + +### 6. Write and verify the durable plan + +Immediately before writing, refresh mutable usage, dependency, contract, provider, retention, and operational evidence that could change the confirmed plan. When drift materially changes the strategy, obligation, stage, owner, risk, or completion claim, leave the plan unchanged and return to complete-synthesis confirmation. + +Follow the project's established retirement-record convention. Otherwise choose a stable lowercase capability slug and create `docs/retirements/.md` using the [Retirement Plan Template](assets/retirement-plan-template.md). Preserve unrelated confirmed content and local structure when refining an existing plan. Include only material sections, link authorities and mutable owning systems instead of duplicating them, and distinguish confirmed current evidence, approved future action, and unresolved gated decisions. + +After writing the plan, invoke `$maintain-context` for confirmed design-dependent terminology and any qualifying ADR whose durable-rationale agreement was included in the synthesis. Verify its returned changes. Re-read the plan against product intent, repository and runtime evidence, context and architecture authorities, external research, and the confirmed synthesis. Verify all paths and links; end-to-end dependent coverage; stage entry, exit, evidence, abort, rollback and recovery; owner and obligation coverage; data, integration, communication, support, security and compliance treatment; removal of temporary paths; and observable completion. State evidence limitations and never describe unknown or future evidence as established. + +Report the plan path, selected strategy, affected parties and obligations, stage and completion outline, confirmed context changes, supporting evidence, verification, unresolved gated decisions, and limitations. Stop after the approved retirement plan and applicable context records are written and verified. Do not create tracker items, invoke downstream lifecycle skills, implement or decommission the capability, disable or delete code, data, configuration or infrastructure, contact consumers, commit, push, open a pull request, release, deploy, or continuously monitor. Later sessions own `$create-tickets`, `$implement`, `$commit`, and `$pull-request`. diff --git a/skills/plan-retirement/agents/openai.yaml b/skills/plan-retirement/agents/openai.yaml new file mode 100644 index 0000000..a8405bc --- /dev/null +++ b/skills/plan-retirement/agents/openai.yaml @@ -0,0 +1,6 @@ +interface: + display_name: 'Plan Retirement' + short_description: 'Plan a safe evidence-led capability retirement' + default_prompt: 'Use $plan-retirement to create an approved plan for safely retiring this obsolete capability.' +policy: + allow_implicit_invocation: false diff --git a/skills/plan-retirement/assets/retirement-plan-template.md b/skills/plan-retirement/assets/retirement-plan-template.md new file mode 100644 index 0000000..0705159 --- /dev/null +++ b/skills/plan-retirement/assets/retirement-plan-template.md @@ -0,0 +1,71 @@ +# {Capability name} retirement plan + + + +## Outcome and authorities + + + +## Scope and non-goals + + + +## Current capability and evidence + + + +## Affected parties and obligations + + + +## Options and selected strategy + + + +## Retirement stages + + + +### {Stage name} + +- **Outcome:** {Bounded state this stage establishes.} +- **Entry evidence and prerequisites:** {What must be true before the stage starts.} +- **Affected contracts or state:** {Consumers, interfaces, data, integrations, controls, or operations affected.} +- **Owner:** {Accountable role or confirmed owner.} +- **Verification and counter-signals:** {Observable evidence that distinguishes success, partial completion, hidden dependence, and harm.} +- **Abort conditions and authority:** {Signals that stop progression and who decides.} +- **Rollback or recovery:** {Proportionate route back to a safe state and required assets.} +- **Exit evidence:** {What must be observed before the next stage or completion.} +- **Residual obligations:** {What remains after the stage and its exit condition.} + +## Migration and compatibility + + + +## Communication and support + + + +## Data, archival, access and recovery + + + +## Integrations and providers + + + +## Operations, security and compliance + + + +## Risks, decisions and prerequisites + + + +## Completion and closure + + + +## Later delivery handoff + + From 0290ac7fcf4da500bf9708a53c95983163c10863 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:23:00 +0100 Subject: [PATCH 41/79] refactor(skills): remove obsolete context elicitation --- SKILLS_PLAN.md | 2 +- skills/elicit-with-context/SKILL.md | 15 --------------- skills/elicit-with-context/agents/openai.yaml | 5 ----- 3 files changed, 1 insertion(+), 21 deletions(-) delete mode 100644 skills/elicit-with-context/SKILL.md delete mode 100644 skills/elicit-with-context/agents/openai.yaml diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 72ae9ed..6c94586 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -417,7 +417,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `respond-to-incident` skill - [x] Create `review-incident` skill - [x] Create `plan-retirement` skill -- [ ] Delete `elicit-with-context` skill +- [x] Delete `elicit-with-context` skill - [ ] Update the README - [ ] Run the suite-wide acceptance pass diff --git a/skills/elicit-with-context/SKILL.md b/skills/elicit-with-context/SKILL.md deleted file mode 100644 index e14c444..0000000 --- a/skills/elicit-with-context/SKILL.md +++ /dev/null @@ -1,15 +0,0 @@ ---- -name: elicit-with-context -description: Elicits shared understanding while maintaining project context. Use when an interview should update language and qualifying architecture decisions. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Elicit with Context - -Elicits shared understanding while maintaining project context. - -## Process - -Invoke `$elicit` for questioning, applying `$maintain-context` throughout. diff --git a/skills/elicit-with-context/agents/openai.yaml b/skills/elicit-with-context/agents/openai.yaml deleted file mode 100644 index 46f2204..0000000 --- a/skills/elicit-with-context/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Elicit with Context' - short_description: 'Elicit while maintaining project context' -policy: - allow_implicit_invocation: false From 95c2d76d1af302d74ea47239b24237b94441d29d Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:27:01 +0100 Subject: [PATCH 42/79] docs: document the software delivery lifecycle --- README.md | 148 +++++++++++++++++++++++++++++++++++++++++-------- SKILLS_PLAN.md | 2 +- 2 files changed, 125 insertions(+), 25 deletions(-) diff --git a/README.md b/README.md index 5bc3e80..ce8e28f 100644 --- a/README.md +++ b/README.md @@ -2,7 +2,11 @@ # Propulsion -Propulsion is a compact skill set for agentic coding. It gives coding agents a stronger workflow. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to planning work, implementation, refinement, debugging, incident response, and retirement. + +Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. + +Substantial work narrows breadth-first over several sessions: product intent becomes a feature specification, then a solution design, implementation-ready tickets, and finally one implemented body of work per ticket. Small, understood work can begin directly with `$implement`. ## Installation @@ -27,48 +31,144 @@ ln -s /absolute/path/to/propulsion/skills/elicit ~/.agents/skills/elicit Repeat the link for each selected skill. Codex and OpenCode both discover skills from `~/.agents/skills`; edits in the clone are available through the links without reinstalling or publishing a new version. -## Usage +## Where to start + +| When you need to… | Start with… | Session outcome | +| --- | --- | --- | +| Establish a new product | `$define-product` | Durable product documentation and shared language | +| Define a substantial feature | `$specify-feature` | An approved, observable feature specification | +| Make a small understood change | `$implement` | A minimal, reviewed, verified local change | +| Route incoming requests | `$triage-work` | Verified dispositions and lifecycle routes | +| Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | +| Stabilise a production incident | `$respond-to-incident` | Verified mitigation or recovery and current service state | +| Review another contributor's PR | `$review-pull-request` | Independent findings and verification evidence | +| Remove an obsolete capability | `$plan-retirement` | An approved, safe retirement plan | + +## Lifecycle skills + +The lifecycle areas are useful entry points, not mandatory phase gates. + +### Establish + +- `$define-product` creates the durable product foundation in `PRODUCT.md` and establishes its ubiquitous language in `CONTEXT.md`. +- `$define-architecture` creates the enduring technical foundation in `ARCHITECTURE.md` and records only the focused architectural decisions that need durable rationale. + +These are separate sessions. Product definition ends with product documentation; architecture begins later from that confirmed foundation. + +### Define + +- `$specify-feature` turns one capability, enhancement, or agreed idea into a feature specification describing externally observable intent, without choosing its implementation. -Propulsion skills are independently invocable. For most feature and change requests, use this recommended workflow: +### Plan -1. Shape the idea with `$elicit-with-context` until the request is understood and confirmed. +- `$design-feature` selects a buildable, system-specific technical solution for an approved feature specification. +- `$create-tickets` turns an approved specification and solution design into verified, dependency-aware work in the project's task-management tool. - ```text - $elicit-with-context Help me work through an idea for . - ``` +### Refine -2. Implement the confirmed request with `$implement`. +- `$triage-work` closes or routes a bounded set of incoming requests without performing the routed work. +- `$refine-backlog` keeps selected backlog work current, prioritised, and honest about readiness, dependencies, and blockers. +- `$plan-iteration` selects feasible ready work in the project's native sprint, cycle, milestone, or next-work mechanism. +- `$review-architecture` reports a few high-value improvements for an explicitly bounded area without changing it or creating tickets. - ```text - $implement the request we just confirmed. - ``` +### Deliver + +- `$implement` produces one minimal, reviewed, verified local change from a ticket or other confirmed work. +- `$commit` records eligible reviewed work as coherent Conventional Commits while preserving unrelated changes. +- `$pull-request` commits eligible remaining work when needed, pushes the branch, and creates or updates one accurate pull request for the whole branch. +- `$review-pull-request` independently assesses someone else's pinned pull request without changing their branch or publishing a review unless requested. + +The author's normal delivery flow does not invoke `$review-pull-request`: `$implement` already includes independent code review. The separate pull-request review skill is for work submitted by others. + +### Debug and Respond + +- `$debug` establishes an evidence-backed root cause and, when repair is requested, produces a minimal reviewed and verified local change. +- `$respond-to-incident` coordinates evidence-led stabilisation of an active production incident and records the resulting service state. +- `$review-incident` later explains a stabilised incident without blame and creates owned follow-up work through the normal lifecycle. + +### Retire + +- `$plan-retirement` defines the deprecation, migration, communication, removal, archival, recovery, and completion obligations for one obsolete capability. Execution happens in later planning and delivery sessions. + +## Representative flows + +Establish a product: + +```text +define-product + → later define-architecture +``` -3. Review the working-tree changes yourself. Ask the agent to explain or adjust anything necessary, and repeat until you are satisfied with the result. +Deliver a substantial feature: -4. Commit the reviewed changes, then create the pull request. +```text +specify-feature + → later design-feature + → later create-tickets + → implement each selected ticket in a separate session + → commit whenever a coherent unit is ready + → pull-request when the branch is ready +``` - ```text - $commit the reviewed changes. - Create a $pr for the current branch. - ``` +Deliver a small understood change: -### Alternative entry points +```text +implement → commit → pull-request +``` -For a bug, start with `$debug` instead of elicitation and implementation: +Refine team work: ```text -$debug Fix . +triage-work → refine-backlog → optionally plan-iteration ``` -After the repair, rejoin the recommended workflow at human review, followed by `$commit` and `$pr`. +Debug, respond, or retire: + +```text +debug → commit → pull-request + +respond-to-incident → later review-incident → route follow-up work + +plan-retirement → later create-tickets → normal delivery sessions +``` -For codebase improvements, start with `$review-architecture` to produce an architecture report: +Review another contributor's work: ```text -$review-architecture Review for high-value architecture improvements. +review-pull-request ``` -Review the report, then take each accepted recommendation through the recommended workflow separately, beginning with `$elicit-with-context`. +## Reusable utilities + +Utilities are independently invokable and are also composed by lifecycle skills when their trigger applies. + +| Utility | Use it to… | +| --- | --- | +| `$elicit` | Resolve material user-held information and decisions one question at a time | +| `$research` | Answer one material question with critically appraised evidence and persist a cited report | +| `$prototype` | Create a disposable visual experiment that resolves one bounded uncertainty | +| `$maintain-context` | Keep confirmed ubiquitous language and rare qualifying decisions current | +| `$modular-design` | Assess boundaries, ownership, contracts, dependencies, and change propagation | +| `$tdd` | Implement observable behaviour through red-green-refactor when its prerequisites hold | +| `$verify-change` | Run applicable project and risk-triggered quality harnesses against a fixed change | +| `$code-review` | Independently inspect a fixed change for behaviour, regressions, minimality, conventions, and maintainability | +| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | + +`$write-skill` develops and evaluates Agent Skills themselves; it is not part of the software-delivery lifecycle. + +## Task-management tools + +Tracker-backed skills use the project tool named in the root `AGENTS.md` and represent work using that tool's native issues, relationships, statuses, priorities, sprints, cycles, milestones, or queues. If no tool is named, the skill asks which one the project uses, records the preference through `$maintain-agents`, and resumes. It never guesses or introduces a Propulsion configuration layer. + +## Quality during implementation + +`$implement` uses TDD when an existing runnable suite can exercise the change through a meaningful observable boundary. It always subjects the fixed candidate to independent `$code-review` and `$verify-change`, using the repository's required checks plus additional harnesses only when the change's risks justify them. + +The workflow does not install unrelated test or quality infrastructure merely to satisfy its process. When meaningful evidence is unavailable, it reports the limitation instead of manufacturing confidence. Implementation stops with a reviewed, verified local change; committing and publication remain separate user-controlled sessions. + +## Scope + +Propulsion does not release, deploy, merge, or continuously monitor software. Those activities remain project-specific. It also avoids duplicating discoverable code, configuration, CI, tracker, or production state in prose documentation. ## Acknowledgements diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 6c94586..2dea2a4 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -418,7 +418,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `review-incident` skill - [x] Create `plan-retirement` skill - [x] Delete `elicit-with-context` skill -- [ ] Update the README +- [x] Update the README - [ ] Run the suite-wide acceptance pass Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. From 53b4d02524b382f67e7f48801ffab83cb795d7de Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:29:18 +0100 Subject: [PATCH 43/79] fix(create-tickets): accept retirement authorities --- skills/create-tickets/SKILL.md | 33 ++++++++++++++---------- skills/create-tickets/agents/openai.yaml | 4 +-- 2 files changed, 21 insertions(+), 16 deletions(-) diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md index e54e730..fd69e12 100644 --- a/skills/create-tickets/SKILL.md +++ b/skills/create-tickets/SKILL.md @@ -1,6 +1,6 @@ --- name: create-tickets -description: Creates verified native tracker items from an approved feature specification and solution design. Use when a feature needs implementation-ready, dependency-aware tickets. +description: Creates verified native tracker items from an approved feature solution or retirement plan. Use when feature delivery or capability retirement needs implementation-ready, dependency-aware tickets. metadata: invocation: user disable-model-invocation: true @@ -8,11 +8,16 @@ disable-model-invocation: true # Create Tickets -Turns one approved feature into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules and examples into observable acceptance evidence. +Turns one approved feature solution or retirement plan into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules, examples, stages, and completion conditions into observable acceptance evidence. ## Prerequisites -Require the approved feature specification and its approved solution design. Inspect both before continuing. When either document is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent to a later `$specify-feature` session and missing technical resolution to a later `$design-feature` session; do not invoke either or manufacture its outcome. +Require one applicable approved authority package and inspect it before continuing: + +- For feature delivery, require the approved feature specification and its approved solution design. When either document is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent to a later `$specify-feature` session and missing technical resolution to a later `$design-feature` session; do not invoke either or manufacture its outcome. +- For capability retirement, require one approved retirement plan that resolves the retirement outcome; affected parties, dependencies, interfaces, data, and integrations; applicable deprecation, migration, coexistence, disablement, removal, archival, communication, support, and recovery obligations; material risks and rollback or recovery routes; and observable stage and final completion conditions sufficiently for honest decomposition. When a needed retirement decision is missing or contradictory, identify the exact gap and stop for a later `$plan-retirement` session; do not force retirement through feature specification or solution design or manufacture the plan's outcome. + +A retirement plan may deliberately depend on a later feature specification and solution design for a replacement capability. Do not create implementation work for that unresolved replacement until those authorities are approved. Retirement-only obligations may be decomposed when they are independently confirmed and do not require the unresolved replacement decision. Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user directly which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool or required account is unavailable or not writable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. @@ -20,15 +25,15 @@ Require the project's task-management tool to be named in the root `AGENTS.md`. ### 1. Fix the authorities and native surface -Inspect the request, repository guidance, approved specification and solution design, their linked product, context, architecture, ADR, and research authorities, and only the implementation evidence needed to understand existing work. Treat the specification as authority for observable intent, the solution design as authority for the selected implementation and genuine technical prerequisites, and the task-management tool as authority for live work state. Discover obvious facts before asking the user. +Inspect the request, repository guidance, applicable approved authority package, its linked product, context, architecture, ADR, and research authorities, and only the implementation evidence needed to understand existing work. For feature delivery, treat the specification as authority for observable intent and the solution design as authority for the selected implementation and genuine technical prerequisites. For retirement, treat the retirement plan as authority for the approved outcome, obligations, stages, risks, recovery, and completion conditions. Treat separately approved replacement feature documents as authority only for replacement work they actually resolve, and treat the task-management tool as authority for live work state. Discover obvious facts before asking the user. Inspect the configured tool's available schema and capabilities before planning any mutation: native item types, bodies, links, parent or grouping concepts, dependency direction, duplicate or related relationships, and read-back operations. Use only concepts the tool actually supports. Keep the tracker read-only through proposal confirmation. -The source authorities, tracker scope, native capabilities, and access boundary are explicit. +The applicable authority package, tracker scope, native capabilities, and access boundary are explicit. ### 2. Reconcile existing work -Read the relevant open and closed tracker state before decomposing. Search by authoritative document links or identifiers, feature and outcome language, affected project concepts, and related or parent items. Classify candidates as exact matches, partial overlaps, related work, or unrelated work. Reuse an exact match, propose a bounded reconciliation for a partial overlap, preserve merely related work, and avoid creating a second representation of the same outcome. +Read the relevant open and closed tracker state before decomposing. Search by authoritative document links or identifiers, feature or retirement outcome language, affected project concepts, and related or parent items. Classify candidates as exact matches, partial overlaps, related work, or unrelated work. Reuse an exact match, propose a bounded reconciliation for a partial overlap, preserve merely related work, and avoid creating a second representation of the same outcome. Record the classification and evidence internally. Do not copy live tracker state into repository documentation. The later proposal can distinguish create, update, relationship-only, and no-op reuse actions without inventing a parallel backlog. @@ -36,21 +41,21 @@ The decomposition starts from a current, duplicate-aware view of native work. ### 3. Form vertical outcomes -Build an internal two-way coverage map from every material specification behaviour, rule, state, scenario, constraint, and acceptance condition, plus every necessary solution responsibility, transition, and verification seam, to proposed tracker outcomes. Each outcome must trace back to an approved source, and every approved requirement necessary to deliver the feature must have an owner. +Build an internal two-way coverage map from the applicable authority package to proposed tracker outcomes. For feature delivery, cover every material specification behaviour, rule, state, scenario, constraint, and acceptance condition plus every necessary solution responsibility, transition, and verification seam. For retirement, cover every material affected party, dependency, interface, data and integration obligation, applicable stage and transition, communication and support obligation, risk control, abort, rollback or recovery route, residual obligation, and stage and final completion condition. Each outcome must trace back to an approved source, and every approved requirement or obligation necessary for the requested delivery or retirement scope must have an owner. -Form the smallest set of coherent vertical outcomes. Each item spans every affected layer needed to produce one observable result for an actor, consumer, published contract, migration, or operational boundary. Include a shared enabling change in the earliest outcome that needs it unless the approved solution establishes a separately observable and independently verifiable prerequisite. Do not split work by frontend, backend, database, service, class, file, architecture layer, test type, or another implementation inventory merely because those parts exist. One item is correct when the feature is already one bounded independently implementable outcome. +Form the smallest set of coherent vertical outcomes. Each item spans every affected layer needed to produce one observable result or fulfilled obligation for an actor, consumer, published contract, migration, retirement stage, or operational boundary. Include a shared enabling change in the earliest outcome that needs it unless an approved authority establishes a separately observable and independently verifiable prerequisite. Do not split work by frontend, backend, database, service, class, file, architecture layer, test type, retirement phase label, or another implementation inventory merely because those parts exist. One item is correct when the approved package is already one bounded independently implementable outcome. -For each item, define a concise native title, one coherent outcome, enough bounded specification and solution context for a separate implementation session, links to both authoritative feature documents, observable acceptance evidence, and applicable relationships. Derive acceptance evidence from approved rules and discriminating examples; use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. +For each item, define a concise native title, one coherent outcome, enough bounded authority context for a separate implementation session, links to the applicable approved authority package, observable acceptance evidence, and applicable relationships. A feature item links both authoritative feature documents. A retirement item links the retirement plan and any separately applicable approved replacement specification and solution design when that item includes replacement work. Derive acceptance evidence from approved rules and discriminating examples for feature delivery, or from approved stage outcomes, exit evidence, counter-signals, recovery obligations, and completion conditions for retirement. Use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. Check each item directly for independent value, bounded size, and testability. Minimise coupling between items while representing unavoidable prerequisites honestly. Infer a blocker only when the approved solution or required state transition establishes that one outcome cannot begin or complete before another; do not convert user-journey order, convenient sequencing, or tracker display order into dependency. Interpret the configured tool's relationship direction explicitly and reject cycles. -When a question exposes missing feature intent or technical design, stop with the owning source gap. Invoke `$elicit` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. +When a question exposes missing feature intent or technical design, stop with the owning source gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification and solution design exist; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. The draft items are vertically coherent, source-complete, independently actionable where possible, and linked only by genuine acyclic prerequisites. ### 4. Challenge and propose the set -Trace every proposed item back to its approved need and every approved need forward to exactly the necessary items. Remove unapproved scope, duplicated outcomes, speculative abstractions, horizontal work, incidental dependencies, and detail that belongs only in code. Confirm that an implementation agent could complete each item in one bounded session without rediscovering feature intent or selecting a missing technical design. +Trace every proposed item back to its approved need or obligation and every approved need or obligation forward to exactly the necessary items. Remove unapproved scope, duplicated outcomes, speculative abstractions, horizontal work, incidental dependencies, and detail that belongs only in code. Confirm that an implementation agent could complete each item in one bounded session without rediscovering observable intent, selecting a missing technical or retirement decision, or inventing unresolved replacement behaviour. Repeat the live duplicate search using the final titles, outcomes, authority links, and concepts. Reconcile the result into one complete proposal containing every native create, update, relationship-only, and no-op reuse action; the exact item contents; relationship types and directions; and any native capability limitation or residual mismatch. Do not include estimates, priorities, sprint or cycle assignments, milestones, target dates, releases, implementation sequencing beyond genuine blockers, or speculative future work. @@ -68,8 +73,8 @@ The configured tracker contains exactly the agreed native representation or an e ### 6. Read back, verify, and stop -Read every created or updated item and every affected relationship back from the tool. Compare titles, outcomes, bounded context, authoritative links, acceptance evidence, relationship types and directions, and create, update, or reuse disposition with the confirmed proposal. Correct an in-scope discrepancy when the agreed value and affected record are unambiguous; stop and report a mismatch when correction would require a new user decision, unsupported capability, or mutation of unrelated work. +Read every created or updated item and every affected relationship back from the tool. Compare titles, outcomes, bounded context, applicable authority-package links, acceptance evidence, relationship types and directions, and create, update, or reuse disposition with the confirmed proposal. Correct an in-scope discrepancy when the agreed value and affected record are unambiguous; stop and report a mismatch when correction would require a new user decision, unsupported capability, or mutation of unrelated work. -Repeat the relevant duplicate search and verify two-way source coverage, vertical coherence, independent actionability, observable completion, genuine acyclic blockers, and preservation of unrelated native state. Report exact item links, created, updated, related, and reused records, verification performed, unsupported capabilities, concurrent-state or duplicate uncertainty, and unresolved source limitations. +Repeat the relevant duplicate search and verify two-way authority-package coverage, vertical coherence, independent actionability, observable completion, genuine acyclic blockers, and preservation of unrelated native state. For retirement, verify that stage ordering became a blocker only where an approved entry condition or required state transition makes it genuine, that retirement-only work does not depend on an unresolved replacement without authority, and that every item links its applicable approved authorities. Report exact item links, created, updated, related, and reused records, verification performed, unsupported capabilities, concurrent-state or duplicate uncertainty, and unresolved source limitations. -Stop after the native items and relationships are read back and verified. Do not change product, feature, solution, architecture, or context documents; select or schedule an iteration; estimate or prioritise work; scaffold or implement code; commit or publish changes; release; or deploy. +Stop after the native items and relationships are read back and verified. Do not change product, feature, solution, retirement, architecture, or context documents; select or schedule an iteration; estimate or prioritise work; scaffold or implement code; commit or publish changes; release; or deploy. diff --git a/skills/create-tickets/agents/openai.yaml b/skills/create-tickets/agents/openai.yaml index eb40b46..65486ce 100644 --- a/skills/create-tickets/agents/openai.yaml +++ b/skills/create-tickets/agents/openai.yaml @@ -1,6 +1,6 @@ interface: display_name: 'Create Tickets' - short_description: 'Create implementation-ready tracker work' - default_prompt: 'Use $create-tickets to turn this approved feature specification and solution design into implementation-ready tracker items.' + short_description: 'Create delivery or retirement tracker work' + default_prompt: 'Use $create-tickets to turn this approved feature solution or retirement plan into implementation-ready tracker items.' policy: allow_implicit_invocation: false From 5ba3bfe694a1df75d94c55ffa4079009f6d019bd Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 16 Aug 2026 12:37:26 +0100 Subject: [PATCH 44/79] fix(skills): close suite acceptance gaps --- README.md | 2 +- SKILLS_PLAN.md | 8 ++++---- skills/respond-to-incident/SKILL.md | 2 +- skills/review-pull-request/SKILL.md | 4 ++-- 4 files changed, 8 insertions(+), 8 deletions(-) diff --git a/README.md b/README.md index ce8e28f..c88d017 100644 --- a/README.md +++ b/README.md @@ -62,7 +62,7 @@ These are separate sessions. Product definition ends with product documentation; ### Plan - `$design-feature` selects a buildable, system-specific technical solution for an approved feature specification. -- `$create-tickets` turns an approved specification and solution design into verified, dependency-aware work in the project's task-management tool. +- `$create-tickets` turns an approved feature specification and solution design, or an approved retirement plan, into verified, dependency-aware work in the project's task-management tool. ### Refine diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 2dea2a4..2befbc6 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -136,9 +136,9 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `create-tickets` -- **Outcome:** An approved feature is represented by implementation-ready, dependency-aware work in the project's task-management tool. -- **Inputs:** The approved feature specification and solution design, plus relevant project guidance and current tracker state. -- **Output:** Native tracker items linked to their authoritative feature documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. +- **Outcome:** An approved feature solution or retirement plan is represented by implementation-ready, dependency-aware work in the project's task-management tool. +- **Inputs:** Either the approved feature specification and solution design or an approved retirement plan, plus relevant project guidance and current tracker state. +- **Output:** Native tracker items linked to their applicable authoritative documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. - **Composition:** Conditionally invokes `elicit` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. - **Stops:** After the created items and relationships are read back and verified. It does not select an iteration or begin implementation. @@ -419,7 +419,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `plan-retirement` skill - [x] Delete `elicit-with-context` skill - [x] Update the README -- [ ] Run the suite-wide acceptance pass +- [x] Run the suite-wide acceptance pass Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. diff --git a/skills/respond-to-incident/SKILL.md b/skills/respond-to-incident/SKILL.md index e71a598..8cccd29 100644 --- a/skills/respond-to-incident/SKILL.md +++ b/skills/respond-to-incident/SKILL.md @@ -38,7 +38,7 @@ Record the action and actor, execute only the agreed operation, read back the mu Invoke `$debug` only when a bounded causal diagnosis or local repair is necessary to stabilise this incident. Supply the exact signal, production boundary, time constraint, and stopping condition; retain ownership of impact, mitigation, production actions, communications, and recovery. Do not let diagnosis become open-ended or delay a safe mitigation. Invoke `$verify-change` for a frozen local code or configuration repair, or a fixed operational change that existing harnesses can appraise; its verdict supplements but never replaces current production recovery evidence. Invoke `$research` only for one material external question whose evidence must survive the session. -Do not silently commit, push, publish, release, or deploy a repair. Use a project-native route only when the requested incident outcome and current operational authority explicitly include it. When stabilisation requires publication or deployment but no authorised project-native mechanism is available, stop that route with the exact handoff, owner, artefact, and service risk rather than improvising a Propulsion release process. +Do not commit, push, publish, release, or deploy a repair. When stabilisation requires one of those project-specific actions, stop that route with the exact external handoff, owner, artefact, required mechanism, and service risk rather than performing or improvising it inside Propulsion. ### 5. Establish the service state diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index bd8e06e..a1e4249 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -30,7 +30,7 @@ Invoke `$modular-design` only when the complete candidate presents a material ow Invoke `$code-review` with the frozen base and head SHAs, complete candidate, intended-behaviour and Standards authorities, and a read-only boundary. Require its independent Standards and applicable Spec reviews. Preserve their findings, distinctions, priorities, evidence, ordering, omitted axes, and limitations without merging, suppressing, cross-axis reranking, adjudicating, or remediating them. -Re-read the remote head after the review. If it differs from the pinned head, retain the result as stale for its original revision. Pin the new base and head and repeat the complete review only when the user's requested outcome still identifies the same pull request unambiguously; otherwise stop and ask for direction. The current code-review result refers to one explicit candidate. +Re-read the remote base and head after the review. If either differs from its pinned SHA, retain the result as stale for its original candidate. Pin the new base and head and repeat the complete review only when the user's requested outcome still identifies the same pull request unambiguously; otherwise stop and ask for direction. The current code-review result refers to one explicit candidate. ### 4. Verify the same candidate @@ -38,7 +38,7 @@ Invoke `$verify-change` on exactly the base and head SHAs reviewed in step 3. Su Use existing CI evidence only when its provenance ties it to the pinned head and its execution boundary and result are inspectable. When local verification cannot run safely or the repository lacks faithful infrastructure, retain the exact `not run`, partial, or unverified evidence; do not install tools, add tests or ad hoc harnesses, alter source, tests, snapshots, baselines, suppressions, dependencies, or checks, or treat static inspection as execution. Passing checks do not erase a code-review finding, and verification never certifies the pull request. -Re-read the remote head after verification. If it changed, preserve both results as stale for their original revision and repeat steps 2–4 on the newly pinned revision only while the requested outcome remains clear. Stop with the stale results and exact blocker when repeated drift prevents one stable current assessment. The completed review and verification evidence refer to identical current SHAs or are explicitly stale. +Re-read the remote base and head after verification. If either differs from its pinned SHA, preserve both results as stale for their original candidate and repeat steps 2–4 on the newly pinned candidate only while the requested outcome remains clear. Stop with the stale results and exact blocker when repeated drift prevents one stable current assessment. The completed review and verification evidence refer to identical current base and head SHAs or are explicitly stale. ### 5. Report the assessment From 88c573e9ad13d0e2fe7a933aeb45b8e8330dea85 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Mon, 17 Aug 2026 21:44:30 +0100 Subject: [PATCH 45/79] feat(write-skill): replace authoring workflow --- bun.lock | 6 + package.json | 4 +- skills/code-review/SKILL.md | 2 +- skills/maintain-context/SKILL.md | 2 +- .../assets/research-report-template.md | 2 +- skills/review-pull-request/SKILL.md | 2 +- skills/write-skill/SKILL.md | 102 ++- skills/write-skill/agents/openai.yaml | 2 +- skills/write-skill/assets/skill-template.md | 13 + skills/write-skill/references/CRAFT.md | 43 - skills/write-skill/references/SECTIONS.md | 60 +- skills/write-skill/references/TESTING.md | 33 - skills/write-skill/scripts/measure-context.js | 347 ++++++++ skills/write-skill/scripts/validate-skill.js | 752 ++++++++++++------ 14 files changed, 966 insertions(+), 404 deletions(-) create mode 100644 skills/write-skill/assets/skill-template.md delete mode 100644 skills/write-skill/references/CRAFT.md delete mode 100644 skills/write-skill/references/TESTING.md create mode 100644 skills/write-skill/scripts/measure-context.js diff --git a/bun.lock b/bun.lock index 050f596..8d0c7df 100644 --- a/bun.lock +++ b/bun.lock @@ -8,6 +8,8 @@ "oxfmt": "^0.59.0", "oxlint": "^1.74.0", "oxlint-tsgolint": "^0.24.0", + "semver": "7.8.5", + "tiktoken": "1.0.22", }, }, }, @@ -106,6 +108,10 @@ "oxlint-tsgolint": ["oxlint-tsgolint@0.24.0", "", { "optionalDependencies": { "@oxlint-tsgolint/darwin-arm64": "0.24.0", "@oxlint-tsgolint/darwin-x64": "0.24.0", "@oxlint-tsgolint/linux-arm64": "0.24.0", "@oxlint-tsgolint/linux-x64": "0.24.0", "@oxlint-tsgolint/win32-arm64": "0.24.0", "@oxlint-tsgolint/win32-x64": "0.24.0" }, "bin": { "tsgolint": "bin/tsgolint.js" } }, "sha512-giCk5sEvG02d5tzPmFMX3hem8ndzEEu1xvGYS5OwNfO2WGl6ZVxt5LjE0yiMDoz94INI7XkXwgFAQiydPvVHDw=="], + "semver": ["semver@7.8.5", "", { "bin": { "semver": "bin/semver.js" } }, "sha512-Y7/KDsb8LjooZpwaqGyulO6DQlksgCncchHGk+sZIY4SBvUocMBEFH5Ur1fI4dV+Jvl0w6cjvucaIi40puRioA=="], + + "tiktoken": ["tiktoken@1.0.22", "", {}, "sha512-PKvy1rVF1RibfF3JlXBSP0Jrcw2uq3yXdgcEXtKTYn3QJ/cBRBHDnrJ5jHky+MENZ6DIPwNUGWpkVx+7joCpNA=="], + "tinypool": ["tinypool@2.1.0", "", {}, "sha512-Pugqs6M0m7Lv1I7FtxN4aoyToKg1C4tu+/381vH35y8oENM/Ai7f7C4StcoK4/+BSw9ebcS8jRiVrORFKCALLw=="], } } diff --git a/package.json b/package.json index 55f6f9c..13a19a5 100644 --- a/package.json +++ b/package.json @@ -11,7 +11,9 @@ "devDependencies": { "oxfmt": "^0.59.0", "oxlint": "^1.74.0", - "oxlint-tsgolint": "^0.24.0" + "oxlint-tsgolint": "^0.24.0", + "semver": "7.8.5", + "tiktoken": "1.0.22" }, "packageManager": "bun@1.3.11" } diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index b78992b..7605eca 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,6 +1,6 @@ --- name: code-review -description: Independently diagnoses a fixed code change against intended behaviour and engineering standards. Use when reviewing code directly or during implementation, debugging, or pull-request review. +description: Diagnoses a fixed code change independently against intended behaviour and engineering standards. Use when reviewing code directly or during implementation, debugging, or pull-request review. metadata: invocation: model disable-model-invocation: false diff --git a/skills/maintain-context/SKILL.md b/skills/maintain-context/SKILL.md index b726c30..d2aaeec 100644 --- a/skills/maintain-context/SKILL.md +++ b/skills/maintain-context/SKILL.md @@ -1,6 +1,6 @@ --- name: maintain-context -description: Continually maintains confirmed project terminology and rare ADRs. Use when project language is introduced, changed, misused, or contradicted, or an accepted decision needs durable rationale. +description: Maintains confirmed project terminology and rare ADRs continuously. Use when project language is introduced, changed, misused, or contradicted, or an accepted decision needs durable rationale. metadata: invocation: model disable-model-invocation: false diff --git a/skills/research/assets/research-report-template.md b/skills/research/assets/research-report-template.md index c5fee5d..4ccfb10 100644 --- a/skills/research/assets/research-report-template.md +++ b/skills/research/assets/research-report-template.md @@ -42,4 +42,4 @@ status: current ## Sources -- [{Source title}]({URL or repository-relative path}) — {publisher or owner, version or publication date, accessed YYYY-MM-DD, and evidential role} +- {Markdown link to source} — {publisher or owner, version or publication date, accessed YYYY-MM-DD, and evidential role} diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index a1e4249..9c95e7d 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -1,6 +1,6 @@ --- name: review-pull-request -description: Independently reviews someone else's pinned pull request with code findings and verification evidence. Use when assessing an external contributor's pull request. +description: Reviews someone else's pinned pull request independently with code findings and verification evidence. Use when assessing an external contributor's pull request. metadata: invocation: user disable-model-invocation: true diff --git a/skills/write-skill/SKILL.md b/skills/write-skill/SKILL.md index e665315..5fdde55 100644 --- a/skills/write-skill/SKILL.md +++ b/skills/write-skill/SKILL.md @@ -1,57 +1,107 @@ --- name: write-skill -description: Creates and updates compact agent skills through confirmed behavioural contracts, evidence-backed choices, and fresh-agent evaluation. Use when authoring or revising a skill. -metadata: - invocation: user +description: Creates and revises effective agent skills through confirmed requirements, minimum-sufficient design, and baseline-controlled evaluation. Use when authoring or maintaining a reusable agent skill. disable-model-invocation: true --- # Write Skill -Creates or updates the smallest skill bundle that reliably teaches a confirmed process while leaving project-dependent results to runtime evidence and user decisions. +Creates minimum-sufficient agent skills whose observed behaviour justifies their context and execution cost. ## Process -### 1. Establish the behavioural contract +### 1. Confirm the contract -Inspect the request, complete target bundle, direct callers and callees, and host conventions. Invoke `$elicit` to confirm the skill's purpose, trigger, required inputs, common successful path, observable result, and resource needs. Retain an exception only when representative evidence, the primary behaviour, or a necessary safety, permission, or prerequisite boundary requires it. Existing and new skills reach one explicit behavioural contract with a dominant path and only its material exceptions. +Inspect the request, an existing skill and its direct composition when revising, and task-scoped evidence without changing the target. Invoke `$elicit` to confirm the recurring job or capability gap; user and model triggers plus non-triggers; inputs and prerequisites; required actions, material branches, authority, and stopping boundary; expected output or handoff; direct composition; necessary resources; and observable success evidence. Keep the target unchanged until `$elicit` returns the confirmed synthesis. The authoring contract is explicit and closed. -### 2. Compare governing methodologies +### 2. Establish the matched baseline -Use the confirmed behavioural contract as the fixed scope and decision authority for both research passes; evidence selects how to teach the confirmed behaviour. Invoke `$research` to compare credible governing methodologies and a methodology-free process until further evidence is unlikely to change the decision set. Analyse the report against the contract. Call a recommendation strong only when one option materially fits better than every credible alternative; otherwise state that no strong recommendation exists. Present the supported options, behavioural consequences, trade-offs, and recommendation strength for the user to choose. The user's selection, including no methodology, is explicit. +Freeze one representative common-path task, harness, tools, permissions, and task-scoped inputs from the confirmed success evidence. Before designing the candidate, run isolated fresh agents: -### 3. Select supporting concepts +- New skill: no skill. +- Revised skill: no skill and the current skill. -Invoke `$research` to find established principles, theories, methods, or techniques that reinforce distinct concerns without competing with the selected methodology or replacing an evidence-backed absence. Analyse whether each name changes behaviour, overlaps another concept, or is more precise in plain language. Apply the same recommendation-strength rule, explain each candidate's behavioural job, and let the user decide. Zero concepts is valid; every retained concept earns a distinct role. +Record each result, loaded path, output, tool calls, and iterations. Design only against observed failure or excess burden. Stop when no skill fulfils the confirmed behaviour and no current-skill defect, contract change, or burden remains. -### 4. Confirm the design +### 3. Bound one capability -Present one complete synthesis of the behavioural contract, selected methodology or absence, supporting concepts, structure, resources, common successful path, material exceptions, constraint levels, and observable success conditions. Obtain explicit confirmation before following the remaining process. +Apply the **KISS principle**: retain the common path and material branches; remove speculative edge cases, unnecessary artefacts, and ceremony that cannot change the evaluated outcome. Keep coupled actions and artefacts together when they serve one outcome. Split independently useful capabilities when their trigger families, outcomes, authority, or success boundaries differ. The candidate has one coherent behavioural purpose. -### 5. Write the bundle +Choose invocation from actual use. Make regularly useful skills model-invoked. Make occasional, deliberate, or composition-only skills user-invoked. Set `disable-model-invocation` and `agents/openai.yaml` consistently. The skill enters context only through its intended route. -Create or update through one path. Treat the confirmed design as closed: encode its common successful path and retained material exceptions without adding behaviour during drafting. Give every `SKILL.md` frontmatter, one H1, a concise introduction, and exactly one `## Process`; when the user selected a governing methodology, place it in bold where it fits naturally in the introduction. Bold each supporting concept at its first behaviour-governing use. Consult [Skill Sections](references/SECTIONS.md) for invocation metadata, optional sections, and resource placement. The bundle expresses the confirmed design in the fewest instructions that reliably change behaviour. +### 4. Design the loading path -### 6. Place runtime information +Use a 1-64 character lowercase name with single hyphen separators and match the directory name. Write the description as one line that starts with a third-person action verb, says what the skill does, includes `Use when`, `Use for`, or `Use to`, and front-loads natural trigger words. Add a human-readable display name and short description to `agents/openai.yaml`. -Apply **progressive disclosure** to the Agent Skills loading model: keep discovery conditions in metadata, instructions required on every invocation in `SKILL.md`, and conditional detail in task-needed resources. Link every resource beside the condition that loads it. The common path remains complete without loading an irrelevant branch. +Use **progressive disclosure**: -### 7. Calibrate the instructions +- Put concise discovery and invocation metadata in frontmatter and `agents/openai.yaml`. +- Put every-run decisions, actions, authority, branches, and stopping conditions in `SKILL.md`. +- Link conditional detail inline at the branch that needs it. +- Put reusable output material in `assets/`. +- Put deterministic transformations and checks in `scripts/`. -Apply **degrees of freedom** to each instruction and evaluation criterion. Keep project-dependent reasoning and results flexible; make required sequences, safety, permissions, fixed formats, schemas, scripts, and other fragile deterministic behaviour correspondingly exact. Give each step one coherent action and observable outcome, keep each rule in one authoritative location, state the desired action first, and pair an essential prohibition with the permitted route. Consult [Skill Craft](references/CRAFT.md) when recommendation strength, wording, structure, authority, or composition remains unclear. Every constraint is no stronger than the confirmed behaviour requires. +Keep required coupled guidance together. Prefer direct one-level references. Remove unreachable and unlinked resources. Trace direct composition across trigger, inputs, authority, result, owner, and stopping boundary. The smallest sufficient context is reachable when needed. -### 8. Validate the mechanics +### 5. Draft the isolated candidate -Run [scripts/validate-skill.js](scripts/validate-skill.js), inspect every bundled script, and execute each within a disposable filesystem using inert fixtures, isolated credentials, and an environment incapable of external mutation. When that boundary is unavailable, leave the script unexecuted and report the limitation. The mechanical contract passes within the safe execution boundary. +Create the candidate outside the requested destination. For revisions, start from the current bundle and preserve behaviour outside the confirmed change. -### 9. Verify direct composition +Use [the skill template](assets/skill-template.md). Load [the section guide](references/SECTIONS.md) only when optional sections or subsection layout are needed. Write commands, not essays: -Trace every direct caller and callee across trigger, supplied inputs and authority, promised result, responsibility owner, and stopping boundary. Remove overlap, omissions, contradictions, and invented obligations. Exercise the least restrictive connected path that would expose a genuine incompatibility. Direct compositions preserve the confirmed contract. +- Use direct verbs such as _Load_, _Check_, _Run_, _Fix_, _Stop_, and _Return_. +- Give each step or subsection one coherent behavioural idea and one observable outcome. +- Use short concrete words and clear fragments; expand conditions, risks, and handoffs that compression could hide. +- State each normative meaning once. Remove filler, hedging, generic knowledge, and repeated enforcement. +- State the intended action first. Retain prohibitions only for essential safety, permission, scope, factual, or deterministic boundaries, paired with the safe action. +- Use a recognised method, principle, theory, or technique only when its canonical name replaces explanation or sharpens a decision, action, or stopping condition. Bold key concepts where they occur naturally. Remove terms that forward testing shows are decorative. +- Match **degrees of freedom** to variability and fragility. Use exact commands or scripts only when the operation requires them. +- Apply **context isolation** to every independently actionable step. Write an explicit fresh-agent step when task-scoped inputs are sufficient and main-thread context is irrelevant or could bias the result; pass only the required inputs and retain only its result. Keep shared understanding, user authority, accumulated state, and process continuity in the main thread. Name the execution boundary instead of leaving delegation optional. +- Give agents a direct result and stopping condition. Add review or feedback only in response to observable evidence, never as recurring ceremony. -### 10. Evaluate the behaviour +The candidate contains only behaviour-changing instructions and reachable resources. -Give a fresh agent only the finished bundle, a realistic request, and task-local evidence. Compare its observable process and result with the confirmed contract. Judge variable results by their correctness and suitability for the supplied evidence; inspect exact content only where the contract fixes it. Evaluate the common successful path and the smallest scenario for each retained material exception. Consult [Fresh-Agent Behavioural Evaluation](references/TESTING.md) for isolation, evidence, and repair. Repair the smallest steering cause, recompress, revalidate, and reevaluate until the contract holds. +### 6. Validate and measure -### 11. Report the result +Run [scripts/validate-skill.js](scripts/validate-skill.js) and [scripts/measure-context.js](scripts/measure-context.js) with Bun: -Return the changed files, research evidence, mechanical and composition results, behavioural-evaluation scenarios and outcomes, unexecuted scripts, and remaining uncertainty. The user receives the finished bundle and evidence that its contract holds. +```sh +bun scripts/validate-skill.js /path/to/candidate-skill +bun scripts/measure-context.js /path/to/skill \ + --include references/loaded-reference.md \ + --output /path/to/agent-output.md +``` + +Fix mechanical failures. Repeat `--include` for every reference an arm loaded and `--output` for every arm output; omit unused options. Tokenize a no-skill output through an evaluated bundle and record its discovery and loaded-path cost as zero. Treat exact `o200k_base` counts as comparative evidence, not a target or quality score. Keep discovery metadata, selected path, generated output, and bundle inventory separate. + +### 7. Complete the forward comparison + +Run the candidate on the frozen common-path task in an isolated fresh agent with the same harness, tools, permissions, and task-scoped inputs as the recorded arms. Record its result, loaded path, agent output, tool calls, and iterations. + +Give the anonymised no-skill and candidate results, plus the current-skill result for revisions, to a separate fresh evaluator with only the confirmed behaviour, expected outcome, and efficiency criteria. Require a meaningful behavioural shift over no skill; for revisions, also require preserved or improved behaviour over the current skill. Among behaviourally equivalent candidates, prefer the lower context and execution burden. The candidate's effect is independently evidenced. + +### 8. Refine and publish + +Change the candidate only to correct a specific evaluator failure or remove measured burden, then rerun the affected mechanical and forward checks. Stop when the candidate passes, a smaller equivalent candidate wins, or further change lacks observed justification. + +Publish the passing candidate to the requested destination and rerun validation and measurement there. When the candidate is not justified, keep the existing destination unchanged and return the evidence. The published bundle is the smallest evaluated version that fulfils the confirmed contract. + +## Rules + +- Preserve exact technical meaning: required keywords, paths, commands, API names, error text, safety boundaries, and ordering constraints. +- Treat output tokens, tool calls, and iterations as part of skill cost. Accept extra work only when it changes the required outcome. +- Keep mechanical checks deterministic. Judge semantic force, routing language, useful concepts, and unnecessary ceremony through the forward comparison. + +## Handoff + +Return the destination, confirmed contract, validation result, published files, evaluator verdict, residual uncertainty, and this comparison. Use `n/a` for Current on new skills. Compare Candidate delta with No skill for creation and Current for revision; also state a revision's behavioural shift over No skill. + +| Metric | No skill | Current | Candidate | Candidate delta | +| --- | --- | --- | --- | --- | +| Behavioural result and evaluator evidence | | | | | +| Discovery metadata tokens | | | | | +| Loaded skill-path tokens | | | | | +| Agent output tokens | | | | | +| Tool calls | | | | | +| Iterations | | | | | +| Total measured token burden | | | | | diff --git a/skills/write-skill/agents/openai.yaml b/skills/write-skill/agents/openai.yaml index bc4b0a4..e1917b9 100644 --- a/skills/write-skill/agents/openai.yaml +++ b/skills/write-skill/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Write Skill' - short_description: 'Create and update compact, evidence-tested skills' + short_description: 'Author skills through controlled evaluation' policy: allow_implicit_invocation: false diff --git a/skills/write-skill/assets/skill-template.md b/skills/write-skill/assets/skill-template.md new file mode 100644 index 0000000..196bd21 --- /dev/null +++ b/skills/write-skill/assets/skill-template.md @@ -0,0 +1,13 @@ +--- +name: skill-name +description: Creates the intended outcome. Use when the skill's confirmed trigger applies. +disable-model-invocation: true +--- + +# Skill Name + +States what the skill enables in one sentence. + +## Process + +Describe the minimum-sufficient behaviour in the form best suited to the work. diff --git a/skills/write-skill/references/CRAFT.md b/skills/write-skill/references/CRAFT.md deleted file mode 100644 index c718f44..0000000 --- a/skills/write-skill/references/CRAFT.md +++ /dev/null @@ -1,43 +0,0 @@ -# Skill Craft - -Use this reference when recommendation strength, supporting concepts, wording, structure, authority, or direct composition remains unclear. - -## Judge recommendations - -Research candidates against the confirmed behaviour rather than selecting one for familiarity. Compare each credible option, including no methodology, through: - -- the process each method would impose; -- the behaviours it strengthens or suppresses; -- its fit across the common successful path and retained material exceptions; -- the local adaptations it would require; and -- the evidence supporting its published meaning. - -Stop when further evidence is unlikely to change the decision set. Call a recommendation strong only when one option materially fits the contract better than every credible alternative. Otherwise state that no strong recommendation exists. Present the evidence, adaptations, consequences, and uncertainty without manufacturing a ranking; the user decides. - -## Select supporting concepts - -A supporting concept earns inclusion only when its established meaning governs a distinct unresolved concern. Name its behavioural job in one sentence, test whether deleting its name changes execution, and compare it with plain language and overlapping concepts. Retain as many as distinct concerns require, including zero, and keep a short local mapping where the canonical name is broader than the required behaviour. - -## Write the smallest effective bundle - -Assume the agent already knows general software practice. Retain a word only when it changes behaviour, preserves a condition, or improves navigation. Prefer a canonical term when it replaces explanation and sharpens a decision; remove it when the local rule must restate the entire useful meaning. - -Bold a selected governing methodology naturally within the introduction. Bold a supporting concept where it first governs behaviour, then use plain text unless renewed emphasis changes execution. - -## Shape coherent instructions - -Give a step or subsection one coherent behavioural idea and one observable outcome. Split it when any sentence could be acted on, tested, reordered, or completed independently. Keep supporting detail together when separating it would make the action harder to understand or create headings without meaningful content. - -Match precision to the confirmed behaviour. Leave valid project-dependent choices open. Specify exact sequences, permissions, formats, assets, scripts, and other fragile or deterministic behaviour only to the degree the contract fixes them. - -## Preserve one authority - -Keep each rule and definition in one authoritative location. Allow deliberate summaries, emphasis, and caller-facing pointers when they navigate to that authority without becoming competing versions. Treat the environment as authoritative for readily discoverable commands, configuration, and structure; document only what inspection cannot reveal safely. - -## State the target behaviour - -State the desired action or result first. Retain a prohibition when the excluded behaviour defines safety, permission, scope, or a deterministic contract, and pair it with the permitted route. Prefer a checkable bound over vague exhortations. - -## Trace direct composition - -For each direct caller and callee, trace the trigger, inputs, authority, result, responsibility owner, and stopping boundary. Compare the real obligations rather than repeated wording. Remove gaps, contradictions, duplicated responsibility, and behaviour neither contract owns. Use the least restrictive check that still exposes an incompatible connection. diff --git a/skills/write-skill/references/SECTIONS.md b/skills/write-skill/references/SECTIONS.md index bd91528..aa9f854 100644 --- a/skills/write-skill/references/SECTIONS.md +++ b/skills/write-skill/references/SECTIONS.md @@ -1,51 +1,21 @@ # Skill Sections -Use this reference when invocation metadata, optional sections, or bundled resource placement needs more detail than the fixed skill spine. +Use a section only when it changes execution, constrains the finished result, or transfers responsibility. -## Frontmatter +| Order | Section | Use when | Contains | +| --- | --- | --- | --- | +| 1 | `# Skill Name` | Always | Human-readable skill name. | +| 2 | Introduction | Always | One sentence stating what the skill enables. | +| 3 | `## Prerequisites` | Execution must check conditions before starting | Required state and the action taken when it is absent. | +| 4 | `## Process` | Always | The behaviour that produces the result. | +| 5 | `## Rules` | One invariant governs several steps or the finished result | Cross-cutting constraints stated once. | +| 6 | `## Handoff` | An artefact, decision, or responsibility meaningfully transfers | Recipient, transferred result, evidence, and remaining responsibility. | -Include the skill's discovery and invocation contract: +Add another H2 only for a distinct behavioural concern that does not fit this spine. Never create an empty or decorative section. -- `name` matches its directory, uses lowercase letters, digits, and single hyphens, and reads naturally when invoked. -- `description` is one action-oriented line containing the capability and natural trigger conditions. -- `metadata.invocation` records `user` or `model`. -- `disable-model-invocation` and `agents/openai.yaml` use the matching policy. +Within `## Process`: -Use user invocation by default. Use model invocation only when autonomous discovery would naturally help during ordinary coding work often enough to earn the permanent description context. - -| Invocation | `disable-model-invocation` | `policy.allow_implicit_invocation` | -| ---------- | -------------------------- | ---------------------------------- | -| `user` | `true` | `false` | -| `model` | `false` | `true` | - -Add `agents/openai.yaml` with a human-readable `interface.display_name`, a 25–64-character `interface.short_description`, and the matching policy. - -## Introduction - -The fixed spine and emphasis rules in the main workflow are authoritative. State the skill's purpose concisely. When the user selected a governing methodology, explain only its context-specific adaptation; a methodology-free skill needs no substitute label. - -## Process details - -Within `## Process`, use numbered H3 headings when order matters, descriptive H3 headings for distinct non-sequential concerns, or direct prose for a truly thin process. For branches, use descriptive H3 headings and numbered H4 steps only when the nested sequence improves execution. Each section carries one coherent behavioural idea; each ordered step ends in an observable postcondition. - -## Optional sections - -Add an H2 when it communicates the content more clearly than placement beside the process instruction it governs: - -- `## Prerequisites` states external conditions and the safe route when absent. -- `## Rules` holds invariants that constrain multiple instructions or the finished result. -- `## Handoff` states a meaningful transfer, its evidence, and unresolved uncertainty. - -Rename or combine optional sections when that improves the confirmed behaviour. Every optional section earns its place through clearer execution. - -## Bundled resources - -Use progressive disclosure through the Agent Skills loading hierarchy: - -- Put discovery conditions in frontmatter metadata. -- Keep instructions required on every invocation in `SKILL.md`. -- Put conditional or extensive runtime guidance in `references/`. -- Put files consumed or copied into generated output in `assets/`. -- Put deterministic, repeated, or fragile operations in `scripts/`. - -Add a resource only when retained conditional depth improves execution. Link every resource directly from `SKILL.md` beside the condition that loads it. Keep references one level deep and each meaning in one authoritative location. References contain runtime guidance rather than general concept explanations. +- Use an ordered list or numbered H3s when order changes the result. +- Use H3s when actions can be tested, reordered, or completed independently. +- Keep supporting detail with its action when separation would hide meaning or create a hollow heading. +- Use bullets for independent items, tables for fixed mappings or comparisons, blockquotes for exact authority, and code fences for literal content. diff --git a/skills/write-skill/references/TESTING.md b/skills/write-skill/references/TESTING.md deleted file mode 100644 index 123370d..0000000 --- a/skills/write-skill/references/TESTING.md +++ /dev/null @@ -1,33 +0,0 @@ -# Fresh-Agent Behavioural Evaluation - -Use this reference when scenario selection, context isolation, result variability, or observable pass evidence needs more detail than the common successful path. - -## Preserve the evaluation boundary - -Give the fresh agent the finished bundle, a realistic user request, and only task-local evidence. Keep the intended answer, design rationale, suspected failure, prior output, and hidden hints outside the evaluation context. Use an inert workspace or read-only artefacts; permit safe local writes only when the scenario requires them. An evaluation demonstrates transferable steering when the bundle and supplied evidence account for the process and result. - -## Select scenarios - -Run the common successful path for every created or rewritten skill. Add the smallest scenario for a retained material exception when it distinguishes: - -- an evidenced invocation that changes the required process or result; -- a necessary safety, permission, or prerequisite boundary; -- an optional reference needed by the retained exception; -- a fragile script or deterministic output contract; or -- wording whose effect depends on a selected concept or precise instruction. - -Prefer one scenario that distinguishes several competing behaviours when failure remains diagnosable. The scenario set contains only confirmed behaviour or material risk rather than quotas or hypothetical combinations. - -## Define evidence - -Translate the confirmed contract into observable pass conditions before reading the result. Inspect the agent's actions, resource reads, decisions, boundaries, result, and postconditions. - -For variable work, judge whether the agent followed the confirmed process, used the correct authorities and resources, respected user decisions and stopping boundaries, and produced a correct result suitable for the supplied evidence. Do not require one hidden solution, route, or phrasing. - -For fixed assets, schemas, scripts, formats, or deterministic behaviour, inspect the exact properties promised by the contract. Pass only when the degree of scrutiny matches the claim the skill makes. - -## Repair the smallest cause - -Trace each failure to the smallest instruction, pointer, section boundary, or missing resource that explains it. Repair that cause, then rerun the failed scenario and a common-path scenario. Recompress and mechanically revalidate after every material change. - -Stop when every confirmed scenario passes or reducing the remaining variance would change the confirmed contract. Report unresolved variance plainly. diff --git a/skills/write-skill/scripts/measure-context.js b/skills/write-skill/scripts/measure-context.js new file mode 100644 index 0000000..77e2f8c --- /dev/null +++ b/skills/write-skill/scripts/measure-context.js @@ -0,0 +1,347 @@ +#!/usr/bin/env bun + +import fs from 'node:fs'; +import path from 'node:path'; +import process from 'node:process'; +import { pathToFileURL } from 'node:url'; + +const args = process.argv.slice(2); +const target = args.shift(); + +if (!target) { + console.error( + 'Usage: bun scripts/measure-context.js [--include relative/path ...] [--output file ...] [--json]', + ); + process.exit(2); +} + +const includes = []; +const outputs = []; +let json = false; + +while (args.length) { + const option = args.shift(); + if (option === '--include') includes.push(args.shift()); + else if (option === '--output') outputs.push(args.shift()); + else if (option === '--json') json = true; + else { + console.error(`Unknown option: ${option}`); + process.exit(2); + } +} + +if ([...includes, ...outputs].some((value) => !value)) { + console.error('--include and --output require a path.'); + process.exit(2); +} + +const root = path.resolve(target); +const rootReal = fs.existsSync(root) ? fs.realpathSync(root) : root; +const skillPath = path.join(root, 'SKILL.md'); +if (!fs.existsSync(skillPath) || !fs.statSync(skillPath).isFile()) { + console.error(`Missing SKILL.md: ${skillPath}`); + process.exit(1); +} +if (!isWithinRoot(skillPath, true)) { + console.error(`SKILL.md resolves outside the skill bundle: ${skillPath}`); + process.exit(1); +} + +function measureText(label, filePath, text) { + return { + label, + path: filePath, + kind: 'text', + tokens: encoder.encode(text).length, + bytes: Buffer.byteLength(text, 'utf8'), + characters: [...text].length, + lines: + text === '' + ? 0 + : text.split(/\r?\n/).length - (/\r?\n$/.test(text) ? 1 : 0), + }; +} + +function isWithinRoot(candidate, resolveSymlinks = false) { + const comparable = resolveSymlinks ? fs.realpathSync(candidate) : candidate; + const base = resolveSymlinks ? rootReal : root; + return comparable === base || comparable.startsWith(`${base}${path.sep}`); +} + +function isInside(base, candidate) { + const relativePath = path.relative(base, candidate); + return relativePath !== '' && !relativePath.startsWith('..'); +} + +function nearestAncestor(startDirectory, matches) { + let directory = path.resolve(startDirectory); + + while (true) { + if (matches(directory)) return directory; + const parent = path.dirname(directory); + if (parent === directory) return null; + directory = parent; + } +} + +function propulsionRoot() { + return nearestAncestor(import.meta.dir, (directory) => { + const manifestPath = path.join(directory, 'package.json'); + if (!fs.existsSync(manifestPath)) return false; + const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8')); + return manifest.name === 'propulsion'; + }); +} + +function executionProjectRoot() { + const gitRoot = nearestAncestor(process.cwd(), (directory) => + fs.existsSync(path.join(directory, '.git')), + ); + if (gitRoot) return gitRoot; + + return nearestAncestor(process.cwd(), (directory) => + fs.existsSync(path.join(directory, 'package.json')), + ); +} + +function installedTokenizerEntry() { + const searchRoots = [propulsionRoot(), executionProjectRoot()].filter( + Boolean, + ); + const visited = new Set(); + const foundVersions = new Set(); + + for (const searchRoot of searchRoots) { + if (visited.has(searchRoot)) continue; + visited.add(searchRoot); + const packageRoot = path.join(searchRoot, 'node_modules', 'tiktoken'); + const manifestPath = path.join(packageRoot, 'package.json'); + if (!fs.existsSync(manifestPath)) continue; + + const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8')); + if (manifest.version !== '1.0.22') { + foundVersions.add(manifest.version ?? 'an unversioned package'); + continue; + } + + const canonicalRoot = fs.realpathSync(packageRoot); + const entry = fs.realpathSync( + path.join(canonicalRoot, manifest.main ?? 'tiktoken.cjs'), + ); + + if (!isInside(canonicalRoot, entry) || !fs.statSync(entry).isFile()) { + throw new Error('Refusing an invalid installed tiktoken entry.'); + } + + return pathToFileURL(entry).href; + } + + const found = foundVersions.size + ? `; found ${[...foundVersions].join(', ')}` + : ''; + throw new Error( + `Exact context measurement requires tiktoken@1.0.22 installed in Propulsion or the execution project${found}.`, + ); +} + +function readArtifact(absolutePath) { + const buffer = fs.readFileSync(absolutePath); + const binaryExtensions = new Set([ + '.7z', + '.avi', + '.avif', + '.bin', + '.docx', + '.gif', + '.gz', + '.ico', + '.jpeg', + '.jpg', + '.m4a', + '.mov', + '.mp3', + '.mp4', + '.otf', + '.pdf', + '.png', + '.pptx', + '.tar', + '.tgz', + '.ttf', + '.wasm', + '.webp', + '.woff', + '.woff2', + '.xlsx', + '.zip', + ]); + if (binaryExtensions.has(path.extname(absolutePath).toLowerCase())) + return { kind: 'binary', buffer }; + if (buffer.includes(0)) return { kind: 'binary', buffer }; + try { + return { + kind: 'text', + text: new TextDecoder('utf-8', { fatal: true }).decode(buffer), + }; + } catch { + return { kind: 'binary', buffer }; + } +} + +function measureFile(label, absolutePath, displayPath = absolutePath) { + const artifact = readArtifact(absolutePath); + if (artifact.kind === 'binary') { + return { + label, + path: displayPath, + kind: 'binary', + tokens: null, + bytes: artifact.buffer.length, + }; + } + return measureText(label, displayPath, artifact.text); +} + +function readPath(value, base = root, requireInsideRoot = false) { + const absolutePath = path.resolve(base, value); + if (requireInsideRoot && !isWithinRoot(absolutePath)) { + throw new Error( + `Selected-path file is outside the skill bundle: ${absolutePath}`, + ); + } + if (!fs.existsSync(absolutePath) || !fs.statSync(absolutePath).isFile()) { + throw new Error(`File not found: ${absolutePath}`); + } + if (requireInsideRoot && !isWithinRoot(absolutePath, true)) { + throw new Error( + `Selected-path file resolves outside the skill bundle: ${absolutePath}`, + ); + } + return { absolutePath, artifact: readArtifact(absolutePath) }; +} + +function walkFiles(directory) { + const files = []; + for (const entry of fs.readdirSync(directory, { withFileTypes: true })) { + if (['.git', 'node_modules'].includes(entry.name)) continue; + const absolutePath = path.join(directory, entry.name); + if (entry.isDirectory()) files.push(...walkFiles(absolutePath)); + if (entry.isFile()) files.push(absolutePath); + } + return files; +} + +let encoder; + +try { + const tokenizerEntry = installedTokenizerEntry(); + const { get_encoding: getEncoding } = await import(tokenizerEntry); + encoder = getEncoding('o200k_base'); + const skillText = fs.readFileSync(skillPath, 'utf8'); + const frontmatterMatch = skillText.match( + /^---\r?\n([\s\S]*?)\r?\n---(?:\r?\n|$)/, + ); + if (!frontmatterMatch) + throw new Error('SKILL.md has no valid frontmatter block.'); + const frontmatter = Bun.YAML.parse(frontmatterMatch[1]); + const discoveryText = `${frontmatter.name}\n${frontmatter.description}`; + const skillBody = skillText.slice(frontmatterMatch[0].length); + + const selectedFiles = [ + measureText('SKILL.md body', 'SKILL.md', skillBody), + ...includes.map((relativePath) => { + const { absolutePath, artifact } = readPath( + relativePath, + root, + true, + ); + if (artifact.kind === 'binary') { + throw new Error( + `Selected-path file is binary and cannot be tokenized: ${relativePath}`, + ); + } + return measureText( + relativePath, + path.relative(root, absolutePath), + artifact.text, + ); + }), + ]; + + const outputFiles = outputs.map((filePath) => { + const { absolutePath } = readPath(filePath, process.cwd()); + return measureFile(path.basename(absolutePath), absolutePath); + }); + + const bundleFiles = walkFiles(root).map((absolutePath) => + measureFile( + path.relative(root, absolutePath), + absolutePath, + path.relative(root, absolutePath), + ), + ); + + const result = { + encoding: 'o200k_base', + scope: 'Exact o200k_base counts for text artefacts; binary artefacts, host wrappers, and hidden instructions are excluded.', + discovery: measureText( + 'name + description', + 'SKILL.md frontmatter', + discoveryText, + ), + selected_path: { + files: selectedFiles, + tokens: selectedFiles.reduce((sum, item) => sum + item.tokens, 0), + }, + generated_output: { + files: outputFiles, + tokens: outputFiles.reduce( + (sum, item) => sum + (item.tokens ?? 0), + 0, + ), + }, + bundle_inventory: { + files: bundleFiles, + tokens: bundleFiles.reduce( + (sum, item) => sum + (item.tokens ?? 0), + 0, + ), + binary_files: bundleFiles.filter((item) => item.kind === 'binary') + .length, + }, + }; + + if (json) { + console.log(JSON.stringify(result, null, 2)); + } else { + console.log(`Encoding: ${result.encoding}`); + console.log(`Scope: ${result.scope}`); + console.log(`Discovery metadata: ${result.discovery.tokens} tokens`); + console.log(`Selected path: ${result.selected_path.tokens} tokens`); + for (const file of result.selected_path.files) + console.log(` ${file.path}: ${file.tokens}`); + if (outputFiles.length) { + console.log( + `Generated output: ${result.generated_output.tokens} tokens`, + ); + for (const file of outputFiles) { + console.log( + ` ${file.path}: ${file.tokens ?? `binary, ${file.bytes} bytes`}`, + ); + } + } + console.log( + `Bundle inventory: ${result.bundle_inventory.tokens} tokens`, + ); + if (result.bundle_inventory.binary_files) { + console.log( + `Binary artefacts excluded: ${result.bundle_inventory.binary_files}`, + ); + } + } +} catch (error) { + console.error(`Measurement failed: ${error.message}`); + process.exitCode = 1; +} finally { + encoder?.free(); +} diff --git a/skills/write-skill/scripts/validate-skill.js b/skills/write-skill/scripts/validate-skill.js index 0878316..418b06b 100644 --- a/skills/write-skill/scripts/validate-skill.js +++ b/skills/write-skill/scripts/validate-skill.js @@ -1,328 +1,578 @@ #!/usr/bin/env bun -import { existsSync, readFileSync, readdirSync, statSync } from 'node:fs'; -import { basename, join, relative, resolve, sep } from 'node:path'; - -const errors = []; -const warnings = []; - -function addError(message) { - errors.push(message); +import fs from 'node:fs'; +import path from 'node:path'; +import process from 'node:process'; +import { pathToFileURL } from 'node:url'; + +function nearestAncestor(startDirectory, matches) { + let directory = path.resolve(startDirectory); + + while (true) { + if (matches(directory)) return directory; + const parent = path.dirname(directory); + if (parent === directory) return null; + directory = parent; + } } -function isRecord(value) { - return value !== null && typeof value === 'object' && !Array.isArray(value); +function propulsionRoot() { + return nearestAncestor(import.meta.dir, (directory) => { + const manifestPath = path.join(directory, 'package.json'); + if (!fs.existsSync(manifestPath)) return false; + const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8')); + return manifest.name === 'propulsion'; + }); } -function parseYaml(raw, label) { - try { - const value = Bun.YAML.parse(raw); - - if (!isRecord(value)) { - addError(`Make ${label} a YAML mapping.`); - return {}; - } +function executionProjectRoot() { + const gitRoot = nearestAncestor(process.cwd(), (directory) => + fs.existsSync(path.join(directory, '.git')), + ); + if (gitRoot) return gitRoot; - return value; - } catch (error) { - addError(`Parse ${label} as valid YAML: ${error.message}`); - return {}; - } + return nearestAncestor(process.cwd(), (directory) => + fs.existsSync(path.join(directory, 'package.json')), + ); } -function parseFrontmatter(content) { - const match = content.match(/^---\r?\n([\s\S]*?)\r?\n---\r?\n?/); +function installedPackageEntry(packageName, expectedVersion) { + const searchRoots = [propulsionRoot(), executionProjectRoot()].filter( + Boolean, + ); + const visited = new Set(); + const foundVersions = new Set(); + + for (const searchRoot of searchRoots) { + if (visited.has(searchRoot)) continue; + visited.add(searchRoot); + const packageRoot = path.join(searchRoot, 'node_modules', packageName); + const manifestPath = path.join(packageRoot, 'package.json'); + if (!fs.existsSync(manifestPath)) continue; + + const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8')); + if (manifest.version !== expectedVersion) { + foundVersions.add(manifest.version ?? 'an unversioned package'); + continue; + } - if (!match) { - addError('Add YAML frontmatter at the start of SKILL.md.'); - return { body: content, data: {}, raw: '' }; + const canonicalRoot = fs.realpathSync(packageRoot); + const entry = fs.realpathSync( + path.join(canonicalRoot, manifest.main ?? 'index.js'), + ); + const relativeEntry = path.relative(canonicalRoot, entry); + + if ( + relativeEntry === '' || + relativeEntry.startsWith('..') || + !fs.statSync(entry).isFile() + ) { + throw new Error( + `Refusing an invalid installed ${packageName} entry.`, + ); + } + + return pathToFileURL(entry).href; } - return { - body: content.slice(match[0].length), - data: parseYaml(match[1], 'SKILL.md frontmatter'), - raw: match[1], - }; + const found = foundVersions.size + ? `; found ${[...foundVersions].join(', ')}` + : ''; + throw new Error( + `Validation requires ${packageName}@${expectedVersion} installed in Propulsion or the execution project${found}.`, + ); } -function validateName(frontmatter, skillPath) { - const { name } = frontmatter; +const semverEntry = installedPackageEntry('semver', '7.8.5'); +const { satisfies, validRange } = await import(semverEntry); - if (typeof name !== 'string' || !name) { - addError('Add the skill name to frontmatter.'); - return; - } +const target = process.argv[2]; - if (!/^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(name)) { - addError( - `Use lowercase letters, digits, and single hyphens for name: ${name}`, - ); - } - - if (name.length > 64) { - addError(`Keep name at most 64 characters; found ${name.length}.`); - } - - if (name !== basename(skillPath)) { - addError(`Match name "${name}" to directory "${basename(skillPath)}".`); - } +if (!target) { + console.error('Usage: bun scripts/validate-skill.js '); + process.exit(2); } -function validateDescription(frontmatter, raw) { - const { description } = frontmatter; - const descriptionLine = raw - .split(/\r?\n/) - .find((line) => line.startsWith('description:')); +const root = path.resolve(target); +const rootReal = fs.existsSync(root) ? fs.realpathSync(root) : root; +const errors = []; - if (typeof description !== 'string' || !description) { - addError('Add a one-line description to frontmatter.'); - return; - } +function fail(message) { + errors.push(message); +} - if (descriptionLine?.match(/^description:\s*[|>]/)) { - addError('Write description on one YAML line.'); +function readRequired(relativePath) { + const absolutePath = path.join(root, relativePath); + if (!fs.existsSync(absolutePath) || !fs.statSync(absolutePath).isFile()) { + fail(`Missing required file: ${relativePath}`); + return null; } - - if (description.length > 200) { - addError( - `Keep description at most 200 characters; found ${description.length}.`, - ); - } - - if (!/\bUse (?:when|for|to)\b/.test(description)) { - addError( - 'State invocation conditions with Use when, Use for, or Use to.', + if (!isWithinRoot(absolutePath, true)) { + fail( + `Required file resolves outside the skill bundle: ${relativePath}`, ); + return null; } + return fs.readFileSync(absolutePath, 'utf8'); +} - if (/\b(?:I|me|my|mine|we|us|our|ours)\b/i.test(description)) { - addError( - 'Write description in third-person, action-oriented language.', - ); +function parseYaml(source, relativePath) { + if (source === null) return null; + try { + const value = Bun.YAML.parse(source); + if (!value || typeof value !== 'object' || Array.isArray(value)) { + fail(`${relativePath} must contain a YAML mapping.`); + return null; + } + return value; + } catch (error) { + fail(`${relativePath} contains invalid YAML: ${error.message}`); + return null; } } -function validateInvocation(frontmatter, openai) { - const invocation = isRecord(frontmatter.metadata) - ? frontmatter.metadata.invocation - : null; - const disableModel = frontmatter['disable-model-invocation']; - const allowImplicit = isRecord(openai.policy) - ? openai.policy.allow_implicit_invocation - : null; - - if (!['user', 'model'].includes(invocation)) { - addError('Set metadata.invocation to user or model.'); - return; - } +function analyseMarkdown(source) { + const lines = source.split(/\r?\n/); + let fence = null; + return lines.map((line) => { + const opening = line.match(/^ {0,3}(`{3,}|~{3,})(.*)$/); + if (!fence && opening) { + fence = { character: opening[1][0], length: opening[1].length }; + return { raw: line, structural: null, content: false }; + } + const closing = line.match(/^ {0,3}(`{3,}|~{3,})[ \t]*$/); + if ( + fence && + closing && + closing[1][0] === fence.character && + closing[1].length >= fence.length + ) { + fence = null; + return { raw: line, structural: null, content: false }; + } + if (fence) + return { raw: line, structural: null, content: line.trim() !== '' }; + return { + raw: line, + structural: line, + content: line.trim() !== '' && !/^#{1,6}\s/.test(line), + }; + }); +} - const expectedDisable = invocation === 'user'; - const expectedImplicit = invocation === 'model'; +function isWithinRoot(candidate, resolveSymlinks = false) { + const comparable = resolveSymlinks ? fs.realpathSync(candidate) : candidate; + const base = resolveSymlinks ? rootReal : root; + return comparable === base || comparable.startsWith(`${base}${path.sep}`); +} - if (disableModel !== expectedDisable) { - addError( - `Set disable-model-invocation to ${expectedDisable} for ${invocation} invocation.`, - ); - } +function stripNonRenderedText(source) { + const withoutComments = source.replace(/|$)/g, ''); + return withoutComments + .split(/\r?\n/) + .map((line) => { + let result = line; + const runs = [...result.matchAll(/`+/g)]; + for (let index = 0; index + 1 < runs.length; index += 2) { + const opening = runs[index]; + const closing = runs + .slice(index + 1) + .find( + (candidate) => + candidate[0].length === opening[0].length, + ); + if (!closing) break; + const start = opening.index; + const end = closing.index + closing[0].length; + result = `${result.slice(0, start)}${' '.repeat(end - start)}${result.slice(end)}`; + } + return result; + }) + .join('\n'); +} - if (allowImplicit !== expectedImplicit) { - addError( - `Set policy.allow_implicit_invocation to ${expectedImplicit} for ${invocation} invocation.`, - ); +function markdownLinkTargets(line) { + const targets = []; + for (let index = 0; index < line.length; index += 1) { + if (line[index] !== '[') continue; + let backslashes = 0; + for ( + let cursor = index - 1; + cursor >= 0 && line[cursor] === '\\'; + cursor -= 1 + ) + backslashes += 1; + if (backslashes % 2 === 1) continue; + const labelEnd = line.indexOf('](', index + 1); + if (labelEnd < 0) continue; + let depth = 1; + let cursor = labelEnd + 2; + let escaped = false; + for (; cursor < line.length; cursor += 1) { + const character = line[cursor]; + if (escaped) { + escaped = false; + continue; + } + if (character === '\\') { + escaped = true; + continue; + } + if (character === '(') depth += 1; + if (character === ')') depth -= 1; + if (depth === 0) break; + } + if (depth !== 0) continue; + const inside = line.slice(labelEnd + 2, cursor).trim(); + const destination = inside.startsWith('<') + ? inside.slice(1, inside.indexOf('>')) + : inside.match(/^(?:\\.|[^\s])+/)?.[0]; + if (destination) targets.push(destination.replace(/\\([ ()])/g, '$1')); + index = cursor; } + return targets; } -function validateOpenaiYaml(skillPath, frontmatter) { - const openaiPath = join(skillPath, 'agents', 'openai.yaml'); +function equalStringMaps(left, right) { + const leftKeys = Object.keys(left).toSorted(); + const rightKeys = Object.keys(right).toSorted(); + return ( + leftKeys.length === rightKeys.length && + leftKeys.every( + (key, index) => + key === rightKeys[index] && left[key] === right[key], + ) + ); +} - if (!existsSync(openaiPath)) { - addError('Add agents/openai.yaml.'); - return ''; +function walkFiles(directory) { + const files = []; + for (const entry of fs.readdirSync(directory, { withFileTypes: true })) { + if (['.git', 'node_modules'].includes(entry.name)) continue; + const absolutePath = path.join(directory, entry.name); + if (entry.isDirectory()) files.push(...walkFiles(absolutePath)); + if (entry.isFile()) files.push(absolutePath); } + return files; +} - const openaiRaw = readFileSync(openaiPath, 'utf8'); - const openai = parseYaml(openaiRaw, 'agents/openai.yaml'); - const skillInterface = isRecord(openai.interface) ? openai.interface : {}; - const displayName = skillInterface.display_name; - const shortDescription = skillInterface.short_description; +const skillSource = readRequired('SKILL.md'); +const openaiSource = readRequired('agents/openai.yaml'); - if (typeof displayName !== 'string' || !displayName) - addError('Set interface.display_name in agents/openai.yaml.'); +let frontmatter = null; +let body = ''; - if (typeof shortDescription !== 'string' || !shortDescription) { - addError('Set interface.short_description in agents/openai.yaml.'); - } else if (shortDescription.length < 25 || shortDescription.length > 64) { - addError( - `Keep interface.short_description between 25 and 64 characters; found ${shortDescription.length}.`, - ); +if (skillSource !== null) { + const match = skillSource.match(/^---\r?\n([\s\S]*?)\r?\n---(?:\r?\n|$)/); + if (!match) { + fail('SKILL.md must start with YAML frontmatter delimited by ---.'); + } else { + frontmatter = parseYaml(match[1], 'SKILL.md frontmatter'); + body = skillSource.slice(match[0].length); + + const descriptionLine = match[1] + .split(/\r?\n/) + .find((line) => /^description\s*:/.test(line)); + if ( + !descriptionLine || + /^description\s*:\s*[>|]/.test(descriptionLine) + ) { + fail('description must be a single-line YAML scalar.'); + } } - - validateInvocation(frontmatter, openai); - return openai; } -function collectMarkdownHeadings(lines) { - const headings = []; - let fence = null; - - for (const line of lines) { - const trimmed = line.trim(); - const fenceMatch = trimmed.match(/^(`{3,}|~{3,})/); - - if (fenceMatch) { - const marker = fenceMatch[1]; - - if (fence === null) { - fence = marker; - } else if ( - marker[0] === fence[0] && - marker.length >= fence.length - ) { - fence = null; - } +if (frontmatter) { + const { name, description } = frontmatter; + if (typeof name !== 'string' || !/^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(name)) { + fail( + 'name must use lowercase letters or digits with single hyphen separators.', + ); + } else { + if (name.length > 64) fail('name must be 1-64 characters.'); + if (name !== path.basename(root)) + fail('name must match the parent directory name.'); + } - continue; + if (typeof description !== 'string' || description.trim() === '') { + fail('description must be a non-empty string.'); + } else { + if (!/^[A-Z][A-Za-z-]*s\b/.test(description)) { + fail('description must start with a third-person action verb.'); } - - if (fence === null && /^#{1,6}\s+\S/.test(trimmed)) { - headings.push(trimmed); + if (!/\bUse (when|for|to)\b/.test(description)) { + fail( + "description must include 'Use when', 'Use for', or 'Use to'.", + ); } } - return headings; + if (typeof frontmatter['disable-model-invocation'] !== 'boolean') { + fail('disable-model-invocation must be a boolean.'); + } } -function validateBody(body) { - const lines = body.split(/\r?\n/); - const headings = collectMarkdownHeadings(lines); - const firstContentIndex = lines.findIndex((line) => line.trim()); - const firstContent = lines[firstContentIndex]?.trim(); - - if (!firstContent?.match(/^#\s+\S/)) { - addError('Start the skill body with a human-readable H1.'); - return; +const openai = parseYaml(openaiSource, 'agents/openai.yaml'); +if (openai) { + if ( + typeof openai.interface?.display_name !== 'string' || + openai.interface.display_name.trim() === '' + ) { + fail('agents/openai.yaml requires interface.display_name.'); } - - const h1Headings = headings.filter((line) => /^#\s+\S/.test(line)); - - if (h1Headings.length !== 1) { - addError('Add exactly one H1 heading to the skill body.'); + const shortDescription = openai.interface?.short_description; + if ( + typeof shortDescription !== 'string' || + shortDescription.length < 25 || + shortDescription.length > 64 + ) { + fail('interface.short_description must be 25-64 characters.'); } - - const firstH2Index = lines.findIndex((line) => - /^##\s+\S/.test(line.trim()), - ); - const introEnd = firstH2Index === -1 ? lines.length : firstH2Index; - const introduction = lines - .slice(firstContentIndex + 1, introEnd) - .map((line) => line.trim()) - .find((line) => line && !line.startsWith('#')); - - if (!introduction) { - addError('Follow the H1 with a concise introductory paragraph.'); + if (typeof openai.policy?.allow_implicit_invocation !== 'boolean') { + fail( + 'agents/openai.yaml requires boolean policy.allow_implicit_invocation.', + ); } - - const processHeadings = headings.filter((line) => line === '## Process'); - - if (processHeadings.length !== 1) { - addError('Add exactly one ## Process heading to the skill body.'); + if ( + frontmatter && + typeof frontmatter['disable-model-invocation'] === 'boolean' && + typeof openai.policy?.allow_implicit_invocation === 'boolean' && + frontmatter['disable-model-invocation'] === + openai.policy.allow_implicit_invocation + ) { + fail( + 'Claude and OpenAI invocation settings must describe the same invocation mode.', + ); } } -function collectFiles(directory, prefix) { - if (!existsSync(directory)) return []; +if (fs.existsSync(path.join(root, 'agents/openai.yml'))) { + fail('Use agents/openai.yaml, not agents/openai.yml.'); +} - return readdirSync(directory, { withFileTypes: true }).flatMap((entry) => { - const absolute = join(directory, entry.name); - const relativePath = `${prefix}/${entry.name}`; +if (skillSource !== null) { + if (body.trim() === '') + fail('SKILL.md must contain a non-empty body after frontmatter.'); - return entry.isDirectory() - ? collectFiles(absolute, relativePath) - : [relativePath]; + const lines = analyseMarkdown(body); + const headings = []; + lines.forEach(({ structural }, index) => { + const match = structural?.match(/^(#{1,6})\s+(\S.*)$/); + if (match) + headings.push({ + level: match[1].length, + text: match[2].trim(), + line: index, + }); }); -} -function validateResources(skillPath, body) { - const resources = ['references', 'assets', 'scripts'].flatMap((directory) => - collectFiles(join(skillPath, directory), directory), - ); - const linked = new Set(); - const linkPattern = /\[[^\]]*\]\(([^)]+)\)/g; - - for (const match of body.matchAll(linkPattern)) { - const rawTarget = match[1].split('#')[0]; - const target = rawTarget.startsWith('./') - ? rawTarget.slice(2) - : rawTarget; - const normalized = target.split('/').join(sep); + const h1s = headings.filter(({ level }) => level === 1); + if (h1s.length !== 1) + fail( + 'SKILL.md must contain exactly one H1 outside fenced code blocks.', + ); - if (!/^(?:references|assets|scripts)\//.test(target)) continue; + const processHeadings = headings.filter( + ({ level, text }) => level === 2 && text === 'Process', + ); + if (processHeadings.length !== 1) + fail('SKILL.md must contain exactly one ## Process heading.'); + + if (h1s.length === 1) { + const firstContentLine = lines.findIndex(({ raw }) => raw.trim()); + if (firstContentLine !== h1s[0].line) + fail('The H1 must be the first body content.'); + + const firstH2 = headings.find(({ level }) => level === 2); + const introduction = lines + .slice(h1s[0].line + 1, firstH2?.line ?? lines.length) + .filter(({ raw }) => raw.trim()); + if ( + introduction.length !== 1 || + introduction[0].structural === null || + introduction[0].raw.startsWith('#') + ) { + fail( + 'The H1 must be followed by one non-empty introduction line before the first H2.', + ); + } + } - linked.add(target); - if (!existsSync(join(skillPath, normalized))) { - addError(`Create linked resource ${target} or update its pointer.`); + const h2s = headings.filter(({ level }) => level === 2); + for (const reserved of ['Prerequisites', 'Process', 'Rules', 'Handoff']) { + if (h2s.filter(({ text }) => text === reserved).length > 1) { + fail(`## ${reserved} may appear at most once.`); } } - for (const resource of resources) { - if (!linked.has(resource)) { - addError( - `Link ${resource} beside the step or branch that uses it.`, + const sectionIndex = Object.fromEntries( + h2s.map(({ text }, index) => [text, index]), + ); + if ( + sectionIndex.Prerequisites !== undefined && + sectionIndex.Prerequisites > sectionIndex.Process + ) { + fail('## Prerequisites must appear before ## Process.'); + } + if ( + sectionIndex.Rules !== undefined && + sectionIndex.Rules < sectionIndex.Process + ) { + fail('## Rules must appear after ## Process.'); + } + if (sectionIndex.Handoff !== undefined) { + const lowerBound = sectionIndex.Rules ?? sectionIndex.Process; + if (sectionIndex.Handoff < lowerBound) + fail( + '## Handoff must appear after ## Process and ## Rules when present.', ); - } } - return resources.length; + for (let index = 0; index < headings.length; index += 1) { + const heading = headings[index]; + const next = headings + .slice(index + 1) + .find(({ level }) => level <= heading.level); + const content = lines + .slice(heading.line + 1, next?.line ?? lines.length) + .some(({ content: hasContent }) => hasContent); + if (!content) fail(`Heading '${heading.text}' has no content.`); + } } -function validateSkill(skillPath) { - if (!existsSync(skillPath)) { - addError(`Provide an existing skill directory: ${skillPath}`); - return null; +if (fs.existsSync(root)) { + const files = walkFiles(root); + const markdownFiles = files.filter((file) => file.endsWith('.md')); + const linkedFromSkill = new Set(); + + for (const file of markdownFiles) { + const source = fs.readFileSync(file, 'utf8'); + const linkSource = file === path.join(root, 'SKILL.md') ? body : source; + for (const { structural } of analyseMarkdown( + stripNonRenderedText(linkSource), + )) { + if (structural === null) continue; + for (const linkTarget of markdownLinkTargets(structural)) { + const rawTarget = linkTarget.trim().split('#')[0]; + if (!rawTarget || /^(?:[a-z]+:|#)/i.test(rawTarget)) continue; + const resolved = path.resolve( + path.dirname(file), + decodeURIComponent(rawTarget), + ); + if (!isWithinRoot(resolved)) { + fail( + `${path.relative(root, file)} links outside the skill bundle: ${rawTarget}`, + ); + continue; + } + if (!fs.existsSync(resolved)) { + fail( + `${path.relative(root, file)} links to missing path: ${rawTarget}`, + ); + continue; + } + if (!isWithinRoot(resolved, true)) { + fail( + `${path.relative(root, file)} links through a symlink outside the skill bundle: ${rawTarget}`, + ); + continue; + } + if (file === path.join(root, 'SKILL.md')) + linkedFromSkill.add(resolved); + } + } } - if (!statSync(skillPath).isDirectory()) { - addError(`Provide a skill directory: ${skillPath}`); - return null; + for (const file of files.filter((candidate) => + candidate.startsWith(path.join(root, 'references') + path.sep), + )) { + if (!linkedFromSkill.has(file)) { + fail( + `Reference must be linked directly from SKILL.md: ${path.relative(root, file)}`, + ); + } } - const skillFile = join(skillPath, 'SKILL.md'); - if (!existsSync(skillFile)) { - addError('Add SKILL.md to the skill directory.'); - return null; + const packagePath = path.join(root, 'package.json'); + if (fs.existsSync(packagePath)) { + let packageJson = null; + try { + packageJson = JSON.parse(fs.readFileSync(packagePath, 'utf8')); + } catch (error) { + fail(`package.json contains invalid JSON: ${error.message}`); + } + const lockPath = path.join(root, 'package-lock.json'); + if (!fs.existsSync(lockPath)) { + fail('package-lock.json is required when package.json exists.'); + } else { + try { + const lock = JSON.parse(fs.readFileSync(lockPath, 'utf8')); + const lockRoot = lock.packages?.['']; + if (![2, 3].includes(lock.lockfileVersion) || !lockRoot) { + fail( + 'package-lock.json must use lockfileVersion 2 or 3 and contain a root package entry.', + ); + } else if (packageJson) { + if ( + lockRoot.name !== packageJson.name || + lockRoot.version !== packageJson.version + ) { + fail( + 'package-lock.json root name and version must match package.json.', + ); + } + const declared = packageJson.dependencies ?? {}; + const locked = lockRoot.dependencies ?? {}; + if (!equalStringMaps(declared, locked)) { + fail( + 'package-lock.json root dependencies must match package.json.', + ); + } + for (const dependency of Object.keys(declared)) { + const entry = + lock.packages?.[`node_modules/${dependency}`]; + if (!entry?.version) { + fail( + `package-lock.json lacks a complete entry for dependency: ${dependency}`, + ); + continue; + } + const range = validRange(declared[dependency]); + if ( + range && + !satisfies(entry.version, range, { + includePrerelease: true, + }) + ) { + fail( + `package-lock.json version for ${dependency} does not satisfy package.json.`, + ); + } + if ( + entry.resolved?.startsWith('http') && + !/^sha(?:256|384|512)-[A-Za-z0-9+/]+={0,2}$/.test( + entry.integrity ?? '', + ) + ) { + fail( + `package-lock.json has invalid integrity for dependency: ${dependency}`, + ); + } + } + } + } catch (error) { + fail( + `package-lock.json contains invalid JSON: ${error.message}`, + ); + } + } } - - const content = readFileSync(skillFile, 'utf8'); - const { body, data, raw } = parseFrontmatter(content); - - validateName(data, skillPath); - validateDescription(data, raw); - validateOpenaiYaml(skillPath, data); - validateBody(body); - const resourceCount = validateResources(skillPath, body); - - return { - resources: resourceCount, - }; } -const cliArguments = process.argv.slice(2); -let target = null; - -if (cliArguments.length !== 1 || cliArguments[0].startsWith('--')) { - addError('Run bun validate-skill.js .'); -} else { - target = resolve(cliArguments[0]); +if (errors.length) { + console.error('Status: invalid'); + for (const error of errors) console.error(`- ${error}`); + process.exit(1); } -const stats = target ? validateSkill(target) : null; -const result = { - path: target ? relative(process.cwd(), target) || '.' : null, - valid: errors.length === 0, - errors, - warnings, - stats, -}; - -console.log(JSON.stringify(result, null, 2)); -process.exitCode = result.valid ? 0 : 1; +console.log(`Status: valid\nSkill: ${root}`); From 19e6fa2ce8d6d8ab414574c02aec17fb21c845ef Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 08:33:57 +0100 Subject: [PATCH 46/79] Partly complete skills --- SKILLS_PLAN.md | 58 +++++++++++++++++++++++++------------------------- 1 file changed, 29 insertions(+), 29 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 2befbc6..10ce40c 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -391,35 +391,35 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ### Implementation checklist - [x] Complete quality-harness research -- [x] Update `elicit` skill -- [x] Update `research` skill -- [x] Create `prototype` skill -- [x] Update `maintain-context` skill -- [x] Update `modular-design` skill -- [x] Update `tdd` skill -- [x] Create `verify-change` skill -- [x] Update `code-review` skill -- [x] Update `maintain-agents` skill -- [x] Update `define-product` skill -- [x] Create `define-architecture` skill -- [x] Create `specify-feature` skill -- [x] Create `design-feature` skill -- [x] Create `create-tickets` skill -- [x] Update `implement` skill -- [x] Update `commit` skill -- [x] Replace `pr` with `pull-request` -- [x] Create `review-pull-request` skill -- [x] Create `triage-work` skill -- [x] Create `refine-backlog` skill -- [x] Create `plan-iteration` skill -- [x] Update `review-architecture` skill -- [x] Update `debug` skill -- [x] Create `respond-to-incident` skill -- [x] Create `review-incident` skill -- [x] Create `plan-retirement` skill -- [x] Delete `elicit-with-context` skill -- [x] Update the README -- [x] Run the suite-wide acceptance pass +- [-] Update `elicit` skill +- [-] Update `research` skill +- [-] Create `prototype` skill +- [-] Update `maintain-context` skill +- [-] Update `modular-design` skill +- [-] Update `tdd` skill +- [-] Create `verify-change` skill +- [-] Update `code-review` skill +- [-] Update `maintain-agents` skill +- [-] Update `define-product` skill +- [-] Create `define-architecture` skill +- [-] Create `specify-feature` skill +- [-] Create `design-feature` skill +- [-] Create `create-tickets` skill +- [-] Update `implement` skill +- [-] Update `commit` skill +- [-] Replace `pr` with `pull-request` +- [-] Create `review-pull-request` skill +- [-] Create `triage-work` skill +- [-] Create `refine-backlog` skill +- [-] Create `plan-iteration` skill +- [-] Update `review-architecture` skill +- [-] Update `debug` skill +- [-] Create `respond-to-incident` skill +- [-] Create `review-incident` skill +- [-] Create `plan-retirement` skill +- [-] Delete `elicit-with-context` skill +- [-] Update the README +- [-] Run the suite-wide acceptance pass Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. From 06cf2c6d96d731f112700ee806120b758764ab36 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 10:50:55 +0100 Subject: [PATCH 47/79] feat(elicit): strengthen information gathering --- SKILLS_PLAN.md | 2 +- skills/elicit/SKILL.md | 16 ++++++++-------- skills/elicit/agents/openai.yaml | 2 +- 3 files changed, 10 insertions(+), 10 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 10ce40c..90b3aa3 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -391,7 +391,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ### Implementation checklist - [x] Complete quality-harness research -- [-] Update `elicit` skill +- [x] Update `elicit` skill - [-] Update `research` skill - [-] Create `prototype` skill - [-] Update `maintain-context` skill diff --git a/skills/elicit/SKILL.md b/skills/elicit/SKILL.md index 1ec0564..54822b3 100644 --- a/skills/elicit/SKILL.md +++ b/skills/elicit/SKILL.md @@ -1,6 +1,6 @@ --- name: elicit -description: Elicits material user-held information and user-authoritative decisions one question at a time until confirmed. Use when shared understanding depends on the user, not safely discoverable facts. +description: Elicits material information and decisions from the user one question at a time until shared understanding is confirmed. Use when a request or consuming skill depends on user-held context, clarification, or choices that cannot be safely discovered. metadata: invocation: model disable-model-invocation: false @@ -8,30 +8,30 @@ disable-model-invocation: false # Elicit -Uses an internal **decision tree** to resolve only the material branches of shared understanding, then applies adapted **theoretical saturation** and obtains explicit confirmation of one complete synthesis. +Builds confirmed shared understanding by resolving a bounded internal **decision tree**, then applying adapted **theoretical saturation**. ## Process ### 1. Establish the basis -Inspect the request, conversation, caller context, and accessible task-scoped evidence before questioning. Derive discoverable facts and identify the user's intended outcome. Keep inspection read-only. When a required fact could be discovered only through a state change, make whether the caller should perform that lookup the active authority decision and leave the lookup to the caller. The outcome and factual basis are explicit. +Inspect the request, conversation, caller context, and available task-scoped evidence before questioning. Use cheap read-only fact-finding and stable, confidently applicable general knowledge to close discoverable branches. Verify uncertain, contentious, or potentially current knowledge before treating it as fact. Leave material research and every state-changing action to the caller. The intended outcome, caller boundary, and factual basis are explicit. ### 2. Maintain the decision tree -Build and continually recompute an internal, dependency-ordered decision tree covering only material outcome, scope, terminology, inputs, outputs, prerequisites, dependencies, constraints, flows, exceptions, permissions, risks, trade-offs, and success conditions. Open a branch only when runtime evidence or an answer exposes it and resolving it could materially change the shared understanding. Preserve compatible answers as branches open, close, become irrelevant, or reopen. Select the highest-impact unresolved branch whose prerequisites are resolved. Keep the tree internal; one active issue is explicit. +Build and continually recompute an internal, dependency-ordered tree containing only branches whose resolution could materially change the bounded outcome. Cover relevant outcome, scope, terminology, inputs, outputs, prerequisites, dependencies, constraints, flows, exceptions, permissions, risks, trade-offs, and success conditions. Preserve compatible answers; close, discard, or reopen branches as evidence changes; and follow directly consequential issues without expanding into unrelated discovery. Select the highest-impact unresolved branch whose prerequisites are resolved. Keep the tree internal and one active issue explicit. ### 3. Ask one question -Ask exactly one explicit question about the active issue, once per turn, in natural conversational language. Explain why an information question matters. For a decision, present only viable positions and their decisive trade-offs, recommending one only when the outcome, evidence, applicable conventions, consequences, and prior answers provide a defensible basis. Leave room for the user's own answer and treat the user as the sole authority over their information and decisions. The active issue has one explicit response. +Ask exactly one explicit question per turn in natural language. Explain why an information question matters. Offer suggested answers only when grounded in the established basis. Apply **Design It Twice**: present at least two genuinely distinct viable options, state the material effect of each, and recommend the strongest when the evidence supports a clear winner. Make those options part of the one question; do not append another choice prompt. Leave room for another answer and keep the user as the sole authority over their information and decisions. The active issue receives one explicit response. ### 4. Adapt and validate -Test each response internally against the root goal, previous decisions, scenarios, counterexamples, conflicts, and consequences. Keep material ambiguity open. When the response is unclear or conflicting, or the user cannot answer, scaffold the same issue through plain-language restatement, clearer choices and trade-offs, examples, scenarios, or prerequisite questions resolved one at a time. Do not offer an escape hatch or decide for the user. Recombine the answers and reopen every affected branch. The active issue is resolved consistently. +Test every response internally against the outcome, established facts, prior answers, implications, contradictions, scenarios, and counterexamples. Surface a challenge only when it could materially change the shared understanding. Close clear answers without ceremonial repetition. For ambiguity, conflict, difficulty, or consequential interpretation, use plain-English paraphrasing, rewording, clearer options and effects, examples, scenarios, or prerequisite questions resolved one at a time. Do not invent an answer or decide for the user. Recompute the tree and reopen every affected branch. The active issue is resolved consistently. ### 5. Reach theoretical saturation -Repeat steps 2–4 until every material branch is resolved. Then make an adapted theoretical-saturation pass across the tree, scenarios, counterexamples, conflicts, and consequences. Saturation is reached only when the pass adds or reopens no branch whose answer could materially change the shared understanding. The decision tree is saturated. +Repeat steps 2–4 until every material branch is resolved. Make an adapted theoretical-saturation pass across the tree, implications, scenarios, counterexamples, and conflicts. Walk each material decision through affected actors and authority, defaults and unchanged behaviour, states and transitions, time boundaries, channels, visibility, exceptions, invalidation, and success evidence; open only consequential gaps. Saturation is reached only when the pass adds or reopens no branch whose answer could materially change the shared understanding. The bounded decision tree is saturated. ### 6. Confirm shared understanding -Present one concise, self-contained synthesis of the outcome, relevant facts, decisions, scope, constraints, and observable success conditions, then ask exactly one question for explicit agreement. Only a new explicit user response confirming the complete synthesis completes elicitation; selecting an earlier option does not. Treat a correction or rejection as new evidence, reopen every affected branch, and continue from step 2 until saturation returns before presenting the revised synthesis. Return only the confirmed synthesis; the caller retains mutation, persistence, verification, downstream action, and stopping ownership. +Present one concise, self-contained synthesis of the relevant facts, information, decisions, scope, constraints, implications, and observable success conditions. Ask exactly one question for explicit agreement. Only a new response confirming the complete synthesis ends elicitation; selecting an earlier option does not. Treat a correction or rejection as new evidence, reopen every affected branch, and repeat the saturation pass before presenting a revised synthesis. Return only the confirmed synthesis; create no durable artefact and leave mutation, persistence, verification, downstream action, and the caller's stopping boundary to the caller. diff --git a/skills/elicit/agents/openai.yaml b/skills/elicit/agents/openai.yaml index e687258..24235cb 100644 --- a/skills/elicit/agents/openai.yaml +++ b/skills/elicit/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Elicit' - short_description: 'Resolve material information and decisions' + short_description: 'Elicit material information and decisions' policy: allow_implicit_invocation: true From d6247f393ab6bc692b24787e30b84d1b11ff8fd3 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 13:16:55 +0100 Subject: [PATCH 48/79] feat(skills): delegate commit and pull request workflows --- AGENTS.md | 1 + SKILLS_PLAN.md | 4 +-- skills/commit/SKILL.md | 24 +++--------------- skills/commit/agents/openai.yaml | 2 +- skills/pull-request/SKILL.md | 34 ++++---------------------- skills/pull-request/agents/openai.yaml | 2 +- 6 files changed, 14 insertions(+), 53 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index 38f53f3..b3d2b1e 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,3 +1,4 @@ - When the user corrects you with a reusable, global rule, ask if they want it added to `AGENTS.md`. - When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. +- Assume the execution harness supports fresh agents; do not add fallback handling for their unavailability. - After implementing changes run `bun run checks` before handoff. diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 90b3aa3..5018c58 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -406,8 +406,8 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Create `design-feature` skill - [-] Create `create-tickets` skill - [-] Update `implement` skill -- [-] Update `commit` skill -- [-] Replace `pr` with `pull-request` +- [x] Update `commit` skill +- [x] Replace `pr` with `pull-request` - [-] Create `review-pull-request` skill - [-] Create `triage-work` skill - [-] Create `refine-backlog` skill diff --git a/skills/commit/SKILL.md b/skills/commit/SKILL.md index bef73c3..36f7b78 100644 --- a/skills/commit/SKILL.md +++ b/skills/commit/SKILL.md @@ -1,6 +1,6 @@ --- name: commit -description: Records eligible reviewed local work in coherent Conventional Commits while preserving unrelated changes. Use when the user asks to commit completed work. +description: Creates atomic Conventional Commits for eligible reviewed work while preserving unrelated changes. Use when the user asks to commit completed work. metadata: invocation: user disable-model-invocation: true @@ -8,26 +8,10 @@ disable-model-invocation: true # Commit -**Conventional Commits** records eligible local work in reviewable history without absorbing or disturbing unrelated changes. +Records eligible local work in coherent commits without disturbing unrelated changes. ## Process -### 1. Establish the eligible scope +Assign the entire process below to one fresh agent with the user's request and repository path; when already inside the fresh agent for an enclosing operation, perform it there. Use only repository evidence to inspect the complete Git state and diffs, group eligible work by intent into atomic **Conventional Commits**, preserve unrelated staged, unstaged, and untracked work, and let commit hooks run. The work is already reviewed and verified; do not run tests, project checks, or implementation verification. Verify only each commit and the final Git status, then return only the hashes, titles, and remaining work. Retain only that result. -Inspect repository instructions, the current branch, recent commit subjects, and the exact staged, unstaged, and untracked state; inspect ignored files when they may explain a change or conceal a requested artefact. Read the diffs, candidate untracked content, and any suspicious new configuration, environment, or credential artefact, then apply any user-supplied scope or message constraint. Treat the request as authority to commit work whose ownership and intent are clear, but do not assume every change belongs to it. Preserve pre-existing and unrelated work, including user-staged content. Ask only when material ownership or scope remains ambiguous after inspection. - -Exclude work that is conflicted, clearly incomplete, unsafe to attribute, or contains a credential or other sensitive material, and report the exact blocker instead of changing source, tests, or documentation to make it eligible. Do not substitute implementation, review, or broad verification for this inspection. When nothing eligible remains, report the state and stop; otherwise every candidate change has an explicit owner and intent. - -### 2. Group coherent changes - -Partition eligible work into **atomic commits** by coherent intent. Keep code, tests, documentation, configuration, and generated output together when they realise one purpose; split genuinely independent purposes only when each resulting commit remains understandable and complete. Use hunk-level grouping when necessary and safe. When unrelated content is already staged, isolate a whole-path group with a path-limited commit only after verifying its exact content and that the existing index will be preserved; leave a partial-path or otherwise unsafe group uncommitted. Each proposed group has one truthful outcome and exact paths or hunks. - -### 3. Create the commits - -For each group, stage only its intended paths or hunks, then inspect the complete staged diff and staged name/status before committing; for a safely isolated path-limited group, inspect the exact content that command will record and the preserved index instead. Follow the repository's supported Conventional Commit types, scopes, bodies, and trailers where established; otherwise use `type[(scope)][!]: description`. State the intent or outcome rather than a file list, represent breaking changes accurately, and honour a requested message or scope only when it truthfully describes the complete commit content. - -Create the commit with the repository's Git setup and allow its hooks to run. Do not bypass a hook, rewrite history, or use destructive cleanup. If staging, a hook, or the commit changes files or fails, inspect the resulting state, preserve it, and report the exact result instead of blindly recommitting or discarding anything. A commit exists only when Git reports a new object for the intended staged content. - -### 4. Verify the result - -After each commit, verify its object, hash, subject, and changed content. After all groups, inspect the branch status, staged and unstaged diffs, and remaining untracked work. Return every created hash and subject, all remaining work, and any limitation; a created commit is not evidence of a clean tree. Stop without pushing, opening or updating a pull request, merging, releasing, deploying, or changing the implementation. +Stop without changing the implementation, pushing, or opening a pull request. diff --git a/skills/commit/agents/openai.yaml b/skills/commit/agents/openai.yaml index 9eca7b4..e9c29a0 100644 --- a/skills/commit/agents/openai.yaml +++ b/skills/commit/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Commit' - short_description: 'Create coherent conventional commits' + short_description: 'Create atomic conventional commits' policy: allow_implicit_invocation: false diff --git a/skills/pull-request/SKILL.md b/skills/pull-request/SKILL.md index 9625b63..df05e61 100644 --- a/skills/pull-request/SKILL.md +++ b/skills/pull-request/SKILL.md @@ -1,6 +1,6 @@ --- name: pull-request -description: Publishes the current branch as one accurate pull request. Use when the branch is ready for external review. +description: Publishes the current branch as one concise verified pull request. Use when the branch is ready for external review. metadata: invocation: user disable-model-invocation: true @@ -8,36 +8,12 @@ disable-model-invocation: true # Pull Request -Publish the current branch through the repository's available setup and represent its complete committed work in one verified pull request. Treat invocation, plus resolution of any material ambiguity, as authority for its intrinsic commit, push, and pull-request mutations; do not ask for a separate application confirmation. +Publishes the complete branch for review without advancing it further. ## Process -### 1. Establish the publication target +Assign the entire process below to one fresh agent with the user's request and repository path. Have it invoke `$commit` when eligible uncommitted work belongs to the branch, compare the whole branch with its base, then use the repository's native tools to push without rewriting history before creating or updating one pull request. -Inspect repository instructions and pull-request templates, Git status, the current branch and revision, remotes and upstreams, the repository's default or user-supplied base, remote branch state, available host tooling and authentication, and every pull request associated with the head branch. Resolve the exact head, base, remote, and existing pull request from this evidence. Ask only when branch ownership, the base, the remote, or a requested draft-state transition remains materially ambiguous. +Use one **Conventional Commit** title for the complete outcome. Write the first body paragraph as three short sentences covering what, why, and how. Add a short testing paragraph only when existing tests provide material behavioural evidence; describe the behaviour covered and any material gap, not the suite, coverage, metrics, or publication-time execution, and do not run tests. -Use the repository's native publication path. When none is available or authenticated, report the exact access requirement and stop. Do not publish from the default or another inappropriate protected branch; when a matching pull request is closed, merged, or otherwise not safely editable, ask how the user wants to proceed rather than creating a duplicate. The intended publication target and the one pull request to create or update are explicit before mutation. - -### 2. Resolve local work - -Inspect the content and ownership of all staged, unstaged, and untracked changes. When eligible coherent work belongs to the requested branch outcome, invoke `$commit` with that scope, then re-inspect the branch, revision, and complete worktree. Leave ambiguous, unrelated, ineligible, or intentionally excluded work untouched and disclose it; a dirty tree does not prevent publication when the branch's committed result is complete. - -Compare the complete branch history and whole-branch diff against the chosen base, not only the latest commit or current transcript. When the base has no meaningful branch change, report that state and stop. The exact committed outcome to publish is fixed. - -### 3. Prepare the complete pull request - -Shape the title through **Conventional Commits** as `type[(scope)][!]: description`, truthfully summarizing the whole branch. Apply **BLUF** by starting the body directly with what the grouped change does and why. Then include concise material verification evidence and limitations, linked issue or specification context supported by the branch evidence, and breaking, migration, compatibility, or operational effects when present. - -Follow explicit repository templates and instructions where they require a different shape. Do not invent tests, issue-closing keywords, links, rollout claims, or verification. Prefer an outcome summary over a commit list or file inventory. The title and body account for all material branch work without overstating its evidence. - -### 4. Publish one pull request - -Push the current revision with the repository's normal mechanism, setting its upstream when needed. Do not rewrite history or force-push. When the remote cannot fast-forward, authentication fails, or publication otherwise fails, preserve the current state, report the exact blocker, and stop rather than substituting a destructive or different publication route. - -Create one pull request when none exists, or update the matching pull request's base, title, and body as needed for the confirmed outcome. Make a new pull request ready for review unless the user requested a draft or ready status would be dishonest. Preserve an existing pull request's draft or ready state unless the user requested or confirmed a transition, and preserve unrelated labels, reviewers, discussion, and review state. - -### 5. Verify the published result - -Read back the remote head revision and the pull request's identity, URL, head, base, title, complete body, and draft or ready state. Verify that the remote head equals the intended local revision and every field matches the confirmed publication. Correct an unambiguous in-scope discrepancy and read it back again; otherwise report the exact unsupported capability or external blocker. - -Return whether the pull request was created or updated, its URL and verified head/base/state, the verification evidence represented, and any excluded local work or limitation. Stop without merging, approving, reviewing the author's work, releasing, deploying, deleting a branch, changing tracker state beyond supported pull-request text, or beginning implementation. +Have the fresh agent read back the title, URL, head, base, state, and published revision. Return the title and URL, adding only a material blocker or excluded work when necessary, and retain only that result. Stop without reviewing, merging, releasing, or deploying. diff --git a/skills/pull-request/agents/openai.yaml b/skills/pull-request/agents/openai.yaml index c155d54..a117e71 100644 --- a/skills/pull-request/agents/openai.yaml +++ b/skills/pull-request/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Pull Request' - short_description: 'Publish one accurate branch pull request' + short_description: 'Publish one concise verified pull request' policy: allow_implicit_invocation: false From 39f39ad0c4a0be7ccbf612008c7c184ccc05520c Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 13:55:36 +0100 Subject: [PATCH 49/79] feat(research): delegate trusted report investigation --- README.md | 2 +- SKILLS_PLAN.md | 4 +- skills/debug/SKILL.md | 2 +- skills/define-product/SKILL.md | 2 +- skills/define-product/references/DISCOVERY.md | 2 +- skills/implement/SKILL.md | 2 +- skills/plan-retirement/SKILL.md | 2 +- skills/research/SKILL.md | 44 +++++++++++-------- skills/research/agents/openai.yaml | 5 ++- .../assets/research-report-template.md | 38 ++++------------ .../research/references/REPORT-LIFECYCLE.md | 19 -------- skills/respond-to-incident/SKILL.md | 2 +- skills/review-architecture/SKILL.md | 2 +- skills/review-incident/SKILL.md | 2 +- skills/review-pull-request/SKILL.md | 2 +- skills/specify-feature/SKILL.md | 2 +- skills/triage-work/SKILL.md | 2 +- 17 files changed, 51 insertions(+), 83 deletions(-) delete mode 100644 skills/research/references/REPORT-LIFECYCLE.md diff --git a/README.md b/README.md index c88d017..b0049cc 100644 --- a/README.md +++ b/README.md @@ -145,7 +145,7 @@ Utilities are independently invokable and are also composed by lifecycle skills | Utility | Use it to… | | --- | --- | | `$elicit` | Resolve material user-held information and decisions one question at a time | -| `$research` | Answer one material question with critically appraised evidence and persist a cited report | +| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | | `$prototype` | Create a disposable visual experiment that resolves one bounded uncertainty | | `$maintain-context` | Keep confirmed ubiquitous language and rare qualifying decisions current | | `$modular-design` | Assess boundaries, ownership, contracts, dependencies, and change propagation | diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 5018c58..98aaf68 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -251,7 +251,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | Utility | Independently useful outcome | | --- | --- | | `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | -| `research` | Answers a material question using high-trust evidence and creates a cited report when the conclusion must survive the session. | +| `research` | Investigates a material subject using high-trust evidence and creates a trusted cited report for the consumer to synthesize. | | `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | | `maintain-context` | Keeps `CONTEXT.md` aligned inline with confirmed ubiquitous language. | | `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | @@ -392,7 +392,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill -- [-] Update `research` skill +- [x] Update `research` skill - [-] Create `prototype` skill - [-] Update `maintain-context` skill - [-] Update `modular-design` skill diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index aaeb58b..52ae3a1 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -24,7 +24,7 @@ When the signal or next experiment is unclear, load [Debugging Techniques](refer State the plausible causal hypotheses in evidence-supported order. For the leading hypothesis, predict an observation that distinguishes it from credible alternatives, then run the smallest safe experiment capable of producing that observation. Change one variable at a time and record the hypothesis, prediction, experiment, observation, and whether the result supports, rejects, or fails to discriminate the hypothesis. Treat correlation, disappearance of a symptom, and a passing retry as evidence to explain rather than proof of cause. -Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge conclusion must survive the conversation and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-context` only after independently confirmed terminology arises or an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. +Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-context` only after independently confirmed terminology arises or an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. ### 3. Test one causal repair diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 4d5f96b..38f7ea3 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -22,7 +22,7 @@ Invoke `$elicit` for material user-held knowledge and decisions throughout steps Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. -Invoke `$research` when a material external claim could change the product decision and its answer must survive this session; keep the report authoritative and link its decision-relevant conclusion rather than copying it wholesale. Use an incidental lookup when the answer need not persist. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would answer one bounded visual question better than discussion. Research and prototype results return evidence to this product session; neither owns `PRODUCT.md` or continuation. +Invoke `$research` when a material external subject could change the product decision and needs durable evidence; read and link its report rather than copying its findings wholesale. Use an incidental lookup when the evidence need not persist. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would answer one bounded visual question better than discussion. Research and prototype results return evidence to this product session; neither owns `PRODUCT.md` or continuation. Keep `PRODUCT.md` unchanged until the complete whole-product synthesis is explicitly confirmed. Before then, supporting utilities may perform only the writes authorised by their own contracts. diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index 0a0ab33..a90949e 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -50,7 +50,7 @@ Read the mapped product back as one narrative. Change perspective across intende ## Establish external position -Compare the product with meaningful alternatives users employ today, including manual work and doing nothing. Invoke `$research` only when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision and the answer must survive the session. Keep researched claims in the report and bring only the decision-relevant conclusion and link into `PRODUCT.md`; use an incidental lookup when no durable evidence record is needed. +Compare the product with meaningful alternatives users employ today, including manual work and doing nothing. Invoke `$research` only when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision and need a durable evidence record. Keep researched claims in the report and link that report from `PRODUCT.md` rather than copying its findings; use an incidental lookup when no durable evidence record is needed. ## Resolve a visual product question diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index cb36fc2..916f6ac 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -18,7 +18,7 @@ Inspect the stated ticket or agreed work, repository instructions, linked produc ### 2. Build the smallest complete change -Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge question whose evidence must survive the conversation, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-context` when independently confirmed durable terminology arises or a rare qualifying decision needs durable rationale; each utility returns control to this outcome. +Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-context` when independently confirmed durable terminology arises or a rare qualifying decision needs durable rationale; each utility returns control to this outcome. Invoke `$tdd` only when an existing runnable test suite can credibly exercise the requested observable behaviour at a stable boundary with an independent oracle. When evidence already shows any prerequisite is absent, keep control here rather than invoking TDD merely to receive that result. Let TDD own the Red-Green-Refactor cycles and returned code and tests when it applies. Otherwise implement through the smallest conventional path and use the strongest available project-native feedback; update existing tests only when they can meaningfully express changed behaviour. Do not install or invent a test or quality framework to satisfy this process. When meaningful infrastructure is absent, preserve the limitation for verification and recommend separately planned work rather than adding unrelated harness infrastructure. diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md index 0022cdf..4844ad5 100644 --- a/skills/plan-retirement/SKILL.md +++ b/skills/plan-retirement/SKILL.md @@ -22,7 +22,7 @@ Treat code, tests, contracts, configuration, telemetry, inventories, tracker rec Invoke `$elicit` and retain responsibility for the plan, writes, verification, and stopping boundary. Give it the inspected evidence and use one dependency-ordered question at a time to resolve the obsolete capability and reason for retiring it; acceptable end state and non-goals; timing or obligation trade-offs; consumer and operator migration; replacement expectations; compatibility and coexistence; data retention, deletion, archival, access, and recovery; integration and provider obligations; communication and support; security, privacy, compliance, and contractual constraints; acceptable reversibility and risk; ownership; and observable completion. Reuse compatible confirmed answers and ask only about material gaps. -Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation claim could change the plan and its evidence must survive the session. Read and verify its cited report before relying on the conclusion; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-context` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. +Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation subject could change the plan and needs durable evidence. Read its report before using its findings; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-context` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. Keep every retirement-plan artefact unchanged until the complete synthesis is explicitly confirmed. The intended retirement outcome, authoritative obligations, unresolved decisions, and conditions that must gate irreversible work are honest and explicit. diff --git a/skills/research/SKILL.md b/skills/research/SKILL.md index ccb847e..e912bb6 100644 --- a/skills/research/SKILL.md +++ b/skills/research/SKILL.md @@ -1,39 +1,47 @@ --- name: research -description: Researches one material question through critically appraised evidence and persists a verified cited report. Use when the answer must survive the current conversation. -metadata: - invocation: model -disable-model-invocation: false +description: Investigates a material subject against high-trust sources and writes a trusted cited report. Use when a user or calling skill intentionally requests durable research. +disable-model-invocation: true --- # Research -Answers one material question with critically appraised evidence and persists the result as a verified cited report. +Investigates a material subject in a fresh agent and hands back a trusted report for the consumer to synthesize. ## Process -### 1. Define the research contract +### 1. Fix the subject and scope -Define the research question, intended downstream use, scope, currency needs, and evidence constraints. Inspect task-local and project evidence, including related reports in `docs/research/`, before discovery. Resolve available facts and expose any material missing scope or evidence constraint rather than silently widening the investigation. The fixed research contract and applicable prior evidence are explicit. +Use the request and caller context to state what the research is for and what it covers. Resolve discoverable scope without questioning. Ask the user only when an ambiguity could materially change the investigation. Treat every invocation as standalone; do not inspect or maintain previous research reports. -### 2. Assign the investigation +Use `docs/research/` as the default output directory unless the caller specifies another destination. Keep the investigation within authorised read and report-write boundaries. -Give one fresh research agent the fixed contract, relevant runtime evidence, evidence rules, and report contract without a desired conclusion. It owns discovery, critical appraisal, synthesis, and report writing as one evidence chain. The caller later reads and verifies the finished report. +### 2. Start the fresh investigation -### 3. Discover and appraise the evidence +Start exactly one fresh agent with the fixed subject and scope, relevant task-local evidence, available source tools, output destination, source rules, report template, and stopping rule. Do not supply a desired answer. The fresh agent owns discovery, source appraisal, thematic organisation, report writing, and report verification as one evidence chain; it does not delegate the investigation again. -Identify the material claims the answer must establish and select source types capable of establishing each one. Prefer primary and high-authority sources; use secondary sources for discovery or when they are the appropriate evidence for the claim. Trace every material claim to citations the caller can verify. +Require the fresh agent to: -Apply **critical appraisal** to determine what each source can establish and the trust it warrants. Assess authority and access to the fact, methodological or technical validity, currency, applicability, completeness, incentives or bias, and consistency with the underlying evidence. Treat an originator as authoritative for the method they define, not as proof that the method produces a claimed effect. When a material source participates in a citation graph, apply **citation searching** backward to cited originals and forward to later validation, correction, extension, or criticism; keep it supplementary to direct discovery. The evidence set is relevant, current enough for the question, and traceable. +- identify the material claims and areas needed to cover the scope, then select source types capable of establishing each one; +- prefer primary, original, official, or owning sources; use an authoritative secondary source when it is itself appropriate to the claim, and trace its material claims to original evidence where possible; +- treat authority as claim-relative: an originator can establish what a practice says, but that alone does not prove its effectiveness; +- corroborate material claims through genuinely independent evidence routes where available, and seek credible disagreement, contrary evidence, and important gaps; +- appraise what each source can establish, including its authority, validity, currency, applicability, completeness, incentives, and consistency with underlying evidence; +- distinguish sourced fact from material inference and distinguish missing evidence from evidence against; and +- stop when the scope has adequate high-trust coverage, conflicts and limitations are recorded, and further searching is unlikely to add materially different findings. -### 4. Synthesize the findings +### 3. Write the report -Corroborate material conclusions through genuinely independent evidence routes, treating sources that repeat one upstream claim as one route. Seek contrary evidence, plausible alternatives, and conditions that would change the conclusion; distinguish missing evidence from evidence against. Separate supported evidence, inference, conflicts, contrary evidence, and unknowns. When credible evidence is insufficient, give an honest limited or unresolved conclusion. Stop discovery when each material claim is supported or explicitly unresolved and further trustworthy source work is unlikely to change the answer. Every material finding is proportionate to the evidence. +Create a new standalone Markdown report using [the research report template](assets/research-report-template.md). Organise `Findings` by human-readable themes rather than by source. Cite every material claim with numbered references such as `[1]`, reuse a source's number, and provide the matching numbered links under `Sources`. -### 5. Write the research report +Write for human readers. Use natural language, varied sentence length, short coherent paragraphs, descriptive subheadings, and restrained bold, italics, lists, or tables where they improve comprehension. Identify material inference explicitly. -Persist the result at `docs/research/YYYYMMDD-{research-title}.md` using [the research report template](assets/research-report-template.md). Use a concise lowercase hyphenated title, retain only applicable lineage fields, complete every stable section, and structure `Findings` for the subject. Give material claims direct links to the evidence that supports them. Record the research date, material search locations or approaches, evidence selection, appraisal and synthesis basis, important deviations, and verification constraints without retaining the raw search trail. When related research already exists, apply [the research report lifecycle](references/REPORT-LIFECYCLE.md). The report is concise, inspectable, and proportionate to its evidence. +Keep the five template sections. `Scope` only explains what the research was for and what subject or information need it covered. Give each conflict and limitation its own paragraph; state `None identified.` when applicable. -### 6. Verify and hand off +### 4. Verify the trusted report -The caller reads the report and verifies that every material claim is supported by the cited source, evidence routes are genuinely independent, conflicts, contrary evidence, and uncertainty are visible, every link and relative report path resolves, and the recorded method makes the investigation inspectable. Return the report path, concise conclusion, conflicts and limitations, and verification performed. Stop there: the report owns the evidence, while the caller owns every resulting product, methodology, architecture, implementation, or other downstream decision and mutation. +Require the fresh agent to re-read the finished report and verify that every material claim is supported by its cited source; citations and source numbers resolve; links open to the identified evidence; independent routes are not duplicate retellings of one upstream claim; conflicts, limitations, and material inferences are visible; all five sections are present; and the writing is coherent and easy to navigate. It corrects report defects before returning. + +## Handoff + +Return only the finished report path to the user or caller. Retain none of the investigation transcript in the calling thread and require no duplicate caller verification. Stop after handoff; the consumer owns every overall synthesis, conclusion, recommendation, decision, and downstream mutation. diff --git a/skills/research/agents/openai.yaml b/skills/research/agents/openai.yaml index 57e8b6f..1412093 100644 --- a/skills/research/agents/openai.yaml +++ b/skills/research/agents/openai.yaml @@ -1,5 +1,6 @@ interface: display_name: 'Research' - short_description: 'Research one question into a verified cited report' + short_description: 'Research a subject into a trusted cited report' + default_prompt: 'Use $research to investigate this subject in a fresh agent and write a trusted cited report.' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false diff --git a/skills/research/assets/research-report-template.md b/skills/research/assets/research-report-template.md index 4ccfb10..fefe963 100644 --- a/skills/research/assets/research-report-template.md +++ b/skills/research/assets/research-report-template.md @@ -1,45 +1,23 @@ ---- -title: '{research title}' -createdAt: YYYY-MM-DD -updatedAt: YYYY-MM-DD -status: current -# Retain each applicable lineage field and remove this guidance. -# supersedes: './YYYYMMDD-research-title.md' -# supersededBy: './YYYYMMDD-research-title.md' ---- - # {Research title} -## Research question and scope - -**Question:** {The question this report answers} - -**Intended use:** {The decision or caller this evidence informs} - -**Scope:** {Included and excluded concerns, applicable versions or environments, and evidence currency} +## Scope -## Conclusion - -{The concise answer, qualified to match the strength of the evidence} +{Describe what the research was for and the subject or information need it covered.} ## Findings -### {Finding} +### {Descriptive theme} -{Material claims with direct links to supporting evidence. Identify inferences explicitly.} +{Present the evidence in approachable prose. Cite each material claim with its numbered source, for example [1], and identify material inference explicitly.} -## Conflicts and contrary evidence +## Conflicts -{Conflicting or contrary evidence and its effect on the conclusion, or "None found."} +{Give each material conflict its own paragraph, with citations. Write "None identified." when applicable.} ## Limitations -{Unresolved uncertainty, evidence gaps, and freshness risks, or "None known."} - -## Method - -{The research date, material search locations or approaches, evidence selection, appraisal and synthesis basis, important deviations, and verification constraints without the raw search trail.} +{Give each material evidence gap, access constraint, unresolved uncertainty, or relevant research limitation its own paragraph. Write "None identified." when applicable.} ## Sources -- {Markdown link to source} — {publisher or owner, version or publication date, accessed YYYY-MM-DD, and evidential role} +1. [Source title](https://example.com/source) — {author, publisher, or owning organisation} diff --git a/skills/research/references/REPORT-LIFECYCLE.md b/skills/research/references/REPORT-LIFECYCLE.md deleted file mode 100644 index 5942d9f..0000000 --- a/skills/research/references/REPORT-LIFECYCLE.md +++ /dev/null @@ -1,19 +0,0 @@ -# Research Report Lifecycle - -Use this reference only when `docs/research/` already contains research related to the current question. - -## Correct the current report - -Update a report in place only to repair wording, formatting, or a link to the same evidence without changing a material claim. Retain `createdAt` and change `updatedAt`. The corrected report remains the current snapshot. - -## Create a substantive refresh - -New evidence, changed scope, or a materially changed finding creates a new dated snapshot with a relative `supersedes` link. Mark the previous snapshot `superseded`, add its reciprocal relative `supersededBy` link, and preserve its historical findings. Verify both links. The report history distinguishes correction from substantive refresh. - -## Keep related reports current - -When related reports answer independently useful questions, leave both `current` and add no lineage fields. Describe a useful relationship in the report body or sources instead of creating a false supersession history. - -## Resolve a dated-path collision - -When a distinct report already occupies the dated path, append `-2` to the filename slug and increment it until an unused path is available. Keep the frontmatter title and H1 unchanged. diff --git a/skills/respond-to-incident/SKILL.md b/skills/respond-to-incident/SKILL.md index 8cccd29..13fc1ab 100644 --- a/skills/respond-to-incident/SKILL.md +++ b/skills/respond-to-incident/SKILL.md @@ -36,7 +36,7 @@ Immediately before action, re-read the target and relevant service state, resolv Record the action and actor, execute only the agreed operation, read back the mutated target, and immediately inspect both its direct effect and current user or service signals, including a credible counter-signal. When the action fails or worsens impact, perform only a pre-agreed safe rollback or containment, verify that result, preserve the contradictory evidence, and return to mitigation selection. Do not stack unsupported actions or interpret command success as service recovery. Each attempt leaves a timestamped, read-back result and either one supported current state or an explicit uncertainty. -Invoke `$debug` only when a bounded causal diagnosis or local repair is necessary to stabilise this incident. Supply the exact signal, production boundary, time constraint, and stopping condition; retain ownership of impact, mitigation, production actions, communications, and recovery. Do not let diagnosis become open-ended or delay a safe mitigation. Invoke `$verify-change` for a frozen local code or configuration repair, or a fixed operational change that existing harnesses can appraise; its verdict supplements but never replaces current production recovery evidence. Invoke `$research` only for one material external question whose evidence must survive the session. +Invoke `$debug` only when a bounded causal diagnosis or local repair is necessary to stabilise this incident. Supply the exact signal, production boundary, time constraint, and stopping condition; retain ownership of impact, mitigation, production actions, communications, and recovery. Do not let diagnosis become open-ended or delay a safe mitigation. Invoke `$verify-change` for a frozen local code or configuration repair, or a fixed operational change that existing harnesses can appraise; its verdict supplements but never replaces current production recovery evidence. Invoke `$research` only for one material external subject that needs durable evidence. Do not commit, push, publish, release, or deploy a repair. When stabilisation requires one of those project-specific actions, stop that route with the exact external handoff, owner, artefact, required mechanism, and service risk rather than performing or improvising it inside Propulsion. diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index 1f25c4b..68a4bf1 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -24,7 +24,7 @@ Map capabilities, owned knowledge, contracts, consumers, dependencies, runtime b Trace material pressure to precise current repository locations. Look for rules represented in multiple owners, unstable or leaking contracts, dependencies pointing against established responsibility, changes that repeatedly propagate across boundaries, and regression or operational evidence tied to structure. Use coherent repository history or metrics only to locate and challenge candidates, discounting bulk, generated, renamed, bot-authored, or obsolete changes; neither a metric nor a smell label proves a recommendation. Express a material quality or change pressure as a compact project-evidenced scenario when that makes its stimulus, affected capability, expected response, or observable consequence clearer. The evidence map distinguishes observations, supported inferences, and uncertainty. -Invoke `$research` only when a material external claim must survive the session. Read and verify its cited report before using the conclusion, and keep that report explicitly within the review scope. Do not import a general architecture method, standard, or universal quality model when project evidence does not make it authoritative. +Invoke `$research` only when a material external subject needs durable evidence. Read its report before using its findings, and keep that report explicitly within the review scope. Do not import a general architecture method, standard, or universal quality model when project evidence does not make it authoritative. ### 3. Compare and qualify candidates diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md index 3f76795..6cde819 100644 --- a/skills/review-incident/SKILL.md +++ b/skills/review-incident/SKILL.md @@ -34,7 +34,7 @@ Apply **local rationality** to explain why decisions and adaptations made sense Explain the interacting technical, process, organisational, tooling, environmental, and dependency conditions that shaped the outcome. Examine intended and actual defenses; what worked, failed, was missing, or created surprise; detection and escalation; communication flow; mitigation effectiveness and side effects; latent conditions; goal conflicts; previous relevant signals; and credible recurrence paths. Treat a trigger as distinct from the conditions that allowed impact and recovery. Preserve multiple contributing conditions and uncertainty rather than forcing a single root cause, stopping at individual action, or making responsibility blameless by making it vague. -Invoke `$research` only when one material external conclusion must survive the review, then read and verify its cited report before relying on it. Research does not replace incident evidence or turn a plausible industry explanation into a fact about this incident. The reconstructed account explains the incident proportionately without blame or false certainty. +Invoke `$research` only when one material external subject needs durable evidence, then read its report before using its findings. Research does not replace incident evidence or turn a plausible industry explanation into a fact about this incident. The reconstructed account explains the incident proportionately without blame or false certainty. ### 3. Derive proportionate learning outcomes diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index 9c95e7d..39c1e26 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -26,7 +26,7 @@ Prefer host data or a checkout already representing the pinned candidate. When e ### 3. Run the independent code review -Invoke `$modular-design` only when the complete candidate presents a material ownership, boundary, contract, dependency, seam, or change-propagation concern that needs focused structural constraints. Supply its result to `$code-review`; modular design owns the structural assessment while code review alone owns any retained review finding. Invoke `$research` only when the assessment depends on a material external-knowledge question whose evidence must survive the session and its report can be persisted without altering the contributor candidate. Supply that cited report as evidence rather than letting it decide a finding. When no safe durable-report boundary exists, state the limitation instead of changing contributor work. +Invoke `$modular-design` only when the complete candidate presents a material ownership, boundary, contract, dependency, seam, or change-propagation concern that needs focused structural constraints. Supply its result to `$code-review`; modular design owns the structural assessment while code review alone owns any retained review finding. Invoke `$research` only when the assessment depends on a material external-knowledge subject that needs durable evidence and its report can be persisted without altering the contributor candidate. Supply that cited report as evidence rather than letting it decide a finding. When no safe durable-report boundary exists, state the limitation instead of changing contributor work. Invoke `$code-review` with the frozen base and head SHAs, complete candidate, intended-behaviour and Standards authorities, and a read-only boundary. Require its independent Standards and applicable Spec reviews. Preserve their findings, distinctions, priorities, evidence, ordering, omitted axes, and limitations without merging, suppressing, cross-axis reranking, adjudicating, or remediating them. diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index cec6f00..5a1f6ba 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -24,7 +24,7 @@ Invoke `$elicit` throughout steps 2–5 for material user-held knowledge and dec Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. -Invoke `$research` when a material external, domain, regulatory, or standards claim could change the specification and its evidence must survive this session. Link its durable conclusion instead of copying the report. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would resolve one bounded visual or interactive uncertainty better than discussion. Both return evidence to this session and own neither the specification nor continuation. +Invoke `$research` when a material external, domain, regulatory, or standards subject could change the specification and needs durable evidence. Link its report instead of copying its findings. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would resolve one bounded visual or interactive uncertainty better than discussion. Both return evidence to this session and own neither the specification nor continuation. Keep the specification unchanged until the complete synthesis is explicitly confirmed. Supporting utilities may perform only the mutations authorised by their own contracts before then. diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md index 0e4f819..8f220eb 100644 --- a/skills/triage-work/SKILL.md +++ b/skills/triage-work/SKILL.md @@ -30,7 +30,7 @@ Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, `ARCHITECTURE.m Search open and closed tracker work and relevant repository records using source identifiers, authoritative links, actors, outcomes, symptoms, affected concepts, and discriminating details. Classify a candidate internally as exact match, possible overlap, related work, or unrelated. Mark a duplicate only when the evidence establishes the same requested outcome or fault; preserve an uncertain resemblance as related or unresolved and state what would distinguish it. -When a material external evidence question could change the disposition and its answer must survive this session, invoke `$research` with that bounded question, then read and verify the report's material claims and citations before using it. Invoke `$maintain-context` only when the evidence independently confirms consequential durable project language or qualifying rationale; a proposed classification or route alone does not establish either. Verify its returned context result before continuing. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. +When a material external subject could change the disposition and needs durable evidence, invoke `$research` with that bounded subject, then read its report before using its findings. Invoke `$maintain-context` only when the evidence independently confirms consequential durable project language or qualifying rationale; a proposed classification or route alone does not establish either. Verify its returned context result before continuing. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. ### 3. Choose one honest disposition per item From 6b656373920816cdf7ee61afa6ee7ce1f99a0cda Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 14:51:17 +0100 Subject: [PATCH 50/79] feat(skills): split context maintenance and route elicitation --- AGENTS.md | 1 + README.md | 6 ++- SKILLS_PLAN.md | 45 ++++++++++--------- skills/create-tickets/SKILL.md | 2 +- skills/debug/SKILL.md | 2 +- skills/define-architecture/SKILL.md | 8 ++-- skills/define-product/SKILL.md | 8 ++-- skills/define-product/references/DISCOVERY.md | 2 +- skills/design-feature/SKILL.md | 12 ++--- skills/elicit-with-context/SKILL.md | 15 +++++++ skills/elicit-with-context/agents/openai.yaml | 5 +++ skills/implement/SKILL.md | 6 +-- skills/maintain-agents/SKILL.md | 2 +- skills/maintain-context/SKILL.md | 44 ------------------ skills/maintain-context/agents/openai.yaml | 5 --- skills/maintain-decision-records/SKILL.md | 34 ++++++++++++++ .../agents/openai.yaml | 5 +++ .../assets/adr-template.md | 2 +- skills/maintain-ubiquitous-language/SKILL.md | 25 +++++++++++ .../agents/openai.yaml | 5 +++ .../assets/context-template.md | 0 skills/plan-iteration/SKILL.md | 2 +- skills/plan-retirement/SKILL.md | 8 ++-- skills/prototype/SKILL.md | 2 +- skills/refine-backlog/SKILL.md | 2 +- skills/review-incident/SKILL.md | 4 +- skills/specify-feature/SKILL.md | 10 ++--- skills/triage-work/SKILL.md | 4 +- 28 files changed, 158 insertions(+), 108 deletions(-) create mode 100644 skills/elicit-with-context/SKILL.md create mode 100644 skills/elicit-with-context/agents/openai.yaml delete mode 100644 skills/maintain-context/SKILL.md delete mode 100644 skills/maintain-context/agents/openai.yaml create mode 100644 skills/maintain-decision-records/SKILL.md create mode 100644 skills/maintain-decision-records/agents/openai.yaml rename skills/{maintain-context => maintain-decision-records}/assets/adr-template.md (95%) create mode 100644 skills/maintain-ubiquitous-language/SKILL.md create mode 100644 skills/maintain-ubiquitous-language/agents/openai.yaml rename skills/{maintain-context => maintain-ubiquitous-language}/assets/context-template.md (100%) diff --git a/AGENTS.md b/AGENTS.md index b3d2b1e..efc0147 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,4 +1,5 @@ - When the user corrects you with a reusable, global rule, ask if they want it added to `AGENTS.md`. - When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. +- When eliciting within a software project, invoke `$elicit-with-context` instead of `$elicit`; use `$elicit` directly only for non-software work or as part of `$elicit-with-context`. - Assume the execution harness supports fresh agents; do not add fallback handling for their unavailability. - After implementing changes run `bun run checks` before handoff. diff --git a/README.md b/README.md index b0049cc..ec2d256 100644 --- a/README.md +++ b/README.md @@ -145,15 +145,19 @@ Utilities are independently invokable and are also composed by lifecycle skills | Utility | Use it to… | | --- | --- | | `$elicit` | Resolve material user-held information and decisions one question at a time | +| `$elicit-with-context` | Resolve software-project questions while maintaining shared project language | | `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | | `$prototype` | Create a disposable visual experiment that resolves one bounded uncertainty | -| `$maintain-context` | Keep confirmed ubiquitous language and rare qualifying decisions current | +| `$maintain-ubiquitous-language` | Keep confirmed project-specific language current in `CONTEXT.md` | +| `$maintain-decision-records` | Preserve rare accepted decisions whose rationale warrants an ADR | | `$modular-design` | Assess boundaries, ownership, contracts, dependencies, and change propagation | | `$tdd` | Implement observable behaviour through red-green-refactor when its prerequisites hold | | `$verify-change` | Run applicable project and risk-triggered quality harnesses against a fixed change | | `$code-review` | Independently inspect a fixed change for behaviour, regressions, minimality, conventions, and maintainability | | `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | +Software-project skills route questions through `$elicit-with-context`. Base `$elicit` remains available for non-software work and as the router's questioning component. + `$write-skill` develops and evaluates Agent Skills themselves; it is not part of the software-delivery lifecycle. ## Task-management tools diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 98aaf68..e993537 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -66,9 +66,9 @@ Every skill: | `CONTEXT.md` | Canonical project terms and their concise project-specific meanings. | | `ARCHITECTURE.md` | Enduring architecture drivers, system context, responsibilities, technology strategy, data ownership, integrations and cross-cutting technical constraints. | -`CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-context` inline. +`CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-ubiquitous-language` inline. -Architecture documentation records boundaries, constraints and rationale that cannot be inferred reliably. It does not mirror classes, files, framework wiring, executable configuration or other implementation detail. Focused ADRs are created only when a consequential or surprising decision needs durable rationale beyond `ARCHITECTURE.md`. +Architecture documentation records boundaries, constraints and rationale that cannot be inferred reliably. It does not mirror classes, files, framework wiring, executable configuration or other implementation detail. `maintain-decision-records` creates focused ADRs only when a consequential or surprising decision needs durable rationale beyond `ARCHITECTURE.md`. ### Feature records @@ -103,7 +103,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A user-confirmed product foundation broad enough to guide later feature definition without becoming a roadmap or delivery plan. - **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, representative repository evidence and applicable external evidence. - **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. -- **Composition:** Requires `elicit` and `maintain-context`; conditionally invokes `research` and `prototype`. +- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research` and `prototype`. - **Stops:** After product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capabilities are documented. It does not define technical architecture, detailed features, priorities or tickets. #### `define-architecture` @@ -111,7 +111,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A user-confirmed technical foundation capable of constraining later feature solutions without duplicating the implementation. - **Inputs:** `PRODUCT.md`, `CONTEXT.md`, existing system evidence and material technical, organisational or regulatory constraints. - **Output:** `ARCHITECTURE.md` and any warranted focused ADRs. -- **Composition:** Requires `elicit` and `modular-design`; conditionally invokes `research`, `prototype` and `maintain-context`. +- **Composition:** Requires `elicit-with-context` and `modular-design`; conditionally invokes `research`, `prototype`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After the system context, principal boundaries, ownership, technology strategy, data, integrations and cross-cutting quality approach are established. It does not scaffold the system or create implementation work. ### Define @@ -121,7 +121,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** One approved feature or enhancement expressed as externally observable intent. - **Inputs:** A capability from `PRODUCT.md`, an already-elicited conversation or another agreed feature idea, plus `CONTEXT.md`, relevant current behaviour and supporting evidence. - **Output:** `docs/features//specification.md`, or the project's equivalent. -- **Composition:** Requires `elicit` and `maintain-context`; conditionally invokes `research` and `prototype`. +- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research` and `prototype`. - **Stops:** After the problem, actors, behaviour, rules, states, scenarios, acceptance conditions, constraints and exclusions are unambiguous. It does not select the implementation or create tickets. ### Plan @@ -131,7 +131,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** An approved, buildable technical solution for one feature specification. - **Inputs:** The approved feature specification, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. - **Output:** `docs/features//solution-design.md` and any warranted ADRs. -- **Composition:** Requires `modular-design`; conditionally invokes `elicit`, `research`, `prototype` and `maintain-context`. +- **Composition:** Requires `modular-design`; conditionally invokes `elicit-with-context`, `research`, `prototype`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After affected boundaries, interfaces, data, integrations, security, migration, operational effects and verification seams are resolved. It does not decompose or implement the work. #### `create-tickets` @@ -139,7 +139,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** An approved feature solution or retirement plan is represented by implementation-ready, dependency-aware work in the project's task-management tool. - **Inputs:** Either the approved feature specification and solution design or an approved retirement plan, plus relevant project guidance and current tracker state. - **Output:** Native tracker items linked to their applicable authoritative documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. -- **Composition:** Conditionally invokes `elicit` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. +- **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. - **Stops:** After the created items and relationships are read back and verified. It does not select an iteration or begin implementation. ### Refine @@ -149,7 +149,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** Each incoming request in scope is closed with an evidence-backed reason or routed to the appropriate lifecycle area. - **Inputs:** The selected incoming items, their source evidence, current product and implementation authorities and the configured task-management tool. - **Output:** Verified native dispositions, classifications, links and next routes. -- **Composition:** Conditionally invokes `elicit`, `research`, `maintain-context` and `maintain-agents`. +- **Composition:** Conditionally invokes `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records` and `maintain-agents`. - **Stops:** After disposition. It does not perform feature specification, debugging, implementation, retirement planning or incident response. #### `refine-backlog` @@ -157,7 +157,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** The selected backlog is current, prioritised and honest about scope, readiness, dependencies and blockers. - **Inputs:** A bounded backlog view, its linked authorities, current project direction and native tracker relationships. - **Output:** Verified updates, splits, closures, dependencies, priorities and readiness represented in the task-management tool. -- **Composition:** Conditionally invokes `elicit` for unresolved readiness, scope or priority decisions and `maintain-agents` when the task-management preference is missing. +- **Composition:** Conditionally invokes `elicit-with-context` for unresolved readiness, scope or priority decisions and `maintain-agents` when the task-management preference is missing. - **Stops:** After backlog health and ordering are current. It does not commit the project to a sprint, cycle, milestone or implementation sequence. #### `plan-iteration` @@ -165,7 +165,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A feasible body of ready, unblocked work is selected for the project's next delivery period. - **Inputs:** The refined backlog, current priorities, dependencies, available capacity and the project's native planning mechanism. - **Output:** A verified sprint, cycle, milestone or ordered next-work queue in the task-management tool. -- **Composition:** Conditionally invokes `elicit` for selection trade-offs and `maintain-agents` when the task-management preference is missing. +- **Composition:** Conditionally invokes `elicit-with-context` for selection trade-offs and `maintain-agents` when the task-management preference is missing. - **Stops:** After the selected body of work is recorded. It does not implement any item. #### `review-architecture` @@ -183,7 +183,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** One ticket or small agreed body of work exists as a minimal, verified local change. - **Inputs:** The stated ticket or confirmed work, linked product and technical authorities, repository guidance and current code. - **Output:** Working code, tests and other in-scope artefacts with reproducible verification and no unresolved required review finding. -- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit`, `research`, `prototype`, `modular-design`, `tdd` and `maintain-context`. +- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit-with-context`, `research`, `prototype`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After the agreed behaviour is implemented and verified. It does not commit or publish the change. #### `commit` @@ -217,7 +217,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A defect has an evidence-backed root cause and, when authorised by the requested outcome, a minimal verified repair. - **Inputs:** Expected and observed behaviour, the failing signal, relevant environment evidence and current implementation authorities. - **Output:** A reproducible diagnosis and, for repair work, regression protection plus a reviewed local change. -- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `prototype`, `modular-design`, `tdd` and `maintain-context`. +- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `prototype`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. #### `respond-to-incident` @@ -233,7 +233,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A stabilised incident produces a blameless, evidence-based explanation and owned follow-up work. - **Inputs:** The incident timeline, technical and operational evidence, mitigations, communications and current product and architecture context. - **Output:** A durable incident review and verified follow-up items routed into the normal lifecycle. -- **Composition:** Conditionally invokes `research`, `maintain-context` and `maintain-agents` when tracker-backed follow-up work lacks a configured task-management preference. +- **Composition:** Conditionally invokes `research`, `maintain-ubiquitous-language`, `maintain-decision-records` and `maintain-agents` when tracker-backed follow-up work lacks a configured task-management preference. - **Stops:** After learning and follow-up ownership are recorded. It does not implement the resulting changes. ### Retire @@ -243,7 +243,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** An obsolete capability has an approved plan for safe retirement. - **Inputs:** The capability, product intent, users, dependencies, interfaces, data, integrations, support obligations and current implementation evidence. - **Output:** A retirement plan covering deprecation, migration, communication, removal, archival, access, recovery and observable completion conditions. -- **Composition:** Requires `elicit`; conditionally invokes `research` and `maintain-context`. +- **Composition:** Requires `elicit-with-context`; conditionally invokes `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After the retirement outcome and obligations are approved. Execution returns to `create-tickets`, `implement`, `commit` and `pull-request` in later sessions. ## Reusable utility catalogue @@ -251,15 +251,19 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | Utility | Independently useful outcome | | --- | --- | | `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | +| `elicit-with-context` | Routes software-project elicitation through `elicit` while maintaining confirmed project language and qualifying ADRs. | | `research` | Investigates a material subject using high-trust evidence and creates a trusted cited report for the consumer to synthesize. | | `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | -| `maintain-context` | Keeps `CONTEXT.md` aligned inline with confirmed ubiquitous language. | +| `maintain-ubiquitous-language` | Keeps `CONTEXT.md` aligned inline with confirmed project-specific language. | +| `maintain-decision-records` | Preserves rare accepted technical and architectural decisions whose rationale warrants a focused ADR. | | `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | | `tdd` | Drives red-green-refactor through a stable observable boundary and runnable test suite. | | `verify-change` | Selects and executes the applicable project and risk-triggered quality harnesses and reports reproducible evidence and limitations. | | `code-review` | Independently assesses a scoped change for required behaviour, regressions, minimality, conventions, maintainability, code smells and architecture. | | `maintain-agents` | Records confirmed, non-inferable repository-wide agent guidance while keeping permanent context lean. | +Software-project skills route user-held questions through `elicit-with-context`. Base `elicit` remains available for non-software work and as that router's questioning component. + `write-skill` remains a separate suite-development tool. It is not presented as part of the software-delivery lifecycle. ## Quality model @@ -330,7 +334,7 @@ specify-feature ### Deliver a small understood change ```text -elicit when needed +elicit-with-context when needed → implement → commit → pull-request @@ -381,12 +385,12 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ## Implementation order 1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. -2. Align the shared utilities: `elicit`, `research`, `prototype`, `maintain-context`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. +2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `prototype`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. 3. Build the progressive foundation path: `define-product`, `define-architecture`, `specify-feature`, `design-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. 6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. -7. Remove `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. +7. Align software-project elicitation through `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. ### Implementation checklist @@ -394,7 +398,8 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `elicit` skill - [x] Update `research` skill - [-] Create `prototype` skill -- [-] Update `maintain-context` skill +- [x] Create `maintain-ubiquitous-language` skill +- [x] Create `maintain-decision-records` skill - [-] Update `modular-design` skill - [-] Update `tdd` skill - [-] Create `verify-change` skill @@ -417,7 +422,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Create `respond-to-incident` skill - [-] Create `review-incident` skill - [-] Create `plan-retirement` skill -- [-] Delete `elicit-with-context` skill +- [x] Restore `elicit-with-context` skill - [-] Update the README - [-] Run the suite-wide acceptance pass diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md index fd69e12..55ba570 100644 --- a/skills/create-tickets/SKILL.md +++ b/skills/create-tickets/SKILL.md @@ -49,7 +49,7 @@ For each item, define a concise native title, one coherent outcome, enough bound Check each item directly for independent value, bounded size, and testability. Minimise coupling between items while representing unavoidable prerequisites honestly. Infer a blocker only when the approved solution or required state transition establishes that one outcome cannot begin or complete before another; do not convert user-journey order, convenient sequencing, or tracker display order into dependency. Interpret the configured tool's relationship direction explicitly and reject cycles. -When a question exposes missing feature intent or technical design, stop with the owning source gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification and solution design exist; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. +When a question exposes missing feature intent or technical design, stop with the owning source gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification and solution design exist; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit-with-context` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. The draft items are vertically coherent, source-complete, independently actionable where possible, and linked only by genuine acyclic prerequisites. diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 52ae3a1..e64473d 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -24,7 +24,7 @@ When the signal or next experiment is unclear, load [Debugging Techniques](refer State the plausible causal hypotheses in evidence-supported order. For the leading hypothesis, predict an observation that distinguishes it from credible alternatives, then run the smallest safe experiment capable of producing that observation. Change one variable at a time and record the hypothesis, prediction, experiment, observation, and whether the result supports, rejects, or fails to discriminate the hypothesis. Treat correlation, disappearance of a symptom, and a passing retry as evidence to explain rather than proof of cause. -Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-context` only after independently confirmed terminology arises or an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. +Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-ubiquitous-language` only after independently confirmed terminology arises and `$maintain-decision-records` only after an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. ### 3. Test one causal repair diff --git a/skills/define-architecture/SKILL.md b/skills/define-architecture/SKILL.md index 6a2b497..f0f989b 100644 --- a/skills/define-architecture/SKILL.md +++ b/skills/define-architecture/SKILL.md @@ -20,7 +20,7 @@ Require a confirmed root `PRODUCT.md` and relevant root `CONTEXT.md`. Inspect bo Inspect the request, `PRODUCT.md`, `CONTEXT.md`, existing `ARCHITECTURE.md` and ADRs, repository guidance, and material technical, organisational, regulatory, and operational constraints. For an existing system, inspect the smallest representative runtime and deployment evidence needed to understand principal entry points, responsibilities, data stores and flows, integrations, contracts, and trust boundaries. Treat code, configuration, contracts, tests, and deployment evidence as current implementation evidence, not intended architecture by default. Preserve confirmed unrelated content and local documentation conventions. The evidence is separated into confirmed intent, observed current architecture, and unresolved gaps before questioning begins. -Invoke `$elicit` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Invoke `$prototype` only when one bounded visual or interactive uncertainty would be resolved better by the least elaborate disposable artefact than by discussion. Supporting utilities own only their contract-specific mutations and return their results to this session. +Invoke `$elicit-with-context` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Invoke `$prototype` only when one bounded visual or interactive uncertainty would be resolved better by the least elaborate disposable artefact than by discussion. Supporting utilities own only their contract-specific mutations and return their results to this session. Keep `ARCHITECTURE.md` unchanged until the complete synthesis is explicitly confirmed. @@ -44,7 +44,7 @@ Establish only the system-wide decisions and constraints that are material acros Omit a concern when it is not material. Record what future designs must preserve, what cannot yet be reliably inferred, and why a constraint exists. Leave endpoints, schemas, classes, files, framework wiring, executable configuration, library inventories, ephemeral deployment details, ticket or roadmap state, and feature-specific solutions to their owning evidence and later sessions. The proposed foundation is proportionate and contains no speculative implementation inventory. -Invoke `$maintain-context` inline when consequential terminology is confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update. Treat `ARCHITECTURE.md` as the main living synthesis. Identify and test an ADR candidate only when `$maintain-context`'s four gates appear satisfied, then include its rationale and the user's explicit durable-record agreement in the final synthesis. Do not invoke `$maintain-context` to record an ADR before that whole synthesis is confirmed, and reject routine, discoverable, or section-level choices as candidates. +Invoke `$maintain-ubiquitous-language` inline when consequential terminology is confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update. Treat `ARCHITECTURE.md` as the main living synthesis. Identify and test an ADR candidate only when `$maintain-decision-records`' four gates appear satisfied, then include its rationale and the user's explicit durable-record agreement in the final synthesis. Do not invoke `$maintain-decision-records` before that whole synthesis is confirmed, and reject routine, discoverable, or section-level choices as candidates. ### 5. Walk representative change @@ -52,13 +52,13 @@ Walk representative current and likely capability changes from product intent th ### 6. Confirm the complete synthesis -Have `$elicit` present one concise, complete synthesis covering drivers and context, responsibilities and dependency direction, technology and data strategy, integrations and material cross-cutting constraints, current-versus-target distinctions, trade-offs, evolution constraints, unresolved material risks, and every proposed ADR rationale. Obtain explicit user confirmation of the whole synthesis. Create no ADR record from any candidate until the complete synthesis, or a revised synthesis after rejection, is confirmed. If the user rejects or corrects it, leave `ARCHITECTURE.md` unchanged, reopen only affected evidence-backed branches, repeat the relevant change walks, and reconfirm the complete revised synthesis. The confirmed synthesis supplies authority for its agreed document writes without a second permission prompt. +Have `$elicit-with-context` present one concise, complete synthesis covering drivers and context, responsibilities and dependency direction, technology and data strategy, integrations and material cross-cutting constraints, current-versus-target distinctions, trade-offs, evolution constraints, unresolved material risks, and every proposed ADR rationale. Obtain explicit user confirmation of the whole synthesis. Create no ADR record from any candidate until the complete synthesis, or a revised synthesis after rejection, is confirmed. If the user rejects or corrects it, leave `ARCHITECTURE.md` unchanged, reopen only affected evidence-backed branches, repeat the relevant change walks, and reconfirm the complete revised synthesis. The confirmed synthesis supplies authority for its agreed document writes without a second permission prompt. ### 7. Record the architecture Create or update the single root `ARCHITECTURE.md` using the [Architecture Foundation Template](assets/architecture-template.md) when the project has no stronger convention. Preserve unrelated confirmed content and local structure when it differs only stylistically. Omit empty or immaterial sections rather than filling boilerplate. Use prose, tables, or repository-native diagrams only when they materially clarify relationships. Keep canonical term meanings in `CONTEXT.md`, external evidence in research reports, and exceptional rationale in qualifying ADRs instead of duplicating them. -After writing `ARCHITECTURE.md`, invoke `$maintain-context` for every confirmed qualifying ADR and for confirmed terminology not already applied. The prior architecture confirmation supplies explicit agreement only for ADR rationales included in that synthesis. +After writing `ARCHITECTURE.md`, invoke `$maintain-decision-records` for every confirmed qualifying ADR and `$maintain-ubiquitous-language` for confirmed terminology not already applied. The prior architecture confirmation supplies explicit agreement only for ADR rationales included in that synthesis. ### 8. Verify and stop diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 38f7ea3..e536b72 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -18,9 +18,9 @@ Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, ap ### 2. Establish the product frame -Invoke `$elicit` for material user-held knowledge and decisions throughout steps 2–5. Reuse the confirmed synthesis from a prior free-form `$elicit` session as resolved branches and ask only about gaps exposed by the product evidence. Establish concise executive orientation: purpose and vision; intended users, actors, and their needs and desired outcomes; value and meaningful alternatives or market position when applicable; product and business outcomes with observable success signals; business model or pricing only when material; and durable boundaries and non-goals. Derive discoverable facts before asking. The concise product frame is complete enough to orient the whole-product pass. +Invoke `$elicit-with-context` for material user-held knowledge and decisions throughout steps 2–5. Reuse the confirmed synthesis from a prior free-form `$elicit-with-context` session as resolved branches and ask only about gaps exposed by the product evidence. Establish concise executive orientation: purpose and vision; intended users, actors, and their needs and desired outcomes; value and meaningful alternatives or market position when applicable; product and business outcomes with observable success signals; business model or pricing only when material; and durable boundaries and non-goals. Derive discoverable facts before asking. The concise product frame is complete enough to orient the whole-product pass. -Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. +Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. Invoke `$research` when a material external subject could change the product decision and needs durable evidence; read and link its report rather than copying its findings wholesale. Use an incidental lookup when the evidence need not persist. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would answer one bounded visual question better than discussion. Research and prototype results return evidence to this product session; neither owns `PRODUCT.md` or continuation. @@ -36,7 +36,7 @@ Clarify every mapped capability only to mini-brief depth: who benefits and the v ### 5. Confirm the whole-product synthesis -Walk every journey and capability across the relevant user perspectives. Resolve material omissions, duplicates, false groupings, broken transitions, contradictions, and capability descriptions that conceal more than one product behaviour. Have `$elicit` present one complete synthesis of the product frame, journeys, and capability catalogue and obtain the user's explicit confirmation; if the user rejects or corrects it, reopen the affected product work and reconfirm the complete revised synthesis. The confirmed foundation is coherent and ready to persist. +Walk every journey and capability across the relevant user perspectives. Resolve material omissions, duplicates, false groupings, broken transitions, contradictions, and capability descriptions that conceal more than one product behaviour. Have `$elicit-with-context` present one complete synthesis of the product frame, journeys, and capability catalogue and obtain the user's explicit confirmation; if the user rejects or corrects it, reopen the affected product work and reconfirm the complete revised synthesis. The confirmed foundation is coherent and ready to persist. ### 6. Write the product definition @@ -44,4 +44,4 @@ After explicit confirmation, create or update the single root `PRODUCT.md` using ### 7. Verify the definition -Verify current claims against inspected evidence, direction and ideas against user confirmation, and material external claims against linked research. Check coverage from frame to journeys to every capability mini-brief, coherence across user perspectives, consistency with `CONTEXT.md`, preservation of unrelated confirmed content, and absence of architecture, detailed feature behaviour, priorities, roadmap, releases, tickets, or implementation detail. Verify every `$maintain-context` result and report changed files, evidence, and limitations. Stop after `PRODUCT.md` and applicable ubiquitous language are updated and verified; later architecture and feature definition remain separate sessions. +Verify current claims against inspected evidence, direction and ideas against user confirmation, and material external claims against linked research. Check coverage from frame to journeys to every capability mini-brief, coherence across user perspectives, consistency with `CONTEXT.md`, preservation of unrelated confirmed content, and absence of architecture, detailed feature behaviour, priorities, roadmap, releases, tickets, or implementation detail. Report changed files, evidence, and limitations. Stop after `PRODUCT.md` and applicable ubiquitous language are updated and verified; later architecture and feature definition remain separate sessions. diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index a90949e..79b3368 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -1,6 +1,6 @@ # Discovery Techniques -Load only the section needed by the active discovery branch. `$elicit` owns material questioning and final synthesis confirmation; `$maintain-context` owns glossary persistence; `$research` owns durable external evidence; `$prototype` owns a bounded disposable visual experiment. `define-product` retains the product outcome and stopping boundary. +Load only the section needed by the active discovery branch. `$elicit-with-context` coordinates material questioning, final synthesis confirmation, and glossary maintenance; its composed skills retain their own results. `$research` owns durable external evidence; `$prototype` owns a bounded disposable visual experiment. `define-product` retains the product outcome and stopping boundary. ## Establish an existing product diff --git a/skills/design-feature/SKILL.md b/skills/design-feature/SKILL.md index d9632a1..ce973ca 100644 --- a/skills/design-feature/SKILL.md +++ b/skills/design-feature/SKILL.md @@ -24,9 +24,9 @@ Inspect the request, prerequisite authorities, existing paired solution design, Separate confirmed future behaviour, architecture constraints, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide decision is absent or the feature would require changing the architecture, identify the exact architecture gap and stop rather than embedding a new foundation in this design. -Invoke `$elicit` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Invoke `$prototype` only when one bounded visual, interactive, or technically observable uncertainty can be discriminated by its disposable artefact. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. +Invoke `$elicit-with-context` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Invoke `$prototype` only when one bounded visual, interactive, or technically observable uncertainty can be discriminated by its disposable artefact. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. -Invoke `$maintain-context` inline whenever consequential project terminology is independently confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update; this skill retains the solution-design outcome and verification. Apply a glossary change before whole-design confirmation only when its meaning is confirmed independently of the proposed solution. Keep terminology that depends on an unconfirmed design choice unchanged until that choice is confirmed. +Invoke `$maintain-ubiquitous-language` inline whenever consequential project terminology is independently confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update; this skill retains the solution-design outcome and verification. Apply a glossary change before whole-design confirmation only when its meaning is confirmed independently of the proposed solution. Keep terminology that depends on an unconfirmed design choice unchanged until that choice is confirmed. The solution-design authority, runtime evidence, unresolved gaps, and permitted decision routes are explicit before solution selection begins. @@ -42,7 +42,7 @@ The affected architecture, design forces, risks, and two-way traceability are co Compare viable system-specific alternatives for every material choice. Prefer the smallest coherent complete solution that follows established project and framework idioms, keeps knowledge with its natural owner, offers deep stable contracts, localises plausible change, and avoids speculative abstraction. Consider the consequences for the relevant forces, failure modes, migration, operations, verification, and likely evolution. Do not let current code win by default when it conflicts with the confirmed target, and do not introduce a new seam or component without evidenced variation, ownership, boundary, migration, or observation value. -Resolve material trade-offs with the user through `$elicit` when invoked. Select one internally coherent solution and record each material selected option and its consequences. Retain rejected alternatives only when they are needed to understand a consequential trade-off, risk, or qualifying ADR; avoid alternative essays. +Resolve material trade-offs with the user through `$elicit-with-context` when invoked. Select one internally coherent solution and record each material selected option and its consequences. Retain rejected alternatives only when they are needed to understand a consequential trade-off, risk, or qualifying ADR; avoid alternative essays. The selected solution is justified by runtime constraints and confirmed decisions rather than familiarity or template completeness. @@ -74,7 +74,7 @@ The selected solution survives its material operating, change, failure, transiti Present one concise, self-contained synthesis covering authorities, drivers, selected solution, affected responsibilities and dependency direction, interfaces and contracts, data, integrations, material cross-cutting concerns, migration and compatibility, operations, verification, implementation dependencies, risks, trade-offs, and both traceability directions. Include every proposed terminology update and ADR rationale. -Identify an ADR candidate only when `$maintain-context`'s four gates appear to pass and the user agrees that the durable rationale is warranted. Most feature choices belong only in the solution design. Obtain a new explicit response confirming the whole synthesis; selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the solution design and ADRs unchanged, leave terminology proposed only within the rejected synthesis unapplied, reopen every affected evidence-backed choice, repeat the applicable traces and scenarios, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid after rejection. +Identify an ADR candidate only when `$maintain-decision-records`' four gates appear to pass and the user agrees that the durable rationale is warranted. Most feature choices belong only in the solution design. Obtain a new explicit response confirming the whole synthesis; selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the solution design and ADRs unchanged, leave terminology proposed only within the rejected synthesis unapplied, reopen every affected evidence-backed choice, repeat the applicable traces and scenarios, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid after rejection. The confirmed synthesis supplies authority for the agreed writes without another application prompt. @@ -84,7 +84,7 @@ Follow the project's established feature-record convention. Otherwise write besi Record the selected how and its consequences at design precision. Use a small example only when indispensable to make a contract unambiguous. Do not rewrite the specification's what or why, duplicate live implementation state, prescribe file-by-file or method-by-method work, or include tickets, code scaffolding, estimates, priorities, releases, or deployment execution. -After writing the solution design, invoke `$maintain-context` for every confirmed qualifying ADR and reconcile confirmed terminology not already applied inline. The prior whole-design confirmation supplies authority for design-dependent terminology and explicit durable-rationale agreement only for ADRs included in that synthesis. It alone owns `CONTEXT.md` and ADR files; this skill retains the solution-design outcome and verification. +After writing the solution design, invoke `$maintain-decision-records` for every confirmed qualifying ADR and `$maintain-ubiquitous-language` for confirmed terminology not already applied inline. The prior whole-design confirmation supplies authority for design-dependent terminology and explicit durable-rationale agreement only for ADRs included in that synthesis. Each utility owns only its context artifact; this skill retains the solution-design outcome and verification. The approved solution and only its warranted durable context are recorded in their owning artifacts. @@ -92,6 +92,6 @@ The approved solution and only its warranted durable context are recorded in the Verify every material specification behaviour and acceptance condition has a design responsibility and verification seam, and every introduced responsibility, component, contract, data shape, migration, integration, operational obligation, and verification mechanism traces back to a specification or architecture need. Rewalk the main, adverse, evolution, migration, and rollback scenarios that apply. Check responsibility ownership, contract compatibility, dependency direction, data lifecycle, integration failure handling, cross-cutting constraints, operational diagnosability, safe transition, and implementation decomposability for internal coherence. -Verify statements against their authorities, links and paired path, preservation of unrelated content, and every `$maintain-context` result. Confirm the design has not changed product behaviour, smuggled in a system-wide architecture decision, duplicated live state, created tickets, or become code or file-level implementation instructions. +Verify statements against their authorities, links and paired path, preservation of unrelated content, and every context-maintenance result. Confirm the design has not changed product behaviour, smuggled in a system-wide architecture decision, duplicated live state, created tickets, or become code or file-level implementation instructions. Report changed files, authorities and supporting evidence, selected solution and material trade-offs, verification performed, and unresolved limitations. Stop after the confirmed solution design and applicable context records are written and verified; do not create tickets, estimate, prioritise, scaffold, implement, commit, publish, release, or deploy. diff --git a/skills/elicit-with-context/SKILL.md b/skills/elicit-with-context/SKILL.md new file mode 100644 index 0000000..814fd85 --- /dev/null +++ b/skills/elicit-with-context/SKILL.md @@ -0,0 +1,15 @@ +--- +name: elicit-with-context +description: Elicits shared understanding while maintaining software-project context. Use when a software-project skill needs user-held information or decisions. +metadata: + invocation: user +disable-model-invocation: true +--- + +# Elicit with Context + +Elicits shared understanding while maintaining project context. + +## Process + +Invoke `$elicit` for questioning, applying `$maintain-ubiquitous-language` and `$maintain-decision-records` throughout. diff --git a/skills/elicit-with-context/agents/openai.yaml b/skills/elicit-with-context/agents/openai.yaml new file mode 100644 index 0000000..46f2204 --- /dev/null +++ b/skills/elicit-with-context/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Elicit with Context' + short_description: 'Elicit while maintaining project context' +policy: + allow_implicit_invocation: false diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 916f6ac..796d97d 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -14,11 +14,11 @@ A tailored **Software Formal Inspection** lifecycle implements one confirmed bod ### 1. Fix the outcome -Inspect the stated ticket or agreed work, repository instructions, linked product and technical authorities, current code and tests, and the complete repository state. A clear small request is sufficient authority; do not manufacture a ticket, specification, or other ceremony. Invoke `$elicit` one question at a time only when user-held ambiguity could materially change behaviour, contracts, architecture, scope, or acceptance. State the observable outcome, exclusions, applicable structural constraints, material risks, and evidence expected before editing. The implementation basis is explicit and bounded. +Inspect the stated ticket or agreed work, repository instructions, linked product and technical authorities, current code and tests, and the complete repository state. A clear small request is sufficient authority; do not manufacture a ticket, specification, or other ceremony. Invoke `$elicit-with-context` one question at a time only when user-held ambiguity could materially change behaviour, contracts, architecture, scope, or acceptance. State the observable outcome, exclusions, applicable structural constraints, material risks, and evidence expected before editing. The implementation basis is explicit and bounded. ### 2. Build the smallest complete change -Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-context` when independently confirmed durable terminology arises or a rare qualifying decision needs durable rationale; each utility returns control to this outcome. +Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-ubiquitous-language` when independently confirmed durable terminology arises and `$maintain-decision-records` when a rare qualifying decision needs durable rationale; each utility returns control to this outcome. Invoke `$tdd` only when an existing runnable test suite can credibly exercise the requested observable behaviour at a stable boundary with an independent oracle. When evidence already shows any prerequisite is absent, keep control here rather than invoking TDD merely to receive that result. Let TDD own the Red-Green-Refactor cycles and returned code and tests when it applies. Otherwise implement through the smallest conventional path and use the strongest available project-native feedback; update existing tests only when they can meaningfully express changed behaviour. Do not install or invent a test or quality framework to satisfy this process. When meaningful infrastructure is absent, preserve the limitation for verification and recommend separately planned work rather than adding unrelated harness infrastructure. @@ -35,7 +35,7 @@ Validate every Standards and Spec finding against the confirmed request, its aut - A rejected finding is falsified, unsupported, superseded, or outside the confirmed scope. - A user decision is required when accepting or rejecting the finding would change behaviour, contracts, architecture, or scope, or when its evidence cannot be obtained safely. -For a user-decision finding, invoke `$elicit` and stop mutation, further adjudication, and verification until its confirmed synthesis returns. Resume from the fixed outcome; when the decision changes the candidate or any review authority, freeze the resulting basis and repeat independent review. +For a user-decision finding, invoke `$elicit-with-context` and stop mutation, further adjudication, and verification until its confirmed synthesis returns. Resume from the fixed outcome; when the decision changes the candidate or any review authority, freeze the resulting basis and repeat independent review. Treat priority as evidence of impact and ordering, not remediation authority. Apply every accepted correction through `$tdd` when its prerequisite applies and otherwise through the conventional path. After any candidate change, freeze the complete new candidate and repeat independent review. Continue until every finding has a supported disposition, no required finding remains unresolved, and the complete current candidate has been independently reviewed. The reviewed candidate is fixed and current. diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index cb6567e..3b957f4 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -14,7 +14,7 @@ disable-model-invocation: false ### 1. Establish the instruction surface -Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Treat a caller handoff carrying an agreed preference as authority; do not ask for duplicate permission to record it. Inspect repository evidence only far enough to recover intent, scope, inferability, conflicts, and referenced commands. Invoke `$elicit` only when material authority, scope, or conflict remains unresolved. The authorised candidates and active constraints are explicit. +Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Treat a caller handoff carrying an agreed preference as authority; do not ask for duplicate permission to record it. Inspect repository evidence only far enough to recover intent, scope, inferability, conflicts, and referenced commands. Invoke `$elicit-with-context` only when material authority, scope, or conflict remains unresolved. The authorised candidates and active constraints are explicit. ### 2. Allocate the guidance diff --git a/skills/maintain-context/SKILL.md b/skills/maintain-context/SKILL.md deleted file mode 100644 index d2aaeec..0000000 --- a/skills/maintain-context/SKILL.md +++ /dev/null @@ -1,44 +0,0 @@ ---- -name: maintain-context -description: Maintains confirmed project terminology and rare ADRs continuously. Use when project language is introduced, changed, misused, or contradicted, or an accepted decision needs durable rationale. -metadata: - invocation: model -disable-model-invocation: false ---- - -# Maintain Context - -Adapted **Ubiquitous Language** keeps confirmed project-specific terminology consistent. Maintain one concise root `CONTEXT.md` glossary inline during ordinary work and, exceptionally, preserve the rationale for a qualifying accepted decision in an ADR. - -## Process - -### 1. Inspect the language - -When project language is introduced, changed, misused, or contradicted, inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the code or product evidence needed to understand that language. Challenge inconsistent uses of established terms. The candidate term, current meaning, and any contradiction are explicit without loading unrelated project context. - -### 2. Resolve the meaning - -Accept terminology already established by authoritative project evidence or explicitly agreed with the user. A caller's agreed in-scope update supplies authority to persist that meaning without another permission prompt. When a material meaning, conflict, rename, or deprecation remains unresolved, return it to an active `$elicit` caller or invoke `$elicit` when none is active, and leave durable context, including the existing entry, unchanged until agreement. One meaning is confirmed or the unresolved decision returns to the caller. - -### 3. Update the glossary inline - -As confirmed language arises, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and the document's local format; edit only affected entries and prefer changing an existing entry to adding a duplicate. Reconcile a confirmed change across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence, and remove obsolete wording only when its replacement is confirmed and losing it will not obscure that evidence. - -Give each canonical term one short plain-language definition, normally one or two sentences, using project-native wording. Include only durable project-specific terms, distinctions, roles, boundaries, or aliases that future agents need to interpret the project consistently; group them naturally when useful. `CONTEXT.md` explains what the language means. It does not store skill procedures, generic guidance, task status, roadmaps, volatile implementation facts, architecture inventories, schemas or endpoints, acceptance criteria, or anything reliably discoverable from code. Keep product intent in `PRODUCT.md`, system structure in `ARCHITECTURE.md`, feature behaviour in its specification, and implementation detail in code. The glossary is current without duplicating another authority. - -### 4. Gate a rare ADR - -Consider an ADR only for an accepted technical or architectural decision that passes every gate: - -- changing it later has meaningful cost; -- a future reader would find it surprising without context; and -- viable alternatives created a genuine trade-off; and -- no existing durable authority is the better owner, including `PRODUCT.md`, `ARCHITECTURE.md`, a feature document, or a dedicated governance record. - -When every gate passes, obtain the user's explicit agreement to preserve the rationale durably. Explicit agreement recorded in the relevant conversation or caller handoff satisfies this gate without another prompt. A decision that misses a gate, lacks that agreement, or is routine or discoverable produces no ADR. The exceptional record has a confirmed owner and purpose. - -### 5. Record and verify the result - -When the user agrees, follow the project's existing ADR convention; when none exists, create `docs/architecture/decisions/` lazily. Write the next record from the [ADR template](assets/adr-template.md) and name it `NNNN-short-title.md`. Use `0001` when no sibling filename matches `NNNN-*.md`; otherwise add one to the maximum four-digit `NNNN` among matching siblings. Keep exactly three substantive sections in order: `Decision`, `Context`, and `Ramifications`. Record the accepted choice, why it was needed, only the rejected or constrained alternatives necessary to explain its trade-offs, and its material consequences. - -Re-read each changed context artifact against the confirmed meaning, accepted decision, and relevant evidence. Return the changed files or why no durable update was warranted, plus unresolved terminology for the caller to resolve. Change no caller-owned product, architecture, feature, code, external state, or verification outcome. The caller receives a concise context update and retains its surrounding outcome. diff --git a/skills/maintain-context/agents/openai.yaml b/skills/maintain-context/agents/openai.yaml deleted file mode 100644 index 8ff35b5..0000000 --- a/skills/maintain-context/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Maintain Context' - short_description: 'Maintain project language and rare ADRs' -policy: - allow_implicit_invocation: true diff --git a/skills/maintain-decision-records/SKILL.md b/skills/maintain-decision-records/SKILL.md new file mode 100644 index 0000000..eff61ef --- /dev/null +++ b/skills/maintain-decision-records/SKILL.md @@ -0,0 +1,34 @@ +--- +name: maintain-decision-records +description: Maintains rare architecture decision records. Use when explicitly invoked to test and preserve an accepted technical or architectural decision's durable rationale. +disable-model-invocation: true +--- + +# Maintain Decision Records + +Preserves the rationale for exceptional accepted technical or architectural decisions in concise Architecture Decision Records. + +## Process + +### 1. Inspect the decision + +Inspect the accepted decision, relevant conversation or caller handoff, existing `docs/adr/` records, applicable repository guidance, and only the product, architecture, feature, governance, or implementation evidence needed to establish its owner and rationale. Do not scan unrelated decisions. The candidate decision, alternatives, consequences, existing coverage, and user authority are explicit. + +### 2. Gate the record + +Create an ADR only when the accepted technical or architectural decision passes every gate: + +- changing it later has meaningful cost; +- a future reader would find it surprising without context; +- viable alternatives created a genuine trade-off; and +- no existing durable authority already records the decision or is its better owner, including `PRODUCT.md`, `ARCHITECTURE.md`, a feature document, or a dedicated governance record. + +Require the user's explicit agreement to preserve the rationale durably. Agreement recorded in the relevant conversation or caller handoff satisfies this gate without another prompt. When every other gate passes but agreement is absent, ask the user one direct question. A decision that misses a gate, lacks agreement, or is routine or discoverable produces no ADR. + +### 3. Write the ADR + +Create `docs/adr/` lazily and write the next record from the [ADR template](assets/adr-template.md). Name it `NNNN-decision-shaped-slug.md`: use `0001` when no sibling matches `NNNN-*.md`; otherwise add one to the largest four-digit prefix. Keep exactly three substantive sections in order: `Decision`, `Context`, and `Consequences`. Record the accepted choice, why it was needed, only the rejected or constrained alternatives necessary to explain its trade-offs, and its material consequences. + +### 4. Verify the result + +Re-read the ADR against the accepted decision, every gate, the user's agreement, and relevant evidence. Return the created file or why no ADR was warranted. Change no caller-owned product, architecture, feature, code, external state, or verification outcome. diff --git a/skills/maintain-decision-records/agents/openai.yaml b/skills/maintain-decision-records/agents/openai.yaml new file mode 100644 index 0000000..29d96ca --- /dev/null +++ b/skills/maintain-decision-records/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Maintain Decision Records' + short_description: 'Maintain rare architecture decisions' +policy: + allow_implicit_invocation: false diff --git a/skills/maintain-context/assets/adr-template.md b/skills/maintain-decision-records/assets/adr-template.md similarity index 95% rename from skills/maintain-context/assets/adr-template.md rename to skills/maintain-decision-records/assets/adr-template.md index a9f181a..bf22119 100644 --- a/skills/maintain-context/assets/adr-template.md +++ b/skills/maintain-decision-records/assets/adr-template.md @@ -8,6 +8,6 @@ {Explain why the decision was needed and only the rejected or constrained alternatives necessary to understand its trade-offs.} -## Ramifications +## Consequences {State the significant consequences, trade-offs, and conditions that could trigger reconsideration.} diff --git a/skills/maintain-ubiquitous-language/SKILL.md b/skills/maintain-ubiquitous-language/SKILL.md new file mode 100644 index 0000000..1e4d9da --- /dev/null +++ b/skills/maintain-ubiquitous-language/SKILL.md @@ -0,0 +1,25 @@ +--- +name: maintain-ubiquitous-language +description: Maintains confirmed project-specific language in CONTEXT.md. Use when explicitly invoked to add, change, challenge, rename, deprecate, or reconcile a project term. +disable-model-invocation: true +--- + +# Maintain Ubiquitous Language + +Adapted **Ubiquitous Language** keeps project-specific terminology consistent in one root `CONTEXT.md` glossary. + +## Process + +### 1. Inspect the language + +Inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the project evidence needed to understand the introduced, changed, misused, or contradicted language. Challenge inconsistent uses of established terms. Make the candidate term, current meaning, and any contradiction explicit without loading unrelated project context. + +### 2. Resolve the meaning + +Accept terminology established by authoritative project evidence or explicitly agreed with the user. A caller's agreed in-scope update supplies authority to persist that meaning without another permission prompt. When a material meaning, conflict, rename, or deprecation remains unresolved, ask the user one direct question that distinguishes the current and proposed meanings and recommends one only when the evidence supports it. Leave the existing glossary unchanged until the answer confirms one meaning. + +### 3. Update the glossary inline + +As confirmed language arises, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and the document's local format; edit only affected entries and prefer changing an existing entry to adding a duplicate. Reconcile a confirmed change across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence, and remove obsolete wording only when its replacement is confirmed and losing it will not obscure that evidence. + +Keep `CONTEXT.md` as a glossary of project-specific domain language. Give each canonical term one short paragraph, normally one or two sentences, using project-native wording; group terms by domain only when useful. Leave implementation and specifications to their authoritative sources. diff --git a/skills/maintain-ubiquitous-language/agents/openai.yaml b/skills/maintain-ubiquitous-language/agents/openai.yaml new file mode 100644 index 0000000..3a8cfa7 --- /dev/null +++ b/skills/maintain-ubiquitous-language/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Maintain Ubiquitous Language' + short_description: 'Maintain shared project terminology' +policy: + allow_implicit_invocation: false diff --git a/skills/maintain-context/assets/context-template.md b/skills/maintain-ubiquitous-language/assets/context-template.md similarity index 100% rename from skills/maintain-context/assets/context-template.md rename to skills/maintain-ubiquitous-language/assets/context-template.md diff --git a/skills/plan-iteration/SKILL.md b/skills/plan-iteration/SKILL.md index 9002954..72cdd9e 100644 --- a/skills/plan-iteration/SKILL.md +++ b/skills/plan-iteration/SKILL.md @@ -30,7 +30,7 @@ The planning horizon, native representation, existing membership, and capability Read current project-native evidence for the horizon: comparable historical throughput or sizing where it exists, current commitments, team availability, specialist constraints, dependency timing, operational load, and other work that consumes the same capacity. Never invent velocity, estimates, dates, availability, or precision, and do not treat points or item counts as interchangeable capacity unless the project's own evidence supports that use. -When evidence cannot resolve a material capacity, horizon, or trade-off decision, invoke `$elicit` with the evidence and viable consequences. Let it ask one user-authoritative question at a time and use only its confirmed synthesis; retain ownership of selection, mutation, verification, and stopping. State the weakest honest capacity bound when precise capacity is unavailable rather than manufacturing a number. +When evidence cannot resolve a material capacity, horizon, or trade-off decision, invoke `$elicit-with-context` with the evidence and viable consequences. Let it ask one user-authoritative question at a time and use only its confirmed synthesis; retain ownership of selection, mutation, verification, and stopping. State the weakest honest capacity bound when precise capacity is unavailable rather than manufacturing a number. Refresh each candidate's observable scope, current priority, readiness for its next implementation step, dependencies, blockers, and current native state. Treat a relationship as blocking only when the dependency meaning and evidence show the item cannot begin or complete without it. Do not confuse preference, display order, or convenient sequence with a blocker. The eligible set, exclusions with reasons, and capacity assumptions and limitations are current and explicit. diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md index 4844ad5..859fb5e 100644 --- a/skills/plan-retirement/SKILL.md +++ b/skills/plan-retirement/SKILL.md @@ -20,9 +20,9 @@ Treat code, tests, contracts, configuration, telemetry, inventories, tracker rec ### 2. Resolve retirement intent and obligations -Invoke `$elicit` and retain responsibility for the plan, writes, verification, and stopping boundary. Give it the inspected evidence and use one dependency-ordered question at a time to resolve the obsolete capability and reason for retiring it; acceptable end state and non-goals; timing or obligation trade-offs; consumer and operator migration; replacement expectations; compatibility and coexistence; data retention, deletion, archival, access, and recovery; integration and provider obligations; communication and support; security, privacy, compliance, and contractual constraints; acceptable reversibility and risk; ownership; and observable completion. Reuse compatible confirmed answers and ask only about material gaps. +Invoke `$elicit-with-context` and retain responsibility for the plan, writes, verification, and stopping boundary. Give it the inspected evidence and use one dependency-ordered question at a time to resolve the obsolete capability and reason for retiring it; acceptable end state and non-goals; timing or obligation trade-offs; consumer and operator migration; replacement expectations; compatibility and coexistence; data retention, deletion, archival, access, and recovery; integration and provider obligations; communication and support; security, privacy, compliance, and contractual constraints; acceptable reversibility and risk; ownership; and observable completion. Reuse compatible confirmed answers and ask only about material gaps. -Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation subject could change the plan and needs durable evidence. Read its report before using its findings; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-context` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. +Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation subject could change the plan and needs durable evidence. Read its report before using its findings; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-ubiquitous-language` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. Keep every retirement-plan artefact unchanged until the complete synthesis is explicitly confirmed. The intended retirement outcome, authoritative obligations, unresolved decisions, and conditions that must gate irreversible work are honest and explicit. @@ -49,7 +49,7 @@ Trace every stage and obligation back to retirement intent, a consumer or depend ### 5. Confirm the complete synthesis -Have `$elicit` present one concise, self-contained synthesis covering the evidence boundary and unknown use, retirement outcome and non-goals, affected parties and obligations, alternatives, selected strategy, stages and ownership, migration and compatibility, communications and support, data and integrations, operations, security and compliance, verification, abort and recovery, completion conditions, unresolved risks, context changes, durable destination, and exact proposed write. Obtain explicit confirmation of the whole synthesis; selecting an earlier option is not confirmation. +Have `$elicit-with-context` present one concise, self-contained synthesis covering the evidence boundary and unknown use, retirement outcome and non-goals, affected parties and obligations, alternatives, selected strategy, stages and ownership, migration and compatibility, communications and support, data and integrations, operations, security and compliance, verification, abort and recovery, completion conditions, unresolved risks, context changes, durable destination, and exact proposed write. Obtain explicit confirmation of the whole synthesis; selecting an earlier option is not confirmation. If the user rejects or revises it, leave the retirement plan and design-dependent context unchanged, reopen every affected evidence, obligation, option, stage, owner, and completion branch, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid. The confirmed synthesis supplies authority for its agreed plan and context writes without another application prompt. @@ -59,6 +59,6 @@ Immediately before writing, refresh mutable usage, dependency, contract, provide Follow the project's established retirement-record convention. Otherwise choose a stable lowercase capability slug and create `docs/retirements/.md` using the [Retirement Plan Template](assets/retirement-plan-template.md). Preserve unrelated confirmed content and local structure when refining an existing plan. Include only material sections, link authorities and mutable owning systems instead of duplicating them, and distinguish confirmed current evidence, approved future action, and unresolved gated decisions. -After writing the plan, invoke `$maintain-context` for confirmed design-dependent terminology and any qualifying ADR whose durable-rationale agreement was included in the synthesis. Verify its returned changes. Re-read the plan against product intent, repository and runtime evidence, context and architecture authorities, external research, and the confirmed synthesis. Verify all paths and links; end-to-end dependent coverage; stage entry, exit, evidence, abort, rollback and recovery; owner and obligation coverage; data, integration, communication, support, security and compliance treatment; removal of temporary paths; and observable completion. State evidence limitations and never describe unknown or future evidence as established. +After writing the plan, invoke `$maintain-ubiquitous-language` for confirmed design-dependent terminology and `$maintain-decision-records` for any qualifying ADR whose durable-rationale agreement was included in the synthesis. Re-read the plan against product intent, repository and runtime evidence, context and architecture authorities, external research, and the confirmed synthesis. Verify all paths and links; end-to-end dependent coverage; stage entry, exit, evidence, abort, rollback and recovery; owner and obligation coverage; data, integration, communication, support, security and compliance treatment; removal of temporary paths; and observable completion. State evidence limitations and never describe unknown or future evidence as established. Report the plan path, selected strategy, affected parties and obligations, stage and completion outline, confirmed context changes, supporting evidence, verification, unresolved gated decisions, and limitations. Stop after the approved retirement plan and applicable context records are written and verified. Do not create tracker items, invoke downstream lifecycle skills, implement or decommission the capability, disable or delete code, data, configuration or infrastructure, contact consumers, commit, push, open a pull request, release, deploy, or continuously monitor. Later sessions own `$create-tickets`, `$implement`, `$commit`, and `$pull-request`. diff --git a/skills/prototype/SKILL.md b/skills/prototype/SKILL.md index 1a6ed84..dd02e72 100644 --- a/skills/prototype/SKILL.md +++ b/skills/prototype/SKILL.md @@ -14,7 +14,7 @@ Creates the least elaborate temporary visual experiment that can answer one boun ### 1. Bound the experiment -Inspect the request, caller outcome and authority, repository guidance, relevant authentic runtime context, intended evaluators, and safely available data. State one question and the observable reaction, comparison, or result that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the decision criteria and keep the choice with the user. Invoke `$elicit` only when user-held information is needed to bound this visual question; return broader discovery to the caller. When a visual artefact cannot discriminate the question, stop and return that boundary without prototyping. The question, audience, criteria, decision authority, and discriminators are explicit. +Inspect the request, caller outcome and authority, repository guidance, relevant authentic runtime context, intended evaluators, and safely available data. State one question and the observable reaction, comparison, or result that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the decision criteria and keep the choice with the user. Invoke `$elicit-with-context` only when user-held information is needed to bound this visual question; return broader discovery to the caller. When a visual artefact cannot discriminate the question, stop and return that boundary without prototyping. The question, audience, criteria, decision authority, and discriminators are explicit. ### 2. Confirm presentation and disposal diff --git a/skills/refine-backlog/SKILL.md b/skills/refine-backlog/SKILL.md index 3abddfa..77fa5a8 100644 --- a/skills/refine-backlog/SKILL.md +++ b/skills/refine-backlog/SKILL.md @@ -50,7 +50,7 @@ Every item has an evidence-led health assessment, relative-priority basis or ope Prepare the smallest native change set that makes the bounded backlog current. It may clarify an item, split supported scope, reconcile or close obsolete, duplicate, invalid, or fully superseded work with evidence, correct relationships and blocker directions, set an agreed native priority or order, and represent honest readiness. Preserve a no-op when current state already expresses the supported result. -Invoke `$elicit` only when a material scope, priority, or readiness decision depends on user-held knowledge or authority. Supply inspected evidence, viable options, and consequences; let it ask one question at a time and use only its confirmed synthesis. Retain responsibility for the complete proposal, mutation, verification, and stopping. +Invoke `$elicit-with-context` only when a material scope, priority, or readiness decision depends on user-held knowledge or authority. Supply inspected evidence, viable options, and consequences; let it ask one question at a time and use only its confirmed synthesis. Retain responsibility for the complete proposal, mutation, verification, and stopping. Recheck duplicates and affected relationships, then present one complete proposal containing every create, update, split, reconciliation, closure, link, priority, readiness, and no-op action; exact relationship meanings and directions; evidence-led reasons; later routes; and native capability limitations. Keep the tracker unchanged while decisions remain open. Obtain explicit user confirmation of the whole set. Selecting one earlier option is not confirmation. If the user rejects or revises any part, reopen every affected assessment and relationship, then present the complete revised set. Confirmation authorises exactly the agreed native mutations without another application prompt. diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md index 6cde819..b1fbd22 100644 --- a/skills/review-incident/SKILL.md +++ b/skills/review-incident/SKILL.md @@ -40,7 +40,7 @@ Invoke `$research` only when one material external subject needs durable evidenc Identify only follow-up outcomes that directly address an evidenced contributing condition, recurrence path, defense gap, response weakness, or material learning unknown. For each candidate, state the incident evidence and rationale, the desired outcome rather than an unsupported implementation prescription, one accountable owner, the appropriate later lifecycle route, and observable evidence that will show completion and, where meaningful, effectiveness. Distinguish reducing recurrence likelihood, reducing impact, improving detection or response, and resolving an unknown. -Route a behaviour or product change to a later `$specify-feature` session; a technical solution or decomposition to later `$design-feature` or `$create-tickets` sessions once their authorities exist; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal design path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology or rare accepted rationale to `$maintain-context`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. +Route a behaviour or product change to a later `$specify-feature` session; a technical solution or decomposition to later `$design-feature` or `$create-tickets` sessions once their authorities exist; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal design path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. Challenge candidates against recurrence value, impact reduction, evidence strength, ownership, observability, cost, and overlap with current work. Reuse or link an exact existing outcome, preserve related work as related, merge duplicates, and remove generic reminders, `be more careful` actions, speculative backlog dumping, and action inflation. Zero follow-ups is valid when the evidence supports no useful new outcome. A critical learning action without an owner remains an explicit completion blocker rather than becoming anonymous backlog work. @@ -54,7 +54,7 @@ Choose the owning incident system or established project convention for the dura Present one complete synthesis for confirmation: the factual account, supported interpretations, contradictions and unknowns, contributing conditions and recurrence paths, defenses and response learning, proposed durable owner and exact write, sensitive-evidence treatment, and every follow-up create, update, link, reuse, owner, route, and effectiveness measure. Keep all durable systems unchanged while material review conclusions or ownership remain open. -Obtain explicit user confirmation of the complete synthesis and mutation plan. If the user rejects or revises it, reopen every affected timeline, explanation, condition, action, owner, route, sensitivity, and persistence decision and present the complete revised synthesis. Confirmation authorises exactly the agreed review and tracker writes without another application prompt. Independently confirmed durable project terminology may invoke `$maintain-context` inline; terminology or rationale established by this review waits until the complete synthesis is confirmed. Verify its returned change before continuing. +Obtain explicit user confirmation of the complete synthesis and mutation plan. If the user rejects or revises it, reopen every affected timeline, explanation, condition, action, owner, route, sensitivity, and persistence decision and present the complete revised synthesis. Confirmation authorises exactly the agreed review and tracker writes without another application prompt. Independently confirmed durable project terminology may invoke `$maintain-ubiquitous-language` inline; terminology or rationale established by this review waits until the complete synthesis is confirmed, then invokes its respective context skill. The user and agent share one exact, blameless explanation and owned persistence plan. diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index 5a1f6ba..d3bc3e6 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -16,13 +16,13 @@ Defines one substantial feature or enhancement without selecting its implementat Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, the existing feature specification and paired solution design when present, and the project's feature-record convention. Inspect only the relevant current public contracts, documentation, tests, UI, research, and implementation evidence needed to understand observable behaviour. Treat `PRODUCT.md` as authority for the product foundation, the user as authority for the intended change, and the current system as evidence of present behaviour rather than intended future behaviour. Distinguish current, confirmed future, and unresolved claims; surface contradictions and unknowns without silently resolving them. -Accept one capability from `PRODUCT.md`, a confirmed synthesis returned by a prior `$elicit` session, or another agreed feature idea whose product context is sufficient. Do not manufacture a missing product foundation when the requested specification depends on one; report the exact prerequisite and stop. A direct agreed idea may proceed without a formal catalogue when its actors, value, and product boundary are sufficiently grounded. The source, authority, current evidence, convention, and unresolved feature decisions are explicit. +Accept one capability from `PRODUCT.md`, a confirmed synthesis returned by a prior `$elicit-with-context` session, or another agreed feature idea whose product context is sufficient. Do not manufacture a missing product foundation when the requested specification depends on one; report the exact prerequisite and stop. A direct agreed idea may proceed without a formal catalogue when its actors, value, and product boundary are sufficiently grounded. The source, authority, current evidence, convention, and unresolved feature decisions are explicit. ### 2. Bound one proportionate feature -Invoke `$elicit` throughout steps 2–5 for material user-held knowledge and decisions. Reuse every compatible branch from a prior confirmed synthesis and ask only about gaps exposed by inspection. Bound one coherent feature or enhancement, its problem or opportunity, beneficiaries, desired outcome, relationship to the product capability, and exclusions that separate adjacent behaviour. When inspection shows that the change is small, already understood, and needs no durable intent record, explain why a specification would add ceremony, name `$implement` as the proportionate next session, and stop without writing or invoking it. Invocation alone does not bypass this gate; proceed only when the user has separately established that a durable feature specification is required. The session owns exactly one warranted feature and a clear reason for specifying it. +Invoke `$elicit-with-context` throughout steps 2–5 for material user-held knowledge and decisions. Reuse every compatible branch from a prior confirmed synthesis and ask only about gaps exposed by inspection. Bound one coherent feature or enhancement, its problem or opportunity, beneficiaries, desired outcome, relationship to the product capability, and exclusions that separate adjacent behaviour. When inspection shows that the change is small, already understood, and needs no durable intent record, explain why a specification would add ceremony, name `$implement` as the proportionate next session, and stop without writing or invoking it. Invocation alone does not bypass this gate; proceed only when the user has separately established that a durable feature specification is required. The session owns exactly one warranted feature and a clear reason for specifying it. -Invoke `$maintain-context` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. +Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. Invoke `$research` when a material external, domain, regulatory, or standards subject could change the specification and needs durable evidence. Link its report instead of copying its findings. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would resolve one bounded visual or interactive uncertainty better than discussion. Both return evidence to this session and own neither the specification nor continuation. @@ -46,7 +46,7 @@ Walk the feature across every material actor goal, interaction, state, rule, mai ### 6. Confirm one full synthesis -Have `$elicit` present one concise, self-contained synthesis of the problem or opportunity, desired outcome, actors and goals, scope and exclusions, observable behaviour and rules, states and transitions, material scenarios, acceptance conditions, external constraints, traceability, and unresolved limitations. Obtain explicit confirmation of the whole synthesis. Selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the specification unchanged, reopen every affected evidence-backed branch, repeat the applicable walks and counterexamples, and confirm the complete revised synthesis. The confirmed synthesis supplies authority for the agreed specification write without another application prompt. +Have `$elicit-with-context` present one concise, self-contained synthesis of the problem or opportunity, desired outcome, actors and goals, scope and exclusions, observable behaviour and rules, states and transitions, material scenarios, acceptance conditions, external constraints, traceability, and unresolved limitations. Obtain explicit confirmation of the whole synthesis. Selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the specification unchanged, reopen every affected evidence-backed branch, repeat the applicable walks and counterexamples, and confirm the complete revised synthesis. The confirmed synthesis supplies authority for the agreed specification write without another application prompt. ### 7. Write the specification @@ -54,6 +54,6 @@ Follow the project's established feature-record convention. Otherwise choose a s ### 8. Verify and stop -Verify every statement against its authority: product foundation against `PRODUCT.md`; intended change against the confirmed synthesis; terminology against `CONTEXT.md`; current behaviour against inspected evidence; and material external claims against linked research. Rewalk traceability, actors, rules, states, scenarios, acceptance conditions, exclusions, and constraints for coherence and missing branches. Apply the rebuild test and confirm that acceptance remains observable without CSS classes, DOM or private internals, architecture, endpoints unless externally contractual, detailed data models, algorithms, frameworks, deployment, ticket breakdown, estimates, priorities, roadmaps, or releases. Verify the path, links, preservation of unrelated content, separation from any solution design, and every `$maintain-context` result. +Verify every statement against its authority: product foundation against `PRODUCT.md`; intended change against the confirmed synthesis; terminology against `CONTEXT.md`; current behaviour against inspected evidence; and material external claims against linked research. Rewalk traceability, actors, rules, states, scenarios, acceptance conditions, exclusions, and constraints for coherence and missing branches. Apply the rebuild test and confirm that acceptance remains observable without CSS classes, DOM or private internals, architecture, endpoints unless externally contractual, detailed data models, algorithms, frameworks, deployment, ticket breakdown, estimates, priorities, roadmaps, or releases. Verify the path, links, preservation of unrelated content, and separation from any solution design. Report changed files, supporting evidence, verification, and unresolved limitations. Stop after the approved specification and applicable ubiquitous language are written and verified; do not select a solution, create tickets, scaffold, implement, publish, prioritise, or continue into another lifecycle outcome. diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md index 8f220eb..6352d24 100644 --- a/skills/triage-work/SKILL.md +++ b/skills/triage-work/SKILL.md @@ -30,7 +30,7 @@ Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, `ARCHITECTURE.m Search open and closed tracker work and relevant repository records using source identifiers, authoritative links, actors, outcomes, symptoms, affected concepts, and discriminating details. Classify a candidate internally as exact match, possible overlap, related work, or unrelated. Mark a duplicate only when the evidence establishes the same requested outcome or fault; preserve an uncertain resemblance as related or unresolved and state what would distinguish it. -When a material external subject could change the disposition and needs durable evidence, invoke `$research` with that bounded subject, then read its report before using its findings. Invoke `$maintain-context` only when the evidence independently confirms consequential durable project language or qualifying rationale; a proposed classification or route alone does not establish either. Verify its returned context result before continuing. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. +When a material external subject could change the disposition and needs durable evidence, invoke `$research` with that bounded subject, then read its report before using its findings. Invoke `$maintain-ubiquitous-language` only when the evidence independently confirms consequential durable project language and `$maintain-decision-records` only when it confirms qualifying rationale; a proposed classification or route alone does not establish either. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. ### 3. Choose one honest disposition per item @@ -44,7 +44,7 @@ Classify each item by the lifecycle outcome it actually needs: - link actionable work whose intent and solution are already defined to its existing appropriate work without creating another specification, design, or ticket; or - close answered, invalid, not-actionable, or out-of-scope requests only with a concise reason grounded in inspected evidence. -When evidence cannot yet support one of those outcomes, retain an honest native needs-information or unresolved disposition when available and identify the minimum missing fact and who can supply it. Invoke `$elicit` only when that fact or a materially different classification or disposition depends on user-held knowledge or authority. Supply the evidence and viable consequences, use only its confirmed synthesis, and retain ownership of proposal, mutation, verification, and stopping. Do not ask the user for repository or tracker facts that inspection can establish. +When evidence cannot yet support one of those outcomes, retain an honest native needs-information or unresolved disposition when available and identify the minimum missing fact and who can supply it. Invoke `$elicit-with-context` only when that fact or a materially different classification or disposition depends on user-held knowledge or authority. Supply the evidence and viable consequences, use only its confirmed synthesis, and retain ownership of proposal, mutation, verification, and stopping. Do not ask the user for repository or tracker facts that inspection can establish. For each item, prepare a proposed native disposition containing its classification, concise evidence-led reason, exact duplicate or related links, supported state and field changes, precise next route and required input, and any native capability limitation. Do not add priority, readiness, estimates, iteration placement, ticket decomposition, root cause, solution, or speculative downstream work. Every selected item has one supported disposition or one explicit evidence blocker. From fbdd5a029485e81909338766fd33db423bbb6422 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 15:13:17 +0100 Subject: [PATCH 51/79] chore(skills): remove prototype skill --- README.md | 1 - SKILLS_PLAN.md | 16 ++++---- skills/debug/SKILL.md | 2 +- skills/define-architecture/SKILL.md | 2 +- skills/define-product/SKILL.md | 2 +- skills/define-product/references/DISCOVERY.md | 6 +-- skills/design-feature/SKILL.md | 2 +- skills/implement/SKILL.md | 2 +- skills/prototype/SKILL.md | 41 ------------------- skills/prototype/agents/openai.yaml | 5 --- skills/specify-feature/SKILL.md | 2 +- 11 files changed, 14 insertions(+), 67 deletions(-) delete mode 100644 skills/prototype/SKILL.md delete mode 100644 skills/prototype/agents/openai.yaml diff --git a/README.md b/README.md index ec2d256..9996945 100644 --- a/README.md +++ b/README.md @@ -147,7 +147,6 @@ Utilities are independently invokable and are also composed by lifecycle skills | `$elicit` | Resolve material user-held information and decisions one question at a time | | `$elicit-with-context` | Resolve software-project questions while maintaining shared project language | | `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | -| `$prototype` | Create a disposable visual experiment that resolves one bounded uncertainty | | `$maintain-ubiquitous-language` | Keep confirmed project-specific language current in `CONTEXT.md` | | `$maintain-decision-records` | Preserve rare accepted decisions whose rationale warrants an ADR | | `$modular-design` | Assess boundaries, ownership, contracts, dependencies, and change propagation | diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index e993537..693f354 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -103,7 +103,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A user-confirmed product foundation broad enough to guide later feature definition without becoming a roadmap or delivery plan. - **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, representative repository evidence and applicable external evidence. - **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. -- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research` and `prototype`. +- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research`. - **Stops:** After product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capabilities are documented. It does not define technical architecture, detailed features, priorities or tickets. #### `define-architecture` @@ -111,7 +111,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A user-confirmed technical foundation capable of constraining later feature solutions without duplicating the implementation. - **Inputs:** `PRODUCT.md`, `CONTEXT.md`, existing system evidence and material technical, organisational or regulatory constraints. - **Output:** `ARCHITECTURE.md` and any warranted focused ADRs. -- **Composition:** Requires `elicit-with-context` and `modular-design`; conditionally invokes `research`, `prototype`, `maintain-ubiquitous-language` and `maintain-decision-records`. +- **Composition:** Requires `elicit-with-context` and `modular-design`; conditionally invokes `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After the system context, principal boundaries, ownership, technology strategy, data, integrations and cross-cutting quality approach are established. It does not scaffold the system or create implementation work. ### Define @@ -121,7 +121,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** One approved feature or enhancement expressed as externally observable intent. - **Inputs:** A capability from `PRODUCT.md`, an already-elicited conversation or another agreed feature idea, plus `CONTEXT.md`, relevant current behaviour and supporting evidence. - **Output:** `docs/features//specification.md`, or the project's equivalent. -- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research` and `prototype`. +- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research`. - **Stops:** After the problem, actors, behaviour, rules, states, scenarios, acceptance conditions, constraints and exclusions are unambiguous. It does not select the implementation or create tickets. ### Plan @@ -131,7 +131,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** An approved, buildable technical solution for one feature specification. - **Inputs:** The approved feature specification, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. - **Output:** `docs/features//solution-design.md` and any warranted ADRs. -- **Composition:** Requires `modular-design`; conditionally invokes `elicit-with-context`, `research`, `prototype`, `maintain-ubiquitous-language` and `maintain-decision-records`. +- **Composition:** Requires `modular-design`; conditionally invokes `elicit-with-context`, `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After affected boundaries, interfaces, data, integrations, security, migration, operational effects and verification seams are resolved. It does not decompose or implement the work. #### `create-tickets` @@ -183,7 +183,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** One ticket or small agreed body of work exists as a minimal, verified local change. - **Inputs:** The stated ticket or confirmed work, linked product and technical authorities, repository guidance and current code. - **Output:** Working code, tests and other in-scope artefacts with reproducible verification and no unresolved required review finding. -- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit-with-context`, `research`, `prototype`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. +- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit-with-context`, `research`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After the agreed behaviour is implemented and verified. It does not commit or publish the change. #### `commit` @@ -217,7 +217,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Outcome:** A defect has an evidence-backed root cause and, when authorised by the requested outcome, a minimal verified repair. - **Inputs:** Expected and observed behaviour, the failing signal, relevant environment evidence and current implementation authorities. - **Output:** A reproducible diagnosis and, for repair work, regression protection plus a reviewed local change. -- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `prototype`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. +- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. #### `respond-to-incident` @@ -253,7 +253,6 @@ The skill uses the selected tool's native issues, relationships, statuses, prior | `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | | `elicit-with-context` | Routes software-project elicitation through `elicit` while maintaining confirmed project language and qualifying ADRs. | | `research` | Investigates a material subject using high-trust evidence and creates a trusted cited report for the consumer to synthesize. | -| `prototype` | Creates a deliberately disposable experiment or mock-up to resolve one bounded uncertainty and preserves its conclusion. | | `maintain-ubiquitous-language` | Keeps `CONTEXT.md` aligned inline with confirmed project-specific language. | | `maintain-decision-records` | Preserves rare accepted technical and architectural decisions whose rationale warrants a focused ADR. | | `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | @@ -385,7 +384,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ## Implementation order 1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. -2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `prototype`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. +2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. 3. Build the progressive foundation path: `define-product`, `define-architecture`, `specify-feature`, `design-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. @@ -397,7 +396,6 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Complete quality-harness research - [x] Update `elicit` skill - [x] Update `research` skill -- [-] Create `prototype` skill - [x] Create `maintain-ubiquitous-language` skill - [x] Create `maintain-decision-records` skill - [-] Update `modular-design` skill diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index e64473d..915cbe7 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -24,7 +24,7 @@ When the signal or next experiment is unclear, load [Debugging Techniques](refer State the plausible causal hypotheses in evidence-supported order. For the leading hypothesis, predict an observation that distinguishes it from credible alternatives, then run the smallest safe experiment capable of producing that observation. Change one variable at a time and record the hypothesis, prediction, experiment, observation, and whether the result supports, rejects, or fails to discriminate the hypothesis. Treat correlation, disappearance of a symptom, and a passing retry as evidence to explain rather than proof of cause. -Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$prototype` only when seeing or interacting with a disposable artefact would resolve one bounded diagnostic uncertainty. Invoke `$maintain-ubiquitous-language` only after independently confirmed terminology arises and `$maintain-decision-records` only after an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. +Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$maintain-ubiquitous-language` only after independently confirmed terminology arises and `$maintain-decision-records` only after an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. ### 3. Test one causal repair diff --git a/skills/define-architecture/SKILL.md b/skills/define-architecture/SKILL.md index f0f989b..5ab425d 100644 --- a/skills/define-architecture/SKILL.md +++ b/skills/define-architecture/SKILL.md @@ -20,7 +20,7 @@ Require a confirmed root `PRODUCT.md` and relevant root `CONTEXT.md`. Inspect bo Inspect the request, `PRODUCT.md`, `CONTEXT.md`, existing `ARCHITECTURE.md` and ADRs, repository guidance, and material technical, organisational, regulatory, and operational constraints. For an existing system, inspect the smallest representative runtime and deployment evidence needed to understand principal entry points, responsibilities, data stores and flows, integrations, contracts, and trust boundaries. Treat code, configuration, contracts, tests, and deployment evidence as current implementation evidence, not intended architecture by default. Preserve confirmed unrelated content and local documentation conventions. The evidence is separated into confirmed intent, observed current architecture, and unresolved gaps before questioning begins. -Invoke `$elicit-with-context` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Invoke `$prototype` only when one bounded visual or interactive uncertainty would be resolved better by the least elaborate disposable artefact than by discussion. Supporting utilities own only their contract-specific mutations and return their results to this session. +Invoke `$elicit-with-context` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Supporting utilities own only their contract-specific mutations and return their results to this session. Keep `ARCHITECTURE.md` unchanged until the complete synthesis is explicitly confirmed. diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index e536b72..2c0d48f 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -22,7 +22,7 @@ Invoke `$elicit-with-context` for material user-held knowledge and decisions thr Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. -Invoke `$research` when a material external subject could change the product decision and needs durable evidence; read and link its report rather than copying its findings wholesale. Use an incidental lookup when the evidence need not persist. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would answer one bounded visual question better than discussion. Research and prototype results return evidence to this product session; neither owns `PRODUCT.md` or continuation. +Invoke `$research` when a material external subject could change the product decision and needs durable evidence; read and link its report rather than copying its findings wholesale. Use an incidental lookup when the evidence need not persist. Research returns evidence to this product session and owns neither `PRODUCT.md` nor continuation. Keep `PRODUCT.md` unchanged until the complete whole-product synthesis is explicitly confirmed. Before then, supporting utilities may perform only the writes authorised by their own contracts. diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index 79b3368..65359a0 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -1,6 +1,6 @@ # Discovery Techniques -Load only the section needed by the active discovery branch. `$elicit-with-context` coordinates material questioning, final synthesis confirmation, and glossary maintenance; its composed skills retain their own results. `$research` owns durable external evidence; `$prototype` owns a bounded disposable visual experiment. `define-product` retains the product outcome and stopping boundary. +Load only the section needed by the active discovery branch. `$elicit-with-context` coordinates material questioning, final synthesis confirmation, and glossary maintenance; its composed skills retain their own results. `$research` owns durable external evidence. `define-product` retains the product outcome and stopping boundary. ## Establish an existing product @@ -51,7 +51,3 @@ Read the mapped product back as one narrative. Change perspective across intende ## Establish external position Compare the product with meaningful alternatives users employ today, including manual work and doing nothing. Invoke `$research` only when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision and need a durable evidence record. Keep researched claims in the report and link that report from `PRODUCT.md` rather than copying its findings; use an incidental lookup when no durable evidence record is needed. - -## Resolve a visual product question - -Invoke `$prototype` only when one bounded uncertainty is better resolved by seeing, comparing, or interacting with a visual artefact than by discussion. Give it the question, audience, authentic constraints, and product decision it informs. Use its returned evidence in elicitation after the disposable artefact is removed; do not preserve prototype material in the product foundation. diff --git a/skills/design-feature/SKILL.md b/skills/design-feature/SKILL.md index ce973ca..ea27e3d 100644 --- a/skills/design-feature/SKILL.md +++ b/skills/design-feature/SKILL.md @@ -24,7 +24,7 @@ Inspect the request, prerequisite authorities, existing paired solution design, Separate confirmed future behaviour, architecture constraints, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide decision is absent or the feature would require changing the architecture, identify the exact architecture gap and stop rather than embedding a new foundation in this design. -Invoke `$elicit-with-context` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Invoke `$prototype` only when one bounded visual, interactive, or technically observable uncertainty can be discriminated by its disposable artefact. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. +Invoke `$elicit-with-context` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. Invoke `$maintain-ubiquitous-language` inline whenever consequential project terminology is independently confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update; this skill retains the solution-design outcome and verification. Apply a glossary change before whole-design confirmation only when its meaning is confirmed independently of the proposed solution. Keep terminology that depends on an unconfirmed design choice unchanged until that choice is confirmed. diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 796d97d..c9f430e 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -18,7 +18,7 @@ Inspect the stated ticket or agreed work, repository instructions, linked produc ### 2. Build the smallest complete change -Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Invoke `$prototype` only for a bounded uncertainty best answered by seeing or interacting. Use `$maintain-ubiquitous-language` when independently confirmed durable terminology arises and `$maintain-decision-records` when a rare qualifying decision needs durable rationale; each utility returns control to this outcome. +Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Use `$maintain-ubiquitous-language` when independently confirmed durable terminology arises and `$maintain-decision-records` when a rare qualifying decision needs durable rationale; each utility returns control to this outcome. Invoke `$tdd` only when an existing runnable test suite can credibly exercise the requested observable behaviour at a stable boundary with an independent oracle. When evidence already shows any prerequisite is absent, keep control here rather than invoking TDD merely to receive that result. Let TDD own the Red-Green-Refactor cycles and returned code and tests when it applies. Otherwise implement through the smallest conventional path and use the strongest available project-native feedback; update existing tests only when they can meaningfully express changed behaviour. Do not install or invent a test or quality framework to satisfy this process. When meaningful infrastructure is absent, preserve the limitation for verification and recommend separately planned work rather than adding unrelated harness infrastructure. diff --git a/skills/prototype/SKILL.md b/skills/prototype/SKILL.md deleted file mode 100644 index dd02e72..0000000 --- a/skills/prototype/SKILL.md +++ /dev/null @@ -1,41 +0,0 @@ ---- -name: prototype -description: Creates disposable visual experiments that answer one bounded question. Use when seeing, comparing, or interacting with an artefact would resolve uncertainty better than discussion. -metadata: - invocation: model -disable-model-invocation: false ---- - -# Prototype - -Creates the least elaborate temporary visual experiment that can answer one bounded question, then returns the evidence and removes all prototype material. - -## Process - -### 1. Bound the experiment - -Inspect the request, caller outcome and authority, repository guidance, relevant authentic runtime context, intended evaluators, and safely available data. State one question and the observable reaction, comparison, or result that would answer it before changing anything. For an objective question, include an observation that would support the proposed answer and one that would refute or materially qualify it; for a subjective choice, state the decision criteria and keep the choice with the user. Invoke `$elicit-with-context` only when user-held information is needed to bound this visual question; return broader discovery to the caller. When a visual artefact cannot discriminate the question, stop and return that boundary without prototyping. The question, audience, criteria, decision authority, and discriminators are explicit. - -### 2. Confirm presentation and disposal - -Inspect the current project state, available rendering or viewing capability, and a safe reversible workspace. Confirm a real presentation path that can expose a live view or rendered capture directly to the user, and record the pre-existing state needed to verify cleanup. If the result cannot be rendered and presented, stop before building, clean any preliminary attempt, and report the missing capability. Identify every file, route, component, process, dependency, fixture, branch, data source, and external effect the experiment could touch. The experiment has a presentation path and an exact disposable boundary before construction begins. - -### 3. Choose the representation - -Choose the simplest trustworthy sketch, static mock-up, interactive stub, or other visual manifestation that exposes the discriminators. Identify the qualities it must show, what it may omit, and whether an omission could distort the answer. Use interaction only when behaviour, sequence, state, or experience is material. When comparing alternatives, show the materially different options under the same content, data, and constraints. The selected form exposes the decisive qualities without implying production completeness. - -### 4. Build the artefact - -Build only the fidelity and behaviour needed to expose the discriminators. Use authentic project layout, components, styling, content, or safe read-only data only when they materially affect interpretation and remain behind the reversible boundary; otherwise use isolated temporary material. Keep live systems read-only, stub external writes and irreversible actions, do not expose credentials, and distinguish simulated behaviour from implemented capability. Do not widen the experiment into production architecture, reusable abstractions, tests, hardening, or unrelated polish. One faithful disposable artefact is ready to render. - -### 5. Render and present it - -Render the artefact in its intended medium and inspect its visual output and relevant interactions. Verify that the question, alternatives or scenarios, simulated boundaries, omissions, and decisive observations are understandable to the intended audience; correct anything that could make the experiment misleading. Present the live view or rendered capture directly to the user before cleanup, keep it available while evaluation requires it, and provide only the controls needed to compare or explore it. Source inspection or a path that will disappear before viewing is not presentation. The user has an opportunity to inspect the visual result. - -### 6. Iterate to a conclusion - -Revise the same artefact only when another pass could change the answer to the bounded question. Let the agreed observations settle an objective question and obtain the user's choice for a subjective one. Stop when the question is answered or further useful iteration cannot discriminate it. If the user instead requests a durable or production outcome, stop the prototype and return that changed outcome to the appropriate caller or downstream skill rather than retaining or converting the experiment. The experiment has a supported conclusion or an exact remaining uncertainty and missing discriminator. - -### 7. Remove and return - -After the user has had the opportunity to inspect the result and the experiment has concluded, remove every prototype artefact, dependency, fixture, branch, process, and temporary change. Revert only prototype-specific adjustments, then compare with the recorded starting state and verify that all pre-existing work remains intact. Preserve no prototype report, code, data, branch, or capture by default. Return the answer or explicitly inconclusive result in the current session or caller handoff with the evidence, limitations, remaining uncertainty, and missing discriminator where applicable. The caller retains every product, specification, architecture, implementation, and durable-update decision; it receives the result with no prototype residue. diff --git a/skills/prototype/agents/openai.yaml b/skills/prototype/agents/openai.yaml deleted file mode 100644 index 71e000c..0000000 --- a/skills/prototype/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Prototype' - short_description: 'Resolve uncertainty with disposable visual experiments' -policy: - allow_implicit_invocation: true diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index d3bc3e6..45d8b25 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -24,7 +24,7 @@ Invoke `$elicit-with-context` throughout steps 2–5 for material user-held know Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. -Invoke `$research` when a material external, domain, regulatory, or standards subject could change the specification and needs durable evidence. Link its report instead of copying its findings. Invoke `$prototype` only when seeing, comparing, or interacting with the least elaborate disposable artefact would resolve one bounded visual or interactive uncertainty better than discussion. Both return evidence to this session and own neither the specification nor continuation. +Invoke `$research` when a material external, domain, regulatory, or standards subject could change the specification and needs durable evidence. Link its report instead of copying its findings. Research returns evidence to this session and owns neither the specification nor continuation. Keep the specification unchanged until the complete synthesis is explicitly confirmed. Supporting utilities may perform only the mutations authorised by their own contracts before then. From 57b0eeba0e4c05b30bb13f1b5d3c5ee34573ce0d Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Tue, 18 Aug 2026 15:19:26 +0100 Subject: [PATCH 52/79] feat(skills): finalize maintain-agents --- AGENTS.md | 2 +- SKILLS_PLAN.md | 2 +- skills/maintain-agents/SKILL.md | 30 +++++++++++------------ skills/maintain-agents/agents/openai.yaml | 4 +-- 4 files changed, 19 insertions(+), 19 deletions(-) diff --git a/AGENTS.md b/AGENTS.md index efc0147..49d92d4 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,4 +1,4 @@ -- When the user corrects you with a reusable, global rule, ask if they want it added to `AGENTS.md`. +- When a user correction establishes a reusable repository-wide rule, ask whether to invoke `$maintain-agents` to add it. - When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. - When eliciting within a software project, invoke `$elicit-with-context` instead of `$elicit`; use `$elicit` directly only for non-software work or as part of `$elicit-with-context`. - Assume the execution harness supports fresh agents; do not add fallback handling for their unavailability. diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 693f354..cf6f8f5 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -402,7 +402,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Update `tdd` skill - [-] Create `verify-change` skill - [-] Update `code-review` skill -- [-] Update `maintain-agents` skill +- [x] Update `maintain-agents` skill - [-] Update `define-product` skill - [-] Create `define-architecture` skill - [-] Create `specify-feature` skill diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index 3b957f4..569044c 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -1,33 +1,33 @@ --- name: maintain-agents -description: Maintains lean root AGENTS.md files containing confirmed repository-wide guidance. Use when creating, updating, or compressing project-wide agent instructions. +description: Maintains concise root AGENTS.md instructions. Use when creating the file or adding, refining, or assessing repository-wide agent guidance. metadata: - invocation: model -disable-model-invocation: false + invocation: user +disable-model-invocation: true --- # Maintain AGENTS.md -**Progressive Disclosure** keeps root `AGENTS.md` guidance limited to confirmed, non-inferable behaviour that governs the repository as a whole, while narrower guidance stays with its discoverable owner. Adapt **Minimalist Instruction** and **DRY** to preserve each behavioural policy once in the fewest words that retain its conditions. +Uses **Progressive Disclosure** to keep the repository's always-loaded agent guidance relevant to every request. ## Process -### 1. Establish the instruction surface +### 1. Establish the file -Locate the repository root. Read its `AGENTS.md` when present, the applicable instruction chain, and the existing or proposed guidance confirmed by the user or caller. Treat a caller handoff carrying an agreed preference as authority; do not ask for duplicate permission to record it. Inspect repository evidence only far enough to recover intent, scope, inferability, conflicts, and referenced commands. Invoke `$elicit-with-context` only when material authority, scope, or conflict remains unresolved. The authorised candidates and active constraints are explicit. +Locate the repository root and read its `AGENTS.md` when present. Create it when absent. Ensure this line appears exactly once at the top, replacing or moving equivalent wording: -### 2. Allocate the guidance +```markdown +- When a user correction establishes a reusable repository-wide rule, ask whether to invoke `$maintain-agents` to add it. +``` -Keep a rule at root only when it is confirmed, cannot be inferred reliably from repository evidence, governs the repository as a whole, and materially changes agent behaviour. Apply every gate equally to checks, tools, preferences, and overrides. A paired-repository relationship or the user's exact task-management tool choice can qualify; mere tool or script existence, framework convention, code inventory, task status, product language, workflow procedure, or directory-only rule cannot. +### 2. Assess the guidance -Ensure the root file itself contains one immediate correction loop: when a user correction establishes a reusable repository-wide rule, ask whether to add it to `AGENTS.md`. Preserve project-native wording when an equivalent confirmed rule already exists; do not add a duplicate. +Recommend an instruction when it is user-confirmed, cannot be inferred reliably from repository evidence, governs the repository as a whole, and materially changes agent behaviour. Treat a composing caller's confirmed instruction as user-confirmed. When an instruction fails a gate, explain the concern and ask for the user's final decision; follow that decision. -Remove stale, inferable, explanatory, or narrower guidance. Report useful narrower guidance with its smallest discoverable proposed owner: an invocable skill for a reusable workflow, a scoped instruction file for a directory, executable enforcement for a mechanical rule, or the applicable product, context, or architecture authority. Leave those owners unchanged. Every candidate is retained, displaced, or removed for an explicit reason. +### 3. Write the instruction -### 3. Write the root instructions +Apply only the requested addition, refinement, or assessment. Use the shortest direct imperative sentence that preserves the required behaviour and conditions. State each policy once; update equivalent wording instead of adding a semantic duplicate. Leave unrelated instructions unchanged. -Create or update only the root `AGENTS.md`, preserving unrelated user work. If no candidate survives allocation or the current wording is already equivalent, make no change. Use direct imperative lines and headings only when they navigate multiple instruction groups. Retain only words that require behaviour or preserve a necessary condition. State each repository policy once: merge instructions only when the same policy change requires both statements to change, and keep independently changeable rules separate. Do not add a manifest, lifecycle state, task-tool adapter, exhaustive command list, architecture or code documentation, or skill procedure. The root file is lean and unambiguous. +### 4. Verify the result -### 4. Verify and hand off - -Re-read every line against authority, non-inferability, repository-wide scope, behavioural value, conflicts, and semantic duplication. Verify each proposed owner and every referenced command definition before deciding whether to run it. Run applicable checks required by active guidance only when their effects, including aggregate command writes, are safe and authorised; otherwise report the limitation without weakening or replacing the check. Return the changed root file or no-op, verification evidence or limitations, displaced guidance with proposed destinations, and unresolved conflicts. Invocation authorises the agreed root edit without another application prompt. The user receives a lean verified root file without unrequested changes to narrower owners. +Re-read `AGENTS.md`. Confirm the canonical first line, requested outcome, minimal wording, and absence of semantic duplication. Return the change, no-op, or requested advice with any concerns. diff --git a/skills/maintain-agents/agents/openai.yaml b/skills/maintain-agents/agents/openai.yaml index 25e0133..2b4bbb9 100644 --- a/skills/maintain-agents/agents/openai.yaml +++ b/skills/maintain-agents/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Maintain AGENTS.md' - short_description: 'Maintain confirmed repository-wide agent guidance' + short_description: 'Write concise repository-wide agent guidance' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false From 1695b845b2e36bc31d5792427cd8b0b744165195 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Wed, 19 Aug 2026 10:50:39 +0100 Subject: [PATCH 53/79] docs(agents): clarify composed skill boundaries --- AGENTS.md | 1 + 1 file changed, 1 insertion(+) diff --git a/AGENTS.md b/AGENTS.md index 49d92d4..5f96e9d 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,5 +1,6 @@ - When a user correction establishes a reusable repository-wide rule, ask whether to invoke `$maintain-agents` to add it. - When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. - When eliciting within a software project, invoke `$elicit-with-context` instead of `$elicit`; use `$elicit` directly only for non-software work or as part of `$elicit-with-context`. +- When composing skills, rely on each invoked skill's public contract; do not restate or steer its internal logic. - Assume the execution harness supports fresh agents; do not add fallback handling for their unavailability. - After implementing changes run `bun run checks` before handoff. From aed22bed0a539acf4a9c8b39df9196a2124099a7 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Wed, 19 Aug 2026 10:51:08 +0100 Subject: [PATCH 54/79] feat(skills): consolidate product and architecture definition --- README.md | 8 +- SKILLS_PLAN.md | 35 ++---- skills/define-architecture/SKILL.md | 67 ---------- skills/define-architecture/agents/openai.yaml | 5 - .../assets/architecture-template.md | 39 ------ skills/define-product/SKILL.md | 54 +++++--- skills/define-product/agents/openai.yaml | 2 +- .../define-product/assets/product-template.md | 116 ++++++++++++++---- skills/define-product/references/DISCOVERY.md | 52 ++------ skills/design-feature/SKILL.md | 10 +- .../assets/solution-design-template.md | 2 +- skills/maintain-decision-records/SKILL.md | 2 +- skills/plan-retirement/SKILL.md | 2 +- skills/review-architecture/SKILL.md | 2 +- skills/triage-work/SKILL.md | 2 +- 15 files changed, 164 insertions(+), 234 deletions(-) delete mode 100644 skills/define-architecture/SKILL.md delete mode 100644 skills/define-architecture/agents/openai.yaml delete mode 100644 skills/define-architecture/assets/architecture-template.md diff --git a/README.md b/README.md index 9996945..754bf7c 100644 --- a/README.md +++ b/README.md @@ -35,7 +35,7 @@ Repeat the link for each selected skill. Codex and OpenCode both discover skills | When you need to… | Start with… | Session outcome | | --- | --- | --- | -| Establish a new product | `$define-product` | Durable product documentation and shared language | +| Establish or revise a product | `$define-product` | Confirmed product and system requirements | | Define a substantial feature | `$specify-feature` | An approved, observable feature specification | | Make a small understood change | `$implement` | A minimal, reviewed, verified local change | | Route incoming requests | `$triage-work` | Verified dispositions and lifecycle routes | @@ -50,10 +50,7 @@ The lifecycle areas are useful entry points, not mandatory phase gates. ### Establish -- `$define-product` creates the durable product foundation in `PRODUCT.md` and establishes its ubiquitous language in `CONTEXT.md`. -- `$define-architecture` creates the enduring technical foundation in `ARCHITECTURE.md` and records only the focused architectural decisions that need durable rationale. - -These are separate sessions. Product definition ends with product documentation; architecture begins later from that confirmed foundation. +- `$define-product` turns a new or existing product into the durable breadth-first product and system requirements foundation in `PRODUCT.md`, while `elicit-with-context` maintains applicable project language and qualifying decision records through its own contract. ### Define @@ -96,7 +93,6 @@ Establish a product: ```text define-product - → later define-architecture ``` Deliver a substantial feature: diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index cf6f8f5..c6aa835 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -29,7 +29,7 @@ Propulsion is not a workflow platform. It does not carry a product automatically | Area | Purpose | | --- | --- | -| Establish | Create the durable product and technical foundations for a new or existing system. | +| Establish | Create the durable product and system requirements foundation for a new or existing system. | | Define | Turn one product capability or enhancement into an approved statement of intended behaviour. | | Plan | Select a buildable solution and decompose it into implementation-ready work. | | Refine | Keep incoming and existing work current, prioritised, ready and feasible for the next delivery period. | @@ -62,13 +62,12 @@ Every skill: | Authority | Owns | | --- | --- | -| `PRODUCT.md` | Product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capability catalogue. | +| `PRODUCT.md` | Product purpose, users, needs, value, outcomes, boundaries, journeys, high-level capabilities and system-wide technical requirements. | | `CONTEXT.md` | Canonical project terms and their concise project-specific meanings. | -| `ARCHITECTURE.md` | Enduring architecture drivers, system context, responsibilities, technology strategy, data ownership, integrations and cross-cutting technical constraints. | `CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-ubiquitous-language` inline. -Architecture documentation records boundaries, constraints and rationale that cannot be inferred reliably. It does not mirror classes, files, framework wiring, executable configuration or other implementation detail. `maintain-decision-records` creates focused ADRs only when a consequential or surprising decision needs durable rationale beyond `ARCHITECTURE.md`. +`PRODUCT.md` records requirements-first system constraints that later feature solutions must respect without selecting their internal implementation. `maintain-decision-records` creates focused ADRs only when a consequential or surprising technical or architectural decision needs durable rationale beyond the product foundation or feature records. ### Feature records @@ -100,19 +99,11 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `define-product` -- **Outcome:** A user-confirmed product foundation broad enough to guide later feature definition without becoming a roadmap or delivery plan. -- **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, representative repository evidence and applicable external evidence. +- **Outcome:** A user-confirmed product and system requirements foundation broad enough to guide later feature specification and solution design without becoming either. +- **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, applicable ADRs, representative repository evidence and applicable external evidence. - **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. -- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research`. -- **Stops:** After product purpose, users, needs, value, outcomes, boundaries, journeys and high-level capabilities are documented. It does not define technical architecture, detailed features, priorities or tickets. - -#### `define-architecture` - -- **Outcome:** A user-confirmed technical foundation capable of constraining later feature solutions without duplicating the implementation. -- **Inputs:** `PRODUCT.md`, `CONTEXT.md`, existing system evidence and material technical, organisational or regulatory constraints. -- **Output:** `ARCHITECTURE.md` and any warranted focused ADRs. -- **Composition:** Requires `elicit-with-context` and `modular-design`; conditionally invokes `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. -- **Stops:** After the system context, principal boundaries, ownership, technology strategy, data, integrations and cross-cutting quality approach are established. It does not scaffold the system or create implementation work. +- **Composition:** Requires `elicit-with-context`; conditionally invokes `research`. +- **Stops:** After purpose, users, needs, value, outcomes, boundaries, journeys, high-level capabilities and material quality, interface, data, integration, security, compliance, platform and operational requirements are documented. It does not define detailed feature behaviour, select feature-specific implementation or architecture, create priorities or create tickets. ### Define @@ -129,7 +120,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `design-feature` - **Outcome:** An approved, buildable technical solution for one feature specification. -- **Inputs:** The approved feature specification, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. +- **Inputs:** The approved feature specification, `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. - **Output:** `docs/features//solution-design.md` and any warranted ADRs. - **Composition:** Requires `modular-design`; conditionally invokes `elicit-with-context`, `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. - **Stops:** After affected boundaries, interfaces, data, integrations, security, migration, operational effects and verification seams are resolved. It does not decompose or implement the work. @@ -171,7 +162,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `review-architecture` - **Outcome:** A bounded, evidence-backed assessment identifies a small number of high-value architecture or maintainability improvements. -- **Inputs:** An explicitly scoped codebase area, `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. +- **Inputs:** An explicitly scoped codebase area, `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. - **Output:** An independently useful architecture review report with prioritised findings, evidence, benefits and trade-offs. - **Composition:** Requires `modular-design`; conditionally invokes `research`. - **Stops:** After reporting. It does not refactor the code or create speculative tickets automatically. @@ -314,10 +305,9 @@ Review findings are corrected through the relevant implementation and verificati ```text define-product - → later define-architecture ``` -Product and architecture planning are separate sessions and stop after their respective documents. +Product definition stops after the confirmed breadth-first product and system requirements foundation. ### Deliver a substantial feature @@ -385,7 +375,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. 2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. -3. Build the progressive foundation path: `define-product`, `define-architecture`, `specify-feature`, `design-feature` and `create-tickets`. +3. Build the progressive foundation path: `define-product`, `specify-feature`, `design-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. 6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. @@ -403,8 +393,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Create `verify-change` skill - [-] Update `code-review` skill - [x] Update `maintain-agents` skill -- [-] Update `define-product` skill -- [-] Create `define-architecture` skill +- [x] Update `define-product` skill - [-] Create `specify-feature` skill - [-] Create `design-feature` skill - [-] Create `create-tickets` skill diff --git a/skills/define-architecture/SKILL.md b/skills/define-architecture/SKILL.md deleted file mode 100644 index 5ab425d..0000000 --- a/skills/define-architecture/SKILL.md +++ /dev/null @@ -1,67 +0,0 @@ ---- -name: define-architecture -description: Establishes a durable technical foundation for a new or existing system. Use when the product foundation is confirmed and later feature solutions need system-wide architecture constraints. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Define Architecture - -Establishes one user-confirmed, system-specific technical foundation breadth-first, then tests it against representative change before recording it. Selective **arc42 architecture-description concepts** keep drivers, constraints, context, responsibility views, cross-cutting concerns, runtime, deployment, and risks inspectable without adopting its full template. The result constrains later feature solutions without duplicating current implementation or designing an individual feature. - -## Prerequisites - -Require a confirmed root `PRODUCT.md` and relevant root `CONTEXT.md`. Inspect both before continuing. When either is missing, or the product foundation is too incomplete or contradictory to establish its technical drivers, report the exact prerequisite and stop; leave product definition to a later `$define-product` invocation rather than manufacturing it here. - -## Process - -### 1. Establish the evidence boundary - -Inspect the request, `PRODUCT.md`, `CONTEXT.md`, existing `ARCHITECTURE.md` and ADRs, repository guidance, and material technical, organisational, regulatory, and operational constraints. For an existing system, inspect the smallest representative runtime and deployment evidence needed to understand principal entry points, responsibilities, data stores and flows, integrations, contracts, and trust boundaries. Treat code, configuration, contracts, tests, and deployment evidence as current implementation evidence, not intended architecture by default. Preserve confirmed unrelated content and local documentation conventions. The evidence is separated into confirmed intent, observed current architecture, and unresolved gaps before questioning begins. - -Invoke `$elicit-with-context` throughout steps 2–6 for material user-held information and technical trade-offs. Derive discoverable facts first and ask exactly one material decision at a time. Invoke `$research` when a material external technical or regulatory claim could change a decision and its evidence must survive the session; use obvious local facts directly and an incidental lookup for external facts that need not persist. Supporting utilities own only their contract-specific mutations and return their results to this session. - -Keep `ARCHITECTURE.md` unchanged until the complete synthesis is explicitly confirmed. - -### 2. Establish drivers and context - -Establish the architecture goals, product and organisational drivers, highest-priority material quality needs, fixed constraints, external actors and systems, scope, and trust boundaries. Express quality needs with enough observable context to compare consequences without manufacturing precision. Distinguish an existing system's observed current state, confirmed target direction, and unresolved gap wherever conflating them could mislead later design. The forces that the technical foundation must balance are explicit before a solution structure is proposed. - -### 3. Establish responsibilities and boundaries - -Map the principal system capabilities or responsibility boundaries, their coherent owners, information hidden behind each boundary, dependency and communication direction, and major contracts future feature designs must respect. Invoke `$modular-design` with the confirmed behaviour, drivers, constraints, and current evidence; use its structural findings as passive constraints while retaining ownership of architecture alternatives, user decisions, and this document. Prefer the smallest justified boundaries and deep, stable contracts that localise plausible change. Compare viable alternatives and their consequences for each material structural trade-off, then leave the choice to the user. The proposed foundation has coherent ownership and explicit dependency rules without mirroring the source tree. - -### 4. Complete the technical foundation - -Establish only the system-wide decisions and constraints that are material across the following concerns: - -- technology strategy and justified constraints; -- data domains, ownership, lifecycle, consistency, privacy, and retention; -- integrations, external dependencies, protocols, and failure boundaries; -- security, reliability and resilience, observability, performance and scalability, accessibility and compatibility, operability, and testing or verification strategy; and -- deployment and runtime topology at an enduring conceptual level. - -Omit a concern when it is not material. Record what future designs must preserve, what cannot yet be reliably inferred, and why a constraint exists. Leave endpoints, schemas, classes, files, framework wiring, executable configuration, library inventories, ephemeral deployment details, ticket or roadmap state, and feature-specific solutions to their owning evidence and later sessions. The proposed foundation is proportionate and contains no speculative implementation inventory. - -Invoke `$maintain-ubiquitous-language` inline when consequential terminology is confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update. Treat `ARCHITECTURE.md` as the main living synthesis. Identify and test an ADR candidate only when `$maintain-decision-records`' four gates appear satisfied, then include its rationale and the user's explicit durable-record agreement in the final synthesis. Do not invoke `$maintain-decision-records` before that whole synthesis is confirmed, and reject routine, discoverable, or section-level choices as candidates. - -### 5. Walk representative change - -Walk representative current and likely capability changes from product intent through actors, boundaries, contracts, data, integrations, cross-cutting constraints, and runtime consequences. Use these walks to expose contradictions, missing ownership, unstable dependency direction, or a foundation that would force scattered change. Reopen only the evidenced architecture branch needed to resolve a failure, compare the viable consequences, and avoid elaborating the feature itself. The foundation can constrain plausible later solution designs without pretending to settle them. - -### 6. Confirm the complete synthesis - -Have `$elicit-with-context` present one concise, complete synthesis covering drivers and context, responsibilities and dependency direction, technology and data strategy, integrations and material cross-cutting constraints, current-versus-target distinctions, trade-offs, evolution constraints, unresolved material risks, and every proposed ADR rationale. Obtain explicit user confirmation of the whole synthesis. Create no ADR record from any candidate until the complete synthesis, or a revised synthesis after rejection, is confirmed. If the user rejects or corrects it, leave `ARCHITECTURE.md` unchanged, reopen only affected evidence-backed branches, repeat the relevant change walks, and reconfirm the complete revised synthesis. The confirmed synthesis supplies authority for its agreed document writes without a second permission prompt. - -### 7. Record the architecture - -Create or update the single root `ARCHITECTURE.md` using the [Architecture Foundation Template](assets/architecture-template.md) when the project has no stronger convention. Preserve unrelated confirmed content and local structure when it differs only stylistically. Omit empty or immaterial sections rather than filling boilerplate. Use prose, tables, or repository-native diagrams only when they materially clarify relationships. Keep canonical term meanings in `CONTEXT.md`, external evidence in research reports, and exceptional rationale in qualifying ADRs instead of duplicating them. - -After writing `ARCHITECTURE.md`, invoke `$maintain-decision-records` for every confirmed qualifying ADR and `$maintain-ubiquitous-language` for confirmed terminology not already applied. The prior architecture confirmation supplies explicit agreement only for ADR rationales included in that synthesis. - -### 8. Verify and stop - -Verify every recorded statement against the confirmed synthesis and its authority: product intent against `PRODUCT.md`; terms against `CONTEXT.md`; current claims against inspected system evidence; target direction and trade-offs against user confirmation; and external claims against cited research. Check that responsibilities, dependency directions, data ownership, integration failure boundaries, and material quality constraints agree with one another and that the representative changes still fit. Confirm preservation of unrelated content and absence of implementation inventory, feature design, scaffolding, priorities, tickets, releases, or deployment execution. - -Report changed files, evidence and limitations, current-versus-target gaps, unresolved material risks, and the next available later session without starting it. Stop after the foundation and warranted context records are written and verified; do not scaffold code, select an individual feature solution, create work items, implement, release, or deploy. diff --git a/skills/define-architecture/agents/openai.yaml b/skills/define-architecture/agents/openai.yaml deleted file mode 100644 index 7067e7a..0000000 --- a/skills/define-architecture/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Define Architecture' - short_description: 'Establish an enduring technical foundation' -policy: - allow_implicit_invocation: false diff --git a/skills/define-architecture/assets/architecture-template.md b/skills/define-architecture/assets/architecture-template.md deleted file mode 100644 index 24c3d37..0000000 --- a/skills/define-architecture/assets/architecture-template.md +++ /dev/null @@ -1,39 +0,0 @@ -# Architecture - - - -## Orientation - - - -## System context - - - -## Responsibilities and boundaries - - - -## Technology strategy - - - -## Data - - - -## Integrations - - - -## Cross-cutting constraints - - - -## Runtime and deployment - - - -## Evolution, trade-offs, and risks - - diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 2c0d48f..f8f03b7 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -1,6 +1,6 @@ --- name: define-product -description: Establishes a durable whole-product foundation for later feature definition. Use when externalising a new product idea or redefining an existing product. +description: Defines a breadth-first product and system requirements foundation. Use when externalising a new product idea or deliberately revising PRODUCT.md. metadata: invocation: user disable-model-invocation: true @@ -8,40 +8,56 @@ disable-model-invocation: true # Define Product -Establishes one user-confirmed product foundation using an adapted **Product Vision Board** for strategic completeness and **User Story Mapping** for a mile-wide, inch-deep view of the whole product. The result guides later feature definition without becoming a feature specification, roadmap, or technical design. +Turns a product idea or existing system into one confirmed `PRODUCT.md` that supplies product intent and system-wide requirements to later feature specification and solution design. ## Process -### 1. Inspect the product evidence +### 1. Establish the evidence boundary -Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, applicable research and decisions, and, for an existing system, the smallest representative public contracts, entry points, data boundaries, code, tests, and configuration. Treat the repository as evidence of current behaviour and the user as authority on intended product direction. Surface contradictions, distinguish observed behaviour from stated direction, and treat missing evidence as unknown rather than absence. Do not turn the inspection into an implementation inventory. When an existing product is difficult to establish, load [Discovery Techniques](references/DISCOVERY.md). The known product evidence, its provenance, and the unresolved product decisions are explicit. +Locate the project root; read its repository guidance and check root `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and research. For an existing system, also inspect the smallest representative public contracts, entry points, data, tests, configuration, and operational evidence needed to understand the product. Record absent foundation files as absent evidence rather than skipping the check. -### 2. Establish the product frame +Treat the user as authority for intended direction, repository and runtime evidence as evidence of current behaviour, and cited external sources as authority only for the claims they establish. -Invoke `$elicit-with-context` for material user-held knowledge and decisions throughout steps 2–5. Reuse the confirmed synthesis from a prior free-form `$elicit-with-context` session as resolved branches and ask only about gaps exposed by the product evidence. Establish concise executive orientation: purpose and vision; intended users, actors, and their needs and desired outcomes; value and meaningful alternatives or market position when applicable; product and business outcomes with observable success signals; business model or pricing only when material; and durable boundaries and non-goals. Derive discoverable facts before asking. The concise product frame is complete enough to orient the whole-product pass. +Separate confirmed target requirements, observed current behaviour, assumptions, and unknowns before questioning. Preserve unrelated confirmed content and local conventions when revising the foundation. When the current product is difficult to reconstruct, load [Discovery Techniques](references/DISCOVERY.md). The evidence, provenance, contradictions, and unresolved decisions are explicit. -Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It owns only the root `CONTEXT.md` glossary update; this skill retains the product outcome and stopping boundary. +### 2. Elicit the breadth-first foundation -Invoke `$research` when a material external subject could change the product decision and needs durable evidence; read and link its report rather than copying its findings wholesale. Use an incidental lookup when the evidence need not persist. Research returns evidence to this product session and owns neither `PRODUCT.md` nor continuation. +Invoke `$elicit-with-context` for material user-held information and decisions throughout steps 2–4. Reuse compatible confirmed answers and derive discoverable facts before questioning. Invoke `$research` when a material external claim could change a requirement and its evidence must persist; link the resulting report rather than copying it. -Keep `PRODUCT.md` unchanged until the complete whole-product synthesis is explicitly confirmed. Before then, supporting utilities may perform only the writes authorised by their own contracts. +Establish the foundation from problem to design inputs: -### 3. Map the whole product +1. authority and executive orientation; +2. purpose, problem, vision, users, actors, stakeholders, and needs; +3. value, product and business goals, observable success measures, scope, boundaries, and non-goals; +4. current situation and confirmed target direction when an existing system makes the distinction material; +5. end-to-end journeys from entry through value, recovery, support, and completion or exit; +6. a whole-product functional capability catalogue; and +7. system-wide quality, context, interface, data, integration, security, privacy, safety, compliance, technology, platform, operational, and lifecycle requirements. -Map the product breadth-first before exploring any one capability. Walk each intended user from entry through recurring value, recovery, and exit or completion where applicable; use those journeys to expose the natural user-facing product areas and their high-level capabilities. Add commercial, account, trust, support, and genuinely cross-cutting areas only where the product needs them. Use the big-picture guidance in [Discovery Techniques](references/DISCOVERY.md) when the product surface is difficult to expose. The whole product is visible at low resolution without stories, screens, releases, or priorities. +Map the whole product before deepening any capability or technical concern. Omit a foundation concern only after establishing that it is immaterial. Record commercial or market detail only when it changes requirements. The product surface and every material system-wide design input are visible without becoming a feature specification or solution design. -### 4. Clarify the capability catalogue +### 3. Shape traceable requirements -Clarify every mapped capability only to mini-brief depth: who benefits and the value they receive, the externally observable high-level behaviour, and meaningful boundaries that prevent a later feature session from assuming adjacent behaviour. For an existing product, distinguish `Current` observed behaviour, `Confirmed direction`, and uncommitted `Idea` where the distinction prevents conflation; these describe product truth, not delivery progress or priority. Do not manufacture current claims for a blank product. Leave detailed rules, states, scenarios, acceptance conditions, technical solution, architecture, sequencing, releases, estimates, and tickets to their owning later sessions. Each capability is independently understandable without becoming a feature specification. +Group capabilities under natural product areas. Give each a stable `CAP-*` identifier and a mini-brief containing its actors and value, high-level externally observable responsibility, meaningful inclusions and exclusions, and source need, goal, journey, obligation, or evidence. Do not add routine delivery or current-state metadata. Leave user stories, detailed functional and non-functional feature requirements, business rules, states, scenarios, examples, and acceptance criteria to later feature specifications. -### 5. Confirm the whole-product synthesis +Give every material technical requirement a stable semantic identifier, such as `QUAL-*`, `INT-*`, `DATA-*`, `SEC-*`, `TECH-*`, or `OPS-*`. State the requirement, rationale and source, affected scope, and observable measure or later-verifiable response when applicable. Distinguish hard constraints from preferences. Record mandated external contracts, platforms, runtimes, database families or drivers, hosting limits, and standards as constraints; do not select internal boundaries, components, schemas, endpoints, payloads, algorithms, frameworks, deployment topology, controls, or other implementation mechanisms merely to fill the document. Later solution designs own the feature-specific technical how, and ADRs own qualifying architecture decisions and rationale. -Walk every journey and capability across the relevant user perspectives. Resolve material omissions, duplicates, false groupings, broken transitions, contradictions, and capability descriptions that conceal more than one product behaviour. Have `$elicit-with-context` present one complete synthesis of the product frame, journeys, and capability catalogue and obtain the user's explicit confirmation; if the user rejects or corrects it, reopen the affected product work and reconfirm the complete revised synthesis. The confirmed foundation is coherent and ready to persist. +Record assumptions, dependencies, constraints, risks, and open questions distinctly. A non-blocking unresolved item states its evidence, consequence if wrong, affected requirements, owner, and resolution trigger. Do not confirm the foundation while an unresolved item makes the product boundary unsafe or prevents responsible feature specification. -### 6. Write the product definition +### 4. Confirm readiness -After explicit confirmation, create or update the single root `PRODUCT.md` using the [Product Foundation Template](assets/product-template.md) where the project has no established convention; the invocation already supplies authority for this agreed write. Preserve unrelated confirmed content and local structure when it differs only stylistically. Record the concise frame and explicit key journeys, then use natural user-facing areas as `##` headings and capabilities as `###` headings without a redundant umbrella. Preserve confirmed status distinctions only where material. Link external claims to durable research and keep canonical term meanings in `CONTEXT.md`. `PRODUCT.md` is the sole product authority, not a backlog, status tracker, assumptions ledger, or implementation mirror. +Walk every intended user and lifecycle stakeholder across the journeys, capabilities, and applicable technical requirements. Resolve material omissions, duplicates, broken transitions, contradictory ownership, untraceable requirements, hidden feature depth, and requirements that prescribe unconfirmed design. Check that every normative requirement is necessary, feasible enough for this stage, unambiguous at its level, source-traceable, and capable of later verification. -### 7. Verify the definition +Have `$elicit-with-context` present one complete synthesis covering the entire proposed foundation, non-blocking unresolved items, and exact document effect. Obtain explicit confirmation of the whole synthesis. If the user rejects or corrects it, leave `PRODUCT.md` unchanged, reopen every affected branch, repeat the relevant coverage walk, and confirm the complete revision. The confirmed synthesis authorises its agreed `PRODUCT.md` write without another permission prompt. -Verify current claims against inspected evidence, direction and ideas against user confirmation, and material external claims against linked research. Check coverage from frame to journeys to every capability mini-brief, coherence across user perspectives, consistency with `CONTEXT.md`, preservation of unrelated confirmed content, and absence of architecture, detailed feature behaviour, priorities, roadmap, releases, tickets, or implementation detail. Report changed files, evidence, and limitations. Stop after `PRODUCT.md` and applicable ubiquitous language are updated and verified; later architecture and feature definition remain separate sessions. +### 5. Write the foundation + +Create or update the single root `PRODUCT.md` using the [Product and System Requirements Template](assets/product-template.md). Keep one authority even when some sections are omitted as immaterial. Preserve unrelated confirmed content and stylistic local structure. Use stable identifiers and direct links so later feature specifications and solution designs can trace to the foundation without copying it. Assign identifiers to capabilities, material technical requirements, product-level acceptance, and unresolved records that need downstream reference; do not number every narrative section, goal, measure, or journey by default. + +State each fact and requirement once, then link its identifier. Keep journeys, readiness, acceptance, and evidence from restating the catalogue; add no traceability matrix when inline source traces already establish coverage. + +### 6. Verify and stop + +Verify intended direction against the confirmed synthesis, current claims against inspected evidence, terminology against `CONTEXT.md`, and material external claims against cited research. Check authority, scope, journeys, capability and technical-requirement coverage, identifier uniqueness, source traceability, internal consistency, explicit non-blocking uncertainty, product-level acceptance, and foundation readiness. Confirm the absence of delivery status, priorities, roadmap, releases, tickets, detailed feature acceptance, selected feature implementation, internal architecture inventory, and unsupported precision. + +Report changed files, supporting evidence, verification, non-blocking unresolved items, and limitations. Stop after `PRODUCT.md` and the context changes owned by invoked skills are verified. Do not create feature specifications, solution designs, ADR decisions, work items, scaffolding, implementation, releases, or deployments. diff --git a/skills/define-product/agents/openai.yaml b/skills/define-product/agents/openai.yaml index 967400c..9b960e1 100644 --- a/skills/define-product/agents/openai.yaml +++ b/skills/define-product/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Define Product' - short_description: 'Establish a durable whole-product foundation' + short_description: 'Define product and system requirements' policy: allow_implicit_invocation: false diff --git a/skills/define-product/assets/product-template.md b/skills/define-product/assets/product-template.md index 8f09af1..8f72b23 100644 --- a/skills/define-product/assets/product-template.md +++ b/skills/define-product/assets/product-template.md @@ -1,45 +1,119 @@ # {Product name} -{Concise executive orientation stating who the product serves, the outcome it enables, and what distinguishes its approach. Include overall maturity only when it materially improves orientation.} +> **Authority:** This document owns confirmed whole-product intent and system-wide requirements. Feature specifications and solution designs refine it without silently changing it. -## Purpose and vision + -{The product's purpose, enduring vision, and positive change, independent of a particular implementation.} +## Executive summary -## Intended users and needs +{Orient readers with intended users, problem, product outcome, distinguishing approach, and consequential boundaries or constraints.} -- **{User or customer}**: {Needs and desired outcomes.} +## Purpose, problem, and vision -## Value proposition and market position +{Describe the problem, why it matters, the product's purpose, and the positive change sought without prescribing implementation.} -{Why intended users would choose this product over meaningful alternatives, manual work, or doing nothing. Describe supported differentiation and link material external claims to the applicable research report.} +## Users, actors, stakeholders, and needs -## Outcomes and success signals +- **{Role or group}:** {Context, needs, pain, and desired outcomes.} -{Describe the product and business outcomes that justify investment and the observable signals of progress. Include the business model and pricing only when material.} +## Value, goals, and success measures -## Product boundaries and non-goals +{Record goals with observable measures, baselines, direction or threshold, measurement source, and horizon where known. Include commercial or market position only when it changes requirements.} -- {A durable boundary or non-goal and the focus it preserves.} +## Scope, boundaries, and non-goals -## Key user journeys +### In scope + +- {Included outcome or responsibility.} + +### Non-goals + +- {Durable exclusion and the focus or ownership boundary it preserves.} + +## Current situation and target direction + + + +{Current evidence, target direction, and material gaps.} + +## Key journeys ### {Journey name} -**Actors:** {Intended users or other consequential actors.} +**Actors:** {Users, stakeholders, or external systems.} + +{Narrate entry, outcome, major stages, consequential interactions, recovery or support, and completion or exit. Reference capabilities without repeating their mini-briefs or adding feature scenarios.} + +## Functional capabilities + +### {Natural product area} + +#### CAP-01 — {Capability name} + +- **Actors and value:** {Beneficiaries, need, and enabled outcome.} +- **High-level responsibility:** {One coherent statement of externally observable product behaviour.} +- **Boundaries:** {Important inclusions, exclusions, and adjacent capabilities.} +- **Source trace:** {Need, goal, journey, obligation, research, or current evidence.} + + + + + +### {PREFIX}-01 — {Requirement name} + +{One direct normative requirement.} + +- **Source:** {Confirmed or cited rationale and evidence.} +- **Scope:** {Journeys, capabilities, actors, data, interfaces, or environments governed.} + +## Quality requirements + +{Use `QUAL-*` records for material qualities in their operating context.} + +## System context and external interfaces + +{Briefly describe people and external systems, ownership, exchanges, direction, and trust boundaries. Use `INT-*` records for binding interaction, protocol, standard, ownership, and failure expectations without inventing endpoints, payloads, or adapters. Add a context diagram only when it materially clarifies relationships.} + +## Data and information requirements + +{Use `DATA-*` records for material purpose, categories, sources or sinks, ownership, sensitivity, lifecycle, retention, residency, provenance, scale, or transition obligations.} + +## Integrations and external dependencies + +{Use `INT-*` records for external parties or systems, purpose, exchanges, ownership, mandated contracts, support expectations, failure consequences, or substitution constraints. Link the applicable interface records instead of repeating them.} + +## Security, privacy, safety, and compliance + +{Use `SEC-*` records for protection or assurance outcomes, actor or data scope, trust boundaries, audit obligations, applicable instruments and versions, and risk tolerance.} + +## Technology and platform constraints + +{Use `TECH-*` records for mandated or prohibited platforms, runtimes, languages, database families or drivers, hosting, environments, standards, supported clients, licences, or vendors. State hard versus preferred.} + +## Operational and lifecycle requirements + +{Use `OPS-*` records for availability, support, recovery, capacity, observability, audit, deployment restrictions, maintainability, export, archival, continuity, or retirement outcomes.} + +## Assumptions, dependencies, constraints, risks, and open questions + + -{Concise journey from entry through recurring value, recovery, and exit or completion where applicable. Reference the user-facing areas and capabilities involved without adding screen flows or acceptance criteria.} +- **Assumption `ASM-01` — {claim}:** {Evidence; consequence if wrong and affected IDs; owner and resolution trigger.} +- **Dependency `DEP-01` — {condition or party}:** {Evidence; consequence and affected IDs; owner and trigger.} +- **Constraint `CON-01` — {limit}:** {Source and rationale; affected scope; change owner or trigger.} +- **Risk `RSK-01` — {uncertainty}:** {Evidence; consequential effect and affected IDs; owner and review trigger.} +- **Open question `OQ-01` — {question}:** {Why non-blocking; consequence and affected IDs; decision owner and trigger.} - +## Product-level acceptance and foundation readiness -## {Capability area} +### Product-level acceptance -{Optional one-sentence orientation when the area's purpose is not clear from its name.} +- **ACC-01:** {Whole-product journey, system-wide outcome, or conformance obligation that one feature cannot own cleanly.} -### {Capability name} +### Foundation readiness - +{State evidence that the foundation is coherent and safe for downstream feature specification, plus residual limitations. Do not restate the catalogue, self-certify with a checklist, or claim uninspected evidence.} -_Status: {Current | Confirmed direction | Idea}._ +## Evidence and traceability -{Concise prose stating who benefits and why, externally observable high-level behaviour, and meaningful boundaries. Leave detailed rules, states, scenarios, acceptance conditions, architecture, implementation, priorities, roadmaps, backlog items, sequencing, releases, estimates, and tickets to their owning later work. Link material external claims to their authoritative research report.} +{Link supplied or inspected research, current-system evidence, applicable ADRs, and `CONTEXT.md`. Do not invent dates, roles, or provenance. Do not add a traceability matrix when the identifiers and inline source traces already establish coverage.} diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index 65359a0..095ce6a 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -1,53 +1,19 @@ # Discovery Techniques -Load only the section needed by the active discovery branch. `$elicit-with-context` coordinates material questioning, final synthesis confirmation, and glossary maintenance; its composed skills retain their own results. `$research` owns durable external evidence. `define-product` retains the product outcome and stopping boundary. +Load only the branch needed when direct inspection and ordinary elicitation do not expose the foundation. `$elicit-with-context` owns shared-understanding questioning; `$research` owns durable external evidence; `define-product` retains `PRODUCT.md` and the stopping boundary. -## Establish an existing product +## Reconstruct an existing product -Use **repository archaeology** to recover available behaviour before asking the user. Start with root documentation and manifests, then sample user entry points, public contracts, data boundaries, tests, and operational configuration. Follow evidence until the product areas and feature candidates stabilise. Treat absence as unknown rather than proof, distinguish shipped behaviour from abandoned or planned code, and ask the user to resolve contradictions with stated intent. +Inspect the smallest end-to-end slices that reveal actors, product boundaries, current behaviour, data ownership, external systems, and consequential constraints. Prefer public documentation and contracts, entry points, representative tests, schemas, configuration, and runtime evidence over an implementation inventory. Sample one normal path, one refusal or failure, one recovery path, and one operational or support path where applicable. -## Complete the product frame +Distinguish shipped behaviour from planned or abandoned code. Treat names, code structure, framework choices, and deployment configuration as evidence of the present system, not durable target requirements. Reconcile contradictory evidence with the user instead of selecting a source silently. Use legacy product or architecture documents as evidence only until their statements are confirmed into the current authority. -Use the **Product Vision Board** as a completeness check for the concise opening: +## Expose the whole product -1. What positive change and enduring purpose define the vision? -2. Which users, customers, and influential actors matter, and what outcomes do they need? -3. Why would they choose this product over meaningful alternatives, manual work, or doing nothing? -4. What market position or differentiation matters, when applicable? -5. Which product or business outcomes and observable success signals matter, and are business model or pricing decisions material now? -6. Which boundaries and non-goals keep the product coherent? +Walk each intended user from entry through recurring value, consequential interaction, failure or refusal, recovery or support, and completion or exit. Change perspective to customers, operators, administrators, support, external-system owners, affected parties, and regulators when material. Derive natural product areas and capability mini-briefs from those journeys, then check commercial, account, trust, data, integration, operational, and lifecycle surfaces that the main journey may hide. -Keep this frame proportionate to its supporting role. Move into whole-product capability discovery once these dimensions are clear. +Use meaningful alternatives, manual work, and doing nothing to clarify value and boundaries. Invoke `$research` only when external position, regulation, standards, providers, or user evidence could materially change a requirement and the evidence must survive the session. -## Map the big picture +## Expose system-wide requirements -Use the **mile-wide, inch-deep** pass before local detail: - -1. Name the product's main users and their entry points. -2. Narrate how each user moves from first contact through recurring value and eventual exit or completion. -3. Include recurring value, recovery, and exit or completion where they apply. -4. Record the user-facing product areas along those journeys. -5. Beneath each area, inventory capabilities at one-line resolution. -6. For an existing product, distinguish observed current behaviour, confirmed direction, and uncommitted ideas where material. -7. Add commercial, account, trust, support, and other cross-cutting capabilities that the main journeys did not expose. - -This pass creates a capability surface, not cards, stories, priorities, screens, releases, or architecture. - -## Clarify a capability - -Resolve only the information required by the mini-brief: - -- **User value**: who benefits and what becomes possible or easier? -- **High-level behaviour**: what does the product do from the user's perspective? -- **Meaningful boundaries**: what nearby behaviour could a later feature session reasonably but incorrectly assume belongs to it? -- **Status when material**: for an existing product, is it observed `Current` behaviour, `Confirmed direction`, or an uncommitted `Idea`? - -Use a concise example only when high-level behaviour remains ambiguous. Stop when a later feature-definition session can understand the capability's product intent without receiving detailed rules, states, scenarios, acceptance conditions, implementation, or delivery design. - -## Walk the complete catalogue - -Read the mapped product back as one narrative. Change perspective across intended users and check entry, recurring-use, recovery, commercial, trust, support, and exit paths where applicable. Look for missing transitions, duplicated capabilities, false groupings, contradictions, and mini-briefs that hide more than one distinct product behaviour. Return each material finding to elicitation, then repeat the walk until it exposes nothing new. - -## Establish external position - -Compare the product with meaningful alternatives users employ today, including manual work and doing nothing. Invoke `$research` only when competitor capabilities, market conditions, pricing, standards, regulation, or user evidence could materially change a product decision and need a durable evidence record. Keep researched claims in the report and link that report from `PRODUCT.md` rather than copying its findings; use an incidental lookup when no durable evidence record is needed. +For each journey and capability, ask what must remain true across likely implementations: operating conditions, measurable qualities, external ownership and exchanges, information duties, trust and harm boundaries, mandated technology, provider or organisational constraints, support and recovery, and eventual export, archival, or retirement. Record only material design inputs. Leave the selected internal structure and mechanisms to later solution design. diff --git a/skills/design-feature/SKILL.md b/skills/design-feature/SKILL.md index ea27e3d..49651eb 100644 --- a/skills/design-feature/SKILL.md +++ b/skills/design-feature/SKILL.md @@ -12,7 +12,7 @@ Designs one approved feature specification into one buildable technical solution ## Prerequisites -Require the approved feature `specification.md`, root `ARCHITECTURE.md`, and relevant root `CONTEXT.md` and ADRs. Inspect them before continuing. When an authority is missing, inadequate, or contradictory in a way that prevents a complete solution, report the exact prerequisite or gap and stop. Leave missing feature intent to a later `$specify-feature` session and missing system-wide constraints to a later `$define-architecture` session; do not invoke either or manufacture its outcome. +Require the approved feature `specification.md`, root `PRODUCT.md`, and relevant root `CONTEXT.md` and ADRs. Inspect them before continuing. When an authority is missing, inadequate, or contradictory in a way that prevents a complete solution, report the exact prerequisite or gap and stop. Leave missing feature intent to a later `$specify-feature` session and missing system-wide requirements to a later `$define-product` session; do not invoke either or manufacture its outcome. Refine an existing paired solution design when requested, preserving unrelated confirmed content and local conventions. @@ -20,9 +20,9 @@ Refine an existing paired solution design when requested, preserving unrelated c ### 1. Freeze authorities and evidence -Inspect the request, prerequisite authorities, existing paired solution design, repository guidance, and the project's feature-record convention. Inspect only the relevant code, tests, contracts, schemas, configuration, runtime and deployment evidence, similar implemented patterns, and material external technical or regulatory evidence needed to design this feature. Treat the specification as authority for what and why, `ARCHITECTURE.md` and applicable ADRs as authority for system-wide constraints, `CONTEXT.md` as authority for project language, and current implementation as evidence of current constraints and idioms rather than automatic intended design. The user remains authority for material system-specific trade-offs, feature scope, and any proposed departure from the architecture. +Inspect the request, prerequisite authorities, existing paired solution design, repository guidance, and the project's feature-record convention. Inspect only the relevant code, tests, contracts, schemas, configuration, runtime and deployment evidence, similar implemented patterns, and material external technical or regulatory evidence needed to design this feature. Treat the specification as authority for detailed feature intent, `PRODUCT.md` as authority for its product foundation and system-wide requirements, applicable ADRs as authority for their accepted decisions, `CONTEXT.md` as authority for project language, and current implementation as evidence of current constraints and idioms rather than automatic intended design. The user remains authority for material system-specific trade-offs, feature scope, and any proposed departure from those authorities. -Separate confirmed future behaviour, architecture constraints, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide decision is absent or the feature would require changing the architecture, identify the exact architecture gap and stop rather than embedding a new foundation in this design. +Separate confirmed future behaviour, system-wide requirements and accepted decisions, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide requirement is absent or the feature would require changing the product foundation, identify the exact foundation gap and stop rather than embedding it in this design. Invoke `$elicit-with-context` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. @@ -32,9 +32,9 @@ The solution-design authority, runtime evidence, unresolved gaps, and permitted ### 2. Map behaviour to design forces -Map every material behaviour, rule, state, scenario, external constraint, and acceptance condition in the specification to affected system responsibilities and a credible verification seam. Identify the feature's entry points, trust boundaries, owned knowledge, data and integration flows, current-versus-target differences, and material quality, migration, operational, compatibility, and evolution forces. Trace each proposed new component, contract, data shape, migration, or operational obligation back to a specification behaviour or architecture need; remove anything with no such origin. +Map every material behaviour, rule, state, scenario, external constraint, and acceptance condition in the specification to affected system responsibilities and a credible verification seam. Identify the feature's entry points, trust boundaries, owned knowledge, data and integration flows, current-versus-target differences, and material quality, migration, operational, compatibility, and evolution forces. Trace each proposed new component, contract, data shape, migration, or operational obligation back to a specification behaviour, system-wide requirement, or applicable ADR; remove anything with no such origin. -Invoke `$modular-design` with the confirmed behaviour, architecture constraints, project language, and current evidence. Use its structural findings as passive constraints on ownership, information hiding, cohesion, contracts, dependency direction, seams, and change locality while retaining responsibility for alternatives, user decisions, the complete solution, verification, and stopping. +Invoke `$modular-design` with the confirmed behaviour, system-wide requirements, applicable ADRs, project language, and current evidence. Use its structural findings as passive constraints on ownership, information hiding, cohesion, contracts, dependency direction, seams, and change locality while retaining responsibility for alternatives, user decisions, the complete solution, verification, and stopping. The affected architecture, design forces, risks, and two-way traceability are complete enough to compare coherent solutions without selecting implementation detail prematurely. diff --git a/skills/design-feature/assets/solution-design-template.md b/skills/design-feature/assets/solution-design-template.md index af59bcc..eae37ed 100644 --- a/skills/design-feature/assets/solution-design-template.md +++ b/skills/design-feature/assets/solution-design-template.md @@ -4,7 +4,7 @@ ## Summary and authorities - + ## Design drivers diff --git a/skills/maintain-decision-records/SKILL.md b/skills/maintain-decision-records/SKILL.md index eff61ef..f1c916a 100644 --- a/skills/maintain-decision-records/SKILL.md +++ b/skills/maintain-decision-records/SKILL.md @@ -21,7 +21,7 @@ Create an ADR only when the accepted technical or architectural decision passes - changing it later has meaningful cost; - a future reader would find it surprising without context; - viable alternatives created a genuine trade-off; and -- no existing durable authority already records the decision or is its better owner, including `PRODUCT.md`, `ARCHITECTURE.md`, a feature document, or a dedicated governance record. +- no existing durable authority already records the decision or is its better owner, including `PRODUCT.md`, a feature document, or a dedicated governance record. Require the user's explicit agreement to preserve the rationale durably. Agreement recorded in the relevant conversation or caller handoff satisfies this gate without another prompt. When every other gate passes but agreement is absent, ask the user one direct question. A decision that misses a gate, lacks agreement, or is routine or discoverable produces no ADR. diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md index 859fb5e..f93d394 100644 --- a/skills/plan-retirement/SKILL.md +++ b/skills/plan-retirement/SKILL.md @@ -14,7 +14,7 @@ Turns evidence and user-authoritative decisions into one proportionate, reversib ### 1. Establish the retirement evidence -Inspect the request, repository guidance, product intent, `CONTEXT.md`, `ARCHITECTURE.md`, applicable ADRs, prior decisions and plans, and the smallest sufficient current implementation and runtime evidence. Trace the capability through entry points; users, tenants, callers, and operators; internal and external dependencies; public and private interfaces, APIs, events, schemas, configuration and flags; data ownership, flows, access, retention, archival, and recovery; integrations and providers; permissions and trust boundaries; observability; runbooks, support and communications; documentation; and known replacements. +Inspect the request, repository guidance, product intent and system-wide requirements in `PRODUCT.md`, `CONTEXT.md`, applicable ADRs, prior decisions and plans, and the smallest sufficient current implementation and runtime evidence. Trace the capability through entry points; users, tenants, callers, and operators; internal and external dependencies; public and private interfaces, APIs, events, schemas, configuration and flags; data ownership, flows, access, retention, archival, and recovery; integrations and providers; permissions and trust boundaries; observability; runbooks, support and communications; documentation; and known replacements. Treat code, tests, contracts, configuration, telemetry, inventories, tracker records, and operational systems as evidence within their actual scope and freshness. Distinguish confirmed use, confirmed non-use, and unknown use. Absence of observed traffic, repository references, or known consumers is not proof of no dependence. Identify lifecycle authorities and contradictions without copying mutable operational or tracker state into the plan. The capability boundary, evidence provenance, affected parties, obligations, current unknowns, and accessible authority are explicit. diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index 68a4bf1..1ee21b1 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -16,7 +16,7 @@ Produces one canonical Markdown assessment and a user-friendly disposable HTML e Inspect the request, repository guidance, available tools, and current repository state. Establish the explicit codebase scope, exclusions, review purpose, applicable architecture or maintainability qualities, and any user-held priority that could change qualification or rank. Do not silently expand an ambiguous scope to the whole repository. Ask only for missing scope or priority that cannot be discovered and would materially change the result. -Capture the reviewed revision and starting worktree state. Read root `ARCHITECTURE.md`, `CONTEXT.md`, applicable ADRs and product constraints when present, then inspect only the scoped implementation, tests, contracts, schemas, configuration, history, and operational evidence needed to assess the area. Treat missing or conflicting authorities as limitations unless they prevent a responsible comparison, in which case expose the exact gap. Separate intended constraints, observed implementation, inferences, and unknowns. The review boundary, authorities, evidence baseline, and limitations are explicit before findings are formed. +Capture the reviewed revision and starting worktree state. Read root `PRODUCT.md`, `CONTEXT.md` and applicable ADRs when present, then inspect only the scoped implementation, tests, contracts, schemas, configuration, history, and operational evidence needed to assess the area. Treat missing or conflicting authorities as limitations unless they prevent a responsible comparison, in which case expose the exact gap. Separate intended constraints, observed implementation, inferences, and unknowns. The review boundary, authorities, evidence baseline, and limitations are explicit before findings are formed. ### 2. Map structural evidence diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md index 6352d24..398bf97 100644 --- a/skills/triage-work/SKILL.md +++ b/skills/triage-work/SKILL.md @@ -26,7 +26,7 @@ For each selected item, read its description, attachments, discussion, source re ### 2. Establish enough evidence to classify -Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, `ARCHITECTURE.md`, ADRs, documentation, public contracts, tests, code, operational evidence, and external authorities only as far as each item needs. Distinguish intended behaviour, current observable behaviour, reported observation, and unsupported inference. Seek enough corroboration to decide a disposition, not a root cause or solution. A safe focused observation may establish that a defect report is credible; full reproduction, causal investigation, mitigation, repair, or feature discovery belongs to its later routed session. +Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, ADRs, documentation, public contracts, tests, code, operational evidence, and external authorities only as far as each item needs. Distinguish intended behaviour, current observable behaviour, reported observation, and unsupported inference. Seek enough corroboration to decide a disposition, not a root cause or solution. A safe focused observation may establish that a defect report is credible; full reproduction, causal investigation, mitigation, repair, or feature discovery belongs to its later routed session. Search open and closed tracker work and relevant repository records using source identifiers, authoritative links, actors, outcomes, symptoms, affected concepts, and discriminating details. Classify a candidate internally as exact match, possible overlap, related work, or unrelated. Mark a duplicate only when the evidence establishes the same requested outcome or fault; preserve an uncertain resemblance as related or unresolved and state what would distinguish it. From f8aeda91e22b0bbcd44665ca195f45aac5b404ce Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Thu, 20 Aug 2026 09:29:19 +0100 Subject: [PATCH 55/79] feat(skills): consolidate feature specification workflow --- AGENTS.md | 1 + README.md | 10 +- SKILLS_PLAN.md | 44 +++--- skills/create-tickets/SKILL.md | 16 +-- skills/create-tickets/agents/openai.yaml | 2 +- skills/define-product/SKILL.md | 10 +- .../define-product/assets/product-template.md | 2 +- skills/define-product/references/DISCOVERY.md | 2 +- skills/design-feature/SKILL.md | 97 ------------- skills/design-feature/agents/openai.yaml | 5 - .../assets/solution-design-template.md | 69 ---------- skills/refine-backlog/SKILL.md | 6 +- skills/review-incident/SKILL.md | 2 +- skills/specify-feature/SKILL.md | 50 +++---- skills/specify-feature/agents/openai.yaml | 2 +- .../assets/specification-template.md | 129 ++++++++++++++---- skills/triage-work/SKILL.md | 4 +- 17 files changed, 163 insertions(+), 288 deletions(-) delete mode 100644 skills/design-feature/SKILL.md delete mode 100644 skills/design-feature/agents/openai.yaml delete mode 100644 skills/design-feature/assets/solution-design-template.md diff --git a/AGENTS.md b/AGENTS.md index 5f96e9d..46edf4d 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -2,5 +2,6 @@ - When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. - When eliciting within a software project, invoke `$elicit-with-context` instead of `$elicit`; use `$elicit` directly only for non-software work or as part of `$elicit-with-context`. - When composing skills, rely on each invoked skill's public contract; do not restate or steer its internal logic. +- Default to a user-agent workflow; introduce team roles, named accountable people, approval bodies, or third-party coordination only when the user or project evidence requires them. - Assume the execution harness supports fresh agents; do not add fallback handling for their unavailability. - After implementing changes run `bun run checks` before handoff. diff --git a/README.md b/README.md index 754bf7c..7a29ee8 100644 --- a/README.md +++ b/README.md @@ -6,7 +6,7 @@ Propulsion is a compact, composable set of Agent Skills for practical software d Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. -Substantial work narrows breadth-first over several sessions: product intent becomes a feature specification, then a solution design, implementation-ready tickets, and finally one implemented body of work per ticket. Small, understood work can begin directly with `$implement`. +Substantial work narrows breadth-first over several sessions: product intent becomes a decision-complete feature specification, then implementation-ready tickets, and finally one implemented body of work per ticket. Small, understood work can begin directly with `$implement`. ## Installation @@ -36,7 +36,7 @@ Repeat the link for each selected skill. Codex and OpenCode both discover skills | When you need to… | Start with… | Session outcome | | --- | --- | --- | | Establish or revise a product | `$define-product` | Confirmed product and system requirements | -| Define a substantial feature | `$specify-feature` | An approved, observable feature specification | +| Define a substantial feature | `$specify-feature` | An approved, decision-complete feature specification | | Make a small understood change | `$implement` | A minimal, reviewed, verified local change | | Route incoming requests | `$triage-work` | Verified dispositions and lifecycle routes | | Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | @@ -54,12 +54,11 @@ The lifecycle areas are useful entry points, not mandatory phase gates. ### Define -- `$specify-feature` turns one capability, enhancement, or agreed idea into a feature specification describing externally observable intent, without choosing its implementation. +- `$specify-feature` turns one high-level request into a single feature specification containing approved feature intent and its selected buildable solution. ### Plan -- `$design-feature` selects a buildable, system-specific technical solution for an approved feature specification. -- `$create-tickets` turns an approved feature specification and solution design, or an approved retirement plan, into verified, dependency-aware work in the project's task-management tool. +- `$create-tickets` turns an approved feature specification or retirement plan into verified, dependency-aware work in the project's task-management tool. ### Refine @@ -99,7 +98,6 @@ Deliver a substantial feature: ```text specify-feature - → later design-feature → later create-tickets → implement each selected ticket in a separate session → commit whenever a coherent unit is ready diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index c6aa835..9ccd9c0 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -15,7 +15,7 @@ Propulsion is not a workflow platform. It does not carry a product automatically ## Design principles 1. **One bounded session outcome.** Each lifecycle skill represents a natural working session and ends with one independently useful outcome. -2. **Narrow progressively.** Product intent narrows into a feature specification, solution design, implementation-ready tickets and, finally, one implemented body of work. +2. **Narrow progressively.** Product intent narrows into a decision-complete feature specification, implementation-ready tickets and, finally, one implemented body of work. 3. **Enter proportionately.** Substantial work uses the complete applicable chain. Small, understood work may begin directly with implementation. 4. **Compose within the requested outcome.** A skill may invoke another skill when its outcome is required to complete the requested session. It must not silently advance into a later lifecycle outcome. 5. **Keep authorities separate.** Product documents own intent, code and tests own implemented behaviour, Git owns change history and the task-management tool owns work state. @@ -30,8 +30,8 @@ Propulsion is not a workflow platform. It does not carry a product automatically | Area | Purpose | | --- | --- | | Establish | Create the durable product and system requirements foundation for a new or existing system. | -| Define | Turn one product capability or enhancement into an approved statement of intended behaviour. | -| Plan | Select a buildable solution and decompose it into implementation-ready work. | +| Define | Turn one high-level feature request into an approved statement of intended behaviour and its selected buildable solution. | +| Plan | Decompose an approved specification into implementation-ready work. | | Refine | Keep incoming and existing work current, prioritised, ready and feasible for the next delivery period. | | Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | | Debug and Respond | Diagnose and repair defects, stabilise active incidents and learn from them afterwards. | @@ -41,7 +41,7 @@ These areas are entry points, not mandatory phase gates. Release, deployment and ## Session and authority model -Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture a missing feature specification or solution design. +Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture a missing or incomplete feature specification. Reusable utilities are user-invokable and may be model-invoked when their documented trigger is present. A caller remains responsible for verifying the utility's result and completing its own outcome. @@ -74,12 +74,10 @@ Every skill: A substantial feature uses the project convention where one exists, otherwise: ```text -docs/features// -├── specification.md -└── solution-design.md +docs/features//specification.md ``` -The feature specification owns externally observable intent and constraints. The solution design owns the selected system-specific implementation. A useful test is whether a statement would remain true if the feature were rebuilt using a different architecture: if so, it normally belongs in the specification. +The feature specification contains two visibly distinct semantic parts inside one authority. Feature intent owns the problem, outcomes, scope, actors, observable behaviour, qualities, scenarios and acceptance. Selected solution owns the system-specific responsibilities, boundaries, contracts, data, integrations, controls, migration, operational effects and verification seams. The distinction prevents implementation ideas from masquerading as requirements without creating another document or session handoff. Small, understood changes do not require feature documents merely to enter implementation. @@ -99,7 +97,7 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `define-product` -- **Outcome:** A user-confirmed product and system requirements foundation broad enough to guide later feature specification and solution design without becoming either. +- **Outcome:** A user-confirmed product and system requirements foundation broad enough to guide later feature specification without becoming one. - **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, applicable ADRs, representative repository evidence and applicable external evidence. - **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. - **Composition:** Requires `elicit-with-context`; conditionally invokes `research`. @@ -109,26 +107,18 @@ The skill uses the selected tool's native issues, relationships, statuses, prior #### `specify-feature` -- **Outcome:** One approved feature or enhancement expressed as externally observable intent. -- **Inputs:** A capability from `PRODUCT.md`, an already-elicited conversation or another agreed feature idea, plus `CONTEXT.md`, relevant current behaviour and supporting evidence. +- **Outcome:** One approved, decision-complete feature specification containing both feature intent and its selected buildable solution. +- **Inputs:** Any high-level feature request, conversation, product capability or existing specification, plus whatever project and current-system evidence is available. No earlier lifecycle artefact is mandatory. - **Output:** `docs/features//specification.md`, or the project's equivalent. -- **Composition:** Requires `elicit-with-context` and `maintain-ubiquitous-language`; conditionally invokes `research`. -- **Stops:** After the problem, actors, behaviour, rules, states, scenarios, acceptance conditions, constraints and exclusions are unambiguous. It does not select the implementation or create tickets. +- **Composition:** Requires `elicit-with-context`; conditionally invokes `research` and `modular-design`. +- **Stops:** After the problem, actors, use cases, behaviour, rules, states, qualities, experience, scenarios, acceptance, selected responsibilities, contracts, data, integrations, controls, migration, operational effects and verification seams are resolved enough for ticket decomposition. It does not create tickets or implement the feature. ### Plan -#### `design-feature` - -- **Outcome:** An approved, buildable technical solution for one feature specification. -- **Inputs:** The approved feature specification, `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. -- **Output:** `docs/features//solution-design.md` and any warranted ADRs. -- **Composition:** Requires `modular-design`; conditionally invokes `elicit-with-context`, `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. -- **Stops:** After affected boundaries, interfaces, data, integrations, security, migration, operational effects and verification seams are resolved. It does not decompose or implement the work. - #### `create-tickets` -- **Outcome:** An approved feature solution or retirement plan is represented by implementation-ready, dependency-aware work in the project's task-management tool. -- **Inputs:** Either the approved feature specification and solution design or an approved retirement plan, plus relevant project guidance and current tracker state. +- **Outcome:** An approved feature specification or retirement plan is represented by implementation-ready, dependency-aware work in the project's task-management tool. +- **Inputs:** Either one approved, decision-complete feature specification or an approved retirement plan, plus relevant project guidance and current tracker state. - **Output:** Native tracker items linked to their applicable authoritative documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. - **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. - **Stops:** After the created items and relationships are read back and verified. It does not select an iteration or begin implementation. @@ -313,7 +303,6 @@ Product definition stops after the confirmed breadth-first product and system re ```text specify-feature - → later design-feature → later create-tickets → implement each selected ticket in a separate session → commit whenever a coherent unit is ready @@ -375,7 +364,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. 2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. -3. Build the progressive foundation path: `define-product`, `specify-feature`, `design-feature` and `create-tickets`. +3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. 6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. @@ -394,8 +383,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Update `code-review` skill - [x] Update `maintain-agents` skill - [x] Update `define-product` skill -- [-] Create `specify-feature` skill -- [-] Create `design-feature` skill +- [x] Create `specify-feature` skill - [-] Create `create-tickets` skill - [-] Update `implement` skill - [x] Update `commit` skill @@ -421,7 +409,7 @@ The revised suite is ready when: - every common lifecycle task has one obvious user-facing entry point - every lifecycle skill produces one bounded outcome and stops at its documented boundary -- substantial work narrows cleanly from product foundation through specification, design, tickets and per-item implementation +- substantial work narrows cleanly from product foundation through decision-complete specification, tickets and per-item implementation - small, clear work can enter implementation without ceremonial documents or tickets - required and conditional composition is explicit and does not duplicate supporting skill instructions - task-management skills use the project tool named in `AGENTS.md` and its native concepts without Propulsion configuration diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md index 55ba570..3de8ced 100644 --- a/skills/create-tickets/SKILL.md +++ b/skills/create-tickets/SKILL.md @@ -1,6 +1,6 @@ --- name: create-tickets -description: Creates verified native tracker items from an approved feature solution or retirement plan. Use when feature delivery or capability retirement needs implementation-ready, dependency-aware tickets. +description: Creates verified native tracker items from an approved feature specification or retirement plan. Use when feature delivery or capability retirement needs implementation-ready, dependency-aware tickets. metadata: invocation: user disable-model-invocation: true @@ -8,16 +8,16 @@ disable-model-invocation: true # Create Tickets -Turns one approved feature solution or retirement plan into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules, examples, stages, and completion conditions into observable acceptance evidence. +Turns one approved feature specification or retirement plan into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules, examples, stages, and completion conditions into observable acceptance evidence. ## Prerequisites Require one applicable approved authority package and inspect it before continuing: -- For feature delivery, require the approved feature specification and its approved solution design. When either document is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent to a later `$specify-feature` session and missing technical resolution to a later `$design-feature` session; do not invoke either or manufacture its outcome. -- For capability retirement, require one approved retirement plan that resolves the retirement outcome; affected parties, dependencies, interfaces, data, and integrations; applicable deprecation, migration, coexistence, disablement, removal, archival, communication, support, and recovery obligations; material risks and rollback or recovery routes; and observable stage and final completion conditions sufficiently for honest decomposition. When a needed retirement decision is missing or contradictory, identify the exact gap and stop for a later `$plan-retirement` session; do not force retirement through feature specification or solution design or manufacture the plan's outcome. +- For feature delivery, require one approved feature specification containing decision-complete feature intent and selected solution. When it is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent or technical resolution to a later `$specify-feature` session; do not invoke it or manufacture its outcome. +- For capability retirement, require one approved retirement plan that resolves the retirement outcome; affected parties, dependencies, interfaces, data, and integrations; applicable deprecation, migration, coexistence, disablement, removal, archival, communication, support, and recovery obligations; material risks and rollback or recovery routes; and observable stage and final completion conditions sufficiently for honest decomposition. When a needed retirement decision is missing or contradictory, identify the exact gap and stop for a later `$plan-retirement` session; do not force retirement through feature specification or manufacture the plan's outcome. -A retirement plan may deliberately depend on a later feature specification and solution design for a replacement capability. Do not create implementation work for that unresolved replacement until those authorities are approved. Retirement-only obligations may be decomposed when they are independently confirmed and do not require the unresolved replacement decision. +A retirement plan may deliberately depend on a later feature specification for a replacement capability. Do not create implementation work for that unresolved replacement until the specification is approved. Retirement-only obligations may be decomposed when they are independently confirmed and do not require the unresolved replacement decision. Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user directly which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool or required account is unavailable or not writable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. @@ -25,7 +25,7 @@ Require the project's task-management tool to be named in the root `AGENTS.md`. ### 1. Fix the authorities and native surface -Inspect the request, repository guidance, applicable approved authority package, its linked product, context, architecture, ADR, and research authorities, and only the implementation evidence needed to understand existing work. For feature delivery, treat the specification as authority for observable intent and the solution design as authority for the selected implementation and genuine technical prerequisites. For retirement, treat the retirement plan as authority for the approved outcome, obligations, stages, risks, recovery, and completion conditions. Treat separately approved replacement feature documents as authority only for replacement work they actually resolve, and treat the task-management tool as authority for live work state. Discover obvious facts before asking the user. +Inspect the request, repository guidance, applicable approved authority package, its linked product, context, ADR, and research authorities, and only the implementation evidence needed to understand existing work. For feature delivery, treat the specification's feature-intent part as authority for observable behaviour and its selected-solution part as authority for technical decisions and genuine prerequisites. For retirement, treat the retirement plan as authority for the approved outcome, obligations, stages, risks, recovery, and completion conditions. Treat a separately approved replacement feature specification as authority only for replacement work it actually resolves, and treat the task-management tool as authority for live work state. Discover obvious facts before asking the user. Inspect the configured tool's available schema and capabilities before planning any mutation: native item types, bodies, links, parent or grouping concepts, dependency direction, duplicate or related relationships, and read-back operations. Use only concepts the tool actually supports. Keep the tracker read-only through proposal confirmation. @@ -45,11 +45,11 @@ Build an internal two-way coverage map from the applicable authority package to Form the smallest set of coherent vertical outcomes. Each item spans every affected layer needed to produce one observable result or fulfilled obligation for an actor, consumer, published contract, migration, retirement stage, or operational boundary. Include a shared enabling change in the earliest outcome that needs it unless an approved authority establishes a separately observable and independently verifiable prerequisite. Do not split work by frontend, backend, database, service, class, file, architecture layer, test type, retirement phase label, or another implementation inventory merely because those parts exist. One item is correct when the approved package is already one bounded independently implementable outcome. -For each item, define a concise native title, one coherent outcome, enough bounded authority context for a separate implementation session, links to the applicable approved authority package, observable acceptance evidence, and applicable relationships. A feature item links both authoritative feature documents. A retirement item links the retirement plan and any separately applicable approved replacement specification and solution design when that item includes replacement work. Derive acceptance evidence from approved rules and discriminating examples for feature delivery, or from approved stage outcomes, exit evidence, counter-signals, recovery obligations, and completion conditions for retirement. Use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. +For each item, define a concise native title, one coherent outcome, enough bounded authority context for a separate implementation session, links to the applicable approved authority package, observable acceptance evidence, and applicable relationships. A feature item links the authoritative feature specification. A retirement item links the retirement plan and any separately applicable approved replacement specification when that item includes replacement work. Derive acceptance evidence from approved rules and discriminating examples for feature delivery, or from approved stage outcomes, exit evidence, counter-signals, recovery obligations, and completion conditions for retirement. Use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. Check each item directly for independent value, bounded size, and testability. Minimise coupling between items while representing unavoidable prerequisites honestly. Infer a blocker only when the approved solution or required state transition establishes that one outcome cannot begin or complete before another; do not convert user-journey order, convenient sequencing, or tracker display order into dependency. Interpret the configured tool's relationship direction explicitly and reject cycles. -When a question exposes missing feature intent or technical design, stop with the owning source gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification and solution design exist; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit-with-context` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. +When a question exposes missing feature intent or selected solution, stop with the owning specification gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification exists; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit-with-context` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. The draft items are vertically coherent, source-complete, independently actionable where possible, and linked only by genuine acyclic prerequisites. diff --git a/skills/create-tickets/agents/openai.yaml b/skills/create-tickets/agents/openai.yaml index 65486ce..42f7453 100644 --- a/skills/create-tickets/agents/openai.yaml +++ b/skills/create-tickets/agents/openai.yaml @@ -1,6 +1,6 @@ interface: display_name: 'Create Tickets' short_description: 'Create delivery or retirement tracker work' - default_prompt: 'Use $create-tickets to turn this approved feature solution or retirement plan into implementation-ready tracker items.' + default_prompt: 'Use $create-tickets to turn this approved feature specification or retirement plan into implementation-ready tracker items.' policy: allow_implicit_invocation: false diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index f8f03b7..51acb60 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -8,7 +8,7 @@ disable-model-invocation: true # Define Product -Turns a product idea or existing system into one confirmed `PRODUCT.md` that supplies product intent and system-wide requirements to later feature specification and solution design. +Turns a product idea or existing system into one confirmed `PRODUCT.md` that supplies product intent and system-wide requirements to later feature specifications. ## Process @@ -34,13 +34,13 @@ Establish the foundation from problem to design inputs: 6. a whole-product functional capability catalogue; and 7. system-wide quality, context, interface, data, integration, security, privacy, safety, compliance, technology, platform, operational, and lifecycle requirements. -Map the whole product before deepening any capability or technical concern. Omit a foundation concern only after establishing that it is immaterial. Record commercial or market detail only when it changes requirements. The product surface and every material system-wide design input are visible without becoming a feature specification or solution design. +Map the whole product before deepening any capability or technical concern. Omit a foundation concern only after establishing that it is immaterial. Record commercial or market detail only when it changes requirements. The product surface and every material system-wide design input are visible without becoming a feature specification. ### 3. Shape traceable requirements Group capabilities under natural product areas. Give each a stable `CAP-*` identifier and a mini-brief containing its actors and value, high-level externally observable responsibility, meaningful inclusions and exclusions, and source need, goal, journey, obligation, or evidence. Do not add routine delivery or current-state metadata. Leave user stories, detailed functional and non-functional feature requirements, business rules, states, scenarios, examples, and acceptance criteria to later feature specifications. -Give every material technical requirement a stable semantic identifier, such as `QUAL-*`, `INT-*`, `DATA-*`, `SEC-*`, `TECH-*`, or `OPS-*`. State the requirement, rationale and source, affected scope, and observable measure or later-verifiable response when applicable. Distinguish hard constraints from preferences. Record mandated external contracts, platforms, runtimes, database families or drivers, hosting limits, and standards as constraints; do not select internal boundaries, components, schemas, endpoints, payloads, algorithms, frameworks, deployment topology, controls, or other implementation mechanisms merely to fill the document. Later solution designs own the feature-specific technical how, and ADRs own qualifying architecture decisions and rationale. +Give every material technical requirement a stable semantic identifier, such as `QUAL-*`, `INT-*`, `DATA-*`, `SEC-*`, `TECH-*`, or `OPS-*`. State the requirement, rationale and source, affected scope, and observable measure or later-verifiable response when applicable. Distinguish hard constraints from preferences. Record mandated external contracts, platforms, runtimes, database families or drivers, hosting limits, and standards as constraints; do not select internal boundaries, components, schemas, endpoints, payloads, algorithms, frameworks, deployment topology, controls, or other implementation mechanisms merely to fill the document. Later feature specifications own the feature-specific technical how, and ADRs own qualifying architecture decisions and rationale. Record assumptions, dependencies, constraints, risks, and open questions distinctly. A non-blocking unresolved item states its evidence, consequence if wrong, affected requirements, owner, and resolution trigger. Do not confirm the foundation while an unresolved item makes the product boundary unsafe or prevents responsible feature specification. @@ -52,7 +52,7 @@ Have `$elicit-with-context` present one complete synthesis covering the entire p ### 5. Write the foundation -Create or update the single root `PRODUCT.md` using the [Product and System Requirements Template](assets/product-template.md). Keep one authority even when some sections are omitted as immaterial. Preserve unrelated confirmed content and stylistic local structure. Use stable identifiers and direct links so later feature specifications and solution designs can trace to the foundation without copying it. Assign identifiers to capabilities, material technical requirements, product-level acceptance, and unresolved records that need downstream reference; do not number every narrative section, goal, measure, or journey by default. +Create or update the single root `PRODUCT.md` using the [Product and System Requirements Template](assets/product-template.md). Keep one authority even when some sections are omitted as immaterial. Preserve unrelated confirmed content and stylistic local structure. Use stable identifiers and direct links so later feature specifications can trace to the foundation without copying it. Assign identifiers to capabilities, material technical requirements, product-level acceptance, and unresolved records that need downstream reference; do not number every narrative section, goal, measure, or journey by default. State each fact and requirement once, then link its identifier. Keep journeys, readiness, acceptance, and evidence from restating the catalogue; add no traceability matrix when inline source traces already establish coverage. @@ -60,4 +60,4 @@ State each fact and requirement once, then link its identifier. Keep journeys, r Verify intended direction against the confirmed synthesis, current claims against inspected evidence, terminology against `CONTEXT.md`, and material external claims against cited research. Check authority, scope, journeys, capability and technical-requirement coverage, identifier uniqueness, source traceability, internal consistency, explicit non-blocking uncertainty, product-level acceptance, and foundation readiness. Confirm the absence of delivery status, priorities, roadmap, releases, tickets, detailed feature acceptance, selected feature implementation, internal architecture inventory, and unsupported precision. -Report changed files, supporting evidence, verification, non-blocking unresolved items, and limitations. Stop after `PRODUCT.md` and the context changes owned by invoked skills are verified. Do not create feature specifications, solution designs, ADR decisions, work items, scaffolding, implementation, releases, or deployments. +Report changed files, supporting evidence, verification, non-blocking unresolved items, and limitations. Stop after `PRODUCT.md` and the context changes owned by invoked skills are verified. Do not create feature specifications, ADR decisions, work items, scaffolding, implementation, releases, or deployments. diff --git a/skills/define-product/assets/product-template.md b/skills/define-product/assets/product-template.md index 8f72b23..6ab3829 100644 --- a/skills/define-product/assets/product-template.md +++ b/skills/define-product/assets/product-template.md @@ -1,6 +1,6 @@ # {Product name} -> **Authority:** This document owns confirmed whole-product intent and system-wide requirements. Feature specifications and solution designs refine it without silently changing it. +> **Authority:** This document owns confirmed whole-product intent and system-wide requirements. Feature specifications refine it without silently changing it. diff --git a/skills/define-product/references/DISCOVERY.md b/skills/define-product/references/DISCOVERY.md index 095ce6a..cdfe354 100644 --- a/skills/define-product/references/DISCOVERY.md +++ b/skills/define-product/references/DISCOVERY.md @@ -16,4 +16,4 @@ Use meaningful alternatives, manual work, and doing nothing to clarify value and ## Expose system-wide requirements -For each journey and capability, ask what must remain true across likely implementations: operating conditions, measurable qualities, external ownership and exchanges, information duties, trust and harm boundaries, mandated technology, provider or organisational constraints, support and recovery, and eventual export, archival, or retirement. Record only material design inputs. Leave the selected internal structure and mechanisms to later solution design. +For each journey and capability, ask what must remain true across likely implementations: operating conditions, measurable qualities, external ownership and exchanges, information duties, trust and harm boundaries, mandated technology, provider or organisational constraints, support and recovery, and eventual export, archival, or retirement. Record only material design inputs. Leave the selected internal structure and mechanisms to later feature specification. diff --git a/skills/design-feature/SKILL.md b/skills/design-feature/SKILL.md deleted file mode 100644 index 49651eb..0000000 --- a/skills/design-feature/SKILL.md +++ /dev/null @@ -1,97 +0,0 @@ ---- -name: design-feature -description: Designs an approved feature specification into a confirmed, buildable, system-specific technical solution. Use when feature behaviour is settled and needs solution design before ticket creation. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Design Feature - -Designs one approved feature specification into one buildable technical solution without redefining the feature or implementing it. A force- and risk-led process uses selective **arc42 viewpoints** to expose the feature's context, boundaries, runtime, deployment, cross-cutting concerns, decisions, and risks; **bidirectional requirements-to-design traceability** and **scenario walkthroughs** keep the result complete and necessary. - -## Prerequisites - -Require the approved feature `specification.md`, root `PRODUCT.md`, and relevant root `CONTEXT.md` and ADRs. Inspect them before continuing. When an authority is missing, inadequate, or contradictory in a way that prevents a complete solution, report the exact prerequisite or gap and stop. Leave missing feature intent to a later `$specify-feature` session and missing system-wide requirements to a later `$define-product` session; do not invoke either or manufacture its outcome. - -Refine an existing paired solution design when requested, preserving unrelated confirmed content and local conventions. - -## Process - -### 1. Freeze authorities and evidence - -Inspect the request, prerequisite authorities, existing paired solution design, repository guidance, and the project's feature-record convention. Inspect only the relevant code, tests, contracts, schemas, configuration, runtime and deployment evidence, similar implemented patterns, and material external technical or regulatory evidence needed to design this feature. Treat the specification as authority for detailed feature intent, `PRODUCT.md` as authority for its product foundation and system-wide requirements, applicable ADRs as authority for their accepted decisions, `CONTEXT.md` as authority for project language, and current implementation as evidence of current constraints and idioms rather than automatic intended design. The user remains authority for material system-specific trade-offs, feature scope, and any proposed departure from those authorities. - -Separate confirmed future behaviour, system-wide requirements and accepted decisions, observed current state, and unresolved claims. Inspect obvious facts before asking questions. When a material product behaviour, rule, or acceptance boundary is missing or contradictory, identify the exact specification gap and stop rather than silently deciding it. When a material system-wide requirement is absent or the feature would require changing the product foundation, identify the exact foundation gap and stop rather than embedding it in this design. - -Invoke `$elicit-with-context` only when material user-held information or a system-specific trade-off remains unresolved. Let it ask one question at a time and, once invoked, use its complete-synthesis confirmation for this design. Otherwise obtain the same explicit whole-design confirmation directly in step 6. Invoke `$research` when a material external technical or regulatory claim could change the design and its evidence must persist; link its report instead of copying it. Each utility returns evidence or decisions to this session and owns neither the solution document nor continuation. - -Invoke `$maintain-ubiquitous-language` inline whenever consequential project terminology is independently confirmed, changed, misused, or contradicted. It alone owns the corresponding `CONTEXT.md` update; this skill retains the solution-design outcome and verification. Apply a glossary change before whole-design confirmation only when its meaning is confirmed independently of the proposed solution. Keep terminology that depends on an unconfirmed design choice unchanged until that choice is confirmed. - -The solution-design authority, runtime evidence, unresolved gaps, and permitted decision routes are explicit before solution selection begins. - -### 2. Map behaviour to design forces - -Map every material behaviour, rule, state, scenario, external constraint, and acceptance condition in the specification to affected system responsibilities and a credible verification seam. Identify the feature's entry points, trust boundaries, owned knowledge, data and integration flows, current-versus-target differences, and material quality, migration, operational, compatibility, and evolution forces. Trace each proposed new component, contract, data shape, migration, or operational obligation back to a specification behaviour, system-wide requirement, or applicable ADR; remove anything with no such origin. - -Invoke `$modular-design` with the confirmed behaviour, system-wide requirements, applicable ADRs, project language, and current evidence. Use its structural findings as passive constraints on ownership, information hiding, cohesion, contracts, dependency direction, seams, and change locality while retaining responsibility for alternatives, user decisions, the complete solution, verification, and stopping. - -The affected architecture, design forces, risks, and two-way traceability are complete enough to compare coherent solutions without selecting implementation detail prematurely. - -### 3. Compare and select the solution - -Compare viable system-specific alternatives for every material choice. Prefer the smallest coherent complete solution that follows established project and framework idioms, keeps knowledge with its natural owner, offers deep stable contracts, localises plausible change, and avoids speculative abstraction. Consider the consequences for the relevant forces, failure modes, migration, operations, verification, and likely evolution. Do not let current code win by default when it conflicts with the confirmed target, and do not introduce a new seam or component without evidenced variation, ownership, boundary, migration, or observation value. - -Resolve material trade-offs with the user through `$elicit-with-context` when invoked. Select one internally coherent solution and record each material selected option and its consequences. Retain rejected alternatives only when they are needed to understand a consequential trade-off, risk, or qualifying ADR; avoid alternative essays. - -The selected solution is justified by runtime constraints and confirmed decisions rather than familiarity or template completeness. - -### 4. Complete the technical design - -Develop only the applicable design depth needed for safe ticket decomposition: - -- map acceptance and behaviour to affected responsibilities, boundaries, ownership, information hiding, dependency direction, and enabling refactors; -- define internal and external interfaces, contracts, events, protocols, data shapes, invariants, ordering, errors, and compatibility at the precision consumers need; -- resolve data ownership, lifecycle, consistency, concurrency, transactions, privacy, retention, migration, backfill, rollback, and recovery; -- resolve integration protocols, authentication, authorization, idempotency, retries, timeouts, partial failure, reconciliation, and ownership of failure; -- resolve material security and threat controls, performance and capacity, resilience, observability, operability, support, and technical accessibility or compatibility concerns; -- define runtime and deployment effects plus rollout or transition constraints only where safe implementation depends on them; and -- identify the project-required and risk-triggered verification seams, evidence, environments, fixtures, or infrastructure needed to establish the designed behaviour. - -Omit immaterial concerns. Keep externally contractual detail already owned by the specification exact; otherwise select internal detail only to the level that constrains implementation meaningfully. Remain specific enough to decompose later without turning the design into code, a file or class inventory, framework wiring, executable configuration, or an exhaustive task list. - -The solution is coherent across structure, data, integrations, cross-cutting concerns, transition, operations, and verification. - -### 5. Walk implementation scenarios - -Walk the solution end to end through the main behaviour, material adverse and recovery paths, likely evolution, and migration and rollback where applicable. Follow each scenario across responsibilities, contracts, data, integrations, trust boundaries, failure ownership, observability, and verification seams. Challenge concurrency, partial execution, retries, incompatible versions, failed migration or backfill, restoration, and operational diagnosis when the runtime evidence makes them material. - -Use these walks to expose missing ownership, contradictions, duplicated rules, unstable dependencies, untraceable design elements, acceptance conditions without verification seams, or a solution that cannot be divided safely. Identify implementation-sized seams, genuine dependencies, enabling refactors, and sequencing constraints only far enough to prove the design is decomposable; do not create tickets, ticket IDs, estimates, priorities, or a delivery plan. Reopen only the affected design choice and repeat the applicable walk after any change. - -The selected solution survives its material operating, change, failure, transition, and verification scenarios and has credible decomposition seams. - -### 6. Confirm the complete design - -Present one concise, self-contained synthesis covering authorities, drivers, selected solution, affected responsibilities and dependency direction, interfaces and contracts, data, integrations, material cross-cutting concerns, migration and compatibility, operations, verification, implementation dependencies, risks, trade-offs, and both traceability directions. Include every proposed terminology update and ADR rationale. - -Identify an ADR candidate only when `$maintain-decision-records`' four gates appear to pass and the user agrees that the durable rationale is warranted. Most feature choices belong only in the solution design. Obtain a new explicit response confirming the whole synthesis; selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the solution design and ADRs unchanged, leave terminology proposed only within the rejected synthesis unapplied, reopen every affected evidence-backed choice, repeat the applicable traces and scenarios, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid after rejection. - -The confirmed synthesis supplies authority for the agreed writes without another application prompt. - -### 7. Write the solution design and context - -Follow the project's established feature-record convention. Otherwise write beside the specification at `docs/features//solution-design.md` using the [Solution Design Template](assets/solution-design-template.md). Preserve unrelated confirmed content and local structure when refining an existing design. Include only material sections and omit empty boilerplate. Link authorities and durable evidence rather than duplicating them. - -Record the selected how and its consequences at design precision. Use a small example only when indispensable to make a contract unambiguous. Do not rewrite the specification's what or why, duplicate live implementation state, prescribe file-by-file or method-by-method work, or include tickets, code scaffolding, estimates, priorities, releases, or deployment execution. - -After writing the solution design, invoke `$maintain-decision-records` for every confirmed qualifying ADR and `$maintain-ubiquitous-language` for confirmed terminology not already applied inline. The prior whole-design confirmation supplies authority for design-dependent terminology and explicit durable-rationale agreement only for ADRs included in that synthesis. Each utility owns only its context artifact; this skill retains the solution-design outcome and verification. - -The approved solution and only its warranted durable context are recorded in their owning artifacts. - -### 8. Verify and stop - -Verify every material specification behaviour and acceptance condition has a design responsibility and verification seam, and every introduced responsibility, component, contract, data shape, migration, integration, operational obligation, and verification mechanism traces back to a specification or architecture need. Rewalk the main, adverse, evolution, migration, and rollback scenarios that apply. Check responsibility ownership, contract compatibility, dependency direction, data lifecycle, integration failure handling, cross-cutting constraints, operational diagnosability, safe transition, and implementation decomposability for internal coherence. - -Verify statements against their authorities, links and paired path, preservation of unrelated content, and every context-maintenance result. Confirm the design has not changed product behaviour, smuggled in a system-wide architecture decision, duplicated live state, created tickets, or become code or file-level implementation instructions. - -Report changed files, authorities and supporting evidence, selected solution and material trade-offs, verification performed, and unresolved limitations. Stop after the confirmed solution design and applicable context records are written and verified; do not create tickets, estimate, prioritise, scaffold, implement, commit, publish, release, or deploy. diff --git a/skills/design-feature/agents/openai.yaml b/skills/design-feature/agents/openai.yaml deleted file mode 100644 index 783b7b2..0000000 --- a/skills/design-feature/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Design Feature' - short_description: 'Design a buildable feature solution' -policy: - allow_implicit_invocation: false diff --git a/skills/design-feature/assets/solution-design-template.md b/skills/design-feature/assets/solution-design-template.md deleted file mode 100644 index eae37ed..0000000 --- a/skills/design-feature/assets/solution-design-template.md +++ /dev/null @@ -1,69 +0,0 @@ -# {Feature name} solution design - - - -## Summary and authorities - - - -## Design drivers - - - -## Selected solution - - - -## Affected responsibilities and boundaries - - - -## Interfaces and contracts - - - -## Data - - - -## Integrations - - - -## Cross-cutting concerns - - - -## Migration and compatibility - - - -## Operations - - - -## Verification strategy and seams - - - -## Implementation and dependency notes - - - -## Risks and trade-offs - - - -## Traceability - -### Specification to design - -| Specification behaviour or acceptance condition | Design responsibility or contract | Verification seam or evidence | -| --- | --- | --- | -| {Use a stable specification label or concise exact reference.} | {Owning design element.} | {Observable verification boundary.} | - -### Design to authority - -| Introduced design element or obligation | Originating specification or architecture need | -| --- | --- | -| {Responsibility, component, contract, data shape, migration, integration, operational change, or verification mechanism.} | {Linked or exact authority reference.} | diff --git a/skills/refine-backlog/SKILL.md b/skills/refine-backlog/SKILL.md index 77fa5a8..75ddd56 100644 --- a/skills/refine-backlog/SKILL.md +++ b/skills/refine-backlog/SKILL.md @@ -40,9 +40,9 @@ Assess each item against the evidence applicable to its lifecycle position: - how its priority compares with the other bounded work under the user's current product direction; and - whether it is honestly ready for its next step using project-native evidence and fields, assessed separately from availability unless the project's native meaning explicitly includes blockers. -Do not derive priority from age, reporter urgency, effort, tracker order, or agent preference. Priority is a relative product decision owned by the user or another explicit project authority. Do not invent missing feature intent or technical design, estimates, acceptance evidence, or readiness claims. Identify the exact source gap, keep the item honestly not ready using a supported native representation, and route resolution to a later `$specify-feature`, `$design-feature`, or other owning session without invoking it. Preserve the distinction between an item that is sufficiently refined but blocked and one whose definition is incomplete; do not lower readiness merely to duplicate a separately represented blocker unless that is the project's confirmed native meaning. +Do not derive priority from age, reporter urgency, effort, tracker order, or agent preference. Priority is a relative product decision owned by the user or another explicit project authority. Do not invent missing feature intent or selected solution, estimates, acceptance evidence, or readiness claims. Identify the exact source gap, keep the item honestly not ready using a supported native representation, and route resolution to a later `$specify-feature` or other owning session without invoking it. Preserve the distinction between an item that is sufficiently refined but blocked and one whose definition is incomplete; do not lower readiness merely to duplicate a separately represented blocker unless that is the project's confirmed native meaning. -Classify matching work as an exact duplicate, partial overlap, related work, or unrelated work from discriminating evidence. Split an existing overbroad implementation item only into coherent outcomes already authorised by its sources. When a feature-level placeholder links approved specification and solution work but has no implementation tickets, do not create children or otherwise perform its initial decomposition: leave that outcome to a later `$create-tickets` session. Renaming the proposed children does not make that decomposition refinement. +Classify matching work as an exact duplicate, partial overlap, related work, or unrelated work from discriminating evidence. Split an existing overbroad implementation item only into coherent outcomes already authorised by its sources. When a feature-level placeholder links an approved specification but has no implementation tickets, do not create children or otherwise perform its initial decomposition: leave that outcome to a later `$create-tickets` session. Renaming the proposed children does not make that decomposition refinement. Every item has an evidence-led health assessment, relative-priority basis or open decision, and honest next-step readiness. @@ -76,4 +76,4 @@ The refined backlog and any partial completion are reproducibly verified in the Report each selected item with its native link, applied or no-op disposition, evidence basis, resulting priority and readiness, duplicates or overlap, dependencies and blockers, later owning route, read-back evidence, limitations, and unresolved decisions. Distinguish completed, blocked, drifted, and unchanged work. -Stop after backlog health and relative ordering are current and verified. Do not select or assign a sprint, cycle, milestone, iteration, release, or delivery body; establish implementation sequence beyond genuine blockers; invoke feature specification, solution design, or ticket creation; implement, debug, commit, publish, release, or deploy. Live backlog state remains in the configured tracker rather than a repository mirror. +Stop after backlog health and relative ordering are current and verified. Do not select or assign a sprint, cycle, milestone, iteration, release, or delivery body; establish implementation sequence beyond genuine blockers; invoke feature specification or ticket creation; implement, debug, commit, publish, release, or deploy. Live backlog state remains in the configured tracker rather than a repository mirror. diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md index b1fbd22..978cce8 100644 --- a/skills/review-incident/SKILL.md +++ b/skills/review-incident/SKILL.md @@ -40,7 +40,7 @@ Invoke `$research` only when one material external subject needs durable evidenc Identify only follow-up outcomes that directly address an evidenced contributing condition, recurrence path, defense gap, response weakness, or material learning unknown. For each candidate, state the incident evidence and rationale, the desired outcome rather than an unsupported implementation prescription, one accountable owner, the appropriate later lifecycle route, and observable evidence that will show completion and, where meaningful, effectiveness. Distinguish reducing recurrence likelihood, reducing impact, improving detection or response, and resolving an unknown. -Route a behaviour or product change to a later `$specify-feature` session; a technical solution or decomposition to later `$design-feature` or `$create-tickets` sessions once their authorities exist; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal design path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. +Route a behaviour or product change and its selected solution to a later `$specify-feature` session; decomposition to `$create-tickets` once an approved specification exists; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal specification path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. Challenge candidates against recurrence value, impact reduction, evidence strength, ownership, observability, cost, and overlap with current work. Reuse or link an exact existing outcome, preserve related work as related, merge duplicates, and remove generic reminders, `be more careful` actions, speculative backlog dumping, and action inflation. Zero follow-ups is valid when the evidence supports no useful new outcome. A critical learning action without an owner remains an explicit completion blocker rather than becoming anonymous backlog work. diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index 45d8b25..1b6157f 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -1,6 +1,6 @@ --- name: specify-feature -description: Defines one substantial feature or enhancement as approved, observable product intent. Use when a product capability or agreed feature idea needs a durable specification before solution design. +description: Creates a decision-complete specification for one feature. Use when a high-level feature request or capability needs collaborative definition before ticket creation. metadata: invocation: user disable-model-invocation: true @@ -8,52 +8,36 @@ disable-model-invocation: true # Specify Feature -Defines one substantial feature or enhancement without selecting its implementation. **Goal-oriented use cases** connect actors and beneficiaries to complete observable journeys; selective **Specification by Example** turns ambiguous rules and acceptance boundaries into discriminating examples rather than an implementation test inventory. +Turns one high-level feature description into one user-confirmed specification containing both feature intent and the selected buildable solution. ## Process -### 1. Inspect the feature evidence +### 1. Build shared understanding -Inspect the request, root `PRODUCT.md` and `CONTEXT.md`, repository guidance, the existing feature specification and paired solution design when present, and the project's feature-record convention. Inspect only the relevant current public contracts, documentation, tests, UI, research, and implementation evidence needed to understand observable behaviour. Treat `PRODUCT.md` as authority for the product foundation, the user as authority for the intended change, and the current system as evidence of present behaviour rather than intended future behaviour. Distinguish current, confirmed future, and unresolved claims; surface contradictions and unknowns without silently resolving them. +Inspect the request and the project evidence relevant to the feature. Resolve discoverable facts before asking the user, and distinguish evidence of the current system from intended future behaviour. -Accept one capability from `PRODUCT.md`, a confirmed synthesis returned by a prior `$elicit-with-context` session, or another agreed feature idea whose product context is sufficient. Do not manufacture a missing product foundation when the requested specification depends on one; report the exact prerequisite and stop. A direct agreed idea may proceed without a formal catalogue when its actors, value, and product boundary are sufficiently grounded. The source, authority, current evidence, convention, and unresolved feature decisions are explicit. +Invoke `$elicit-with-context` whenever material user-held information or a decision is needed. Bring it the evidence, gaps, viable options and consequences. The user decides the feature's direction; make informed recommendations, challenge ambiguity and expose overlooked paths without turning a suggestion into a decision. -### 2. Bound one proportionate feature +Invoke `$research` when durable external evidence is needed to resolve the feature. Use its report as evidence, not as authority for a product or solution decision. -Invoke `$elicit-with-context` throughout steps 2–5 for material user-held knowledge and decisions. Reuse every compatible branch from a prior confirmed synthesis and ask only about gaps exposed by inspection. Bound one coherent feature or enhancement, its problem or opportunity, beneficiaries, desired outcome, relationship to the product capability, and exclusions that separate adjacent behaviour. When inspection shows that the change is small, already understood, and needs no durable intent record, explain why a specification would add ceremony, name `$implement` as the proportionate next session, and stop without writing or invoking it. Invocation alone does not bypass this gate; proceed only when the user has separately established that a durable feature specification is required. The session owns exactly one warranted feature and a clear reason for specifying it. +### 2. Make the feature decision-complete -Invoke `$maintain-ubiquitous-language` inline whenever consequential project language is confirmed, changed, misused, or contradicted. It alone owns the root `CONTEXT.md` glossary update; this skill retains the specification outcome and verification. +Establish **feature intent** before selecting the **solution**, and visibly reopen the affected intent when solution work exposes a behavioural gap. Use the [Feature Specification Template](assets/specification-template.md) as a coverage guide rather than a checklist: depth follows ambiguity, risk, external obligation and what ticket decomposition needs. -Invoke `$research` when a material external, domain, regulatory, or standards subject could change the specification and needs durable evidence. Link its report instead of copying its findings. Research returns evidence to this session and owns neither the specification nor continuation. +Use goal-oriented use cases to connect actors to complete outcomes. Add triggers, preconditions, alternatives, failures and recovery where they reveal material behaviour. Use concrete context-event-outcome scenarios only where they make a rule, boundary or acceptance outcome testable through the application's interface or another stable observable boundary. Do not add user stories when they contribute no further meaning. -Keep the specification unchanged until the complete synthesis is explicitly confirmed. Supporting utilities may perform only the mutations authorised by their own contracts before then. +Keep feature intent independent of internal implementation. Put consequential system-specific choices in the selected solution: enough responsibility, boundary, contract, data, integration, failure, quality, migration, operational and verification detail to constrain implementation, but not files, classes, framework wiring, exhaustive tests or ticket breakdown. Invoke `$modular-design` when a material structural choice needs resolving. -### 3. Establish observable behaviour +Resolve every decision that could change scope, observable behaviour, acceptance or the material solution. An uncertainty may remain only when ticket decomposition is still safe; record its evidence, consequence, affected decisions and the condition or later user-agent step that will resolve it. Do not invent team roles, owners, approval bodies or third-party coordination. -Establish only the applicable feature-level detail: actors and their goals; scope and boundaries; externally observable behaviour and business or domain rules; material states, transitions, and lifecycle; main, adverse, recovery, and edge scenarios; and external product, regulatory, or compatibility constraints. Walk each actor from entry through goal, consequential interactions with other actors or systems, failure or refusal, recovery, and completion where applicable. Distinguish current behaviour from intended change wherever conflating them could misstate the feature. +### 3. Write the specification -Use concrete examples when an abstract rule, boundary, state transition, or scenario remains open to materially different interpretations. Include an example only when it discriminates valid from invalid behaviour or exposes a missing branch; do not expand the specification into exhaustive cases. Resolve contradictions and material ambiguity as they emerge. The feature's complete observable behaviour is understandable without assuming a technical solution. +Once shared understanding covers the complete feature, create `docs/features//specification.md` from the template. Record only decisions established with the user and evidence-supported facts, keeping feature intent and selected solution visibly distinct. Preserve the stable section spine, omit immaterial conditional content and link supporting product, context, decision, research, design and current-system evidence rather than duplicating it. -### 4. Establish acceptance and traceability +Use stable semantic identifiers only where they make later reference and traceability clearer. Add a concise Mermaid diagram when a state model, interaction, boundary or runtime relationship is materially easier to understand visually. -State acceptance conditions as unambiguous outcomes and evidence observable by an actor, consumer, or published contract. Cover the material behaviour, rules, state transitions, adverse paths, exclusions, and external constraints established for this feature. Use exact protocol or schema detail only when it is itself a published external contract; otherwise leave endpoints, components, services, tables, internal events, data models, algorithms, frameworks, libraries, deployment, and verification implementation to solution design. +### 4. Verify and stop -Apply the rebuild test to every proposed statement: it normally belongs in the specification only when it should remain true if the feature were rebuilt using a different architecture. Retain a system-specific statement only when it is an externally binding product, regulatory, or compatibility boundary. Link applicable product capability, research, context terms, current evidence, and stable material dependencies or assumptions without duplicating their authorities. The acceptance boundary is observable, testable, traceable, and independent of implementation. +Re-read the finished document against the established decisions and cited evidence. Check that it describes one coherent feature; states each rule and decision once; covers the material use cases, adverse paths, qualities and acceptance boundaries; keeps acceptance observable; distinguishes intent from solution; traces consequential solution elements to their reason; and represents uncertainty honestly. -### 5. Challenge the complete feature - -Walk the feature across every material actor goal, interaction, state, rule, main path, adverse path, recovery, edge condition, acceptance outcome, and exclusion. Test counterexamples at the feature boundary and look for hidden adjacent features, conflicting rules, impossible transitions, undefined ownership, unsupported current claims, and acceptance conditions that inspect private internals or prescribe a solution. Reopen only evidence-backed branches that could materially change the specification. Stop elaborating when another case would not change externally observable intent. The proposed feature is coherent and complete at feature depth without becoming solution design or an implementation backlog. - -### 6. Confirm one full synthesis - -Have `$elicit-with-context` present one concise, self-contained synthesis of the problem or opportunity, desired outcome, actors and goals, scope and exclusions, observable behaviour and rules, states and transitions, material scenarios, acceptance conditions, external constraints, traceability, and unresolved limitations. Obtain explicit confirmation of the whole synthesis. Selecting an earlier option is not confirmation. If the user rejects or corrects it, leave the specification unchanged, reopen every affected evidence-backed branch, repeat the applicable walks and counterexamples, and confirm the complete revised synthesis. The confirmed synthesis supplies authority for the agreed specification write without another application prompt. - -### 7. Write the specification - -Follow the project's established feature-record convention. Otherwise choose a stable lowercase feature slug and create `docs/features//specification.md` using the [Feature Specification Template](assets/specification-template.md). When refining an existing specification, preserve unrelated confirmed content and its local structure when differences are stylistic. Include only material sections and omit empty boilerplate. Keep canonical meanings in `CONTEXT.md`, external evidence in research reports, and implementation choices in a distinct `solution-design.md` when one exists. The durable specification records exactly the confirmed observable intent. - -### 8. Verify and stop - -Verify every statement against its authority: product foundation against `PRODUCT.md`; intended change against the confirmed synthesis; terminology against `CONTEXT.md`; current behaviour against inspected evidence; and material external claims against linked research. Rewalk traceability, actors, rules, states, scenarios, acceptance conditions, exclusions, and constraints for coherence and missing branches. Apply the rebuild test and confirm that acceptance remains observable without CSS classes, DOM or private internals, architecture, endpoints unless externally contractual, detailed data models, algorithms, frameworks, deployment, ticket breakdown, estimates, priorities, roadmaps, or releases. Verify the path, links, preservation of unrelated content, and separation from any solution design. - -Report changed files, supporting evidence, verification, and unresolved limitations. Stop after the approved specification and applicable ubiquitous language are written and verified; do not select a solution, create tickets, scaffold, implement, publish, prioritise, or continue into another lifecycle outcome. +The specification is ready when `$create-tickets` can decompose it without inventing product behaviour or selecting a material solution. Report the path, evidence used, verification performed and any non-blocking uncertainty. Do not create tickets or implement the feature. diff --git a/skills/specify-feature/agents/openai.yaml b/skills/specify-feature/agents/openai.yaml index 3303890..27fdbe8 100644 --- a/skills/specify-feature/agents/openai.yaml +++ b/skills/specify-feature/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Specify Feature' - short_description: 'Define one feature as observable product intent' + short_description: 'Create one decision-complete feature specification' policy: allow_implicit_invocation: false diff --git a/skills/specify-feature/assets/specification-template.md b/skills/specify-feature/assets/specification-template.md index efd7ef8..5c65cb4 100644 --- a/skills/specify-feature/assets/specification-template.md +++ b/skills/specify-feature/assets/specification-template.md @@ -1,53 +1,128 @@ # {Feature name} - +> **Authority:** This document owns the approved intent and selected solution for one feature. It is ready for ticket decomposition only when its readiness statement is satisfied. -## Summary + - +## Feature overview -## Actors and goals +{Orient the reader with the feature, its source request or capability, current context, intended beneficiaries, and concise outcome.} -- **{Actor or beneficiary}:** {Goal or outcome sought from this feature.} +## Problem, outcomes, and success measures -## Scope and boundaries +{State the evidenced problem or opportunity, desired user or product outcomes, and only applicable measures with their source, baseline, direction or threshold, and horizon when known.} -### Included +## Scope and non-goals -- {Observable behaviour within this feature.} +### In scope -### Excluded +- {Included feature outcome or observable responsibility.} -- {Adjacent behaviour deliberately outside this feature.} +### Non-goals -## Observable behaviour and rules +- {Excluded adjacent behaviour and the boundary it preserves.} - +## Actors, goals, and use cases -## States and transitions +### {ACT-01 — Actor or beneficiary} - +{Goal and value sought from this feature.} -## Scenarios +### {UC-01 — Use-case name} -### {Main, adverse, recovery, or edge scenario} +- **Actor:** {Primary actor.} +- **Trigger:** {Observable event that starts the use case.} +- **Preconditions:** {Material conditions, if any.} +- **Successful outcome:** {Externally observable completion.} +- **Alternatives and exceptions:** {Only material refusal, adverse, recovery, or edge paths.} - +## Feature intent -## Acceptance conditions +### Requirements and rules -- {An unambiguous, testable outcome observable by an actor, consumer, or published external contract.} +- **REQ-01 — {Requirement name}:** {One necessary, unambiguous, verifiable normative statement.} -## External constraints +### States and transitions - + -## Dependencies and assumptions +### Experience and design requirements - + -## Evidence and traceability +### Quality and external constraints -- **Product foundation:** {Link to the applicable PRODUCT.md capability or state the agreed direct feature source.} -- **Current behaviour:** {Link to relevant public contracts, docs, tests, UI evidence, or other inspected evidence when material.} -- **Research:** {Link material durable external evidence when applicable.} + + +### Scenarios and acceptance + +#### SC-01 — {Scenario name} + +- **Given:** {Observable starting context.} +- **When:** {Actor action or external event.} +- **Then:** {User-, assistive-technology-, contract-, or operating-interface outcome.} + +**Covers:** {REQ-01 and any distinct acceptance identifiers evidenced by this scenario.} + +#### Acceptance conditions + + + +- **ACC-01:** {Observable acceptance boundary stated once.} + +## Selected solution + +### Design drivers and current context + +{Link the requirements, constraints, accepted decisions, and inspected current-system forces that shape the solution. Distinguish current evidence from selected direction.} + +### Strategy and material decisions + +{State the selected coherent approach and only the alternatives or trade-offs needed to understand a consequential choice.} + +### Responsibilities and boundaries + +{Describe affected capabilities, owned knowledge, contracts, dependency direction, and enabling changes without a file or class inventory. Use a small Mermaid context or runtime diagram when it clarifies the design.} + +### Interfaces and data contracts + + + +### Data and lifecycle + + + +### Integrations and failure handling + + + +### Cross-cutting and operational design + + + +### Verification seams + +{Identify the public or stable contract boundaries and evidence capable of establishing the selected behaviour. Do not prescribe an exhaustive test inventory.} + +## Dependencies, assumptions, risks, and unresolved items + + + +- **DEP-01 — {Dependency}:** {Evidence, consequence, and affected identifiers.} +- **ASM-01 — {Assumption}:** {Evidence, consequence if wrong, and affected identifiers.} +- **RSK-01 — {Risk}:** {Evidence, consequence, mitigation or design response, and affected identifiers.} +- **UNR-01 — {Non-blocking uncertainty}:** {Evidence, consequence if wrong, affected decisions, and resolution condition or later user-agent step; explain why ticket decomposition remains safe.} + +## Traceability and readiness + +| Intent or acceptance identifier | Selected responsibility or contract | Verification seam or evidence | +| --- | --- | --- | +| {REQ-01 or ACC-01} | {Design responsibility or contract.} | {Observable verification boundary.} | + +| Introduced design element | Originating intent, constraint, evidence, or accepted decision | +| --- | --- | +| {Material responsibility, contract, data, integration, migration, operational, or verification element.} | {Exact identifier or linked authority.} | + +### Readiness statement + +{State concisely why later ticket creation can proceed without inventing observable behaviour or selecting a material solution, plus any non-blocking limitation. Do not recap the specification.} diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md index 398bf97..c3fb2b5 100644 --- a/skills/triage-work/SKILL.md +++ b/skills/triage-work/SKILL.md @@ -41,7 +41,7 @@ Classify each item by the lifecycle outcome it actually needs: - route active production impact requiring stabilisation to a later `$respond-to-incident` session; - route removal or deprecation intent to a later `$plan-retirement` session; - link an exact existing representation and use the supported duplicate or equivalent native resolution; -- link actionable work whose intent and solution are already defined to its existing appropriate work without creating another specification, design, or ticket; or +- link actionable work whose intent and solution are already defined to its existing appropriate work without creating another specification or ticket; or - close answered, invalid, not-actionable, or out-of-scope requests only with a concise reason grounded in inspected evidence. When evidence cannot yet support one of those outcomes, retain an honest native needs-information or unresolved disposition when available and identify the minimum missing fact and who can supply it. Invoke `$elicit-with-context` only when that fact or a materially different classification or disposition depends on user-held knowledge or authority. Supply the evidence and viable consequences, use only its confirmed synthesis, and retain ownership of proposal, mutation, verification, and stopping. Do not ask the user for repository or tracker facts that inspection can establish. @@ -66,4 +66,4 @@ Read back every affected record and relationship. Verify source preservation, cl Report every selected item with its native link, applied or no-op disposition, evidence basis, duplicate or related work, next lifecycle route or missing input, read-back evidence, limitations, and unresolved uncertainty. Distinguish completed, blocked, and unchanged items so a blocker on one does not overstate or erase safe batch progress. -Stop after the native dispositions and routes are verified. Do not invoke the routed lifecycle skill or perform feature elicitation or specification, solution design, ticket decomposition, backlog scope or readiness refinement, priority or iteration planning, root-cause debugging, repair, incident stabilisation, retirement planning, implementation, commit, publication, release, or deployment. Live work state remains in the configured tracker rather than a repository mirror. +Stop after the native dispositions and routes are verified. Do not invoke the routed lifecycle skill or perform feature elicitation or specification, ticket decomposition, backlog scope or readiness refinement, priority or iteration planning, root-cause debugging, repair, incident stabilisation, retirement planning, implementation, commit, publication, release, or deployment. Live work state remains in the configured tracker rather than a repository mirror. From 8b44da37ea1d003db162b06fbaf6532fe98ac655 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Thu, 20 Aug 2026 10:20:12 +0100 Subject: [PATCH 56/79] chore(skills): remove tracker workflow skills --- README.md | 14 +---- SKILLS_PLAN.md | 49 ++------------- skills/plan-iteration/SKILL.md | 65 ------------------- skills/plan-iteration/agents/openai.yaml | 6 -- skills/refine-backlog/SKILL.md | 79 ------------------------ skills/refine-backlog/agents/openai.yaml | 6 -- skills/review-incident/SKILL.md | 2 +- skills/triage-work/SKILL.md | 69 --------------------- skills/triage-work/agents/openai.yaml | 6 -- 9 files changed, 9 insertions(+), 287 deletions(-) delete mode 100644 skills/plan-iteration/SKILL.md delete mode 100644 skills/plan-iteration/agents/openai.yaml delete mode 100644 skills/refine-backlog/SKILL.md delete mode 100644 skills/refine-backlog/agents/openai.yaml delete mode 100644 skills/triage-work/SKILL.md delete mode 100644 skills/triage-work/agents/openai.yaml diff --git a/README.md b/README.md index 7a29ee8..4d35d40 100644 --- a/README.md +++ b/README.md @@ -2,7 +2,7 @@ # Propulsion -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to planning work, implementation, refinement, debugging, incident response, and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, debugging, incident response, and retirement. Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. @@ -38,7 +38,6 @@ Repeat the link for each selected skill. Codex and OpenCode both discover skills | Establish or revise a product | `$define-product` | Confirmed product and system requirements | | Define a substantial feature | `$specify-feature` | An approved, decision-complete feature specification | | Make a small understood change | `$implement` | A minimal, reviewed, verified local change | -| Route incoming requests | `$triage-work` | Verified dispositions and lifecycle routes | | Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | | Stabilise a production incident | `$respond-to-incident` | Verified mitigation or recovery and current service state | | Review another contributor's PR | `$review-pull-request` | Independent findings and verification evidence | @@ -62,9 +61,6 @@ The lifecycle areas are useful entry points, not mandatory phase gates. ### Refine -- `$triage-work` closes or routes a bounded set of incoming requests without performing the routed work. -- `$refine-backlog` keeps selected backlog work current, prioritised, and honest about readiness, dependencies, and blockers. -- `$plan-iteration` selects feasible ready work in the project's native sprint, cycle, milestone, or next-work mechanism. - `$review-architecture` reports a few high-value improvements for an explicitly bounded area without changing it or creating tickets. ### Deliver @@ -110,12 +106,6 @@ Deliver a small understood change: implement → commit → pull-request ``` -Refine team work: - -```text -triage-work → refine-backlog → optionally plan-iteration -``` - Debug, respond, or retire: ```text @@ -155,7 +145,7 @@ Software-project skills route questions through `$elicit-with-context`. Base `$e ## Task-management tools -Tracker-backed skills use the project tool named in the root `AGENTS.md` and represent work using that tool's native issues, relationships, statuses, priorities, sprints, cycles, milestones, or queues. If no tool is named, the skill asks which one the project uses, records the preference through `$maintain-agents`, and resumes. It never guesses or introduces a Propulsion configuration layer. +Tracker-backed skills use the project tool named in the root `AGENTS.md` and represent work using that tool's native items, relationships, and statuses. If no tool is named, the skill asks which one the project uses, records the preference through `$maintain-agents`, and resumes. It never guesses or introduces a Propulsion configuration layer. ## Quality during implementation diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9ccd9c0..4e6031d 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -8,7 +8,7 @@ The existing suite may be changed or replaced where necessary. Skill authoring r ## Objective -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, backlog refinement, implementation, pull-request publication and review, debugging, incident response and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, debugging, incident response and retirement. Propulsion is not a workflow platform. It does not carry a product automatically through its lifecycle, maintain its own work state or require a setup framework. A user may enter at any point where the necessary inputs already exist and leave when that session's outcome is complete. @@ -32,7 +32,7 @@ Propulsion is not a workflow platform. It does not carry a product automatically | Establish | Create the durable product and system requirements foundation for a new or existing system. | | Define | Turn one high-level feature request into an approved statement of intended behaviour and its selected buildable solution. | | Plan | Decompose an approved specification into implementation-ready work. | -| Refine | Keep incoming and existing work current, prioritised, ready and feasible for the next delivery period. | +| Refine | Assess a bounded area of the current architecture and identify high-value improvements. | | Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | | Debug and Respond | Diagnose and repair defects, stabilise active incidents and learn from them afterwards. | | Retire | Plan the safe removal of an obsolete capability and route its execution through normal delivery. | @@ -89,7 +89,7 @@ Research reports, architecture reviews, incident reviews and retirement plans ar The project's task-management tool must be named in `AGENTS.md`. A tracker-backed skill that cannot find this preference asks the user which tool the project uses, invokes `maintain-agents` to record it and then resumes its original outcome. -The skill uses the selected tool's native issues, relationships, statuses, priorities, sprints, cycles, milestones or queues. If the named tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. +The skill uses the selected tool's native items, relationships and statuses. If the named tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. ## Lifecycle skill catalogue @@ -121,34 +121,10 @@ The skill uses the selected tool's native issues, relationships, statuses, prior - **Inputs:** Either one approved, decision-complete feature specification or an approved retirement plan, plus relevant project guidance and current tracker state. - **Output:** Native tracker items linked to their applicable authoritative documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. - **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. -- **Stops:** After the created items and relationships are read back and verified. It does not select an iteration or begin implementation. +- **Stops:** After the created items and relationships are read back and verified. It does not schedule the work or begin implementation. ### Refine -#### `triage-work` - -- **Outcome:** Each incoming request in scope is closed with an evidence-backed reason or routed to the appropriate lifecycle area. -- **Inputs:** The selected incoming items, their source evidence, current product and implementation authorities and the configured task-management tool. -- **Output:** Verified native dispositions, classifications, links and next routes. -- **Composition:** Conditionally invokes `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records` and `maintain-agents`. -- **Stops:** After disposition. It does not perform feature specification, debugging, implementation, retirement planning or incident response. - -#### `refine-backlog` - -- **Outcome:** The selected backlog is current, prioritised and honest about scope, readiness, dependencies and blockers. -- **Inputs:** A bounded backlog view, its linked authorities, current project direction and native tracker relationships. -- **Output:** Verified updates, splits, closures, dependencies, priorities and readiness represented in the task-management tool. -- **Composition:** Conditionally invokes `elicit-with-context` for unresolved readiness, scope or priority decisions and `maintain-agents` when the task-management preference is missing. -- **Stops:** After backlog health and ordering are current. It does not commit the project to a sprint, cycle, milestone or implementation sequence. - -#### `plan-iteration` - -- **Outcome:** A feasible body of ready, unblocked work is selected for the project's next delivery period. -- **Inputs:** The refined backlog, current priorities, dependencies, available capacity and the project's native planning mechanism. -- **Output:** A verified sprint, cycle, milestone or ordered next-work queue in the task-management tool. -- **Composition:** Conditionally invokes `elicit-with-context` for selection trade-offs and `maintain-agents` when the task-management preference is missing. -- **Stops:** After the selected body of work is recorded. It does not implement any item. - #### `review-architecture` - **Outcome:** A bounded, evidence-backed assessment identifies a small number of high-value architecture or maintainability improvements. @@ -318,16 +294,6 @@ elicit-with-context when needed → pull-request ``` -### Refine team work - -```text -triage-work - → refine-backlog - → plan-iteration when the project uses a delivery period -``` - -The skills may be invoked independently. Triage may route a valid defect directly to `debug` or a new capability to `specify-feature`. - ### Debug a known defect ```text @@ -366,7 +332,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. 3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. -5. Build recurring team workflows: `triage-work`, `refine-backlog`, `plan-iteration` and `review-architecture`. +5. Build the architecture-assessment path: `review-architecture`. 6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. 7. Align software-project elicitation through `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. @@ -389,9 +355,6 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` - [-] Create `review-pull-request` skill -- [-] Create `triage-work` skill -- [-] Create `refine-backlog` skill -- [-] Create `plan-iteration` skill - [-] Update `review-architecture` skill - [-] Update `debug` skill - [-] Create `respond-to-incident` skill @@ -428,7 +391,7 @@ Propulsion will not: - provide a lifecycle navigator or end-to-end delivery orchestrator - require a Propulsion manifest, setup skill, state model, completion horizon or tracker adapter -- impose Scrum, sprints or another project-management method +- impose a project-management method - create separate skills for individual test types, metrics or quality commands - require discovery, architecture or feature documents for small work that is already sufficiently understood - duplicate code, CI, tracker or production state in prose documents diff --git a/skills/plan-iteration/SKILL.md b/skills/plan-iteration/SKILL.md deleted file mode 100644 index 72cdd9e..0000000 --- a/skills/plan-iteration/SKILL.md +++ /dev/null @@ -1,65 +0,0 @@ ---- -name: plan-iteration -description: Selects and records a feasible body of ready, unblocked work in the project's native planning mechanism. Use when planning the next delivery period or ordered next-work queue from a refined backlog. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Plan Iteration - -Selects a coherent body of work that current evidence makes feasible for the project's next delivery period, then records and verifies that selection in the configured task-management tool. The result is a forecast of selected work, not a promise of completion or a release plan. - -## Prerequisites - -Require an explicitly bounded backlog whose candidate items have already been refined. An eligible item has observable bounded scope, a current priority, enough approved intent and technical direction for its next implementation step, and no unresolved blocker. Exclude an item that fails any condition, preserve the exact reason, and route it to a later `$refine-backlog` session; do not refine, estimate, split, or rescope it here. - -Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or introduce a Propulsion manifest, adapter, configuration, or state model. When the configured tool, account, bounded backlog, or required read or write operation is unavailable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. - -## Process - -### 1. Establish the native planning surface - -Inspect the request, repository guidance, bounded backlog, configured tool, and current planning state before proposing any mutation. Discover the tool's native sprint, cycle, milestone, or ordered-queue concepts; period identity and dates; membership and ordering; dependency direction; priority, readiness, sizing, capacity, and assignment fields; existing selections; and read-back operations. Use only concepts with the meanings established by the project and tool. - -Determine the next planning horizon from project evidence or user authority. A formal-period project uses its established or explicitly confirmed native period. A project without a formal period uses only its existing native ordered next-work mechanism; tool support alone does not establish that the project uses sprints, cycles, or milestones. When neither a native period nor an ordered next-work mechanism can represent the outcome, report the capability gap and agree only a representation the configured tool actually supports. Do not create a repository mirror. - -The planning horizon, native representation, existing membership, and capability limits are explicit. - -### 2. Establish honest capacity and eligibility - -Read current project-native evidence for the horizon: comparable historical throughput or sizing where it exists, current commitments, team availability, specialist constraints, dependency timing, operational load, and other work that consumes the same capacity. Never invent velocity, estimates, dates, availability, or precision, and do not treat points or item counts as interchangeable capacity unless the project's own evidence supports that use. - -When evidence cannot resolve a material capacity, horizon, or trade-off decision, invoke `$elicit-with-context` with the evidence and viable consequences. Let it ask one user-authoritative question at a time and use only its confirmed synthesis; retain ownership of selection, mutation, verification, and stopping. State the weakest honest capacity bound when precise capacity is unavailable rather than manufacturing a number. - -Refresh each candidate's observable scope, current priority, readiness for its next implementation step, dependencies, blockers, and current native state. Treat a relationship as blocking only when the dependency meaning and evidence show the item cannot begin or complete without it. Do not confuse preference, display order, or convenient sequence with a blocker. The eligible set, exclusions with reasons, and capacity assumptions and limitations are current and explicit. - -### 3. Form one feasible selection - -Consider eligible items in current priority order. Select a coherent body subject to the capacity evidence, whole-item fit, genuine dependencies, specialist availability, operational commitments, and material delivery risk. Include a prerequisite only when it is ready, necessary, and feasible in the same horizon; exclude work blocked by an external or out-of-horizon prerequisite. - -Do not fill capacity automatically, add lower-priority work merely to maximise utilisation, or promise that selected work will finish. Preserve a high-priority item that does not fit as an exclusion rather than splitting, rescoping, or lowering its readiness standard. Keep intentionally unused capacity visible when uncertainty, operational load, item shape, or a constrained capability makes that the honest plan. - -The proposed body is priority-led, dependency-valid, coherent, and feasible within the stated evidence without changing the backlog. - -### 4. Confirm the complete native plan - -Present one complete proposal containing the planning horizon, selected items, native membership and order, exclusions and reasons, included prerequisites, capacity evidence and assumptions, existing assignments or commitments being preserved, exact intended mutations, capability limitations, and residual feasibility risk. Keep the tracker unchanged while resolving material decisions. - -Obtain explicit user confirmation of the complete proposal. If the user rejects or revises it, reopen every affected capacity, eligibility, dependency, coherence, ordering, and representation decision and present the complete revised proposal. Confirmation supplies authority for exactly the agreed native mutations without another application question. - -The user and agent share one exact, feasible mutation plan. - -### 5. Refresh and apply the agreed selection - -Immediately before writing, re-read the target period or queue, every selected or displaced membership and order, relevant capacity state, priorities, readiness, blockers, and dependencies. When drift materially changes eligibility, feasibility, membership, order, or an agreed mutation, keep the affected state unchanged and return to complete-proposal confirmation. Preserve an independent agreed mutation only when the drift cannot affect its feasibility or meaning. - -Apply exactly the confirmed native assignments, membership, and order. Preserve unrelated period, queue, backlog, field, relationship, and assignment state. Use the tool's native operations and meanings; when an agreed capability is unsupported, leave that part unapplied and report the limitation rather than simulating it with a misleading field, relationship, or prose record. - -The configured tracker contains exactly the safely applicable agreed selection. - -### 6. Read back, verify, and stop - -Read back the target period or queue, selected membership and order, affected relationships and assignments, and every changed field. Compare them with the confirmed proposal and verify that selected items remain ready and unblocked, prerequisites and dependency directions are valid, capacity assumptions still hold within their stated precision, and unrelated native state was preserved. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop and report the mismatch or unsupported capability. - -Report the verified native period or queue, selected and excluded items with reasons, applied and no-op mutations, capacity evidence and limitations, read-back evidence, unsupported capabilities, drift, and residual feasibility risk. Stop after the selected body is recorded and verified. Do not refine or estimate backlog work, change specifications or designs, create tickets, implement, commit, publish, release, or deploy. diff --git a/skills/plan-iteration/agents/openai.yaml b/skills/plan-iteration/agents/openai.yaml deleted file mode 100644 index d7b5000..0000000 --- a/skills/plan-iteration/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Plan Iteration' - short_description: 'Select a feasible body of ready work' - default_prompt: 'Use $plan-iteration to select and record a feasible body of ready work for the next delivery period.' -policy: - allow_implicit_invocation: false diff --git a/skills/refine-backlog/SKILL.md b/skills/refine-backlog/SKILL.md deleted file mode 100644 index 75ddd56..0000000 --- a/skills/refine-backlog/SKILL.md +++ /dev/null @@ -1,79 +0,0 @@ ---- -name: refine-backlog -description: Refines a bounded software backlog into verified current, prioritised, and honestly ready native tracker work. Use for recurring backlog health and ordering sessions before iteration planning. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Refine Backlog - -Keeps a selected backlog aligned, coherent, and usable for later delivery decisions without committing work to an iteration or manufacturing missing product or technical definition. - -## Prerequisites - -Require an explicitly bounded native backlog view or item set. Resolve only the selection when scope is unclear; do not inspect or mutate work outside it. - -Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, treat their answer as authority to record that confirmed preference, invoke `$maintain-agents`, verify the root instruction, and resume without a separate permission question. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool, selected backlog, linked authority, or required read or write operation is unavailable, stop the affected work with the exact access, connection, or evidence needed rather than falling back to Markdown or another tracker. - -## Process - -### 1. Establish the refinement surface - -Inspect the request, repository guidance, configured tool, bounded backlog, and current project direction before mutation. Inspect the tool's native item types, fields, statuses, priorities or ordering, parent and split concepts, duplicate and related relationships, blocker direction, readiness signals, and read-back capabilities. Use only meanings supported by the current project and tool; do not invent a universal workflow or Definition of Ready. - -Read every selected item, its discussion and evidence, linked feature, product, solution, architecture, or other owning authorities, and its native relationships. Search relevant open and closed work for duplicates and overlap. Inspect only enough current product, documentation, code, tests, or operational evidence to judge whether the item remains current. Treat stale tracker text and reporter wording as claims rather than authority. - -The bounded items, available native operations, current authorities, and access limits are explicit. - -### 2. Assess every selected item - -Assess each item against the evidence applicable to its lifecycle position: - -- whether the work is still needed and aligned with current project direction; -- whether its scope expresses one coherent observable outcome; -- whether linked intent and technical design are sufficient for the next step it claims to be ready for; -- whether its acceptance evidence makes completion observable; -- whether its size and internal coherence support one bounded later session; -- whether it duplicates or overlaps existing work; -- whether dependencies and blockers are genuine, acyclic, and represented in the correct direction; -- how its priority compares with the other bounded work under the user's current product direction; and -- whether it is honestly ready for its next step using project-native evidence and fields, assessed separately from availability unless the project's native meaning explicitly includes blockers. - -Do not derive priority from age, reporter urgency, effort, tracker order, or agent preference. Priority is a relative product decision owned by the user or another explicit project authority. Do not invent missing feature intent or selected solution, estimates, acceptance evidence, or readiness claims. Identify the exact source gap, keep the item honestly not ready using a supported native representation, and route resolution to a later `$specify-feature` or other owning session without invoking it. Preserve the distinction between an item that is sufficiently refined but blocked and one whose definition is incomplete; do not lower readiness merely to duplicate a separately represented blocker unless that is the project's confirmed native meaning. - -Classify matching work as an exact duplicate, partial overlap, related work, or unrelated work from discriminating evidence. Split an existing overbroad implementation item only into coherent outcomes already authorised by its sources. When a feature-level placeholder links an approved specification but has no implementation tickets, do not create children or otherwise perform its initial decomposition: leave that outcome to a later `$create-tickets` session. Renaming the proposed children does not make that decomposition refinement. - -Every item has an evidence-led health assessment, relative-priority basis or open decision, and honest next-step readiness. - -### 3. Form the complete proposal - -Prepare the smallest native change set that makes the bounded backlog current. It may clarify an item, split supported scope, reconcile or close obsolete, duplicate, invalid, or fully superseded work with evidence, correct relationships and blocker directions, set an agreed native priority or order, and represent honest readiness. Preserve a no-op when current state already expresses the supported result. - -Invoke `$elicit-with-context` only when a material scope, priority, or readiness decision depends on user-held knowledge or authority. Supply inspected evidence, viable options, and consequences; let it ask one question at a time and use only its confirmed synthesis. Retain responsibility for the complete proposal, mutation, verification, and stopping. - -Recheck duplicates and affected relationships, then present one complete proposal containing every create, update, split, reconciliation, closure, link, priority, readiness, and no-op action; exact relationship meanings and directions; evidence-led reasons; later routes; and native capability limitations. Keep the tracker unchanged while decisions remain open. Obtain explicit user confirmation of the whole set. Selecting one earlier option is not confirmation. If the user rejects or revises any part, reopen every affected assessment and relationship, then present the complete revised set. Confirmation authorises exactly the agreed native mutations without another application prompt. - -The user and agent share one exact, duplicate-aware refinement plan. - -### 4. Refresh and apply the agreed changes - -Immediately before writing, re-read every affected item, matching candidate, and relationship. When drift materially changes an agreed action, leave that item and its coupled mutations unchanged and return its complete revised proposal for confirmation. Continue only independent batch items whose agreed changes remain safe. - -Apply only the confirmed native mutations and preserve unrelated content, discussion, fields, and metadata. Create only relationships required to express an evidenced split, reconciliation, duplicate, overlap, or genuine blocker; native parent or split structure does not also justify an incidental related link between siblings. Use a different supported native operation only when it preserves the exact confirmed meaning; otherwise leave that part unapplied and report the capability limit. Do not turn a convenient ordering preference into a blocker or extend the mutation to adjacent backlog work. - -The configured tracker contains every still-valid agreed refinement or an explicit item-level blocker. - -### 5. Read back and verify the backlog - -Read every affected record and relationship back from the tool. Compare scope, observable outcome, authority links, acceptance evidence, state, priority or order, readiness, parent or split structure, duplicate disposition, and blocker type, target, and direction with the confirmed proposal. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop the affected item and report the mismatch. - -Repeat the relevant duplicate and overlap search. Verify that each selected item is current, supported at its claimed lifecycle position, coherent, relatively prioritised where authority permits, honest about readiness, and connected only by genuine blockers. Verify closure reasons and preservation of unrelated native state. A blocked item does not erase verified independent progress. - -The refined backlog and any partial completion are reproducibly verified in the owning tracker. - -### 6. Report and stop before planning - -Report each selected item with its native link, applied or no-op disposition, evidence basis, resulting priority and readiness, duplicates or overlap, dependencies and blockers, later owning route, read-back evidence, limitations, and unresolved decisions. Distinguish completed, blocked, drifted, and unchanged work. - -Stop after backlog health and relative ordering are current and verified. Do not select or assign a sprint, cycle, milestone, iteration, release, or delivery body; establish implementation sequence beyond genuine blockers; invoke feature specification or ticket creation; implement, debug, commit, publish, release, or deploy. Live backlog state remains in the configured tracker rather than a repository mirror. diff --git a/skills/refine-backlog/agents/openai.yaml b/skills/refine-backlog/agents/openai.yaml deleted file mode 100644 index 91c58ca..0000000 --- a/skills/refine-backlog/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Refine Backlog' - short_description: 'Keep backlog work current and honestly ready' - default_prompt: 'Use $refine-backlog to refine this bounded backlog into verified current, prioritised work.' -policy: - allow_implicit_invocation: false diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md index 978cce8..97cb0cf 100644 --- a/skills/review-incident/SKILL.md +++ b/skills/review-incident/SKILL.md @@ -40,7 +40,7 @@ Invoke `$research` only when one material external subject needs durable evidenc Identify only follow-up outcomes that directly address an evidenced contributing condition, recurrence path, defense gap, response weakness, or material learning unknown. For each candidate, state the incident evidence and rationale, the desired outcome rather than an unsupported implementation prescription, one accountable owner, the appropriate later lifecycle route, and observable evidence that will show completion and, where meaningful, effectiveness. Distinguish reducing recurrence likelihood, reducing impact, improving detection or response, and resolving an unknown. -Route a behaviour or product change and its selected solution to a later `$specify-feature` session; decomposition to `$create-tickets` once an approved specification exists; an evidenced defect to `$debug`; backlog health or priority to `$refine-backlog`; an architecture hotspot to `$review-architecture` or the normal specification path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. +Route a behaviour or product change and its selected solution to a later `$specify-feature` session; decomposition to `$create-tickets` once an approved specification exists; an evidenced defect to `$debug`; an architecture hotspot to `$review-architecture` or the normal specification path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. Challenge candidates against recurrence value, impact reduction, evidence strength, ownership, observability, cost, and overlap with current work. Reuse or link an exact existing outcome, preserve related work as related, merge duplicates, and remove generic reminders, `be more careful` actions, speculative backlog dumping, and action inflation. Zero follow-ups is valid when the evidence supports no useful new outcome. A critical learning action without an owner remains an explicit completion blocker rather than becoming anonymous backlog work. diff --git a/skills/triage-work/SKILL.md b/skills/triage-work/SKILL.md deleted file mode 100644 index c3fb2b5..0000000 --- a/skills/triage-work/SKILL.md +++ /dev/null @@ -1,69 +0,0 @@ ---- -name: triage-work -description: Triages bounded incoming software requests into verified native tracker dispositions and lifecycle routes. Use for recurring intake review before backlog refinement. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Triage Work - -Determines what each selected incoming request is, records an evidence-led native disposition, and routes it to the right later lifecycle session without performing that work. - -## Prerequisites - -Require an explicitly bounded set of incoming items or queue. When the scope is unclear, resolve only the selection before inspecting or mutating items outside it. - -Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool, account, selected queue, source evidence, or required write operation is unavailable, stop the affected item with the exact access or connection needed rather than falling back to Markdown or another tracker. - -## Process - -### 1. Establish the intake surface - -Inspect the request, repository guidance, configured tool, selected items, and their current native state before mutation. Inspect the tool's available item types, statuses, resolutions, labels or classifications, comments, links, duplicate and related relationships, transfer or routing operations, and read-back capabilities. Use only concepts and meanings the current project and tool actually support. - -For each selected item, read its description, attachments, discussion, source report, prior dispositions, and linked work. Treat reporter wording as a claim rather than the classification. When the source is not yet represented in the tracker, include the smallest native intake record needed to preserve its evidence and route in the later proposal; do not turn it into an implementation ticket. The bounded items, available evidence, native capabilities, and access limits are explicit. - -### 2. Establish enough evidence to classify - -Inspect the current `PRODUCT.md`, `CONTEXT.md`, feature records, ADRs, documentation, public contracts, tests, code, operational evidence, and external authorities only as far as each item needs. Distinguish intended behaviour, current observable behaviour, reported observation, and unsupported inference. Seek enough corroboration to decide a disposition, not a root cause or solution. A safe focused observation may establish that a defect report is credible; full reproduction, causal investigation, mitigation, repair, or feature discovery belongs to its later routed session. - -Search open and closed tracker work and relevant repository records using source identifiers, authoritative links, actors, outcomes, symptoms, affected concepts, and discriminating details. Classify a candidate internally as exact match, possible overlap, related work, or unrelated. Mark a duplicate only when the evidence establishes the same requested outcome or fault; preserve an uncertain resemblance as related or unresolved and state what would distinguish it. - -When a material external subject could change the disposition and needs durable evidence, invoke `$research` with that bounded subject, then read its report before using its findings. Invoke `$maintain-ubiquitous-language` only when the evidence independently confirms consequential durable project language and `$maintain-decision-records` only when it confirms qualifying rationale; a proposed classification or route alone does not establish either. The evidence for classification, duplicate confidence, and remaining uncertainty is explicit. - -### 3. Choose one honest disposition per item - -Classify each item by the lifecycle outcome it actually needs: - -- route a new or changed externally observable capability to a later `$specify-feature` session; -- route incorrect existing behaviour to a later `$debug` session; -- route active production impact requiring stabilisation to a later `$respond-to-incident` session; -- route removal or deprecation intent to a later `$plan-retirement` session; -- link an exact existing representation and use the supported duplicate or equivalent native resolution; -- link actionable work whose intent and solution are already defined to its existing appropriate work without creating another specification or ticket; or -- close answered, invalid, not-actionable, or out-of-scope requests only with a concise reason grounded in inspected evidence. - -When evidence cannot yet support one of those outcomes, retain an honest native needs-information or unresolved disposition when available and identify the minimum missing fact and who can supply it. Invoke `$elicit-with-context` only when that fact or a materially different classification or disposition depends on user-held knowledge or authority. Supply the evidence and viable consequences, use only its confirmed synthesis, and retain ownership of proposal, mutation, verification, and stopping. Do not ask the user for repository or tracker facts that inspection can establish. - -For each item, prepare a proposed native disposition containing its classification, concise evidence-led reason, exact duplicate or related links, supported state and field changes, precise next route and required input, and any native capability limitation. Do not add priority, readiness, estimates, iteration placement, ticket decomposition, root cause, solution, or speculative downstream work. Every selected item has one supported disposition or one explicit evidence blocker. - -### 4. Confirm the mutation plan - -Recheck each proposal against the source evidence, current authorities, possible duplicates, and the native meanings discovered in step 1. Present the complete material mutation plan: records to create, close, update, transfer, comment on, classify, or link; exact relationship meanings and targets; reasons; routes; no-op items; and limitations. Keep the tracker unchanged while resolving material decisions. - -Obtain explicit user confirmation of the complete plan. Selecting an earlier classification option is not confirmation of all mutations. If the user rejects or revises a proposal, reopen every affected evidence, classification, duplicate, and routing decision and present the complete revised plan. Confirmation supplies authority for exactly the agreed mutations, including uncontroversial native fields mechanically entailed by them, without another application prompt. - -### 5. Apply and verify the agreed dispositions - -Immediately before writing, re-read every affected item, relationship, and duplicate candidate. If drift changes an agreed disposition or target, leave that item unchanged and return its revised complete proposal for confirmation; continue independent batch items only when their plans remain safe. - -Apply only the agreed native mutations and relationships. Preserve unrelated content, discussion, fields, and metadata. When the tool lacks an agreed classification, resolution, link, or route, use another supported native operation only when it preserves the exact confirmed meaning; otherwise leave that part unapplied and report the capability limit rather than simulating it with a misleading field or relationship. - -Read back every affected record and relationship. Verify source preservation, classification, reason, state, fields, duplicate or related target and direction, next route, and absence of unintended changes against the confirmed plan. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop that item and report the mismatch. Each applied disposition is verified in its owning system. - -### 6. Report and stop at intake - -Report every selected item with its native link, applied or no-op disposition, evidence basis, duplicate or related work, next lifecycle route or missing input, read-back evidence, limitations, and unresolved uncertainty. Distinguish completed, blocked, and unchanged items so a blocker on one does not overstate or erase safe batch progress. - -Stop after the native dispositions and routes are verified. Do not invoke the routed lifecycle skill or perform feature elicitation or specification, ticket decomposition, backlog scope or readiness refinement, priority or iteration planning, root-cause debugging, repair, incident stabilisation, retirement planning, implementation, commit, publication, release, or deployment. Live work state remains in the configured tracker rather than a repository mirror. diff --git a/skills/triage-work/agents/openai.yaml b/skills/triage-work/agents/openai.yaml deleted file mode 100644 index 710eafa..0000000 --- a/skills/triage-work/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Triage Work' - short_description: 'Triage incoming work into verified routes' - default_prompt: 'Use $triage-work to triage these incoming requests into verified tracker dispositions and lifecycle routes.' -policy: - allow_implicit_invocation: false From 11b49c8c8a4fa084c3612070a7bfb988d848ed0c Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Thu, 20 Aug 2026 20:25:08 +0100 Subject: [PATCH 57/79] feat(skills): finalize create-tickets --- README.md | 6 +- SKILLS_PLAN.md | 24 +++---- skills/create-tickets/SKILL.md | 71 +++++-------------- skills/create-tickets/agents/openai.yaml | 4 +- .../create-tickets/assets/ticket-template.md | 29 ++++++++ .../assets/specification-template.md | 2 +- 6 files changed, 66 insertions(+), 70 deletions(-) create mode 100644 skills/create-tickets/assets/ticket-template.md diff --git a/README.md b/README.md index 4d35d40..d33bdfb 100644 --- a/README.md +++ b/README.md @@ -57,7 +57,7 @@ The lifecycle areas are useful entry points, not mandatory phase gates. ### Plan -- `$create-tickets` turns an approved feature specification or retirement plan into verified, dependency-aware work in the project's task-management tool. +- `$create-tickets` turns an approved work definition—either a sufficiently complete document or confirmed conversation—into agreed, verified, vertically sliced tickets with Fibonacci complexity. ### Refine @@ -143,9 +143,9 @@ Software-project skills route questions through `$elicit-with-context`. Base `$e `$write-skill` develops and evaluates Agent Skills themselves; it is not part of the software-delivery lifecycle. -## Task-management tools +## Ticket destinations -Tracker-backed skills use the project tool named in the root `AGENTS.md` and represent work using that tool's native items, relationships, and statuses. If no tool is named, the skill asks which one the project uses, records the preference through `$maintain-agents`, and resumes. It never guesses or introduces a Propulsion configuration layer. +The root `AGENTS.md` explicitly names the project's ticket destination, such as Linear or local Markdown. If none is named, `$create-tickets` asks the user, records the answer through `$maintain-agents`, and resumes. Local tickets use one file each under `docs/features//tickets/`; external destinations use their native items, estimates, relationships, and statuses. The skill never guesses, silently falls back, or introduces a Propulsion adapter layer. ## Quality during implementation diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 4e6031d..800750c 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -18,7 +18,7 @@ Propulsion is not a workflow platform. It does not carry a product automatically 2. **Narrow progressively.** Product intent narrows into a decision-complete feature specification, implementation-ready tickets and, finally, one implemented body of work. 3. **Enter proportionately.** Substantial work uses the complete applicable chain. Small, understood work may begin directly with implementation. 4. **Compose within the requested outcome.** A skill may invoke another skill when its outcome is required to complete the requested session. It must not silently advance into a later lifecycle outcome. -5. **Keep authorities separate.** Product documents own intent, code and tests own implemented behaviour, Git owns change history and the task-management tool owns work state. +5. **Keep authorities separate.** Product documents own intent, code and tests own implemented behaviour, Git owns change history and the configured ticket system owns work state. 6. **Document what cannot be inferred safely.** Durable documents preserve intent, language, constraints and rationale without copying observable code, configuration, CI or tracker state. 7. **Use native project tools.** Propulsion does not introduce a manifest, tracker adapter, lifecycle state model or completion-horizon schema. 8. **Treat agreement as authority.** Invoking a skill authorises the mutations intrinsic to its agreed outcome. Questions resolve the outcome; they are not followed by a redundant application gate. @@ -41,7 +41,7 @@ These areas are entry points, not mandatory phase gates. Release, deployment and ## Session and authority model -Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture a missing or incomplete feature specification. +Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture an incomplete work definition. Reusable utilities are user-invokable and may be model-invoked when their documented trigger is present. A caller remains responsible for verifying the utility's result and completing its own outcome. @@ -85,11 +85,11 @@ Small, understood changes do not require feature documents merely to enter imple Research reports, architecture reviews, incident reviews and retirement plans are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. -### Task-management tool +### Ticket destination -The project's task-management tool must be named in `AGENTS.md`. A tracker-backed skill that cannot find this preference asks the user which tool the project uses, invokes `maintain-agents` to record it and then resumes its original outcome. +The project's ticket destination must be named in `AGENTS.md`, for example `The project uses Linear for tickets.` or `The project uses local Markdown for tickets.` A ticket-creating skill that cannot find this preference asks the user which destination the project uses, invokes `maintain-agents` to record it and then resumes its original outcome. -The skill uses the selected tool's native items, relationships and statuses. If the named tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. +Local Markdown uses one file per ticket under `docs/features//tickets/`. An external destination uses its native items, estimates, relationships and statuses. If the named external tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. ## Lifecycle skill catalogue @@ -117,11 +117,11 @@ The skill uses the selected tool's native items, relationships and statuses. If #### `create-tickets` -- **Outcome:** An approved feature specification or retirement plan is represented by implementation-ready, dependency-aware work in the project's task-management tool. -- **Inputs:** Either one approved, decision-complete feature specification or an approved retirement plan, plus relevant project guidance and current tracker state. -- **Output:** Native tracker items linked to their applicable authoritative documents, each describing one coherent vertical outcome, acceptance evidence and genuine blocking relationships. -- **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the task-management preference is missing. -- **Stops:** After the created items and relationships are read back and verified. It does not schedule the work or begin implementation. +- **Outcome:** One approved work definition is represented by confirmed, implementation-ready, dependency-aware tickets in the project's configured destination. +- **Inputs:** One sufficiently complete approved document or confirmed conversation, plus relevant project guidance and implementation evidence. +- **Output:** One local Markdown file or external native item per coherent vertical outcome, with bounded context, requirements, applicable constraints, observable acceptance, Fibonacci complexity and genuine blocking relationships. +- **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the ticket destination is missing. +- **Stops:** After the created tickets and relationships are read back and verified. It does not schedule the work or begin implementation. ### Refine @@ -350,7 +350,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `maintain-agents` skill - [x] Update `define-product` skill - [x] Create `specify-feature` skill -- [-] Create `create-tickets` skill +- [x] Create `create-tickets` skill - [-] Update `implement` skill - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` @@ -375,7 +375,7 @@ The revised suite is ready when: - substantial work narrows cleanly from product foundation through decision-complete specification, tickets and per-item implementation - small, clear work can enter implementation without ceremonial documents or tickets - required and conditional composition is explicit and does not duplicate supporting skill instructions -- task-management skills use the project tool named in `AGENTS.md` and its native concepts without Propulsion configuration +- ticket-creating skills use the local Markdown or external destination named in `AGENTS.md` and preserve its exact concepts without Propulsion configuration - documentation preserves enduring intent without copying observable implementation or live tool state - implementation always completes applicable verification and independent review - absent quality infrastructure is reported honestly and never causes unrelated tooling to be installed implicitly diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md index 3de8ced..764c929 100644 --- a/skills/create-tickets/SKILL.md +++ b/skills/create-tickets/SKILL.md @@ -1,80 +1,47 @@ --- name: create-tickets -description: Creates verified native tracker items from an approved feature specification or retirement plan. Use when feature delivery or capability retirement needs implementation-ready, dependency-aware tickets. -metadata: - invocation: user +description: Creates implementation-ready tickets from approved work. Use when a complete work definition needs vertical decomposition in the project's configured ticket system. disable-model-invocation: true --- # Create Tickets -Turns one approved feature specification or retirement plan into the smallest coherent set of implementation-ready work in the project's configured task-management tool. **Vertical slicing** keeps each item end to end, while selected **Example Mapping** and **Specification by Example** techniques turn approved rules, examples, stages, and completion conditions into observable acceptance evidence. +Turns one **approved work definition** into a confirmed, verified set of small vertical tickets. A work definition may be a document or confirmed conversation whose outcome, scope, behaviour or obligations, material constraints, and solution decisions are complete enough to decompose without invention. ## Prerequisites -Require one applicable approved authority package and inspect it before continuing: - -- For feature delivery, require one approved feature specification containing decision-complete feature intent and selected solution. When it is missing, inadequate, or contradictory in a way that prevents honest decomposition, identify the exact gap and stop. Route missing observable intent or technical resolution to a later `$specify-feature` session; do not invoke it or manufacture its outcome. -- For capability retirement, require one approved retirement plan that resolves the retirement outcome; affected parties, dependencies, interfaces, data, and integrations; applicable deprecation, migration, coexistence, disablement, removal, archival, communication, support, and recovery obligations; material risks and rollback or recovery routes; and observable stage and final completion conditions sufficiently for honest decomposition. When a needed retirement decision is missing or contradictory, identify the exact gap and stop for a later `$plan-retirement` session; do not force retirement through feature specification or manufacture the plan's outcome. - -A retirement plan may deliberately depend on a later feature specification for a replacement capability. Do not create implementation work for that unresolved replacement until the specification is approved. Retirement-only obligations may be decomposed when they are independently confirmed and do not require the unresolved replacement decision. - -Require the project's task-management tool to be named in the root `AGENTS.md`. When the preference is absent, ask the user directly which tool the project uses, invoke `$maintain-agents` with the confirmed preference, verify the root instruction, and resume. Never guess a tool or add a Propulsion manifest, adapter, configuration, or state model. When the configured tool or required account is unavailable or not writable, stop with the exact access or connection needed rather than falling back to Markdown or another tracker. +Read the root `AGENTS.md` for an explicit ticket destination such as `The project uses Linear for tickets.` or `The project uses local Markdown for tickets.` When absent, ask the user which destination the project uses, invoke `$maintain-agents` with the answer, verify the instruction, and resume. For an external destination, require its installed integration and writable account; otherwise stop with the exact access needed. Never guess or silently fall back. ## Process -### 1. Fix the authorities and native surface - -Inspect the request, repository guidance, applicable approved authority package, its linked product, context, ADR, and research authorities, and only the implementation evidence needed to understand existing work. For feature delivery, treat the specification's feature-intent part as authority for observable behaviour and its selected-solution part as authority for technical decisions and genuine prerequisites. For retirement, treat the retirement plan as authority for the approved outcome, obligations, stages, risks, recovery, and completion conditions. Treat a separately approved replacement feature specification as authority only for replacement work it actually resolves, and treat the task-management tool as authority for live work state. Discover obvious facts before asking the user. - -Inspect the configured tool's available schema and capabilities before planning any mutation: native item types, bodies, links, parent or grouping concepts, dependency direction, duplicate or related relationships, and read-back operations. Use only concepts the tool actually supports. Keep the tracker read-only through proposal confirmation. - -The applicable authority package, tracker scope, native capabilities, and access boundary are explicit. - -### 2. Reconcile existing work - -Read the relevant open and closed tracker state before decomposing. Search by authoritative document links or identifiers, feature or retirement outcome language, affected project concepts, and related or parent items. Classify candidates as exact matches, partial overlaps, related work, or unrelated work. Reuse an exact match, propose a bounded reconciliation for a partial overlap, preserve merely related work, and avoid creating a second representation of the same outcome. - -Record the classification and evidence internally. Do not copy live tracker state into repository documentation. The later proposal can distinguish create, update, relationship-only, and no-op reuse actions without inventing a parallel backlog. - -The decomposition starts from a current, duplicate-aware view of native work. - -### 3. Form vertical outcomes - -Build an internal two-way coverage map from the applicable authority package to proposed tracker outcomes. For feature delivery, cover every material specification behaviour, rule, state, scenario, constraint, and acceptance condition plus every necessary solution responsibility, transition, and verification seam. For retirement, cover every material affected party, dependency, interface, data and integration obligation, applicable stage and transition, communication and support obligation, risk control, abort, rollback or recovery route, residual obligation, and stage and final completion condition. Each outcome must trace back to an approved source, and every approved requirement or obligation necessary for the requested delivery or retirement scope must have an owner. - -Form the smallest set of coherent vertical outcomes. Each item spans every affected layer needed to produce one observable result or fulfilled obligation for an actor, consumer, published contract, migration, retirement stage, or operational boundary. Include a shared enabling change in the earliest outcome that needs it unless an approved authority establishes a separately observable and independently verifiable prerequisite. Do not split work by frontend, backend, database, service, class, file, architecture layer, test type, retirement phase label, or another implementation inventory merely because those parts exist. One item is correct when the approved package is already one bounded independently implementable outcome. - -For each item, define a concise native title, one coherent outcome, enough bounded authority context for a separate implementation session, links to the applicable approved authority package, observable acceptance evidence, and applicable relationships. A feature item links the authoritative feature specification. A retirement item links the retirement plan and any separately applicable approved replacement specification when that item includes replacement work. Derive acceptance evidence from approved rules and discriminating examples for feature delivery, or from approved stage outcomes, exit evidence, counter-signals, recovery obligations, and completion conditions for retirement. Use a concrete example only when it separates valid from invalid behaviour or exposes a consequential branch. State outcomes at stable observable boundaries without prescribing test implementation. - -Check each item directly for independent value, bounded size, and testability. Minimise coupling between items while representing unavoidable prerequisites honestly. Infer a blocker only when the approved solution or required state transition establishes that one outcome cannot begin or complete before another; do not convert user-journey order, convenient sequencing, or tracker display order into dependency. Interpret the configured tool's relationship direction explicitly and reject cycles. +### 1. Establish the work and destination -When a question exposes missing feature intent or selected solution, stop with the owning specification gap. When it exposes a missing retirement decision, stop for a later `$plan-retirement` session. When a retirement item would implement an unresolved replacement capability, omit that item until its approved feature specification exists; retain retirement-only items only when their outcomes and dependencies are independently confirmed. Invoke `$elicit-with-context` only when the approved authorities support more than one materially different decomposition and the user must choose among them. Supply the evidence and alternatives, let it ask one question at a time, and use its confirmed decision while retaining proposal, tracker mutation, verification, and stopping ownership. +Inspect the request, approved work definition, repository guidance, linked product, context, decision, research, and current-system evidence needed to understand the work. Treat the definition as authority for intended outcomes and settled constraints, and the repository as evidence of the current system. -The draft items are vertically coherent, source-complete, independently actionable where possible, and linked only by genuine acyclic prerequisites. +For an external destination, inspect its native item, body, status, estimate, parent, dependency, and read-back capabilities before planning writes. For local Markdown, use one file per ticket under `docs/features//tickets/`. -### 4. Challenge and propose the set +### 2. Agree the vertical ticket set -Trace every proposed item back to its approved need or obligation and every approved need or obligation forward to exactly the necessary items. Remove unapproved scope, duplicated outcomes, speculative abstractions, horizontal work, incidental dependencies, and detail that belongs only in code. Confirm that an implementation agent could complete each item in one bounded session without rediscovering observable intent, selecting a missing technical or retirement decision, or inventing unresolved replacement behaviour. +Map every material requirement, obligation, constraint, solution responsibility, transition, and acceptance boundary in both directions between the work definition and proposed tickets. Form the smallest coherent set in which each ticket delivers one observable outcome or fulfilled contract across every layer it needs. Absorb shared setup into the earliest outcome that needs it. Create separate enabling, migration, or refactor work only when it has an independently necessary, verifiable boundary or preserves a valid compatible intermediate state. -Repeat the live duplicate search using the final titles, outcomes, authority links, and concepts. Reconcile the result into one complete proposal containing every native create, update, relationship-only, and no-op reuse action; the exact item contents; relationship types and directions; and any native capability limitation or residual mismatch. Do not include estimates, priorities, sprint or cycle assignments, milestones, target dates, releases, implementation sequencing beyond genuine blockers, or speculative future work. +Size each ticket for one fresh implementation session. Assign one project-local **complexity** bucket from `1`, `2`, `3`, `5`, or `8` by comparing the affected behaviour, technical difficulty, integrations, data or migration work, applicable quality constraints, and bounded uncertainty with the other proposed tickets. The buckets are ordered comparisons, not ratios, hours, dates, or duration promises. Split any outcome that would exceed `8`, then estimate the resulting slices again. -Present the complete proposal for explicit user confirmation. Selecting an earlier decomposition option is not confirmation of the full set. If the user rejects or revises it, keep the tracker unchanged, reopen every affected coverage, slicing, relationship, and reconciliation decision, repeat the duplicate search, and present the complete revised proposal. The confirmed proposal supplies authority for exactly its agreed tracker mutations without another application question. +Add a blocker only when one ticket cannot honestly begin or meet acceptance before another reaches its accepted outcome. Keep display order, likely sequence, user-journey order, and merely related work out of the dependency graph. State dependency direction explicitly and reject cycles. -The user and agent share one exact, duplicate-aware mutation plan. +Present only a compact numbered proposal containing each ticket's stable title, vertical outcome, complexity, and `Blocked by` references. Invoke `$elicit-with-context` when supported alternatives require a user decomposition decision. Iterate the compact set until the user confirms the whole proposal; do not synthesize full bodies first. Confirmation authorises the agreed synthesis and writes without another application prompt. If synthesis exposes a material gap, leave the destination unchanged and reopen the affected proposal. -### 5. Apply the confirmed work +### 3. Synthesize and write the tickets -Immediately before writing, refresh every matching candidate and affected relationship. If concurrent change materially alters the confirmed create, update, no-op, or dependency plan, make no tracker mutation and return to complete-proposal confirmation. Otherwise use the configured tool's native operations to create, update, and link exactly the agreed items. Preserve unrelated content and fields on reused items. Use native idempotency, duplicate, parent, related, and blocking concepts where available; never silently substitute a different meaning when a capability is absent. +Use the [Ticket Template](assets/ticket-template.md) as the minimum semantic body, omitting only inapplicable conditional content. Embed the necessary confirmed intent and solution context when no durable authority exists; never tell a fresh implementation agent to recover an unavailable conversation. Explain what must be delivered at stable boundaries without prescribing files, classes, framework wiring, code, or an exhaustive test inventory. -Create in dependency-safe order only where the tool requires identifiers before relationships. Agreement authorises the confirmed writes; do not ask whether to apply them again and do not extend the mutation to convenient adjacent work. +For local Markdown, allocate stable sequential identifiers such as `TKT-001`, use `-.md`, set new tickets to `Todo`, and write reciprocal `Blocked by` and `Blocks` references. The local status vocabulary is `Todo`, `In Progress`, `Done`, and `Cancelled`. -The configured tracker contains exactly the agreed native representation or an explicitly surfaced tool limitation. +For an external destination, use the closest established initial native status and map complexity, blocker, blocked, and parent meanings to exact native fields when available. When an exact capability is absent, retain the meaning explicitly in the body and surface the limitation; never substitute a different relation. Do not set priority, assignee, sprint or cycle, milestone, release, or dates unless the approved definition or user supplies them. -### 6. Read back, verify, and stop +Create and relate exactly the confirmed set. -Read every created or updated item and every affected relationship back from the tool. Compare titles, outcomes, bounded context, applicable authority-package links, acceptance evidence, relationship types and directions, and create, update, or reuse disposition with the confirmed proposal. Correct an in-scope discrepancy when the agreed value and affected record are unambiguous; stop and report a mismatch when correction would require a new user decision, unsupported capability, or mutation of unrelated work. +### 4. Read back, verify, and stop -Repeat the relevant duplicate search and verify two-way authority-package coverage, vertical coherence, independent actionability, observable completion, genuine acyclic blockers, and preservation of unrelated native state. For retirement, verify that stage ordering became a blocker only where an approved entry condition or required state transition makes it genuine, that retirement-only work does not depend on an unresolved replacement without authority, and that every item links its applicable approved authorities. Report exact item links, created, updated, related, and reused records, verification performed, unsupported capabilities, concurrent-state or duplicate uncertainty, and unresolved source limitations. +Read every created ticket and relationship back from its destination. Verify the confirmed titles, outcomes, authority or embedded context, requirements, applicable constraints, acceptance criteria, complexity values and ceiling, status, relationship meanings and directions, acyclic dependencies, resolved local references, and two-way work-definition coverage. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise stop with the mismatch and required decision or capability. -Stop after the native items and relationships are read back and verified. Do not change product, feature, solution, retirement, architecture, or context documents; select or schedule an iteration; estimate or prioritise work; scaffold or implement code; commit or publish changes; release; or deploy. +Return the created ticket links or local paths, verification performed, capability limitations, source limitations, and unresolved external state. Stop without changing the approved work definition, scheduling or prioritising work, implementing, committing, publishing, releasing, or deploying. diff --git a/skills/create-tickets/agents/openai.yaml b/skills/create-tickets/agents/openai.yaml index 42f7453..4f7072c 100644 --- a/skills/create-tickets/agents/openai.yaml +++ b/skills/create-tickets/agents/openai.yaml @@ -1,6 +1,6 @@ interface: display_name: 'Create Tickets' - short_description: 'Create delivery or retirement tracker work' - default_prompt: 'Use $create-tickets to turn this approved feature specification or retirement plan into implementation-ready tracker items.' + short_description: 'Create implementation-ready vertical tickets' + default_prompt: 'Use $create-tickets to turn this approved work definition into confirmed, implementation-ready tickets.' policy: allow_implicit_invocation: false diff --git a/skills/create-tickets/assets/ticket-template.md b/skills/create-tickets/assets/ticket-template.md new file mode 100644 index 0000000..52a9413 --- /dev/null +++ b/skills/create-tickets/assets/ticket-template.md @@ -0,0 +1,29 @@ +# {TKT-001} — {Ticket title} + +- **Status:** {Todo or destination-native status} +- **Complexity:** {1, 2, 3, 5, or 8} +- **Blocked by:** {Ticket references or None} +- **Blocks:** {Ticket references or None} +- **Authority:** {Durable work-definition link and identifiers, or concise embedded authority context} + +## Outcome + +{One independently observable vertical result or fulfilled contract.} + +## Context + +{Only the bounded actors, current state, settled solution constraints, non-goals, and source details a fresh implementation session cannot infer safely.} + +## Functional requirements + +- {Approved behaviour or rule owned by this ticket.} + +## Non-functional requirements + + + +- {Constraint in its operating context with an observable response or measure where established.} + +## Acceptance criteria + +- [ ] {Observable pass/fail condition at a user, assistive-technology, contract, event, migration, persistence, provider, or operating boundary.} diff --git a/skills/specify-feature/assets/specification-template.md b/skills/specify-feature/assets/specification-template.md index 5c65cb4..56baf35 100644 --- a/skills/specify-feature/assets/specification-template.md +++ b/skills/specify-feature/assets/specification-template.md @@ -2,7 +2,7 @@ > **Authority:** This document owns the approved intent and selected solution for one feature. It is ready for ticket decomposition only when its readiness statement is satisfied. - + ## Feature overview From 0c5459f11910b7d3ef00d3b0faaeaae6fe77c9a7 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 08:23:19 +0100 Subject: [PATCH 58/79] refactor(skills): redefine implementation skill suite --- IMPLEMENTATION_SKILLS_PLAN.md | 439 +++++++++++++++++++ SKILLS_PLAN.md | 71 +-- skills/verify-change/SKILL.md | 67 --- skills/verify-change/agents/openai.yaml | 5 - skills/verify-change/references/HARNESSES.md | 43 -- 5 files changed, 480 insertions(+), 145 deletions(-) create mode 100644 IMPLEMENTATION_SKILLS_PLAN.md delete mode 100644 skills/verify-change/SKILL.md delete mode 100644 skills/verify-change/agents/openai.yaml delete mode 100644 skills/verify-change/references/HARNESSES.md diff --git a/IMPLEMENTATION_SKILLS_PLAN.md b/IMPLEMENTATION_SKILLS_PLAN.md new file mode 100644 index 0000000..75d5f83 --- /dev/null +++ b/IMPLEMENTATION_SKILLS_PLAN.md @@ -0,0 +1,439 @@ +# Implementation skills plan + +## Status and authority + +This is the authoritative authoring brief for Propulsion's implementation skill set. [`SKILLS_PLAN.md`](SKILLS_PLAN.md) owns the wider lifecycle catalogue and each skill's public outcome. This document owns the detailed contract, composition, teaching coverage, resource expectations and acceptance evidence for: + +- `implement` +- `tdd` +- `modular-design` +- `quality-harnesses` +- `code-review` + +The existing skill bundles are baseline evidence only. They may be rewritten and their resources replaced, split, combined or removed during their individual `$write-skill` sessions. Each authoring session must still establish matched baselines and prove that its final bundle is the smallest version that produces the behaviour fixed here. + +The evidence basis is [`docs/research/agentic-coding-quality-guardrails-2026-08-20.md`](docs/research/agentic-coding-quality-guardrails-2026-08-20.md). That report supports a relevant portfolio of independent evidence routes rather than a universal gauntlet, score or claim that passing tools proves correctness. + +## Intended outcome + +Together these skills let an agent implement agreed software with enough structural guidance, test discipline, quality evidence and independent challenge that a human can manage the work from its requirements and evidence rather than line-by-line code inspection. + +The system optimises for the **smallest coherent change**: the least production code, test code and supporting machinery that completely satisfies the agreed behaviour while remaining understandable and economical to change. Fewest lines or files is not the target. A small extraction may be necessary to localise knowledge; speculative abstractions, extension points, duplicate paths and unrelated cleanup are not. + +## Composition model + +`implement` owns the implementation session. `tdd`, `modular-design` and `quality-harnesses` are **teaching skills**: they add applicable knowledge to their caller without becoming delegated workers or owning separate workflows. `code-review` is different because context isolation is part of its outcome; it coordinates independent fresh-agent review and returns suggestions without changing the candidate. + +```text +implement + applies modular-design throughout + applies tdd when behaviour changes and a usable suite exists + applies relevant quality-harnesses guidance + builds and locally checks one complete candidate + invokes code-review + Spec fresh agent when behavioural authority is available + Standards fresh agent with the teaching skills + validates and adjudicates both sets of suggestions + applies supported changes and repeats from the affected step +``` + +Teaching skills state what their caller must understand and apply. They do not claim separate mutation authority, hand back a workflow verdict or require the caller to restate their internal instructions. Their direct composition must preserve: + +- the caller's fixed outcome and authority; +- the teaching skill's applicable trigger and knowledge; +- the caller's ownership of edits, commands, evidence and stopping; +- no duplicated internal procedure in the caller; and +- no hidden advancement into commit, publication, release or deployment. + +## Shared quality rules + +All five skill contracts preserve these rules: + +1. **Independent oracle before confidence.** A requirement, accepted example, protocol, invariant, trusted reference, prior accepted behaviour or explicit decision must be capable of disagreeing with the implementation. Implementation-derived expectations and regenerated outputs are not independent evidence. +2. **Behaviour before representation.** Tests and Spec review observe supported outcomes, contracts, state transitions and effects rather than private methods, internal call graphs, incidental DOM structure or storage layout unless that representation is itself promised. +3. **Relevant evidence rather than control count.** Apply each repository requirement and each catalogue harness whose trigger matches the changed risk. Do not run irrelevant tools to create confidence theatre. +4. **Metrics prompt investigation.** Coverage, mutation, complexity, size, churn, coupling and smell counts inform named questions. They are not universal thresholds, composite scores or substitutes for judgement. +5. **Measurement paths are part of the change.** Tests, fixtures, snapshots, baselines, suppressions, exclusions, quality configuration, generators, scripts and thresholds require the same scrutiny as production code. +6. **Unavailable evidence stays visible.** Missing tools, credentials, services, platforms, representative data or specialist judgement are limitations, neither passes nor evidence against the implementation. +7. **Project authority wins.** Repository instructions, accepted specifications, architecture decisions, framework conventions and configured checks override generic guidance where they conflict. +8. **No implicit infrastructure programme.** A caller may improve evidence through existing infrastructure. It does not add a project-wide framework or quality tool unless establishing that capability is explicitly part of the agreed work. + +## `implement` + +### Recurring job + +Turn one implementation-ready ticket or small confirmed body of work into a minimal local change whose applicable tests and quality checks pass and whose current candidate has completed independent Spec and Standards review. + +### Triggers and non-triggers + +Use for a ticket, bug repair handed off from diagnosis, or small agreed software change that is sufficiently defined to build. A ticket or feature document is not mandatory when the request itself is clear. + +Do not use it to discover product direction, conduct exploratory research, create specifications or tickets, commit, publish, release or deploy. It does not invoke `research`; implementation should already have its material external knowledge. If implementation exposes a missing behavioural, contract, architecture, scope or acceptance decision, invoke `elicit-with-context`. That skill owns any resulting ubiquitous-language or decision-record maintenance. + +### Required sequence + +1. **Fix the implementation basis.** Inspect the work authority, repository instructions, current code, tests, relevant contracts and complete repository state. State the observable outcome, exclusions, material risks and completion evidence. Resolve discoverable facts directly; use `elicit-with-context` only for a material user-held gap. +2. **Load the teaching knowledge.** Apply `modular-design` and `quality-harnesses`. Apply `tdd` when the work changes observable behaviour and a usable test suite exists. Keep the knowledge active throughout implementation rather than treating it as a final checklist. +3. **Select the smallest coherent slice.** Trace the next production addition to one current requirement, caller or necessary support responsibility. Use a tracer bullet for the smallest end-to-end observable result. Avoid speculative flexibility, unrelated cleanup and broad refactoring. +4. **Implement test-first when applicable.** Let the caller perform the Red–Green–Refactor cycle taught by `tdd`, carrying the modular constraints and relevant quality risks into test and production design. For work outside TDD's trigger, use the smallest project-native feedback path without claiming TDD. +5. **Maintain structural locality.** Keep each rule with its coherent owner, retain or establish the smallest justified interface, and perform conditional Design It Twice before a material new or changed boundary is fixed. +6. **Apply relevant harnesses.** Review the quality catalogue, load the detailed references for every matching entry, run or perform the applicable repository-native harnesses, and preserve their exact evidence and limitations. Do not silently replace an unavailable harness with a weaker check. +7. **Inspect the complete candidate.** Read the entire diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact and measurement-path change. Remove superseded paths, dead helpers, duplication and speculative code made unnecessary by the final design. +8. **Rationalise retained tests.** After production refactoring has stayed green with unchanged tests, review the suite for obsolete, duplicative, brittle or implementation-coupled cases. Merge, rewrite or delete a test only when every unique promised behaviour and oracle remains protected and the revised suite still detects the realistic defects it is meant to catch. +9. **Freeze and review.** Freeze one exact, complete, locally checked candidate and invoke `code-review`. Do not send an incomplete slice or changing target. +10. **Adjudicate objectively.** Validate each suggestion against its authority, the code and the strongest benign interpretation. Classify it as a required correction, proportionate improvement, rejected finding or user decision. Consequence rank orders attention but never decides the classification. +11. **Correct and repeat.** Apply required corrections and worthwhile in-scope improvements using the teaching skills. Ask the user only when resolution changes agreed behaviour, architecture or scope. Rerun every affected test and harness. Material candidate changes require a new frozen review; continue until no applicable change remains. +12. **Hand off locally.** Report delivered behaviour, changed artefacts, applicable TDD and modular-design decisions, checks and results, unavailable evidence, residual risks, both review results and every disposition. Stop without committing or publishing. + +### Completion boundary + +Successful completion requires: + +- the agreed behaviour exists locally; +- every production addition has a current purpose; +- TDD was applied wherever its trigger held; +- relevant project and catalogue harnesses passed or their unavailable evidence is explicit; +- the current candidate, not an earlier revision, received applicable Spec and Standards review; +- every suggestion was validated and no applicable correction or accepted improvement remains; and +- no in-scope failure is hidden by a weakened test, threshold, baseline, exclusion or suppression. + +### Authoring and evaluation evidence + +The future `$write-skill` session must test at least: + +- a clear ticket needing no elicitation or research; +- a material specification gap discovered during implementation; +- a behaviour change with a usable suite; +- a non-behavioural change for which TDD must not be manufactured; +- a repository missing a relevant harness; +- a review containing a high-ranked false positive and a lower-ranked required correction; +- a material correction that invalidates the previous review; and +- unrelated dirty work that must remain untouched. + +## `tdd` + +### Recurring job + +Teach a caller to implement observable behaviour through disciplined Red–Green–Refactor and leave a small, durable suite whose oracles can detect realistic defects while surviving behaviour-preserving structural change. + +### Applicability + +Apply when both conditions hold: + +1. the work adds, changes or repairs observable behaviour; and +2. an existing usable test suite can exercise that behaviour through a credible seam. + +Do not apply to documentation, human prose, dependency-only updates, configuration-only maintenance without an exercisable effect, generated artefacts or other non-behavioural changes. When behaviour changes but no usable suite exists, the caller proceeds without claiming TDD, uses the strongest available project-native feedback and reports the evidence limitation. Creating a suite or framework is permitted only when it is explicitly part of the agreed work. + +### Required teaching + +#### Establish the slice, seam and oracle + +- Start from one observable outcome and use a **tracer bullet** to select the smallest complete vertical slice. +- Choose the narrowest seam that exposes the promised result, hides structure allowed to change and retains the production risk that matters. +- Derive expected results independently before Green. Prefer requirements, accepted worked examples, contracts, invariants, trusted references, separate models or accepted prior behaviour. +- Treat current output capture as characterization, not proof of correctness. +- Name the realistic production break each retained test must catch and the hidden structural changes it must survive. + +#### Red + +- Write one focused behavioural test before the corresponding production change. +- Arrange only necessary state, act through the supported seam, assert the complete promised outcome and material unchanged state, then clean up acquired resources. +- Run the focused test and require the intended behavioural failure. A pass, syntax error, fixture failure or environment error is not Red. +- For bugs, reproduce the incorrect result and assert the desired one; a crash or implementation-derived expectation is not automatically a valid regression oracle. + +#### Green + +- Add only enough production code to satisfy the current behaviour. +- Do not anticipate later cases, introduce hypothetical options or refactor unrelated code. +- Run the focused test and relevant nearby tests. Fix production code rather than weakening the test when its independent oracle remains valid. + +#### Refactor + +- Refactor production design only while the behaviour is green. +- Keep tests unchanged during behaviour-preserving production refactoring. If hidden structural movement breaks a test, move its observation back to the promised outcome. +- Improve naming, cohesion, duplication and ownership without adding behaviour. +- Rerun the focused tests after each material step. + +#### Rationalise the suite + +- After production refactoring, reassess whether every retained test still protects a unique behaviour, boundary, invariant or risk. +- Remove duplicate examples that add no distinct failure detection; combine related cases with table-driven or property forms when that makes the oracle clearer. +- Remove obsolete implementation-detail assertions and broad snapshots whose only value is change detection. +- Preserve readable specification value and failure diagnosis; fewer tests are better only when no unique protection is lost. +- Apply a mental mutation check to wrong constants, wrong branches, missing effects, empty/default results and missing validation. Use actual mutation or coverage tooling only through applicable `quality-harnesses` guidance. + +### Required example coverage + +The final bundle must use concrete good/bad pairs to teach: + +- an independent literal or invariant versus a tautological expected value; +- a public result versus a private method or internal call graph; +- user-visible UI semantics versus CSS classes, wrapper elements or child indexes; +- persistence through supported write/read behaviour versus private schema inspection; +- a real internal collaborator versus unnecessary mocking; +- a focused fake, stub or mock at an uncontrollable external boundary; +- legitimate protocol interaction assertions versus incidental call choreography; +- a failing bug regression before its repair; +- deterministic time, randomness and cleanup; +- property, model-based, metamorphic and differential tests with their triggers and oracle limitations; and +- consolidation of several brittle examples into a smaller behaviourally equivalent test form. + +Examples should be small and concrete enough for an agent to imitate while explaining the decision independently of one language or framework. Core every-run examples belong on the loaded path; conditional techniques may use direct references. + +### Ownership-aware test-first rule + +When the current agent writes production behaviour before demonstrating Red, discard only that agent-owned implementation and restart test-first. Never delete or overwrite pre-existing or user-owned work. For an existing partial implementation, demonstrate sensitivity against the pre-change revision, a safely disabled behaviour or a controlled known-bad variant. When that cannot be done, label the retained test accurately as characterization or regression evidence rather than claiming TDD. + +### Expected resources + +The authoring session should evaluate a concise every-run `SKILL.md` plus directly linked references covering test design and oracles, doubles and boundaries, conditional generative techniques, and suite rationalisation. Existing reference names are not constraints; merge or split them only when their loading triggers differ materially. + +### Authoring and evaluation evidence + +Forward tests must distinguish this skill from generic “write a failing test first” behaviour. Cases must expose tautological tests, internal mocks, CSS/DOM coupling, invalid Red signals, overbuilt Green code, tests changed during refactoring, duplicated post-refactor tests, absence of a usable suite and pre-existing user code that must not be deleted. + +## `modular-design` + +### Recurring job + +Teach a caller to place behaviour and knowledge in deep, cohesive modules with small stable interfaces so current and plausible evidenced changes remain local. + +### Governing doctrine + +Use **information hiding** and **deep modules** as the backbone: + +- A module is any cohesive capability with an interface and implementation, regardless of whether code expresses it as a function, class, package, process or tier-spanning slice. +- Its interface includes everything callers must know: operations, data, invariants, ordering, errors, configuration, effects and material performance characteristics. +- A deep module gives callers substantial capability through a comparatively simple interface. +- Hide volatile design decisions and keep each material rule, invariant, protocol or representation with one coherent owner. +- Code that changes for the same evidenced reason should live together; one conceptual change should not require scattered knowledge across callers. +- Prefer established language, framework and repository idioms. An interface in the design sense does not require a language `interface`, class hierarchy or wrapper. + +### Required teaching + +#### Assign ownership + +Identify the behaviour, rule, invariant, protocol, representation, sequencing or side effect being introduced or changed. Keep it with its existing natural owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each decision once. + +#### Design the interface + +Minimise caller knowledge rather than raw method count. Hide decisions the caller does not need, choose domain-meaningful inputs and outcomes, make failure and side effects explicit, and avoid leaking framework or storage mechanics across an established domain or application seam. + +Reject shallow forwarding layers that only repeat another contract. Retain adapters or coordinators that own translation, policy, authorisation, compatibility, observation or another evidenced responsibility. + +#### Control coupling and change propagation + +Trace one plausible current-project change through the proposed owner, interface, consumers and tests. Reconsider the design when the same rule remains duplicated, elements that always change together are split, or one change still requires scattered edits. Do not substitute file size, class count or an abstract coupling score for this trace. + +#### Keep abstraction earned + +Keep stable local concrete dependencies direct. Add a seam, port or inversion only when current evidence requires substitution, isolated observation, multiple implementations, staged migration, a volatile external mechanism or separation of stable policy from mechanism. Avoid interface-per-class, universal repositories, generic service layers and hypothetical extension hooks. + +#### Design It Twice + +For a material new or substantially changed public interface, module boundary, seam or dependency direction, produce at least two genuinely different designs before selecting one. Compare them on interface depth, caller burden, cohesion, locality, dependency direction, testability and likely change propagation. Routine changes that fit an established coherent boundary proceed directly. + +The teaching skill supplies the comparison method; the calling agent owns producing and selecting the alternatives. It does not dispatch its own design workflow or require parallel agents. + +#### Apply conditional techniques + +Teach explicit triggers, trade-offs and good/bad examples for: + +- dependency inversion toward stable policy; +- ports and adapters at material technology or ownership boundaries; +- bounded contexts where one term has competing domain meanings; +- conceptual contours where domain language and change patterns reveal the natural capability grain; +- functional core/imperative shell where deterministic decisions are entangled with effects; +- seams for controlled substitution, observation or migration; +- composition and delegation; +- genuine substitutability when inheritance or polymorphism is considered; and +- co-location or extraction when it improves evidenced cohesion and locality. + +SOLID and design patterns are diagnostic tools, not mandatory architecture. Fowler smells belong to `code-review` as prompts for investigation, while this skill supplies the structural mechanisms needed to understand and correct a concrete concern. + +### Required example coverage + +Use concrete contrasted examples for deep versus shallow interfaces, information leakage, shotgun change propagation, pass-through layers, earned versus speculative seams, policy depending on mechanism, functional decisions mixed with effects, competing domain meanings and two alternative interface designs. Examples must show consequences for callers and future change rather than relying on labels alone. + +### Expected resources + +Keep the always-applicable ownership, interface, depth, locality and minimality rules on the loaded path. Link conditional technique detail and Design It Twice at their triggers. The future authoring session may reuse, replace or split the current `TECHNIQUES.md` according to observed loading needs. + +### Authoring and evaluation evidence + +Cases must test a simple change that needs no new abstraction, a material interface needing Design It Twice, a volatile external mechanism, a shallow forwarding layer, competing domain meanings, a large but cohesive module, a short but damaging coordinator and a request that would tempt speculative generality. + +## `quality-harnesses` + +### Recurring job + +Teach a caller to select, apply and interpret the quality harnesses relevant to a fixed change without confusing execution, metrics or control count with proof of correctness. + +### Loading path + +The final `SKILL.md` instructs the caller to: + +1. inspect the changed behaviour, implementation, repository requirements, measurement-path changes and material risks; +2. load and review the complete compact harness catalogue; +3. identify every catalogue entry whose trigger applies; +4. load the detailed reference for each applicable entry; +5. apply that knowledge within the caller's own workflow; and +6. preserve exact results, limitations, unavailable evidence and residual risk. + +The skill does not split entries into baseline and optional sections. A repository-required check is applicable because the repository requires it; another harness is applicable because the change exposes its named risk. Irrelevant entries are not loaded or run. + +### Catalogue contract + +Each catalogue entry contains one short paragraph with: + +- **Trigger:** the concrete repository requirement, change shape or risk cue that makes the harness relevant. +- **Narrow claim:** what an honestly clean result can support. +- **Limitations:** the important properties, environments or failure classes it cannot establish. +- **Anti-gaming checks:** how tautological oracles, exclusions, thresholds, suppressions, stale baselines, unrepresentative inputs, flaky retries or empty selectors could make the result misleading. +- **Reference:** the detailed instructions and examples to load only when the entry applies. + +The catalogue covers at least: + +| Harness family | Applicability and intended teaching | +| --- | --- | +| Build, compilation and packaging | Apply when repository-required or when build inputs, generated output, dependencies, packaging or runtime compilation change. Record the exact environment and artefact; one build does not establish portability or behavioural correctness. | +| Types, linting, formatting and framework diagnostics | Apply configured checks to their actual scope. Inspect baselines, exclusions and changed configuration; a clean result proves only encoded enabled rules. | +| Behavioural and regression suites | Apply focused evidence for changed behaviour and the repository-required broader suite. Distinguish new-behaviour evidence from pass-to-pass regression evidence and appraise the oracle through `tdd`. | +| Coverage | Use to locate changed or risk-relevant paths that no test executes. Never treat execution percentage as correctness, effectiveness or a target without repository authority. | +| Mutation | Use when test sensitivity is materially uncertain and existing tooling can assess a bounded change economically. Appraise surviving mutants and equivalent or irrelevant cases; never treat the score as correctness. | +| Static and security analysis | Apply configured analysis when changed code reaches its rules, and security-focused controls for trust boundaries, permissions, sensitive data, untrusted input, cryptography or dependency risk. Record rules, versions, suppressions and blind spots. | +| Architecture fitness rules | Apply when an authoritative dependency, layer, cycle, import or public-surface constraint is encoded. Check selectors and mappings are non-empty and current; passing rules cannot establish that the architecture decision itself is good. | +| Contracts and compatibility | Apply when APIs, events, schemas, serialization, storage formats, runtimes, browsers, databases or supported versions change. Record consumers and matrix coverage; finite examples cannot establish all consumer semantics. | +| Fuzzing and adversarial input | Apply to parsers, protocols and hostile or high-volume input when existing harnesses have meaningful crash, sanitizer, assertion or invariant oracles. Preserve corpora, seeds and minimized failures; execution alone rarely proves semantic correctness. | +| Accessibility | Apply when visible UI, semantics, focus, input, navigation, contrast or assistive-technology behaviour changes. Combine configured automation with required manual or specialist evidence; automation alone does not establish conformance. | +| Performance and load | Apply to hot paths, queries, I/O, algorithms, memory, caching or explicit budgets. Use representative workloads, pinned baselines, repeated measurements and correctness assertions; record variance and environment. | +| Concurrency and ordering | Apply to shared state, locks, transactions, asynchronous cancellation, retries, queues or idempotency. Use supported race detection, deterministic schedules, models or seeded stress; bounded success does not prove freedom from races. | +| Resilience and fault injection | Apply to timeouts, failover, partial dependency failure, recovery, rollout or operational degradation. Require an explicit hypothesis, safe boundary, recovery oracle and observability; injected faults cannot enumerate production combinations. | +| Migration and data integrity | Apply transformations to representative legacy states and verify invariants, mixed-version compatibility and rollback or forward recovery where promised. Fixtures cannot represent every production distribution or invalid state. | +| Reproducibility and provenance | Apply when toolchains, generated artefacts, signed output or environment-sensitive builds change. Record material inputs and comparison method; same-machine repetition is not cross-environment reproducibility. | +| Manual and specialist QA | Apply when automation cannot faithfully assess visual, physical, usability, domain or contextual outcomes. Preserve steps, inputs, versions, expected and actual results and evaluator limitations. | + +Complexity, cognitive complexity, size, churn, coupling and Fowler smell counts may identify a hotspot for investigation. Preserve repository-owned thresholds, but do not introduce universal thresholds or treat a metric as a finding without a concrete mechanism and consequence. + +### Infrastructure and authority boundary + +The calling workflow owns commands, mutations and remediation. This teaching skill does not run a verifier, return a pass/fail verdict or install tooling. A caller may add better behavioural tests through the existing suite when the implementation requires them. It must not introduce project-wide harness infrastructure unless that work is explicitly authorised. Missing relevant infrastructure is reported with the claim and risk left unassessed. + +### Expected resources + +Create a compact catalogue that is always reviewed when this skill applies. Link each entry to detailed reference material with selection guidance, execution boundaries, honest interpretation, anti-gaming examples and representative commands or procedures. References may group closely coupled harnesses when their triggers and evidence boundaries align; they must remain directly reachable from each catalogue entry. + +### Authoring and evaluation evidence + +Cases must include a small change with only repository-required checks, a security-sensitive change, changed tests that inflate coverage without an oracle, mutation survivors of mixed relevance, an empty architecture selector, a flaky check, unavailable specialist infrastructure and an irrelevant expensive harness that must not be selected. + +## `code-review` + +### Recurring job + +Independently review one fixed code change through isolated Spec and Standards agents and return consequence-ranked, falsifiable suggestions for a caller to validate and adjudicate. + +### Fixed candidate + +Resolve one explicit non-empty diff, revision range, branch comparison, pull-request revision, repair or local candidate. Capture its changed and in-scope untracked files, complete relevant source and tests, repository state and exact revision identifiers where available. If the target changes during review, mark affected results stale rather than silently retargeting them. + +### Review axes + +#### Spec + +Run this axis only when a behavioural authority is reasonably available: a feature specification, ticket, acceptance criteria, confirmed request, contract or another project source that states intended behaviour independently of the implementation. + +Give the fresh Spec agent: + +- the fixed candidate and complete changed files; +- the behavioural authority and relevant affected contracts or consumers; +- the consequence ranking and finding format; and +- a read-only boundary. + +Direct it to account for every applicable requirement in code and retained tests, account for introduced behaviour against authority, and follow states, contracts, data shapes, errors, effects and consumers far enough to find missing, partial, conflicting, excessive or regressed behaviour. + +Do not give it Standards findings, implementation-thread rationale, conversation history or the Standards packet. When no authority is reasonably available, do not infer intent from code or commits and do not stop Standards; report `Spec not performed: no behavioural authority reasonably available.` + +#### Standards + +Give a separate fresh Standards agent: + +- the same fixed candidate and complete changed files; +- repository instructions, conventions, architecture decisions and configured checks; +- relevant surrounding implementation, tests and supplied quality evidence; +- the full Fowler second-edition smell catalogue as diagnostic vocabulary; +- the applicable `modular-design`, `tdd` and `quality-harnesses` teaching paths; +- the consequence ranking and finding format; and +- a read-only boundary. + +Direct it to understand every changed line in its necessary context and assess correctness and regression risks, minimality, project and framework conventions, changed-test validity, modular ownership and interfaces, quality-harness selection and evidence integrity, recognised smells, and specialist risks exposed by the change. + +Do not give it Spec findings, implementation-thread rationale, conversation history or the Spec packet. The teaching skills inform its judgement without becoming nested delegated workflows. + +### Independent execution + +Run the applicable fresh agents in parallel. Their isolation prevents one review axis from anchoring or suppressing the other; a fresh context alone is not treated as proof of statistical independence. + +Each reviewer may run a safe, focused, non-mutating command only when it can confirm or falsify a concrete concern. It does not repeat the implementation's whole quality portfolio, edit code or tests, update snapshots or baselines, install dependencies, modify durable data or repair a finding. + +Before retaining a suggestion, the reviewer must try to disprove it using contradicting authority, existing handling, repository-sanctioned exceptions, surrounding code, a focused counterexample and the strongest benign interpretation. Unsupported generic advice, tooling-enforced nits and speculative best practices are omitted. + +### Consequence ranking + +- **High:** a credible path to materially wrong required behaviour, security or privacy compromise, data loss or corruption, major production or reliability failure, or a structural, test or evidence defect that makes the change untrustworthy. +- **Medium:** a concrete defect, regression risk or significant maintainability, modularity, test-quality or harness-integrity weakness with a bounded material consequence. +- **Low:** a local but evidenced issue whose correction has a concrete benefit. Personal taste, hypothetical flexibility and inconsequential style are not findings. + +Rank communicates consequence and attention only. It never directs the caller to implement a suggestion automatically; a high finding may be false, and a lower-ranked finding may still be a required repository or behavioural correction. + +### Output + +Keep `## Standards` and `## Spec` separate. Do not merge, deduplicate, suppress or cross-axis rerank their results before returning them to the caller. For an applicable clean axis, report `No findings.` For an omitted or stale axis, state that status explicitly. + +Each suggestion contains: + +```markdown +### [high|medium|low] Concise finding + +- Evidence: exact code location, applicable authority or criterion, and observed fact +- Consequence: concrete behavioural or code-health impact +- Suggested direction: the narrow outcome that would address the concern without prescribing an unverified patch +``` + +End with axis-specific counts and review scope. Do not issue an approval, verification or completion verdict. The caller owns validation, deduplication during adjudication, remediation, re-review and any publication decision. + +### Authoring and evaluation evidence + +Cases must include two axes with conflicting-looking but independently valid results, no Spec authority, a standards concern requiring teaching-skill knowledge, a high-ranked false positive, a low-ranked genuine requirement violation, an unsupported generic suggestion that must be removed, a targeted probe, scope drift and a clean axis. + +## Cross-skill acceptance + +The implementation group is ready only when matched fresh-agent evaluation demonstrates: + +- `implement` behaves as a small process owner and does not duplicate the teaching skills; +- the three teaching skills materially change caller behaviour without taking workflow ownership; +- TDD produces tests that fail for the intended reason, rely on independent oracles, survive structural change and can be rationalised without losing behaviour; +- modular design improves ownership, interface depth and change locality without automatic abstraction or class proliferation; +- quality-harness selection is complete for evidenced risks, excludes irrelevant controls and reports absent infrastructure honestly; +- Spec and Standards reviews are isolated, falsifiable and consequence-ranked; +- implementation rejects false positives rather than blindly obeying review; +- a changed candidate invalidates earlier review and affected quality evidence; +- the group composes for local implementation, debug repair and pull-request review without duplicating another review or verification method; and +- all skills stop before commit or publication unless a later lifecycle skill is explicitly invoked. + +## Authoring order + +Refine the skills in dependency order so later evaluations consume published contracts rather than provisional internals: + +1. `modular-design` +2. `tdd` +3. `quality-harnesses` +4. `code-review` +5. `implement` + +For each skill, use `$write-skill` to confirm the closed contract, establish no-skill and current-skill baselines where applicable, draft outside the destination, validate mechanically, compare through isolated fresh agents and publish only the smallest behaviourally justified bundle. Mark its checklist item in [`SKILLS_PLAN.md`](SKILLS_PLAN.md) complete only after that full authoring and evaluation pass. + +## Current cutover boundary + +This planning session creates this brief, updates the lifecycle plan and removes the obsolete `verify-change` bundle. It does not rewrite the five surviving skills, `debug`, `respond-to-incident`, `review-pull-request` or the README. Those files may temporarily describe the previous composition until their scheduled authoring sessions align them with this authority. diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 800750c..8850571 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -135,13 +135,15 @@ Local Markdown uses one file per ticket under `docs/features//tickets ### Deliver +The detailed authoring contract for the implementation skills is maintained in [`IMPLEMENTATION_SKILLS_PLAN.md`](IMPLEMENTATION_SKILLS_PLAN.md). This plan retains their lifecycle outcomes and composition; the implementation plan owns their teaching content, internal sequence, resources and evaluation expectations. + #### `implement` -- **Outcome:** One ticket or small agreed body of work exists as a minimal, verified local change. +- **Outcome:** One ticket or small agreed body of work exists as a minimal, locally checked and independently reviewed change. - **Inputs:** The stated ticket or confirmed work, linked product and technical authorities, repository guidance and current code. -- **Output:** Working code, tests and other in-scope artefacts with reproducible verification and no unresolved required review finding. -- **Composition:** Requires `verify-change` and independent `code-review`; conditionally invokes `elicit-with-context`, `research`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. -- **Stops:** After the agreed behaviour is implemented and verified. It does not commit or publish the change. +- **Output:** Working code, retained tests and other in-scope artefacts, applicable quality evidence and limitations, and no unresolved applicable review change. +- **Composition:** Applies the teaching guidance from `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, requires independent `code-review`, and conditionally invokes `elicit-with-context` when implementation exposes a material missing detail or decision. +- **Stops:** After the agreed behaviour is implemented, applicable checks pass, unavailable evidence is explicit and the current candidate has completed the review-and-correction loop. It does not commit or publish the change. #### `commit` @@ -164,7 +166,7 @@ Local Markdown uses one file per ticket under `docs/features//tickets - **Outcome:** Someone else's pinned pull request receives an independent, evidence-backed assessment. - **Inputs:** The complete pull-request change, originating specification or work item, repository standards and relevant product and technical authorities. - **Output:** Prioritised findings covering behavioural conformance, regressions, minimality, conventions, maintainability, architecture and verification evidence. -- **Composition:** Requires `verify-change` and `code-review`; conditionally invokes `modular-design` and `research`. +- **Composition:** Requires `code-review`, whose Standards axis applies the implementation teaching skills; conditionally invokes `research` when external evidence is material to the review. - **Stops:** After reporting findings. It does not change the contributor's branch or publish comments, approvals or requested changes unless the user's invocation includes that outcome. ### Debug and Respond @@ -174,7 +176,7 @@ Local Markdown uses one file per ticket under `docs/features//tickets - **Outcome:** A defect has an evidence-backed root cause and, when authorised by the requested outcome, a minimal verified repair. - **Inputs:** Expected and observed behaviour, the failing signal, relevant environment evidence and current implementation authorities. - **Output:** A reproducible diagnosis and, for repair work, regression protection plus a reviewed local change. -- **Composition:** Repair requires `verify-change` and `code-review`; conditionally invokes `research`, `modular-design`, `tdd`, `maintain-ubiquitous-language` and `maintain-decision-records`. +- **Composition:** Repair applies `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, and requires `code-review`; conditionally invokes `research` and `elicit-with-context` when their triggers apply. - **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. #### `respond-to-incident` @@ -182,7 +184,7 @@ Local Markdown uses one file per ticket under `docs/features//tickets - **Outcome:** An active production incident is understood sufficiently to stabilise service safely and verify the resulting service state. - **Inputs:** The incident signal, production evidence, affected systems, available operational controls and current communications. - **Output:** A verified mitigation or recovery, concise timeline, current impact and explicit unresolved risks or follow-up needs in the owning incident system. -- **Composition:** Conditionally invokes `debug`, `verify-change` and `research` when they are necessary to stabilise the incident. +- **Composition:** Conditionally invokes `debug` and `research` when they are necessary to stabilise the incident. Any local repair uses the implementation-quality model through `debug`. - **Stops:** When service is stable or the exact external blocker is established. It does not perform the later learning review. #### `review-incident` @@ -212,23 +214,23 @@ Local Markdown uses one file per ticket under `docs/features//tickets | `research` | Investigates a material subject using high-trust evidence and creates a trusted cited report for the consumer to synthesize. | | `maintain-ubiquitous-language` | Keeps `CONTEXT.md` aligned inline with confirmed project-specific language. | | `maintain-decision-records` | Preserves rare accepted technical and architectural decisions whose rationale warrants a focused ADR. | -| `modular-design` | Applies evidence-backed modularity, information hiding, cohesion, coupling and maintainability principles to a scoped design decision. | -| `tdd` | Drives red-green-refactor through a stable observable boundary and runnable test suite. | -| `verify-change` | Selects and executes the applicable project and risk-triggered quality harnesses and reports reproducible evidence and limitations. | -| `code-review` | Independently assesses a scoped change for required behaviour, regressions, minimality, conventions, maintainability, code smells and architecture. | +| `modular-design` | Teaches callers to structure code through information hiding, deep modules, cohesive ownership, stable interfaces, locality and conditional design techniques. | +| `tdd` | Teaches callers to implement observable behaviour through red-green-refactor and retain a small, durable, independently meaningful test suite. | +| `quality-harnesses` | Teaches callers to select and apply relevant quality harnesses from a trigger-led catalogue without overstating their evidence. | +| `code-review` | Independently reviews a fixed change through isolated Spec and Standards agents and returns consequence-ranked suggestions for its caller to adjudicate. | | `maintain-agents` | Records confirmed, non-inferable repository-wide agent guidance while keeping permanent context lean. | Software-project skills route user-held questions through `elicit-with-context`. Base `elicit` remains available for non-software work and as that router's questioning component. `write-skill` remains a separate suite-development tool. It is not presented as part of the software-delivery lifecycle. -## Quality model +## Implementation quality model ### Research prerequisite -Before finalising `implement`, `tdd`, `verify-change` or `code-review`, conduct focused research using primary technical sources, established standards and credible field practices. Persist one cited report that establishes: +Before finalising `implement`, `tdd`, `modular-design`, `quality-harnesses` or `code-review`, conduct focused research using primary technical sources, established standards and credible field practices. Persist one cited report that establishes: -- a small universal quality baseline +- the obligations common to implementation regardless of harness selection - selection triggers for additional harnesses - what each harness can and cannot demonstrate - ways metrics and tests can be gamed @@ -237,9 +239,15 @@ Before finalising `implement`, `tdd`, `verify-change` or `code-review`, conduct The research evaluates practices used successfully in real projects without turning every available harness into mandatory ceremony. -### Verification portfolio +The current evidence is recorded in [`docs/research/agentic-coding-quality-guardrails-2026-08-20.md`](docs/research/agentic-coding-quality-guardrails-2026-08-20.md). The detailed implementation-skill decisions derived from it are fixed in [`IMPLEMENTATION_SKILLS_PLAN.md`](IMPLEMENTATION_SKILLS_PLAN.md). + +### Teaching utilities + +`modular-design`, `tdd` and `quality-harnesses` are teaching skills. They supply the calling agent with applicable knowledge; they do not become delegated workers, mutate independently, return separate workflow verdicts or take ownership from their caller. `implement` uses that knowledge to write and assess code. The Standards agent created by `code-review` uses the same knowledge to review code and evidence. -`verify-change` discovers the repository's existing commands, tests, CI-equivalent checks and QA procedures, then selects the applicable portfolio for the changed behaviour and risk. The portfolio may include: +### Quality-harness catalogue + +`quality-harnesses` directs its caller to inspect the change, repository requirements and material risks, then review a compact catalogue and apply every relevant entry. Each catalogue entry states its trigger, the narrow claim it can support, its limitations, its anti-gaming checks and the detailed reference to load when applicable. The catalogue may include: - unit, acceptance, integration, contract and system tests - executable business examples such as Gherkin where they improve precision @@ -248,22 +256,24 @@ The research evaluates practices used successfully in real projects without turn - security, accessibility, performance, compatibility and resilience checks - reproducible manual or specialist QA procedures -The repository's existing constraints remain authoritative. A skill may strengthen evidence for the current change but must not weaken or skip required checks silently. +The catalogue does not divide controls into a mandatory baseline and optional extras. Applicability governs selection. Complexity, size, coverage, mutation and other metrics remain evidence for a named question rather than universal gates or composite quality scores. + +The repository's existing constraints remain authoritative. A caller may strengthen evidence through the existing test and quality infrastructure but must not weaken or skip required checks silently. -When an existing project cannot support meaningful verification, the skill reports the limitation and recommends explicit planned work. It does not install a project-wide test framework or quality tool merely to satisfy its own process. New projects establish required infrastructure through their architecture and initial delivery tickets. +When an existing project lacks an applicable harness, the caller records the missing evidence and residual risk. It does not install a project-wide test framework or quality tool unless establishing that infrastructure is explicitly part of the agreed work. New projects establish required infrastructure through their architecture and initial delivery tickets. ### Maintainability judgement -Passing checks does not prove that a change is well designed. `code-review` and `modular-design` also require the implementation to: +Passing checks does not prove that a change is well designed. `implement` applies `modular-design` throughout the change, and the independent Standards review uses `modular-design`, `tdd` and `quality-harnesses` to assess whether the implementation: -- contain only the behaviour and code needed for the agreed outcome -- follow the language, framework and repository conventions -- reuse established helpers and abstractions where they fit -- avoid duplication, speculative generality and recognised code smells -- preserve or improve information hiding and module boundaries -- remain understandable and economical to change +- contains only the behaviour and code needed for the agreed outcome +- follows the language, framework and repository conventions +- reuses established helpers and abstractions where they fit +- avoids duplication, speculative generality and recognised code smells +- preserves or improves information hiding and module boundaries +- remains understandable and economical to change -Review findings are corrected through the relevant implementation and verification loop until no required finding remains unresolved. +Spec and Standards suggestions are ranked high, medium or low by consequence. Severity never decides remediation. The implementation owner validates each suggestion against its authority and evidence, applies required corrections and proportionate in-scope improvements, rejects unsupported or out-of-scope findings, and returns material behaviour, architecture or scope decisions to the user. Review repeats after material candidate changes until no applicable change remains. ## Representative flows @@ -328,8 +338,8 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ## Implementation order -1. Complete the quality-harness research and use its conclusions to fix the engineering-quality utility contracts. -2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `verify-change`, `code-review` and `maintain-agents`. +1. Complete the quality-harness research and implementation-skill authoring plan. +2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `quality-harnesses`, `code-review` and `maintain-agents`. 3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build the architecture-assessment path: `review-architecture`. @@ -339,13 +349,14 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre ### Implementation checklist - [x] Complete quality-harness research +- [x] Define the implementation-skill authoring plan - [x] Update `elicit` skill - [x] Update `research` skill - [x] Create `maintain-ubiquitous-language` skill - [x] Create `maintain-decision-records` skill - [-] Update `modular-design` skill - [-] Update `tdd` skill -- [-] Create `verify-change` skill +- [-] Create `quality-harnesses` skill - [-] Update `code-review` skill - [x] Update `maintain-agents` skill - [x] Update `define-product` skill @@ -377,7 +388,7 @@ The revised suite is ready when: - required and conditional composition is explicit and does not duplicate supporting skill instructions - ticket-creating skills use the local Markdown or external destination named in `AGENTS.md` and preserve its exact concepts without Propulsion configuration - documentation preserves enduring intent without copying observable implementation or live tool state -- implementation always completes applicable verification and independent review +- implementation always applies the relevant modular-design, TDD and quality-harness guidance and completes independent review - absent quality infrastructure is reported honestly and never causes unrelated tooling to be installed implicitly - external mutations remain within the outcome authorised by the invoked skill and are read back after application - release, deployment and continuous monitoring remain outside the suite diff --git a/skills/verify-change/SKILL.md b/skills/verify-change/SKILL.md deleted file mode 100644 index 24d7be6..0000000 --- a/skills/verify-change/SKILL.md +++ /dev/null @@ -1,67 +0,0 @@ ---- -name: verify-change -description: Verifies fixed software changes with applicable quality harnesses and reproducible evidence. Use for a diff, revision, pull request, repair, or completed implementation. -metadata: - invocation: model -disable-model-invocation: false ---- - -# Verify Change - -Verifies a fixed software change through the smallest faithful portfolio of repository-required and risk-triggered quality evidence without repairing or certifying the implementation. - -## Process - -### 1. Fix the verification basis - -Read repository instructions and resolve the exact change set from caller context or the user's stated diff, revision, branch comparison, pull request, repair, or other fixed scope. Capture its revisions when applicable, changed and untracked paths, current repository status, intended behaviour, governing specifications and decisions, repository rules, architecture promises, known risks, and relevant pre-existing results. Inspect the patch, affected implementation, tests, quality configuration, and enough surrounding evidence to understand what changed without performing an independent code review. Ask when the scope or expected behaviour is materially ambiguous; when it cannot be fixed, return the exact blocker without running checks. The change, authorities, initial state, and verification target are explicit. - -### 2. Build the evidence map - -Apply **bidirectional verification traceability** for evidence selection and accounting: trace each applicable behaviour, repository rule, architecture promise, and material risk forward to the changed scope and evidence capable of assessing it; trace each material changed path, selected harness, result, limitation, and verdict statement back to its governing claim or rule. Do not use this map to replace the independent conformance and maintainability judgement owned by code review. - -Discover existing repository commands, tests, CI-equivalent gates, and reproducible QA procedures. Every applicable repository-required gate remains selected. Add a risk-triggered harness only when it tests a named claim or exposes a distinct material failure mode; multiple checks resting on the same signal do not create independent assurance. When a specialist risk, unfamiliar harness, or evidential boundary makes selection unclear, consult [Harness Selection](references/HARNESSES.md). Do not add tests, author an ad hoc executable check, generated-input campaign, benchmark, or oracle, install tools, create project-wide infrastructure, or alter CI to fill a gap, even when the addition would be transient. A read-only inspection or probe may characterise a directly observable artefact or diagnose existing evidence, but it does not become a missing behavioural harness. Record the gap for the verdict and recommend explicit planned work. Each material claim has a selected existing evidence route or a stated limitation. - -### 3. Bound the execution - -For each selected command or procedure, establish the revision or diff, environment, configuration, fixtures or data, seed, material tool versions, expected outputs, and local or external effects. Prefer a check, dry-run, read-only, or disposable mode. Ordinary caches, build artefacts, temporary files, disposable services, and isolated local test data are permitted only when their boundary is understood; clean up only effects created and owned by this run. Never remove or overwrite pre-existing user work. - -Do not run an operation that can alter source, tests, snapshots, baselines, suppressions, dependencies, durable project data, or an external system unless the user has separately authorised that exact verification effect and it does not remediate the change. Use a non-mutating mode when available; otherwise record the check as not run and its consequence. The execution plan is reproducible and stays within available authority. - -### 4. Execute and appraise the evidence - -Run the portfolio in widening order where project dependencies permit: focused reproduction or behavioural evidence; configured static, formatting, type, framework, and architecture checks; related component, integration, contract, or system evidence; broader regression and build gates; then activated specialist procedures. A faster or narrower check does not substitute for a slower faithful one. Existing CI or supplied evidence counts only when its provenance ties it to the frozen change and exposes enough of the execution boundary and result to appraise; otherwise run the applicable local equivalent or record the limitation. - -For each executable behavioural check, identify its **test oracle**: the requirement, invariant, worked example, versioned contract, trusted reference, accepted prior behaviour, explicit domain decision, or other authority capable of distinguishing acceptable from incorrect behaviour. Treat implementation-derived expectations, regenerated snapshots, proxy metrics, and execution without a discriminating outcome as weak or absent oracles. Ask what plausible counterexample would refute the claim and what defect in the harness, configuration, inputs, exclusions, or oracle could make a passing result misleading. For any metric, state the claim it informs and how the observed value bears on that claim; never let a proxy erase a failed gate or unsupported behaviour. - -Record one evidence entry per claim with: - -- the exact claim; -- the harness, its repository or procedure provenance, why it applies, and whether it is required or risk-triggered; -- the execution boundary; -- the observed result: `pass`, `fail`, `flaky-inconclusive`, or `not run`, with salient output, exit status, and artefact location where applicable; -- evidence integrity: `current`, or `stale` when post-basis drift applies; -- the inputs, paths, states, rules, environments, or acceptance criteria actually exercised; -- the limitation and residual risk; and -- the in-scope resolution taken, or the planned work or user decision needed for stronger evidence. - -Preserve every retry, seed, failing counterexample, and contradictory observation. A flaky or inconclusive result never becomes a pass through repetition. A required gate that fails remains a failure even when evidence proves it predates the change; state that provenance and whether it masks change-specific signals. The executed portfolio has inspectable claim-level results rather than an undifferentiated green summary. - -### 5. Reinspect scope and harness integrity - -Compare the final repository state and diff with the frozen basis. Account for changes to source, tests, snapshots, dependencies, generated artefacts, quality configuration, baselines, ignore lists, suppressions, and scripts. Distinguish declared temporary effects from material changes, but do not revert an unexpected change when ownership is uncertain. When any in-scope source or measurement-path change occurred after the scope was fixed, identify the paths, retain each observed result, set its evidence integrity to `stale`, treat it as unable to satisfy the affected claim, and stop without silently retargeting the run. The final evidence refers to the frozen change and an intact measurement path, or its invalidation is explicit. - -### 6. Determine the verification verdict - -Use exactly one verification-specific verdict: - -- `passed within stated scope` when every applicable required gate passed, each material claim has faithful evidence, no material result is unreliable or stale, and all remaining limitations are stated; -- `partially verified` when useful evidence passed but a named material claim or risk lacks a faithful harness or reliable result; -- `unverified` when the requested behaviour could not be exercised, an applicable required gate did not run, or stale evidence leaves no adequate basis for the requested scope; or -- `failed` when an applicable required gate or acceptance criterion failed and remains unresolved. - -A failed required gate takes precedence over incomplete evidence. Do not compute a confidence percentage or describe static inspection, compilation, coverage, or any other single route as proof beyond the claim it exercised. The verdict characterises verification evidence only; it does not certify the complete implementation or replace independent code review. - -### 7. Report and stop - -Return the exact frozen scope, authorities, material risks, selected and omitted harnesses, execution boundaries, commands or procedures, claim-level evidence entries, before-and-after repository state, verdict, limitations, residual risks, and any recommended planned work. Name every check that was not executed and why. Do not repair a failure, add a missing harness, weaken a gate, commit, publish, release, deploy, or continuously monitor. The caller receives reproducible evidence and one bounded verdict for its own remediation, review, or lifecycle decision. diff --git a/skills/verify-change/agents/openai.yaml b/skills/verify-change/agents/openai.yaml deleted file mode 100644 index 57b4514..0000000 --- a/skills/verify-change/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Verify Change' - short_description: 'Run proportionate quality harnesses for a fixed change' -policy: - allow_implicit_invocation: true diff --git a/skills/verify-change/references/HARNESSES.md b/skills/verify-change/references/HARNESSES.md deleted file mode 100644 index 2439c1a..0000000 --- a/skills/verify-change/references/HARNESSES.md +++ /dev/null @@ -1,43 +0,0 @@ -# Harness Selection - -Load this reference when a material risk, unfamiliar quality harness, or unclear evidential boundary makes the smallest faithful portfolio difficult to select. Repository rules, specifications, architecture decisions, configured checks, and accepted thresholds remain authoritative. - -## Select by claim and failure mode - -Start with the claim, not a preferred tool. Choose an existing harness only when its observation and oracle can distinguish a failure of that claim within a known environment. Retain every applicable repository-required gate. Add another existing route only when it exposes a distinct material failure mode or reduces a named oracle, environment, or state-space weakness enough to justify its feedback cost. Do not invent a transient test, fuzzer, benchmark, generator, or reference implementation to close a project evidence gap; return that need as planned implementation work. - -Run cheap diagnostic checks early when dependencies permit, but never substitute speed for fidelity. Prefer deterministic, local, narrowly diagnostic, machine-readable procedures for correction feedback. Preserve slower integration, specialist, or manual procedures when they alone exercise a material boundary. A required unreliable check remains unreliable evidence rather than a pass. - -## General-purpose harnesses - -| Harness family | Select when | Evidence supplied | Principal boundary and integrity check | -| --- | --- | --- | --- | -| Focused unit or component test | A stable observable boundary can exercise the changed behaviour economically. | The oracle accepted the exercised nominal, boundary, failure, or off-nominal examples. | It does not establish collaborations or environments outside the boundary. Reject tautological assertions, implementation-derived expectations, and hidden-structure coupling. | -| Executable acceptance or business example | A specification, acceptance criterion, or worked example is authoritative and precision benefits from execution. | The system produced the specified externally observable outcome for the examples. | Examples are not exhaustive. Keep business language tied to real production behaviour rather than step definitions that merely restate code. | -| Integration, contract, or system test | The change crosses a process, service, database, framework, protocol, hardware, or user-facing boundary. | Selected collaborators, interfaces, side effects, or end-to-end paths worked in the exercised configuration. | Contract tests do not prove provider internals or complete collaboration; isolated integration does not prove the deployed system. Record fakes, versions, topology, and omitted real boundaries. | -| Static type, lint, format, framework, or architecture check | The repository configures it, or a named structural or language rule governs the change. | The analysed source satisfied the configured rules in the inspected scope. | It proves only encoded and enabled rules. Inspect baselines, exclusions, suppressions, generated files, database freshness, and changed configuration. | -| Build or package check | Build inputs, dependencies, generated artefacts, packaging, or runtime compilation changed, or the repository requires it. | The declared inputs produced the expected artefact in the recorded environment. | One successful local build does not prove portability or reproducibility. Record dependency resolution, generated changes, environment, and artefact identity. | -| Broader regression suite | Changed behaviour or dependencies could affect established nearby or system behaviour, or the repository requires the suite. | Previously encoded expectations still passed across the suite's actual coverage. | Passing regressions inherit every oracle, exclusion, environment, and flakiness weakness in the suite. Distinguish focused new-behaviour evidence from pass-to-pass regression evidence. | -| Coverage analysis | A coverage gap could reveal changed code or a material path never exercised. | Which configured code, branches, or paths executed under the selected suite. | Coverage is not correctness or test effectiveness. Do not target a percentage without repository authority or treat execution as a discriminating oracle. | -| Changed-code mutation analysis | The sensitivity of changed or high-consequence tests is materially uncertain and existing infrastructure can run it economically. | Selected seeded faults were detected or survived within the mutation scope. | Survivors may be equivalent, irrelevant, unreachable, or genuine gaps. Record operator set, exclusions, timeouts, scope, and survivor appraisal; do not mandate whole-suite mutation. | -| Reproducible manual or specialist procedure | Automation cannot faithfully exercise a visual, physical, usability, domain, or contextual claim. | A named actor followed recorded steps against stated inputs and observed retained outcomes. | One observation may depend on competence and environment. Preserve steps, versions, data, screenshots or logs, expected result, actual result, and specialist limitations. | - -## Risk-triggered specialist harnesses - -| Material risk cue | Candidate existing evidence | What a clean result cannot establish | -| --- | --- | --- | -| Trust boundary, authentication or authorisation, sensitive data, cryptography, untrusted input, command or query construction, file handling, dependency or security configuration | Threat or attack-surface analysis, configured secret/dependency/SAST checks, security-focused tests, DAST, fuzzing, penetration procedure, qualified security review | One scanner or finite attack set cannot establish absence of vulnerabilities. Confirm scope, rule/database version, environment, permissions, suppressions, and untested threats. | -| User interface, semantic structure, keyboard or assistive-technology interaction, contrast, focus, motion, or accessibility requirement | Configured automated accessibility check plus interaction, browser, assistive-technology, or knowledgeable human evaluation as the claim requires | Automated rules cover only machine-detectable conditions; one browser, device, or evaluator does not establish conformance across the target population. | -| Hot path, algorithm, query, serialization, start-up, resource constraint, volume, latency, or throughput requirement | Focused benchmark, profile, load, stress, or soak procedure against an authoritative budget and representative workload | A result does not generalise beyond workload, environment, warm-up, sample size, variance, versions, and competing load. Compare a relevant baseline; do not infer improvement from one noisy run. | -| Supported runtime, browser, device, operating system, database, architecture, locale, time zone, or dependency matrix | Repository matrix, compatibility suite, representative build/run, or targeted manual procedure | Tested combinations do not cover omitted versions or interactions. Use the authoritative support matrix and record exact combinations. | -| Retry, timeout, failure recovery, partial outage, queue, distributed boundary, or degraded dependency | Fault injection, failure-path integration test, resilience procedure, recovery drill, or state-transition assertion | Synthetic faults may not match operational timing or correlated failures. Record injected conditions, recovery oracle, state cleanup, and unexercised modes. | -| Shared mutable state, parallel work, lock, transaction, async ordering, idempotency, or race-sensitive path | Deterministic schedule test when available, repeated stress with preserved seed and trace, invariant check, or focused concurrency procedure | Stress cannot prove absence of races; bounded schedule exploration cannot prove states beyond its model and bounds. Preserve the failing schedule or trace. | -| Schema, migration, backfill, serialization, data transformation, retention, destructive operation, or rollback promise | Representative production-shaped fixtures, forward and reverse migration where required, invariant/reconciliation queries, backup/restore or recovery procedure | Synthetic or sampled data may omit real distributions and invalid legacy states. Record scale, constraints, before/after invariants, failure recovery, and environment fidelity. | -| Environment-sensitive build, generated output, compiler/toolchain change, signed or distributed artefact | Clean build, repeated build under relevant declared variance, artefact comparison, provenance or signature check | Same-machine repetition does not establish reproducibility elsewhere. Record all material inputs, toolchain, environment variation, and comparison method. | -| High-consequence protocol, state machine, concurrency algorithm, or invariant already supported by project modelling infrastructure | Model checking, proof checking, symbolic execution, or other bounded formal analysis | The result applies to the stated model, assumptions, abstraction, and explored bounds; it does not prove model-to-code or environment correspondence without separate traceability. | - -## Guard against self-confirming evidence - -Inspect every changed test, snapshot, baseline, suppression, quality rule, ignore list, generator, fixture, benchmark, and harness script that participates in the verdict. Confirm that its change has specification or repository authority and that its oracle can disagree with the implementation. Treat a weakened or unexplained measurement path as a limitation even when its command is green, and leave independent acceptability judgement to code review. - -For generated or sampled inputs, record the generator domain, distribution, seed, shrink or reduction result, and failing counterexample where applicable. For metrics, record the governing claim, measurement scope, interpretation, variance or threshold authority, and blind spots. Never combine unrelated evidence into a composite score that can hide a required failure or unsupported material risk. From 5fdd2afa719b94aa42750db3004f13e76a05e029 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 09:48:50 +0100 Subject: [PATCH 59/79] docs(agents): constrain skill bundle content --- AGENTS.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/AGENTS.md b/AGENTS.md index 46edf4d..4321b5f 100644 --- a/AGENTS.md +++ b/AGENTS.md @@ -1,5 +1,5 @@ - When a user correction establishes a reusable repository-wide rule, ask whether to invoke `$maintain-agents` to add it. -- When authoring or reviewing a skill, justify its behaviour using only context available to the agent at runtime. +- When authoring or reviewing a skill, include only runtime instructions and knowledge for its agent; exclude authoring rationale, evaluation history, and composition discussion, and justify behaviour only with runtime-available context. - When eliciting within a software project, invoke `$elicit-with-context` instead of `$elicit`; use `$elicit` directly only for non-software work or as part of `$elicit-with-context`. - When composing skills, rely on each invoked skill's public contract; do not restate or steer its internal logic. - Default to a user-agent workflow; introduce team roles, named accountable people, approval bodies, or third-party coordination only when the user or project evidence requires them. From 4af63e912530320c027dc794aedf03c56d6b0c83 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 09:49:10 +0100 Subject: [PATCH 60/79] feat(modular-design): deepen structural design guidance --- SKILLS_PLAN.md | 2 +- skills/modular-design/SKILL.md | 76 ++++++++++++++----- skills/modular-design/agents/openai.yaml | 2 +- .../references/DEPENDENCY-TECHNIQUES.md | 48 ++++++++++++ .../references/DESIGN-IT-TWICE.md | 28 +++++++ .../references/DOMAIN-TECHNIQUES.md | 32 ++++++++ .../references/EFFECT-TECHNIQUES.md | 14 ++++ .../modular-design/references/TECHNIQUES.md | 25 ------ 8 files changed, 183 insertions(+), 44 deletions(-) create mode 100644 skills/modular-design/references/DEPENDENCY-TECHNIQUES.md create mode 100644 skills/modular-design/references/DESIGN-IT-TWICE.md create mode 100644 skills/modular-design/references/DOMAIN-TECHNIQUES.md create mode 100644 skills/modular-design/references/EFFECT-TECHNIQUES.md delete mode 100644 skills/modular-design/references/TECHNIQUES.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 8850571..8489109 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -354,7 +354,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `research` skill - [x] Create `maintain-ubiquitous-language` skill - [x] Create `maintain-decision-records` skill -- [-] Update `modular-design` skill +- [x] Update `modular-design` skill - [-] Update `tdd` skill - [-] Create `quality-harnesses` skill - [-] Update `code-review` skill diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index aa9dcc6..0e68a70 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -1,6 +1,6 @@ --- name: modular-design -description: Applies an evidence-backed structural-maintainability standard. Use when designing or assessing ownership, boundaries, contracts, dependencies, seams, or change propagation. +description: Teaches structural design through information hiding, deep modules, cohesive ownership, stable interfaces, and change locality. Use when writing or assessing code, boundaries, contracts, dependencies, seams, or change propagation. metadata: invocation: model disable-model-invocation: false @@ -8,36 +8,78 @@ disable-model-invocation: false # Modular Design -Supplies structural findings and constraints inside a caller's workflow by locating each material decision with a coherent owner, making the smallest complete change, and keeping caller knowledge proportionate to the capability provided. +Teaches a caller to make the smallest coherent change through deep modules whose owned knowledge and likely changes remain local. ## Process -### 1. Fix the behaviour and evidence +### 1. Fix the structural basis -Start from confirmed scoped behaviour and runtime-available caller and project evidence: project guidance, architecture decisions, domain language, relevant source, contracts, and tests. Explicit project authorities and established language and framework conventions govern when they conflict with this baseline. Keep the assessment structural: ownership, information hiding, cohesion, coupling, contracts, dependencies, seams, invariants, policies, protocols, sequencing, side effects, and change propagation. Leave product authority, behavioural correctness, security, performance, accessibility, formatting, testing, and general style to their applicable authorities except where they materially shape a contract or boundary. The behaviour, evidence, and material structural decision are explicit. +Read the confirmed behaviour, caller outcome, repository guidance, domain language, architecture decisions, relevant source, contracts, and tests. Follow explicit project authorities and established language and framework idioms when they conflict with this baseline. -### 2. Assign the affected knowledge +Treat a **module** as any cohesive capability with an interface and implementation: a function, object, component, package, process, service, or tier-spanning slice. Treat its **interface** as everything callers must know, including operations, data, invariants, ordering, errors, configuration, effects, and material performance characteristics. Do not equate modules with files, classes, layers, or deployment units. -Identify the representation, contract, invariant, policy, protocol, sequencing, side effect, or framework decision the behaviour introduces or changes. Keep it with its natural existing owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each material rule once, and let code that changes for the same project-evidenced reason change together. The affected knowledge has one coherent owner rather than being repeated across callers. +Use confirmed behaviour and applicable quality requirements as constraints on ownership, boundaries, and contracts. Do not redefine those requirements during structural design. -### 3. Make the smallest coherent change +### 2. Assign each decision to one owner -Deliver the complete scoped behaviour without speculative abstraction or unrelated cleanup. Permit a bounded enabling refactor only when the behaviour cannot fit coherently without it; change no more surrounding structure than the new behaviour needs. Keep a stable local concrete dependency direct unless meaningful variation, isolation, observation, migration, a technology boundary, or policy-mechanism separation justifies a seam. +Name the behaviour and every rule, invariant, protocol, representation, sequence, side effect, or framework decision it introduces or changes. Keep each with its natural existing owner when that owner remains coherent; otherwise create the smallest boundary that owns the complete concern. An owner has the knowledge and authority needed to uphold the decision, not merely the nearest file or conventional technical layer. -Realize the design through established project and framework idioms. Objects may own identity, state, invariants, and cohesive behaviour; functions, closures, structural types, language modules, components, and framework primitives are equally valid when they preserve the same ownership and contract. Introduce a language interface only when callers need a stable contract distinct from its realization, and use inheritance only for genuine substitutability or a required framework extension. The result is the smallest idiomatic change that completely delivers the behaviour. +Represent each material decision once. Co-locate elements that change for the same current, evidenced reason. Separate elements when their responsibilities and axes of change differ. Prefer domain and capability boundaries over uniform file sizes or technical-layer symmetry. -### 4. Keep the contract deep +Keep derived knowledge behind the owner of its source representation. When another module needs a fact such as a total, classification, eligibility, or transition derived from owned state, use or add a domain-meaningful owner operation instead of making the caller traverse that state. Let the caller derive it only when the representation is deliberately public and the derivation belongs to the caller's own policy. -Treat a module as a cohesive capability, whether code expresses it as a function, object, package, process, or tier-spanning slice. Its contract is everything callers must know, including behaviour, data, invariants, ordering, errors, configuration, side effects, and material performance characteristics. +Check these contrasts: -Apply **deep modules**: make the caller-visible contract markedly simpler than the cohesive capability and knowledge hidden behind it. Absorb or remove shallow forwarding layers that merely repeat another contract or redistribute navigation; retain a layer when it owns translation or another evidenced boundary responsibility. The contract repays its caller burden without combining unrelated knowledge to appear deeper. +- **Local rule:** put a pricing threshold in the pricing policy that decides it. **Leaked rule:** repeat it in a controller, job, validator, and their tests. +- **Owned derivation:** ask an order for `totalWeightGrams()` when it owns line weights and quantities. **Leaked derivation:** make delivery, fulfilment, and reporting each traverse order lines to reconstruct that fact. +- **Cohesive size:** keep a large parser together when its grammar and invariants change as one concern. **Fragmented size:** split it only to satisfy a line or class count, making one grammar change cross several files. +- **Damaging coordinator:** reconsider even a short coordinator when it knows every participant's order, data shape, and error handling; brevity does not make scattered knowledge cohesive. -### 5. Check locality and applicable techniques +### 3. Design a deep interface -Trace plausible current-project changes through the owner, contract, consumers, and affected code. Reconsider the boundary when one owned decision would still require scattered changes, when several elements must always change together but are split, or when the proposed owner changes mostly for unrelated reasons. Do not substitute file size, class count, method count, dependency scores, abstract metrics, or speculative agent-context savings for this evidence. +Apply **information hiding** and **deep modules**. Give callers substantial cohesive capability through a comparatively simple, domain-meaningful contract. Minimise what callers must know rather than method count. Hide representations, mechanisms, sequencing, framework details, and policies that callers do not need. Make required inputs, outcomes, failures, ordering, and effects explicit. -Load [Modular Design Techniques](references/TECHNIQUES.md) only when current evidence already presents one of its conditions: policy directly depends on a volatile mechanism; an application conversation needs independence from technologies, replacement, isolated execution, or a material external boundary; domain meanings or change contours compete; controlled substitution, observation, or migration is needed; or deterministic decisions are entangled with effects. Do not open the reference to discover whether a condition exists. Apply only the fitting technique. The resulting structure keeps the owned knowledge and its likely changes local without introducing an unevidenced abstraction. +Keep the contract narrower and more stable than the decisions behind it without combining unrelated knowledge merely to appear deep. A design interface does not require a language `interface`, base class, wrapper, or extra file. -## Handoff +Check these contrasts: -Return the structural findings or constraints, their runtime evidence, the responsible owner and contract, the change-locality trace, any triggered technique, and unresolved structural uncertainty. The caller retains elicitation, requirements, architecture and review artifacts, implementation, tests, verification, external mutation, and the decision to continue or stop its workflow. +- **Deep:** `reservation.confirm(command)` owns validation, state transition, persistence sequencing, and the confirmation outcome. **Shallow:** callers invoke `validate`, `markConfirmed`, `save`, and `publish` in the required order. +- **Hidden vendor protocol:** an adapter accepts checkout terms and maps vendor fields, units, and expected failures. **Leaked protocol:** controllers and policies know `vendorAmount`, vendor error classes, and response codes. +- **Owned boundary:** retain a layer that translates, authorises, applies policy, preserves compatibility, observes, or coordinates an application conversation. **Pass-through:** absorb a layer whose only act is forwarding the same arguments and result to another contract. + +### 4. Keep the change coherent and earned + +Make the least production, test, and supporting change that completely delivers the confirmed behaviour and leaves its knowledge local. Fewest lines or files is not the goal. Permit a bounded enabling refactor only when the behaviour cannot fit coherently without it; alter no more surrounding structure than the behaviour needs. + +Keep a stable local concrete dependency direct. Add an abstraction only for current evidence of meaningful variation, isolated observation, replacement, staged migration, a volatile external mechanism, a material technology or ownership boundary, or separation of stable policy from mechanism. Reuse an existing extension point only when its contract fits the present behaviour. Do not add hypothetical implementations, options, generic services, interface-per-class, universal repositories, plugin systems, or unrelated cleanup. + +Use objects for cohesive identity, state, lifecycle, and invariants when they fit. Use functions, closures, structural types, modules, components, composition, and framework primitives when they preserve the same ownership with less machinery. Use inheritance only for genuine substitutability or a required framework extension. + +### 5. Compare material structural choices + +When introducing or substantially changing a public interface, module boundary, seam, or dependency direction, load [Design It Twice](references/DESIGN-IT-TWICE.md). Produce at least two genuinely different responsibility or contract designs, compare them, and select the smallest design supported by current evidence. Routine work that fits an established coherent boundary proceeds directly. + +The caller produces and selects the alternatives inside its own workflow. Do not dispatch a design workflow, require parallel agents, or turn the comparison into a separate artefact unless the caller's outcome requires one. + +### 6. Apply only triggered techniques + +Classify triggers from the evidence already loaded, then read only the smallest applicable reference. An ordinary ownership choice, one domain vocabulary, one conditional, or the mere possibility of future change is not a trigger: + +- Load [Dependency Techniques](references/DEPENDENCY-TECHNIQUES.md) when stable policy directly knows a volatile mechanism; a material technology or ownership boundary needs translation; controlled substitution, observation, replacement, or migration is required; composition competes with inheritance; or polymorphic substitutability is in question. +- Load [Domain Techniques](references/DOMAIN-TECHNIQUES.md) only when one term has evidenced competing meanings or rules, domain language and observed change patterns disagree with the current boundary, or co-location versus extraction remains a material choice after the ownership analysis. Do not load it for a single coherent domain model or an ordinary file-placement choice. +- Load [Decision and Effect Techniques](references/EFFECT-TECHNIQUES.md) only when substantial deterministic policy across multiple cases or transitions is entangled with I/O, time, mutable state, or framework lifecycle effects. Do not load it for one calculation or conditional merely because the surrounding operation performs I/O. + +Do not load references to search for a reason to use a technique. Use one fitting mechanism when triggers overlap. Treat SOLID and named patterns as diagnostic vocabulary, not mandatory architecture; treat smells and metrics only as prompts to investigate a concrete mechanism and consequence. + +### 7. Trace locality + +Trace one plausible current-project change through the proposed owner, interface, consumers, implementation, and tests. Prefer a change suggested by requirements, integrations, repository history, or present volatility over an imagined future. + +Reconsider the design when the same decision remains duplicated, elements that always change together are split, an owner changes mostly for unrelated reasons, callers still coordinate hidden sequencing, or one conceptual change requires scattered edits. Do not substitute file size, class count, method count, churn, coupling scores, complexity thresholds, or speculative agent-context savings for this trace. + +Finish with the owner, caller-visible contract, hidden decisions, locality trace, triggered technique or material comparison, and unresolved structural uncertainty. The caller retains its outcome, edits, commands, tests, verification, review findings, artefacts, and stopping decision. + +## Rules + +- Prefer project evidence over generic architecture advice. +- Preserve the complete confirmed behaviour while improving structure; do not trade correctness or scope for aesthetic purity. diff --git a/skills/modular-design/agents/openai.yaml b/skills/modular-design/agents/openai.yaml index 8e43c61..19cb267 100644 --- a/skills/modular-design/agents/openai.yaml +++ b/skills/modular-design/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Modular Design' - short_description: 'Apply structural maintainability decisions' + short_description: 'Design deep modules with local change' policy: allow_implicit_invocation: true diff --git a/skills/modular-design/references/DEPENDENCY-TECHNIQUES.md b/skills/modular-design/references/DEPENDENCY-TECHNIQUES.md new file mode 100644 index 0000000..0ab6087 --- /dev/null +++ b/skills/modular-design/references/DEPENDENCY-TECHNIQUES.md @@ -0,0 +1,48 @@ +# Dependency Techniques + +Load only for a triggered dependency, technology boundary, seam, composition, or substitutability decision. A contract in design does not require a language `interface`. + +## Invert volatile mechanism knowledge + +Apply **dependency inversion** when stable domain or application policy directly knows a materially volatile framework, device, vendor, persistence, transport, or delivery mechanism. Define the smallest policy-owned contract in policy language and make the mechanism satisfy it. + +- **Good:** invoicing policy asks a `TaxRateSource` for the rate and the vendor adapter owns authentication, wire fields, and response mapping. +- **Bad:** invoicing imports vendor request types and switches on vendor status codes, or adds an interface that repeats one stable local calculator without changing dependency knowledge. + +Cost: an owned contract and composition wiring must be maintained. Keep a stable local concrete dependency direct when inversion creates only interface-per-class indirection. + +## Isolate a material boundary + +Apply **ports and adapters** when an application-purpose conversation must remain independent of a material external technology or ownership boundary, isolated execution, replacement, staged migration, or multiple implementations. State the port in the application's language; let adapters translate UI, persistence, network, vendor, or test technology. Framework types stay at framework-owned edges. + +- **Good:** `PlaceOrder` accepts an application command and returns an application outcome while HTTP and database adapters translate at the edges. +- **Bad:** wrap every framework call, mirror a vendor SDK method-for-method, or add a generic provider registry for one fixed integration. + +Cost: translation, wiring, and sometimes duplicated boundary types. A concrete adapter without a language interface can still be the right port boundary when it hides the protocol and no separate runtime contract is earned. + +## Add a controlled seam + +Add a **seam** when substitution, observation, replacement, or staged migration requires behaviour to vary without editing the calling location. Use the smallest enabling point: a parameter, callback, collaborator, provider, module export, composition-root binding, or framework facility. + +- **Good:** inject a clock where expiry policy needs deterministic time. +- **Bad:** introduce factories and provider interfaces around every deterministic local function in case tests might mock it. + +Cost: another variation point callers and maintainers must understand. Prefer direct behaviour when it is local, stable, and observable through its supported outcome. + +## Prefer composition and delegation + +Use composition or delegation when a capability can be assembled from independently owned behaviours without claiming an is-a relationship. Keep delegation deep: the composing owner must add policy, translation, sequencing, or a simpler capability rather than merely repeat another contract. + +- **Good:** a `Checkout` delegates tax calculation and payment authorisation while owning their ordering and checkout outcome. +- **Bad:** `PaymentService.process()` only forwards to `PaymentProcessor.process()` with identical inputs, outputs, and responsibility. + +Cost: extra navigation and wiring. Absorb a delegation layer that owns no decision. + +## Require genuine substitutability + +Use inheritance or polymorphism only when every subtype can honour the complete base contract for all supported callers without strengthened preconditions, weakened outcomes, surprising effects, or type tests. Prefer composition for optional behaviour or shared implementation that does not form a stable substitutable concept. + +- **Good:** supported storage adapters honour the same explicit read/write/error contract. +- **Bad:** `ReadOnlyRepository extends Repository` but throws from inherited `save`, forcing callers to inspect its concrete type. + +Cost: base-contract evolution constrains every subtype. A required framework extension is valid when its lifecycle contract is honoured, even when composition would otherwise be preferred. diff --git a/skills/modular-design/references/DESIGN-IT-TWICE.md b/skills/modular-design/references/DESIGN-IT-TWICE.md new file mode 100644 index 0000000..4d4443a --- /dev/null +++ b/skills/modular-design/references/DESIGN-IT-TWICE.md @@ -0,0 +1,28 @@ +# Design It Twice + +Load this reference only for a material new or substantially changed public interface, module boundary, seam, or dependency direction. + +## Form two real designs + +Describe at least two alternatives that differ in ownership, caller knowledge, contract, or dependency direction. Renaming, moving the same class, changing interface syntax, or comparing a credible design with an obvious straw design does not count. + +Keep each alternative only detailed enough to decide. For each, state: + +- which module owns the rule, representation, sequence, and effects; +- what callers must know and do; +- what the interface promises, including failure and effects; +- which dependencies point toward which owner; +- how tests observe the promised behaviour; and +- how one plausible evidenced change propagates. + +Compare interface depth, caller burden, cohesion, locality, dependency direction, testability, transition cost, and speculative machinery. Select the smallest coherent design supported by current evidence; record the losing alternative's material disadvantage and any uncertainty that could reverse the choice. + +## Contrast: delivery quote boundary + +Suppose checkout needs a delivery quote from one current carrier. + +**Design A — concrete checkout adapter:** `DeliveryQuotes` depends directly on `FastShipDeliveryQuotes`, which translates FastShip fields and failures into checkout terms. This hides the vendor protocol with no language interface. It is preferable when one implementation is approved, the concrete adapter is replaceable at the composition root, and tests can observe policy without a distinct runtime contract. + +**Design B — policy-owned port:** `DeliveryQuotes` depends on a checkout-owned `DeliveryQuoteSource`; `FastShipDeliveryQuoteSource` implements it. This adds a separate stable contract. It is preferable when isolated execution, multiple implementations, staged replacement, or an ownership boundary currently requires the policy to remain independent of the concrete adapter. + +Both designs isolate vendor translation. Selecting B merely because another carrier might exist someday is speculative; selecting A when application policy must directly import vendor types leaves the boundary incomplete. diff --git a/skills/modular-design/references/DOMAIN-TECHNIQUES.md b/skills/modular-design/references/DOMAIN-TECHNIQUES.md new file mode 100644 index 0000000..7b65338 --- /dev/null +++ b/skills/modular-design/references/DOMAIN-TECHNIQUES.md @@ -0,0 +1,32 @@ +# Domain Techniques + +Load only when domain meanings, observed change contours, or co-location and extraction are materially in question. + +## Separate competing meanings + +Apply **bounded contexts** when the same term has competing meanings, rules, lifecycles, or representations. Keep one ubiquitous language and model inside each evidenced context; translate explicitly at their boundary. + +- **Good:** Sales `Customer` owns prospects and purchasing eligibility while Support `Customer` owns service entitlement, with an explicit identity translation. +- **Bad:** one universal `Customer` accumulates nullable sales, billing, fulfilment, and support state, so every change affects unrelated callers. + +Cost: translation and explicit context ownership. Do not create bounded contexts merely because departments or folders have different names. + +## Follow conceptual contours + +Apply **conceptual contours** when domain language and observed axes of change reveal a more natural capability grain than current technical layers or uniform class sizes. Keep entity identity, state, invariants, and cohesive behaviour together. Use a standalone domain service only for a significant process that belongs naturally to no entity or value object. + +- **Good:** an `Allocation` owns quantity invariants and release behaviour that always change together. +- **Bad:** split `AllocationValidator`, `AllocationUpdater`, and `AllocationRules` by technical action, making one invariant change cross all three. + +Cost: a domain-shaped module may be larger or cross conventional layer boundaries. That is acceptable when its contract remains clear and changes stay local. + +## Co-locate or extract for evidenced locality + +Co-locate code when it shares one owner and changes for the same evidenced reason. Extract when a complete sub-concern has a distinct contract, reuse by real consumers, independent volatility, or enough hidden knowledge to repay navigation. + +- **Good co-location:** keep a component's one-use rendering helper beside the component while they evolve together. +- **Bad extraction:** move every small helper into a global utilities module, coupling unrelated capabilities to a grab bag. +- **Good extraction:** isolate a tax-jurisdiction rule used by checkout and invoicing behind one policy-owned contract. +- **Bad co-location:** duplicate that rule in both consumers to avoid adding a file. + +Cost: extraction adds naming, navigation, and a contract; co-location can overload an owner when change reasons diverge. Decide from ownership and propagation rather than size. diff --git a/skills/modular-design/references/EFFECT-TECHNIQUES.md b/skills/modular-design/references/EFFECT-TECHNIQUES.md new file mode 100644 index 0000000..c29c327 --- /dev/null +++ b/skills/modular-design/references/EFFECT-TECHNIQUES.md @@ -0,0 +1,14 @@ +# Decision and Effect Techniques + +Load only when substantial deterministic decisions are entangled with I/O, time, mutable state, or framework lifecycle effects. + +## Form a functional core and imperative shell + +Apply **functional core, imperative shell** when explicit input values can drive substantial deterministic policy while a narrow edge performs I/O, time access, mutation, or framework integration. The core returns decisions or intended effects; the shell obtains inputs, executes effects, and maps failures. + +- **Good:** a renewal function receives subscription state and `now`, then returns `renew`, `expire`, or `no-op`; a job handler loads state, calls it, and performs the selected effect. +- **Bad:** policy reads the system clock, database, queue, and configuration throughout nested branches, so every case requires effectful setup and policy changes alter orchestration. + +Preserve stateful ownership when identity, lifecycle, effect ordering, transactions, or invariant enforcement is intrinsic. Do not turn an application into a bag of pure functions with an unencapsulated shell. + +Cost: explicit values and decision types cross the core boundary, and the shell must translate effect failures honestly. Use the technique only when the deterministic policy is substantial enough to repay that contract. diff --git a/skills/modular-design/references/TECHNIQUES.md b/skills/modular-design/references/TECHNIQUES.md deleted file mode 100644 index 96b351f..0000000 --- a/skills/modular-design/references/TECHNIQUES.md +++ /dev/null @@ -1,25 +0,0 @@ -# Modular Design Techniques - -Load this reference only when repository evidence satisfies a technique's condition. The direct process in `SKILL.md` remains authoritative; use one mechanism to satisfy overlapping conditions when possible rather than adding duplicate layers. - -## Direct dependencies towards policy - -Apply **dependency inversion** when stable domain or application policy directly knows a materially volatile framework, device, vendor, persistence, transport, or delivery mechanism. Define the smallest policy-owned contract in the policy's language and make the mechanism satisfy it. Keep a stable local concrete dependency direct when inversion would only create an interface-per-class and wiring. - -## Isolate an application conversation - -Apply **ports and adapters** when an application-purpose conversation must remain independent of multiple technologies, replacement, isolated execution, or a material external boundary. Let the port state the application's need and let adapters translate UI, persistence, network, vendor, or test technology. Keep framework types at framework-owned edges and avoid wrapping every framework call. - -## Separate domain models - -Apply **bounded contexts** when the same term or concept has competing meanings, rules, or models. Keep one ubiquitous language inside each evidenced context and translate explicitly between them. - -Apply **conceptual contours** when domain language and observed axes of change reveal a more natural capability grain than technical layers or uniform class sizes. Keep cohesive entity behaviour and invariants together; use a standalone domain service only for a significant process that belongs to no entity or value object. - -## Create controlled variation or observation - -Add a **seam** when substitution, observation, replacement, or staged migration needs behaviour to vary without editing the calling location. Supply the smallest enabling point that selects the behaviour, such as a parameter, collaborator, provider, module export, or framework facility. The caller owns the reason and verification; direct behaviour remains preferable when it is already local, stable, and observable. - -## Separate decisions from effects - -Apply **functional core, imperative shell** when substantial deterministic decisions are entangled with I/O, time, mutable state, or framework lifecycle behaviour. Express the decisions as transformations of explicit values and keep effects at a narrow edge. Preserve stateful ownership when identity, lifecycle, effect ordering, or invariant enforcement is intrinsic; do not leave an unencapsulated application in the shell. From 29faed9bfe9ce235a94dc30e2a0930e3384676bd Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 10:31:52 +0100 Subject: [PATCH 61/79] feat(tdd): finalize test-driven development guidance --- SKILLS_PLAN.md | 2 +- skills/tdd/SKILL.md | 44 +++++++---- skills/tdd/agents/openai.yaml | 2 +- skills/tdd/references/GENERATIVE-TESTING.md | 32 ++++++++ skills/tdd/references/TEST-DESIGN.md | 79 +++++++++++++++++++ skills/tdd/references/TEST-DOUBLES.md | 56 +++++++------- skills/tdd/references/TEST-QUALITY.md | 84 --------------------- 7 files changed, 174 insertions(+), 125 deletions(-) create mode 100644 skills/tdd/references/GENERATIVE-TESTING.md create mode 100644 skills/tdd/references/TEST-DESIGN.md delete mode 100644 skills/tdd/references/TEST-QUALITY.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 8489109..c8a743c 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -355,7 +355,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `maintain-ubiquitous-language` skill - [x] Create `maintain-decision-records` skill - [x] Update `modular-design` skill -- [-] Update `tdd` skill +- [x] Update `tdd` skill - [-] Create `quality-harnesses` skill - [-] Update `code-review` skill - [x] Update `maintain-agents` skill diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 81a6cf6..0265b81 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -1,6 +1,6 @@ --- name: tdd -description: Implements observable features and bug fixes through red-green-refactor with durable tests. Use when an existing runnable test suite can exercise the change at a stable observable boundary. +description: Teaches test-driven development through red-green-refactor with durable behavioural tests. Use when a usable suite can exercise changed behaviour through a credible seam. metadata: invocation: model disable-model-invocation: false @@ -8,32 +8,50 @@ disable-model-invocation: false # Test-Driven Development -Builds one observable behaviour at a time through red-green-refactor. Retained tests detect promised behaviour changes while surviving changes to hidden structure. +Teaches a caller to implement one observable behaviour at a time and retain a small suite that detects promised-behaviour defects without resisting structural change. -## Prerequisite +## Prerequisites -TDD applies only when an existing runnable test suite can credibly exercise the requested behaviour through a stable observable boundary. Otherwise return control with the exact missing condition. Do not install or invent a test framework or add a harness to satisfy this prerequisite; new-project test infrastructure belongs in architecture and initial delivery work. Apply TDD to observable behaviour, including configuration with observable effects; leave documentation, configuration-only maintenance without an exercisable effect, and behaviour-preserving refactors with the caller. +Apply TDD only when the work adds, changes, or repairs observable behaviour and an existing usable suite can exercise it through a credible seam. Otherwise return the missing condition to the caller; proceed through the strongest project-native feedback without claiming TDD. Create a suite or framework only when that infrastructure is part of the agreed work. ## Process -### 1. Establish the baseline +### 1. Fix the slice, seam, and oracle -Read repository instructions, caller-confirmed behaviour and acceptance evidence, applicable modular constraints, relevant production code, and current tests. Identify the test command and run the smallest relevant focused baseline. Record baseline failures without attributing a pre-existing failure to the change; continue only when the focused Red and Green signals remain distinguishable, otherwise return the blocker. Use **tracer bullets** to select the smallest end-to-end behaviour that produces one requested observable result. Apply caller-supplied modular constraints, or invoke `$modular-design` only when the slice presents a new material structural-maintainability decision. Choose the narrowest boundary that exposes the result, hides structure allowed to change, and credibly exercises production behaviour. Identify a **test oracle**—a requirement, accepted example, published protocol, independent invariant, trusted reference or separate model, accepted prior behaviour, or explicit domain decision—capable of disagreeing with the implementation; when none exists, return the missing behavioural decision. Consult [Test Quality](references/TEST-QUALITY.md) when the boundary, oracle, or proposed assertion needs deeper guidance. The baseline, behaviour, boundary, oracle, and applicable modular constraints are explicit. +Read the confirmed behaviour, repository instructions, applicable modular constraints, relevant production code, and current tests. Identify the focused command and run its baseline. Keep pre-existing failures visible; continue only when they cannot hide the next Red or Green signal. + +Select a **tracer bullet**: the smallest complete vertical slice that produces one requested observable outcome. Choose the narrowest seam that exposes the promise, hides changeable structure, and retains the production risk that matters. Derive the expected result before Green from a requirement, accepted example, contract, invariant, trusted reference, separate model, or accepted prior behaviour. Current output is characterization, not proof of correctness. Load [Test Design](references/TEST-DESIGN.md) and apply its boundary, oracle, determinism, and durability rules. + +If the caller already wrote this production behaviour, discard only caller-owned uncommitted implementation and restart test-first. Preserve pre-existing and user-owned work. For an existing partial implementation, demonstrate sensitivity against the pre-change revision, a safely disabled behaviour, or a controlled known-bad variant. When none is safe, retain an independently meaningful test as regression or characterization evidence without calling the work TDD. ### 2. Red -Add one focused test: arrange only the necessary state, act once through the selected boundary, assert the complete promised outcome and any material unchanged state, and clean up resources the test acquires. Name one plausible promised-behaviour defect that must make the test fail and one hidden structural change that must leave it unchanged. Keep internal collaborators real. When an uncontrollable boundary must be controlled or observed, choose the least powerful **Test Double** that supplies the required evidence; consult [Test Doubles](references/TEST-DOUBLES.md) before introducing a double or interaction assertion. +Write one focused behavioural test before its production change. Arrange only necessary state, act once through the supported seam, assert the complete promised outcome, and clean up acquired resources. Assert unchanged state only when its preservation is promised and the action can actually affect the observed value. Name one realistic production break the test must detect and one hidden structural change it must survive. + +Keep internal collaborators real. When an uncontrollable boundary needs substitution or observation, load [Test Doubles](references/TEST-DOUBLES.md), choose the least powerful double, and assert only contractually meaningful effects. -For a bug, reproduce the incorrect behaviour; adopt an already-failing regression test only when it independently specifies the desired behaviour. Run the focused test and confirm that it fails for the expected behavioural reason. When it fails because of the test or environment, remain in Red: correct an in-scope defect or report the blocker, then rerun until the intended failure is observed. Meaningful red evidence exists before Green begins. +For a bug, reproduce the incorrect result and assert the independently established desired result. Structure the test so failure is reported against the promise: an unexpected exception that aborts before any oracle assertion is not Red unless the test explicitly establishes that successful completion is the promise. When count or order at a system boundary is promised, assert it directly rather than inferring it from final state. + +Run the focused test. Red exists only when an oracle fails for the intended missing or incorrect behaviour; a pass, syntax error, fixture failure, or environment error is invalid. Correct an in-scope test or environment defect and rerun, or return the blocker. Preserve the exact Red command and failure. ### 3. Green -Implement only enough production code to satisfy the behaviour using project and framework conventions, then run the focused test and relevant nearby tests. The new behaviour passes without speculative production code or hidden baseline failures. +Add only enough production code for the current behaviour. Treat confirmed preconditions as inputs to this slice; do not invent validation, errors, fallbacks, or branches for their violation. Do not anticipate later cases, add hypothetical options, or refactor unrelated code. Run the focused test and relevant nearby tests. Fix production code while the independent oracle remains valid; never weaken the test merely to obtain Green. Preserve the exact Green command and result. + +### 4. Refactor while green + +Improve production naming, cohesion, duplication, ownership, and interfaces without adding behaviour. Keep behavioural tests unchanged through production refactors. When structural movement alone breaks a test, move its observation back to the promised outcome. Clarify test names or fixtures only without weakening the oracle or counterfactuals. Run the focused tests after each material step. + +If refactoring exposes a new material structural decision not resolved by the caller's modular constraints, return it to the caller for resolution before continuing. + +### 5. Rationalise the retained suite + +After production refactoring, identify the unique behaviour, boundary, invariant, or risk protected by each affected test. Remove obsolete implementation-detail assertions, broad snapshots used only as change alarms, and examples that add no distinct failure detection. Consolidate equivalent cases with a table or property only when the resulting oracle and failure diagnosis remain clear. -### 4. Refactor +Mentally mutate wrong constants, branches, validation, effects, and empty or default results. Strengthen the smallest test or input that would let a realistic defect survive. Use actual mutation or coverage tooling only when the caller's quality-harness guidance selects it. -Improve the production design behind the selected boundary while keeping behaviour fixed. Preserve the test across changes to algorithms, collaborators, storage, rendering, or other hidden structure; when structure alone breaks it, move the observation back to the promised outcome. Clarify test code only without weakening its oracle or quality counterfactuals. Run the focused tests after each material change until the design is clear and green. The cycle ends without a refactor regression. +### 6. Complete and return -### 5. Complete the cycles +Repeat Red, Green, Refactor, and rationalisation for each remaining confirmed behaviour. When example cases cannot economically cover a large input or state space, or a direct oracle is unavailable, load [Generative Testing](references/GENERATIVE-TESTING.md) and use only the triggered technique. Run the complete relevant suite. -Repeat Red, Green, and Refactor for each remaining confirmed behaviour, then run the complete relevant suite. Report the behaviours delivered, exact Red and Green evidence, refactors, commands and results, unresolved baseline failures, and other limitations. The requested behaviour and durable retained tests are implemented; the caller owns the overall work outcome, broader quality portfolio, later `$verify-change`, independent `$code-review`, commit or artefact decisions, and whether its workflow can stop. +Return the behaviours delivered; retained tests and the unique protection each adds; exact Red and Green evidence; refactors; commands and results; unresolved baseline failures; fidelity limits; and other missing evidence. The caller retains ownership of broader quality harnesses, independent review, the complete work outcome, commits, and publication. diff --git a/skills/tdd/agents/openai.yaml b/skills/tdd/agents/openai.yaml index 1386203..bbba392 100644 --- a/skills/tdd/agents/openai.yaml +++ b/skills/tdd/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Test-Driven Development' - short_description: 'Build behaviour with red-green-refactor' + short_description: 'Build durable behaviour through red-green-refactor' policy: allow_implicit_invocation: true diff --git a/skills/tdd/references/GENERATIVE-TESTING.md b/skills/tdd/references/GENERATIVE-TESTING.md new file mode 100644 index 0000000..ff67d8f --- /dev/null +++ b/skills/tdd/references/GENERATIVE-TESTING.md @@ -0,0 +1,32 @@ +# Generative Testing + +Use an example test by default. Select one of these techniques only when its trigger applies and its oracle is independently justified. + +| Technique | Trigger | Oracle and limitation | +| --- | --- | --- | +| **Property-based** | Stable invariants over many values; parsers, serializers, algebra, collections, boundary-rich pure logic | Assert an independent invariant. Biased generators and tautological properties can pass indefinitely. | +| **Model-based** | Stateful workflows or protocols where sequences and invalid transitions matter | Compare with a smaller independent state model. A model copied from production repeats its defects and can suffer state explosion. | +| **Metamorphic** | Direct expected outputs are unavailable but relationships between executions are known | Assert relations such as round-trip, permutation invariance, monotonicity, or scaling. A relation detects inconsistency, not necessarily the uniquely correct result. | +| **Differential** | Two genuinely comparable implementations, versions, backends, or conforming tools can receive the same inputs | Compare normalized outcomes. Agreement may be jointly wrong; shared code or assumptions weaken independence. | + +## Apply the selected technique + +1. State the rule and its authority before writing the generator or model. +2. Constrain generated data to valid and deliberately invalid domains that exercise the promised behaviour. +3. Reject vacuous properties, unreachable model states, shared oracle logic, and normalization that erases material differences. +4. Run with a recorded seed. Minimize a failure and retain the smallest counterexample as a focused regression test. +5. Report the sampled scope and the oracle's limitation; a passing generated run is not proof over the full space. + +```typescript +// Tautological: repeats the implementation's sorting comparison. +property(array(integer()), (values) => + expect(sort(values)).toEqual(values.toSorted(productionComparator)), +); + +// Independent metamorphic relation: permutation cannot change the multiset. +property(array(integer()), (values) => + expect(frequencies(sort(values))).toEqual(frequencies(values)), +); +``` + +For stateful behaviour, keep the model smaller than production and express only authoritative states and transitions. For differential testing, first identify allowed variation and normalize only that variation. Preserve every meaningful discrepancy until it is explained as a candidate defect, an oracle defect, or permitted divergence. diff --git a/skills/tdd/references/TEST-DESIGN.md b/skills/tdd/references/TEST-DESIGN.md new file mode 100644 index 0000000..ecc7791 --- /dev/null +++ b/skills/tdd/references/TEST-DESIGN.md @@ -0,0 +1,79 @@ +# Test Design + +Apply this guide to every retained test. + +## Select a durable seam + +Use the narrowest seam that is: + +- **Observable:** exposes a supported result, state, error, effect, persistence, navigation, or user-perceivable outcome. +- **Stable:** hides algorithms, collaborator graphs, storage layouts, rendering wrappers, generated selectors, and other replaceable decisions. +- **Predictive:** exercises the behaviour and technology whose production failure matters. + +Add a broader companion test only when the narrower seam cannot retain a material integration risk. Test a pure rule through its public function, a database guarantee against the real engine when its semantics matter, and an end-user interaction through the rendered application when wiring or accessibility is part of the promise. + +| Concern | Brittle or circular test | Durable behavioural test | +| --- | --- | --- | +| Expected value | Repeats the production reduction to calculate `expected` | Uses the accepted example `10 + 5 = 15` or an independent invariant | +| Domain result | Calls a private pricing method or asserts helper delegation | Calls the supported quote operation and asserts its public result | +| UI | Asserts CSS classes, wrapper tags, child indexes, or DOM depth | Interacts by role or label and asserts visible text, value, focus, state, or navigation | +| Persistence | Inspects private tables, columns, or ORM calls | Writes and reads through supported interfaces; reloads when durability matters | +| Error | Asserts a private branch or stack shape | Asserts caller-visible type, code, promised message, recovery, and unchanged state | +| Snapshot | Accepts a broad incidental object or markup snapshot | Asserts selected semantic fields, unless exact bytes or markup are the contract | +| Bug | Treats the existing crash or output as the expected result | Reproduces the wrong result, then asserts a requirement-derived correction | + +Representation is valid only when that representation is itself a published contract. + +## Build an independent oracle + +Prefer, in order: + +1. a confirmed requirement, accepted worked example, published protocol, or domain decision; +2. a law, invariant, or contract independent of the production algorithm; +3. a trusted external reference or separately implemented model; or +4. accepted prior behaviour when preservation is the requirement. + +Encode the expectation as an independently reasoned literal or predicate. Never call the production helper, copy its algorithm, approve newly captured output, or assert only that execution completed unless completion is the entire promise. When no credible oracle exists, return the missing behavioural decision. + +## Make the test deterministic and complete + +Arrange local explicit state, perform one meaningful action, assert the complete promised outcome, then release acquired resources. Add a negative or unchanged-state assertion only when the requirement makes it material and the action has a real causal path to the observed value. Do not clone, replace, or otherwise isolate that value before the action and then claim its preservation as evidence. One behaviour may need several cohesive assertions; split multiple independent actions or rules. + +Control time and randomness through an injected clock or deterministic source at the system boundary. Use fixed instants and seeds and preserve a minimized failing counterexample. Give each test isolated data and restore process state, files, transactions, services, timers, and environment values it changes. Avoid sleeps, execution order dependencies, live external services, and shared mutable fixtures. + +```typescript +// Brittle: wall-clock timing and an assertion too weak to catch the wrong expiry. +expect(issueToken().expiresAt).toBeGreaterThan(Date.now()); + +// Durable: controlled time and the exact promised interval. +expect(issueToken({ now: fixedClock('2030-01-01T00:00:00Z') }).expiresAt).toBe( + '2030-01-01T01:00:00Z', +); +``` + +## Challenge and rationalise + +Ask both counterfactuals before retaining a test: + +- If promised behaviour broke, would this test fail? +- If only hidden structure changed, would this test remain unchanged and pass? + +After refactoring, map every affected test to unique protection. A focused example remains useful when it specifies a distinct rule or gives a clearer failure than a consolidated form. Otherwise remove duplicates or combine them: + +```typescript +// Repetition with no distinct structure. +expect(priceFor(1)).toBe(10); +expect(priceFor(2)).toBe(20); +expect(priceFor(3)).toBe(30); + +// One rule with the same protection and clearer extension. +for (const [quantity, expected] of [ + [1, 10], + [2, 20], + [3, 30], +]) { + expect(priceFor(quantity)).toBe(expected); +} +``` + +Do not consolidate cases with materially different setup, outcome, boundary, or diagnostic value. Fewer tests are better only when no unique defect detection is lost. diff --git a/skills/tdd/references/TEST-DOUBLES.md b/skills/tdd/references/TEST-DOUBLES.md index 3c90b73..bb899bc 100644 --- a/skills/tdd/references/TEST-DOUBLES.md +++ b/skills/tdd/references/TEST-DOUBLES.md @@ -1,35 +1,29 @@ # Test Doubles -Use this guide before replacing a collaborator or asserting calls. Keep collaborators inside the system real; substitute a boundary when the real dependency is slow, unavailable, non-deterministic, externally mutating, or otherwise uncontrollable in the test. +Keep collaborators inside the system real. Substitute a boundary only when the real dependency is slow, unavailable, destructive, non-deterministic, externally mutating, or otherwise uncontrollable in this test. ## Choose the least powerful double -Start with the real collaborator, then introduce only the capability the test needs: - -| Double | Use it to | Verification | +| Double | Supply | Verify | | --- | --- | --- | -| Dummy | Fill an unused required parameter | None | -| Stub | Supply a controlled indirect input | Assert the public outcome | -| Fake | Run a working, simplified boundary implementation | Assert the public outcome or recorded public effect | -| Spy | Record an otherwise invisible boundary effect | Inspect only promised boundary facts | -| Mock | Specify a required external interaction protocol | Verify only contractually material calls | - -Prefer state verification: act through the selected observable boundary, then inspect its result or a recorded boundary effect. Use interaction verification when the interaction is itself observable behaviour, such as one idempotency-keyed payment request or committing only after a durable write. - -## Keep the contract visible +| Dummy | An unused required value | Nothing | +| Stub | A controlled indirect input | The public outcome | +| Fake | A small working boundary implementation | The public outcome or recorded public effect | +| Spy | A record of an otherwise invisible boundary effect | Only promised boundary facts | +| Mock | An expected external interaction protocol | Only contractually material calls | -Specify the external fact that matters and leave the internal route free to change. +Prefer state verification. Interaction verification is warranted when the interaction is the promise: for example, one idempotency-keyed payment request or publishing only after a durable write. ```typescript -// Couples the test to internal delegation. -expect(pricingService.lookup).toHaveBeenCalledTimes(1); -expect(discountCalculator.apply).toHaveBeenCalledBefore(taxCalculator.apply); +// Brittle: replaces internal policy and specifies its call graph. +expect(pricing.lookup).toHaveBeenCalledTimes(1); +expect(discounts.apply).toHaveBeenCalledBefore(tax.apply); -// Observes the public result with real internal collaborators. +// Durable: real internal collaborators produce the supported result. expect(await quoteOrder(order)).toEqual({ total: 108, currency: 'GBP' }); ``` -At an uncontrollable boundary, record the promised effect without specifying internal calls: +At an uncontrollable external boundary, record only the promised effect: ```typescript const mailer = new RecordingMailer(); @@ -38,11 +32,9 @@ await registerUser({ email: 'ada@example.com' }, { mailer }); expect(mailer.sent).toEqual([{ to: 'ada@example.com', template: 'welcome' }]); ``` -Use an expectation mock when the external protocol is the outcome: +Use a mock when interaction is the provider-facing contract: ```typescript -await submitPayment(order, paymentGateway); - expect(paymentGateway.charge).toHaveBeenCalledOnceWith({ amount: 108, currency: 'GBP', @@ -50,10 +42,22 @@ expect(paymentGateway.charge).toHaveBeenCalledOnceWith({ }); ``` -Here the amount, currency, single request, and idempotency key are provider-facing promises. Do not add expectations for logging, helper calls, object construction, or other internal routing. +When sequencing across system boundaries is promised, record those boundary events and assert their public order: -## Preserve boundary fidelity +```typescript +const events: string[] = []; +const payments = new RecordingGateway({ events, result: { paymentId: 'p-1' } }); +const orders = new RecordingOrders({ events, initial: pendingOrder }); -A double can make an impossible system look correct. Keep its behaviour smaller than the production boundary and derive responses from the provider contract rather than copied client logic. Where feasible, run focused contract tests against the real boundary to confirm that the fake, stub, or recorded request still matches it. Otherwise report the unverified fidelity as a limitation. +await checkout(pendingOrder.id, { payments, orders }); + +expect(events).toEqual(['charge-succeeded', 'order-saved']); +``` + +This protects a payment-before-persistence promise. Ordering between pricing helpers, mappers, or other internal collaborators remains hidden structure. + +Do not assert logging, helper calls, object construction, or internal order unless those facts are externally promised. + +## Preserve boundary fidelity -Control time and randomness by injecting a clock or deterministic source at the system boundary. Prefer a real test database or filesystem in an isolated disposable environment when its semantics are material; use a fake only when its behavioural differences cannot invalidate the test's claim. +Derive double responses from the external contract, not copied client logic. Keep the double smaller than the real boundary. Prefer a real database or filesystem in an isolated disposable environment when its semantics matter. When feasible, use caller-selected contract evidence against the real boundary to check that a fake, stub, or recorded request remains compatible; otherwise report the untested fidelity. diff --git a/skills/tdd/references/TEST-QUALITY.md b/skills/tdd/references/TEST-QUALITY.md deleted file mode 100644 index 7657103..0000000 --- a/skills/tdd/references/TEST-QUALITY.md +++ /dev/null @@ -1,84 +0,0 @@ -# Test Quality - -Use this guide when choosing an observable boundary or oracle, or when an assertion may couple the test to representation rather than promised behaviour. - -## Choose the observable boundary - -Test the narrowest boundary that satisfies all three conditions: - -- **Observable:** it exposes the requested return value, public state, error, emitted effect, persistence, navigation, or user-perceivable result. -- **Stable:** it hides algorithms, collaborator graphs, storage layouts, rendering wrappers, generated selectors, and other decisions that may change while behaviour remains fixed. -- **Predictive:** exercising it provides credible evidence that the behaviour will work in production. - -Use a coarser companion test only when the narrow boundary cannot predict a material integration outcome. A substitutable private hook is not a suitable assertion boundary merely because it is convenient to replace. - -## Construct the oracle - -Derive the expected outcome before implementing Green. Prefer, in order: - -1. A requirement, accepted example, published protocol, or explicit domain decision. -2. A law, invariant, or contract independent of the production algorithm. -3. A trusted external reference or separately implemented model. -4. Prior accepted behaviour when preserving that behaviour is the requirement. - -Encode the oracle as an independently reasoned literal or predicate. A small calculation is suitable when it expresses a different trusted rule; reusing the production helper or repeating its algorithm can only reproduce the same defect. - -```typescript -// Repeats the implementation's likely algorithm. -const expected = items.reduce((sum, item) => sum + item.price, 0); -expect(calculateTotal(items)).toBe(expected); - -// Uses the accepted worked example as an independent oracle. -expect(calculateTotal([{ price: 10 }, { price: 5 }])).toBe(15); -``` - -When no credible oracle exists, return the missing behavioural decision instead of inventing the expected result. Treat output captured from the current implementation as characterization, not proof of correctness. - -## Structure one behaviour - -Keep the test's phases explicit: - -1. **Arrange** only the state and collaborators needed for the behaviour. -2. **Act** once through the selected observable boundary. -3. **Assert** the complete promised outcome and material unchanged state. -4. **Clean up** resources or external state acquired by the test. - -One behaviour may require several cohesive assertions. Split a test when it contains multiple independent Acts or when its name cannot state one behavioural rule. - -## Challenge the test's quality - -Before retaining a test, name one plausible promised-behaviour defect that it must detect and one hidden structural change that it must survive. Require deterministic, readable assertions and credible evidence for production behaviour. Treat speed, isolation, and fidelity as trade-offs: a slower test can earn its cost through prediction, while a faster test still needs credible production evidence. - -Use two counterfactuals: - -- If promised behaviour changed, would the test result change? -- If only hidden structure changed, would the test and result remain unchanged? - -A useful test answers yes to both. Move a structure-sensitive observation outward to the supported result; move a behaviour-insensitive assertion toward the actual promise. - -## Observe promised outcomes - -| Concern | Durable observation | Replace structure coupling with | Representation is valid when | -| --- | --- | --- | --- | -| Domain result | Public operation against an independently justified example or invariant | A literal, contract, or independently reasoned predicate instead of the production algorithm | The representation is part of the domain contract | -| Public state | Supported command followed by a documented query or return value | Public state instead of private fields, methods, or patched internals | Diagnostic state is an intentional supported interface | -| Error | Caller-visible type, code, material message, and promised recovery or unchanged state | Stable error meaning instead of a private branch, stack shape, or helper call | Exact wording is a documented user or API contract | -| UI | Role, label, text, displayed value, accessible state, focus, or navigation | User-perceivable outcome instead of classes, wrapper tags, child indexes, or DOM depth | Visual appearance or semantic markup is the promised behaviour | -| Persistence | Write and read through supported interfaces, with reload or restart when durability matters | Public retrieval instead of private tables, columns, ORM calls, or storage layout | The schema is a published integration contract | -| External effect | Recorded boundary request or fake mailbox, queue, or provider outcome | Promised payload and result instead of internal delegation | Count, order, or arguments are part of the external protocol | -| Snapshot or serialization | Small reviewed public artifact or selected semantic fields | Focused compatibility facts instead of broad structural snapshots | Exact bytes, markup, or object shape are the published format | -| Configuration | Configured system exercised through its observable effect | Runtime result instead of parser calls or incidental internal objects | Generated configuration text is itself the public artifact | - -For UI behaviour, interact as a user would and assert what a user or assistive technology can perceive: - -```typescript -// Couples the test to generated styling and DOM structure. -expect(button.className).toBe('btn btn-primary px-4'); -expect(container.children[0]).toBe(button); - -// Observes the supported interaction and visible result. -await user.click(screen.getByRole('button', { name: 'Save' })); -expect(screen.getByText('Changes saved')).toBeVisible(); -``` - -For side effects, observe the system boundary rather than the internal route. Assert call count or order only when duplicate suppression, transaction ordering, or protocol sequencing is the behaviour. See [Test Doubles](TEST-DOUBLES.md) for boundary substitutes and interaction assertions. From adc4cee846ae2ab8dd42e2bae0066b3bcb66eee5 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 11:26:57 +0100 Subject: [PATCH 62/79] feat(quality-harnesses): teach risk-led quality selection --- SKILLS_PLAN.md | 2 +- skills/quality-harnesses/SKILL.md | 68 ++++++++++++++++++ skills/quality-harnesses/agents/openai.yaml | 5 ++ .../references/BEHAVIOURAL-EVIDENCE.md | 39 ++++++++++ .../references/BUILD-STATIC-STRUCTURE.md | 31 ++++++++ .../quality-harnesses/references/CATALOGUE.md | 71 +++++++++++++++++++ .../references/CONTRACTS-DATA.md | 21 ++++++ .../references/HUMAN-SPECIALIST.md | 19 +++++ .../references/RUNTIME-QUALITIES.md | 25 +++++++ .../references/SECURITY-ADVERSARIAL.md | 19 +++++ 10 files changed, 299 insertions(+), 1 deletion(-) create mode 100644 skills/quality-harnesses/SKILL.md create mode 100644 skills/quality-harnesses/agents/openai.yaml create mode 100644 skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md create mode 100644 skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md create mode 100644 skills/quality-harnesses/references/CATALOGUE.md create mode 100644 skills/quality-harnesses/references/CONTRACTS-DATA.md create mode 100644 skills/quality-harnesses/references/HUMAN-SPECIALIST.md create mode 100644 skills/quality-harnesses/references/RUNTIME-QUALITIES.md create mode 100644 skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index c8a743c..a249f25 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -356,7 +356,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `maintain-decision-records` skill - [x] Update `modular-design` skill - [x] Update `tdd` skill -- [-] Create `quality-harnesses` skill +- [x] Create `quality-harnesses` skill - [-] Update `code-review` skill - [x] Update `maintain-agents` skill - [x] Update `define-product` skill diff --git a/skills/quality-harnesses/SKILL.md b/skills/quality-harnesses/SKILL.md new file mode 100644 index 0000000..005f9fd --- /dev/null +++ b/skills/quality-harnesses/SKILL.md @@ -0,0 +1,68 @@ +--- +name: quality-harnesses +description: Teaches risk-triggered selection and honest interpretation of software quality harnesses for a fixed change. Use when implementing or assessing code, tests, configuration, migrations, generated artefacts, or other software changes. +metadata: + invocation: model +disable-model-invocation: false +--- + +# Quality Harnesses + +Teaches a caller to apply every repository-required and changed-risk harness that can provide faithful evidence without treating green tools, metrics, or control count as proof. + +## Process + +### 1. Fix the evidence boundary + +Read the caller's outcome, exact candidate or change scope, intended behaviour and other authorities, repository instructions, changed implementation and tests, available quality infrastructure, and material environment constraints. Identify: + +- behaviour, contracts, data, dependencies, generated output, configuration, and operational qualities changed or relied upon; +- repository-required checks and their actual scopes; +- material failure modes introduced or altered; and +- measurement-path changes, including tests, assertions, snapshots, fixtures, generators, seeds, selectors, exclusions, suppressions, thresholds, baselines, scripts, tool configuration, and cached or generated evidence. + +Treat earlier evidence as stale when its candidate, authority, harness, configuration, inputs, environment, or measurement path has materially changed. The fixed evidence boundary and every plausible changed risk are explicit. + +### 2. Select through the complete catalogue + +Read the complete [Harness Catalogue](references/CATALOGUE.md) on every invocation. For each entry, compare its trigger with the fixed boundary. Select an entry when repository authority requires it or the change exposes its named risk; neither familiarity nor low cost is required. Reject entries with no trigger even when their tooling exists. + +Load the linked detail for every selected entry and only selected entries. When entries share one reference, read it once and apply each selected section. Preserve repository-owned commands, versions, matrices, thresholds, and procedures; generic guidance does not replace them. + +The catalogue routes selected entries to [Build, Static, and Structural Harnesses](references/BUILD-STATIC-STRUCTURE.md), [Behavioural Evidence](references/BEHAVIOURAL-EVIDENCE.md), [Security and Adversarial Harnesses](references/SECURITY-ADVERSARIAL.md), [Contracts and Data](references/CONTRACTS-DATA.md), [Runtime Qualities](references/RUNTIME-QUALITIES.md), and [Human and Specialist Evidence](references/HUMAN-SPECIALIST.md). Do not load these references until their catalogue trigger applies. + +### 3. Apply the selected knowledge + +Apply the loaded guidance inside the caller's workflow. The caller owns commands, edits, remediation, evidence, and stopping. Where dependencies allow, use fast focused feedback before broader or specialist evidence, but never substitute a cheaper harness for the faithful one a claim requires. + +Selection precedes feasibility. Keep every triggered entry selected even when its existing harness is unreliable, unrepresentative, unsafe, unavailable, or outside the execution boundary; record the result as inconclusive or not run and leave the claim unassessed. Do not relabel an applicable entry as optional or irrelevant merely because faithful evidence is expensive or missing. + +Before trusting a result, confirm that the intended target, cases, rules, inputs, and environments actually ran and that the oracle can distinguish the material failure. Treat empty selectors, tautological expectations, stale baselines, unexplained exclusions or suppressions, unrepresentative inputs, hidden retries, and changed thresholds as measurement defects rather than clean evidence. + +Use existing infrastructure. Add or improve behavioural tests through the existing suite when the implementation requires them and the caller has authority. Do not install or establish project-wide harness infrastructure unless it is part of the agreed work. Do not weaken or skip a repository-required check silently. When an applicable harness is absent, unsafe, unavailable, or outside scope, leave the claim unassessed and record the residual risk. + +### 4. Preserve claim-level evidence + +For each selected harness, retain: + +- **Claim:** the exact behaviour, repository rule, contract, invariant, or risk assessed; +- **Provenance:** why the harness applies and the repository command or reproducible procedure used; +- **Boundary:** candidate, environment, configuration, fixtures or data, seed, tool version, and external-state constraints where material; +- **Result:** pass, fail, flaky or inconclusive, or not run, with salient output and artefact location; +- **Coverage:** the cases, paths, rules, versions, platforms, states, or criteria actually exercised; +- **Limitation:** what the result cannot establish and any measurement weakness; and +- **Residual risk:** the consequence still unsupported and the in-scope correction, later work, or user decision needed for stronger evidence. + +Record a field only when it materially changes interpretation or reproduction for that harness; omit immaterial metadata. Use bounded claims such as “the selected examples passed in this environment.” Do not promote compilation, execution, coverage, mutation, static analysis, or a collection of green controls into proof of correctness, a confidence percentage, or an overall verification verdict. + +### 5. Return control + +Return the selected and rejected catalogue entries with their triggers, applicable commands or procedures, claim-level evidence, unavailable evidence, and residual risks. End by stating that execution, evidence capture, remediation, and completion remain with the caller. Stop after the guidance has been applied within the caller's current boundary. Do not take ownership of implementation, independent review, commit, publication, release, deployment, or the caller's completion decision. + +## Rules + +- Treat complexity, cognitive complexity, size, churn, coupling, coverage, mutation, and smell counts as evidence for a named question, never as universal quality gates or findings without a concrete mechanism and consequence. +- Preserve flaky or conflicting results as unreliable evidence; never turn them into a pass through unrecorded retries, averaging, suppression, or selective omission. +- Prefer several independent evidence routes over repeated controls that share the same oracle, configuration, implementation assumption, or failure mode. +- Never state or imply that the candidate is ready, not ready, approved, rejected, verified, unverified, blocked, or clear to continue. Report claim-level evidence and return the completion decision to the caller. +- Return decisions to the caller or an authority already named by the caller or project. Do not invent owners, reviewers, specialists, approval bodies, or coordination. diff --git a/skills/quality-harnesses/agents/openai.yaml b/skills/quality-harnesses/agents/openai.yaml new file mode 100644 index 0000000..f869ab1 --- /dev/null +++ b/skills/quality-harnesses/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Quality Harnesses' + short_description: 'Select risk-matched software quality evidence' +policy: + allow_implicit_invocation: true diff --git a/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md b/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md new file mode 100644 index 0000000..c4c20f1 --- /dev/null +++ b/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md @@ -0,0 +1,39 @@ +# Behavioural Evidence + +## Behavioural and regression suites + +Map each changed observable outcome to an oracle capable of disagreeing with the implementation. When `$tdd` applies, use its public result for Red/Green sensitivity, durable boundary, and oracle evidence rather than recreating its workflow here. Treat captured current output as characterization until an independent authority accepts it. + +Run the narrowest faithful focused target while changing behaviour, then relevant nearby tests and every repository-required broader suite. Distinguish: + +- new-behaviour or defect evidence from focused changed tests; +- integration evidence that retains the production technology or boundary at risk; and +- pass-to-pass regression evidence from previously passing tests. + +Confirm discovery and selected-case counts, intended failure sensitivity, complete assertions, negative or unchanged outcomes, fixture representativeness, cleanup, deterministic time and randomness, and meaningful boundary interactions. Inspect changed snapshots, fixtures, doubles, skips, quarantines, retries, timeouts, and test configuration. A shared implementation helper, regenerated snapshot, “does not throw,” empty selector, or incidental call assertion is not a faithful oracle unless that exact representation or interaction is the contract. + +Preserve the exact command, result, case count, environment, and failing counterexample. Report flaky evidence as flaky or inconclusive; diagnose or rerun only through an explicit recorded procedure, never until green. + +## Coverage + +Use coverage to locate changed or risk-relevant statements, branches, conditions, states, and error paths that the selected tests never execute. Start from the claim and ask whether an uncovered path can alter it. Add a test only when an independent oracle and credible seam can make that path observable; executed code without a discriminating assertion is not stronger evidence. + +Inspect the measured denominator, included packages, changed-code mapping, branch versus line criteria, exclusions, generated code, merged reports, stale artefacts, and coverage configuration changes. Preserve repository-owned thresholds, but do not invent a target or raise execution merely to improve a number. + +Representative contrast: use a report to find that the duplicate-conflict branch never ran, then exercise its externally visible settlement result. Do not add a test that calls the branch and asserts only that it returned. + +## Mutation + +Use existing mutation tooling only when the bounded behaviour and suite are stable, test sensitivity remains materially uncertain, and execution is safe and economical. Target the changed or risk-bearing area rather than an unrelated whole repository unless authority requires otherwise. + +Interpret each category: + +- **Killed:** the suite detected that particular non-equivalent injected change. +- **Survived:** determine whether the mutant represents a plausible promised-behaviour defect; strengthen a faithful oracle only when it does. +- **Uncovered:** use coverage and claim tracing to decide whether the location affects material behaviour. +- **Equivalent or irrelevant:** retain a concrete justification; do not count it as detected. +- **Timed out or errored:** record unreliable or missing evidence and diagnose when in scope. + +Inspect operator selection, changed-code mapping, exclusions, baseline failures, timeouts, retries, mutant filters, and whether a proposed test merely couples to implementation. Never use the aggregate score as correctness, a universal threshold, or permission to add low-value assertions. + +Representative contrast: a surviving change that removes the uniqueness guard requires a test of duplicate settlement behaviour; a mutant that changes an unreachable diagnostic string may be irrelevant to the fixed claim. diff --git a/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md b/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md new file mode 100644 index 0000000..4477548 --- /dev/null +++ b/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md @@ -0,0 +1,31 @@ +# Build, Static, and Structural Harnesses + +## Build, compilation, and packaging + +Use the repository-owned command, target, production mode, and dependency state. Record the source revision, environment, toolchain, inputs, output artefact, exit status, and material warnings. Confirm that the command rebuilt the intended target rather than accepting stale or cached output; inspect changed scripts, lockfiles, generators, ignored failures, omitted workspaces, and warning policy. + +A clean result supports only that the recorded inputs compiled or assembled in that environment. It does not execute promised behaviour or establish portability, compatibility, deployment, or reproducibility. When packaging changes, inspect the produced archive or manifest for required files, entry points, permissions, metadata, and absence of unintended material rather than relying on exit status alone. + +Representative procedure: resolve the authoritative build command; clear or invalidate only project-sanctioned disposable output when needed; run the exact target; verify a fresh artefact and its manifest; preserve salient output and environment. If cleanup is unsafe or unauthorised, record the cache limitation. + +## Types, linting, formatting, and framework diagnostics + +Resolve the configured check command and its real file, project, rule, and mode scope. Run repository-required checks even when a narrower invocation already passed. For changed configuration, compare selection, rule severity, exclusions, suppressions, baselines, generated-code treatment, and formatter check/write mode before interpreting the result. + +A clean type check supports only relationships encoded by that type system and configuration. A linter or framework diagnostic supports only enabled rules. Formatting supports only the selected canonical representation. Confirm a non-empty target and preserve warnings or diagnostics that the command reports without failing. + +Do not add suppressions, widen ignores, lower severity, regenerate a baseline, or format unrelated files merely to make a check green. An authorised exception remains an explicit limitation with its scope and rationale. + +## Architecture fitness rules + +First identify the authoritative structural decision the rule encodes. Run the existing rule only when the change can affect its dependency, layer, cycle, import, ownership, or public-surface selection. Confirm that source selectors, module mappings, and expected violations are non-empty and still match renamed or moved code. Inspect changed exemptions, baseline files, generated-source treatment, and rule configuration. + +A pass supports conformance to the encoded rule for the selected source. It cannot decide whether the rule is desirable, whether omitted boundaries are healthy, or whether runtime, data, semantic, or organisational coupling is acceptable. Use `$modular-design` for structural judgement when its public trigger applies; do not turn a new architectural preference into a gate without project authority. + +Representative contrast: a dependency rule that scans 86 current modules and finds no forbidden imports is evidence; the same green command after its package selector matches zero modules is not. + +## Reproducibility and provenance + +Identify the promised equivalence: byte-identical artefacts, canonical content, generated-source stability, dependency provenance, or signed-output traceability. Record source revision, toolchain, resolved dependencies, environment, locale, time inputs, network access, generation command, and comparison method. Repeat under the relevant independent variance rather than only rerunning against the same caches and machine state. + +Inspect embedded timestamps and paths, undeclared state, caches, dependency drift, ignored comparison fields, signatures, attestations, and whether regenerated output was reviewed and checked into the intended boundary. A successful same-machine repeat is useful determinism evidence but not cross-environment reproducibility. Equivalent artefacts do not establish behavioural correctness or trustworthy provenance by themselves. diff --git a/skills/quality-harnesses/references/CATALOGUE.md b/skills/quality-harnesses/references/CATALOGUE.md new file mode 100644 index 0000000..b784274 --- /dev/null +++ b/skills/quality-harnesses/references/CATALOGUE.md @@ -0,0 +1,71 @@ +# Harness Catalogue + +Review every entry against the fixed change. Applicability comes from repository authority or the stated trigger, not from a baseline/optional tier. + +## Build, compilation, and packaging + +**Trigger:** A repository-required build, or changed build inputs, generated output, dependencies, packaging, compilation mode, or runtime compilation. **Narrow claim:** The selected inputs produced the identified artefact or compiled successfully in the recorded environment. **Limitations:** One build does not establish behavioural correctness, portability, reproducibility, deployability, or runtime compatibility. **Anti-gaming checks:** Confirm the intended target and configuration ran; inspect stale or cached output, ignored errors, warnings-as-errors policy, omitted packages, and changed build scripts. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#build-compilation-and-packaging). + +## Types, linting, formatting, and framework diagnostics + +**Trigger:** Repository-required diagnostics, or changed code or configuration within an existing tool's scope. **Narrow claim:** The checked files satisfy the enabled rules under the recorded configuration. **Limitations:** Clean diagnostics establish only encoded properties and cannot prove behaviour, security, accessibility, or maintainability. **Anti-gaming checks:** Confirm file and rule selection is non-empty and current; inspect changed ignores, baselines, suppressions, severity, generated-file treatment, and formatter write-versus-check mode. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#types-linting-formatting-and-framework-diagnostics). + +## Behavioural and regression suites + +**Trigger:** Changed observable behaviour, a defect repair, changed tests or test infrastructure, or a repository-required suite. **Narrow claim:** The exercised examples produced their independently specified outcomes; a broader unchanged suite also supplies pass-to-pass regression evidence. **Limitations:** Tests say nothing about unexercised cases, environments, or risks and are only as sound as their seams and oracles. **Anti-gaming checks:** Inspect test discovery, focused-case counts, assertions, snapshots, fixtures, doubles, retries, skips, quarantines, and whether expected values derive from production. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#behavioural-and-regression-suites). + +## Coverage + +**Trigger:** Execution of changed or risk-relevant paths is uncertain, coverage configuration changed, or repository authority requires the measurement. **Narrow claim:** The selected tests executed the reported code elements. **Limitations:** Execution percentage does not establish correct assertions, fault detection, requirement coverage, or overall quality. **Anti-gaming checks:** Inspect denominator and scope changes, exclusions, generated code, branch versus line criteria, merged reports, stale artefacts, and tests that execute without discriminating outcomes. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#coverage). + +## Mutation + +**Trigger:** Test sensitivity is materially uncertain and existing mutation tooling can assess a bounded changed area economically, or repository authority requires it. **Narrow claim:** Killed non-equivalent mutants show sensitivity to those particular injected changes. **Limitations:** Operators omit fault classes; survivors may be equivalent or irrelevant; timeouts and uncovered mutants are distinct; no score proves correctness. **Anti-gaming checks:** Inspect changed-code selection, mutant categories, exclusions, operator configuration, baseline stability, timeouts, and whether fixes merely assert implementation details. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#mutation). + +## Static and security analysis + +**Trigger:** Configured analysis covers the changed code, repository authority requires it, or the change affects trust boundaries, permissions, sensitive data, untrusted input, cryptography, secrets, or dependency risk. **Narrow claim:** The enabled rules reported no unresolved findings for their scanned inputs, or produced findings for appraisal. **Limitations:** Analysis approximates selected weakness classes and cannot establish absence of vulnerabilities or sound security design. **Anti-gaming checks:** Record tool and rule versions; inspect scan scope, baselines, suppressions, severity changes, ignored paths, generated findings, and dependency database age. **Reference:** [Security and Adversarial Harnesses](SECURITY-ADVERSARIAL.md#static-and-security-analysis). + +## Architecture fitness rules + +**Trigger:** An authoritative dependency, layer, cycle, import, ownership, or public-surface rule is already encoded and the change can affect it. **Narrow claim:** The selected source conforms to the encoded structural rule. **Limitations:** Passing cannot establish that the rule is a good architecture decision or cover omitted runtime, data, or organisational coupling. **Anti-gaming checks:** Confirm selectors and mappings match non-empty current source; inspect exemptions, generated code, renamed modules, baseline changes, and rules disabled or weakened by the change. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#architecture-fitness-rules). + +## Contracts and compatibility + +**Trigger:** Changed or relied-upon APIs, events, schemas, serialization, storage formats, runtimes, browsers, databases, protocols, generated clients, or supported versions. **Narrow claim:** Named producers, consumers, examples, or matrix cells conform to the exercised contract. **Limitations:** Finite examples and environments cannot establish every consumer interpretation, historical payload, or supported combination. **Anti-gaming checks:** Confirm consumer and version inventory, matrix cells, backward/forward direction, optional and unknown fields, real serialization, stale examples, normalisation, and provider/consumer independence. **Reference:** [Contracts and Data](CONTRACTS-DATA.md#contracts-and-compatibility). + +## Fuzzing and adversarial input + +**Trigger:** Parsers, protocols, decoders, validators, or hostile/high-volume inputs changed, and an existing harness has meaningful crash, sanitizer, assertion, or invariant oracles. **Narrow claim:** No selected oracle failed for the exercised corpus and generated inputs, or a preserved counterexample exposed a defect. **Limitations:** Execution alone rarely establishes semantic correctness; generators and corpora sample only part of the space. **Anti-gaming checks:** Preserve corpus, seed, duration, minimized failures, oracle and sanitizer configuration; confirm non-empty input selection and inspect timeouts, rejected-input bias, coverage plateaus, and swallowed failures. **Reference:** [Security and Adversarial Harnesses](SECURITY-ADVERSARIAL.md#fuzzing-and-adversarial-input). + +## Accessibility + +**Trigger:** Changed visible interface, semantics, focus, input, navigation, content, contrast, zoom, motion, or assistive-technology behaviour, or repository authority requires it. **Narrow claim:** The exercised automated rules and recorded manual interactions satisfy their named criteria in the tested environment. **Limitations:** Automation alone cannot establish conformance, usability, or behaviour across assistive technologies and user contexts. **Anti-gaming checks:** Inspect tested route and state selection, hidden or disabled elements, rule suppressions, keyboard path completeness, focus order, zoom/contrast conditions, and actual assistive-technology coverage. **Reference:** [Human and Specialist Evidence](HUMAN-SPECIALIST.md#accessibility). + +## Performance and load + +**Trigger:** Changed hot paths, algorithms, queries, I/O, payloads, caching, memory, concurrency, capacity assumptions, or an explicit latency, throughput, or resource budget. **Narrow claim:** The representative workload met or differed from the pinned budget or baseline by the reported distribution in the recorded environment. **Limitations:** Synthetic or local results may not predict production workload, tail latency, contention, cost, or capacity. **Anti-gaming checks:** Preserve workload, data scale, warm-up, repetitions, variance, environment, correctness assertions, baseline revision, outlier policy, and measurement overhead. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#performance-and-load). + +## Concurrency and ordering + +**Trigger:** Changed shared state, locks, transactions, parallelism, asynchronous cancellation, retries, queues, ordering, or idempotency. **Narrow claim:** The exercised schedules or stresses preserved the named invariants and outcomes. **Limitations:** Bounded schedules and successful stress runs cannot prove freedom from races, deadlocks, starvation, or production-specific timing defects. **Anti-gaming checks:** Confirm the intended scenario and a relevant execution count are non-empty; preserve a schedule or seed when used; assert outcomes as well as absence of crashes; inspect hidden serialization, disabled race detection, retries, timeouts, and environment fidelity. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#concurrency-and-ordering). + +## Resilience and fault injection + +**Trigger:** Changed timeouts, retries as recovery policy, failover, partial dependency failure, recovery, rollout, or operational degradation. **Narrow claim:** The injected fault occurred and the observed system met the explicit recovery or degradation hypothesis within the exercised boundary. **Limitations:** Injected scenarios cannot enumerate production combinations or establish operational readiness outside the observed environment. **Anti-gaming checks:** Verify fault delivery, safe isolation, observability, recovery oracle, cleanup, steady state before and after, blast-radius controls, and that retries do not hide failure. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#resilience-and-fault-injection). + +## Migration and data integrity + +**Trigger:** Changed schemas, transformations, backfills, imports, storage semantics, retention, mixed-version operation, rollback, or recovery promises. **Narrow claim:** Representative source states transformed or recovered while preserving the asserted invariants. **Limitations:** Fixtures cannot represent every production distribution, invalid state, scale, lock interaction, or irreversible operational consequence. **Anti-gaming checks:** Inspect fixture provenance and diversity, row and aggregate invariants, idempotence/resume, mixed versions, failure interruption, rollback or forward recovery, constraint timing, and excluded legacy states. **Reference:** [Contracts and Data](CONTRACTS-DATA.md#migration-and-data-integrity). + +## Reproducibility and provenance + +**Trigger:** Changed toolchains, dependency resolution, generated artefacts, signed output, provenance metadata, or environment-sensitive builds, or an explicit reproducibility promise. **Narrow claim:** Recorded material inputs produced equivalent artefacts under the exercised comparison and variance. **Limitations:** Same-machine repetition is not cross-environment reproducibility; equivalent bytes do not establish behaviour or supply-chain trust. **Anti-gaming checks:** Record toolchain, dependencies, source revision, environment, timestamps and comparison method; vary relevant inputs; inspect caches, undeclared network/state, embedded paths, signatures, and ignored differences. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#reproducibility-and-provenance). + +## Manual and specialist QA + +**Trigger:** Automation cannot faithfully assess a material visual, physical, usability, domain, operational, data, compliance, or contextual outcome. **Narrow claim:** The named procedure produced the recorded observation under its inputs and evaluator capability. **Limitations:** Results are bounded by procedure, environment, evaluator expertise and subjectivity, and do not generalise automatically. **Anti-gaming checks:** Preserve steps, inputs, versions, expected and actual results, captures, evaluator limitation, repetitions where needed, and deviations from the procedure. **Reference:** [Human and Specialist Evidence](HUMAN-SPECIALIST.md#manual-and-specialist-qa). + +## Diagnostic metrics + +Complexity, cognitive complexity, size, churn, coupling, and smell counts apply only when repository authority requires them or a changed hotspot needs investigation. Preserve repository thresholds. Trace the concrete control flow, ownership, dependency, or change-propagation consequence before treating a signal as actionable; a metric alone is neither a defect nor evidence that a refactor improves the change. diff --git a/skills/quality-harnesses/references/CONTRACTS-DATA.md b/skills/quality-harnesses/references/CONTRACTS-DATA.md new file mode 100644 index 0000000..8d60331 --- /dev/null +++ b/skills/quality-harnesses/references/CONTRACTS-DATA.md @@ -0,0 +1,21 @@ +# Contracts and Data + +## Contracts and compatibility + +Name the contract, producer and consumer directions, supported versions, serialization or transport, optional and unknown-field policy, and compatibility promise. Apply existing consumer/provider tests, schema checks, golden examples, matrix runs, or real round trips to every material changed or relied-upon cell. + +Confirm that examples are current, independently owned where possible, and serialized through the real boundary rather than reconstructed from the same implementation. Inspect defaulting, ordering, precision, encoding, unknown fields, errors, feature negotiation, generated clients, and changed normalisation. Record every producer, consumer, version, runtime, browser, database, or platform actually exercised and every promised cell omitted. + +A clean result supports only those named examples or matrix cells. Agreement between implementations that share code or assumptions can be jointly wrong. Unchanged public syntax does not establish unchanged semantics. + +Representative contrast: current and previous event examples passing supports those two payloads; it does not support all historical `v2` events or unknown downstream interpretations. + +## Migration and data integrity + +Identify source states, target invariants, scale and distribution assumptions, mixed-version behaviour, interruption points, and the promised rollback or forward-recovery path. Use maintained fixtures or safe representative copies containing ordinary, boundary, duplicate, null, malformed, legacy, and partially processed states that are plausible under project authority. + +For each exercised state, check row and aggregate counts, identifiers, uniqueness, referential integrity, semantic values, ordering, permissions, timestamps, and other domain invariants before and after transformation. Exercise idempotent rerun or resume, constraint timing, mixed application versions, interrupted execution, and rollback or forward recovery when promised. Record engine version, transaction mode, data scale, duration, locks, warnings, and cleanup. + +Never run destructive migration experiments against durable or production data without explicit authority and a safe boundary. Passing fixtures cannot represent every production distribution, invalid state, scale, lock interaction, or irreversible operational effect. When automation cannot faithfully appraise operational safety or domain meaning, also apply [Manual and Specialist QA](HUMAN-SPECIALIST.md#manual-and-specialist-qa); if unavailable, preserve the gap. + +Representative contrast: a deterministic backfill over ordinary and duplicate legacy fixtures is evidence for those states; absent null and malformed fixtures leave their transformation unassessed. diff --git a/skills/quality-harnesses/references/HUMAN-SPECIALIST.md b/skills/quality-harnesses/references/HUMAN-SPECIALIST.md new file mode 100644 index 0000000..e8bb83b --- /dev/null +++ b/skills/quality-harnesses/references/HUMAN-SPECIALIST.md @@ -0,0 +1,19 @@ +# Human and Specialist Evidence + +## Accessibility + +Resolve the applicable product and accessibility authority, changed routes and states, supported input modes, browsers, and assistive technologies. Run configured automation against the actual rendered state, then perform the manual interactions the change exposes: keyboard navigation, visible focus, focus order and restoration, labels and semantics, error identification, zoom and reflow, contrast, motion, and relevant screen-reader or other assistive-technology behaviour. + +Record route, state, content, viewport, zoom, theme, browser, platform, assistive technology and version, automated rules, manual steps, expected outcome, actual outcome, and captures where useful. Inspect hidden states, disabled elements, overlays, asynchronous updates, rule suppressions, partial keyboard paths, and tests against non-production markup. + +Automation supports only enabled rule checks. Manual results support only the exercised environment and evaluator capability. Neither alone establishes conformance or usability across users and technologies; unavailable required coverage remains an explicit gap. + +## Manual and specialist QA + +Use a reproducible manual procedure when automation cannot faithfully observe a material visual, physical, usability, domain, data, compliance, or contextual result. Use specialist assessment only when the claim requires knowledge the current caller and executable harnesses do not supply; do not simulate or invent that expertise. + +Define the claim, inputs, initial state, environment, steps, expected result from an independent authority, observation method, cleanup, and acceptable variation before execution. Preserve actual results, captures or measurements, versions, repetitions where variability matters, deviations, and the evaluator's relevant capability or limitation. + +A recorded observation supports only that procedure and context. Subjectivity, evaluator knowledge, unavailable equipment or data, and untested states remain limitations. When the required person, environment, service, device, data, or authority is unavailable, mark the claim unassessed and return the residual risk to the caller; do not install infrastructure, fabricate a result, or convert generic inspection into specialist approval. + +Representative contrast: a database specialist can assess lock duration and operational backfill safety from production-scale evidence; an implementation agent reading the migration can identify questions but cannot claim that specialist assessment occurred. diff --git a/skills/quality-harnesses/references/RUNTIME-QUALITIES.md b/skills/quality-harnesses/references/RUNTIME-QUALITIES.md new file mode 100644 index 0000000..e237d42 --- /dev/null +++ b/skills/quality-harnesses/references/RUNTIME-QUALITIES.md @@ -0,0 +1,25 @@ +# Runtime Qualities + +## Performance and load + +Start from an authoritative budget, SLO, pre-change baseline, or concrete regression risk. Use a representative workload, data scale, concurrency, cache state, warm-up, duration, repetitions, and environment. Assert behavioural correctness during the measurement; a fast wrong result is not evidence. + +Preserve distributions and tail values rather than one average, variance and outlier policy, resource consumption, baseline revision, tool overhead, and environmental noise. Compare like with like and rerun enough to distinguish a material effect from noise. Inspect reduced workloads, changed data, warmed caches, hidden retries, omitted errors, and thresholds moved after seeing results. + +A clean local benchmark supports only the exercised workload and environment. It does not establish production tail latency, capacity, contention, cost, or scalability beyond measured ranges. + +## Concurrency and ordering + +Name the invariant and realistic failure: duplicate effect, lost update, stale read, reordered event, deadlock, starvation, cancellation leak, retry amplification, or broken idempotency. Use existing race detection, deterministic schedule exploration, model checking, controlled interleavings, or seeded stress at the narrowest boundary that retains the shared state and transaction semantics at risk. + +Assert final outcomes, intermediate invariants where observable, and absence of duplicate or missing effects; do not rely only on no crash. Confirm that the intended scenario and its relevant execution count are non-empty. Preserve the schedule or seed and environment details only when material to interpretation or reproduction. Inspect accidental serialization, mocks that remove shared state, selectors that match zero cases, disabled detectors, swallowed task errors, and retries that mask nondeterminism. + +A bounded clean run supports only its exercised schedules and environment. Stress cannot prove race freedom. When production-like scheduling or specialist concurrency assessment is unavailable, record the remaining failure mode rather than increasing an arbitrary iteration count. + +## Resilience and fault injection + +State an explicit hypothesis: which dependency or resource fails, how the system should degrade or recover, what state must remain safe, and what users or operators observe. Use a safe isolated boundary with an existing fault mechanism and current observability. Establish steady state, inject the fault, confirm it actually occurred, assert degradation and recovery outcomes, remove the fault, and verify cleanup and restored state. + +Record fault type, timing, duration, affected calls or nodes, environment, observability signals, recovery window, data invariants, retries, and residual effects. Inspect no-op injection, fallbacks that bypass the target, hidden retries, insufficient blast-radius controls, missing recovery oracle, and cleanup failures. + +One successful experiment cannot enumerate production combinations or prove rollout readiness. Release, deployment, rollback execution, and production monitoring remain with their owning workflows. diff --git a/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md b/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md new file mode 100644 index 0000000..cf4ca65 --- /dev/null +++ b/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md @@ -0,0 +1,19 @@ +# Security and Adversarial Harnesses + +## Static and security analysis + +Map the change through trust boundaries, identities, permissions, untrusted inputs, sensitive data, secrets, cryptography, deserialization, dependencies, and external effects. Apply repository-required or already configured analysis to the code it actually covers. Add risk-focused negative or abuse-case procedures only for the changed surface and established authority. + +Record tool, rule set and vulnerability database versions, scanned inputs, build mode, findings and suppressions. Inspect changed baselines, ignored paths, generated-code treatment, severity thresholds, stale dependency data, secret-history scope, and whether the scan completed. Appraise findings against reachable code and concrete consequences; do not suppress a false positive without preserving why it is false. + +A clean scan supports only absence of findings from enabled rules over scanned inputs. It cannot establish safe design, complete data flow, correct authorization, absence of unknown weaknesses, or runtime protection. When specialist penetration or threat assessment is materially required but unavailable, record the unassessed claim and risk. + +Representative procedure: identify the changed trust-boundary claim; run existing configured analysis; exercise one authorised and one unauthorised or malformed case through the supported boundary where in scope; inspect data and effects; preserve blind spots. + +## Fuzzing and adversarial input + +Use an existing fuzz or generative harness for changed parsers, protocols, validators, decoders, or hostile/high-volume input only when it has a meaningful oracle: crash, sanitizer, invariant, differential result, resource bound, or assertion. Define the target and rejection semantics before execution. + +Preserve corpus provenance, seed, generator distribution, duration or iteration budget, sanitizer and assertion configuration, coverage or state progress, timeouts, and every minimized failing input. Confirm the selected corpus and generated-case count are non-empty. Inspect swallowed exceptions, over-filtered inputs, biased generators, duplicate cases, resource exhaustion mistaken for a product defect, and failures discarded after minimization. + +No crash establishes robustness only for exercised inputs and enabled detectors. It rarely establishes semantic correctness. A preserved counterexample becomes regression evidence only after an independent oracle identifies the correct outcome. From e83e43bce41b5a106b5ffd3f249109dd6ce8d803 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 11:41:42 +0100 Subject: [PATCH 63/79] refactor(skills): remove incident workflows and invocation metadata --- IMPLEMENTATION_SKILLS_PLAN.md | 2 +- README.md | 11 +-- SKILLS_PLAN.md | 37 ++------- skills/code-review/SKILL.md | 2 - skills/commit/SKILL.md | 2 - skills/debug/SKILL.md | 6 +- skills/define-product/SKILL.md | 2 - skills/elicit-with-context/SKILL.md | 2 - skills/elicit/SKILL.md | 2 - skills/implement/SKILL.md | 2 - skills/maintain-agents/SKILL.md | 2 - skills/modular-design/SKILL.md | 2 - skills/plan-retirement/SKILL.md | 2 - skills/pull-request/SKILL.md | 2 - skills/quality-harnesses/SKILL.md | 2 - skills/respond-to-incident/SKILL.md | 53 ------------- skills/respond-to-incident/agents/openai.yaml | 6 -- skills/review-architecture/SKILL.md | 2 - skills/review-incident/SKILL.md | 75 ------------------- skills/review-incident/agents/openai.yaml | 6 -- skills/review-pull-request/SKILL.md | 2 - skills/specify-feature/SKILL.md | 2 - skills/tdd/SKILL.md | 2 - 23 files changed, 11 insertions(+), 215 deletions(-) delete mode 100644 skills/respond-to-incident/SKILL.md delete mode 100644 skills/respond-to-incident/agents/openai.yaml delete mode 100644 skills/review-incident/SKILL.md delete mode 100644 skills/review-incident/agents/openai.yaml diff --git a/IMPLEMENTATION_SKILLS_PLAN.md b/IMPLEMENTATION_SKILLS_PLAN.md index 75d5f83..019cf1b 100644 --- a/IMPLEMENTATION_SKILLS_PLAN.md +++ b/IMPLEMENTATION_SKILLS_PLAN.md @@ -436,4 +436,4 @@ For each skill, use `$write-skill` to confirm the closed contract, establish no- ## Current cutover boundary -This planning session creates this brief, updates the lifecycle plan and removes the obsolete `verify-change` bundle. It does not rewrite the five surviving skills, `debug`, `respond-to-incident`, `review-pull-request` or the README. Those files may temporarily describe the previous composition until their scheduled authoring sessions align them with this authority. +This planning session creates this brief, updates the lifecycle plan and removes the obsolete `verify-change` bundle. It does not rewrite the five surviving skills, `debug`, `review-pull-request` or the README. Those files may temporarily describe the previous composition until their scheduled authoring sessions align them with this authority. diff --git a/README.md b/README.md index d33bdfb..235dc4c 100644 --- a/README.md +++ b/README.md @@ -2,7 +2,7 @@ # Propulsion -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, debugging, incident response, and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, debugging, and retirement. Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. @@ -39,7 +39,6 @@ Repeat the link for each selected skill. Codex and OpenCode both discover skills | Define a substantial feature | `$specify-feature` | An approved, decision-complete feature specification | | Make a small understood change | `$implement` | A minimal, reviewed, verified local change | | Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | -| Stabilise a production incident | `$respond-to-incident` | Verified mitigation or recovery and current service state | | Review another contributor's PR | `$review-pull-request` | Independent findings and verification evidence | | Remove an obsolete capability | `$plan-retirement` | An approved, safe retirement plan | @@ -72,11 +71,9 @@ The lifecycle areas are useful entry points, not mandatory phase gates. The author's normal delivery flow does not invoke `$review-pull-request`: `$implement` already includes independent code review. The separate pull-request review skill is for work submitted by others. -### Debug and Respond +### Debug - `$debug` establishes an evidence-backed root cause and, when repair is requested, produces a minimal reviewed and verified local change. -- `$respond-to-incident` coordinates evidence-led stabilisation of an active production incident and records the resulting service state. -- `$review-incident` later explains a stabilised incident without blame and creates owned follow-up work through the normal lifecycle. ### Retire @@ -106,13 +103,11 @@ Deliver a small understood change: implement → commit → pull-request ``` -Debug, respond, or retire: +Debug or retire: ```text debug → commit → pull-request -respond-to-incident → later review-incident → route follow-up work - plan-retirement → later create-tickets → normal delivery sessions ``` diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index a249f25..d2df274 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -8,7 +8,7 @@ The existing suite may be changed or replaced where necessary. Skill authoring r ## Objective -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, debugging, incident response and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, debugging and retirement. Propulsion is not a workflow platform. It does not carry a product automatically through its lifecycle, maintain its own work state or require a setup framework. A user may enter at any point where the necessary inputs already exist and leave when that session's outcome is complete. @@ -34,7 +34,7 @@ Propulsion is not a workflow platform. It does not carry a product automatically | Plan | Decompose an approved specification into implementation-ready work. | | Refine | Assess a bounded area of the current architecture and identify high-value improvements. | | Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | -| Debug and Respond | Diagnose and repair defects, stabilise active incidents and learn from them afterwards. | +| Debug | Diagnose and repair defects. | | Retire | Plan the safe removal of an obsolete capability and route its execution through normal delivery. | These areas are entry points, not mandatory phase gates. Release, deployment and continuous production monitoring are outside Propulsion's core lifecycle. @@ -83,7 +83,7 @@ Small, understood changes do not require feature documents merely to enter imple ### Bounded supporting records -Research reports, architecture reviews, incident reviews and retirement plans are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. +Research reports, architecture reviews and retirement plans are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. ### Ticket destination @@ -169,7 +169,7 @@ The detailed authoring contract for the implementation skills is maintained in [ - **Composition:** Requires `code-review`, whose Standards axis applies the implementation teaching skills; conditionally invokes `research` when external evidence is material to the review. - **Stops:** After reporting findings. It does not change the contributor's branch or publish comments, approvals or requested changes unless the user's invocation includes that outcome. -### Debug and Respond +### Debug #### `debug` @@ -179,22 +179,6 @@ The detailed authoring contract for the implementation skills is maintained in [ - **Composition:** Repair applies `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, and requires `code-review`; conditionally invokes `research` and `elicit-with-context` when their triggers apply. - **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. -#### `respond-to-incident` - -- **Outcome:** An active production incident is understood sufficiently to stabilise service safely and verify the resulting service state. -- **Inputs:** The incident signal, production evidence, affected systems, available operational controls and current communications. -- **Output:** A verified mitigation or recovery, concise timeline, current impact and explicit unresolved risks or follow-up needs in the owning incident system. -- **Composition:** Conditionally invokes `debug` and `research` when they are necessary to stabilise the incident. Any local repair uses the implementation-quality model through `debug`. -- **Stops:** When service is stable or the exact external blocker is established. It does not perform the later learning review. - -#### `review-incident` - -- **Outcome:** A stabilised incident produces a blameless, evidence-based explanation and owned follow-up work. -- **Inputs:** The incident timeline, technical and operational evidence, mitigations, communications and current product and architecture context. -- **Output:** A durable incident review and verified follow-up items routed into the normal lifecycle. -- **Composition:** Conditionally invokes `research`, `maintain-ubiquitous-language`, `maintain-decision-records` and `maintain-agents` when tracker-backed follow-up work lacks a configured task-management preference. -- **Stops:** After learning and follow-up ownership are recorded. It does not implement the resulting changes. - ### Retire #### `plan-retirement` @@ -312,14 +296,6 @@ debug → pull-request ``` -### Respond to an incident - -```text -respond-to-incident - → later review-incident - → follow-up work enters Define, Plan, Refine or Deliver -``` - ### Retire a capability ```text @@ -343,7 +319,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build the architecture-assessment path: `review-architecture`. -6. Build the exceptional paths: `debug`, `respond-to-incident`, `review-incident` and `plan-retirement`. +6. Build the remaining paths: `debug` and `plan-retirement`. 7. Align software-project elicitation through `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. ### Implementation checklist @@ -368,8 +344,6 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [-] Create `review-pull-request` skill - [-] Update `review-architecture` skill - [-] Update `debug` skill -- [-] Create `respond-to-incident` skill -- [-] Create `review-incident` skill - [-] Create `plan-retirement` skill - [x] Restore `elicit-with-context` skill - [-] Update the README @@ -408,5 +382,4 @@ Propulsion will not: - duplicate code, CI, tracker or production state in prose documents - release, deploy or continuously monitor software - merge a pull request merely because it has been created or reviewed -- turn incident response and incident learning into one session - perform retirement implementation or production decommissioning inside `plan-retirement` diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 7605eca..c78a6a2 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,8 +1,6 @@ --- name: code-review description: Diagnoses a fixed code change independently against intended behaviour and engineering standards. Use when reviewing code directly or during implementation, debugging, or pull-request review. -metadata: - invocation: model disable-model-invocation: false --- diff --git a/skills/commit/SKILL.md b/skills/commit/SKILL.md index 36f7b78..394f29e 100644 --- a/skills/commit/SKILL.md +++ b/skills/commit/SKILL.md @@ -1,8 +1,6 @@ --- name: commit description: Creates atomic Conventional Commits for eligible reviewed work while preserving unrelated changes. Use when the user asks to commit completed work. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 915cbe7..ce5d1a2 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -1,8 +1,6 @@ --- name: debug -description: Diagnoses defects and conditionally makes verified repairs. Use for failures, regressions, runtime errors, performance faults, incorrect behaviour, or causal debugging called by incident response. -metadata: - invocation: user +description: Diagnoses defects and conditionally makes verified repairs. Use for failures, regressions, runtime errors, performance faults, incorrect behaviour, or causal debugging. disable-model-invocation: true --- @@ -48,4 +46,4 @@ When verification exposes an in-scope failure, return to the causal repair path, ## Handoff -Report the requested outcome; expected and observed behaviour; authority and environment boundary; original signal; hypotheses, predictions, experiments, and contradictory evidence; root cause and causal mechanism or exact unresolved evidence boundary; reverted attempts; retained change and regression protection; review scope and finding dispositions; verification scope, commands, evidence, verdict, limitations, residual risks, and next discriminating experiment or planned work. For diagnosis-only work, state that no implementation was changed. Stop after the diagnosis or verified local repair without committing, pushing, opening a pull request, changing tracker state, owning incident mitigation or communications, performing an incident review, releasing, deploying, monitoring, or adding speculative documentation. +Report the requested outcome; expected and observed behaviour; authority and environment boundary; original signal; hypotheses, predictions, experiments, and contradictory evidence; root cause and causal mechanism or exact unresolved evidence boundary; reverted attempts; retained change and regression protection; review scope and finding dispositions; verification scope, commands, evidence, verdict, limitations, residual risks, and next discriminating experiment or planned work. For diagnosis-only work, state that no implementation was changed. Stop after the diagnosis or verified local repair without committing, pushing, opening a pull request, changing tracker state, releasing, deploying, monitoring, or adding speculative documentation. diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 51acb60..9e4651e 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -1,8 +1,6 @@ --- name: define-product description: Defines a breadth-first product and system requirements foundation. Use when externalising a new product idea or deliberately revising PRODUCT.md. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/elicit-with-context/SKILL.md b/skills/elicit-with-context/SKILL.md index 814fd85..841f651 100644 --- a/skills/elicit-with-context/SKILL.md +++ b/skills/elicit-with-context/SKILL.md @@ -1,8 +1,6 @@ --- name: elicit-with-context description: Elicits shared understanding while maintaining software-project context. Use when a software-project skill needs user-held information or decisions. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/elicit/SKILL.md b/skills/elicit/SKILL.md index 54822b3..e5c2813 100644 --- a/skills/elicit/SKILL.md +++ b/skills/elicit/SKILL.md @@ -1,8 +1,6 @@ --- name: elicit description: Elicits material information and decisions from the user one question at a time until shared understanding is confirmed. Use when a request or consuming skill depends on user-held context, clarification, or choices that cannot be safely discovered. -metadata: - invocation: model disable-model-invocation: false --- diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index c9f430e..e249ad8 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -1,8 +1,6 @@ --- name: implement description: Implements one ticket or small agreed software change with proportionate quality evidence. Use when the user asks to build confirmed work locally. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index 569044c..db21b8c 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -1,8 +1,6 @@ --- name: maintain-agents description: Maintains concise root AGENTS.md instructions. Use when creating the file or adding, refining, or assessing repository-wide agent guidance. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index 0e68a70..dedb864 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -1,8 +1,6 @@ --- name: modular-design description: Teaches structural design through information hiding, deep modules, cohesive ownership, stable interfaces, and change locality. Use when writing or assessing code, boundaries, contracts, dependencies, seams, or change propagation. -metadata: - invocation: model disable-model-invocation: false --- diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md index f93d394..13a9d5a 100644 --- a/skills/plan-retirement/SKILL.md +++ b/skills/plan-retirement/SKILL.md @@ -1,8 +1,6 @@ --- name: plan-retirement description: Plans the safe retirement of one obsolete capability. Use when deprecation, migration, disablement, removal, archival, or recovery need an approved durable plan before delivery work. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/pull-request/SKILL.md b/skills/pull-request/SKILL.md index df05e61..4e29f4a 100644 --- a/skills/pull-request/SKILL.md +++ b/skills/pull-request/SKILL.md @@ -1,8 +1,6 @@ --- name: pull-request description: Publishes the current branch as one concise verified pull request. Use when the branch is ready for external review. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/quality-harnesses/SKILL.md b/skills/quality-harnesses/SKILL.md index 005f9fd..021d4cf 100644 --- a/skills/quality-harnesses/SKILL.md +++ b/skills/quality-harnesses/SKILL.md @@ -1,8 +1,6 @@ --- name: quality-harnesses description: Teaches risk-triggered selection and honest interpretation of software quality harnesses for a fixed change. Use when implementing or assessing code, tests, configuration, migrations, generated artefacts, or other software changes. -metadata: - invocation: model disable-model-invocation: false --- diff --git a/skills/respond-to-incident/SKILL.md b/skills/respond-to-incident/SKILL.md deleted file mode 100644 index 13fc1ab..0000000 --- a/skills/respond-to-incident/SKILL.md +++ /dev/null @@ -1,53 +0,0 @@ ---- -name: respond-to-incident -description: Stabilises active production incidents through coordinated, evidence-led mitigation and verified service recovery. Use when current production impact requires an emergency response. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Respond to Incident - -An **adapted Incident Command System** scales clear command, operations, and communications ownership to the active incident so responders can reduce user impact safely and establish the resulting service state. - -## Process - -### 1. Establish the incident basis - -Inspect the request, repository guidance, current incident and communication records, operational authorities, runbooks, available tools and controls, relevant product and architecture constraints, and current production evidence. Resolve the incident identity; source-reported time; current impact and affected users, tenants, regions, systems, data, and dependencies; known recent changes; safety, privacy, security, compliance, and support constraints; current responders and owners; evidence sources and freshness; and accessible mutation boundary. Use the project's severity meaning only when it changes the response; do not invent a severity scheme or guess current production state. - -Define the signals that can distinguish continued impact, partial recovery, and credible stability before taking action. Prefer the owning incident, observability, and operational systems over copied state. When evidence is stale, contradictory, or unavailable, seek the smallest safe current observation; if no trustworthy signal or safe observation exists, keep production unchanged and establish the exact evidence or access blocker and next owner. The active impact, response authority, and evidence boundary are explicit. - -### 2. Establish proportionate command - -Map the project's current incident roles onto three responsibilities: incident command owns impact, priorities, mutation decisions, coordination, escalation, and the stopping decision; operations owns bounded technical investigation and controlled action; communications owns factual updates and the response timeline. One responder may retain every responsibility; delegate only when incident complexity or concurrent work requires it, and preserve existing project-native command and handoff authority. - -Adopt the owning incident record and established communication channel where available. Keep current impact, status, mitigation, command and action owners, next communication time, blockers, and unresolved risks current there. Maintain one concise timeline throughout: timestamp, observation or decision, action and actor, expected result, observed result, and reversal when applicable. When the owning record is inaccessible, retain a bounded session timeline and identify the exact record-update handoff rather than creating a competing repository record. The response has one visible command owner and one current factual account. - -### 3. Select the safest useful mitigation - -Prioritise reducing user impact over broad root-cause work. Start with applicable runbook or automated recovery actions and compare credible mitigation options by expected impact reduction, time, reversibility, blast radius, operational and data risk, dependencies, and signals capable of verifying direct effect and user or service health. Prefer the reversible, low-blast-radius option that is likely to stabilise service while preserving evidence. A successful command, quiet alert, or plausible diagnosis is not recovery evidence. - -Ask the incident commander or user to resolve a material trade-off when an option changes intended product behaviour, risks data loss or integrity, affects security or compliance, has a material irreversible effect, exceeds accessible authority, or requires a genuinely command-owned priority. Present the live evidence, viable options, consequences, exact target, intended effect, and available rollback or containment. Once the mitigation is agreed, the invocation supplies authority for its intrinsic mutations without a redundant application prompt; still satisfy required external confirmations and obtain any distinct operational authority the environment requires. One authorised mitigation and its expected evidence are explicit. - -### 4. Execute and appraise one controlled action - -Immediately before action, re-read the target and relevant service state, resolve its exact identity and scope, and record the expected result plus safe rollback or containment. Use least privilege and the narrowest native operational control. Preserve logs, traces, data, and unrelated user work; avoid destructive, broad, recursively targeted, glob-selected, or evidence-erasing operations. For a security or data-integrity incident, restrict exposure and preserve forensic evidence; defer destructive eradication, rebuild, or data repair until its targets, consequences, authority, and recovery path are exact. - -Record the action and actor, execute only the agreed operation, read back the mutated target, and immediately inspect both its direct effect and current user or service signals, including a credible counter-signal. When the action fails or worsens impact, perform only a pre-agreed safe rollback or containment, verify that result, preserve the contradictory evidence, and return to mitigation selection. Do not stack unsupported actions or interpret command success as service recovery. Each attempt leaves a timestamped, read-back result and either one supported current state or an explicit uncertainty. - -Invoke `$debug` only when a bounded causal diagnosis or local repair is necessary to stabilise this incident. Supply the exact signal, production boundary, time constraint, and stopping condition; retain ownership of impact, mitigation, production actions, communications, and recovery. Do not let diagnosis become open-ended or delay a safe mitigation. Invoke `$verify-change` for a frozen local code or configuration repair, or a fixed operational change that existing harnesses can appraise; its verdict supplements but never replaces current production recovery evidence. Invoke `$research` only for one material external subject that needs durable evidence. - -Do not commit, push, publish, release, or deploy a repair. When stabilisation requires one of those project-specific actions, stop that route with the exact external handoff, owner, artefact, required mechanism, and service risk rather than performing or improvising it inside Propulsion. - -### 5. Establish the service state - -After an effective action, observe the direct effect, user and service health, and counter-signals across the evidence window warranted by the system, runbook, traffic pattern, and incident risk. Do not invent an arbitrary duration. Check for partial or regional recovery, tenant or cohort skew, retries and hidden backlog, delayed processing, data damage, dependency degradation, recurrence, and changed alert or metric coverage when applicable. Distinguish mitigated impact, restored service, and remaining unknowns; silence alone does not establish stability. - -Keep the incident record and communications current as evidence changes. State observed facts, affected scope, action underway, user guidance, and next update without speculative root cause, blame, fabricated metrics, or unsupported recovery times. Service is stable only when trustworthy current signals and counter-signals support that conclusion within the stated scope and limitations. - -### 6. Conclude the active response - -Conclude with exactly one terminal state: verified mitigation or recovery with bounded ongoing observation and ownership, or an exact external blocker, escalation, or handoff. Read back the owning incident record where accessible and report current impact, stability evidence and its window, timeline, actions and reversals, communications, unresolved data or service risk, remaining observation and owner, access or authority limitations, and follow-up needs. Do not call the incident resolved merely because symptoms quieted or a mitigation executed. - -Stop before the later learning session. Do not write a postmortem or root-cause narrative, create speculative follow-up tickets, perform broad cleanup, plan a release, continuously monitor beyond bounded recovery evidence, retire a capability, or invoke `$review-incident`. The active incident response ends with a credible service state or precise handoff; later `$review-incident` owns blameless learning and follow-up routing. diff --git a/skills/respond-to-incident/agents/openai.yaml b/skills/respond-to-incident/agents/openai.yaml deleted file mode 100644 index 1ce243b..0000000 --- a/skills/respond-to-incident/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Respond to Incident' - short_description: 'Stabilise an active production incident safely' - default_prompt: 'Use $respond-to-incident to coordinate a safe mitigation and verify the resulting production service state.' -policy: - allow_implicit_invocation: false diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index 1ee21b1..d06fc04 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -1,8 +1,6 @@ --- name: review-architecture description: Assesses an explicitly bounded codebase area for a few high-value architecture improvements and reports the evidence without changing implementation. Use for architecture or maintainability reviews. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/review-incident/SKILL.md b/skills/review-incident/SKILL.md deleted file mode 100644 index 97cb0cf..0000000 --- a/skills/review-incident/SKILL.md +++ /dev/null @@ -1,75 +0,0 @@ ---- -name: review-incident -description: Reviews stabilised production incidents into durable, blameless explanations and owned follow-up work. Use when active incident response has ended. -metadata: - invocation: user -disable-model-invocation: true ---- - -# Review Incident - -Explains one stabilised incident from evidence and records the smallest useful set of owned learning outcomes without reopening active response or implementing the resulting work. - -## Prerequisites - -Require one fixed incident identity, its owning incident record, and current evidence that production is stable. Re-read the latest service state and its evidence window rather than trusting a resolved label or an earlier handoff. When impact is still active, stability is uncertain, or a bounded observation remains under incident command, stop the review and route the incident to a later `$respond-to-incident` session with the exact current signal, owner, and missing evidence. Do not operate production, change communications, or perform learning analysis as a substitute for stabilisation. - -Require enough incident evidence to distinguish what happened from what is merely believed: the timeline and its revisions, impact, detections, actions and reversals, mitigations, communications, current state, unresolved data or service risks, and relevant technical, operational, product, architecture, and context authorities. When a material gap prevents a responsible explanation, identify the evidence, source, and owner needed and stop without manufacturing a narrative. A gap that can remain explicit without distorting the review is an unknown, not a reason to invent certainty. - -## Process - -### 1. Establish the review and evidence boundary - -Inspect the request, repository guidance, owning incident system and project review convention, complete response handoff, current product and architecture context, available tools, evidence revisions, participants, owners, and access limits. Preserve the incident system as authority for live incident state and the project convention as authority for the durable review. Use a repository record only when durable learning belongs there; it must not become a mirror of volatile operational or tracker state. - -Set the incident scope, affected period, systems and populations, review participants, evidence cut-off, and the sensitivity of logs, communications, customer or personnel data, security details, and forensic material. Retain secrets, credentials, personal data, exploit detail, and other restricted evidence only in an appropriate access-controlled owner; cite a stable evidence identity or a safe summary elsewhere. When no available owner can preserve a required sensitive conclusion safely, report that exact persistence blocker instead of exposing or erasing it. - -Record each material statement as an observed fact with its source and time, an interpretation supported by named evidence, or an unknown. Preserve conflicting sources and superseded timeline entries with their provenance until the conflict is resolved; a later edit is evidence, not permission to rewrite history silently. The review has one bounded, current, and inspectable evidence base. - -### 2. Reconstruct the incident - -Build a timestamped sequence from initial conditions through impact, detection, escalation, response, mitigation, recovery, and the verified stable state. Account for who or what observed each event, the information available at that moment, decisions and actions, expected and observed results, communications, reversals, and revisions. Distinguish detection delay, response delay, recovery delay, and incomplete visibility when the evidence supports them. Do not infer an event time from record order or collapse contradictory clocks into false precision. - -Apply **local rationality** to explain why decisions and adaptations made sense to participants with the goals, pressures, signals, tools, permissions, and uncertainty available then. Compare **work-as-done** with documented or assumed work only when the difference reveals a material operating condition; neither deviation nor human error is a sufficient causal conclusion. Counter hindsight by asking what evidence would have made a different action reasonable at the time. State a counterfactual only when evidence supports both its changed condition and plausible consequence. - -Explain the interacting technical, process, organisational, tooling, environmental, and dependency conditions that shaped the outcome. Examine intended and actual defenses; what worked, failed, was missing, or created surprise; detection and escalation; communication flow; mitigation effectiveness and side effects; latent conditions; goal conflicts; previous relevant signals; and credible recurrence paths. Treat a trigger as distinct from the conditions that allowed impact and recovery. Preserve multiple contributing conditions and uncertainty rather than forcing a single root cause, stopping at individual action, or making responsibility blameless by making it vague. - -Invoke `$research` only when one material external subject needs durable evidence, then read its report before using its findings. Research does not replace incident evidence or turn a plausible industry explanation into a fact about this incident. The reconstructed account explains the incident proportionately without blame or false certainty. - -### 3. Derive proportionate learning outcomes - -Identify only follow-up outcomes that directly address an evidenced contributing condition, recurrence path, defense gap, response weakness, or material learning unknown. For each candidate, state the incident evidence and rationale, the desired outcome rather than an unsupported implementation prescription, one accountable owner, the appropriate later lifecycle route, and observable evidence that will show completion and, where meaningful, effectiveness. Distinguish reducing recurrence likelihood, reducing impact, improving detection or response, and resolving an unknown. - -Route a behaviour or product change and its selected solution to a later `$specify-feature` session; decomposition to `$create-tickets` once an approved specification exists; an evidenced defect to `$debug`; an architecture hotspot to `$review-architecture` or the normal specification path; operational runbook work to its project-native owner; repository-wide agent guidance to `$maintain-agents` only when its allocation gates hold; durable terminology to `$maintain-ubiquitous-language`; rare accepted rationale to `$maintain-decision-records`; and removal to `$plan-retirement`. Record the route without invoking downstream lifecycle work. - -Challenge candidates against recurrence value, impact reduction, evidence strength, ownership, observability, cost, and overlap with current work. Reuse or link an exact existing outcome, preserve related work as related, merge duplicates, and remove generic reminders, `be more careful` actions, speculative backlog dumping, and action inflation. Zero follow-ups is valid when the evidence supports no useful new outcome. A critical learning action without an owner remains an explicit completion blocker rather than becoming anonymous backlog work. - -When agreed follow-ups need tracker mutations, require the task-management tool named in the root `AGENTS.md`. If absent, ask the user which tool the project uses, invoke `$maintain-agents` with that confirmed preference, verify the root instruction, and resume. Inspect relevant open and closed native tracker state, schema, relationships, and read-back operations before proposing changes. When the configured tool or account is unavailable or not writable, preserve the review and report the exact tracker blocker; do not fall back to Markdown, another tracker, or a repository backlog mirror. - -The candidate set is small, evidence-traceable, owned, observable, duplicate-aware, and routed without prescribing or performing later work. - -### 4. Confirm the complete review - -Choose the owning incident system or established project convention for the durable review. When neither exists and a repository record is appropriate and safe, propose `docs/incidents/YYYYMMDD-{incident-slug}-review.md`; preserve an unrelated existing path with a numeric suffix unless replacement is explicit. Keep a repository review durable and concise: incident identity and status, impact, evidence boundary and limitations, timestamped sequence, contributing conditions, defenses and response learning, follow-up outcomes and routes, and unresolved unknowns. Link to the owning incident and tracker records rather than copying their changing state. - -Present one complete synthesis for confirmation: the factual account, supported interpretations, contradictions and unknowns, contributing conditions and recurrence paths, defenses and response learning, proposed durable owner and exact write, sensitive-evidence treatment, and every follow-up create, update, link, reuse, owner, route, and effectiveness measure. Keep all durable systems unchanged while material review conclusions or ownership remain open. - -Obtain explicit user confirmation of the complete synthesis and mutation plan. If the user rejects or revises it, reopen every affected timeline, explanation, condition, action, owner, route, sensitivity, and persistence decision and present the complete revised synthesis. Confirmation authorises exactly the agreed review and tracker writes without another application prompt. Independently confirmed durable project terminology may invoke `$maintain-ubiquitous-language` inline; terminology or rationale established by this review waits until the complete synthesis is confirmed, then invokes its respective context skill. - -The user and agent share one exact, blameless explanation and owned persistence plan. - -### 5. Persist and verify the agreed outcome - -Immediately before mutation, re-read the current incident state, material evidence revisions, target review record, matching tracker items, owners, and affected relationships. When drift makes the incident active or stability uncertain, leave the review and tracker unchanged and route back to `$respond-to-incident`. When drift materially changes the explanation, follow-up set, owner, route, or target, leave coupled mutations unchanged and return to complete-review confirmation. Continue an independent agreed write only when the drift cannot change its meaning. - -Write the confirmed review in its owning system or project location and apply exactly the confirmed native tracker creates, updates, links, owners, and routes that remain valid. Preserve unrelated record content, access controls, discussion, fields, and relationships. Use only native concepts whose meanings match the agreed plan; leave an unsupported part unapplied and report the capability limit rather than simulating it with a misleading field or document. - -Read the durable review, each affected tracker item, owner, link, and relationship back from its owning system. Verify the incident identity and stable status, evidence labels and provenance, timeline and contradictions, impact, contributing conditions, defenses, mitigation effects, limitations, follow-up rationale, ownership, lifecycle route, observable completion or effectiveness evidence, relationship meaning and direction, sensitive-data boundary, and preservation of unrelated state. Correct an exact in-scope discrepancy when the agreed value and target are unambiguous; otherwise report the mismatch and partial completion without widening authority. - -The owning systems contain the confirmed review and every safely applicable follow-up, verified against current state. - -### 6. Report and stop after learning - -Return the durable review link or path, incident and evidence boundary, stabilisation evidence, concise explanation, contradictions and unknowns, follow-up links with owners and routes, no-op or zero-action result, read-back verification, partial completion, access or capability limitations, and any unresolved critical action. Distinguish observed fact, supported interpretation, and remaining uncertainty in the handoff. - -Stop without operating production, resuming incident command or communications, implementing fixes, changing code, specifying or designing features, decomposing general delivery work, refining priorities, selecting an iteration, committing, pushing, opening or reviewing a pull request, releasing, deploying, or continuously monitoring. Later lifecycle sessions own every routed outcome. diff --git a/skills/review-incident/agents/openai.yaml b/skills/review-incident/agents/openai.yaml deleted file mode 100644 index f6ca86c..0000000 --- a/skills/review-incident/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Review Incident' - short_description: 'Review a stable incident and own its lessons' - default_prompt: 'Use $review-incident to explain this stabilised incident and record owned, evidence-led follow-up work.' -policy: - allow_implicit_invocation: false diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index 39c1e26..aa42770 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -1,8 +1,6 @@ --- name: review-pull-request description: Reviews someone else's pinned pull request independently with code findings and verification evidence. Use when assessing an external contributor's pull request. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index 1b6157f..8a012e9 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -1,8 +1,6 @@ --- name: specify-feature description: Creates a decision-complete specification for one feature. Use when a high-level feature request or capability needs collaborative definition before ticket creation. -metadata: - invocation: user disable-model-invocation: true --- diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 0265b81..0160a1a 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -1,8 +1,6 @@ --- name: tdd description: Teaches test-driven development through red-green-refactor with durable behavioural tests. Use when a usable suite can exercise changed behaviour through a credible seam. -metadata: - invocation: model disable-model-invocation: false --- From 7c0f0c902251b9415ca15e4e63e143456791415d Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 13:26:00 +0100 Subject: [PATCH 64/79] feat(code-review): refine independent review contract --- SKILLS_PLAN.md | 2 +- skills/code-review/SKILL.md | 70 +++++++++++++++------------ skills/code-review/agents/openai.yaml | 2 +- 3 files changed, 41 insertions(+), 33 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index d2df274..9e61cb7 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -333,7 +333,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `modular-design` skill - [x] Update `tdd` skill - [x] Create `quality-harnesses` skill -- [-] Update `code-review` skill +- [x] Update `code-review` skill - [x] Update `maintain-agents` skill - [x] Update `define-product` skill - [x] Create `specify-feature` skill diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index c78a6a2..d25fa6e 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,69 +1,77 @@ --- name: code-review -description: Diagnoses a fixed code change independently against intended behaviour and engineering standards. Use when reviewing code directly or during implementation, debugging, or pull-request review. +description: Reviews a fixed code change through isolated specification and engineering-standards agents. Use for direct diff review or when another delivery skill invokes independent review. +metadata: + invocation: model disable-model-invocation: false --- # Code Review -A tailored **Software Formal Inspection** independently diagnoses a fixed code change through isolated Spec and Standards reviews, returning evidence for the caller without verifying, approving, or changing the work. +Independently diagnoses one fixed code change against intended behaviour and engineering standards, returning falsifiable suggestions for its caller to adjudicate. ## Process -### 1. Freeze the change +### 1. Fix the candidate and authorities -Use a caller-supplied frozen basis when available; otherwise resolve the exact uncommitted diff, revision range, branch comparison, pull-request change, repair, or other scoped work product. Read repository instructions, resolve every revision, and capture the patch, changed paths, complete in-scope changed and untracked files, and repository state once. Ask when scope is materially ambiguous. When it is invalid or empty, return the exact blocker without reviewing. The review has one explicit, non-empty work product. +Use the caller's pinned candidate when supplied; otherwise resolve one exact non-empty diff, revision range, branch comparison, pull-request revision, repair, or local candidate. Capture its patch, changed and in-scope untracked paths, complete changed files, relevant source and tests, repository state, and exact revision identifiers where available. Ask when the target is materially ambiguous. Return the exact blocker when it is empty or cannot be fixed. -### 2. Resolve the authorities and context +Resolve intended behaviour from an implementation-independent authority such as a specification, ticket, acceptance criteria, confirmed request, contract, or another project source. When none is reasonably available, omit Spec and continue Standards without inferring intent from code, commits, or implementation rationale. -Identify intended behaviour from the caller's fixed authority, such as a specification, ticket, acceptance criteria, confirmed request, or another project source. Extract the authoritative statement without carrying surrounding conversation into inspection. When no intended-behaviour authority can be found, ask the user to confirm a Standards-only review unless the caller already supplied that confirmation. +Resolve applicable repository instructions, conventions, architecture decisions, configured checks, affected contracts and consumers, and supplied quality evidence. The review basis is explicit and sufficient to understand every changed line in context. -Independently resolve applicable repository instructions, architecture decisions, language and framework policies, configured checks, and established local conventions. Read every changed file in full, relevant tests, affected contracts and consumers, and enough surrounding implementation to understand each change. Reuse relevant caller-supplied `$modular-design` constraints, or invoke `$modular-design` when ownership, boundaries, dependencies, contracts, or change propagation present a material structural concern. +### 2. Prepare isolated review packets -Apply the universal Standards criteria directly: correctness and regression risk; scoped minimality; language, framework, and repository conventions, including established helpers; changed-test validity and durability; maintainability and economy; recognised smells; and modular architecture. Activate security, performance, accessibility, compatibility, resilience, concurrency, migration, or another specialist concern only when task evidence or project authority exposes it. Use ISO/IEC 25010 only when an authority adopts it. For an implicated supported-language or Web-security construct, consult the current official CERT rule or stable versioned OWASP ASVS requirement narrowly and include its exact applicability and exceptions. When a concrete maintainability shape still needs recognised diagnostic vocabulary, consult [Code Smells](references/CODE-SMELLS.md); a smell prompts investigation and is never finding authority by itself. The two axes have explicit, task-relevant evidence without a universal external checklist. +Give both applicable packets the fixed candidate, complete changed files and necessary surrounding source, their own authorities and criteria, the consequence ranking and finding format below, and a read-only boundary. -### 3. Prepare the inspection packets +Direct the Spec reviewer to account for every applicable requirement in code and retained tests, account for every introduced behaviour against authority, and follow affected contracts, states, data shapes, errors, effects, and consumers far enough to find missing, partial, conflicting, excessive, or regressed behaviour. Trace distinct success, failure, retry, and concurrent paths when the authority distinguishes them. When an external effect precedes durable state or acknowledgement, trace failure after the effect and before that record or response, including what a retry repeats. Report separate suggestions when a different triggering state, violated requirement, consequence, or narrow corrective outcome survives review, even if the evidence overlaps another suggestion. -Create one self-contained, operational packet per applicable axis. Both packets receive the frozen scope, complete changed files and necessary surrounding source, their applicable authorities and criteria, the priority definitions and output schema below, and the read-only boundary. Exclude conversation history, undocumented implementation rationale, `$verify-change` results or conclusions, and the other reviewer's materials. +Direct the Standards reviewer to understand every changed line in its necessary context and assess: -The Spec packet directs its reviewer to: +- correctness and regression risk; +- the smallest coherent scope and absence of speculative or superseded work; +- repository, language, and framework conventions; +- changed-test validity, independent oracles, behavioural durability, and meaningful failure detection; +- modular ownership, interface depth, dependencies, and change locality; +- repository-required and changed-risk harness selection, measurement-path integrity, and honest evidence limits; and +- recognised maintainability shapes and specialist risks exposed by the change. -1. account for every applicable statement of intended behaviour in the changed implementation and relevant tests; -2. account for every introduced or altered behaviour against an identified authority; and -3. follow affected contracts, states, data shapes, side effects, and consumers far enough to expose missing, partial, conflicting, excess, and regressed behaviour. +Require the Standards reviewer to load and apply `$modular-design` and `$quality-harnesses`, plus `$tdd` when its prerequisites hold. These skills supply teaching knowledge within the Standards review; they do not become delegated workflows or own findings, commands, remediation, or a verdict. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. -The Standards packet directs its reviewer to: +### 3. Run independent reviews -1. understand every changed line and the necessary whole-file and system context; -2. follow affected contracts, states, data shapes, side effects, callers, and dependencies far enough to expose correctness and regression risks; -3. apply the universal and triggered criteria resolved in step 2; and -4. when tests changed, ask whether they detect a promised-behaviour defect, survive behaviour-preserving changes to hidden structure, remain deterministic and readable, and credibly predict the promised result. +Give each applicable packet to a separate fresh agent and run them in parallel where possible. Do not substitute coordinator self-review. If fresh-agent execution is unavailable, mark the affected axis not performed and state the exact limitation. -Each packet tells the reviewer how to inspect its axis rather than supplying labels alone. +Each reviewer may run a safe, focused, non-mutating command only to confirm or falsify a concrete concern. It does not repeat the implementation's whole quality portfolio, edit code or tests, update snapshots or baselines, install dependencies, modify durable data, or repair a finding. -### 4. Run the independent reviews +Before retaining a suggestion, try to disprove it through contradicting authority, existing handling, repository-sanctioned exceptions, surrounding code, a focused counterexample, and the strongest benign interpretation. Omit unsupported generic advice, tooling-enforced trivia, speculative best practice, and a pre-existing issue unless the candidate introduces, worsens, or makes it newly consequential. -Give each applicable packet to a separate fresh agent, in parallel when possible and otherwise sequentially. Do not substitute the coordinating agent; when fresh-agent delegation is unavailable, mark the affected axis not performed and state the exact limitation. +Rank each surviving suggestion by consequence: -Each reviewer owns candidate discovery, code and authority validation, consequence analysis, and priority validation for its axis. Before retaining a candidate, actively try to disprove it through contradicting authority, an already-handled path, a repository-sanctioned exception, surrounding code, and the strongest benign interpretation. Run a targeted probe only when its command and boundary demonstrate that it cannot mutate the checkout, repository state, durable data, or an external system; otherwise omit it and record the limitation. Report a pre-existing issue only when the fixed change introduces or worsens it, makes it newly consequential, or cannot conform because of it. Return only findings that survive these checks. +- **High:** a credible path to materially wrong required behaviour, including a duplicated, lost, or misdirected external effect; security or privacy compromise; data loss or corruption; major production or reliability failure; or a structural, test, or evidence defect that makes the change untrustworthy. +- **Medium:** a concrete defect, regression risk, or significant maintainability, modularity, test-quality, or harness-integrity weakness with a bounded material consequence. +- **Low:** a local evidenced issue whose narrow correction has a concrete benefit. -Use consequence-based priority within each axis: `critical` for immediate data loss, security compromise, or production failure; `high` for incorrect required behaviour or major security, reliability, or maintainability risk; `medium` for a concrete defect or significant code, design, or test weakness; and `low` for a local but worthwhile issue. Priority communicates impact and order, not remediation authority. +Priority orders attention only; it does not direct remediation. -### 5. Preserve the independent results +### 4. Preserve the results and candidate -Validate only that each reviewer returned the required fields, returning an incomplete result to that same reviewer for structural completion from the original packet. Reinspect the scoped revisions, diff, paths, and in-scope untracked content. When the frozen work product drifted, preserve each affected axis's frozen result, prefix it with `Status: stale` and the changed paths and basis mismatch, and stop without silently retargeting the review. +Validate only that each result uses the required fields. Return an incomplete result to that same reviewer for structural completion from its original packet. Do not substantively re-review, merge, deduplicate, suppress, or cross-axis rerank the independent results. -Present Standards and Spec separately without substantive re-review, merging, deduplication, suppression, or cross-axis reranking. Preserve each reviewer's findings and ordering. The caller receives independent diagnostic evidence for its own adjudication. +Reinspect the fixed revisions, diff, paths, and in-scope untracked content. When the candidate drifted, preserve each affected frozen result, prefix it with `Status: stale`, state the changed paths and basis mismatch, and stop without retargeting it. ## Handoff -State the exact frozen scope, intended-behaviour authority or user-confirmed absence, Standards authorities, any omitted or unperformed axis, every probe run or omitted, and any scope drift. Return `## Standards` and `## Spec`; use `No findings.` only for a current clean axis, never for a stale clean result without its `Status: stale` prefix, and state when Spec was omitted or either axis was not performed. Format each finding as: +State the exact frozen scope, behavioural authority or reason Spec was omitted, Standards authorities, probes and limitations, unperformed axes, and scope drift. Return `## Standards` and `## Spec` separately; for an applicable current clean axis use `No findings.` and for an omitted, unperformed, or stale axis state that status explicitly. + +Format each suggestion as: ```markdown -### [priority] Concise finding +### [high|medium|low] Concise finding -- Evidence: exact code `path:line`, applicable authority or criterion, and observed fact +- Evidence: exact code location, applicable authority or criterion, and observed fact - Consequence: concrete behavioural or code-health impact +- Suggested direction: narrow outcome that addresses the concern without prescribing an unverified patch ``` -End with `## Summary` and the finding count for each completed axis, labelling an affected frozen count `stale`. The caller owns adjudication, remediation, re-review, verification, and any publication or pull-request decision. Do not repair the code, adjudicate findings, issue a verification or approval verdict, publish review comments, or alter pull-request state. +End with axis-specific counts and review scope. Return the reviewers' substance and ordering unchanged. Do not repair the candidate, adjudicate suggestions, issue a verification, completion, or approval verdict, publish comments, or alter pull-request state. The caller owns validation, adjudication, remediation, re-review, verification, and publication. diff --git a/skills/code-review/agents/openai.yaml b/skills/code-review/agents/openai.yaml index 4f4a4fa..6a11c8e 100644 --- a/skills/code-review/agents/openai.yaml +++ b/skills/code-review/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Code Review' - short_description: 'Independently diagnose a fixed code change' + short_description: 'Independently review a fixed code change' policy: allow_implicit_invocation: true From 0da6eb04a7c63f458e68583fcfabd4e00e65d64c Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 14:05:42 +0100 Subject: [PATCH 65/79] feat(implement): refine implementation workflow --- SKILLS_PLAN.md | 2 +- skills/implement/SKILL.md | 44 +++++++++++++---------------- skills/implement/agents/openai.yaml | 2 +- 3 files changed, 22 insertions(+), 26 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9e61cb7..4dfa270 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -338,7 +338,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `define-product` skill - [x] Create `specify-feature` skill - [x] Create `create-tickets` skill -- [-] Update `implement` skill +- [x] Update `implement` skill - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` - [-] Create `review-pull-request` skill diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index e249ad8..4a6d00c 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -1,48 +1,44 @@ --- name: implement -description: Implements one ticket or small agreed software change with proportionate quality evidence. Use when the user asks to build confirmed work locally. +description: Implements one ticket or small confirmed software change as a minimal, locally checked, independently reviewed candidate. Use when the user asks to build agreed work locally. disable-model-invocation: true --- # Implement -A tailored **Software Formal Inspection** lifecycle implements one confirmed body of work, independently reviews a fixed candidate, and verifies that same candidate. TDD supplies the implementation method only when its prerequisites hold. +Turns one confirmed body of software work into the smallest coherent local change with applicable quality evidence and independent review. ## Process -### 1. Fix the outcome +### 1. Fix the implementation basis -Inspect the stated ticket or agreed work, repository instructions, linked product and technical authorities, current code and tests, and the complete repository state. A clear small request is sufficient authority; do not manufacture a ticket, specification, or other ceremony. Invoke `$elicit-with-context` one question at a time only when user-held ambiguity could materially change behaviour, contracts, architecture, scope, or acceptance. State the observable outcome, exclusions, applicable structural constraints, material risks, and evidence expected before editing. The implementation basis is explicit and bounded. +Inspect the work authority, repository instructions, linked product and technical authorities, current code, tests, relevant contracts, available tools, and complete repository state. State the observable outcome, exclusions, material risks, and completion evidence. Resolve discoverable facts directly. Invoke `$elicit-with-context` only when a material user-held behaviour, contract, architecture, scope, or acceptance decision remains. -### 2. Build the smallest complete change +Keep the outcome to one implementation-ready ticket, diagnosed repair, or small confirmed change. Preserve unrelated work. Stop when the work is not sufficiently defined to implement and elicitation cannot resolve it within the requested outcome. -Follow established language, framework, and repository conventions and reuse fitting helpers. Add only the production code, tests, documentation, configuration, generated output, or other artefacts required by the outcome. Avoid speculative flexibility, unrelated cleanup, broad refactors, and generated churn. Invoke `$modular-design` when the work presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, then carry its constraints through the change. Invoke `$research` only for a material external-knowledge subject that needs durable evidence, and treat its cited report as an explicit in-scope artefact. Use `$maintain-ubiquitous-language` when independently confirmed durable terminology arises and `$maintain-decision-records` when a rare qualifying decision needs durable rationale; each utility returns control to this outcome. +### 2. Load the teaching knowledge -Invoke `$tdd` only when an existing runnable test suite can credibly exercise the requested observable behaviour at a stable boundary with an independent oracle. When evidence already shows any prerequisite is absent, keep control here rather than invoking TDD merely to receive that result. Let TDD own the Red-Green-Refactor cycles and returned code and tests when it applies. Otherwise implement through the smallest conventional path and use the strongest available project-native feedback; update existing tests only when they can meaningfully express changed behaviour. Do not install or invent a test or quality framework to satisfy this process. When meaningful infrastructure is absent, preserve the limitation for verification and recommend separately planned work rather than adding unrelated harness infrastructure. +Invoke `$modular-design` and `$quality-harnesses` against the fixed work. Invoke `$tdd` when the change affects observable behaviour and an existing usable suite can exercise it through a credible seam. Retain their applicable guidance throughout implementation; they teach within this workflow while this skill retains ownership of edits, commands, evidence, and completion. -Inspect the resulting diff and every in-scope untracked file. Account for each production addition against the agreed outcome or a necessary supporting change, and identify every change to tests, snapshots, baselines, suppressions, dependencies, quality configuration, scripts, or generated artefacts as part of the measurement path. The candidate is minimal, coherent, and ready for independent inspection. +### 3. Build and check the candidate -### 3. Review a fixed candidate +Implement the smallest coherent change that completely satisfies the confirmed outcome. Follow the selected modular constraints, TDD cycles, and quality-harness guidance without adding speculative flexibility, unrelated cleanup, or project-wide quality infrastructure outside the agreed work. -Freeze the complete candidate, including its exact diff or revisions, changed and untracked paths, repository state, intended-behaviour authority, applicable structural constraints, and named risks. Invoke `$code-review` independently on that basis. If the basis drifts or either applicable review axis is stale or unperformed, re-freeze and repeat the review when the scope and ownership remain clear; otherwise stop with the exact blocker. Do not treat implementation self-inspection, tests, or verification as a substitute for the independent review. +Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run every applicable repository-required and risk-triggered existing harness selected through `$quality-harnesses`. Preserve exact results, unavailable evidence, and residual risks. Freeze one complete candidate only after its local checks are current. -Validate every Standards and Spec finding against the confirmed request, its authorities, and current code, then give it one technical disposition: +### 4. Review and adjudicate -- A required correction is a verified violation of confirmed behaviour, the specification, a repository rule, or a necessary safety boundary, or another defect that leaves the change untrustworthy. -- A proportional improvement is valid but non-blocking; accept it only when its concrete benefit justifies the added change and complexity. -- A rejected finding is falsified, unsupported, superseded, or outside the confirmed scope. -- A user decision is required when accepting or rejecting the finding would change behaviour, contracts, architecture, or scope, or when its evidence cannot be obtained safely. +Invoke `$code-review` on the frozen candidate and its authorities. Validate every Spec and Standards suggestion against the strongest available authority and evidence, then disposition it as: -For a user-decision finding, invoke `$elicit-with-context` and stop mutation, further adjudication, and verification until its confirmed synthesis returns. Resume from the fixed outcome; when the decision changes the candidate or any review authority, freeze the resulting basis and repeat independent review. +- **Required correction:** a confirmed violation, regression, safety defect, or evidence weakness that leaves the change untrustworthy. +- **Proportionate improvement:** a valid in-scope benefit worth its added change and complexity. +- **Rejected finding:** falsified, unsupported, superseded, or outside the confirmed scope. +- **User decision:** resolution would change agreed behaviour, contracts, architecture, or scope, or needs evidence that cannot be obtained safely. -Treat priority as evidence of impact and ordering, not remediation authority. Apply every accepted correction through `$tdd` when its prerequisite applies and otherwise through the conventional path. After any candidate change, freeze the complete new candidate and repeat independent review. Continue until every finding has a supported disposition, no required finding remains unresolved, and the complete current candidate has been independently reviewed. The reviewed candidate is fixed and current. - -### 4. Verify the reviewed candidate - -Invoke `$verify-change` on the exact reviewed candidate, supplying its authorities, named risks, review dispositions, relevant implementation evidence, and identified measurement-path changes. Let it select and run every applicable repository-required and risk-triggered existing harness and return claim-level evidence, reproducible commands, limitations, residual risks, and one honest verdict. Metrics support only their named claims; they never prove quality alone. - -When verification exposes an in-scope failure, repair it through the applicable TDD or conventional path, then repeat independent review and verification on the new fixed candidate. When verification evidence is stale because source or measurement-path state changed, establish ownership and scope, then re-freeze, re-review, and re-verify; do not silently retarget old evidence. Preserve a pre-existing or out-of-scope required-gate failure as a blocker rather than broadening the work. When missing infrastructure leaves the result partially verified or unverified and no in-scope repair can supply faithful evidence, report that limitation and explicit planned work without manufacturing a harness or claiming successful completion. +Apply required corrections and accepted improvements through the applicable teaching skills. Invoke `$elicit-with-context` for a user decision. Rerun every affected harness after a candidate or measurement-path change. Freeze and review the complete new candidate after any material change. Continue until every finding has a supported disposition and no applicable correction or accepted improvement remains. ### 5. Hand off the local result -Report the agreed behaviour delivered, changed artefacts, review scope and every finding disposition, verification scope and verdict, exact commands and material results, limitations, residual risks, and any blocker or planned follow-up. Claim successful completion only when the behaviour exists locally, applicable tests and artefacts are present, the complete current candidate has a current independent review with no unresolved required finding, and verification passed within its stated scope. A partial, unverified, or failed verdict is an explicit incomplete or blocked handoff. Stop without committing, pushing, opening a pull request, releasing, deploying, changing tracker state, or authoring unrelated product, specification, or design documents. +Report the delivered behaviour, changed artefacts, applicable TDD and modular-design decisions, exact checks and results, unavailable evidence, residual risks, review scope and results, and every finding disposition. Successful completion requires the agreed behaviour locally, every production addition serving a current purpose, applicable evidence passed or explicitly unavailable, and current Spec and Standards review with no unresolved applicable change. + +Stop without committing, pushing, opening or updating a pull request, changing tracker state, releasing, or deploying. Do not conduct exploratory research, create specifications or tickets, or add unrelated quality infrastructure inside this outcome. diff --git a/skills/implement/agents/openai.yaml b/skills/implement/agents/openai.yaml index 100f67c..5421aeb 100644 --- a/skills/implement/agents/openai.yaml +++ b/skills/implement/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Implement' - short_description: 'Implement and verify one bounded change' + short_description: 'Implement and review one local change' policy: allow_implicit_invocation: false From f883f35a21ae865eb60052edda71b3156bbcb97f Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 15:08:10 +0100 Subject: [PATCH 66/79] feat(review-pull-request): finalize read-only review workflow --- SKILLS_PLAN.md | 2 +- skills/review-pull-request/SKILL.md | 52 ++++--------------- skills/review-pull-request/agents/openai.yaml | 2 +- 3 files changed, 13 insertions(+), 43 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 4dfa270..c306714 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -341,7 +341,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `implement` skill - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` -- [-] Create `review-pull-request` skill +- [x] Create `review-pull-request` skill - [-] Update `review-architecture` skill - [-] Update `debug` skill - [-] Create `plan-retirement` skill diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index aa42770..7349343 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -1,59 +1,29 @@ --- name: review-pull-request -description: Reviews someone else's pinned pull request independently with code findings and verification evidence. Use when assessing an external contributor's pull request. +description: Reviews a supplied pull request through independent specification and engineering-standards assessment. Use when the user wants a read-only pull-request report. disable-model-invocation: true --- # Review Pull Request -Coordinates `$code-review`'s tailored **Software Formal Inspection** and `$verify-change`'s **bidirectional verification traceability** against one frozen pull-request revision, then reports the independent findings and verification evidence without changing the contributor's work. +Resolves one pull request and routes its complete change through `$code-review`, then returns the independent findings without changing the pull request or repository. ## Process -### 1. Resolve the review target +### 1. Resolve the pull request -Inspect the request, repository guidance, available host tooling and authentication, pull-request metadata and discussion, repository state, and relevant product, architecture, context, ADR, specification, ticket, and acceptance authorities. Resolve one exact repository and pull-request identity. Confirm intent when the selected pull request belongs to the user or current author's branch; this skill normally reviews work submitted by someone else rather than duplicating `$implement`'s author-side review. +Resolve the supplied pull-request identifier against the current repository unless the user names another repository. Use available native host tooling to inspect its metadata, commits, complete patch and changed files, relevant surrounding code, discussion, existing check evidence, repository guidance, linked specification or work item, and applicable product and technical authorities. -Identify the intended-behaviour authority without inferring intent from the implementation. When none exists, follow `$code-review`'s Standards-only confirmation rule. The review target, authorship, authorities, and requested report-only or publication outcome are explicit. +Fix one exact, non-empty pull-request candidate with its repository, number or URL, base and head revisions, complete changed paths, and authorities. Return the exact blocker when the pull request is ambiguous, inaccessible, or incomplete. Do not infer intended behaviour from the implementation, commit messages, or implementation rationale. -### 2. Freeze the complete candidate +### 2. Invoke the independent review -Resolve and record the pull request's exact base SHA and head SHA, full patch, commits and revisions, changed paths and complete changed files, relevant current code, discussion, and check evidence. Review the whole pull request rather than its latest commit, comment, or visible excerpt. +Invoke `$code-review` with the fixed candidate, complete pull-request context, intended-behaviour authorities when available, applicable Standards authorities, existing check evidence, and a read-only boundary. When no implementation-independent behavioural authority exists, let `$code-review` omit Spec and continue Standards. -Prefer host data or a checkout already representing the pinned candidate. When execution needs another filesystem view, use an isolated disposable worktree or temporary checkout only when its boundary and ownership are explicit. Do not switch, reset, clean, or overwrite the user's working tree. Stop with the exact access or isolation requirement when no safe complete view is available. The candidate is complete, immutable for the review, and safe to inspect. +Treat existing checks as supplied evidence. Leave focused review probes to `$code-review`; do not run a separate review, verification workflow, or full project check suite. -### 3. Run the independent code review +### 3. Return the report -Invoke `$modular-design` only when the complete candidate presents a material ownership, boundary, contract, dependency, seam, or change-propagation concern that needs focused structural constraints. Supply its result to `$code-review`; modular design owns the structural assessment while code review alone owns any retained review finding. Invoke `$research` only when the assessment depends on a material external-knowledge subject that needs durable evidence and its report can be persisted without altering the contributor candidate. Supply that cited report as evidence rather than letting it decide a finding. When no safe durable-report boundary exists, state the limitation instead of changing contributor work. +State the pull-request identity and exact reviewed scope, then return `$code-review`'s separate Standards and Spec results, priorities, findings, evidence, limitations, omitted axes, and counts without substantive rewriting or adjudication. -Invoke `$code-review` with the frozen base and head SHAs, complete candidate, intended-behaviour and Standards authorities, and a read-only boundary. Require its independent Standards and applicable Spec reviews. Preserve their findings, distinctions, priorities, evidence, ordering, omitted axes, and limitations without merging, suppressing, cross-axis reranking, adjudicating, or remediating them. - -Re-read the remote base and head after the review. If either differs from its pinned SHA, retain the result as stale for its original candidate. Pin the new base and head and repeat the complete review only when the user's requested outcome still identifies the same pull request unambiguously; otherwise stop and ask for direction. The current code-review result refers to one explicit candidate. - -### 4. Verify the same candidate - -Invoke `$verify-change` on exactly the base and head SHAs reviewed in step 3. Supply the intended-behaviour and repository authorities, material risks exposed by the complete change, relevant existing CI evidence, and the isolated execution boundary. Let it select and appraise applicable existing required and risk-triggered harnesses and return claim-level evidence, exact commands or procedures, limitations, residual risk, and its verification-specific verdict. - -Use existing CI evidence only when its provenance ties it to the pinned head and its execution boundary and result are inspectable. When local verification cannot run safely or the repository lacks faithful infrastructure, retain the exact `not run`, partial, or unverified evidence; do not install tools, add tests or ad hoc harnesses, alter source, tests, snapshots, baselines, suppressions, dependencies, or checks, or treat static inspection as execution. Passing checks do not erase a code-review finding, and verification never certifies the pull request. - -Re-read the remote base and head after verification. If either differs from its pinned SHA, preserve both results as stale for their original candidate and repeat steps 2–4 on the newly pinned candidate only while the requested outcome remains clear. Stop with the stale results and exact blocker when repeated drift prevents one stable current assessment. The completed review and verification evidence refer to identical current base and head SHAs or are explicitly stale. - -### 5. Report the assessment - -Report: - -- the repository, pull request, frozen base and head SHAs, full-change scope, and any drift; -- the intended-behaviour authority or confirmed Standards-only basis, plus applicable repository and technical authorities; -- the independent Standards and Spec findings in their returned sections and order, including `No findings.` only for a genuinely current completed axis; -- the verification verdict and claim-level evidence, exact commands or procedures, execution boundaries, and material results; -- existing pull-request check evidence whose provenance was accepted; -- every omitted, unsafe, unavailable, unperformed, or stale review or verification axis; and -- limitations, residual risks, and a concise finding-count and evidence summary. - -Do not add an approval, request-changes, merge, or other disposition merely from green checks or the absence of findings. Provide a disposition only when the user's requested outcome includes it and the complete current evidence supports it. - -### 6. Publish only an included outcome - -Stop after the chat report by default. Do not publish comments, approve, request changes, merge, close, edit pull-request metadata, change tracker state, modify the contributor branch, commit, push, release, or deploy. - -When the invocation explicitly includes publishing review feedback or a host review state, present the exact supported comments and disposition for confirmation as part of resolving that outcome. After agreement, publish only that content through the repository's native host tooling without another application gate, then read back the comments and review state. Never publish an unsupported conclusion or merge, fix, or otherwise advance the pull request. +Stop after the report. Do not modify files or branches, publish comments, change pull-request or tracker state, approve, request changes, merge, commit, push, release, or deploy. diff --git a/skills/review-pull-request/agents/openai.yaml b/skills/review-pull-request/agents/openai.yaml index 66d5bd3..f2d504f 100644 --- a/skills/review-pull-request/agents/openai.yaml +++ b/skills/review-pull-request/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Review Pull Request' - short_description: 'Independently assess an external pull request' + short_description: 'Return a read-only pull-request review' policy: allow_implicit_invocation: false From 7cbeab5375af962a6ee6bee0de252bf3dd3cc3be Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 15:14:33 +0100 Subject: [PATCH 67/79] chore(skills): remove plan-retirement skill --- README.md | 11 +-- SKILLS_PLAN.md | 27 +------ skills/plan-retirement/SKILL.md | 62 ---------------- skills/plan-retirement/agents/openai.yaml | 6 -- .../assets/retirement-plan-template.md | 71 ------------------- 5 files changed, 5 insertions(+), 172 deletions(-) delete mode 100644 skills/plan-retirement/SKILL.md delete mode 100644 skills/plan-retirement/agents/openai.yaml delete mode 100644 skills/plan-retirement/assets/retirement-plan-template.md diff --git a/README.md b/README.md index 235dc4c..9346ebb 100644 --- a/README.md +++ b/README.md @@ -2,7 +2,7 @@ # Propulsion -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, debugging, and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, and debugging. Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. @@ -40,7 +40,6 @@ Repeat the link for each selected skill. Codex and OpenCode both discover skills | Make a small understood change | `$implement` | A minimal, reviewed, verified local change | | Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | | Review another contributor's PR | `$review-pull-request` | Independent findings and verification evidence | -| Remove an obsolete capability | `$plan-retirement` | An approved, safe retirement plan | ## Lifecycle skills @@ -75,10 +74,6 @@ The author's normal delivery flow does not invoke `$review-pull-request`: `$impl - `$debug` establishes an evidence-backed root cause and, when repair is requested, produces a minimal reviewed and verified local change. -### Retire - -- `$plan-retirement` defines the deprecation, migration, communication, removal, archival, recovery, and completion obligations for one obsolete capability. Execution happens in later planning and delivery sessions. - ## Representative flows Establish a product: @@ -103,12 +98,10 @@ Deliver a small understood change: implement → commit → pull-request ``` -Debug or retire: +Debug a known defect: ```text debug → commit → pull-request - -plan-retirement → later create-tickets → normal delivery sessions ``` Review another contributor's work: diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index c306714..9ffd599 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -8,7 +8,7 @@ The existing suite may be changed or replaced where necessary. Skill authoring r ## Objective -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, debugging and retirement. +Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, and debugging. Propulsion is not a workflow platform. It does not carry a product automatically through its lifecycle, maintain its own work state or require a setup framework. A user may enter at any point where the necessary inputs already exist and leave when that session's outcome is complete. @@ -35,7 +35,6 @@ Propulsion is not a workflow platform. It does not carry a product automatically | Refine | Assess a bounded area of the current architecture and identify high-value improvements. | | Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | | Debug | Diagnose and repair defects. | -| Retire | Plan the safe removal of an obsolete capability and route its execution through normal delivery. | These areas are entry points, not mandatory phase gates. Release, deployment and continuous production monitoring are outside Propulsion's core lifecycle. @@ -83,7 +82,7 @@ Small, understood changes do not require feature documents merely to enter imple ### Bounded supporting records -Research reports, architecture reviews and retirement plans are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. +Research reports and architecture reviews are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. ### Ticket destination @@ -179,16 +178,6 @@ The detailed authoring contract for the implementation skills is maintained in [ - **Composition:** Repair applies `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, and requires `code-review`; conditionally invokes `research` and `elicit-with-context` when their triggers apply. - **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. -### Retire - -#### `plan-retirement` - -- **Outcome:** An obsolete capability has an approved plan for safe retirement. -- **Inputs:** The capability, product intent, users, dependencies, interfaces, data, integrations, support obligations and current implementation evidence. -- **Output:** A retirement plan covering deprecation, migration, communication, removal, archival, access, recovery and observable completion conditions. -- **Composition:** Requires `elicit-with-context`; conditionally invokes `research`, `maintain-ubiquitous-language` and `maintain-decision-records`. -- **Stops:** After the retirement outcome and obligations are approved. Execution returns to `create-tickets`, `implement`, `commit` and `pull-request` in later sessions. - ## Reusable utility catalogue | Utility | Independently useful outcome | @@ -296,14 +285,6 @@ debug → pull-request ``` -### Retire a capability - -```text -plan-retirement - → later create-tickets - → normal implementation and publication sessions -``` - ### Review another contributor's work ```text @@ -319,7 +300,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre 3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. 4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. 5. Build the architecture-assessment path: `review-architecture`. -6. Build the remaining paths: `debug` and `plan-retirement`. +6. Build the remaining path: `debug`. 7. Align software-project elicitation through `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. ### Implementation checklist @@ -344,7 +325,6 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Create `review-pull-request` skill - [-] Update `review-architecture` skill - [-] Update `debug` skill -- [-] Create `plan-retirement` skill - [x] Restore `elicit-with-context` skill - [-] Update the README - [-] Run the suite-wide acceptance pass @@ -382,4 +362,3 @@ Propulsion will not: - duplicate code, CI, tracker or production state in prose documents - release, deploy or continuously monitor software - merge a pull request merely because it has been created or reviewed -- perform retirement implementation or production decommissioning inside `plan-retirement` diff --git a/skills/plan-retirement/SKILL.md b/skills/plan-retirement/SKILL.md deleted file mode 100644 index 13a9d5a..0000000 --- a/skills/plan-retirement/SKILL.md +++ /dev/null @@ -1,62 +0,0 @@ ---- -name: plan-retirement -description: Plans the safe retirement of one obsolete capability. Use when deprecation, migration, disablement, removal, archival, or recovery need an approved durable plan before delivery work. -disable-model-invocation: true ---- - -# Plan Retirement - -Turns evidence and user-authoritative decisions into one proportionate, reversible plan for retiring an obsolete capability. The plan reduces total system complexity without treating missing usage evidence as proof that no one depends on the capability or crossing into retirement execution. - -## Process - -### 1. Establish the retirement evidence - -Inspect the request, repository guidance, product intent and system-wide requirements in `PRODUCT.md`, `CONTEXT.md`, applicable ADRs, prior decisions and plans, and the smallest sufficient current implementation and runtime evidence. Trace the capability through entry points; users, tenants, callers, and operators; internal and external dependencies; public and private interfaces, APIs, events, schemas, configuration and flags; data ownership, flows, access, retention, archival, and recovery; integrations and providers; permissions and trust boundaries; observability; runbooks, support and communications; documentation; and known replacements. - -Treat code, tests, contracts, configuration, telemetry, inventories, tracker records, and operational systems as evidence within their actual scope and freshness. Distinguish confirmed use, confirmed non-use, and unknown use. Absence of observed traffic, repository references, or known consumers is not proof of no dependence. Identify lifecycle authorities and contradictions without copying mutable operational or tracker state into the plan. The capability boundary, evidence provenance, affected parties, obligations, current unknowns, and accessible authority are explicit. - -### 2. Resolve retirement intent and obligations - -Invoke `$elicit-with-context` and retain responsibility for the plan, writes, verification, and stopping boundary. Give it the inspected evidence and use one dependency-ordered question at a time to resolve the obsolete capability and reason for retiring it; acceptable end state and non-goals; timing or obligation trade-offs; consumer and operator migration; replacement expectations; compatibility and coexistence; data retention, deletion, archival, access, and recovery; integration and provider obligations; communication and support; security, privacy, compliance, and contractual constraints; acceptable reversibility and risk; ownership; and observable completion. Reuse compatible confirmed answers and ask only about material gaps. - -Invoke `$research` only when a material external legal, regulatory, provider, protocol, standards, or deprecation subject could change the plan and needs durable evidence. Read its report before using its findings; treat research as evidence, not authority to invent a project obligation. Invoke `$maintain-ubiquitous-language` inline when independently confirmed consequential project language is introduced, changed, misused, or contradicted; defer any design-dependent terminology or qualifying ADR until the complete retirement synthesis is confirmed. Each utility owns only its contract-specific result and returns control here. - -Keep every retirement-plan artefact unchanged until the complete synthesis is explicitly confirmed. The intended retirement outcome, authoritative obligations, unresolved decisions, and conditions that must gate irreversible work are honest and explicit. - -### 3. Compare proportionate retirement strategies - -Compare the status quo, replacement or migration, disablement without removal, staged retirement, and immediate removal where each is credible. Assess intended benefit, affected parties, consumer and operational risk, evidence gaps, cost and complexity, residual obligations, reversibility, recovery, and the condition that ends any temporary compatibility path. Prefer the least irreversible strategy that can produce the required outcome and evidence. Do not introduce permanent coexistence, shims, flags, duplicate data, or compatibility scaffolding without an explicit exit condition. - -Apply supporting techniques only when their runtime trigger is present: - -- Use **API deprecation and sunset semantics** for an externally consumed contract, distinguishing notice of discouragement from the planned end of availability and retaining direct consumer evidence, migration guidance, compatibility obligations, and completion signals. -- Use **Parallel Change** when an incompatible interface or schema must coexist while consumers or data migrate; plan the expansion, observed migration, and contraction gates without assuming the final removal is safe. -- Use the **Strangler Fig pattern** only when a substantial capability is being replaced incrementally across a stable routing or responsibility seam; keep ownership, coexistence, and the removal condition explicit. -- Use **controlled stop and recovery** when use is dormant, uncertain, or operationally mediated; plan a reversible disablement, evidence window, counter-signals, tested recovery, abort authority, and reassessment before removal. - -Choose direct atomic removal only when current evidence demonstrates that the change is local and reversible and that no consumer, data, public contract, integration, support, security, compliance, or recovery obligation requires a staged transition. Otherwise select only the applicable stages from deprecate, migrate or coexist, disable, observe, remove, and archive or close. The selected strategy is justified by evidence and confirmed trade-offs rather than ceremony or a preferred migration pattern. - -### 4. Build the complete retirement plan - -Define the selected end state and explicit non-goals, affected parties, prerequisites, owners, contract and deprecation policy, migration and compatibility path, communication and support plan, data disposition and access rules, integration and provider handling, operational and monitoring effects, security and compliance checks, archival and recovery, unresolved risks and decisions, and observable completion conditions. - -For each applicable stage, state its bounded outcome, entry evidence and prerequisites, affected contracts or state, owner, verification and counter-signals, abort conditions and authority, rollback or recovery route, residual obligations, and evidence required to exit. Use confirmed dates, deadlines, notice periods, adoption claims, retention rules, and production mechanisms only when their authority exists; otherwise preserve the unknown and its resolution gate. Ensure the final removal and closure eliminate temporary paths and stale obligations rather than merely hiding the capability. - -Trace every stage and obligation back to retirement intent, a consumer or dependent, current evidence, an external authority, or an explicit user decision. Compare the selected plan with the credible alternatives and retain only the rejected choices necessary to understand a material consequence or reversal. The proposed plan is safe to decompose later without becoming a file-by-file implementation plan, task list, estimate, sprint, release, or deployment procedure. - -### 5. Confirm the complete synthesis - -Have `$elicit-with-context` present one concise, self-contained synthesis covering the evidence boundary and unknown use, retirement outcome and non-goals, affected parties and obligations, alternatives, selected strategy, stages and ownership, migration and compatibility, communications and support, data and integrations, operations, security and compliance, verification, abort and recovery, completion conditions, unresolved risks, context changes, durable destination, and exact proposed write. Obtain explicit confirmation of the whole synthesis; selecting an earlier option is not confirmation. - -If the user rejects or revises it, leave the retirement plan and design-dependent context unchanged, reopen every affected evidence, obligation, option, stage, owner, and completion branch, and confirm the complete revised synthesis. Independently confirmed glossary updates already applied inline remain valid. The confirmed synthesis supplies authority for its agreed plan and context writes without another application prompt. - -### 6. Write and verify the durable plan - -Immediately before writing, refresh mutable usage, dependency, contract, provider, retention, and operational evidence that could change the confirmed plan. When drift materially changes the strategy, obligation, stage, owner, risk, or completion claim, leave the plan unchanged and return to complete-synthesis confirmation. - -Follow the project's established retirement-record convention. Otherwise choose a stable lowercase capability slug and create `docs/retirements/.md` using the [Retirement Plan Template](assets/retirement-plan-template.md). Preserve unrelated confirmed content and local structure when refining an existing plan. Include only material sections, link authorities and mutable owning systems instead of duplicating them, and distinguish confirmed current evidence, approved future action, and unresolved gated decisions. - -After writing the plan, invoke `$maintain-ubiquitous-language` for confirmed design-dependent terminology and `$maintain-decision-records` for any qualifying ADR whose durable-rationale agreement was included in the synthesis. Re-read the plan against product intent, repository and runtime evidence, context and architecture authorities, external research, and the confirmed synthesis. Verify all paths and links; end-to-end dependent coverage; stage entry, exit, evidence, abort, rollback and recovery; owner and obligation coverage; data, integration, communication, support, security and compliance treatment; removal of temporary paths; and observable completion. State evidence limitations and never describe unknown or future evidence as established. - -Report the plan path, selected strategy, affected parties and obligations, stage and completion outline, confirmed context changes, supporting evidence, verification, unresolved gated decisions, and limitations. Stop after the approved retirement plan and applicable context records are written and verified. Do not create tracker items, invoke downstream lifecycle skills, implement or decommission the capability, disable or delete code, data, configuration or infrastructure, contact consumers, commit, push, open a pull request, release, deploy, or continuously monitor. Later sessions own `$create-tickets`, `$implement`, `$commit`, and `$pull-request`. diff --git a/skills/plan-retirement/agents/openai.yaml b/skills/plan-retirement/agents/openai.yaml deleted file mode 100644 index a8405bc..0000000 --- a/skills/plan-retirement/agents/openai.yaml +++ /dev/null @@ -1,6 +0,0 @@ -interface: - display_name: 'Plan Retirement' - short_description: 'Plan a safe evidence-led capability retirement' - default_prompt: 'Use $plan-retirement to create an approved plan for safely retiring this obsolete capability.' -policy: - allow_implicit_invocation: false diff --git a/skills/plan-retirement/assets/retirement-plan-template.md b/skills/plan-retirement/assets/retirement-plan-template.md deleted file mode 100644 index 0705159..0000000 --- a/skills/plan-retirement/assets/retirement-plan-template.md +++ /dev/null @@ -1,71 +0,0 @@ -# {Capability name} retirement plan - - - -## Outcome and authorities - - - -## Scope and non-goals - - - -## Current capability and evidence - - - -## Affected parties and obligations - - - -## Options and selected strategy - - - -## Retirement stages - - - -### {Stage name} - -- **Outcome:** {Bounded state this stage establishes.} -- **Entry evidence and prerequisites:** {What must be true before the stage starts.} -- **Affected contracts or state:** {Consumers, interfaces, data, integrations, controls, or operations affected.} -- **Owner:** {Accountable role or confirmed owner.} -- **Verification and counter-signals:** {Observable evidence that distinguishes success, partial completion, hidden dependence, and harm.} -- **Abort conditions and authority:** {Signals that stop progression and who decides.} -- **Rollback or recovery:** {Proportionate route back to a safe state and required assets.} -- **Exit evidence:** {What must be observed before the next stage or completion.} -- **Residual obligations:** {What remains after the stage and its exit condition.} - -## Migration and compatibility - - - -## Communication and support - - - -## Data, archival, access and recovery - - - -## Integrations and providers - - - -## Operations, security and compliance - - - -## Risks, decisions and prerequisites - - - -## Completion and closure - - - -## Later delivery handoff - - From 9c699422beca079b17bec436b6865d41881d6b53 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 15:44:49 +0100 Subject: [PATCH 68/79] feat(review-architecture): finalize whole-codebase review --- SKILLS_PLAN.md | 2 +- skills/review-architecture/SKILL.md | 68 ++---- skills/review-architecture/agents/openai.yaml | 2 +- .../assets/report-template.html | 213 ++++++++++++++++++ .../scripts/validate-report.js | 99 ++++++++ 5 files changed, 330 insertions(+), 54 deletions(-) create mode 100644 skills/review-architecture/assets/report-template.html create mode 100644 skills/review-architecture/scripts/validate-report.js diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 9ffd599..30ce05e 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -323,7 +323,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Update `commit` skill - [x] Replace `pr` with `pull-request` - [x] Create `review-pull-request` skill -- [-] Update `review-architecture` skill +- [x] Update `review-architecture` skill - [-] Update `debug` skill - [x] Restore `elicit-with-context` skill - [-] Update the README diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index d06fc04..a49cdda 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -1,75 +1,39 @@ --- name: review-architecture -description: Assesses an explicitly bounded codebase area for a few high-value architecture improvements and reports the evidence without changing implementation. Use for architecture or maintainability reviews. +description: Reviews a whole codebase for high-impact evidence-backed modular-design improvements and writes an interactive HTML report. Use for architecture or structural maintainability reviews. disable-model-invocation: true --- # Review Architecture -Produces one canonical Markdown assessment and a user-friendly disposable HTML explanation of the highest-value structural improvements in an explicitly bounded codebase area. It preserves the reviewed implementation and stops before detailed redesign, tickets, or delivery work. +Finds the most consequential current opportunities to improve a codebase's modular structure and explains them in one interactive HTML report. ## Process -### 1. Bound and freeze the review +### 1. Establish whole-codebase evidence -Inspect the request, repository guidance, available tools, and current repository state. Establish the explicit codebase scope, exclusions, review purpose, applicable architecture or maintainability qualities, and any user-held priority that could change qualification or rank. Do not silently expand an ambiguous scope to the whole repository. Ask only for missing scope or priority that cannot be discovered and would materially change the result. +Inspect the request, repository guidance, root `PRODUCT.md` and `CONTEXT.md`, applicable ADRs, current revision, and starting worktree state. Treat the whole repository as the review scope while honouring explicit exclusions. Map its capabilities and inspect the source, tests, contracts, schemas, configuration, and operational evidence needed to understand current structure. Exclude dependencies, generated output, vendored code, and obsolete paths from recommendation coverage. -Capture the reviewed revision and starting worktree state. Read root `PRODUCT.md`, `CONTEXT.md` and applicable ADRs when present, then inspect only the scoped implementation, tests, contracts, schemas, configuration, history, and operational evidence needed to assess the area. Treat missing or conflicting authorities as limitations unless they prevent a responsible comparison, in which case expose the exact gap. Separate intended constraints, observed implementation, inferences, and unknowns. The review boundary, authorities, evidence baseline, and limitations are explicit before findings are formed. +Inspect relevant Git history after mapping the repository. Use repeated cross-file changes and hotspots as supporting evidence and ranking signals without excluding quieter areas. Distinguish project authority, current observation, inference, and unknowns. -### 2. Map structural evidence +Invoke `$modular-design` with the authorities and current evidence. Apply its structural teaching while retaining ownership of investigation, candidate qualification, ranking, the report, and stopping. Invoke `$research` only when a material candidate depends on external framework, platform, or technical evidence that the repository cannot establish; read its report before deciding whether the candidate qualifies. -Map capabilities, owned knowledge, contracts, consumers, dependencies, runtime boundaries, and verification seams instead of equating directories, classes, packages, services, or deployment units with modules. Invoke `$modular-design` with the fixed scope, authorities, and current evidence. Use its findings as passive structural authority for ownership, information hiding, cohesion, coupling, contracts, dependency direction, seams, and change locality; retain responsibility for investigation, candidate comparison, prioritisation, both reports, and stopping. +### 2. Retain only top-level candidates -Trace material pressure to precise current repository locations. Look for rules represented in multiple owners, unstable or leaking contracts, dependencies pointing against established responsibility, changes that repeatedly propagate across boundaries, and regression or operational evidence tied to structure. Use coherent repository history or metrics only to locate and challenge candidates, discounting bulk, generated, renamed, bot-authored, or obsolete changes; neither a metric nor a smell label proves a recommendation. Express a material quality or change pressure as a compact project-evidenced scenario when that makes its stimulus, affected capability, expected response, or observable consequence clearer. The evidence map distinguishes observations, supported inferences, and uncertainty. +Require each candidate to have precise current repository evidence, a credible structural cause, a material present benefit, and visible trade-offs or uncertainty. State only the recommended structural direction, why it matters, and what it improves. Claim only improvements caused by that structural direction; do not imply that it establishes correctness, safety, or another quality it cannot demonstrate. Reject cosmetic cleanup, unsupported smell labels, speculative abstractions, and changes whose benefit is minor or hypothetical. -Invoke `$research` only when a material external subject needs durable evidence. Read its report before using its findings, and keep that report explicitly within the review scope. Do not import a general architecture method, standard, or universal quality model when project evidence does not make it authoritative. +Keep only the highest-impact qualifying candidates. Use no fixed count. Order them by explicit qualitative judgement of present consequence, evidence, likely benefit, and feasibility; preserve that fixed order and identify the top recommendation. Zero candidates is a valid result. Record reviewed coverage and meaningful non-qualifying areas so absence of recommendations remains trustworthy. -### 3. Compare and qualify candidates +### 3. Write the interactive report -For every serious candidate, apply **Design It Twice**: compare the current design with at least one materially different responsibility, boundary, contract, or dependency arrangement. Consider the status quo and another credible alternative when either could reasonably win. Do not count naming, file placement, interface syntax, or a straw design as an alternative, and do not elaborate any option beyond what the review decision requires. +Write only `docs/architecture/YYYYMMDD-full-codebase-architecture-review.html`, using `-2`, `-3`, and later numeric suffixes when a same-day path exists unless replacement was explicit. Copy and complete [the report template](assets/report-template.html). Escape repository-derived text before placing it in HTML and remove every template marker. -Retain a recommendation only when it has all of: +Keep each candidate's recommendation, reason, and improvement always visible. Put exact file and history evidence, trade-offs, and uncertainty in one native expandable section. Do not add filters, reordering, scores, detailed target interfaces, file-by-file changes, migration steps, test prescriptions, tickets, or an implementation plan. -- a material benefit to an explicit constraint or quality; -- exact current repository evidence and a credible causal explanation; -- a clear current and target responsibility, contract, or dependency difference; -- a bounded transition seam, smallest useful step, and containment or recovery route; -- observable verification or fitness evidence for the claimed benefit; and -- costs, trade-offs, risks, alternatives, and material uncertainty visible enough to support a decision. +Add a current-versus-recommended visual only when it makes the structural relationship materially easier to understand. Design it specifically for that candidate with accessible HTML or inline SVG; add another CDN library only when it clearly improves the explanation. Omit a weak, generic, or decorative visual. -Reject local cosmetic cleanup, a smell without consequence, speculative abstraction, and a redesign whose expected benefit or transition cannot be evidenced. Zero recommendations is a valid successful result. +### 4. Verify and stop -Select only the small decision-useful shortlist with the greatest consequence or quality priority, evidence reach, confidence, expected benefit, and migration feasibility. Rank it qualitatively with an explained ordering; do not calculate composite scores or confidence percentages. Record reviewed areas and lower-value candidates that were not retained only to the degree needed to make coverage honest. The shortlist contains the highest-value supported recommendations, not an inventory of conceivable redesigns. +Run the bundled `scripts/validate-report.js` with Bun and the report path. Re-read the report against the fixed order and current evidence; verify every cited repository path and material claim, the top recommendation, coverage, limitations, HTML escaping, and absence of leftover markers. Recheck the revision for material drift and confirm that only the report changed apart from pre-existing work. Do not open the report or perform rendered visual review. -### 4. Resolve the report outcome - -Present the material scope, limitations, concise outcome, and proposed ordered shortlist before persistence when a user-authoritative trade-off could still change them. Compare the live choices and let the user decide; do not convert an unavailable preference into an architectural conclusion. Once shared understanding is reached, the invocation authorises both report artifacts without a second application prompt. - -### 5. Write the canonical review - -Follow the project's existing architecture-review convention. Otherwise write `docs/architecture/YYYYMMDD-{scope-slug}-architecture-review.md`, preserving an unrelated existing file with a numeric suffix unless replacement was explicit. Use **plain language** for the outcome, finding titles, and short `Issue`, `Change`, and `Benefit` summaries while retaining precise project and architecture terms. - -Make the report independently useful with: - -1. scope, exclusions, reviewed revision, authorities, qualities, and concise outcome; -2. an ordered finding index, or an explicit zero-finding result; -3. for each numbered finding, precise evidence; current and target responsibilities and contracts; benefit; trade-offs, cost, risk, and uncertainty; the rejected alternative or status quo; transition seam, smallest useful step, coexistence or containment; and observable verification or fitness evidence; -4. reviewed coverage, candidates not retained when material, limitations, stale-evidence risk, and any invoked research. - -Use a small repository-native diagram only when it materially clarifies a complex ownership, dependency, runtime, or transition relationship. Keep the report at architecture-decision depth: it can support a later decision or body of work, but it is not a file-by-file implementation plan. - -### 6. Present the explanatory view - -Create a self-contained disposable HTML file in the operating system's temporary directory and open it for the user with an available browser tool. Treat Markdown as canonical and HTML as a linear explanatory view, not a second independently maintained report. - -Show the outcome and linked finding index first. Give each numbered finding one section whose always-visible content includes the same title and core `Issue`, `Change`, and `Benefit` claims as Markdown. Add a paired current-and-target visual only when it clarifies the material relationship; stack it without losing correspondence at narrow widths. Put ancillary evidence, alternatives, transition detail, and limitations in native `details` disclosures only when hiding them improves scanning; keep information required to understand or decide the recommendation visible. Use no filters, tabs, nested disclosure, dashboard cards, or JavaScript-dependent controls for the small shortlist. - -Keep the page readable as one column, semantic and keyboard-operable, usable without colour, and printable with all content available. Make printing open every native disclosure through a self-contained `beforeprint` behaviour or an equivalent that works while the on-screen disclosures start closed; do not assume print CSS can expose a closed `details` subtree. Escape repository-derived content for HTML, avoid unverified external dependencies, and include the canonical Markdown path, reviewed revision, scope, and limitations. A zero-finding view states the result directly and shows coverage and limitations without empty recommendation furniture. - -### 7. Verify and stop - -Re-read both artifacts against the frozen evidence. Verify every repository path and material claim, the Markdown structure and links, the minimal matching finding numbers, order, titles, and core claims across artifacts, and the separation between observation, inference, and uncertainty. Recheck the source revision and implementation paths; when material source drift occurred, reassess affected findings or mark the evidence stale rather than presenting it as current. - -Inspect the actual HTML in an available browser at ordinary and narrow widths. Exercise native disclosures by pointer and keyboard, check focus, reflow, overflow, diagrams, colour independence, and a print render with every disclosure open, and correct material defects. When browser inspection or printing is unavailable, perform structural and source checks and report the exact visual limitation. - -Confirm the reviewed implementation and its existing tests, configuration, architecture, context, product, and feature documents remain unchanged apart from concurrent pre-existing work. Return both report paths, scope and exclusions, finding count and order, invoked research, verification performed, limitations, and any stale evidence. Stop without refactoring, implementing, editing project authorities, creating tickets or specifications, scheduling work, committing, publishing, releasing, or deploying. +Return the report path, ordered candidate titles or the zero-candidate result, reviewed revision, invoked research, verification, and limitations. Stop without changing implementation or project authorities, elaborating a detailed design, creating specifications or tickets, committing, or publishing. diff --git a/skills/review-architecture/agents/openai.yaml b/skills/review-architecture/agents/openai.yaml index 2ead7b8..21441f7 100644 --- a/skills/review-architecture/agents/openai.yaml +++ b/skills/review-architecture/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Review Architecture' - short_description: 'Find high-value architecture improvements' + short_description: 'Find the biggest structural improvements' policy: allow_implicit_invocation: false diff --git a/skills/review-architecture/assets/report-template.html b/skills/review-architecture/assets/report-template.html new file mode 100644 index 0000000..0274bc8 --- /dev/null +++ b/skills/review-architecture/assets/report-template.html @@ -0,0 +1,213 @@ + + + + + + %%PROJECT_NAME%% architecture review + + + + + +
+
+

+ Architecture review · %%REVIEW_DATE%% +

+

+ %%PROJECT_NAME%% +

+

+ %%OUTCOME_SUMMARY%% +

+
+
+
Reviewed revision
+
+ %%REVISION%% +
+
+
+
Recommendations
+
%%CANDIDATE_COUNT%%
+
+
+
Scope
+
+ Whole codebase%%SCOPE_EXCLUSIONS%% +
+
+
+
+ + + +
+
+
+

+ Prioritised shortlist +

+

+ Highest-impact opportunities +

+
+ +
+ +
+ +
+

+ Recommendation 01 · top recommendation +

+

+ %%CANDIDATE_TITLE%% +

+
+
+
+ Recommendation +
+
+ %%RECOMMENDATION%% +
+
+
+
+ Why it matters +
+
+ %%WHY_IT_MATTERS%% +
+
+
+
+ What it improves +
+
+ %%WHAT_IT_IMPROVES%% +
+
+
+ + + +
+ + Evidence and trade-offs + + +
+
+

+ Evidence +

+
%%EVIDENCE%%
+
+
+

+ Trade-offs and uncertainty +

+
%%TRADEOFFS%%
+
+
+
+
+ %%ADDITIONAL_CANDIDATES_OR_ZERO_RESULT%% +
+
+ +
+

+ Review record +

+

+ Coverage and limitations +

+
+
+

Reviewed

+
%%REVIEWED_COVERAGE%%
+
+
+

+ Limitations +

+
%%LIMITATIONS%%
+
+
+
+
+ + diff --git a/skills/review-architecture/scripts/validate-report.js b/skills/review-architecture/scripts/validate-report.js new file mode 100644 index 0000000..10a0a01 --- /dev/null +++ b/skills/review-architecture/scripts/validate-report.js @@ -0,0 +1,99 @@ +#!/usr/bin/env bun + +import fs from 'node:fs'; +import path from 'node:path'; + +const reportPath = process.argv[2]; +const errors = []; + +if (!reportPath) { + console.error('Usage: bun scripts/validate-report.js '); + process.exit(2); +} + +const absolutePath = path.resolve(reportPath); +if (!fs.existsSync(absolutePath) || !fs.statSync(absolutePath).isFile()) { + errors.push(`Report not found: ${absolutePath}`); +} + +let html = ''; +if (!errors.length) html = fs.readFileSync(absolutePath, 'utf8'); + +const normalized = absolutePath.split(path.sep).join('/'); +if ( + !normalized.includes('/docs/architecture/') || + path.extname(absolutePath) !== '.html' +) { + errors.push('Report must be an HTML file under docs/architecture/.'); +} +if ( + !/^\d{8}-full-codebase-architecture-review(?:-\d+)?\.html$/.test( + path.basename(absolutePath), + ) +) { + errors.push( + 'Report filename must follow the dated full-codebase convention.', + ); +} + +const required = [ + ['HTML document', //i], + ['language', /]*\blang=/i], + ['UTF-8 metadata', /]*charset=["']?utf-8/i], + ['responsive viewport', /]*name=["']viewport["']/i], + [ + 'Tailwind browser CDN', + /cdn\.jsdelivr\.net\/npm\/@tailwindcss\/browser@4/, + ], + [ + 'deferred Alpine CDN', + /]*\bdefer\b[^>]*alpinejs@3\.x\.x\/dist\/cdn\.min\.js/i, + ], + ['review root', /\bdata-architecture-review\b/], + ['reviewed revision', /\bdata-reviewed-revision\b/], + ['coverage', /\bdata-coverage\b/], +]; + +for (const [label, pattern] of required) { + if (!pattern.test(html)) errors.push(`Missing ${label}.`); +} + +if (/%%[A-Z0-9_-]+%%/.test(html)) + errors.push('Unresolved template marker remains.'); + +const candidates = [ + ...html.matchAll(/]*\bdata-recommendation=["'](\d+)["']/gi), +]; +const numbers = candidates.map((match) => Number(match[1])); +if (numbers.some((number, index) => number !== index + 1)) { + errors.push( + 'Recommendation numbers must be contiguous and in fixed document order.', + ); +} + +if (candidates.length) { + if (!/\bdata-top-recommendation\b/.test(html)) + errors.push('Missing top recommendation.'); + for (const marker of [ + 'data-recommendation-summary', + 'data-field="recommendation"', + 'data-field="why"', + 'data-field="improves"', + 'data-evidence', + 'data-tradeoffs', + ]) { + if (!html.includes(marker)) + errors.push(`Missing candidate field: ${marker}.`); + } +} else if (!/\bdata-zero-result\b/.test(html)) { + errors.push('A zero-candidate report must contain data-zero-result.'); +} + +if (errors.length) { + for (const error of errors) console.error(`- ${error}`); + process.exit(1); +} + +console.log( + `Report valid: ${absolutePath} (${candidates.length} recommendations)`, +); From 0a16623a4e1a1d59582cd2b360a08c6646c4861d Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 16:41:59 +0100 Subject: [PATCH 69/79] feat(skills): finalize systematic debugging --- SKILLS_PLAN.md | 2 +- skills/debug/SKILL.md | 56 ++++++++++++------- skills/debug/agents/openai.yaml | 2 +- skills/debug/references/BOUNDARY-EVIDENCE.md | 23 -------- skills/debug/references/BOUNDARY-ISOLATION.md | 17 ++++++ skills/debug/references/CHANGE-BISECTION.md | 17 ++++++ .../debug/references/COMPARATIVE-DEBUGGING.md | 15 +++++ skills/debug/references/DELTA-DEBUGGING.md | 17 ++++++ .../references/HYPOTHESIS-EXPERIMENTS.md | 23 ++++++++ .../references/INSTRUMENTATION-DEBUGGERS.md | 24 ++++++++ .../references/NONDETERMINISTIC-CONCURRENT.md | 24 ++++++++ skills/debug/references/NONDETERMINISTIC.md | 37 ------------ skills/debug/references/ORIGIN-TRACKING.md | 23 ++++++++ .../debug/references/PERFORMANCE-RESOURCE.md | 25 +++++++++ skills/debug/references/PERFORMANCE.md | 28 ---------- skills/debug/references/TECHNIQUES.md | 52 ----------------- 16 files changed, 224 insertions(+), 161 deletions(-) delete mode 100644 skills/debug/references/BOUNDARY-EVIDENCE.md create mode 100644 skills/debug/references/BOUNDARY-ISOLATION.md create mode 100644 skills/debug/references/CHANGE-BISECTION.md create mode 100644 skills/debug/references/COMPARATIVE-DEBUGGING.md create mode 100644 skills/debug/references/DELTA-DEBUGGING.md create mode 100644 skills/debug/references/HYPOTHESIS-EXPERIMENTS.md create mode 100644 skills/debug/references/INSTRUMENTATION-DEBUGGERS.md create mode 100644 skills/debug/references/NONDETERMINISTIC-CONCURRENT.md delete mode 100644 skills/debug/references/NONDETERMINISTIC.md create mode 100644 skills/debug/references/ORIGIN-TRACKING.md create mode 100644 skills/debug/references/PERFORMANCE-RESOURCE.md delete mode 100644 skills/debug/references/PERFORMANCE.md delete mode 100644 skills/debug/references/TECHNIQUES.md diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 30ce05e..5d1b9b1 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -324,7 +324,7 @@ The author's normal flow does not invoke `review-pull-request`; `implement` alre - [x] Replace `pr` with `pull-request` - [x] Create `review-pull-request` skill - [x] Update `review-architecture` skill -- [-] Update `debug` skill +- [x] Update `debug` skill - [x] Restore `elicit-with-context` skill - [-] Update the README - [-] Run the suite-wide acceptance pass diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index ce5d1a2..3cb36cf 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -1,49 +1,67 @@ --- name: debug -description: Diagnoses defects and conditionally makes verified repairs. Use for failures, regressions, runtime errors, performance faults, incorrect behaviour, or causal debugging. +description: Diagnoses software defects and conditionally makes causal, verified repairs. Use for failures, regressions, runtime errors, flaky behaviour, or performance faults. disable-model-invocation: true --- # Debug -**Scientific debugging** establishes an evidence-backed causal explanation through falsifiable hypotheses, predicted observations, and discriminating experiments. It stops at diagnosis when requested and permits a repair only when repair is part of the agreed outcome. +Diagnoses a defect through discriminating evidence and, when requested, produces one minimal reviewed local repair. ## Process ### 1. Fix the outcome and signal -Read repository instructions, the requested outcome, expected and observed behaviour, environment evidence, current implementation and tests, and applicable product, feature, architecture, and operational authorities. Resolve whether the outcome is diagnosis only or diagnosis plus repair; ask when that distinction is materially ambiguous. Record the authorised local, external, and production boundaries before probing or editing. +Read the request, repository instructions, expected and observed behaviour, failing evidence, environment, relevant authorities, current code, and tests. Resolve whether the requested outcome is diagnosis only or diagnosis plus repair. Invoke `$elicit-with-context` only when a material user-held fact or decision remains. -Define the smallest reliable signal whose verdict distinguishes expected from observed behaviour. Use a focused test, runtime error, trace or log, captured artefact, controlled benchmark, safe external observation, or production-boundary evidence according to the fault. Record its invocation or probe, input, environment, expected verdict, observed verdict, and fidelity limitations. Do not claim a local reproduction when only external or captured evidence exists. Run an external or production probe only when it is read-only and within available authority; obtain explicit authority before any probe that can mutate durable or external state, expose sensitive data, or materially affect service. Preserve the original unminimised signal for the final causal check. +Keep production and external systems read-only and within explicit authority. When diagnosis or mitigation needs external mutation, report the required action and stop; production mitigation, deployment, and monitoring remain outside this skill. -When the signal or next experiment is unclear, load [Debugging Techniques](references/TECHNIQUES.md). For an intermittent, flaky, timing-sensitive, order-dependent, or concurrent fault, load [Nondeterministic Faults](references/NONDETERMINISTIC.md). For a component, process, service, environment, or production boundary, load [Boundary Evidence](references/BOUNDARY-EVIDENCE.md). For a latency, throughput, resource, query, or scale regression, load [Performance Faults](references/PERFORMANCE.md); also load the nondeterministic guide when its classifier is noisy. If no reliable signal can be established, leave implementation unchanged and stop with the exact evidence gap and next smallest discriminating experiment. The outcome, authority, and evidence boundary are explicit. +Define the best available **signal**: a focused test, command, error, trace, captured artefact, controlled benchmark, safe observation, or measured failure rate. Record its invocation or provenance, input, environment, expected result, observed result, and fidelity limits. Run it before changing implementation when safe. Prefer a sharp repeatable reproduction, but accept authorised captured, external, or probabilistic evidence when that is the honest boundary. Preserve the original unminimised signal through the investigation. If no evidence can distinguish the reported defect, leave implementation unchanged and return the exact gap and next smallest experiment. -### 2. Establish the causal explanation +### 2. Narrow the cause -State the plausible causal hypotheses in evidence-supported order. For the leading hypothesis, predict an observation that distinguishes it from credible alternatives, then run the smallest safe experiment capable of producing that observation. Change one variable at a time and record the hypothesis, prediction, experiment, observation, and whether the result supports, rejects, or fails to discriminate the hypothesis. Treat correlation, disappearance of a symptom, and a passing retry as evidence to explain rather than proof of cause. +Build a small model of the relevant path, contracts, state, boundaries, and recent code, configuration, dependency, environment, data, or workload changes. Treat recent change, suspicious location, correlation, and a passing retry as hypotheses or observations, never as proof. -Repeat until the evidence supports a root cause and causal mechanism and accounts for the strongest alternatives, or until a named evidence boundary prevents the next discriminating experiment. Preserve failed hypotheses and contradictory observations. Keep diagnosis-only experiments non-mutating with respect to the implementation; when the next discriminating experiment requires a source or configuration change, stop with that exact expanded outcome rather than performing it. Remove temporary local or external effects owned by the diagnostic run. Invoke `$research` only when a material external-knowledge subject needs durable evidence and its cited report is explicitly in scope; use proportionate transient documentation lookup otherwise. Invoke `$maintain-ubiquitous-language` only after independently confirmed terminology arises and `$maintain-decision-records` only after an accepted decision reaches its rare ADR gate. For diagnosis-only work, stop here without implementation mutation and follow the Handoff. +Maintain one compact investigation record containing observations, a small ranked set of causal hypotheses, the active prediction, the experiment, its result, its implication, and contradictory evidence. Test one prediction at a time with the smallest safe experiment that distinguishes credible alternatives. Prefer changing one factor while preserving the signal's material conditions. -### 3. Test one causal repair +Select the reference whose trigger matches the current uncertainty and load only that reference. Return its evidence to this causal loop before selecting another: -For a repair outcome, select the smallest change that addresses the supported cause. Invoke `$modular-design` only when the repair presents a material ownership, boundary, contract, dependency, seam, or change-propagation decision, and carry its constraints into the repair. Exclude unrelated cleanup, broad refactoring, speculative hardening, and other plausible defects. +| Evidence shape | Load | +| --- | --- | +| Multiple plausible explanations remain | [Hypothesis-Driven Experiments](references/HYPOTHESIS-EXPERIMENTS.md) | +| A component, pipeline, request, or data path has observable boundaries | [Boundary Isolation](references/BOUNDARY-ISOLATION.md) | +| Reliable known-good and known-bad ordered states exist | [Change Bisection](references/CHANGE-BISECTION.md) | +| A large input, state, sequence, trace, or change set can be reduced | [Delta Debugging](references/DELTA-DEBUGGING.md) | +| Invalid state appears downstream of where it originated | [Origin Tracking](references/ORIGIN-TRACKING.md) | +| A credible working case can be compared with the failing case | [Comparative Debugging](references/COMPARATIVE-DEBUGGING.md) | +| Existing evidence cannot distinguish the active hypotheses | [Instrumentation and Debuggers](references/INSTRUMENTATION-DEBUGGERS.md) | +| Outcome depends on timing, order, randomness, load, or shared state | [Nondeterministic and Concurrent Faults](references/NONDETERMINISTIC-CONCURRENT.md) | +| The signal is latency, throughput, contention, exhaustion, leakage, or resource cost | [Performance and Resource Faults](references/PERFORMANCE-RESOURCE.md) | -Invoke `$tdd` only when an existing runnable suite can exercise the desired observable behaviour through a stable boundary with an independent oracle. An established reproduction may supply TDD's Red only when it independently specifies that desired behaviour and fails for the expected behavioural reason; a crash, log, trace, performance symptom, or implementation-derived expectation is not automatically a retained test or valid Red. Otherwise apply one conventional minimal repair and add or update regression protection only when existing project infrastructure can meaningfully express the changed behaviour. Do not install or invent a test or quality framework for this repair. +When the shape is unclear, verify the signal and inspect recent changes, then compare a working case, isolate boundaries, reduce the reproducer, trace the first invalid state, and finally add targeted probes. Enter nondeterministic or performance guidance immediately when its trigger applies. Combine techniques only when each answers a distinct unresolved question. -Run the original signal against the attempt. Retain the repair provisionally only when the signal changes as the causal model predicted. When it does not, record the contradiction and revert only changes introduced and owned by that attempt when their exact ownership is known, preserving the original signal, diagnostic evidence, and all pre-existing user work. Do not use destructive reset or checkout, discard uncertain state, or stack another unsupported fix. Remove owned temporary instrumentation and throwaway harnesses before freezing a supported candidate; retain a diagnostic only when the requested scope and project authority make it durable. Return to the causal hypotheses with either one supported candidate or no repair residue. +Continue until the evidence explains the causal mechanism and accounts proportionately for the strongest alternatives. A root cause may be a set of jointly necessary conditions rather than one deepest line. If the next discriminating experiment is unsafe, unavailable, outside authority, or no longer economical, stop with the supported findings, uncertainty, and exact next experiment. Invoke `$research` only when a material external-knowledge question requires a durable cited report. -### 4. Review the fixed candidate +### 3. Close diagnosis or repair one cause -Freeze the complete supported candidate: exact diff or revisions, changed and untracked paths, repository state, intended-behaviour authority, causal account, original signal, applicable structural constraints, and named risks. Invoke `$code-review` independently on that basis. If the candidate drifts or an applicable review axis is stale or unperformed, re-freeze and repeat the review when scope and ownership remain clear; otherwise stop with the exact blocker. +Remove owned temporary probes, captures, fixtures, and experimental edits, retaining a diagnostic only when project authority makes it durable and safe. State the supported cause, mechanism, trigger conditions, evidence for and against the material alternatives, contradictions, confidence boundary, and remaining unknowns. For diagnosis-only work, state that implementation was not changed and stop. -Adjudicate every finding against the requested outcome, project authorities, causal evidence, and current code. Correct each supported required finding through the applicable TDD or conventional repair path; reject a finding only with concrete contrary evidence. When a finding requires a user-authoritative change to behaviour, contract, architecture, or scope, ask one material question and pause mutation until the answer is confirmed. After any candidate change, re-run the original signal, freeze the new candidate, and repeat independent review. Continue only when the complete current candidate has no unresolved required finding. +For an authorised repair, apply `$modular-design` to keep the change with the natural owner and limit propagation. Select one minimal change directed at the supported cause; exclude unrelated cleanup, broad refactoring, speculative hardening, and other plausible defects. -### 5. Verify the reviewed repair +Apply `$tdd` when its prerequisites hold. The established signal supplies Red only when it independently specifies the desired observable behaviour and fails for that reason; a crash, trace, log, profile, or implementation-derived expectation does not automatically qualify. Otherwise use the strongest existing project-native feedback without claiming TDD or creating project-wide test infrastructure. -Invoke `$verify-change` on the exact reviewed candidate. Supply the intended behaviour, causal claim, original unminimised signal, focused regression evidence, applicable nearby and repository requirements, named risks, review dispositions, and any changed tests, snapshots, baselines, suppressions, scripts, configuration, dependencies, or generated artefacts. A passing regression alone does not prove the root cause; require the original signal to change as predicted and explain the causal mechanism and evidence against material alternatives. +Run the original signal against the repair. Retain the attempt only when the result changes as the causal account predicted. When it does not, record the contradiction, revert only the exact changes owned by that attempt without disturbing pre-existing work, and return to the causal loop. Do not stack another unsupported fix. -When verification exposes an in-scope failure, return to the causal repair path, then independently re-review and freshly verify the new fixed candidate. When source or measurement-path drift makes evidence stale, establish ownership, re-freeze, re-review, and re-verify rather than retargeting old evidence. Preserve a required pre-existing or out-of-scope failure as a blocker instead of widening the repair. Claim a completed repair only when the current review has no unresolved required finding and verification returns `passed within stated scope`; treat `partially verified`, `unverified`, or `failed` as incomplete or blocked and report it honestly. +Apply `$quality-harnesses` to the fixed candidate and preserve its claim-level evidence and missing-evidence limits. Add or retain regression protection only through existing infrastructure and a credible behavioural seam. + +### 4. Review and verify the local repair + +Freeze the complete candidate, intended-behaviour authority, causal account, original signal, regression evidence, selected quality evidence, changed and untracked paths, and repository state. Invoke `$code-review` on that exact candidate. Adjudicate every finding against the request, authorities, causal evidence, and current code. Correct supported required findings, ask the user about material behaviour, architecture, or scope decisions, and reject unsupported or out-of-scope findings with concrete evidence. + +After any candidate change, rerun the original signal and affected harnesses, re-freeze the candidate, and repeat independent review. Complete only when the current candidate has no unresolved required finding; the focused regression passes; the original unminimised signal changes as predicted; applicable repository and risk-triggered checks have been run; and unavailable or inconclusive evidence remains explicit. Do not turn green checks into universal proof. ## Handoff -Report the requested outcome; expected and observed behaviour; authority and environment boundary; original signal; hypotheses, predictions, experiments, and contradictory evidence; root cause and causal mechanism or exact unresolved evidence boundary; reverted attempts; retained change and regression protection; review scope and finding dispositions; verification scope, commands, evidence, verdict, limitations, residual risks, and next discriminating experiment or planned work. For diagnosis-only work, state that no implementation was changed. Stop after the diagnosis or verified local repair without committing, pushing, opening a pull request, changing tracker state, releasing, deploying, monitoring, or adding speculative documentation. +Return the requested outcome; expected and observed behaviour; authority and environment boundary; original signal; hypotheses, predictions, experiments, contradictions, and reverted attempts; causal mechanism or exact evidence gap; retained change and regression protection; quality evidence and limitations; review scope and dispositions; residual risk; and next discriminating experiment or follow-up. Include one brief prevention observation only when it follows directly from the established cause; do not create an automatic postmortem or wider programme. + +For diagnosis-only work, state that implementation was unchanged. Stop after the diagnosis or reviewed verified local repair without committing, changing tracker state, publishing, deploying, mutating production, or monitoring. diff --git a/skills/debug/agents/openai.yaml b/skills/debug/agents/openai.yaml index 445ebaf..412d7c6 100644 --- a/skills/debug/agents/openai.yaml +++ b/skills/debug/agents/openai.yaml @@ -1,5 +1,5 @@ interface: display_name: 'Debug' - short_description: 'Find root causes and verify repairs' + short_description: 'Find causal defects and verify local repairs' policy: allow_implicit_invocation: false diff --git a/skills/debug/references/BOUNDARY-EVIDENCE.md b/skills/debug/references/BOUNDARY-EVIDENCE.md deleted file mode 100644 index 35ad9aa..0000000 --- a/skills/debug/references/BOUNDARY-EVIDENCE.md +++ /dev/null @@ -1,23 +0,0 @@ -# Boundary Evidence - -Use this reference when the fault crosses components, processes, services, environments, or an authorised production boundary. The goal is to identify the earliest boundary whose observed output or invariant diverges, using the least sensitive evidence that distinguishes the hypotheses. - -## Trace one event - -Map the smallest relevant path and select one request, event, job, or transaction. Preserve its correlation identifier and environment or version context. At only the boundaries that distinguish the leading hypotheses, record applicable input, output, state, configuration, timing, status, and correlation. - -Apply **backward causal tracing** from the symptom until the earliest invalid transition is visible. Continue inside that component rather than widening instrumentation across the whole system. Treat missing, sampled, or uncorrelated telemetry as an evidence limitation rather than a healthy verdict. - -## Capture and replay safely - -Capture the smallest authorised request, event, trace, payload, or state slice. Redact secrets and unnecessary personal or production data before retaining it; preserve an external pointer or correlation identifier when the workspace is not authorised to store the artefact. - -Replay at the nearest stable seam only when the replay preserves the hypothesis-relevant environment, ordering, dependencies, identity, and state. Record every fidelity gap. When replay would erase the suspected cause, prefer a targeted external probe or temporary boundary instrumentation within the user's authority. - -When the next discriminating observation requires new access, privileged instrumentation, or production mutation, leave implementation unchanged and report the required evidence and authority as a blocker. - -## Clean up and verify - -Tag temporary probes so they can be found mechanically. Remove them after diagnosis, or retain them deliberately as documented diagnostics with an explicit data and access boundary. Re-run the original correlated event or nearest faithful reproduction after repair, then verify focused regression and repository checks. - -If a repair does not move the earliest divergent boundary as predicted, record the result and revert the attempt before revising the causal path. diff --git a/skills/debug/references/BOUNDARY-ISOLATION.md b/skills/debug/references/BOUNDARY-ISOLATION.md new file mode 100644 index 0000000..0809d81 --- /dev/null +++ b/skills/debug/references/BOUNDARY-ISOLATION.md @@ -0,0 +1,17 @@ +# Boundary Isolation + +Use this reference when a failure crosses a short component, process, service, pipeline, request, or data path with observable boundaries. The result is the earliest boundary where a valid state or invariant becomes invalid. + +## Narrow the path + +Map the smallest relevant path and define the expected input, output, state, and invariant at each material boundary. Follow one request, event, job, transaction, or datum using its correlation identity and environment or version context. + +For a short path, walk boundaries in order. For a large path with trustworthy midpoints, observe the midpoint, keep the failing half, and repeat. Inject or replay known data only when doing so preserves the conditions relevant to the hypothesis. + +Capture only the fields needed to distinguish the current hypotheses. Treat absent, sampled, delayed, or uncorrelated telemetry as an evidence limitation rather than a healthy result. Redact secrets and unnecessary personal or production data; retain an external pointer when the workspace cannot store the artefact safely. + +## Transfer the result + +A divergent boundary localises the fault but does not explain the internal cause. Record the last valid state, first invalid state, correlation evidence, replay fidelity gaps, and components excluded. Then return to the main causal loop and use origin tracking, comparison, or a targeted experiment inside the remaining boundary. + +Stop when one component or transition remains, or when the next observation needs unavailable access or external mutation. Remove owned temporary boundary probes after use. If a repair does not move the earliest divergence as predicted, reject the repair hypothesis. diff --git a/skills/debug/references/CHANGE-BISECTION.md b/skills/debug/references/CHANGE-BISECTION.md new file mode 100644 index 0000000..83427f0 --- /dev/null +++ b/skills/debug/references/CHANGE-BISECTION.md @@ -0,0 +1,17 @@ +# Change Bisection + +Use this reference when the same crisp property is reliably absent in one ordered state and present in another. States may be revisions, versions, configurations, datasets, or ordered inputs. The result is the smallest reliably classified transition. + +## Freeze the classifier + +Define one command or observation that classifies a state as good, bad, or untestable against the original signal. Confirm trustworthy endpoints before searching. A flaky, environment-sensitive, or changed classifier invalidates ordinary bisection; stabilise it or route to nondeterministic guidance first. + +Test the midpoint, retain its exact state and verdict, keep the half containing the transition, and repeat. For Git history, prefer the repository's supported build and test procedure; use `git bisect run` only when the command's exit statuses faithfully encode good, bad, and untestable. Preserve the log and exit bisect state when finished. + +Apply any compatibility adjustment consistently and keep it separate from the states under test. Do not discard unrelated work or rewrite a user's worktree to make historical states runnable; use an isolated worktree or disposable copy when necessary. + +## Interpret the boundary + +Skipped or broken states near the boundary may leave several candidate transitions. Non-monotonic behaviour or interacting changes may make the result misleading. Environment drift can cause an old state to receive the wrong verdict. + +The first bad state is localisation evidence, not automatic root cause. Inspect the semantic difference and run a discriminating experiment that connects it to the causal mechanism. Stop when the transition is reliable or when the remaining untestable range is explicit. diff --git a/skills/debug/references/COMPARATIVE-DEBUGGING.md b/skills/debug/references/COMPARATIVE-DEBUGGING.md new file mode 100644 index 0000000..da6bfbe --- /dev/null +++ b/skills/debug/references/COMPARATIVE-DEBUGGING.md @@ -0,0 +1,15 @@ +# Comparative Debugging + +Use this reference when a credible working version, implementation, configuration, platform, tenant, request, or input can be compared with the failing case. The result is the first material divergence relevant to the defect. + +## Control the comparison + +State why the comparator is trustworthy for the property under investigation. Hold inputs and environment stable where possible and list intentional differences before observing internal state. + +Compare external outcomes first. Then place equivalent checkpoints along both paths and narrow toward the first difference in output, state, dependency, configuration, control flow, or timing. When many differences exist, partition them or combine this technique with bisection. Reduce a differing case when a smaller comparison retains the same divergence. + +## Test the divergence + +Either side can be wrong, both can share a defect, and intentional or nondeterministic differences create noise. A difference therefore produces a causal candidate, not a verdict. + +Record the shared conditions, relevant differences, first divergence, and why later differences are consequences rather than causes. Return to origin tracking or a controlled hypothesis experiment that introduces or removes the divergence and predicts the original signal. Stop when the first explainable divergence is found or the comparator proves unreliable. diff --git a/skills/debug/references/DELTA-DEBUGGING.md b/skills/debug/references/DELTA-DEBUGGING.md new file mode 100644 index 0000000..dfc36ee --- /dev/null +++ b/skills/debug/references/DELTA-DEBUGGING.md @@ -0,0 +1,17 @@ +# Delta Debugging + +Use this reference when a large input, request, configuration, dataset, trace, sequence, state, or change set can be classified automatically or cheaply. The result is a smaller case that preserves the same failure. + +## Reduce against one oracle + +Freeze a classifier that distinguishes the exact failure from a pass and from an invalid or unresolved candidate. Remove a large partition, run the unchanged classifier, retain the removal only when the same failure remains, and increase granularity when a coarse partition cannot be removed. Continue with the reduced case. + +Preserve ordering and dependencies that make a candidate valid. Record unresolved candidates separately; do not treat them as passes or failures. For generated or fuzzed input, retain the seed and original counterexample before reducing it. + +Manual minimisation uses the same discipline: remove one coherent dimension at a time, retest, and keep only evidence-preserving reductions. Do not rewrite the oracle to accommodate a smaller case. + +## Preserve representativeness + +The result is minimal only relative to the tried partitions and granularity; it need not be globally smallest. Interdependent elements can prevent further reduction, and a reduced case can omit the interaction that makes the production scenario representative. + +Stop when no remaining unit can be removed at the chosen granularity without losing the exact signal, or when further reduction is more expensive than investigating the bounded case. Use the reduced case to localise or form hypotheses, but keep the original unminimised signal for repair verification. diff --git a/skills/debug/references/HYPOTHESIS-EXPERIMENTS.md b/skills/debug/references/HYPOTHESIS-EXPERIMENTS.md new file mode 100644 index 0000000..d0941cb --- /dev/null +++ b/skills/debug/references/HYPOTHESIS-EXPERIMENTS.md @@ -0,0 +1,23 @@ +# Hypothesis-Driven Experiments + +Use this reference when more than one causal explanation remains. The result is one observation that changes the ranked hypothesis set. + +## Form the prediction + +For each plausible cause, state the mechanism and the observation it predicts. Keep a small alternative set grounded in the current evidence; a file, line, component, or recent commit without an explanation is only a location. + +Choose the experiment that best partitions the credible alternatives for the least cost and risk. Prefer an invariant check, focused input change, controlled environment difference, intermediate value, or boundary observation over a repair attempt. Hold other material conditions stable. + +Record: + +- **Hypothesis:** the proposed causal mechanism. +- **Prediction:** the result expected if it is true and, where useful, if it is false. +- **Experiment:** the one controlled observation or intervention. +- **Result:** the exact observation, including conflicting or inconclusive data. +- **Implication:** supported, rejected, or not discriminated, plus the next ranked alternatives. + +## Interpret honestly + +An experiment supports only what its controls and oracle distinguish. A symptom disappearing after a code change may reflect masking, altered timing, or a bypassed path. A negative result is useful when the experiment was capable of exposing the proposed mechanism. + +Remove an owned experimental change after its observation unless it becomes the separately authorised causal repair. Stop when one causal account explains the full signal and trigger conditions and proportionate alternatives have been excluded, or when the next discriminating experiment crosses a named evidence, safety, or authority boundary. diff --git a/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md b/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md new file mode 100644 index 0000000..4c4561d --- /dev/null +++ b/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md @@ -0,0 +1,24 @@ +# Instrumentation and Debuggers + +Use this reference when existing evidence cannot distinguish the active hypotheses. Tools are probes for a stated question, not the debugging method itself. The result is the smallest new observation that changes the causal model. + +## Select the probe + +Choose the instrument whose observation matches the question: + +| Question | Probe | +| --- | --- | +| Which local state or branch first violates an invariant? | Conditional breakpoint, watchpoint, or stack inspection | +| Where does one request cross a process or service boundary? | Correlated logs, metrics, trace, or narrow state endpoint | +| Which operation consumes CPU, memory, I/O, locks, or time? | Matching profiler, counter, or targeted runtime trace | +| Does a known fault class occur? | Repository-supported detector such as a race, memory, or query diagnostic | + +Place the fewest probes at boundaries, decisions, and state transitions that partition the hypotheses. Capture only the needed input, output, identity, correlation, time, and environment. Establish measurement overhead when the probe can affect the signal. + +## Preserve the evidence boundary + +Pausing can invalidate real-time or concurrent behaviour; logging can change timing; profiles are samples; multiple diagnostic tools can interfere. Prefer non-stopping tracepoints or record/replay only when the project and platform support them. + +Do not log credentials, tokens, keys, connection strings, or unnecessary personal data. Follow project retention and access rules, and keep production systems read-only. Tag temporary probes so they can be found and removed. + +Stop when the observation supports, rejects, or fails to discriminate the active hypotheses. Remove owned probes and sensitive captures after use, or retain a diagnostic deliberately within project authority. If the probe only produces more undirected data, sharpen the prediction before adding another. diff --git a/skills/debug/references/NONDETERMINISTIC-CONCURRENT.md b/skills/debug/references/NONDETERMINISTIC-CONCURRENT.md new file mode 100644 index 0000000..db65ac2 --- /dev/null +++ b/skills/debug/references/NONDETERMINISTIC-CONCURRENT.md @@ -0,0 +1,24 @@ +# Nondeterministic and Concurrent Faults + +Use this reference when nominally identical inputs produce different outcomes or the failure rate changes with timing, order, isolation, load, logging, randomness, clocks, network, shared state, or scheduling. The result is a controlled failure rate or a replayable causal execution, not one passing retry. + +## Measure and partition + +Run the exact trigger repeatedly. Record attempts, failures, seed, input, order, clock, load, resources, environment, and any captured schedule. Establish a baseline failure rate before changing code. + +Vary one likely source at a time: + +- replay or sweep randomness and generated inputs; +- permute test order and isolate shared files, ports, processes, or state; +- control time and wait for observable conditions rather than guessed delays; +- constrain parallelism or use a repository-supported race detector; +- capture, replay, or systematically explore schedules when tooling permits; or +- control load, resource pressure, network, filesystem, and dependency state. + +Use stress or injected delay only to amplify and capture a failure. Preserve the seed, order, trace, schedule, or invariant violation that makes the execution useful. + +## Interpret probability + +A passing sample does not prove absence. Added sleeps, longer timeouts, logging, and reduced parallelism can hide a race without removing its cause. Race detectors and schedule explorers cover only the behaviour and nondeterminism they observe or control. + +Stop reproduction when a failing execution is replayable or the measured rate is stable enough to compare discriminating experiments. Verify a repair first against the preserved failing execution, then repeat the original trigger under the same controls and report failures per attempts and remaining statistical uncertainty. Reject a repair that does not change the causal execution or measured rate as predicted. diff --git a/skills/debug/references/NONDETERMINISTIC.md b/skills/debug/references/NONDETERMINISTIC.md deleted file mode 100644 index ab0ac8d..0000000 --- a/skills/debug/references/NONDETERMINISTIC.md +++ /dev/null @@ -1,37 +0,0 @@ -# Nondeterministic Faults - -Use this reference for intermittent, flaky, timing-sensitive, order-dependent, or concurrent faults. The goal is a controlled failure rate or replayable causal execution, not one passing retry. - -## Measure the fault - -Run the exact trigger repeatedly and record attempts, failures, seed, order, clock, load, resources, environment, and any captured schedule. A useful signal either reproduces at a stable enough rate to compare experiments or preserves the execution that failed. - -Partition the likely nondeterminism before changing code: - -- randomness or generated input; -- test order, shared state, or leaked resources; -- clock, timeout, asynchronous condition, or event order; -- concurrent access or scheduling; -- load, resource pressure, network, filesystem, or environment. - -Vary one dimension and compare the failure rate or execution. Preserve every failing seed, order, input, and schedule that improves repeatability. - -## Select the experiment - -| Evidence | Technique | Causal evidence | -| --- | --- | --- | -| A seed or generated input controls the failure | Seed replay and counterexample reduction | The same input fails repeatedly and a minimised input preserves the verdict. | -| Test order or shared state is suspected | Order permutation and isolation probe | A specific predecessor, state, or unreleased resource changes the rate. | -| An asynchronous condition is suspected | Condition instrumentation and controlled perturbation | An observable state transition, rather than elapsed delay alone, determines success. | -| Unsynchronised access is possible | Repository-supported **race detector** | Conflicting accesses and their call paths identify the shared state to investigate. | -| A particular interleaving is suspected | Systematic, controlled, or recorded schedule | The captured schedule replays the failure and an alternative schedule discriminates the hypothesis. | - -Use stress, injected delay, parallelism, or load only to amplify and capture a failure. Stress without a retained input, schedule, trace, or invariant violation does not prove a cause, and a long passing run does not prove absence. - -Replace guessed delays with **condition-based waiting** only after evidence identifies the condition whose ordering is causal. A longer timeout that merely reduces the observed rate is not a confirmed repair. - -## Verify the repair - -Run the preserved failing seed, order, schedule, or trace first. Then repeat the original trigger under the same controls and compare failures per attempts; keep attempt counts proportionate to the prior rate and report the remaining uncertainty. Run focused regression and repository checks after the causal signal changes as predicted. - -If a repair does not change the preserved execution or measured rate as predicted, record the result and revert the attempt before testing the next hypothesis. diff --git a/skills/debug/references/ORIGIN-TRACKING.md b/skills/debug/references/ORIGIN-TRACKING.md new file mode 100644 index 0000000..5054b75 --- /dev/null +++ b/skills/debug/references/ORIGIN-TRACKING.md @@ -0,0 +1,23 @@ +# Origin Tracking + +Use this reference when a bad value, state transition, side effect, or decision is observed far downstream from where it arose. The result is the earliest demonstrated valid-to-invalid transition and its responsible conditions. + +## Trace backwards + +Start at the first observed invalid state. Identify its data definitions, callers, controlling conditions, prior state, and external inputs. Move backwards through the call and data flow, checking the relevant invariant at each step, until the last valid and first invalid states are adjacent. + +Choose the smallest project-native observation that can expose the transition: + +- a conditional breakpoint for a local state and condition; +- a watchpoint for an unexpected mutation; +- a targeted trace for a call or effect path; +- record/replay or reverse execution when the platform supports it; or +- a focused dynamic or static slice when many statements may influence the value. + +Record both the responsible operation and the inputs, state, ordering, or control condition that made it harmful. An assignment or stack frame alone is not a causal account. + +## Bound the claim + +Static slices may over-approximate, debugger facilities have platform and threading limits, and tracing can perturb execution. If instrumentation changes the failure, route that evidence through nondeterministic guidance. + +Stop at the earliest invalid transition that explains the downstream symptom. Return the transition and conditions to the hypothesis loop and test whether controlling them changes the original signal as predicted. diff --git a/skills/debug/references/PERFORMANCE-RESOURCE.md b/skills/debug/references/PERFORMANCE-RESOURCE.md new file mode 100644 index 0000000..f3d2c64 --- /dev/null +++ b/skills/debug/references/PERFORMANCE-RESOURCE.md @@ -0,0 +1,25 @@ +# Performance and Resource Faults + +Use this reference when the defect is latency, throughput, CPU, memory, allocation, I/O, contention, queueing, exhaustion, leakage, query cost, or scale behaviour. The result is a measured causal connection between responsible work and the original user-visible regression. + +## Establish a representative signal + +Define the workload, environment, known-good baseline or expected threshold, warm-up, measurement interval, repetitions, and distribution that distinguishes the fault. Control input size and shape, concurrency, cache state, runtime version, machine resources, data volume, and other material conditions. Route a noisy classifier through nondeterministic guidance before trusting it. + +For early systemic triage, inspect utilization, saturation, and errors for each relevant resource. Then choose the measurement that matches the suspected cost: + +| Suspected cost | Evidence | +| --- | --- | +| CPU or call-path work | Sampling or instrumenting profile under the controlled workload | +| Memory or allocation | Heap, allocation, retention, or growth measurement | +| I/O, network, queue, or lock pressure | Resource counter and targeted trace | +| Database execution | Actual plan, cardinality, I/O, locks, and query timing | +| Version or configuration regression | Stable differential benchmark or change bisection | + +Connect the hotspot, plan, wait, saturation, or growth to the user-visible signal through a controlled workload or intervention. A hotspot is where cost was observed, not automatically why the system is slow. + +## Verify like with like + +Record measurement overhead and production cost. Long averages can hide bursts, profiles are samples, and one constrained resource can be traded for another. + +After one causal repair, rerun the identical workload and controls. Compare the full relevant distribution and low-level causal metric, then run functional regression evidence. Stop only when both move as predicted under the original representative workload. A faster result obtained by changing behaviour, workload, environment, or measurement does not verify the repair. diff --git a/skills/debug/references/PERFORMANCE.md b/skills/debug/references/PERFORMANCE.md deleted file mode 100644 index d648a72..0000000 --- a/skills/debug/references/PERFORMANCE.md +++ /dev/null @@ -1,28 +0,0 @@ -# Performance Faults - -Use this reference for latency, throughput, resource, query, or scale regressions. Measure a representative workload before selecting a profiler or changing code. - -## Establish the regression - -Define the workload, environment, expected threshold or known-good baseline, measured outcome, and variance that distinguishes the fault. Control input size, data shape, concurrency, cache state, runtime version, machine resources, and other material conditions. Warm up when the runtime requires it, repeat the measurement, and report the distribution rather than one timing. - -When known-good and known-bad states exist, use automated **bisection** with the same stable classifier. Treat noisy or untestable states explicitly; route a flaky classifier through [Nondeterministic Faults](NONDETERMINISTIC.md) before trusting its boundary. - -## Locate responsible work - -After the controlled benchmark proves the regression, choose the smallest instrument that distinguishes the hypotheses: - -| Suspected cost | Technique | Discriminating observation | -| --- | --- | --- | -| CPU or call-path work | Sampling or instrumenting profiler | The responsible stack or operation accounts for the measured difference. | -| Memory, allocation, I/O, lock, or network pressure | Resource counter or targeted trace | The relevant resource changes with the regression under the same workload. | -| Database work | Query plan and execution measurement | The plan, cardinality, I/O, lock, or execution step explains the difference. | -| Version or configuration change | Bisection or differential benchmark | The first ordered boundary preserves the same performance verdict. | - -Record measurement overhead and side effects. Use transactions or inert fixtures when an execution plan can mutate data. Profile the controlled workload; a profile from a different workload does not explain the measured regression. - -## Repair and verify - -Make one minimal change to the responsible work. Remeasure the identical workload and controls, compare the result and variance with the original baseline, then run functional regression and repository checks. A faster result that changes behaviour, workload, or environment does not verify the repair. - -If the measurement does not change as predicted, record the result and revert the attempt before revising the performance hypothesis. diff --git a/skills/debug/references/TECHNIQUES.md b/skills/debug/references/TECHNIQUES.md deleted file mode 100644 index 60d0ad8..0000000 --- a/skills/debug/references/TECHNIQUES.md +++ /dev/null @@ -1,52 +0,0 @@ -# Debugging Techniques - -Use this reference when the repeatable failing signal or next discriminating experiment is not obvious. Identify the runtime situation, select the smallest technique that predicts a discriminating observation, retain its evidence, then return to the scientific-debugging loop. Combine techniques only when each settles a distinct question. - -## Tighten the Signal - -Run the signal at least once and record its command or probe, input, environment, expected verdict, and observed verdict. A useful signal is: - -- **specific:** it reaches the relevant path and asserts the reported symptom, rather than merely completing without an error; -- **repeatable:** it records the fixture, environment, seed, order, schedule, or captured artefact needed to reproduce the verdict; -- **measurable:** it is deterministic, or reports failures per attempts for a nondeterministic fault; -- **tight:** it removes unrelated setup and runs quickly enough to guide the next experiment; -- **runnable:** the agent can execute it unattended when the environment permits; otherwise it uses repeatable captured evidence or a targeted external probe; -- **safe:** production artefacts are minimised, redacted, and handled within the user's permissions. - -A passing retry does not turn an intermittent failure green. Use [Nondeterministic Faults](NONDETERMINISTIC.md) to control and compare the measured failure rate. - -## Construct a Signal - -| Situation | Technique | Observable verdict and retained artefact | -| --- | --- | --- | -| A stable test seam reaches the fault | Focused failing test or minimal harness | Assert the exact behaviour and preserve the smallest fixture. It is a candidate TDD Red only when it independently specifies the desired behaviour and satisfies every TDD prerequisite. | -| The fault is at an HTTP boundary | HTTP request script | Assert the relevant status, body, and headers; retain a redacted request and response rather than relying only on process exit. | -| The fault is a CLI contract | CLI invocation with fixture input | Assert exit status, stdout, and stderr as applicable; record flags, working directory, and relevant environment. | -| Only a production request or event exposes the fault | Capture and replay | Use [Boundary Evidence](BOUNDARY-EVIDENCE.md) to capture the smallest authorised event and assess replay fidelity. | -| The bad input is unknown or combinatorial | Property or fuzz loop | State an executable invariant, preserve the seed and failing input, then minimise the counterexample before diagnosis. | -| The failure is intermittent, order-dependent, concurrent, or timing-sensitive | Repetition, race detector, or controlled schedule | Use [Nondeterministic Faults](NONDETERMINISTIC.md) to measure the rate, partition nondeterminism, and capture a causal execution. | -| The fault crosses components or environments | Correlated boundary probe | Use [Boundary Evidence](BOUNDARY-EVIDENCE.md) to observe only the boundaries that distinguish the hypotheses. | -| The fault is performance | Controlled benchmark | Use [Performance Faults](PERFORMANCE.md) to establish a representative baseline and threshold before profiling. | -| The fault requires browser behaviour | Headless browser assertion or trace | Assert the relevant DOM, accessibility, console, request, response, screenshot, or timing outcome; retain the smallest trace and fixture. Use a structured human-in-the-loop transcript only when automation cannot perform or observe the essential step. | - -## Select an Experiment - -| Evidence | Experiment | Discriminating result | -| --- | --- | --- | -| A failing input, configuration, or sequence can be reduced | **Minimal reproducible example** or **delta debugging** | Remove partitions while preserving the exact verdict; the remaining elements bound the causal search space. | -| Known-good and known-bad states form an ordered space | **Binary search** or automated bisection | Use a stable good, bad, and untestable classifier across commits, versions, datasets, inputs, or configurations; record the first boundary found. | -| A comparable case works | **Differential testing** | Run the same input through both cases and isolate the smallest output, state, dependency, or configuration difference. Treat a difference as evidence to test, not proof by itself. | -| The symptom appears far from the bad value or action | **Backward causal tracing** | Follow the call and data flow from symptom to the earliest divergence, recording where the value entered and which invariant first failed. | -| The system crosses process or component boundaries | **Boundary instrumentation** | Follow [Boundary Evidence](BOUNDARY-EVIDENCE.md) and record hypothesis-relevant observations at the few boundaries that distinguish the candidates. | -| A value changes unexpectedly during execution | **Breakpoint**, **watchpoint**, or targeted trace | Pause at the earliest mutation or invariant violation and capture the responsible call path and state. | -| Timing, order, or scheduling is suspected | Controlled perturbation | Follow [Nondeterministic Faults](NONDETERMINISTIC.md) and vary one dimension while comparing the measured rate or captured schedule. | -| A controlled benchmark proves a regression | Profiler, query plan, or resource trace | Follow [Performance Faults](PERFORMANCE.md), identify the responsible work, and remeasure the identical workload after repair. | - -An experiment is complete when its observation confirms or rejects a stated hypothesis. If it only produces more data, sharpen the prediction or choose a different experiment. - -## Preserve Useful Evidence - -- Promote a minimised reproducer to regression protection at the strongest stable seam only when it independently specifies desired observable behaviour. Invoke `$tdd` only when its complete prerequisite holds. -- Remove temporary instrumentation and throwaway harnesses after use, or retain them deliberately as documented diagnostics. Tag temporary probes so cleanup is mechanically checkable. -- Store only sanitised captures and fixtures that the repository is authorised to retain. Report external artefacts without copying sensitive data into the workspace. -- Preserve the original reproduction across repair attempts. When a repair fails its prediction, retain the observation but revert the attempt before testing the next hypothesis. From 85b11099b4006b51f92abad1b9b41d64a936e550 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 17:55:14 +0100 Subject: [PATCH 70/79] refactor(skills): require explicit invocation --- SKILLS_PLAN.md | 4 +--- skills/code-review/SKILL.md | 4 +--- skills/code-review/agents/openai.yaml | 2 +- skills/modular-design/SKILL.md | 2 +- skills/modular-design/agents/openai.yaml | 2 +- skills/quality-harnesses/SKILL.md | 2 +- skills/quality-harnesses/agents/openai.yaml | 2 +- skills/tdd/SKILL.md | 2 +- skills/tdd/agents/openai.yaml | 2 +- 9 files changed, 9 insertions(+), 13 deletions(-) diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md index 5d1b9b1..f1917ed 100644 --- a/SKILLS_PLAN.md +++ b/SKILLS_PLAN.md @@ -40,9 +40,7 @@ These areas are entry points, not mandatory phase gates. Release, deployment and ## Session and authority model -Lifecycle skills are explicitly user-invokable. A lifecycle skill may call another lifecycle skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture an incomplete work definition. - -Reusable utilities are user-invokable and may be model-invoked when their documented trigger is present. A caller remains responsible for verifying the utility's result and completing its own outcome. +Propulsion skills are explicitly user-invokable. Only `elicit` may also be model-invoked when its documented trigger is present. Entry skills fan out through their documented composition; a skill may call another skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture an incomplete work definition. A caller remains responsible for verifying a composed skill's result and completing its own outcome. Invoking a skill grants authority for its expected repository and external mutations once the skill has reached shared understanding. Material product, architecture, priority or scope questions remain with the user, but an agreed decision does not require a second permission prompt before it is applied. Work outside the invoked outcome remains unauthorised. diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index d25fa6e..40355f6 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,9 +1,7 @@ --- name: code-review description: Reviews a fixed code change through isolated specification and engineering-standards agents. Use for direct diff review or when another delivery skill invokes independent review. -metadata: - invocation: model -disable-model-invocation: false +disable-model-invocation: true --- # Code Review diff --git a/skills/code-review/agents/openai.yaml b/skills/code-review/agents/openai.yaml index 6a11c8e..6c2a385 100644 --- a/skills/code-review/agents/openai.yaml +++ b/skills/code-review/agents/openai.yaml @@ -2,4 +2,4 @@ interface: display_name: 'Code Review' short_description: 'Independently review a fixed code change' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index dedb864..e005c07 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -1,7 +1,7 @@ --- name: modular-design description: Teaches structural design through information hiding, deep modules, cohesive ownership, stable interfaces, and change locality. Use when writing or assessing code, boundaries, contracts, dependencies, seams, or change propagation. -disable-model-invocation: false +disable-model-invocation: true --- # Modular Design diff --git a/skills/modular-design/agents/openai.yaml b/skills/modular-design/agents/openai.yaml index 19cb267..d97ee31 100644 --- a/skills/modular-design/agents/openai.yaml +++ b/skills/modular-design/agents/openai.yaml @@ -2,4 +2,4 @@ interface: display_name: 'Modular Design' short_description: 'Design deep modules with local change' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false diff --git a/skills/quality-harnesses/SKILL.md b/skills/quality-harnesses/SKILL.md index 021d4cf..d4ab88c 100644 --- a/skills/quality-harnesses/SKILL.md +++ b/skills/quality-harnesses/SKILL.md @@ -1,7 +1,7 @@ --- name: quality-harnesses description: Teaches risk-triggered selection and honest interpretation of software quality harnesses for a fixed change. Use when implementing or assessing code, tests, configuration, migrations, generated artefacts, or other software changes. -disable-model-invocation: false +disable-model-invocation: true --- # Quality Harnesses diff --git a/skills/quality-harnesses/agents/openai.yaml b/skills/quality-harnesses/agents/openai.yaml index f869ab1..c04dca6 100644 --- a/skills/quality-harnesses/agents/openai.yaml +++ b/skills/quality-harnesses/agents/openai.yaml @@ -2,4 +2,4 @@ interface: display_name: 'Quality Harnesses' short_description: 'Select risk-matched software quality evidence' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 0160a1a..7f1b48a 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -1,7 +1,7 @@ --- name: tdd description: Teaches test-driven development through red-green-refactor with durable behavioural tests. Use when a usable suite can exercise changed behaviour through a credible seam. -disable-model-invocation: false +disable-model-invocation: true --- # Test-Driven Development diff --git a/skills/tdd/agents/openai.yaml b/skills/tdd/agents/openai.yaml index bbba392..d3df8d3 100644 --- a/skills/tdd/agents/openai.yaml +++ b/skills/tdd/agents/openai.yaml @@ -2,4 +2,4 @@ interface: display_name: 'Test-Driven Development' short_description: 'Build durable behaviour through red-green-refactor' policy: - allow_implicit_invocation: true + allow_implicit_invocation: false From 99af05f4a315573c93e2d7a9dde8b9a7c4b2022c Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 18:58:29 +0100 Subject: [PATCH 71/79] docs: document skill entry points --- README.md | 153 +++++++++++++++++++++++++++++++----------------------- 1 file changed, 87 insertions(+), 66 deletions(-) diff --git a/README.md b/README.md index 9346ebb..2cb34d5 100644 --- a/README.md +++ b/README.md @@ -4,7 +4,7 @@ Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports the realistic lifecycle around coding—from establishing a product and defining features to creating tickets, implementation, and debugging. -Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Propulsion does not require an end-to-end orchestrator, manifest, setup framework, state machine, or tracker adapter. +Each skill runs one bounded working session and stops with an independently useful outcome. You can enter wherever the necessary inputs already exist and leave when that outcome is complete. Substantial work narrows breadth-first over several sessions: product intent becomes a decision-complete feature specification, then implementation-ready tickets, and finally one implemented body of work per ticket. Small, understood work can begin directly with `$implement`. @@ -31,122 +31,143 @@ ln -s /absolute/path/to/propulsion/skills/elicit ~/.agents/skills/elicit Repeat the link for each selected skill. Codex and OpenCode both discover skills from `~/.agents/skills`; edits in the clone are available through the links without reinstalling or publishing a new version. -## Where to start - -| When you need to… | Start with… | Session outcome | -| --- | --- | --- | -| Establish or revise a product | `$define-product` | Confirmed product and system requirements | -| Define a substantial feature | `$specify-feature` | An approved, decision-complete feature specification | -| Make a small understood change | `$implement` | A minimal, reviewed, verified local change | -| Diagnose or repair a defect | `$debug` | A causal diagnosis and, when requested, a verified repair | -| Review another contributor's PR | `$review-pull-request` | Independent findings and verification evidence | - ## Lifecycle skills -The lifecycle areas are useful entry points, not mandatory phase gates. +The lifecycle areas below are independent entry points, not mandatory phase gates. Invoke the skill for the outcome you need, later stages do not start automatically. ### Establish -- `$define-product` turns a new or existing product into the durable breadth-first product and system requirements foundation in `PRODUCT.md`, while `elicit-with-context` maintains applicable project language and qualifying decision records through its own contract. +#### `$define-product` + +Enter here to establish or deliberately revise the product and system requirements foundation: + +```text +$define-product +``` + +The session produces a user-confirmed `PRODUCT.md` and the corresponding project language in `CONTEXT.md`. It stops before defining individual features, selecting feature-specific architecture, or creating tickets. ### Define -- `$specify-feature` turns one high-level request into a single feature specification containing approved feature intent and its selected buildable solution. +#### `$specify-feature` -### Plan +Enter here when a high-level feature request needs enough behavioural and technical definition for planning: -- `$create-tickets` turns an approved work definition—either a sufficiently complete document or confirmed conversation—into agreed, verified, vertically sliced tickets with Fibonacci complexity. +```text +$specify-feature +``` -### Refine +The session produces one approved, decision-complete feature specification containing visibly distinct feature intent and selected solution sections. No earlier lifecycle artefact is mandatory, and the skill stops before ticket creation or implementation. -- `$review-architecture` reports a few high-value improvements for an explicitly bounded area without changing it or creating tickets. +### Plan -### Deliver +#### `$create-tickets` -- `$implement` produces one minimal, reviewed, verified local change from a ticket or other confirmed work. -- `$commit` records eligible reviewed work as coherent Conventional Commits while preserving unrelated changes. -- `$pull-request` commits eligible remaining work when needed, pushes the branch, and creates or updates one accurate pull request for the whole branch. -- `$review-pull-request` independently assesses someone else's pinned pull request without changing their branch or publishing a review unless requested. +Enter here when an approved document or confirmed conversation is complete enough to decompose without inventing behaviour or material solution decisions: -The author's normal delivery flow does not invoke `$review-pull-request`: `$implement` already includes independent code review. The separate pull-request review skill is for work submitted by others. +```text +$create-tickets +``` -### Debug +The session creates and verifies small, vertically sliced, dependency-aware tickets with Fibonacci complexity in the destination named by the project's root `AGENTS.md`. It stops before scheduling or implementing them. -- `$debug` establishes an evidence-backed root cause and, when repair is requested, produces a minimal reviewed and verified local change. +Local Markdown destinations use one file per ticket under `docs/features//tickets/`. External destinations use their native items, estimates, relationships, and statuses. If `AGENTS.md` does not name the destination, the skill asks and records it. -## Representative flows +### Refine -Establish a product: +#### `$review-architecture` + +Enter here to assess an explicitly bounded part of the current architecture: ```text -define-product +$review-architecture ``` -Deliver a substantial feature: +The session produces an evidence-backed report containing a small number of prioritised, high-value improvements and their trade-offs. It does not refactor the code or create speculative tickets. + +### Deliver + +#### `$implement` + +Enter here with one ticket or another small, confirmed body of work: ```text -specify-feature - → later create-tickets - → implement each selected ticket in a separate session - → commit whenever a coherent unit is ready - → pull-request when the branch is ready +$implement ``` -Deliver a small understood change: +The session produces a minimal local change with retained tests, applicable quality evidence, and completed independent review. Small, understood work can begin here without a product definition, feature specification, or ticket. Implementation does not commit or publish the change. + +`$implement` applies `$modular-design` and `$quality-harnesses`, uses `$tdd` for behaviour-changing work when a usable suite can exercise it, and requires an independent `$code-review`. It uses existing project infrastructure and reports unavailable evidence instead of installing unrelated tooling or manufacturing confidence. + +#### `$commit` + +Enter here when eligible reviewed work is ready to be recorded: ```text -implement → commit → pull-request +$commit ``` -Debug a known defect: +The session creates and verifies coherent Conventional Commits while preserving unrelated staged, unstaged, and untracked work. It does not push or open a pull request. + +#### `$pull-request` + +Enter here when the current branch is ready to publish for human review: ```text -debug → commit → pull-request +$pull-request ``` -Review another contributor's work: +The session commits eligible remaining work when needed, pushes the branch, and creates or updates one pull request describing the complete branch. It does not merge, release, or deploy the work. + +A substantial feature commonly moves through separate sessions: ```text -review-pull-request +$specify-feature + → later $create-tickets + → $implement each selected ticket + → $commit whenever a coherent unit is ready + → $pull-request when the branch is ready ``` -## Reusable utilities +A small, understood change may start later: -Utilities are independently invokable and are also composed by lifecycle skills when their trigger applies. +```text +$implement → $commit → $pull-request +``` -| Utility | Use it to… | -| --- | --- | -| `$elicit` | Resolve material user-held information and decisions one question at a time | -| `$elicit-with-context` | Resolve software-project questions while maintaining shared project language | -| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | -| `$maintain-ubiquitous-language` | Keep confirmed project-specific language current in `CONTEXT.md` | -| `$maintain-decision-records` | Preserve rare accepted decisions whose rationale warrants an ADR | -| `$modular-design` | Assess boundaries, ownership, contracts, dependencies, and change propagation | -| `$tdd` | Implement observable behaviour through red-green-refactor when its prerequisites hold | -| `$verify-change` | Run applicable project and risk-triggered quality harnesses against a fixed change | -| `$code-review` | Independently inspect a fixed change for behaviour, regressions, minimality, conventions, and maintainability | -| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | +### Debug -Software-project skills route questions through `$elicit-with-context`. Base `$elicit` remains available for non-software work and as the router's questioning component. +#### `$debug` -`$write-skill` develops and evaluates Agent Skills themselves; it is not part of the software-delivery lifecycle. +Enter here with observed and expected behaviour or another usable failure signal: -## Ticket destinations +```text +$debug +``` -The root `AGENTS.md` explicitly names the project's ticket destination, such as Linear or local Markdown. If none is named, `$create-tickets` asks the user, records the answer through `$maintain-agents`, and resumes. Local tickets use one file each under `docs/features//tickets/`; external destinations use their native items, estimates, relationships, and statuses. The skill never guesses, silently falls back, or introduces a Propulsion adapter layer. +The session establishes an evidence-backed root cause and stops there when diagnosis is the requested outcome. When repair is authorised, it produces a minimal, reviewed local change with regression evidence and stops before commit or publication. -## Quality during implementation +```text +$debug → $commit → $pull-request +``` -`$implement` uses TDD when an existing runnable suite can exercise the change through a meaningful observable boundary. It always subjects the fixed candidate to independent `$code-review` and `$verify-change`, using the repository's required checks plus additional harnesses only when the change's risks justify them. +## Utility skills -The workflow does not install unrelated test or quality infrastructure merely to satisfy its process. When meaningful evidence is unavailable, it reports the limitation instead of manufacturing confidence. Implementation stops with a reviewed, verified local change; committing and publication remain separate user-controlled sessions. +These skills are independently invokable outside the main lifecycle path and may also be composed by lifecycle skills when their trigger applies. -## Scope +| Skill | Invoke it to… | +| --- | --- | +| `$review-pull-request` | Independently assess someone else's pinned pull request without changing their branch or publishing a review unless requested | +| `$elicit` | Resolve material user-held information and decisions one question at a time | +| `$elicit-with-context` | Resolve software-project questions through `$elicit`, applying `$maintain-ubiquitous-language` and `$maintain-decision-records` when their triggers apply | +| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | +| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | -Propulsion does not release, deploy, merge, or continuously monitor software. Those activities remain project-specific. It also avoids duplicating discoverable code, configuration, CI, tracker, or production state in prose documentation. +Use `$elicit-with-context` to work through questions or decisions within a software project. For non-software work, use `$elicit`. ## Acknowledgements -Propulsion is heavily inspired by: +Propulsion is heavily inspired by other great skill sets: -- [mattpocock/skills](https://github.com/mattpocock/skills) for brevity, wording discipline, and the question-by-question discovery style +- [obra/superpowers](https://github.com/obra/superpowers) +- [mattpocock/skills](https://github.com/mattpocock/skills) From 087ead820e59cec54bd5e05406f03e87995db8b0 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 19:03:55 +0100 Subject: [PATCH 72/79] docs: remove unused terminology --- CONTEXT.md | 4 ---- 1 file changed, 4 deletions(-) diff --git a/CONTEXT.md b/CONTEXT.md index d72d9c6..384e58a 100644 --- a/CONTEXT.md +++ b/CONTEXT.md @@ -12,10 +12,6 @@ **Invocation policy**: Client-specific metadata within a skill bundle that controls whether an agent may select that skill implicitly. The same intent may require different policy fields in different clients. -**Skill-only distribution**: Distributing Propulsion directly as Agent Skills, using the skills installer for remote sources and filesystem links for local sources. Propulsion does not require client-specific plugin packaging. - ### Composition -**Composable skill**: An independently useful skill that may be invoked alone or coordinated by another skill without requiring the rest of a prescribed workflow. - **Router skill**: A lightweight skill that invokes other skills to produce a combined outcome. It contains only coordination unique to that outcome; called skills remain authoritative and the router neither repeats nor overrides their context. From 8aad1722d19b1424c3624fec1558ca712e60c8d8 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 19:12:38 +0100 Subject: [PATCH 73/79] chore(skills): remove completed plans --- IMPLEMENTATION_SKILLS_PLAN.md | 439 ---------------------------------- SKILLS_PLAN.md | 362 ---------------------------- 2 files changed, 801 deletions(-) delete mode 100644 IMPLEMENTATION_SKILLS_PLAN.md delete mode 100644 SKILLS_PLAN.md diff --git a/IMPLEMENTATION_SKILLS_PLAN.md b/IMPLEMENTATION_SKILLS_PLAN.md deleted file mode 100644 index 019cf1b..0000000 --- a/IMPLEMENTATION_SKILLS_PLAN.md +++ /dev/null @@ -1,439 +0,0 @@ -# Implementation skills plan - -## Status and authority - -This is the authoritative authoring brief for Propulsion's implementation skill set. [`SKILLS_PLAN.md`](SKILLS_PLAN.md) owns the wider lifecycle catalogue and each skill's public outcome. This document owns the detailed contract, composition, teaching coverage, resource expectations and acceptance evidence for: - -- `implement` -- `tdd` -- `modular-design` -- `quality-harnesses` -- `code-review` - -The existing skill bundles are baseline evidence only. They may be rewritten and their resources replaced, split, combined or removed during their individual `$write-skill` sessions. Each authoring session must still establish matched baselines and prove that its final bundle is the smallest version that produces the behaviour fixed here. - -The evidence basis is [`docs/research/agentic-coding-quality-guardrails-2026-08-20.md`](docs/research/agentic-coding-quality-guardrails-2026-08-20.md). That report supports a relevant portfolio of independent evidence routes rather than a universal gauntlet, score or claim that passing tools proves correctness. - -## Intended outcome - -Together these skills let an agent implement agreed software with enough structural guidance, test discipline, quality evidence and independent challenge that a human can manage the work from its requirements and evidence rather than line-by-line code inspection. - -The system optimises for the **smallest coherent change**: the least production code, test code and supporting machinery that completely satisfies the agreed behaviour while remaining understandable and economical to change. Fewest lines or files is not the target. A small extraction may be necessary to localise knowledge; speculative abstractions, extension points, duplicate paths and unrelated cleanup are not. - -## Composition model - -`implement` owns the implementation session. `tdd`, `modular-design` and `quality-harnesses` are **teaching skills**: they add applicable knowledge to their caller without becoming delegated workers or owning separate workflows. `code-review` is different because context isolation is part of its outcome; it coordinates independent fresh-agent review and returns suggestions without changing the candidate. - -```text -implement - applies modular-design throughout - applies tdd when behaviour changes and a usable suite exists - applies relevant quality-harnesses guidance - builds and locally checks one complete candidate - invokes code-review - Spec fresh agent when behavioural authority is available - Standards fresh agent with the teaching skills - validates and adjudicates both sets of suggestions - applies supported changes and repeats from the affected step -``` - -Teaching skills state what their caller must understand and apply. They do not claim separate mutation authority, hand back a workflow verdict or require the caller to restate their internal instructions. Their direct composition must preserve: - -- the caller's fixed outcome and authority; -- the teaching skill's applicable trigger and knowledge; -- the caller's ownership of edits, commands, evidence and stopping; -- no duplicated internal procedure in the caller; and -- no hidden advancement into commit, publication, release or deployment. - -## Shared quality rules - -All five skill contracts preserve these rules: - -1. **Independent oracle before confidence.** A requirement, accepted example, protocol, invariant, trusted reference, prior accepted behaviour or explicit decision must be capable of disagreeing with the implementation. Implementation-derived expectations and regenerated outputs are not independent evidence. -2. **Behaviour before representation.** Tests and Spec review observe supported outcomes, contracts, state transitions and effects rather than private methods, internal call graphs, incidental DOM structure or storage layout unless that representation is itself promised. -3. **Relevant evidence rather than control count.** Apply each repository requirement and each catalogue harness whose trigger matches the changed risk. Do not run irrelevant tools to create confidence theatre. -4. **Metrics prompt investigation.** Coverage, mutation, complexity, size, churn, coupling and smell counts inform named questions. They are not universal thresholds, composite scores or substitutes for judgement. -5. **Measurement paths are part of the change.** Tests, fixtures, snapshots, baselines, suppressions, exclusions, quality configuration, generators, scripts and thresholds require the same scrutiny as production code. -6. **Unavailable evidence stays visible.** Missing tools, credentials, services, platforms, representative data or specialist judgement are limitations, neither passes nor evidence against the implementation. -7. **Project authority wins.** Repository instructions, accepted specifications, architecture decisions, framework conventions and configured checks override generic guidance where they conflict. -8. **No implicit infrastructure programme.** A caller may improve evidence through existing infrastructure. It does not add a project-wide framework or quality tool unless establishing that capability is explicitly part of the agreed work. - -## `implement` - -### Recurring job - -Turn one implementation-ready ticket or small confirmed body of work into a minimal local change whose applicable tests and quality checks pass and whose current candidate has completed independent Spec and Standards review. - -### Triggers and non-triggers - -Use for a ticket, bug repair handed off from diagnosis, or small agreed software change that is sufficiently defined to build. A ticket or feature document is not mandatory when the request itself is clear. - -Do not use it to discover product direction, conduct exploratory research, create specifications or tickets, commit, publish, release or deploy. It does not invoke `research`; implementation should already have its material external knowledge. If implementation exposes a missing behavioural, contract, architecture, scope or acceptance decision, invoke `elicit-with-context`. That skill owns any resulting ubiquitous-language or decision-record maintenance. - -### Required sequence - -1. **Fix the implementation basis.** Inspect the work authority, repository instructions, current code, tests, relevant contracts and complete repository state. State the observable outcome, exclusions, material risks and completion evidence. Resolve discoverable facts directly; use `elicit-with-context` only for a material user-held gap. -2. **Load the teaching knowledge.** Apply `modular-design` and `quality-harnesses`. Apply `tdd` when the work changes observable behaviour and a usable test suite exists. Keep the knowledge active throughout implementation rather than treating it as a final checklist. -3. **Select the smallest coherent slice.** Trace the next production addition to one current requirement, caller or necessary support responsibility. Use a tracer bullet for the smallest end-to-end observable result. Avoid speculative flexibility, unrelated cleanup and broad refactoring. -4. **Implement test-first when applicable.** Let the caller perform the Red–Green–Refactor cycle taught by `tdd`, carrying the modular constraints and relevant quality risks into test and production design. For work outside TDD's trigger, use the smallest project-native feedback path without claiming TDD. -5. **Maintain structural locality.** Keep each rule with its coherent owner, retain or establish the smallest justified interface, and perform conditional Design It Twice before a material new or changed boundary is fixed. -6. **Apply relevant harnesses.** Review the quality catalogue, load the detailed references for every matching entry, run or perform the applicable repository-native harnesses, and preserve their exact evidence and limitations. Do not silently replace an unavailable harness with a weaker check. -7. **Inspect the complete candidate.** Read the entire diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact and measurement-path change. Remove superseded paths, dead helpers, duplication and speculative code made unnecessary by the final design. -8. **Rationalise retained tests.** After production refactoring has stayed green with unchanged tests, review the suite for obsolete, duplicative, brittle or implementation-coupled cases. Merge, rewrite or delete a test only when every unique promised behaviour and oracle remains protected and the revised suite still detects the realistic defects it is meant to catch. -9. **Freeze and review.** Freeze one exact, complete, locally checked candidate and invoke `code-review`. Do not send an incomplete slice or changing target. -10. **Adjudicate objectively.** Validate each suggestion against its authority, the code and the strongest benign interpretation. Classify it as a required correction, proportionate improvement, rejected finding or user decision. Consequence rank orders attention but never decides the classification. -11. **Correct and repeat.** Apply required corrections and worthwhile in-scope improvements using the teaching skills. Ask the user only when resolution changes agreed behaviour, architecture or scope. Rerun every affected test and harness. Material candidate changes require a new frozen review; continue until no applicable change remains. -12. **Hand off locally.** Report delivered behaviour, changed artefacts, applicable TDD and modular-design decisions, checks and results, unavailable evidence, residual risks, both review results and every disposition. Stop without committing or publishing. - -### Completion boundary - -Successful completion requires: - -- the agreed behaviour exists locally; -- every production addition has a current purpose; -- TDD was applied wherever its trigger held; -- relevant project and catalogue harnesses passed or their unavailable evidence is explicit; -- the current candidate, not an earlier revision, received applicable Spec and Standards review; -- every suggestion was validated and no applicable correction or accepted improvement remains; and -- no in-scope failure is hidden by a weakened test, threshold, baseline, exclusion or suppression. - -### Authoring and evaluation evidence - -The future `$write-skill` session must test at least: - -- a clear ticket needing no elicitation or research; -- a material specification gap discovered during implementation; -- a behaviour change with a usable suite; -- a non-behavioural change for which TDD must not be manufactured; -- a repository missing a relevant harness; -- a review containing a high-ranked false positive and a lower-ranked required correction; -- a material correction that invalidates the previous review; and -- unrelated dirty work that must remain untouched. - -## `tdd` - -### Recurring job - -Teach a caller to implement observable behaviour through disciplined Red–Green–Refactor and leave a small, durable suite whose oracles can detect realistic defects while surviving behaviour-preserving structural change. - -### Applicability - -Apply when both conditions hold: - -1. the work adds, changes or repairs observable behaviour; and -2. an existing usable test suite can exercise that behaviour through a credible seam. - -Do not apply to documentation, human prose, dependency-only updates, configuration-only maintenance without an exercisable effect, generated artefacts or other non-behavioural changes. When behaviour changes but no usable suite exists, the caller proceeds without claiming TDD, uses the strongest available project-native feedback and reports the evidence limitation. Creating a suite or framework is permitted only when it is explicitly part of the agreed work. - -### Required teaching - -#### Establish the slice, seam and oracle - -- Start from one observable outcome and use a **tracer bullet** to select the smallest complete vertical slice. -- Choose the narrowest seam that exposes the promised result, hides structure allowed to change and retains the production risk that matters. -- Derive expected results independently before Green. Prefer requirements, accepted worked examples, contracts, invariants, trusted references, separate models or accepted prior behaviour. -- Treat current output capture as characterization, not proof of correctness. -- Name the realistic production break each retained test must catch and the hidden structural changes it must survive. - -#### Red - -- Write one focused behavioural test before the corresponding production change. -- Arrange only necessary state, act through the supported seam, assert the complete promised outcome and material unchanged state, then clean up acquired resources. -- Run the focused test and require the intended behavioural failure. A pass, syntax error, fixture failure or environment error is not Red. -- For bugs, reproduce the incorrect result and assert the desired one; a crash or implementation-derived expectation is not automatically a valid regression oracle. - -#### Green - -- Add only enough production code to satisfy the current behaviour. -- Do not anticipate later cases, introduce hypothetical options or refactor unrelated code. -- Run the focused test and relevant nearby tests. Fix production code rather than weakening the test when its independent oracle remains valid. - -#### Refactor - -- Refactor production design only while the behaviour is green. -- Keep tests unchanged during behaviour-preserving production refactoring. If hidden structural movement breaks a test, move its observation back to the promised outcome. -- Improve naming, cohesion, duplication and ownership without adding behaviour. -- Rerun the focused tests after each material step. - -#### Rationalise the suite - -- After production refactoring, reassess whether every retained test still protects a unique behaviour, boundary, invariant or risk. -- Remove duplicate examples that add no distinct failure detection; combine related cases with table-driven or property forms when that makes the oracle clearer. -- Remove obsolete implementation-detail assertions and broad snapshots whose only value is change detection. -- Preserve readable specification value and failure diagnosis; fewer tests are better only when no unique protection is lost. -- Apply a mental mutation check to wrong constants, wrong branches, missing effects, empty/default results and missing validation. Use actual mutation or coverage tooling only through applicable `quality-harnesses` guidance. - -### Required example coverage - -The final bundle must use concrete good/bad pairs to teach: - -- an independent literal or invariant versus a tautological expected value; -- a public result versus a private method or internal call graph; -- user-visible UI semantics versus CSS classes, wrapper elements or child indexes; -- persistence through supported write/read behaviour versus private schema inspection; -- a real internal collaborator versus unnecessary mocking; -- a focused fake, stub or mock at an uncontrollable external boundary; -- legitimate protocol interaction assertions versus incidental call choreography; -- a failing bug regression before its repair; -- deterministic time, randomness and cleanup; -- property, model-based, metamorphic and differential tests with their triggers and oracle limitations; and -- consolidation of several brittle examples into a smaller behaviourally equivalent test form. - -Examples should be small and concrete enough for an agent to imitate while explaining the decision independently of one language or framework. Core every-run examples belong on the loaded path; conditional techniques may use direct references. - -### Ownership-aware test-first rule - -When the current agent writes production behaviour before demonstrating Red, discard only that agent-owned implementation and restart test-first. Never delete or overwrite pre-existing or user-owned work. For an existing partial implementation, demonstrate sensitivity against the pre-change revision, a safely disabled behaviour or a controlled known-bad variant. When that cannot be done, label the retained test accurately as characterization or regression evidence rather than claiming TDD. - -### Expected resources - -The authoring session should evaluate a concise every-run `SKILL.md` plus directly linked references covering test design and oracles, doubles and boundaries, conditional generative techniques, and suite rationalisation. Existing reference names are not constraints; merge or split them only when their loading triggers differ materially. - -### Authoring and evaluation evidence - -Forward tests must distinguish this skill from generic “write a failing test first” behaviour. Cases must expose tautological tests, internal mocks, CSS/DOM coupling, invalid Red signals, overbuilt Green code, tests changed during refactoring, duplicated post-refactor tests, absence of a usable suite and pre-existing user code that must not be deleted. - -## `modular-design` - -### Recurring job - -Teach a caller to place behaviour and knowledge in deep, cohesive modules with small stable interfaces so current and plausible evidenced changes remain local. - -### Governing doctrine - -Use **information hiding** and **deep modules** as the backbone: - -- A module is any cohesive capability with an interface and implementation, regardless of whether code expresses it as a function, class, package, process or tier-spanning slice. -- Its interface includes everything callers must know: operations, data, invariants, ordering, errors, configuration, effects and material performance characteristics. -- A deep module gives callers substantial capability through a comparatively simple interface. -- Hide volatile design decisions and keep each material rule, invariant, protocol or representation with one coherent owner. -- Code that changes for the same evidenced reason should live together; one conceptual change should not require scattered knowledge across callers. -- Prefer established language, framework and repository idioms. An interface in the design sense does not require a language `interface`, class hierarchy or wrapper. - -### Required teaching - -#### Assign ownership - -Identify the behaviour, rule, invariant, protocol, representation, sequencing or side effect being introduced or changed. Keep it with its existing natural owner when that owner remains coherent; otherwise establish the smallest justified boundary around the complete concern. Represent each decision once. - -#### Design the interface - -Minimise caller knowledge rather than raw method count. Hide decisions the caller does not need, choose domain-meaningful inputs and outcomes, make failure and side effects explicit, and avoid leaking framework or storage mechanics across an established domain or application seam. - -Reject shallow forwarding layers that only repeat another contract. Retain adapters or coordinators that own translation, policy, authorisation, compatibility, observation or another evidenced responsibility. - -#### Control coupling and change propagation - -Trace one plausible current-project change through the proposed owner, interface, consumers and tests. Reconsider the design when the same rule remains duplicated, elements that always change together are split, or one change still requires scattered edits. Do not substitute file size, class count or an abstract coupling score for this trace. - -#### Keep abstraction earned - -Keep stable local concrete dependencies direct. Add a seam, port or inversion only when current evidence requires substitution, isolated observation, multiple implementations, staged migration, a volatile external mechanism or separation of stable policy from mechanism. Avoid interface-per-class, universal repositories, generic service layers and hypothetical extension hooks. - -#### Design It Twice - -For a material new or substantially changed public interface, module boundary, seam or dependency direction, produce at least two genuinely different designs before selecting one. Compare them on interface depth, caller burden, cohesion, locality, dependency direction, testability and likely change propagation. Routine changes that fit an established coherent boundary proceed directly. - -The teaching skill supplies the comparison method; the calling agent owns producing and selecting the alternatives. It does not dispatch its own design workflow or require parallel agents. - -#### Apply conditional techniques - -Teach explicit triggers, trade-offs and good/bad examples for: - -- dependency inversion toward stable policy; -- ports and adapters at material technology or ownership boundaries; -- bounded contexts where one term has competing domain meanings; -- conceptual contours where domain language and change patterns reveal the natural capability grain; -- functional core/imperative shell where deterministic decisions are entangled with effects; -- seams for controlled substitution, observation or migration; -- composition and delegation; -- genuine substitutability when inheritance or polymorphism is considered; and -- co-location or extraction when it improves evidenced cohesion and locality. - -SOLID and design patterns are diagnostic tools, not mandatory architecture. Fowler smells belong to `code-review` as prompts for investigation, while this skill supplies the structural mechanisms needed to understand and correct a concrete concern. - -### Required example coverage - -Use concrete contrasted examples for deep versus shallow interfaces, information leakage, shotgun change propagation, pass-through layers, earned versus speculative seams, policy depending on mechanism, functional decisions mixed with effects, competing domain meanings and two alternative interface designs. Examples must show consequences for callers and future change rather than relying on labels alone. - -### Expected resources - -Keep the always-applicable ownership, interface, depth, locality and minimality rules on the loaded path. Link conditional technique detail and Design It Twice at their triggers. The future authoring session may reuse, replace or split the current `TECHNIQUES.md` according to observed loading needs. - -### Authoring and evaluation evidence - -Cases must test a simple change that needs no new abstraction, a material interface needing Design It Twice, a volatile external mechanism, a shallow forwarding layer, competing domain meanings, a large but cohesive module, a short but damaging coordinator and a request that would tempt speculative generality. - -## `quality-harnesses` - -### Recurring job - -Teach a caller to select, apply and interpret the quality harnesses relevant to a fixed change without confusing execution, metrics or control count with proof of correctness. - -### Loading path - -The final `SKILL.md` instructs the caller to: - -1. inspect the changed behaviour, implementation, repository requirements, measurement-path changes and material risks; -2. load and review the complete compact harness catalogue; -3. identify every catalogue entry whose trigger applies; -4. load the detailed reference for each applicable entry; -5. apply that knowledge within the caller's own workflow; and -6. preserve exact results, limitations, unavailable evidence and residual risk. - -The skill does not split entries into baseline and optional sections. A repository-required check is applicable because the repository requires it; another harness is applicable because the change exposes its named risk. Irrelevant entries are not loaded or run. - -### Catalogue contract - -Each catalogue entry contains one short paragraph with: - -- **Trigger:** the concrete repository requirement, change shape or risk cue that makes the harness relevant. -- **Narrow claim:** what an honestly clean result can support. -- **Limitations:** the important properties, environments or failure classes it cannot establish. -- **Anti-gaming checks:** how tautological oracles, exclusions, thresholds, suppressions, stale baselines, unrepresentative inputs, flaky retries or empty selectors could make the result misleading. -- **Reference:** the detailed instructions and examples to load only when the entry applies. - -The catalogue covers at least: - -| Harness family | Applicability and intended teaching | -| --- | --- | -| Build, compilation and packaging | Apply when repository-required or when build inputs, generated output, dependencies, packaging or runtime compilation change. Record the exact environment and artefact; one build does not establish portability or behavioural correctness. | -| Types, linting, formatting and framework diagnostics | Apply configured checks to their actual scope. Inspect baselines, exclusions and changed configuration; a clean result proves only encoded enabled rules. | -| Behavioural and regression suites | Apply focused evidence for changed behaviour and the repository-required broader suite. Distinguish new-behaviour evidence from pass-to-pass regression evidence and appraise the oracle through `tdd`. | -| Coverage | Use to locate changed or risk-relevant paths that no test executes. Never treat execution percentage as correctness, effectiveness or a target without repository authority. | -| Mutation | Use when test sensitivity is materially uncertain and existing tooling can assess a bounded change economically. Appraise surviving mutants and equivalent or irrelevant cases; never treat the score as correctness. | -| Static and security analysis | Apply configured analysis when changed code reaches its rules, and security-focused controls for trust boundaries, permissions, sensitive data, untrusted input, cryptography or dependency risk. Record rules, versions, suppressions and blind spots. | -| Architecture fitness rules | Apply when an authoritative dependency, layer, cycle, import or public-surface constraint is encoded. Check selectors and mappings are non-empty and current; passing rules cannot establish that the architecture decision itself is good. | -| Contracts and compatibility | Apply when APIs, events, schemas, serialization, storage formats, runtimes, browsers, databases or supported versions change. Record consumers and matrix coverage; finite examples cannot establish all consumer semantics. | -| Fuzzing and adversarial input | Apply to parsers, protocols and hostile or high-volume input when existing harnesses have meaningful crash, sanitizer, assertion or invariant oracles. Preserve corpora, seeds and minimized failures; execution alone rarely proves semantic correctness. | -| Accessibility | Apply when visible UI, semantics, focus, input, navigation, contrast or assistive-technology behaviour changes. Combine configured automation with required manual or specialist evidence; automation alone does not establish conformance. | -| Performance and load | Apply to hot paths, queries, I/O, algorithms, memory, caching or explicit budgets. Use representative workloads, pinned baselines, repeated measurements and correctness assertions; record variance and environment. | -| Concurrency and ordering | Apply to shared state, locks, transactions, asynchronous cancellation, retries, queues or idempotency. Use supported race detection, deterministic schedules, models or seeded stress; bounded success does not prove freedom from races. | -| Resilience and fault injection | Apply to timeouts, failover, partial dependency failure, recovery, rollout or operational degradation. Require an explicit hypothesis, safe boundary, recovery oracle and observability; injected faults cannot enumerate production combinations. | -| Migration and data integrity | Apply transformations to representative legacy states and verify invariants, mixed-version compatibility and rollback or forward recovery where promised. Fixtures cannot represent every production distribution or invalid state. | -| Reproducibility and provenance | Apply when toolchains, generated artefacts, signed output or environment-sensitive builds change. Record material inputs and comparison method; same-machine repetition is not cross-environment reproducibility. | -| Manual and specialist QA | Apply when automation cannot faithfully assess visual, physical, usability, domain or contextual outcomes. Preserve steps, inputs, versions, expected and actual results and evaluator limitations. | - -Complexity, cognitive complexity, size, churn, coupling and Fowler smell counts may identify a hotspot for investigation. Preserve repository-owned thresholds, but do not introduce universal thresholds or treat a metric as a finding without a concrete mechanism and consequence. - -### Infrastructure and authority boundary - -The calling workflow owns commands, mutations and remediation. This teaching skill does not run a verifier, return a pass/fail verdict or install tooling. A caller may add better behavioural tests through the existing suite when the implementation requires them. It must not introduce project-wide harness infrastructure unless that work is explicitly authorised. Missing relevant infrastructure is reported with the claim and risk left unassessed. - -### Expected resources - -Create a compact catalogue that is always reviewed when this skill applies. Link each entry to detailed reference material with selection guidance, execution boundaries, honest interpretation, anti-gaming examples and representative commands or procedures. References may group closely coupled harnesses when their triggers and evidence boundaries align; they must remain directly reachable from each catalogue entry. - -### Authoring and evaluation evidence - -Cases must include a small change with only repository-required checks, a security-sensitive change, changed tests that inflate coverage without an oracle, mutation survivors of mixed relevance, an empty architecture selector, a flaky check, unavailable specialist infrastructure and an irrelevant expensive harness that must not be selected. - -## `code-review` - -### Recurring job - -Independently review one fixed code change through isolated Spec and Standards agents and return consequence-ranked, falsifiable suggestions for a caller to validate and adjudicate. - -### Fixed candidate - -Resolve one explicit non-empty diff, revision range, branch comparison, pull-request revision, repair or local candidate. Capture its changed and in-scope untracked files, complete relevant source and tests, repository state and exact revision identifiers where available. If the target changes during review, mark affected results stale rather than silently retargeting them. - -### Review axes - -#### Spec - -Run this axis only when a behavioural authority is reasonably available: a feature specification, ticket, acceptance criteria, confirmed request, contract or another project source that states intended behaviour independently of the implementation. - -Give the fresh Spec agent: - -- the fixed candidate and complete changed files; -- the behavioural authority and relevant affected contracts or consumers; -- the consequence ranking and finding format; and -- a read-only boundary. - -Direct it to account for every applicable requirement in code and retained tests, account for introduced behaviour against authority, and follow states, contracts, data shapes, errors, effects and consumers far enough to find missing, partial, conflicting, excessive or regressed behaviour. - -Do not give it Standards findings, implementation-thread rationale, conversation history or the Standards packet. When no authority is reasonably available, do not infer intent from code or commits and do not stop Standards; report `Spec not performed: no behavioural authority reasonably available.` - -#### Standards - -Give a separate fresh Standards agent: - -- the same fixed candidate and complete changed files; -- repository instructions, conventions, architecture decisions and configured checks; -- relevant surrounding implementation, tests and supplied quality evidence; -- the full Fowler second-edition smell catalogue as diagnostic vocabulary; -- the applicable `modular-design`, `tdd` and `quality-harnesses` teaching paths; -- the consequence ranking and finding format; and -- a read-only boundary. - -Direct it to understand every changed line in its necessary context and assess correctness and regression risks, minimality, project and framework conventions, changed-test validity, modular ownership and interfaces, quality-harness selection and evidence integrity, recognised smells, and specialist risks exposed by the change. - -Do not give it Spec findings, implementation-thread rationale, conversation history or the Spec packet. The teaching skills inform its judgement without becoming nested delegated workflows. - -### Independent execution - -Run the applicable fresh agents in parallel. Their isolation prevents one review axis from anchoring or suppressing the other; a fresh context alone is not treated as proof of statistical independence. - -Each reviewer may run a safe, focused, non-mutating command only when it can confirm or falsify a concrete concern. It does not repeat the implementation's whole quality portfolio, edit code or tests, update snapshots or baselines, install dependencies, modify durable data or repair a finding. - -Before retaining a suggestion, the reviewer must try to disprove it using contradicting authority, existing handling, repository-sanctioned exceptions, surrounding code, a focused counterexample and the strongest benign interpretation. Unsupported generic advice, tooling-enforced nits and speculative best practices are omitted. - -### Consequence ranking - -- **High:** a credible path to materially wrong required behaviour, security or privacy compromise, data loss or corruption, major production or reliability failure, or a structural, test or evidence defect that makes the change untrustworthy. -- **Medium:** a concrete defect, regression risk or significant maintainability, modularity, test-quality or harness-integrity weakness with a bounded material consequence. -- **Low:** a local but evidenced issue whose correction has a concrete benefit. Personal taste, hypothetical flexibility and inconsequential style are not findings. - -Rank communicates consequence and attention only. It never directs the caller to implement a suggestion automatically; a high finding may be false, and a lower-ranked finding may still be a required repository or behavioural correction. - -### Output - -Keep `## Standards` and `## Spec` separate. Do not merge, deduplicate, suppress or cross-axis rerank their results before returning them to the caller. For an applicable clean axis, report `No findings.` For an omitted or stale axis, state that status explicitly. - -Each suggestion contains: - -```markdown -### [high|medium|low] Concise finding - -- Evidence: exact code location, applicable authority or criterion, and observed fact -- Consequence: concrete behavioural or code-health impact -- Suggested direction: the narrow outcome that would address the concern without prescribing an unverified patch -``` - -End with axis-specific counts and review scope. Do not issue an approval, verification or completion verdict. The caller owns validation, deduplication during adjudication, remediation, re-review and any publication decision. - -### Authoring and evaluation evidence - -Cases must include two axes with conflicting-looking but independently valid results, no Spec authority, a standards concern requiring teaching-skill knowledge, a high-ranked false positive, a low-ranked genuine requirement violation, an unsupported generic suggestion that must be removed, a targeted probe, scope drift and a clean axis. - -## Cross-skill acceptance - -The implementation group is ready only when matched fresh-agent evaluation demonstrates: - -- `implement` behaves as a small process owner and does not duplicate the teaching skills; -- the three teaching skills materially change caller behaviour without taking workflow ownership; -- TDD produces tests that fail for the intended reason, rely on independent oracles, survive structural change and can be rationalised without losing behaviour; -- modular design improves ownership, interface depth and change locality without automatic abstraction or class proliferation; -- quality-harness selection is complete for evidenced risks, excludes irrelevant controls and reports absent infrastructure honestly; -- Spec and Standards reviews are isolated, falsifiable and consequence-ranked; -- implementation rejects false positives rather than blindly obeying review; -- a changed candidate invalidates earlier review and affected quality evidence; -- the group composes for local implementation, debug repair and pull-request review without duplicating another review or verification method; and -- all skills stop before commit or publication unless a later lifecycle skill is explicitly invoked. - -## Authoring order - -Refine the skills in dependency order so later evaluations consume published contracts rather than provisional internals: - -1. `modular-design` -2. `tdd` -3. `quality-harnesses` -4. `code-review` -5. `implement` - -For each skill, use `$write-skill` to confirm the closed contract, establish no-skill and current-skill baselines where applicable, draft outside the destination, validate mechanically, compare through isolated fresh agents and publish only the smallest behaviourally justified bundle. Mark its checklist item in [`SKILLS_PLAN.md`](SKILLS_PLAN.md) complete only after that full authoring and evaluation pass. - -## Current cutover boundary - -This planning session creates this brief, updates the lifecycle plan and removes the obsolete `verify-change` bundle. It does not rewrite the five surviving skills, `debug`, `review-pull-request` or the README. Those files may temporarily describe the previous composition until their scheduled authoring sessions align them with this authority. diff --git a/SKILLS_PLAN.md b/SKILLS_PLAN.md deleted file mode 100644 index f1917ed..0000000 --- a/SKILLS_PLAN.md +++ /dev/null @@ -1,362 +0,0 @@ -# Propulsion skills plan - -## Status - -This is the authoritative implementation brief for the Propulsion lifecycle skill set. It fixes the intended outcomes, composition, boundaries and acceptance conditions without prescribing the finished wording or detailed process of each skill. - -The existing suite may be changed or replaced where necessary. Skill authoring remains responsible for researching credible governing methodologies, allowing a methodology-free design when none fits, testing the common successful path and handling only material exceptions. - -## Objective - -Propulsion is a compact, composable set of Agent Skills for practical software delivery. It supports work from initial product discovery through feature definition, planning, implementation, pull-request publication and review, and debugging. - -Propulsion is not a workflow platform. It does not carry a product automatically through its lifecycle, maintain its own work state or require a setup framework. A user may enter at any point where the necessary inputs already exist and leave when that session's outcome is complete. - -## Design principles - -1. **One bounded session outcome.** Each lifecycle skill represents a natural working session and ends with one independently useful outcome. -2. **Narrow progressively.** Product intent narrows into a decision-complete feature specification, implementation-ready tickets and, finally, one implemented body of work. -3. **Enter proportionately.** Substantial work uses the complete applicable chain. Small, understood work may begin directly with implementation. -4. **Compose within the requested outcome.** A skill may invoke another skill when its outcome is required to complete the requested session. It must not silently advance into a later lifecycle outcome. -5. **Keep authorities separate.** Product documents own intent, code and tests own implemented behaviour, Git owns change history and the configured ticket system owns work state. -6. **Document what cannot be inferred safely.** Durable documents preserve intent, language, constraints and rationale without copying observable code, configuration, CI or tracker state. -7. **Use native project tools.** Propulsion does not introduce a manifest, tracker adapter, lifecycle state model or completion-horizon schema. -8. **Treat agreement as authority.** Invoking a skill authorises the mutations intrinsic to its agreed outcome. Questions resolve the outcome; they are not followed by a redundant application gate. -9. **Make quality evidence cumulative.** Implementation satisfies project checks and adds proportionate evidence without weakening existing constraints or installing unrelated infrastructure merely to make a check available. -10. **Keep the public surface clear.** Lifecycle skills are the primary entry points. Reusable utilities remain independently useful and may be composed where their explicit trigger applies. - -## Lifecycle coverage - -| Area | Purpose | -| --- | --- | -| Establish | Create the durable product and system requirements foundation for a new or existing system. | -| Define | Turn one high-level feature request into an approved statement of intended behaviour and its selected buildable solution. | -| Plan | Decompose an approved specification into implementation-ready work. | -| Refine | Assess a bounded area of the current architecture and identify high-value improvements. | -| Deliver | Implement agreed work, record it, publish it for review and review work submitted by others. | -| Debug | Diagnose and repair defects. | - -These areas are entry points, not mandatory phase gates. Release, deployment and continuous production monitoring are outside Propulsion's core lifecycle. - -## Session and authority model - -Propulsion skills are explicitly user-invokable. Only `elicit` may also be model-invoked when its documented trigger is present. Entry skills fan out through their documented composition; a skill may call another skill only when the subordinate outcome is necessary to finish the invoked session. For example, `pull-request` may call `commit` for eligible uncommitted work, while `create-tickets` must not silently manufacture an incomplete work definition. A caller remains responsible for verifying a composed skill's result and completing its own outcome. - -Invoking a skill grants authority for its expected repository and external mutations once the skill has reached shared understanding. Material product, architecture, priority or scope questions remain with the user, but an agreed decision does not require a second permission prompt before it is applied. Work outside the invoked outcome remains unauthorised. - -Every skill: - -1. Inspects the request, repository guidance, existing artefacts and available tools before asking questions. -2. Reuses current authoritative inputs rather than recreating earlier lifecycle work. -3. Invokes required and triggered supporting skills. -4. Persists its outcome in the owning document or system. -5. Verifies the resulting artefact or external state. -6. Stops at its stated boundary and identifies the next available session without starting it automatically. - -## Documentation and tool authorities - -### Living project foundations - -| Authority | Owns | -| --- | --- | -| `PRODUCT.md` | Product purpose, users, needs, value, outcomes, boundaries, journeys, high-level capabilities and system-wide technical requirements. | -| `CONTEXT.md` | Canonical project terms and their concise project-specific meanings. | - -`CONTEXT.md` remains live throughout the lifecycle. Any skill that establishes or changes consequential domain language invokes `maintain-ubiquitous-language` inline. - -`PRODUCT.md` records requirements-first system constraints that later feature solutions must respect without selecting their internal implementation. `maintain-decision-records` creates focused ADRs only when a consequential or surprising technical or architectural decision needs durable rationale beyond the product foundation or feature records. - -### Feature records - -A substantial feature uses the project convention where one exists, otherwise: - -```text -docs/features//specification.md -``` - -The feature specification contains two visibly distinct semantic parts inside one authority. Feature intent owns the problem, outcomes, scope, actors, observable behaviour, qualities, scenarios and acceptance. Selected solution owns the system-specific responsibilities, boundaries, contracts, data, integrations, controls, migration, operational effects and verification seams. The distinction prevents implementation ideas from masquerading as requirements without creating another document or session handoff. - -Small, understood changes do not require feature documents merely to enter implementation. - -### Bounded supporting records - -Research reports and architecture reviews are repository artefacts only when their conclusions and rationale need to survive the session. They must not duplicate live operational, tracker or implementation state. - -### Ticket destination - -The project's ticket destination must be named in `AGENTS.md`, for example `The project uses Linear for tickets.` or `The project uses local Markdown for tickets.` A ticket-creating skill that cannot find this preference asks the user which destination the project uses, invokes `maintain-agents` to record it and then resumes its original outcome. - -Local Markdown uses one file per ticket under `docs/features//tickets/`. An external destination uses its native items, estimates, relationships and statuses. If the named external tool is unavailable or not writable, the skill stops with the exact access requirement. It does not guess or fall back silently to Markdown or another tracker. - -## Lifecycle skill catalogue - -### Establish - -#### `define-product` - -- **Outcome:** A user-confirmed product and system requirements foundation broad enough to guide later feature specification without becoming one. -- **Inputs:** The product idea, existing `PRODUCT.md` and `CONTEXT.md`, applicable ADRs, representative repository evidence and applicable external evidence. -- **Output:** `PRODUCT.md` and the corresponding ubiquitous language in `CONTEXT.md`. -- **Composition:** Requires `elicit-with-context`; conditionally invokes `research`. -- **Stops:** After purpose, users, needs, value, outcomes, boundaries, journeys, high-level capabilities and material quality, interface, data, integration, security, compliance, platform and operational requirements are documented. It does not define detailed feature behaviour, select feature-specific implementation or architecture, create priorities or create tickets. - -### Define - -#### `specify-feature` - -- **Outcome:** One approved, decision-complete feature specification containing both feature intent and its selected buildable solution. -- **Inputs:** Any high-level feature request, conversation, product capability or existing specification, plus whatever project and current-system evidence is available. No earlier lifecycle artefact is mandatory. -- **Output:** `docs/features//specification.md`, or the project's equivalent. -- **Composition:** Requires `elicit-with-context`; conditionally invokes `research` and `modular-design`. -- **Stops:** After the problem, actors, use cases, behaviour, rules, states, qualities, experience, scenarios, acceptance, selected responsibilities, contracts, data, integrations, controls, migration, operational effects and verification seams are resolved enough for ticket decomposition. It does not create tickets or implement the feature. - -### Plan - -#### `create-tickets` - -- **Outcome:** One approved work definition is represented by confirmed, implementation-ready, dependency-aware tickets in the project's configured destination. -- **Inputs:** One sufficiently complete approved document or confirmed conversation, plus relevant project guidance and implementation evidence. -- **Output:** One local Markdown file or external native item per coherent vertical outcome, with bounded context, requirements, applicable constraints, observable acceptance, Fibonacci complexity and genuine blocking relationships. -- **Composition:** Conditionally invokes `elicit-with-context` for unresolved decomposition and `maintain-agents` when the ticket destination is missing. -- **Stops:** After the created tickets and relationships are read back and verified. It does not schedule the work or begin implementation. - -### Refine - -#### `review-architecture` - -- **Outcome:** A bounded, evidence-backed assessment identifies a small number of high-value architecture or maintainability improvements. -- **Inputs:** An explicitly scoped codebase area, `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and current implementation evidence. -- **Output:** An independently useful architecture review report with prioritised findings, evidence, benefits and trade-offs. -- **Composition:** Requires `modular-design`; conditionally invokes `research`. -- **Stops:** After reporting. It does not refactor the code or create speculative tickets automatically. - -### Deliver - -The detailed authoring contract for the implementation skills is maintained in [`IMPLEMENTATION_SKILLS_PLAN.md`](IMPLEMENTATION_SKILLS_PLAN.md). This plan retains their lifecycle outcomes and composition; the implementation plan owns their teaching content, internal sequence, resources and evaluation expectations. - -#### `implement` - -- **Outcome:** One ticket or small agreed body of work exists as a minimal, locally checked and independently reviewed change. -- **Inputs:** The stated ticket or confirmed work, linked product and technical authorities, repository guidance and current code. -- **Output:** Working code, retained tests and other in-scope artefacts, applicable quality evidence and limitations, and no unresolved applicable review change. -- **Composition:** Applies the teaching guidance from `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, requires independent `code-review`, and conditionally invokes `elicit-with-context` when implementation exposes a material missing detail or decision. -- **Stops:** After the agreed behaviour is implemented, applicable checks pass, unavailable evidence is explicit and the current candidate has completed the review-and-correction loop. It does not commit or publish the change. - -#### `commit` - -- **Outcome:** Eligible reviewed work is recorded in one or more coherent Conventional Commits. -- **Inputs:** The current Git state, repository guidance, recent history and any user-supplied scope. -- **Output:** Verified atomic commits that preserve unrelated staged, unstaged and untracked work. -- **Composition:** No required utility. -- **Stops:** After reporting created commits and remaining work. It does not push or create a pull request. - -#### `pull-request` - -- **Outcome:** The current branch and its complete committed work are represented by one accurate pull request. -- **Inputs:** The current branch, its base, full branch history, repository guidance and any existing pull request. -- **Output:** A pushed branch and created or updated pull request whose title and body describe the whole branch. -- **Composition:** Conditionally invokes `commit` when eligible uncommitted work remains. -- **Stops:** When the published head, base, content and review state are verified. It does not merge, release or deploy the work. - -#### `review-pull-request` - -- **Outcome:** Someone else's pinned pull request receives an independent, evidence-backed assessment. -- **Inputs:** The complete pull-request change, originating specification or work item, repository standards and relevant product and technical authorities. -- **Output:** Prioritised findings covering behavioural conformance, regressions, minimality, conventions, maintainability, architecture and verification evidence. -- **Composition:** Requires `code-review`, whose Standards axis applies the implementation teaching skills; conditionally invokes `research` when external evidence is material to the review. -- **Stops:** After reporting findings. It does not change the contributor's branch or publish comments, approvals or requested changes unless the user's invocation includes that outcome. - -### Debug - -#### `debug` - -- **Outcome:** A defect has an evidence-backed root cause and, when authorised by the requested outcome, a minimal verified repair. -- **Inputs:** Expected and observed behaviour, the failing signal, relevant environment evidence and current implementation authorities. -- **Output:** A reproducible diagnosis and, for repair work, regression protection plus a reviewed local change. -- **Composition:** Repair applies `modular-design` and `quality-harnesses`, applies `tdd` for behaviour-changing work when a usable suite exists, and requires `code-review`; conditionally invokes `research` and `elicit-with-context` when their triggers apply. -- **Stops:** At the diagnosis boundary when requested, otherwise after the original failure and regression evidence verify the repair. It does not commit or publish the change. - -## Reusable utility catalogue - -| Utility | Independently useful outcome | -| --- | --- | -| `elicit` | Resolves dependency-ordered material information and user decisions one question at a time until shared understanding is confirmed. | -| `elicit-with-context` | Routes software-project elicitation through `elicit` while maintaining confirmed project language and qualifying ADRs. | -| `research` | Investigates a material subject using high-trust evidence and creates a trusted cited report for the consumer to synthesize. | -| `maintain-ubiquitous-language` | Keeps `CONTEXT.md` aligned inline with confirmed project-specific language. | -| `maintain-decision-records` | Preserves rare accepted technical and architectural decisions whose rationale warrants a focused ADR. | -| `modular-design` | Teaches callers to structure code through information hiding, deep modules, cohesive ownership, stable interfaces, locality and conditional design techniques. | -| `tdd` | Teaches callers to implement observable behaviour through red-green-refactor and retain a small, durable, independently meaningful test suite. | -| `quality-harnesses` | Teaches callers to select and apply relevant quality harnesses from a trigger-led catalogue without overstating their evidence. | -| `code-review` | Independently reviews a fixed change through isolated Spec and Standards agents and returns consequence-ranked suggestions for its caller to adjudicate. | -| `maintain-agents` | Records confirmed, non-inferable repository-wide agent guidance while keeping permanent context lean. | - -Software-project skills route user-held questions through `elicit-with-context`. Base `elicit` remains available for non-software work and as that router's questioning component. - -`write-skill` remains a separate suite-development tool. It is not presented as part of the software-delivery lifecycle. - -## Implementation quality model - -### Research prerequisite - -Before finalising `implement`, `tdd`, `modular-design`, `quality-harnesses` or `code-review`, conduct focused research using primary technical sources, established standards and credible field practices. Persist one cited report that establishes: - -- the obligations common to implementation regardless of harness selection -- selection triggers for additional harnesses -- what each harness can and cannot demonstrate -- ways metrics and tests can be gamed -- how executable evidence and engineering judgement complement each other -- how to report incomplete confidence when required infrastructure is absent - -The research evaluates practices used successfully in real projects without turning every available harness into mandatory ceremony. - -The current evidence is recorded in [`docs/research/agentic-coding-quality-guardrails-2026-08-20.md`](docs/research/agentic-coding-quality-guardrails-2026-08-20.md). The detailed implementation-skill decisions derived from it are fixed in [`IMPLEMENTATION_SKILLS_PLAN.md`](IMPLEMENTATION_SKILLS_PLAN.md). - -### Teaching utilities - -`modular-design`, `tdd` and `quality-harnesses` are teaching skills. They supply the calling agent with applicable knowledge; they do not become delegated workers, mutate independently, return separate workflow verdicts or take ownership from their caller. `implement` uses that knowledge to write and assess code. The Standards agent created by `code-review` uses the same knowledge to review code and evidence. - -### Quality-harness catalogue - -`quality-harnesses` directs its caller to inspect the change, repository requirements and material risks, then review a compact catalogue and apply every relevant entry. Each catalogue entry states its trigger, the narrow claim it can support, its limitations, its anti-gaming checks and the detailed reference to load when applicable. The catalogue may include: - -- unit, acceptance, integration, contract and system tests -- executable business examples such as Gherkin where they improve precision -- static types, linting, formatting and framework diagnostics -- coverage analysis and mutation testing -- security, accessibility, performance, compatibility and resilience checks -- reproducible manual or specialist QA procedures - -The catalogue does not divide controls into a mandatory baseline and optional extras. Applicability governs selection. Complexity, size, coverage, mutation and other metrics remain evidence for a named question rather than universal gates or composite quality scores. - -The repository's existing constraints remain authoritative. A caller may strengthen evidence through the existing test and quality infrastructure but must not weaken or skip required checks silently. - -When an existing project lacks an applicable harness, the caller records the missing evidence and residual risk. It does not install a project-wide test framework or quality tool unless establishing that infrastructure is explicitly part of the agreed work. New projects establish required infrastructure through their architecture and initial delivery tickets. - -### Maintainability judgement - -Passing checks does not prove that a change is well designed. `implement` applies `modular-design` throughout the change, and the independent Standards review uses `modular-design`, `tdd` and `quality-harnesses` to assess whether the implementation: - -- contains only the behaviour and code needed for the agreed outcome -- follows the language, framework and repository conventions -- reuses established helpers and abstractions where they fit -- avoids duplication, speculative generality and recognised code smells -- preserves or improves information hiding and module boundaries -- remains understandable and economical to change - -Spec and Standards suggestions are ranked high, medium or low by consequence. Severity never decides remediation. The implementation owner validates each suggestion against its authority and evidence, applies required corrections and proportionate in-scope improvements, rejects unsupported or out-of-scope findings, and returns material behaviour, architecture or scope decisions to the user. Review repeats after material candidate changes until no applicable change remains. - -## Representative flows - -### Establish a product - -```text -define-product -``` - -Product definition stops after the confirmed breadth-first product and system requirements foundation. - -### Deliver a substantial feature - -```text -specify-feature - → later create-tickets - → implement each selected ticket in a separate session - → commit whenever a coherent unit is ready - → pull-request when the branch is ready for human review -``` - -### Deliver a small understood change - -```text -elicit-with-context when needed - → implement - → commit - → pull-request -``` - -### Debug a known defect - -```text -debug - → commit - → pull-request -``` - -### Review another contributor's work - -```text -review-pull-request -``` - -The author's normal flow does not invoke `review-pull-request`; `implement` already includes independent change review. - -## Implementation order - -1. Complete the quality-harness research and implementation-skill authoring plan. -2. Align the shared utilities: `elicit`, `elicit-with-context`, `research`, `maintain-ubiquitous-language`, `maintain-decision-records`, `modular-design`, `tdd`, `quality-harnesses`, `code-review` and `maintain-agents`. -3. Build the progressive foundation path: `define-product`, `specify-feature` and `create-tickets`. -4. Build the delivery path: `implement`, `commit`, `pull-request` and `review-pull-request`. -5. Build the architecture-assessment path: `review-architecture`. -6. Build the remaining path: `debug`. -7. Align software-project elicitation through `elicit-with-context`, update the README around the lifecycle areas and independently invokable utilities, and run the suite-wide acceptance pass. - -### Implementation checklist - -- [x] Complete quality-harness research -- [x] Define the implementation-skill authoring plan -- [x] Update `elicit` skill -- [x] Update `research` skill -- [x] Create `maintain-ubiquitous-language` skill -- [x] Create `maintain-decision-records` skill -- [x] Update `modular-design` skill -- [x] Update `tdd` skill -- [x] Create `quality-harnesses` skill -- [x] Update `code-review` skill -- [x] Update `maintain-agents` skill -- [x] Update `define-product` skill -- [x] Create `specify-feature` skill -- [x] Create `create-tickets` skill -- [x] Update `implement` skill -- [x] Update `commit` skill -- [x] Replace `pr` with `pull-request` -- [x] Create `review-pull-request` skill -- [x] Update `review-architecture` skill -- [x] Update `debug` skill -- [x] Restore `elicit-with-context` skill -- [-] Update the README -- [-] Run the suite-wide acceptance pass - -Each skill is authored and forward-tested as one coherent work unit. The rewrite branch may contain incomplete groups during development, but the public cutover occurs only when the final catalogue composes coherently and replaces obsolete guidance. - -## Acceptance conditions - -The revised suite is ready when: - -- every common lifecycle task has one obvious user-facing entry point -- every lifecycle skill produces one bounded outcome and stops at its documented boundary -- substantial work narrows cleanly from product foundation through decision-complete specification, tickets and per-item implementation -- small, clear work can enter implementation without ceremonial documents or tickets -- required and conditional composition is explicit and does not duplicate supporting skill instructions -- ticket-creating skills use the local Markdown or external destination named in `AGENTS.md` and preserve its exact concepts without Propulsion configuration -- documentation preserves enduring intent without copying observable implementation or live tool state -- implementation always applies the relevant modular-design, TDD and quality-harness guidance and completes independent review -- absent quality infrastructure is reported honestly and never causes unrelated tooling to be installed implicitly -- external mutations remain within the outcome authorised by the invoked skill and are read back after application -- release, deployment and continuous monitoring remain outside the suite -- each skill passes a fresh-agent common-path test and every material permission, missing-input and stopping boundary identified during authoring -- representative cross-session flows pass without transcript dependency or an orchestrating lifecycle skill -- `bun run checks` and `git diff --check` pass - -## Explicit non-goals - -Propulsion will not: - -- provide a lifecycle navigator or end-to-end delivery orchestrator -- require a Propulsion manifest, setup skill, state model, completion horizon or tracker adapter -- impose a project-management method -- create separate skills for individual test types, metrics or quality commands -- require discovery, architecture or feature documents for small work that is already sufficiently understood -- duplicate code, CI, tracker or production state in prose documents -- release, deploy or continuously monitor software -- merge a pull request merely because it has been created or reviewed From 271bbe39debe25f7a3ca6e08e42cedd053cca741 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Fri, 21 Aug 2026 19:23:30 +0100 Subject: [PATCH 74/79] style(markdown): align tables with oxfmt --- .oxfmtrc.json | 2 +- README.md | 12 +++++----- skills/debug/SKILL.md | 22 +++++++++---------- .../references/INSTRUMENTATION-DEBUGGERS.md | 12 +++++----- .../debug/references/PERFORMANCE-RESOURCE.md | 14 ++++++------ .../assets/specification-template.md | 12 +++++----- skills/tdd/references/GENERATIVE-TESTING.md | 12 +++++----- skills/tdd/references/TEST-DESIGN.md | 18 +++++++-------- skills/tdd/references/TEST-DOUBLES.md | 14 ++++++------ skills/write-skill/SKILL.md | 18 +++++++-------- skills/write-skill/references/SECTIONS.md | 16 +++++++------- 11 files changed, 76 insertions(+), 76 deletions(-) diff --git a/.oxfmtrc.json b/.oxfmtrc.json index 553cf6c..37348b3 100644 --- a/.oxfmtrc.json +++ b/.oxfmtrc.json @@ -2,7 +2,7 @@ "$schema": "./node_modules/oxfmt/configuration_schema.json", "ignorePatterns": ["node_modules/**", ".opencode/node_modules/**"], "printWidth": 80, - "proseWrap": "never", + "proseWrap": "preserve", "tabWidth": 4, "singleQuote": true, "sortImports": { diff --git a/README.md b/README.md index 2cb34d5..630d3b7 100644 --- a/README.md +++ b/README.md @@ -155,13 +155,13 @@ $debug → $commit → $pull-request These skills are independently invokable outside the main lifecycle path and may also be composed by lifecycle skills when their trigger applies. -| Skill | Invoke it to… | -| --- | --- | -| `$review-pull-request` | Independently assess someone else's pinned pull request without changing their branch or publishing a review unless requested | -| `$elicit` | Resolve material user-held information and decisions one question at a time | +| Skill | Invoke it to… | +| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | +| `$review-pull-request` | Independently assess someone else's pinned pull request without changing their branch or publishing a review unless requested | +| `$elicit` | Resolve material user-held information and decisions one question at a time | | `$elicit-with-context` | Resolve software-project questions through `$elicit`, applying `$maintain-ubiquitous-language` and `$maintain-decision-records` when their triggers apply | -| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | -| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | +| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | +| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | Use `$elicit-with-context` to work through questions or decisions within a software project. For non-software work, use `$elicit`. diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 3cb36cf..c194bd4 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -26,17 +26,17 @@ Maintain one compact investigation record containing observations, a small ranke Select the reference whose trigger matches the current uncertainty and load only that reference. Return its evidence to this causal loop before selecting another: -| Evidence shape | Load | -| --- | --- | -| Multiple plausible explanations remain | [Hypothesis-Driven Experiments](references/HYPOTHESIS-EXPERIMENTS.md) | -| A component, pipeline, request, or data path has observable boundaries | [Boundary Isolation](references/BOUNDARY-ISOLATION.md) | -| Reliable known-good and known-bad ordered states exist | [Change Bisection](references/CHANGE-BISECTION.md) | -| A large input, state, sequence, trace, or change set can be reduced | [Delta Debugging](references/DELTA-DEBUGGING.md) | -| Invalid state appears downstream of where it originated | [Origin Tracking](references/ORIGIN-TRACKING.md) | -| A credible working case can be compared with the failing case | [Comparative Debugging](references/COMPARATIVE-DEBUGGING.md) | -| Existing evidence cannot distinguish the active hypotheses | [Instrumentation and Debuggers](references/INSTRUMENTATION-DEBUGGERS.md) | -| Outcome depends on timing, order, randomness, load, or shared state | [Nondeterministic and Concurrent Faults](references/NONDETERMINISTIC-CONCURRENT.md) | -| The signal is latency, throughput, contention, exhaustion, leakage, or resource cost | [Performance and Resource Faults](references/PERFORMANCE-RESOURCE.md) | +| Evidence shape | Load | +| ------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------- | +| Multiple plausible explanations remain | [Hypothesis-Driven Experiments](references/HYPOTHESIS-EXPERIMENTS.md) | +| A component, pipeline, request, or data path has observable boundaries | [Boundary Isolation](references/BOUNDARY-ISOLATION.md) | +| Reliable known-good and known-bad ordered states exist | [Change Bisection](references/CHANGE-BISECTION.md) | +| A large input, state, sequence, trace, or change set can be reduced | [Delta Debugging](references/DELTA-DEBUGGING.md) | +| Invalid state appears downstream of where it originated | [Origin Tracking](references/ORIGIN-TRACKING.md) | +| A credible working case can be compared with the failing case | [Comparative Debugging](references/COMPARATIVE-DEBUGGING.md) | +| Existing evidence cannot distinguish the active hypotheses | [Instrumentation and Debuggers](references/INSTRUMENTATION-DEBUGGERS.md) | +| Outcome depends on timing, order, randomness, load, or shared state | [Nondeterministic and Concurrent Faults](references/NONDETERMINISTIC-CONCURRENT.md) | +| The signal is latency, throughput, contention, exhaustion, leakage, or resource cost | [Performance and Resource Faults](references/PERFORMANCE-RESOURCE.md) | When the shape is unclear, verify the signal and inspect recent changes, then compare a working case, isolate boundaries, reduce the reproducer, trace the first invalid state, and finally add targeted probes. Enter nondeterministic or performance guidance immediately when its trigger applies. Combine techniques only when each answers a distinct unresolved question. diff --git a/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md b/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md index 4c4561d..5f96641 100644 --- a/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md +++ b/skills/debug/references/INSTRUMENTATION-DEBUGGERS.md @@ -6,12 +6,12 @@ Use this reference when existing evidence cannot distinguish the active hypothes Choose the instrument whose observation matches the question: -| Question | Probe | -| --- | --- | -| Which local state or branch first violates an invariant? | Conditional breakpoint, watchpoint, or stack inspection | -| Where does one request cross a process or service boundary? | Correlated logs, metrics, trace, or narrow state endpoint | -| Which operation consumes CPU, memory, I/O, locks, or time? | Matching profiler, counter, or targeted runtime trace | -| Does a known fault class occur? | Repository-supported detector such as a race, memory, or query diagnostic | +| Question | Probe | +| ----------------------------------------------------------- | ------------------------------------------------------------------------- | +| Which local state or branch first violates an invariant? | Conditional breakpoint, watchpoint, or stack inspection | +| Where does one request cross a process or service boundary? | Correlated logs, metrics, trace, or narrow state endpoint | +| Which operation consumes CPU, memory, I/O, locks, or time? | Matching profiler, counter, or targeted runtime trace | +| Does a known fault class occur? | Repository-supported detector such as a race, memory, or query diagnostic | Place the fewest probes at boundaries, decisions, and state transitions that partition the hypotheses. Capture only the needed input, output, identity, correlation, time, and environment. Establish measurement overhead when the probe can affect the signal. diff --git a/skills/debug/references/PERFORMANCE-RESOURCE.md b/skills/debug/references/PERFORMANCE-RESOURCE.md index f3d2c64..79a28c4 100644 --- a/skills/debug/references/PERFORMANCE-RESOURCE.md +++ b/skills/debug/references/PERFORMANCE-RESOURCE.md @@ -8,13 +8,13 @@ Define the workload, environment, known-good baseline or expected threshold, war For early systemic triage, inspect utilization, saturation, and errors for each relevant resource. Then choose the measurement that matches the suspected cost: -| Suspected cost | Evidence | -| --- | --- | -| CPU or call-path work | Sampling or instrumenting profile under the controlled workload | -| Memory or allocation | Heap, allocation, retention, or growth measurement | -| I/O, network, queue, or lock pressure | Resource counter and targeted trace | -| Database execution | Actual plan, cardinality, I/O, locks, and query timing | -| Version or configuration regression | Stable differential benchmark or change bisection | +| Suspected cost | Evidence | +| ------------------------------------- | --------------------------------------------------------------- | +| CPU or call-path work | Sampling or instrumenting profile under the controlled workload | +| Memory or allocation | Heap, allocation, retention, or growth measurement | +| I/O, network, queue, or lock pressure | Resource counter and targeted trace | +| Database execution | Actual plan, cardinality, I/O, locks, and query timing | +| Version or configuration regression | Stable differential benchmark or change bisection | Connect the hotspot, plan, wait, saturation, or growth to the user-visible signal through a controlled workload or intervention. A hotspot is where cost was observed, not automatically why the system is slow. diff --git a/skills/specify-feature/assets/specification-template.md b/skills/specify-feature/assets/specification-template.md index 56baf35..a400d49 100644 --- a/skills/specify-feature/assets/specification-template.md +++ b/skills/specify-feature/assets/specification-template.md @@ -115,13 +115,13 @@ ## Traceability and readiness -| Intent or acceptance identifier | Selected responsibility or contract | Verification seam or evidence | -| --- | --- | --- | -| {REQ-01 or ACC-01} | {Design responsibility or contract.} | {Observable verification boundary.} | +| Intent or acceptance identifier | Selected responsibility or contract | Verification seam or evidence | +| ------------------------------- | ------------------------------------ | ----------------------------------- | +| {REQ-01 or ACC-01} | {Design responsibility or contract.} | {Observable verification boundary.} | -| Introduced design element | Originating intent, constraint, evidence, or accepted decision | -| --- | --- | -| {Material responsibility, contract, data, integration, migration, operational, or verification element.} | {Exact identifier or linked authority.} | +| Introduced design element | Originating intent, constraint, evidence, or accepted decision | +| -------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------- | +| {Material responsibility, contract, data, integration, migration, operational, or verification element.} | {Exact identifier or linked authority.} | ### Readiness statement diff --git a/skills/tdd/references/GENERATIVE-TESTING.md b/skills/tdd/references/GENERATIVE-TESTING.md index ff67d8f..63e43a4 100644 --- a/skills/tdd/references/GENERATIVE-TESTING.md +++ b/skills/tdd/references/GENERATIVE-TESTING.md @@ -2,12 +2,12 @@ Use an example test by default. Select one of these techniques only when its trigger applies and its oracle is independently justified. -| Technique | Trigger | Oracle and limitation | -| --- | --- | --- | -| **Property-based** | Stable invariants over many values; parsers, serializers, algebra, collections, boundary-rich pure logic | Assert an independent invariant. Biased generators and tautological properties can pass indefinitely. | -| **Model-based** | Stateful workflows or protocols where sequences and invalid transitions matter | Compare with a smaller independent state model. A model copied from production repeats its defects and can suffer state explosion. | -| **Metamorphic** | Direct expected outputs are unavailable but relationships between executions are known | Assert relations such as round-trip, permutation invariance, monotonicity, or scaling. A relation detects inconsistency, not necessarily the uniquely correct result. | -| **Differential** | Two genuinely comparable implementations, versions, backends, or conforming tools can receive the same inputs | Compare normalized outcomes. Agreement may be jointly wrong; shared code or assumptions weaken independence. | +| Technique | Trigger | Oracle and limitation | +| ------------------ | ------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------- | +| **Property-based** | Stable invariants over many values; parsers, serializers, algebra, collections, boundary-rich pure logic | Assert an independent invariant. Biased generators and tautological properties can pass indefinitely. | +| **Model-based** | Stateful workflows or protocols where sequences and invalid transitions matter | Compare with a smaller independent state model. A model copied from production repeats its defects and can suffer state explosion. | +| **Metamorphic** | Direct expected outputs are unavailable but relationships between executions are known | Assert relations such as round-trip, permutation invariance, monotonicity, or scaling. A relation detects inconsistency, not necessarily the uniquely correct result. | +| **Differential** | Two genuinely comparable implementations, versions, backends, or conforming tools can receive the same inputs | Compare normalized outcomes. Agreement may be jointly wrong; shared code or assumptions weaken independence. | ## Apply the selected technique diff --git a/skills/tdd/references/TEST-DESIGN.md b/skills/tdd/references/TEST-DESIGN.md index ecc7791..8713180 100644 --- a/skills/tdd/references/TEST-DESIGN.md +++ b/skills/tdd/references/TEST-DESIGN.md @@ -12,15 +12,15 @@ Use the narrowest seam that is: Add a broader companion test only when the narrower seam cannot retain a material integration risk. Test a pure rule through its public function, a database guarantee against the real engine when its semantics matter, and an end-user interaction through the rendered application when wiring or accessibility is part of the promise. -| Concern | Brittle or circular test | Durable behavioural test | -| --- | --- | --- | -| Expected value | Repeats the production reduction to calculate `expected` | Uses the accepted example `10 + 5 = 15` or an independent invariant | -| Domain result | Calls a private pricing method or asserts helper delegation | Calls the supported quote operation and asserts its public result | -| UI | Asserts CSS classes, wrapper tags, child indexes, or DOM depth | Interacts by role or label and asserts visible text, value, focus, state, or navigation | -| Persistence | Inspects private tables, columns, or ORM calls | Writes and reads through supported interfaces; reloads when durability matters | -| Error | Asserts a private branch or stack shape | Asserts caller-visible type, code, promised message, recovery, and unchanged state | -| Snapshot | Accepts a broad incidental object or markup snapshot | Asserts selected semantic fields, unless exact bytes or markup are the contract | -| Bug | Treats the existing crash or output as the expected result | Reproduces the wrong result, then asserts a requirement-derived correction | +| Concern | Brittle or circular test | Durable behavioural test | +| -------------- | -------------------------------------------------------------- | --------------------------------------------------------------------------------------- | +| Expected value | Repeats the production reduction to calculate `expected` | Uses the accepted example `10 + 5 = 15` or an independent invariant | +| Domain result | Calls a private pricing method or asserts helper delegation | Calls the supported quote operation and asserts its public result | +| UI | Asserts CSS classes, wrapper tags, child indexes, or DOM depth | Interacts by role or label and asserts visible text, value, focus, state, or navigation | +| Persistence | Inspects private tables, columns, or ORM calls | Writes and reads through supported interfaces; reloads when durability matters | +| Error | Asserts a private branch or stack shape | Asserts caller-visible type, code, promised message, recovery, and unchanged state | +| Snapshot | Accepts a broad incidental object or markup snapshot | Asserts selected semantic fields, unless exact bytes or markup are the contract | +| Bug | Treats the existing crash or output as the expected result | Reproduces the wrong result, then asserts a requirement-derived correction | Representation is valid only when that representation is itself a published contract. diff --git a/skills/tdd/references/TEST-DOUBLES.md b/skills/tdd/references/TEST-DOUBLES.md index bb899bc..a18429a 100644 --- a/skills/tdd/references/TEST-DOUBLES.md +++ b/skills/tdd/references/TEST-DOUBLES.md @@ -4,13 +4,13 @@ Keep collaborators inside the system real. Substitute a boundary only when the r ## Choose the least powerful double -| Double | Supply | Verify | -| --- | --- | --- | -| Dummy | An unused required value | Nothing | -| Stub | A controlled indirect input | The public outcome | -| Fake | A small working boundary implementation | The public outcome or recorded public effect | -| Spy | A record of an otherwise invisible boundary effect | Only promised boundary facts | -| Mock | An expected external interaction protocol | Only contractually material calls | +| Double | Supply | Verify | +| ------ | -------------------------------------------------- | -------------------------------------------- | +| Dummy | An unused required value | Nothing | +| Stub | A controlled indirect input | The public outcome | +| Fake | A small working boundary implementation | The public outcome or recorded public effect | +| Spy | A record of an otherwise invisible boundary effect | Only promised boundary facts | +| Mock | An expected external interaction protocol | Only contractually material calls | Prefer state verification. Interaction verification is warranted when the interaction is the promise: for example, one idempotency-keyed payment request or publishing only after a durable write. diff --git a/skills/write-skill/SKILL.md b/skills/write-skill/SKILL.md index 5fdde55..ceaf67f 100644 --- a/skills/write-skill/SKILL.md +++ b/skills/write-skill/SKILL.md @@ -96,12 +96,12 @@ Publish the passing candidate to the requested destination and rerun validation Return the destination, confirmed contract, validation result, published files, evaluator verdict, residual uncertainty, and this comparison. Use `n/a` for Current on new skills. Compare Candidate delta with No skill for creation and Current for revision; also state a revision's behavioural shift over No skill. -| Metric | No skill | Current | Candidate | Candidate delta | -| --- | --- | --- | --- | --- | -| Behavioural result and evaluator evidence | | | | | -| Discovery metadata tokens | | | | | -| Loaded skill-path tokens | | | | | -| Agent output tokens | | | | | -| Tool calls | | | | | -| Iterations | | | | | -| Total measured token burden | | | | | +| Metric | No skill | Current | Candidate | Candidate delta | +| ----------------------------------------- | -------- | ------- | --------- | --------------- | +| Behavioural result and evaluator evidence | | | | | +| Discovery metadata tokens | | | | | +| Loaded skill-path tokens | | | | | +| Agent output tokens | | | | | +| Tool calls | | | | | +| Iterations | | | | | +| Total measured token burden | | | | | diff --git a/skills/write-skill/references/SECTIONS.md b/skills/write-skill/references/SECTIONS.md index aa9f854..aa1a6a7 100644 --- a/skills/write-skill/references/SECTIONS.md +++ b/skills/write-skill/references/SECTIONS.md @@ -2,14 +2,14 @@ Use a section only when it changes execution, constrains the finished result, or transfers responsibility. -| Order | Section | Use when | Contains | -| --- | --- | --- | --- | -| 1 | `# Skill Name` | Always | Human-readable skill name. | -| 2 | Introduction | Always | One sentence stating what the skill enables. | -| 3 | `## Prerequisites` | Execution must check conditions before starting | Required state and the action taken when it is absent. | -| 4 | `## Process` | Always | The behaviour that produces the result. | -| 5 | `## Rules` | One invariant governs several steps or the finished result | Cross-cutting constraints stated once. | -| 6 | `## Handoff` | An artefact, decision, or responsibility meaningfully transfers | Recipient, transferred result, evidence, and remaining responsibility. | +| Order | Section | Use when | Contains | +| ----- | ------------------ | --------------------------------------------------------------- | ---------------------------------------------------------------------- | +| 1 | `# Skill Name` | Always | Human-readable skill name. | +| 2 | Introduction | Always | One sentence stating what the skill enables. | +| 3 | `## Prerequisites` | Execution must check conditions before starting | Required state and the action taken when it is absent. | +| 4 | `## Process` | Always | The behaviour that produces the result. | +| 5 | `## Rules` | One invariant governs several steps or the finished result | Cross-cutting constraints stated once. | +| 6 | `## Handoff` | An artefact, decision, or responsibility meaningfully transfers | Recipient, transferred result, evidence, and remaining responsibility. | Add another H2 only for a distinct behavioural concern that does not fit this spine. Never create an empty or decorative section. From 6b8fb309e8b9db67c622afa43dc217ebd9d79ff3 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 22 Aug 2026 18:44:05 +0100 Subject: [PATCH 75/79] refactor(maintain-ubiquitous-language): update resolved terms inline --- skills/maintain-ubiquitous-language/SKILL.md | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/skills/maintain-ubiquitous-language/SKILL.md b/skills/maintain-ubiquitous-language/SKILL.md index 1e4d9da..affcd2f 100644 --- a/skills/maintain-ubiquitous-language/SKILL.md +++ b/skills/maintain-ubiquitous-language/SKILL.md @@ -1,25 +1,25 @@ --- name: maintain-ubiquitous-language -description: Maintains confirmed project-specific language in CONTEXT.md. Use when explicitly invoked to add, change, challenge, rename, deprecate, or reconcile a project term. +description: Maintains project-specific language in CONTEXT.md. Use when explicitly invoked to add, change, challenge, rename, deprecate, or reconcile a project term. disable-model-invocation: true --- # Maintain Ubiquitous Language -Adapted **Ubiquitous Language** keeps project-specific terminology consistent in one root `CONTEXT.md` glossary. +Applies **Ubiquitous Language** by keeping project-specific terminology consistent in one root `CONTEXT.md` glossary. ## Process ### 1. Inspect the language -Inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the project evidence needed to understand the introduced, changed, misused, or contradicted language. Challenge inconsistent uses of established terms. Make the candidate term, current meaning, and any contradiction explicit without loading unrelated project context. +Inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the project evidence needed to understand the language. Treat a stable project-specific meaning as resolved when that context makes it clear. Challenge uses that conflict with an established term. -### 2. Resolve the meaning +### 2. Update inline -Accept terminology established by authoritative project evidence or explicitly agreed with the user. A caller's agreed in-scope update supplies authority to persist that meaning without another permission prompt. When a material meaning, conflict, rename, or deprecation remains unresolved, ask the user one direct question that distinguishes the current and proposed meanings and recommends one only when the evidence supports it. Leave the existing glossary unchanged until the answer confirms one meaning. +Update `CONTEXT.md` as each meaning resolves, before the surrounding work continues. Do not request approval or announce the skill. Add no separate verification or handoff for a routine glossary update; follow the surrounding task's normal checks and reporting. -### 3. Update the glossary inline +When ambiguity, conflict, rename, or deprecation could materially change the meaning, ask one direct question that distinguishes the alternatives. Preserve the existing entry until the answer resolves the meaning. -As confirmed language arises, update `CONTEXT.md` before the surrounding work continues. Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and the document's local format; edit only affected entries and prefer changing an existing entry to adding a duplicate. Reconcile a confirmed change across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence, and remove obsolete wording only when its replacement is confirmed and losing it will not obscure that evidence. +Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and local format. Edit only affected entries, prefer updating an entry to adding a duplicate, and reconcile resolved changes across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence. -Keep `CONTEXT.md` as a glossary of project-specific domain language. Give each canonical term one short paragraph, normally one or two sentences, using project-native wording; group terms by domain only when useful. Leave implementation and specifications to their authoritative sources. +Keep `CONTEXT.md` a glossary, not a specification or implementation record. Give each canonical term one short project-native paragraph; group terms by domain only when useful. From f6415f03b03ac4206c59d1b85d20cfdb9ab4c9d8 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sat, 22 Aug 2026 21:20:31 +0100 Subject: [PATCH 76/79] refactor(skills): remove quality harnesses --- README.md | 2 +- skills/code-review/SKILL.md | 2 +- skills/debug/SKILL.md | 2 - skills/implement/SKILL.md | 6 +- skills/quality-harnesses/SKILL.md | 66 ----------------- skills/quality-harnesses/agents/openai.yaml | 5 -- .../references/BEHAVIOURAL-EVIDENCE.md | 39 ---------- .../references/BUILD-STATIC-STRUCTURE.md | 31 -------- .../quality-harnesses/references/CATALOGUE.md | 71 ------------------- .../references/CONTRACTS-DATA.md | 21 ------ .../references/HUMAN-SPECIALIST.md | 19 ----- .../references/RUNTIME-QUALITIES.md | 25 ------- .../references/SECURITY-ADVERSARIAL.md | 19 ----- skills/tdd/SKILL.md | 4 +- 14 files changed, 7 insertions(+), 305 deletions(-) delete mode 100644 skills/quality-harnesses/SKILL.md delete mode 100644 skills/quality-harnesses/agents/openai.yaml delete mode 100644 skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md delete mode 100644 skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md delete mode 100644 skills/quality-harnesses/references/CATALOGUE.md delete mode 100644 skills/quality-harnesses/references/CONTRACTS-DATA.md delete mode 100644 skills/quality-harnesses/references/HUMAN-SPECIALIST.md delete mode 100644 skills/quality-harnesses/references/RUNTIME-QUALITIES.md delete mode 100644 skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md diff --git a/README.md b/README.md index 630d3b7..0684ab2 100644 --- a/README.md +++ b/README.md @@ -97,7 +97,7 @@ $implement The session produces a minimal local change with retained tests, applicable quality evidence, and completed independent review. Small, understood work can begin here without a product definition, feature specification, or ticket. Implementation does not commit or publish the change. -`$implement` applies `$modular-design` and `$quality-harnesses`, uses `$tdd` for behaviour-changing work when a usable suite can exercise it, and requires an independent `$code-review`. It uses existing project infrastructure and reports unavailable evidence instead of installing unrelated tooling or manufacturing confidence. +`$implement` applies `$modular-design`, uses `$tdd` for behaviour-changing work when a usable suite can exercise it, and requires an independent `$code-review`. It uses existing project infrastructure and reports unavailable evidence instead of installing unrelated tooling or manufacturing confidence. #### `$commit` diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 40355f6..694093f 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -34,7 +34,7 @@ Direct the Standards reviewer to understand every changed line in its necessary - repository-required and changed-risk harness selection, measurement-path integrity, and honest evidence limits; and - recognised maintainability shapes and specialist risks exposed by the change. -Require the Standards reviewer to load and apply `$modular-design` and `$quality-harnesses`, plus `$tdd` when its prerequisites hold. These skills supply teaching knowledge within the Standards review; they do not become delegated workflows or own findings, commands, remediation, or a verdict. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. +Require the Standards reviewer to load and apply `$modular-design`, plus `$tdd` when its prerequisites hold. These skills supply teaching knowledge within the Standards review; they do not become delegated workflows or own findings, commands, remediation, or a verdict. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. ### 3. Run independent reviews diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index c194bd4..5be7e8d 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -52,8 +52,6 @@ Apply `$tdd` when its prerequisites hold. The established signal supplies Red on Run the original signal against the repair. Retain the attempt only when the result changes as the causal account predicted. When it does not, record the contradiction, revert only the exact changes owned by that attempt without disturbing pre-existing work, and return to the causal loop. Do not stack another unsupported fix. -Apply `$quality-harnesses` to the fixed candidate and preserve its claim-level evidence and missing-evidence limits. Add or retain regression protection only through existing infrastructure and a credible behavioural seam. - ### 4. Review and verify the local repair Freeze the complete candidate, intended-behaviour authority, causal account, original signal, regression evidence, selected quality evidence, changed and untracked paths, and repository state. Invoke `$code-review` on that exact candidate. Adjudicate every finding against the request, authorities, causal evidence, and current code. Correct supported required findings, ask the user about material behaviour, architecture, or scope decisions, and reject unsupported or out-of-scope findings with concrete evidence. diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 4a6d00c..624d8c9 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -18,13 +18,13 @@ Keep the outcome to one implementation-ready ticket, diagnosed repair, or small ### 2. Load the teaching knowledge -Invoke `$modular-design` and `$quality-harnesses` against the fixed work. Invoke `$tdd` when the change affects observable behaviour and an existing usable suite can exercise it through a credible seam. Retain their applicable guidance throughout implementation; they teach within this workflow while this skill retains ownership of edits, commands, evidence, and completion. +Invoke `$modular-design` against the fixed work. Invoke `$tdd` when the change affects observable behaviour and an existing usable suite can exercise it through a credible seam. Retain their applicable guidance throughout implementation; they teach within this workflow while this skill retains ownership of edits, commands, evidence, and completion. ### 3. Build and check the candidate -Implement the smallest coherent change that completely satisfies the confirmed outcome. Follow the selected modular constraints, TDD cycles, and quality-harness guidance without adding speculative flexibility, unrelated cleanup, or project-wide quality infrastructure outside the agreed work. +Implement the smallest coherent change that completely satisfies the confirmed outcome. Follow the selected modular constraints and TDD cycles without adding speculative flexibility, unrelated cleanup, or project-wide quality infrastructure outside the agreed work. -Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run every applicable repository-required and risk-triggered existing harness selected through `$quality-harnesses`. Preserve exact results, unavailable evidence, and residual risks. Freeze one complete candidate only after its local checks are current. +Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run applicable repository-required checks. Preserve exact results, unavailable evidence, and residual risks. Freeze one complete candidate only after its local checks are current. ### 4. Review and adjudicate diff --git a/skills/quality-harnesses/SKILL.md b/skills/quality-harnesses/SKILL.md deleted file mode 100644 index d4ab88c..0000000 --- a/skills/quality-harnesses/SKILL.md +++ /dev/null @@ -1,66 +0,0 @@ ---- -name: quality-harnesses -description: Teaches risk-triggered selection and honest interpretation of software quality harnesses for a fixed change. Use when implementing or assessing code, tests, configuration, migrations, generated artefacts, or other software changes. -disable-model-invocation: true ---- - -# Quality Harnesses - -Teaches a caller to apply every repository-required and changed-risk harness that can provide faithful evidence without treating green tools, metrics, or control count as proof. - -## Process - -### 1. Fix the evidence boundary - -Read the caller's outcome, exact candidate or change scope, intended behaviour and other authorities, repository instructions, changed implementation and tests, available quality infrastructure, and material environment constraints. Identify: - -- behaviour, contracts, data, dependencies, generated output, configuration, and operational qualities changed or relied upon; -- repository-required checks and their actual scopes; -- material failure modes introduced or altered; and -- measurement-path changes, including tests, assertions, snapshots, fixtures, generators, seeds, selectors, exclusions, suppressions, thresholds, baselines, scripts, tool configuration, and cached or generated evidence. - -Treat earlier evidence as stale when its candidate, authority, harness, configuration, inputs, environment, or measurement path has materially changed. The fixed evidence boundary and every plausible changed risk are explicit. - -### 2. Select through the complete catalogue - -Read the complete [Harness Catalogue](references/CATALOGUE.md) on every invocation. For each entry, compare its trigger with the fixed boundary. Select an entry when repository authority requires it or the change exposes its named risk; neither familiarity nor low cost is required. Reject entries with no trigger even when their tooling exists. - -Load the linked detail for every selected entry and only selected entries. When entries share one reference, read it once and apply each selected section. Preserve repository-owned commands, versions, matrices, thresholds, and procedures; generic guidance does not replace them. - -The catalogue routes selected entries to [Build, Static, and Structural Harnesses](references/BUILD-STATIC-STRUCTURE.md), [Behavioural Evidence](references/BEHAVIOURAL-EVIDENCE.md), [Security and Adversarial Harnesses](references/SECURITY-ADVERSARIAL.md), [Contracts and Data](references/CONTRACTS-DATA.md), [Runtime Qualities](references/RUNTIME-QUALITIES.md), and [Human and Specialist Evidence](references/HUMAN-SPECIALIST.md). Do not load these references until their catalogue trigger applies. - -### 3. Apply the selected knowledge - -Apply the loaded guidance inside the caller's workflow. The caller owns commands, edits, remediation, evidence, and stopping. Where dependencies allow, use fast focused feedback before broader or specialist evidence, but never substitute a cheaper harness for the faithful one a claim requires. - -Selection precedes feasibility. Keep every triggered entry selected even when its existing harness is unreliable, unrepresentative, unsafe, unavailable, or outside the execution boundary; record the result as inconclusive or not run and leave the claim unassessed. Do not relabel an applicable entry as optional or irrelevant merely because faithful evidence is expensive or missing. - -Before trusting a result, confirm that the intended target, cases, rules, inputs, and environments actually ran and that the oracle can distinguish the material failure. Treat empty selectors, tautological expectations, stale baselines, unexplained exclusions or suppressions, unrepresentative inputs, hidden retries, and changed thresholds as measurement defects rather than clean evidence. - -Use existing infrastructure. Add or improve behavioural tests through the existing suite when the implementation requires them and the caller has authority. Do not install or establish project-wide harness infrastructure unless it is part of the agreed work. Do not weaken or skip a repository-required check silently. When an applicable harness is absent, unsafe, unavailable, or outside scope, leave the claim unassessed and record the residual risk. - -### 4. Preserve claim-level evidence - -For each selected harness, retain: - -- **Claim:** the exact behaviour, repository rule, contract, invariant, or risk assessed; -- **Provenance:** why the harness applies and the repository command or reproducible procedure used; -- **Boundary:** candidate, environment, configuration, fixtures or data, seed, tool version, and external-state constraints where material; -- **Result:** pass, fail, flaky or inconclusive, or not run, with salient output and artefact location; -- **Coverage:** the cases, paths, rules, versions, platforms, states, or criteria actually exercised; -- **Limitation:** what the result cannot establish and any measurement weakness; and -- **Residual risk:** the consequence still unsupported and the in-scope correction, later work, or user decision needed for stronger evidence. - -Record a field only when it materially changes interpretation or reproduction for that harness; omit immaterial metadata. Use bounded claims such as “the selected examples passed in this environment.” Do not promote compilation, execution, coverage, mutation, static analysis, or a collection of green controls into proof of correctness, a confidence percentage, or an overall verification verdict. - -### 5. Return control - -Return the selected and rejected catalogue entries with their triggers, applicable commands or procedures, claim-level evidence, unavailable evidence, and residual risks. End by stating that execution, evidence capture, remediation, and completion remain with the caller. Stop after the guidance has been applied within the caller's current boundary. Do not take ownership of implementation, independent review, commit, publication, release, deployment, or the caller's completion decision. - -## Rules - -- Treat complexity, cognitive complexity, size, churn, coupling, coverage, mutation, and smell counts as evidence for a named question, never as universal quality gates or findings without a concrete mechanism and consequence. -- Preserve flaky or conflicting results as unreliable evidence; never turn them into a pass through unrecorded retries, averaging, suppression, or selective omission. -- Prefer several independent evidence routes over repeated controls that share the same oracle, configuration, implementation assumption, or failure mode. -- Never state or imply that the candidate is ready, not ready, approved, rejected, verified, unverified, blocked, or clear to continue. Report claim-level evidence and return the completion decision to the caller. -- Return decisions to the caller or an authority already named by the caller or project. Do not invent owners, reviewers, specialists, approval bodies, or coordination. diff --git a/skills/quality-harnesses/agents/openai.yaml b/skills/quality-harnesses/agents/openai.yaml deleted file mode 100644 index c04dca6..0000000 --- a/skills/quality-harnesses/agents/openai.yaml +++ /dev/null @@ -1,5 +0,0 @@ -interface: - display_name: 'Quality Harnesses' - short_description: 'Select risk-matched software quality evidence' -policy: - allow_implicit_invocation: false diff --git a/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md b/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md deleted file mode 100644 index c4c20f1..0000000 --- a/skills/quality-harnesses/references/BEHAVIOURAL-EVIDENCE.md +++ /dev/null @@ -1,39 +0,0 @@ -# Behavioural Evidence - -## Behavioural and regression suites - -Map each changed observable outcome to an oracle capable of disagreeing with the implementation. When `$tdd` applies, use its public result for Red/Green sensitivity, durable boundary, and oracle evidence rather than recreating its workflow here. Treat captured current output as characterization until an independent authority accepts it. - -Run the narrowest faithful focused target while changing behaviour, then relevant nearby tests and every repository-required broader suite. Distinguish: - -- new-behaviour or defect evidence from focused changed tests; -- integration evidence that retains the production technology or boundary at risk; and -- pass-to-pass regression evidence from previously passing tests. - -Confirm discovery and selected-case counts, intended failure sensitivity, complete assertions, negative or unchanged outcomes, fixture representativeness, cleanup, deterministic time and randomness, and meaningful boundary interactions. Inspect changed snapshots, fixtures, doubles, skips, quarantines, retries, timeouts, and test configuration. A shared implementation helper, regenerated snapshot, “does not throw,” empty selector, or incidental call assertion is not a faithful oracle unless that exact representation or interaction is the contract. - -Preserve the exact command, result, case count, environment, and failing counterexample. Report flaky evidence as flaky or inconclusive; diagnose or rerun only through an explicit recorded procedure, never until green. - -## Coverage - -Use coverage to locate changed or risk-relevant statements, branches, conditions, states, and error paths that the selected tests never execute. Start from the claim and ask whether an uncovered path can alter it. Add a test only when an independent oracle and credible seam can make that path observable; executed code without a discriminating assertion is not stronger evidence. - -Inspect the measured denominator, included packages, changed-code mapping, branch versus line criteria, exclusions, generated code, merged reports, stale artefacts, and coverage configuration changes. Preserve repository-owned thresholds, but do not invent a target or raise execution merely to improve a number. - -Representative contrast: use a report to find that the duplicate-conflict branch never ran, then exercise its externally visible settlement result. Do not add a test that calls the branch and asserts only that it returned. - -## Mutation - -Use existing mutation tooling only when the bounded behaviour and suite are stable, test sensitivity remains materially uncertain, and execution is safe and economical. Target the changed or risk-bearing area rather than an unrelated whole repository unless authority requires otherwise. - -Interpret each category: - -- **Killed:** the suite detected that particular non-equivalent injected change. -- **Survived:** determine whether the mutant represents a plausible promised-behaviour defect; strengthen a faithful oracle only when it does. -- **Uncovered:** use coverage and claim tracing to decide whether the location affects material behaviour. -- **Equivalent or irrelevant:** retain a concrete justification; do not count it as detected. -- **Timed out or errored:** record unreliable or missing evidence and diagnose when in scope. - -Inspect operator selection, changed-code mapping, exclusions, baseline failures, timeouts, retries, mutant filters, and whether a proposed test merely couples to implementation. Never use the aggregate score as correctness, a universal threshold, or permission to add low-value assertions. - -Representative contrast: a surviving change that removes the uniqueness guard requires a test of duplicate settlement behaviour; a mutant that changes an unreachable diagnostic string may be irrelevant to the fixed claim. diff --git a/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md b/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md deleted file mode 100644 index 4477548..0000000 --- a/skills/quality-harnesses/references/BUILD-STATIC-STRUCTURE.md +++ /dev/null @@ -1,31 +0,0 @@ -# Build, Static, and Structural Harnesses - -## Build, compilation, and packaging - -Use the repository-owned command, target, production mode, and dependency state. Record the source revision, environment, toolchain, inputs, output artefact, exit status, and material warnings. Confirm that the command rebuilt the intended target rather than accepting stale or cached output; inspect changed scripts, lockfiles, generators, ignored failures, omitted workspaces, and warning policy. - -A clean result supports only that the recorded inputs compiled or assembled in that environment. It does not execute promised behaviour or establish portability, compatibility, deployment, or reproducibility. When packaging changes, inspect the produced archive or manifest for required files, entry points, permissions, metadata, and absence of unintended material rather than relying on exit status alone. - -Representative procedure: resolve the authoritative build command; clear or invalidate only project-sanctioned disposable output when needed; run the exact target; verify a fresh artefact and its manifest; preserve salient output and environment. If cleanup is unsafe or unauthorised, record the cache limitation. - -## Types, linting, formatting, and framework diagnostics - -Resolve the configured check command and its real file, project, rule, and mode scope. Run repository-required checks even when a narrower invocation already passed. For changed configuration, compare selection, rule severity, exclusions, suppressions, baselines, generated-code treatment, and formatter check/write mode before interpreting the result. - -A clean type check supports only relationships encoded by that type system and configuration. A linter or framework diagnostic supports only enabled rules. Formatting supports only the selected canonical representation. Confirm a non-empty target and preserve warnings or diagnostics that the command reports without failing. - -Do not add suppressions, widen ignores, lower severity, regenerate a baseline, or format unrelated files merely to make a check green. An authorised exception remains an explicit limitation with its scope and rationale. - -## Architecture fitness rules - -First identify the authoritative structural decision the rule encodes. Run the existing rule only when the change can affect its dependency, layer, cycle, import, ownership, or public-surface selection. Confirm that source selectors, module mappings, and expected violations are non-empty and still match renamed or moved code. Inspect changed exemptions, baseline files, generated-source treatment, and rule configuration. - -A pass supports conformance to the encoded rule for the selected source. It cannot decide whether the rule is desirable, whether omitted boundaries are healthy, or whether runtime, data, semantic, or organisational coupling is acceptable. Use `$modular-design` for structural judgement when its public trigger applies; do not turn a new architectural preference into a gate without project authority. - -Representative contrast: a dependency rule that scans 86 current modules and finds no forbidden imports is evidence; the same green command after its package selector matches zero modules is not. - -## Reproducibility and provenance - -Identify the promised equivalence: byte-identical artefacts, canonical content, generated-source stability, dependency provenance, or signed-output traceability. Record source revision, toolchain, resolved dependencies, environment, locale, time inputs, network access, generation command, and comparison method. Repeat under the relevant independent variance rather than only rerunning against the same caches and machine state. - -Inspect embedded timestamps and paths, undeclared state, caches, dependency drift, ignored comparison fields, signatures, attestations, and whether regenerated output was reviewed and checked into the intended boundary. A successful same-machine repeat is useful determinism evidence but not cross-environment reproducibility. Equivalent artefacts do not establish behavioural correctness or trustworthy provenance by themselves. diff --git a/skills/quality-harnesses/references/CATALOGUE.md b/skills/quality-harnesses/references/CATALOGUE.md deleted file mode 100644 index b784274..0000000 --- a/skills/quality-harnesses/references/CATALOGUE.md +++ /dev/null @@ -1,71 +0,0 @@ -# Harness Catalogue - -Review every entry against the fixed change. Applicability comes from repository authority or the stated trigger, not from a baseline/optional tier. - -## Build, compilation, and packaging - -**Trigger:** A repository-required build, or changed build inputs, generated output, dependencies, packaging, compilation mode, or runtime compilation. **Narrow claim:** The selected inputs produced the identified artefact or compiled successfully in the recorded environment. **Limitations:** One build does not establish behavioural correctness, portability, reproducibility, deployability, or runtime compatibility. **Anti-gaming checks:** Confirm the intended target and configuration ran; inspect stale or cached output, ignored errors, warnings-as-errors policy, omitted packages, and changed build scripts. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#build-compilation-and-packaging). - -## Types, linting, formatting, and framework diagnostics - -**Trigger:** Repository-required diagnostics, or changed code or configuration within an existing tool's scope. **Narrow claim:** The checked files satisfy the enabled rules under the recorded configuration. **Limitations:** Clean diagnostics establish only encoded properties and cannot prove behaviour, security, accessibility, or maintainability. **Anti-gaming checks:** Confirm file and rule selection is non-empty and current; inspect changed ignores, baselines, suppressions, severity, generated-file treatment, and formatter write-versus-check mode. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#types-linting-formatting-and-framework-diagnostics). - -## Behavioural and regression suites - -**Trigger:** Changed observable behaviour, a defect repair, changed tests or test infrastructure, or a repository-required suite. **Narrow claim:** The exercised examples produced their independently specified outcomes; a broader unchanged suite also supplies pass-to-pass regression evidence. **Limitations:** Tests say nothing about unexercised cases, environments, or risks and are only as sound as their seams and oracles. **Anti-gaming checks:** Inspect test discovery, focused-case counts, assertions, snapshots, fixtures, doubles, retries, skips, quarantines, and whether expected values derive from production. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#behavioural-and-regression-suites). - -## Coverage - -**Trigger:** Execution of changed or risk-relevant paths is uncertain, coverage configuration changed, or repository authority requires the measurement. **Narrow claim:** The selected tests executed the reported code elements. **Limitations:** Execution percentage does not establish correct assertions, fault detection, requirement coverage, or overall quality. **Anti-gaming checks:** Inspect denominator and scope changes, exclusions, generated code, branch versus line criteria, merged reports, stale artefacts, and tests that execute without discriminating outcomes. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#coverage). - -## Mutation - -**Trigger:** Test sensitivity is materially uncertain and existing mutation tooling can assess a bounded changed area economically, or repository authority requires it. **Narrow claim:** Killed non-equivalent mutants show sensitivity to those particular injected changes. **Limitations:** Operators omit fault classes; survivors may be equivalent or irrelevant; timeouts and uncovered mutants are distinct; no score proves correctness. **Anti-gaming checks:** Inspect changed-code selection, mutant categories, exclusions, operator configuration, baseline stability, timeouts, and whether fixes merely assert implementation details. **Reference:** [Behavioural Evidence](BEHAVIOURAL-EVIDENCE.md#mutation). - -## Static and security analysis - -**Trigger:** Configured analysis covers the changed code, repository authority requires it, or the change affects trust boundaries, permissions, sensitive data, untrusted input, cryptography, secrets, or dependency risk. **Narrow claim:** The enabled rules reported no unresolved findings for their scanned inputs, or produced findings for appraisal. **Limitations:** Analysis approximates selected weakness classes and cannot establish absence of vulnerabilities or sound security design. **Anti-gaming checks:** Record tool and rule versions; inspect scan scope, baselines, suppressions, severity changes, ignored paths, generated findings, and dependency database age. **Reference:** [Security and Adversarial Harnesses](SECURITY-ADVERSARIAL.md#static-and-security-analysis). - -## Architecture fitness rules - -**Trigger:** An authoritative dependency, layer, cycle, import, ownership, or public-surface rule is already encoded and the change can affect it. **Narrow claim:** The selected source conforms to the encoded structural rule. **Limitations:** Passing cannot establish that the rule is a good architecture decision or cover omitted runtime, data, or organisational coupling. **Anti-gaming checks:** Confirm selectors and mappings match non-empty current source; inspect exemptions, generated code, renamed modules, baseline changes, and rules disabled or weakened by the change. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#architecture-fitness-rules). - -## Contracts and compatibility - -**Trigger:** Changed or relied-upon APIs, events, schemas, serialization, storage formats, runtimes, browsers, databases, protocols, generated clients, or supported versions. **Narrow claim:** Named producers, consumers, examples, or matrix cells conform to the exercised contract. **Limitations:** Finite examples and environments cannot establish every consumer interpretation, historical payload, or supported combination. **Anti-gaming checks:** Confirm consumer and version inventory, matrix cells, backward/forward direction, optional and unknown fields, real serialization, stale examples, normalisation, and provider/consumer independence. **Reference:** [Contracts and Data](CONTRACTS-DATA.md#contracts-and-compatibility). - -## Fuzzing and adversarial input - -**Trigger:** Parsers, protocols, decoders, validators, or hostile/high-volume inputs changed, and an existing harness has meaningful crash, sanitizer, assertion, or invariant oracles. **Narrow claim:** No selected oracle failed for the exercised corpus and generated inputs, or a preserved counterexample exposed a defect. **Limitations:** Execution alone rarely establishes semantic correctness; generators and corpora sample only part of the space. **Anti-gaming checks:** Preserve corpus, seed, duration, minimized failures, oracle and sanitizer configuration; confirm non-empty input selection and inspect timeouts, rejected-input bias, coverage plateaus, and swallowed failures. **Reference:** [Security and Adversarial Harnesses](SECURITY-ADVERSARIAL.md#fuzzing-and-adversarial-input). - -## Accessibility - -**Trigger:** Changed visible interface, semantics, focus, input, navigation, content, contrast, zoom, motion, or assistive-technology behaviour, or repository authority requires it. **Narrow claim:** The exercised automated rules and recorded manual interactions satisfy their named criteria in the tested environment. **Limitations:** Automation alone cannot establish conformance, usability, or behaviour across assistive technologies and user contexts. **Anti-gaming checks:** Inspect tested route and state selection, hidden or disabled elements, rule suppressions, keyboard path completeness, focus order, zoom/contrast conditions, and actual assistive-technology coverage. **Reference:** [Human and Specialist Evidence](HUMAN-SPECIALIST.md#accessibility). - -## Performance and load - -**Trigger:** Changed hot paths, algorithms, queries, I/O, payloads, caching, memory, concurrency, capacity assumptions, or an explicit latency, throughput, or resource budget. **Narrow claim:** The representative workload met or differed from the pinned budget or baseline by the reported distribution in the recorded environment. **Limitations:** Synthetic or local results may not predict production workload, tail latency, contention, cost, or capacity. **Anti-gaming checks:** Preserve workload, data scale, warm-up, repetitions, variance, environment, correctness assertions, baseline revision, outlier policy, and measurement overhead. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#performance-and-load). - -## Concurrency and ordering - -**Trigger:** Changed shared state, locks, transactions, parallelism, asynchronous cancellation, retries, queues, ordering, or idempotency. **Narrow claim:** The exercised schedules or stresses preserved the named invariants and outcomes. **Limitations:** Bounded schedules and successful stress runs cannot prove freedom from races, deadlocks, starvation, or production-specific timing defects. **Anti-gaming checks:** Confirm the intended scenario and a relevant execution count are non-empty; preserve a schedule or seed when used; assert outcomes as well as absence of crashes; inspect hidden serialization, disabled race detection, retries, timeouts, and environment fidelity. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#concurrency-and-ordering). - -## Resilience and fault injection - -**Trigger:** Changed timeouts, retries as recovery policy, failover, partial dependency failure, recovery, rollout, or operational degradation. **Narrow claim:** The injected fault occurred and the observed system met the explicit recovery or degradation hypothesis within the exercised boundary. **Limitations:** Injected scenarios cannot enumerate production combinations or establish operational readiness outside the observed environment. **Anti-gaming checks:** Verify fault delivery, safe isolation, observability, recovery oracle, cleanup, steady state before and after, blast-radius controls, and that retries do not hide failure. **Reference:** [Runtime Qualities](RUNTIME-QUALITIES.md#resilience-and-fault-injection). - -## Migration and data integrity - -**Trigger:** Changed schemas, transformations, backfills, imports, storage semantics, retention, mixed-version operation, rollback, or recovery promises. **Narrow claim:** Representative source states transformed or recovered while preserving the asserted invariants. **Limitations:** Fixtures cannot represent every production distribution, invalid state, scale, lock interaction, or irreversible operational consequence. **Anti-gaming checks:** Inspect fixture provenance and diversity, row and aggregate invariants, idempotence/resume, mixed versions, failure interruption, rollback or forward recovery, constraint timing, and excluded legacy states. **Reference:** [Contracts and Data](CONTRACTS-DATA.md#migration-and-data-integrity). - -## Reproducibility and provenance - -**Trigger:** Changed toolchains, dependency resolution, generated artefacts, signed output, provenance metadata, or environment-sensitive builds, or an explicit reproducibility promise. **Narrow claim:** Recorded material inputs produced equivalent artefacts under the exercised comparison and variance. **Limitations:** Same-machine repetition is not cross-environment reproducibility; equivalent bytes do not establish behaviour or supply-chain trust. **Anti-gaming checks:** Record toolchain, dependencies, source revision, environment, timestamps and comparison method; vary relevant inputs; inspect caches, undeclared network/state, embedded paths, signatures, and ignored differences. **Reference:** [Build, Static, and Structural Harnesses](BUILD-STATIC-STRUCTURE.md#reproducibility-and-provenance). - -## Manual and specialist QA - -**Trigger:** Automation cannot faithfully assess a material visual, physical, usability, domain, operational, data, compliance, or contextual outcome. **Narrow claim:** The named procedure produced the recorded observation under its inputs and evaluator capability. **Limitations:** Results are bounded by procedure, environment, evaluator expertise and subjectivity, and do not generalise automatically. **Anti-gaming checks:** Preserve steps, inputs, versions, expected and actual results, captures, evaluator limitation, repetitions where needed, and deviations from the procedure. **Reference:** [Human and Specialist Evidence](HUMAN-SPECIALIST.md#manual-and-specialist-qa). - -## Diagnostic metrics - -Complexity, cognitive complexity, size, churn, coupling, and smell counts apply only when repository authority requires them or a changed hotspot needs investigation. Preserve repository thresholds. Trace the concrete control flow, ownership, dependency, or change-propagation consequence before treating a signal as actionable; a metric alone is neither a defect nor evidence that a refactor improves the change. diff --git a/skills/quality-harnesses/references/CONTRACTS-DATA.md b/skills/quality-harnesses/references/CONTRACTS-DATA.md deleted file mode 100644 index 8d60331..0000000 --- a/skills/quality-harnesses/references/CONTRACTS-DATA.md +++ /dev/null @@ -1,21 +0,0 @@ -# Contracts and Data - -## Contracts and compatibility - -Name the contract, producer and consumer directions, supported versions, serialization or transport, optional and unknown-field policy, and compatibility promise. Apply existing consumer/provider tests, schema checks, golden examples, matrix runs, or real round trips to every material changed or relied-upon cell. - -Confirm that examples are current, independently owned where possible, and serialized through the real boundary rather than reconstructed from the same implementation. Inspect defaulting, ordering, precision, encoding, unknown fields, errors, feature negotiation, generated clients, and changed normalisation. Record every producer, consumer, version, runtime, browser, database, or platform actually exercised and every promised cell omitted. - -A clean result supports only those named examples or matrix cells. Agreement between implementations that share code or assumptions can be jointly wrong. Unchanged public syntax does not establish unchanged semantics. - -Representative contrast: current and previous event examples passing supports those two payloads; it does not support all historical `v2` events or unknown downstream interpretations. - -## Migration and data integrity - -Identify source states, target invariants, scale and distribution assumptions, mixed-version behaviour, interruption points, and the promised rollback or forward-recovery path. Use maintained fixtures or safe representative copies containing ordinary, boundary, duplicate, null, malformed, legacy, and partially processed states that are plausible under project authority. - -For each exercised state, check row and aggregate counts, identifiers, uniqueness, referential integrity, semantic values, ordering, permissions, timestamps, and other domain invariants before and after transformation. Exercise idempotent rerun or resume, constraint timing, mixed application versions, interrupted execution, and rollback or forward recovery when promised. Record engine version, transaction mode, data scale, duration, locks, warnings, and cleanup. - -Never run destructive migration experiments against durable or production data without explicit authority and a safe boundary. Passing fixtures cannot represent every production distribution, invalid state, scale, lock interaction, or irreversible operational effect. When automation cannot faithfully appraise operational safety or domain meaning, also apply [Manual and Specialist QA](HUMAN-SPECIALIST.md#manual-and-specialist-qa); if unavailable, preserve the gap. - -Representative contrast: a deterministic backfill over ordinary and duplicate legacy fixtures is evidence for those states; absent null and malformed fixtures leave their transformation unassessed. diff --git a/skills/quality-harnesses/references/HUMAN-SPECIALIST.md b/skills/quality-harnesses/references/HUMAN-SPECIALIST.md deleted file mode 100644 index e8bb83b..0000000 --- a/skills/quality-harnesses/references/HUMAN-SPECIALIST.md +++ /dev/null @@ -1,19 +0,0 @@ -# Human and Specialist Evidence - -## Accessibility - -Resolve the applicable product and accessibility authority, changed routes and states, supported input modes, browsers, and assistive technologies. Run configured automation against the actual rendered state, then perform the manual interactions the change exposes: keyboard navigation, visible focus, focus order and restoration, labels and semantics, error identification, zoom and reflow, contrast, motion, and relevant screen-reader or other assistive-technology behaviour. - -Record route, state, content, viewport, zoom, theme, browser, platform, assistive technology and version, automated rules, manual steps, expected outcome, actual outcome, and captures where useful. Inspect hidden states, disabled elements, overlays, asynchronous updates, rule suppressions, partial keyboard paths, and tests against non-production markup. - -Automation supports only enabled rule checks. Manual results support only the exercised environment and evaluator capability. Neither alone establishes conformance or usability across users and technologies; unavailable required coverage remains an explicit gap. - -## Manual and specialist QA - -Use a reproducible manual procedure when automation cannot faithfully observe a material visual, physical, usability, domain, data, compliance, or contextual result. Use specialist assessment only when the claim requires knowledge the current caller and executable harnesses do not supply; do not simulate or invent that expertise. - -Define the claim, inputs, initial state, environment, steps, expected result from an independent authority, observation method, cleanup, and acceptable variation before execution. Preserve actual results, captures or measurements, versions, repetitions where variability matters, deviations, and the evaluator's relevant capability or limitation. - -A recorded observation supports only that procedure and context. Subjectivity, evaluator knowledge, unavailable equipment or data, and untested states remain limitations. When the required person, environment, service, device, data, or authority is unavailable, mark the claim unassessed and return the residual risk to the caller; do not install infrastructure, fabricate a result, or convert generic inspection into specialist approval. - -Representative contrast: a database specialist can assess lock duration and operational backfill safety from production-scale evidence; an implementation agent reading the migration can identify questions but cannot claim that specialist assessment occurred. diff --git a/skills/quality-harnesses/references/RUNTIME-QUALITIES.md b/skills/quality-harnesses/references/RUNTIME-QUALITIES.md deleted file mode 100644 index e237d42..0000000 --- a/skills/quality-harnesses/references/RUNTIME-QUALITIES.md +++ /dev/null @@ -1,25 +0,0 @@ -# Runtime Qualities - -## Performance and load - -Start from an authoritative budget, SLO, pre-change baseline, or concrete regression risk. Use a representative workload, data scale, concurrency, cache state, warm-up, duration, repetitions, and environment. Assert behavioural correctness during the measurement; a fast wrong result is not evidence. - -Preserve distributions and tail values rather than one average, variance and outlier policy, resource consumption, baseline revision, tool overhead, and environmental noise. Compare like with like and rerun enough to distinguish a material effect from noise. Inspect reduced workloads, changed data, warmed caches, hidden retries, omitted errors, and thresholds moved after seeing results. - -A clean local benchmark supports only the exercised workload and environment. It does not establish production tail latency, capacity, contention, cost, or scalability beyond measured ranges. - -## Concurrency and ordering - -Name the invariant and realistic failure: duplicate effect, lost update, stale read, reordered event, deadlock, starvation, cancellation leak, retry amplification, or broken idempotency. Use existing race detection, deterministic schedule exploration, model checking, controlled interleavings, or seeded stress at the narrowest boundary that retains the shared state and transaction semantics at risk. - -Assert final outcomes, intermediate invariants where observable, and absence of duplicate or missing effects; do not rely only on no crash. Confirm that the intended scenario and its relevant execution count are non-empty. Preserve the schedule or seed and environment details only when material to interpretation or reproduction. Inspect accidental serialization, mocks that remove shared state, selectors that match zero cases, disabled detectors, swallowed task errors, and retries that mask nondeterminism. - -A bounded clean run supports only its exercised schedules and environment. Stress cannot prove race freedom. When production-like scheduling or specialist concurrency assessment is unavailable, record the remaining failure mode rather than increasing an arbitrary iteration count. - -## Resilience and fault injection - -State an explicit hypothesis: which dependency or resource fails, how the system should degrade or recover, what state must remain safe, and what users or operators observe. Use a safe isolated boundary with an existing fault mechanism and current observability. Establish steady state, inject the fault, confirm it actually occurred, assert degradation and recovery outcomes, remove the fault, and verify cleanup and restored state. - -Record fault type, timing, duration, affected calls or nodes, environment, observability signals, recovery window, data invariants, retries, and residual effects. Inspect no-op injection, fallbacks that bypass the target, hidden retries, insufficient blast-radius controls, missing recovery oracle, and cleanup failures. - -One successful experiment cannot enumerate production combinations or prove rollout readiness. Release, deployment, rollback execution, and production monitoring remain with their owning workflows. diff --git a/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md b/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md deleted file mode 100644 index cf4ca65..0000000 --- a/skills/quality-harnesses/references/SECURITY-ADVERSARIAL.md +++ /dev/null @@ -1,19 +0,0 @@ -# Security and Adversarial Harnesses - -## Static and security analysis - -Map the change through trust boundaries, identities, permissions, untrusted inputs, sensitive data, secrets, cryptography, deserialization, dependencies, and external effects. Apply repository-required or already configured analysis to the code it actually covers. Add risk-focused negative or abuse-case procedures only for the changed surface and established authority. - -Record tool, rule set and vulnerability database versions, scanned inputs, build mode, findings and suppressions. Inspect changed baselines, ignored paths, generated-code treatment, severity thresholds, stale dependency data, secret-history scope, and whether the scan completed. Appraise findings against reachable code and concrete consequences; do not suppress a false positive without preserving why it is false. - -A clean scan supports only absence of findings from enabled rules over scanned inputs. It cannot establish safe design, complete data flow, correct authorization, absence of unknown weaknesses, or runtime protection. When specialist penetration or threat assessment is materially required but unavailable, record the unassessed claim and risk. - -Representative procedure: identify the changed trust-boundary claim; run existing configured analysis; exercise one authorised and one unauthorised or malformed case through the supported boundary where in scope; inspect data and effects; preserve blind spots. - -## Fuzzing and adversarial input - -Use an existing fuzz or generative harness for changed parsers, protocols, validators, decoders, or hostile/high-volume input only when it has a meaningful oracle: crash, sanitizer, invariant, differential result, resource bound, or assertion. Define the target and rejection semantics before execution. - -Preserve corpus provenance, seed, generator distribution, duration or iteration budget, sanitizer and assertion configuration, coverage or state progress, timeouts, and every minimized failing input. Confirm the selected corpus and generated-case count are non-empty. Inspect swallowed exceptions, over-filtered inputs, biased generators, duplicate cases, resource exhaustion mistaken for a product defect, and failures discarded after minimization. - -No crash establishes robustness only for exercised inputs and enabled detectors. It rarely establishes semantic correctness. A preserved counterexample becomes regression evidence only after an independent oracle identifies the correct outcome. diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 7f1b48a..7712cee 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -46,10 +46,10 @@ If refactoring exposes a new material structural decision not resolved by the ca After production refactoring, identify the unique behaviour, boundary, invariant, or risk protected by each affected test. Remove obsolete implementation-detail assertions, broad snapshots used only as change alarms, and examples that add no distinct failure detection. Consolidate equivalent cases with a table or property only when the resulting oracle and failure diagnosis remain clear. -Mentally mutate wrong constants, branches, validation, effects, and empty or default results. Strengthen the smallest test or input that would let a realistic defect survive. Use actual mutation or coverage tooling only when the caller's quality-harness guidance selects it. +Mentally mutate wrong constants, branches, validation, effects, and empty or default results. Strengthen the smallest test or input that would let a realistic defect survive. ### 6. Complete and return Repeat Red, Green, Refactor, and rationalisation for each remaining confirmed behaviour. When example cases cannot economically cover a large input or state space, or a direct oracle is unavailable, load [Generative Testing](references/GENERATIVE-TESTING.md) and use only the triggered technique. Run the complete relevant suite. -Return the behaviours delivered; retained tests and the unique protection each adds; exact Red and Green evidence; refactors; commands and results; unresolved baseline failures; fidelity limits; and other missing evidence. The caller retains ownership of broader quality harnesses, independent review, the complete work outcome, commits, and publication. +Return the behaviours delivered; retained tests and the unique protection each adds; exact Red and Green evidence; refactors; commands and results; unresolved baseline failures; fidelity limits; and other missing evidence. The caller retains ownership of broader checks, independent review, the complete work outcome, commits, and publication. From 8f098f95c548e158214a57212695c4246c6e3204 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 23 Aug 2026 10:07:23 +0100 Subject: [PATCH 77/79] feat(skills): add code complexity measurement --- skills/code-review/SKILL.md | 2 + skills/implement/SKILL.md | 4 +- skills/measure-code-complexity/SKILL.md | 52 ++ .../agents/openai.yaml | 5 + .../references/CYCLOMATIC-COMPLEXITY.md | 9 + .../references/DUPLICATION.md | 9 + .../references/FUNCTION-SIZE.md | 9 + .../references/NESTING.md | 9 + .../references/PARAMETERS.md | 9 + .../scripts/measure.py | 740 ++++++++++++++++++ .../scripts/vendor/THIRD-PARTY-NOTICES.md | 8 + .../vendor/licenses/lizard/APACHE-2.0.txt | 202 +++++ .../vendor/licenses/lizard/LICENSE.txt | 30 + .../scripts/vendor/licenses/pygments/AUTHORS | 291 +++++++ .../scripts/vendor/licenses/pygments/LICENSE | 25 + .../vendor/lizard-1.24.0-py2.py3-none-any.whl | Bin 0 -> 111418 bytes .../vendor/pygments-2.19.2-py3-none-any.whl | Bin 0 -> 1225217 bytes 17 files changed, 1403 insertions(+), 1 deletion(-) create mode 100644 skills/measure-code-complexity/SKILL.md create mode 100644 skills/measure-code-complexity/agents/openai.yaml create mode 100644 skills/measure-code-complexity/references/CYCLOMATIC-COMPLEXITY.md create mode 100644 skills/measure-code-complexity/references/DUPLICATION.md create mode 100644 skills/measure-code-complexity/references/FUNCTION-SIZE.md create mode 100644 skills/measure-code-complexity/references/NESTING.md create mode 100644 skills/measure-code-complexity/references/PARAMETERS.md create mode 100644 skills/measure-code-complexity/scripts/measure.py create mode 100644 skills/measure-code-complexity/scripts/vendor/THIRD-PARTY-NOTICES.md create mode 100644 skills/measure-code-complexity/scripts/vendor/licenses/lizard/APACHE-2.0.txt create mode 100644 skills/measure-code-complexity/scripts/vendor/licenses/lizard/LICENSE.txt create mode 100644 skills/measure-code-complexity/scripts/vendor/licenses/pygments/AUTHORS create mode 100644 skills/measure-code-complexity/scripts/vendor/licenses/pygments/LICENSE create mode 100644 skills/measure-code-complexity/scripts/vendor/lizard-1.24.0-py2.py3-none-any.whl create mode 100644 skills/measure-code-complexity/scripts/vendor/pygments-2.19.2-py3-none-any.whl diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 694093f..37438f6 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -36,6 +36,8 @@ Direct the Standards reviewer to understand every changed line in its necessary Require the Standards reviewer to load and apply `$modular-design`, plus `$tdd` when its prerequisites hold. These skills supply teaching knowledge within the Standards review; they do not become delegated workflows or own findings, commands, remediation, or a verdict. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. +Require the Standards reviewer to invoke `$measure-code-complexity` against the fixed candidate and comparison base with read-only authority. Treat its current output as diagnostic evidence: inspect each trigger in source and test context, and retain a review suggestion only when a concrete maintainability consequence survives falsification. Report unavailable or incomplete measurement in the review limitations. + ### 3. Run independent reviews Give each applicable packet to a separate fresh agent and run them in parallel where possible. Do not substitute coordinator self-review. If fresh-agent execution is unavailable, mark the affected axis not performed and state the exact limitation. diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 624d8c9..4e42e60 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -24,7 +24,9 @@ Invoke `$modular-design` against the fixed work. Invoke `$tdd` when the change a Implement the smallest coherent change that completely satisfies the confirmed outcome. Follow the selected modular constraints and TDD cycles without adding speculative flexibility, unrelated cleanup, or project-wide quality infrastructure outside the agreed work. -Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run applicable repository-required checks. Preserve exact results, unavailable evidence, and residual risks. Freeze one complete candidate only after its local checks are current. +Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run applicable repository-required checks. Preserve exact results, unavailable evidence, and residual risks. + +After applicable checks are current, invoke `$measure-code-complexity` with the fixed candidate, comparison base, and implementation authority. Resolve its returned dispositions by applying authorised coherent improvements through the applicable teaching skills, retaining supported justifications, or invoking `$elicit-with-context` when resolution would change agreed behaviour, contracts, architecture, or scope. Rerun every affected harness and the measurement after a change. Freeze one complete candidate only when its checks and complexity measurement are current and every trigger is resolved or explicitly retained for handoff. ### 4. Review and adjudicate diff --git a/skills/measure-code-complexity/SKILL.md b/skills/measure-code-complexity/SKILL.md new file mode 100644 index 0000000..448de92 --- /dev/null +++ b/skills/measure-code-complexity/SKILL.md @@ -0,0 +1,52 @@ +--- +name: measure-code-complexity +description: Measures changed-code complexity with pinned cross-language metrics and advisory review triggers. Use when implementing, reviewing, or directly assessing code complexity. +disable-model-invocation: false +--- + +# Measure Code Complexity + +Measures reproducible complexity signals and routes triggered changed code through contextual engineering judgement. + +## Prerequisites + +Fix the Git repository, candidate, comparison base, caller authority, and any explicit repository thresholds. Require Python 3.8 or newer. Report the exact unavailable or incomplete evidence when any prerequisite is absent; never treat a failed or unsupported measurement as clear. + +## Process + +### 1. Measure the fixed change + +Run the bundled [measurement script](scripts/measure.py) without installing dependencies or accessing the network: + +```sh +python3 /path/to/measure-code-complexity/scripts/measure.py \ + --repo /path/to/repository \ + --base +``` + +Pass the corresponding threshold option when an explicit repository policy overrides a bundled default. Use `--whole-repository` only for an explicit repository-wide request and `--all` only when compact evidence cannot resolve a concrete question. + +### 2. Investigate triggered signals + +Read the compact JSON. Its summary accounts for every measured changed function; detailed records default to triggered entities and the raw artifact retains the complete normalized result. + +For each triggered metric, load only its named reference: [Cyclomatic Complexity](references/CYCLOMATIC-COMPLEXITY.md), [Function Size](references/FUNCTION-SIZE.md), [Nesting](references/NESTING.md), [Parameters](references/PARAMETERS.md), or [Duplication](references/DUPLICATION.md). Inspect the cited source, nearby contracts, tests, and clone counterparts. Do not load a metric reference that did not trigger. + +Disposition every trigger as one of: + +- **Improved:** apply an authorised, coherent simplification; rerun affected tests and this measurement. +- **Justified retention:** retain the shape with a concrete code- and test-based reason. +- **Unresolved:** state the missing authority or evidence and leave the trigger open. + +Respect the caller's boundary. An implementation caller may edit and rerun; a review caller remains read-only and returns an evidenced suggestion when the consequence warrants one. + +### 3. Return the evidence + +Return the fixed scope, compact summary, triggered locations and values, baseline deltas when present, each disposition, incomplete evidence, and raw-artifact path. A high value triggers inspection; it does not establish a defect, require refactoring, or supply a code-quality verdict. + +## Rules + +- Keep the bundled metric definitions and bands intact unless explicit repository policy overrides a threshold. +- Preserve all metric components; do not combine them into Maintainability Index, CRAP, or another scalar grade. +- Reject score-only changes that merely move branching, add indirection, hide parameters, or replace one clone with a shallow abstraction. +- Stop after every trigger has a supported disposition or is explicitly unresolved. Do not create repository policy, durable reports, commits, pull requests, or external state unless the caller separately authorises them. diff --git a/skills/measure-code-complexity/agents/openai.yaml b/skills/measure-code-complexity/agents/openai.yaml new file mode 100644 index 0000000..46dd2f9 --- /dev/null +++ b/skills/measure-code-complexity/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Measure Code Complexity' + short_description: 'Measure changed-code complexity and review triggers' +policy: + allow_implicit_invocation: true diff --git a/skills/measure-code-complexity/references/CYCLOMATIC-COMPLEXITY.md b/skills/measure-code-complexity/references/CYCLOMATIC-COMPLEXITY.md new file mode 100644 index 0000000..fba7a8a --- /dev/null +++ b/skills/measure-code-complexity/references/CYCLOMATIC-COMPLEXITY.md @@ -0,0 +1,9 @@ +# Cyclomatic Complexity + +Cyclomatic complexity counts recognized decision points within each function as a control-flow-path proxy. Use the bundled bands as attention routing: `1-6` low, `7-9` moderate, `10-20` high, and `21+` very high. Values differ across measurement engines and languages; this does not measure all executable paths, nesting, domain difficulty, runtime behaviour, or correctness. + +Inspect whether the branches represent one cohesive decision, repeated cases, invalid intermediate states, flag-driven behaviour, or several policies sharing a function. Trace the distinct outcomes and their tests. + +Prefer guard clauses for invalid states, data or lookup tables for declarative cases, and extraction behind a named boundary when a decision has an independent contract. Use polymorphism or strategy objects only when the variants are real concepts with separate change patterns. Do not improve the number by scattering the same decisions across shallow helpers. + +Retain high complexity when an explicit algorithm, parser, state transition, protocol, or irreducible domain rule is clearer locally and its material paths are tested. Record that mechanism and evidence. diff --git a/skills/measure-code-complexity/references/DUPLICATION.md b/skills/measure-code-complexity/references/DUPLICATION.md new file mode 100644 index 0000000..511f65a --- /dev/null +++ b/skills/measure-code-complexity/references/DUPLICATION.md @@ -0,0 +1,9 @@ +# Duplication + +Clone detection normalizes token sequences and reports locations meeting the bundled `70`-token floor. Review new or expanded clone groups that intersect changed code. A clone is syntactic similarity, not proof that occurrences share one concept or should evolve together. + +Compare every occurrence. Inspect whether the same rule, defect fix, invariant, or sequence would need parallel change; also check whether similarities are coincidental, generated, declarative, protocol-defined, or intentionally independent test examples. + +Prefer one owner with a deep interface when the occurrences share knowledge and change together. Use data tables or parameterized tests when only values differ. Do not create a shallow helper that leaks sequencing, accepts mode flags, couples independent features, or obscures a readable example. + +Retain a clone when independent evolution, protocol fidelity, test clarity, or a small stable repetition outweighs shared ownership. Record that reason and the evidence that future changes need not stay synchronized. diff --git a/skills/measure-code-complexity/references/FUNCTION-SIZE.md b/skills/measure-code-complexity/references/FUNCTION-SIZE.md new file mode 100644 index 0000000..bcbefda --- /dev/null +++ b/skills/measure-code-complexity/references/FUNCTION-SIZE.md @@ -0,0 +1,9 @@ +# Function Size + +NLOC counts non-comment source lines within a function; token count is supporting scale evidence. Review NLOC above `100`. Size measures change surface, not value, productivity, correctness, or understandability, and formatting or language syntax can change it. + +Inspect whether the function mixes validation, policy, transformation, persistence, external effects, or several temporal phases. Look for a coherent block with its own inputs, result, invariant, and useful name rather than extracting by line count. + +Prefer a smaller deep helper, a table for repetitive data, or an owned module when that improves locality and hides a real decision. Keep related sequencing together when extraction would force callers to coordinate it or create pass-through methods. + +Retain long linear functions when they are declarative mappings, protocol mirrors, migrations, generated structures, or one readable algorithm whose split would obscure the flow. Verify the retained shape through focused tests and the applicable repository checks. diff --git a/skills/measure-code-complexity/references/NESTING.md b/skills/measure-code-complexity/references/NESTING.md new file mode 100644 index 0000000..b9d29ec --- /dev/null +++ b/skills/measure-code-complexity/references/NESTING.md @@ -0,0 +1,9 @@ +# Nested Control Structures + +The nested-structures metric reports the maximum number of nested control structures in a function. Review values above `3`. It is a syntactic proxy for context carried through nested branches; it is not Cognitive Complexity and may vary with language support. + +Inspect which conditions protect invalid states, which levels represent independent policy, and whether loop, exception, and conditional nesting obscure the active invariant. Trace the deepest path and its tests. + +Prefer guard clauses, `continue` or early completion for invalid cases, named predicates, and extraction of a cohesive nested policy. Flatten only when the resulting order and effects remain explicit. Avoid boolean flags, callback chains, or hidden mutable state that reduce indentation while increasing coordination. + +Retain nesting when tree traversal, parsing, structured resource handling, or transactional cleanup is clearest in its lexical hierarchy. Record why flattening would weaken correctness or locality and identify the tests protecting the deepest path. diff --git a/skills/measure-code-complexity/references/PARAMETERS.md b/skills/measure-code-complexity/references/PARAMETERS.md new file mode 100644 index 0000000..0936b80 --- /dev/null +++ b/skills/measure-code-complexity/references/PARAMETERS.md @@ -0,0 +1,9 @@ +# Parameter Count + +The measurement reports a function parameter count. Verify it against the signature before dispositioning it because partial language parsers can miscount parameters. Review counts of `5` or more. The count can expose a wide interface or recurring data group, but it does not measure parameter meaning, type safety, defaults, or the depth of the function's abstraction. + +Inspect whether parameters travel together, represent one concept, encode flags or modes, or change for unrelated reasons. Check call sites before changing the interface. + +Prefer an existing domain value, an explicit request or options type with a coherent invariant, or separate operations when callers currently select unrelated behaviour. Do not hide unrelated values in a generic bag or move required context into globals, ambient state, or service locators. + +Retain a wider signature for mathematical functions, stable interoperation boundaries, framework callbacks, or flat data construction when grouping would add ceremony without ownership. Record the caller evidence and keep meaningful parameter names and types. diff --git a/skills/measure-code-complexity/scripts/measure.py b/skills/measure-code-complexity/scripts/measure.py new file mode 100644 index 0000000..a537d3c --- /dev/null +++ b/skills/measure-code-complexity/scripts/measure.py @@ -0,0 +1,740 @@ +#!/usr/bin/env python3 +"""Measure changed-code complexity with a pinned, vendored Lizard runtime.""" + +from __future__ import annotations + +import argparse +import hashlib +import itertools +import json +import os +from pathlib import Path +import re +import subprocess +import sys +import tempfile +from typing import Any, Iterable + + +SCHEMA_VERSION = 1 +SCRIPT_ROOT = Path(__file__).resolve().parent +VENDOR_ROOT = SCRIPT_ROOT / "vendor" +WHEELS = ( + ( + "pygments-2.19.2-py3-none-any.whl", + "86540386c03d588bb81d44bc3928634ff26449851e99741617ecb9037ee5ec0b", + ), + ( + "lizard-1.24.0-py2.py3-none-any.whl", + "a688bc607a891ff4a7836826f25742dc9c1bf648da3075dbd495e199e8848602", + ), +) + +DEFAULTS = { + "cyclomaticReview": 10, + "nlocReview": 100, + "nestedStructuresReview": 3, + "parameterCountReview": 5, + "duplicateTokens": 70, +} + +REFERENCES = { + "cyclomaticComplexity": "references/CYCLOMATIC-COMPLEXITY.md", + "nloc": "references/FUNCTION-SIZE.md", + "nestedStructures": "references/NESTING.md", + "parameterCount": "references/PARAMETERS.md", + "duplication": "references/DUPLICATION.md", +} + +EXCLUDED_SEGMENTS = { + ".git", + ".hg", + ".svn", + "build", + "coverage", + "deps", + "dist", + "fixtures", + "generated", + "node_modules", + "snapshots", + "vendor", + "vendors", + "__fixtures__", + "__snapshots__", +} + +TEST_SEGMENTS = {"spec", "specs", "test", "tests", "__tests__"} +NON_SOURCE_SUFFIXES = { + ".css", + ".csv", + ".gif", + ".html", + ".jpeg", + ".jpg", + ".json", + ".lock", + ".md", + ".pdf", + ".png", + ".svg", + ".toml", + ".txt", + ".xml", + ".yaml", + ".yml", +} + + +class MeasurementError(RuntimeError): + """A deterministic measurement prerequisite or execution failed.""" + + +def sha256(path: Path) -> str: + digest = hashlib.sha256() + with path.open("rb") as stream: + for block in iter(lambda: stream.read(1024 * 1024), b""): + digest.update(block) + return digest.hexdigest() + + +def load_analyzer() -> tuple[Any, Any, Any, str]: + if sys.version_info < (3, 8): + raise MeasurementError("Python 3.8 or newer is required.") + + for filename, expected in WHEELS: + wheel = VENDOR_ROOT / filename + if not wheel.is_file(): + raise MeasurementError(f"Missing vendored analyzer wheel: {filename}") + actual = sha256(wheel) + if actual != expected: + raise MeasurementError( + f"Checksum mismatch for {filename}: expected {expected}, got {actual}" + ) + sys.path.insert(0, str(wheel)) + + try: + import lizard # type: ignore + from lizard_ext import lizardduplicate, lizardns, version # type: ignore + except (ImportError, SystemExit) as error: + raise MeasurementError(f"Unable to load vendored Lizard: {error}") from error + + return lizard, lizardns, lizardduplicate, str(version) + + +def git(repo: Path, *arguments: str, check: bool = True) -> bytes: + command = ["git", "-C", str(repo), *arguments] + result = subprocess.run(command, capture_output=True, check=False) + if check and result.returncode != 0: + message = result.stderr.decode("utf-8", "replace").strip() + raise MeasurementError( + f"Command failed ({result.returncode}): {' '.join(command)}: {message}" + ) + return result.stdout + + +def nul_fields(data: bytes) -> list[str]: + return [field.decode("utf-8", "surrogateescape") for field in data.split(b"\0") if field] + + +def normalize_path(value: str) -> str: + return Path(value).as_posix().removeprefix("./") + + +def excluded(path: str) -> bool: + normalized = normalize_path(path) + parts = set(Path(normalized).parts) + name = Path(normalized).name.lower() + return ( + bool(parts & EXCLUDED_SEGMENTS) + or bool(parts & TEST_SEGMENTS) + or name.endswith((".min.js", ".min.css", ".generated.ts", ".generated.js")) + or ".fixture." in name + or ".snap." in name + or name.endswith((".spec.ts", ".spec.tsx", ".test.ts", ".test.tsx")) + ) + + +def parse_name_status(data: bytes) -> dict[str, str | None]: + fields = nul_fields(data) + changed: dict[str, str | None] = {} + index = 0 + while index < len(fields): + status = fields[index] + index += 1 + if status.startswith(("R", "C")): + if index + 1 >= len(fields): + raise MeasurementError("Malformed git rename/copy status output.") + old_path = normalize_path(fields[index]) + new_path = normalize_path(fields[index + 1]) + changed[new_path] = old_path + index += 2 + else: + if index >= len(fields): + raise MeasurementError("Malformed git name-status output.") + path = normalize_path(fields[index]) + index += 1 + if not status.startswith("D"): + changed[path] = path + return changed + + +def all_changed_paths(repo: Path, base: str) -> dict[str, str | None]: + changed = parse_name_status( + git(repo, "diff", "--name-status", "-z", "--find-renames", base, "--") + ) + for path in nul_fields(git(repo, "ls-files", "--others", "--exclude-standard", "-z")): + changed[normalize_path(path)] = None + return {path: old_path for path, old_path in changed.items() if (repo / path).is_file()} + + +def changed_paths(repo: Path, base: str) -> dict[str, str | None]: + return { + path: old_path + for path, old_path in all_changed_paths(repo, base).items() + if not excluded(path) + } + + +def line_ranges(repo: Path, base: str, path: str, is_untracked: bool) -> list[tuple[int, int]]: + if is_untracked: + line_count = sum(1 for _ in (repo / path).open("rb")) + return [(1, max(1, line_count))] + + patch = git(repo, "diff", "--unified=0", "--no-color", base, "--", path).decode( + "utf-8", "replace" + ) + ranges: list[tuple[int, int]] = [] + for match in re.finditer(r"^@@ -\d+(?:,\d+)? \+(\d+)(?:,(\d+))? @@", patch, re.MULTILINE): + start = int(match.group(1)) + count = int(match.group(2) or "1") + if count: + ranges.append((start, start + count - 1)) + return ranges + + +def intersects(start: int, end: int, ranges: Iterable[tuple[int, int]]) -> bool: + return any(start <= range_end and end >= range_start for range_start, range_end in ranges) + + +def current_paths(repo: Path) -> list[str]: + paths = nul_fields(git(repo, "ls-files", "-co", "--exclude-standard", "-z")) + return sorted( + { + normalize_path(path) + for path in paths + if (repo / path).is_file() and not excluded(path) + } + ) + + +def base_paths(repo: Path, base: str) -> list[str]: + return sorted( + path + for path in map( + normalize_path, + nul_fields(git(repo, "ls-tree", "-r", "--name-only", "-z", base)), + ) + if not excluded(path) + ) + + +def source_paths(paths: Iterable[str], root: Path, lizard: Any) -> tuple[list[str], list[str]]: + supported: list[str] = [] + unsupported: list[str] = [] + for path in paths: + if lizard.get_reader_for(str(root / path)): + supported.append(path) + elif Path(path).suffix.lower() not in NON_SOURCE_SUFFIXES: + unsupported.append(path) + return supported, unsupported + + +def extract_base(repo: Path, base: str, paths: Iterable[str], target: Path) -> list[str]: + extracted: list[str] = [] + for path in paths: + result = subprocess.run( + ["git", "-C", str(repo), "show", f"{base}:{path}"], + capture_output=True, + check=False, + ) + if result.returncode != 0: + continue + destination = target / path + destination.parent.mkdir(parents=True, exist_ok=True) + destination.write_bytes(result.stdout) + extracted.append(path) + return extracted + + +def analyze( + root: Path, + paths: list[str], + lizard: Any, + lizardns: Any, + duplicate_module: Any, + duplicate_tokens: int, +) -> tuple[list[dict[str, Any]], list[list[dict[str, Any]]], float]: + nested_extension = lizardns.LizardExtension() + duplicate_extension = duplicate_module.LizardExtension() + extensions = lizard.get_extensions([nested_extension, duplicate_extension]) + absolute_paths = [str(root / path) for path in paths] + file_infos = list(lizard.analyze_files(absolute_paths, threads=1, exts=extensions)) + + functions: list[dict[str, Any]] = [] + for file_info in file_infos: + relative = normalize_path(os.path.relpath(file_info.filename, root)) + for function in file_info.function_list: + functions.append( + { + "file": relative, + "line": int(function.start_line), + "endLine": int(function.end_line), + "symbol": str(function.name), + "longName": str(function.long_name), + "metrics": { + "cyclomaticComplexity": int(function.cyclomatic_complexity), + "nloc": int(function.nloc), + "nestedStructures": int(function.max_nested_structures), + "parameterCount": int(function.parameter_count), + "tokenCount": int(function.token_count), + }, + } + ) + + duplicates: list[list[dict[str, Any]]] = [] + for group in duplicate_extension.get_duplicates(duplicate_tokens): + duplicates.append( + [ + { + "file": normalize_path(os.path.relpath(snippet.file_name, root)), + "line": int(snippet.start_line), + "endLine": int(snippet.end_line), + } + for snippet in group + ] + ) + return functions, duplicates, float(duplicate_extension.duplicate_rate() or 0.0) + + +def cyclomatic_band(value: int) -> str: + if value <= 6: + return "low" + if value <= 9: + return "moderate" + if value <= 20: + return "high" + return "veryHigh" + + +def baseline_match(function: dict[str, Any], base_functions: list[dict[str, Any]], old_path: str | None) -> dict[str, Any] | None: + if old_path is None: + return None + same_name = [ + item + for item in base_functions + if item["file"] == old_path and item["symbol"] == function["symbol"] + ] + if len(same_name) == 1: + return same_name[0] + same_signature = [item for item in same_name if item["longName"] == function["longName"]] + if same_signature: + return min(same_signature, key=lambda item: abs(item["line"] - function["line"])) + if same_name: + return min(same_name, key=lambda item: abs(item["line"] - function["line"])) + return None + + +def metric_value( + value: int, + baseline: int | None, + band: str | None = None, + is_new: bool = False, +) -> dict[str, Any]: + result: dict[str, Any] = {"value": value} + if is_new: + result["baseline"] = "absent" + result["delta"] = "new" + elif baseline is not None: + result["baseline"] = baseline + result["delta"] = value - baseline + if band: + result["band"] = band + return result + + +def duplicate_pairs(groups: list[list[dict[str, Any]]]) -> dict[tuple[str, str], int]: + pairs: dict[tuple[str, str], int] = {} + for group in groups: + for left, right in itertools.combinations(group, 2): + key = tuple(sorted((left["file"], right["file"]))) + span = min( + left["endLine"] - left["line"] + 1, + right["endLine"] - right["line"] + 1, + ) + pairs[key] = max(pairs.get(key, 0), span) + return pairs + + +def duplicate_triggers( + candidate_groups: list[list[dict[str, Any]]], + base_groups: list[list[dict[str, Any]]], + changed: dict[str, str | None], + changed_lines: dict[str, list[tuple[int, int]]], + whole_repository: bool, + minimum_tokens: int, +) -> list[dict[str, Any]]: + base_pair_lengths = duplicate_pairs(base_groups) + results: list[dict[str, Any]] = [] + seen: set[tuple[tuple[str, int, int], ...]] = set() + + for group in candidate_groups: + touches_change = whole_repository or any( + snippet["file"] in changed + and intersects( + snippet["line"], + snippet["endLine"], + changed_lines.get(snippet["file"], []), + ) + for snippet in group + ) + if not touches_change: + continue + + state = "present" if whole_repository else None + baseline_lines: int | None = None + if not whole_repository: + for left, right in itertools.combinations(group, 2): + old_left = changed.get(left["file"], left["file"]) + old_right = changed.get(right["file"], right["file"]) + if old_left is None or old_right is None: + state = "new" + continue + base_key = tuple(sorted((old_left, old_right))) + candidate_span = min( + left["endLine"] - left["line"] + 1, + right["endLine"] - right["line"] + 1, + ) + prior_span = base_pair_lengths.get(base_key) + if prior_span is None: + state = "new" + elif candidate_span > prior_span and state != "new": + state = "expanded" + baseline_lines = prior_span + + if state is None: + continue + identity = tuple( + sorted((item["file"], item["line"], item["endLine"]) for item in group) + ) + if identity in seen: + continue + seen.add(identity) + item: dict[str, Any] = { + "state": state, + "minimumTokens": minimum_tokens, + "locations": group, + "reference": REFERENCES["duplication"], + } + if state == "new": + item["baseline"] = "absent" + item["delta"] = "new" + elif state == "expanded": + item["delta"] = "expanded" + if baseline_lines is not None: + item["baselineSpanLines"] = baseline_lines + results.append(item) + return results + + +def write_raw(payload: dict[str, Any]) -> str: + descriptor, path = tempfile.mkstemp(prefix="code-complexity-", suffix=".json") + with os.fdopen(descriptor, "w", encoding="utf-8") as stream: + json.dump(payload, stream, indent=2, sort_keys=True) + stream.write("\n") + return path + + +def parser() -> argparse.ArgumentParser: + result = argparse.ArgumentParser(description=__doc__) + result.add_argument("--repo", default=".", help="Git repository to measure") + result.add_argument("--base", help="Base revision for changed-code measurement") + result.add_argument( + "--whole-repository", + action="store_true", + help="Measure every supported production source file without a baseline", + ) + result.add_argument("--all", action="store_true", help="Include every normalized record") + result.add_argument("--cyclomatic-review", type=int, default=DEFAULTS["cyclomaticReview"]) + result.add_argument("--nloc-review", type=int, default=DEFAULTS["nlocReview"]) + result.add_argument( + "--nested-structures-review", + type=int, + default=DEFAULTS["nestedStructuresReview"], + ) + result.add_argument( + "--parameter-count-review", + type=int, + default=DEFAULTS["parameterCountReview"], + ) + result.add_argument("--duplicate-tokens", type=int, default=DEFAULTS["duplicateTokens"]) + return result + + +def main() -> int: + args = parser().parse_args() + try: + if args.whole_repository and args.base: + raise MeasurementError("--base and --whole-repository cannot be combined.") + if not args.whole_repository and not args.base: + raise MeasurementError("--base is required unless --whole-repository is set.") + thresholds = { + "cyclomaticReview": args.cyclomatic_review, + "nlocReview": args.nloc_review, + "nestedStructuresReview": args.nested_structures_review, + "parameterCountReview": args.parameter_count_review, + "duplicateTokens": args.duplicate_tokens, + } + if any(value < 1 for value in thresholds.values()): + raise MeasurementError("Every threshold must be a positive integer.") + + repo = Path(args.repo).resolve() + if not repo.is_dir(): + raise MeasurementError(f"Repository directory does not exist: {repo}") + git(repo, "rev-parse", "--show-toplevel") + head = git(repo, "rev-parse", "HEAD").decode().strip() + base = None + if args.base: + base = git(repo, "rev-parse", "--verify", f"{args.base}^{{commit}}").decode().strip() + + lizard, lizardns, duplicate_module, lizard_version = load_analyzer() + all_current_paths = current_paths(repo) + supported_current, _ = source_paths(all_current_paths, repo, lizard) + + if args.whole_repository: + changed: dict[str, str | None] = {path: None for path in supported_current} + changed_lines = { + path: [(1, max(1, sum(1 for _ in (repo / path).open("rb"))))] + for path in supported_current + } + unsupported_changed: list[str] = [] + else: + changed = changed_paths(repo, base or args.base) + changed_lines = { + path: line_ranges(repo, base or args.base, path, old_path is None) + for path, old_path in changed.items() + } + supported_changed, unsupported_changed = source_paths(changed, repo, lizard) + changed = {path: changed[path] for path in supported_changed} + changed_lines = {path: changed_lines[path] for path in supported_changed} + + candidate_functions, candidate_duplicates, candidate_duplicate_rate = analyze( + repo, + supported_current, + lizard, + lizardns, + duplicate_module, + thresholds["duplicateTokens"], + ) + + base_functions: list[dict[str, Any]] = [] + base_duplicates: list[list[dict[str, Any]]] = [] + base_duplicate_rate: float | None = None + if base: + with tempfile.TemporaryDirectory(prefix="code-complexity-base-") as directory: + base_root = Path(directory) + possible_base_paths = base_paths(repo, base) + supported_base, _ = source_paths(possible_base_paths, base_root, lizard) + extracted = extract_base(repo, base, supported_base, base_root) + base_functions, base_duplicates, base_duplicate_rate = analyze( + base_root, + extracted, + lizard, + lizardns, + duplicate_module, + thresholds["duplicateTokens"], + ) + + changed_functions = [ + function + for function in candidate_functions + if function["file"] in changed + and ( + args.whole_repository + or intersects( + function["line"], + function["endLine"], + changed_lines.get(function["file"], []), + ) + ) + ] + + band_counts = {"low": 0, "moderate": 0, "high": 0, "veryHigh": 0} + maxima = { + "cyclomaticComplexity": 0, + "nloc": 0, + "nestedStructures": 0, + "parameterCount": 0, + "tokenCount": 0, + } + function_triggers: list[dict[str, Any]] = [] + + for function in changed_functions: + metrics = function["metrics"] + band = cyclomatic_band(metrics["cyclomaticComplexity"]) + band_counts[band] += 1 + for metric in maxima: + maxima[metric] = max(maxima[metric], metrics[metric]) + + baseline_function = baseline_match( + function, + base_functions, + changed.get(function["file"]), + ) + baseline_metrics = baseline_function["metrics"] if baseline_function else {} + is_new = bool(base and baseline_function is None) + triggered: dict[str, Any] = {} + references: list[str] = [] + + if metrics["cyclomaticComplexity"] >= thresholds["cyclomaticReview"]: + triggered["cyclomaticComplexity"] = metric_value( + metrics["cyclomaticComplexity"], + baseline_metrics.get("cyclomaticComplexity"), + band, + is_new, + ) + references.append(REFERENCES["cyclomaticComplexity"]) + if metrics["nloc"] > thresholds["nlocReview"]: + triggered["nloc"] = metric_value( + metrics["nloc"], baseline_metrics.get("nloc"), is_new=is_new + ) + references.append(REFERENCES["nloc"]) + if metrics["nestedStructures"] > thresholds["nestedStructuresReview"]: + triggered["nestedStructures"] = metric_value( + metrics["nestedStructures"], + baseline_metrics.get("nestedStructures"), + is_new=is_new, + ) + references.append(REFERENCES["nestedStructures"]) + if metrics["parameterCount"] >= thresholds["parameterCountReview"]: + triggered["parameterCount"] = metric_value( + metrics["parameterCount"], + baseline_metrics.get("parameterCount"), + is_new=is_new, + ) + references.append(REFERENCES["parameterCount"]) + + if triggered: + function_triggers.append( + { + "file": function["file"], + "line": function["line"], + "endLine": function["endLine"], + "symbol": function["symbol"], + "metrics": triggered, + "tokenCount": metrics["tokenCount"], + "references": sorted(set(references)), + } + ) + + duplication_triggers = duplicate_triggers( + candidate_duplicates, + base_duplicates, + changed, + changed_lines, + args.whole_repository, + thresholds["duplicateTokens"], + ) + + scope = { + "base": base, + "candidate": f"working-tree@{head}", + "changedProductionFiles": len(changed), + "sourceFilesAnalyzed": len(supported_current), + "changedFunctionsAnalyzed": len(changed_functions), + "excludedChangedPaths": sorted( + path + for path in (all_changed_paths(repo, base or args.base) if base else {}) + if excluded(path) + ), + } + summary = { + "cyclomaticBands": band_counts, + "maxima": maxima, + "observedCloneGroups": len(candidate_duplicates), + "duplicateRate": candidate_duplicate_rate, + "triggeredFunctions": len(function_triggers), + "triggeredCloneGroups": len(duplication_triggers), + } + limitations = [] + if unsupported_changed: + limitations.append( + { + "kind": "unsupported-language", + "paths": sorted(unsupported_changed), + } + ) + + tool = { + "name": "lizard", + "version": lizard_version, + "cloneDetector": { + "name": "lizard-duplicate", + "version": lizard_version, + }, + } + raw_payload = { + "schema": SCHEMA_VERSION, + "tool": tool, + "scope": scope, + "thresholds": thresholds, + "functions": candidate_functions, + "duplicates": candidate_duplicates, + "duplicateRate": candidate_duplicate_rate, + "baseFunctions": base_functions, + "baseDuplicates": base_duplicates, + "baseDuplicateRate": base_duplicate_rate, + "limitations": limitations, + } + raw_artifact = write_raw(raw_payload) + attention = bool(function_triggers or duplication_triggers) + status = "incomplete" if limitations else "attention" if attention else "clear" + output: dict[str, Any] = { + "schema": SCHEMA_VERSION, + "tool": tool, + "status": status, + "scope": scope, + "thresholds": thresholds, + "summary": summary, + "triggers": { + "functions": function_triggers, + "duplicates": duplication_triggers, + }, + "rawArtifact": raw_artifact, + } + if limitations: + output["limitations"] = limitations + if args.all: + output["records"] = { + "functions": candidate_functions, + "duplicates": candidate_duplicates, + } + print(json.dumps(output, indent=2, sort_keys=True)) + return 0 + except MeasurementError as error: + print( + json.dumps( + { + "schema": SCHEMA_VERSION, + "status": "incomplete", + "error": str(error), + }, + indent=2, + sort_keys=True, + ) + ) + return 2 + + +if __name__ == "__main__": + raise SystemExit(main()) diff --git a/skills/measure-code-complexity/scripts/vendor/THIRD-PARTY-NOTICES.md b/skills/measure-code-complexity/scripts/vendor/THIRD-PARTY-NOTICES.md new file mode 100644 index 0000000..f3c85cd --- /dev/null +++ b/skills/measure-code-complexity/scripts/vendor/THIRD-PARTY-NOTICES.md @@ -0,0 +1,8 @@ +# Third-Party Notices + +The measurement script loads these unmodified Python wheels directly from this directory: + +- Lizard 1.24.0, checksum `a688bc607a891ff4a7836826f25742dc9c1bf648da3075dbd495e199e8848602`. Retain its [license](licenses/lizard/LICENSE.txt) and the [Apache License 2.0](licenses/lizard/APACHE-2.0.txt) referenced by licensed source headers. +- Pygments 2.19.2, checksum `86540386c03d588bb81d44bc3928634ff26449851e99741617ecb9037ee5ec0b`. Retain its [license](licenses/pygments/LICENSE) and [authors](licenses/pygments/AUTHORS). + +The wheels remain subject to their included upstream notices and disclaimers. diff --git a/skills/measure-code-complexity/scripts/vendor/licenses/lizard/APACHE-2.0.txt b/skills/measure-code-complexity/scripts/vendor/licenses/lizard/APACHE-2.0.txt new file mode 100644 index 0000000..d645695 --- /dev/null +++ b/skills/measure-code-complexity/scripts/vendor/licenses/lizard/APACHE-2.0.txt @@ -0,0 +1,202 @@ + + Apache License + Version 2.0, January 2004 + http://www.apache.org/licenses/ + + TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION + + 1. Definitions. + + "License" shall mean the terms and conditions for use, reproduction, + and distribution as defined by Sections 1 through 9 of this document. + + "Licensor" shall mean the copyright owner or entity authorized by + the copyright owner that is granting the License. + + "Legal Entity" shall mean the union of the acting entity and all + other entities that control, are controlled by, or are under common + control with that entity. For the purposes of this definition, + "control" means (i) the power, direct or indirect, to cause the + direction or management of such entity, whether by contract or + otherwise, or (ii) ownership of fifty percent (50%) or more of the + outstanding shares, or (iii) beneficial ownership of such entity. + + "You" (or "Your") shall mean an individual or Legal Entity + exercising permissions granted by this License. + + "Source" form shall mean the preferred form for making modifications, + including but not limited to software source code, documentation + source, and configuration files. + + "Object" form shall mean any form resulting from mechanical + transformation or translation of a Source form, including but + not limited to compiled object code, generated documentation, + and conversions to other media types. + + "Work" shall mean the work of authorship, whether in Source or + Object form, made available under the License, as indicated by a + copyright notice that is included in or attached to the work + (an example is provided in the Appendix below). + + "Derivative Works" shall mean any work, whether in Source or Object + form, that is based on (or derived from) the Work and for which the + editorial revisions, annotations, elaborations, or other modifications + represent, as a whole, an original work of authorship. For the purposes + of this License, Derivative Works shall not include works that remain + separable from, or merely link (or bind by name) to the interfaces of, + the Work and Derivative Works thereof. + + "Contribution" shall mean any work of authorship, including + the original version of the Work and any modifications or additions + to that Work or Derivative Works thereof, that is intentionally + submitted to Licensor for inclusion in the Work by the copyright owner + or by an individual or Legal Entity authorized to submit on behalf of + the copyright owner. For the purposes of this definition, "submitted" + means any form of electronic, verbal, or written communication sent + to the Licensor or its representatives, including but not limited to + communication on electronic mailing lists, source code control systems, + and issue tracking systems that are managed by, or on behalf of, the + Licensor for the purpose of discussing and improving the Work, but + excluding communication that is conspicuously marked or otherwise + designated in writing by the copyright owner as "Not a Contribution." + + "Contributor" shall mean Licensor and any individual or Legal Entity + on behalf of whom a Contribution has been received by Licensor and + subsequently incorporated within the Work. + + 2. Grant of Copyright License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + copyright license to reproduce, prepare Derivative Works of, + publicly display, publicly perform, sublicense, and distribute the + Work and such Derivative Works in Source or Object form. + + 3. Grant of Patent License. Subject to the terms and conditions of + this License, each Contributor hereby grants to You a perpetual, + worldwide, non-exclusive, no-charge, royalty-free, irrevocable + (except as stated in this section) patent license to make, have made, + use, offer to sell, sell, import, and otherwise transfer the Work, + where such license applies only to those patent claims licensable + by such Contributor that are necessarily infringed by their + Contribution(s) alone or by combination of their Contribution(s) + with the Work to which such Contribution(s) was submitted. If You + institute patent litigation against any entity (including a + cross-claim or counterclaim in a lawsuit) alleging that the Work + or a Contribution incorporated within the Work constitutes direct + or contributory patent infringement, then any patent licenses + granted to You under this License for that Work shall terminate + as of the date such litigation is filed. + + 4. Redistribution. You may reproduce and distribute copies of the + Work or Derivative Works thereof in any medium, with or without + modifications, and in Source or Object form, provided that You + meet the following conditions: + + (a) You must give any other recipients of the Work or + Derivative Works a copy of this License; and + + (b) You must cause any modified files to carry prominent notices + stating that You changed the files; and + + (c) You must retain, in the Source form of any Derivative Works + that You distribute, all copyright, patent, trademark, and + attribution notices from the Source form of the Work, + excluding those notices that do not pertain to any part of + the Derivative Works; and + + (d) If the Work includes a "NOTICE" text file as part of its + distribution, then any Derivative Works that You distribute must + include a readable copy of the attribution notices contained + within such NOTICE file, excluding those notices that do not + pertain to any part of the Derivative Works, in at least one + of the following places: within a NOTICE text file distributed + as part of the Derivative Works; within the Source form or + documentation, if provided along with the Derivative Works; or, + within a display generated by the Derivative Works, if and + wherever such third-party notices normally appear. The contents + of the NOTICE file are for informational purposes only and + do not modify the License. You may add Your own attribution + notices within Derivative Works that You distribute, alongside + or as an addendum to the NOTICE text from the Work, provided + that such additional attribution notices cannot be construed + as modifying the License. + + You may add Your own copyright statement to Your modifications and + may provide additional or different license terms and conditions + for use, reproduction, or distribution of Your modifications, or + for any such Derivative Works as a whole, provided Your use, + reproduction, and distribution of the Work otherwise complies with + the conditions stated in this License. + + 5. Submission of Contributions. Unless You explicitly state otherwise, + any Contribution intentionally submitted for inclusion in the Work + by You to the Licensor shall be under the terms and conditions of + this License, without any additional terms or conditions. + Notwithstanding the above, nothing herein shall supersede or modify + the terms of any separate license agreement you may have executed + with Licensor regarding such Contributions. + + 6. Trademarks. This License does not grant permission to use the trade + names, trademarks, service marks, or product names of the Licensor, + except as required for reasonable and customary use in describing the + origin of the Work and reproducing the content of the NOTICE file. + + 7. Disclaimer of Warranty. Unless required by applicable law or + agreed to in writing, Licensor provides the Work (and each + Contributor provides its Contributions) on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or + implied, including, without limitation, any warranties or conditions + of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A + PARTICULAR PURPOSE. You are solely responsible for determining the + appropriateness of using or redistributing the Work and assume any + risks associated with Your exercise of permissions under this License. + + 8. Limitation of Liability. In no event and under no legal theory, + whether in tort (including negligence), contract, or otherwise, + unless required by applicable law (such as deliberate and grossly + negligent acts) or agreed to in writing, shall any Contributor be + liable to You for damages, including any direct, indirect, special, + incidental, or consequential damages of any character arising as a + result of this License or out of the use or inability to use the + Work (including but not limited to damages for loss of goodwill, + work stoppage, computer failure or malfunction, or any and all + other commercial damages or losses), even if such Contributor + has been advised of the possibility of such damages. + + 9. Accepting Warranty or Additional Liability. While redistributing + the Work or Derivative Works thereof, You may choose to offer, + and charge a fee for, acceptance of support, warranty, indemnity, + or other liability obligations and/or rights consistent with this + License. However, in accepting such obligations, You may act only + on Your own behalf and on Your sole responsibility, not on behalf + of any other Contributor, and only if You agree to indemnify, + defend, and hold each Contributor harmless for any liability + incurred by, or claims asserted against, such Contributor by reason + of your accepting any such warranty or additional liability. + + END OF TERMS AND CONDITIONS + + APPENDIX: How to apply the Apache License to your work. + + To apply the Apache License to your work, attach the following + boilerplate notice, with the fields enclosed by brackets "[]" + replaced with your own identifying information. (Don't include + the brackets!) The text should be enclosed in the appropriate + comment syntax for the file format. We also recommend that a + file or class name and description of purpose be included on the + same "printed page" as the copyright notice for easier + identification within third-party archives. + + Copyright [yyyy] [name of copyright owner] + + Licensed under the Apache License, Version 2.0 (the "License"); + you may not use this file except in compliance with the License. + You may obtain a copy of the License at + + http://www.apache.org/licenses/LICENSE-2.0 + + Unless required by applicable law or agreed to in writing, software + distributed under the License is distributed on an "AS IS" BASIS, + WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + See the License for the specific language governing permissions and + limitations under the License. diff --git a/skills/measure-code-complexity/scripts/vendor/licenses/lizard/LICENSE.txt b/skills/measure-code-complexity/scripts/vendor/licenses/lizard/LICENSE.txt new file mode 100644 index 0000000..74a3411 --- /dev/null +++ b/skills/measure-code-complexity/scripts/vendor/licenses/lizard/LICENSE.txt @@ -0,0 +1,30 @@ +Copyright Terry Yin and other contributors + +This software consists of voluntary contributions made by many +individuals. For exact contribution history, see the revision history. + +The following license applies to all parts of this software except as +documented below: + +==== + +Permission is hereby granted, free of charge, to any person obtaining +a copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sublicense, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice shall be +included in all copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, +EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND +NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE +LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION +OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION +WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +==== diff --git a/skills/measure-code-complexity/scripts/vendor/licenses/pygments/AUTHORS b/skills/measure-code-complexity/scripts/vendor/licenses/pygments/AUTHORS new file mode 100644 index 0000000..811c66a --- /dev/null +++ b/skills/measure-code-complexity/scripts/vendor/licenses/pygments/AUTHORS @@ -0,0 +1,291 @@ +Pygments is written and maintained by Georg Brandl . + +Major developers are Tim Hatch and Armin Ronacher +. + +Other contributors, listed alphabetically, are: + +* Sam Aaron -- Ioke lexer +* Jean Abou Samra -- LilyPond lexer +* João Abecasis -- JSLT lexer +* Ali Afshar -- image formatter +* Thomas Aglassinger -- Easytrieve, JCL, Rexx, Transact-SQL and VBScript + lexers +* Maxence Ahlouche -- PostgreSQL Explain lexer +* Muthiah Annamalai -- Ezhil lexer +* Nikolay Antipov -- OpenSCAD lexer +* Kumar Appaiah -- Debian control lexer +* Andreas Amann -- AppleScript lexer +* Timothy Armstrong -- Dart lexer fixes +* Jeffrey Arnold -- R/S, Rd, BUGS, Jags, and Stan lexers +* Eiríkr Åsheim -- Uxntal lexer +* Jeremy Ashkenas -- CoffeeScript lexer +* José Joaquín Atria -- Praat lexer +* Stefan Matthias Aust -- Smalltalk lexer +* Lucas Bajolet -- Nit lexer +* Ben Bangert -- Mako lexers +* Max Battcher -- Darcs patch lexer +* Thomas Baruchel -- APL lexer +* Tim Baumann -- (Literate) Agda lexer +* Paul Baumgart, 280 North, Inc. -- Objective-J lexer +* Michael Bayer -- Myghty lexers +* Thomas Beale -- Archetype lexers +* John Benediktsson -- Factor lexer +* David Benjamin, Google LLC -- TLS lexer +* Trevor Bergeron -- mIRC formatter +* Vincent Bernat -- LessCSS lexer +* Christopher Bertels -- Fancy lexer +* Sébastien Bigaret -- QVT Operational lexer +* Jarrett Billingsley -- MiniD lexer +* Adam Blinkinsop -- Haskell, Redcode lexers +* Stéphane Blondon -- Procfile, SGF and Sieve lexers +* Frits van Bommel -- assembler lexers +* Pierre Bourdon -- bugfixes +* Martijn Braam -- Kernel log lexer, BARE lexer +* JD Browne, Google LLC -- GoogleSQL lexer +* Matthias Bussonnier -- ANSI style handling for terminal-256 formatter +* chebee7i -- Python traceback lexer improvements +* Hiram Chirino -- Scaml and Jade lexers +* Mauricio Caceres -- SAS and Stata lexers. +* Michael Camilleri, John Gabriele, sogaiu -- Janet lexer +* Daren Chandisingh -- Gleam lexer +* Ian Cooper -- VGL lexer +* David Corbett -- Inform, Jasmin, JSGF, Snowball, and TADS 3 lexers +* Leaf Corcoran -- MoonScript lexer +* Fraser Cormack -- TableGen lexer +* Gabriel Corona -- ASN.1 lexer +* Christopher Creutzig -- MuPAD lexer +* Daniël W. Crompton -- Pike lexer +* Pete Curry -- bugfixes +* Bryan Davis -- EBNF lexer +* Bruno Deferrari -- Shen lexer +* Walter Dörwald -- UL4 lexer +* Luke Drummond -- Meson lexer +* Giedrius Dubinskas -- HTML formatter improvements +* Owen Durni -- Haxe lexer +* Alexander Dutton, Oxford University Computing Services -- SPARQL lexer +* James Edwards -- Terraform lexer +* Nick Efford -- Python 3 lexer +* Sven Efftinge -- Xtend lexer +* Artem Egorkine -- terminal256 formatter +* Matthew Fernandez -- CAmkES lexer +* Paweł Fertyk -- GDScript lexer, HTML formatter improvements +* Michael Ficarra -- CPSA lexer +* James H. Fisher -- PostScript lexer +* Amanda Fitch, Google LLC -- GoogleSQL lexer +* William S. Fulton -- SWIG lexer +* Carlos Galdino -- Elixir and Elixir Console lexers +* Michael Galloy -- IDL lexer +* Naveen Garg -- Autohotkey lexer +* Simon Garnotel -- FreeFem++ lexer +* Laurent Gautier -- R/S lexer +* Alex Gaynor -- PyPy log lexer +* Richard Gerkin -- Igor Pro lexer +* Alain Gilbert -- TypeScript lexer +* Alex Gilding -- BlitzBasic lexer +* GitHub, Inc -- DASM16, Augeas, TOML, and Slash lexers +* Bertrand Goetzmann -- Groovy lexer +* Krzysiek Goj -- Scala lexer +* Rostyslav Golda -- FloScript lexer +* Andrey Golovizin -- BibTeX lexers +* Matt Good -- Genshi, Cheetah lexers +* Michał Górny -- vim modeline support +* Alex Gosse -- TrafficScript lexer +* Patrick Gotthardt -- PHP namespaces support +* Hubert Gruniaux -- C and C++ lexer improvements +* Olivier Guibe -- Asymptote lexer +* Phil Hagelberg -- Fennel lexer +* Florian Hahn -- Boogie lexer +* Martin Harriman -- SNOBOL lexer +* Matthew Harrison -- SVG formatter +* Steven Hazel -- Tcl lexer +* Dan Michael Heggø -- Turtle lexer +* Aslak Hellesøy -- Gherkin lexer +* Greg Hendershott -- Racket lexer +* Justin Hendrick -- ParaSail lexer +* Jordi Gutiérrez Hermoso -- Octave lexer +* David Hess, Fish Software, Inc. -- Objective-J lexer +* Ken Hilton -- Typographic Number Theory and Arrow lexers +* Varun Hiremath -- Debian control lexer +* Rob Hoelz -- Perl 6 lexer +* Doug Hogan -- Mscgen lexer +* Ben Hollis -- Mason lexer +* Max Horn -- GAP lexer +* Fred Hornsey -- OMG IDL Lexer +* Alastair Houghton -- Lexer inheritance facility +* Tim Howard -- BlitzMax lexer +* Dustin Howett -- Logos lexer +* Ivan Inozemtsev -- Fantom lexer +* Hiroaki Itoh -- Shell console rewrite, Lexers for PowerShell session, + MSDOS session, BC, WDiff +* Brian R. Jackson -- Tea lexer +* Christian Jann -- ShellSession lexer +* Jonas Camillus Jeppesen -- Line numbers and line highlighting for + RTF-formatter +* Dennis Kaarsemaker -- sources.list lexer +* Dmitri Kabak -- Inferno Limbo lexer +* Igor Kalnitsky -- vhdl lexer +* Colin Kennedy - USD lexer +* Alexander Kit -- MaskJS lexer +* Pekka Klärck -- Robot Framework lexer +* Gerwin Klein -- Isabelle lexer +* Eric Knibbe -- Lasso lexer +* Stepan Koltsov -- Clay lexer +* Oliver Kopp - Friendly grayscale style +* Adam Koprowski -- Opa lexer +* Benjamin Kowarsch -- Modula-2 lexer +* Domen Kožar -- Nix lexer +* Oleh Krekel -- Emacs Lisp lexer +* Alexander Kriegisch -- Kconfig and AspectJ lexers +* Marek Kubica -- Scheme lexer +* Jochen Kupperschmidt -- Markdown processor +* Gerd Kurzbach -- Modelica lexer +* Jon Larimer, Google Inc. -- Smali lexer +* Olov Lassus -- Dart lexer +* Matt Layman -- TAP lexer +* Dan Lazin, Google LLC -- GoogleSQL lexer +* Kristian Lyngstøl -- Varnish lexers +* Sylvestre Ledru -- Scilab lexer +* Chee Sing Lee -- Flatline lexer +* Mark Lee -- Vala lexer +* Thomas Linder Puls -- Visual Prolog lexer +* Pete Lomax -- Phix lexer +* Valentin Lorentz -- C++ lexer improvements +* Ben Mabey -- Gherkin lexer +* Angus MacArthur -- QML lexer +* Louis Mandel -- X10 lexer +* Louis Marchand -- Eiffel lexer +* Simone Margaritelli -- Hybris lexer +* Tim Martin - World of Warcraft TOC lexer +* Kirk McDonald -- D lexer +* Gordon McGregor -- SystemVerilog lexer +* Stephen McKamey -- Duel/JBST lexer +* Brian McKenna -- F# lexer +* Charles McLaughlin -- Puppet lexer +* Kurt McKee -- Tera Term macro lexer, PostgreSQL updates, MySQL overhaul, JSON lexer +* Joe Eli McIlvain -- Savi lexer +* Lukas Meuser -- BBCode formatter, Lua lexer +* Cat Miller -- Pig lexer +* Paul Miller -- LiveScript lexer +* Hong Minhee -- HTTP lexer +* Michael Mior -- Awk lexer +* Bruce Mitchener -- Dylan lexer rewrite +* Reuben Morais -- SourcePawn lexer +* Jon Morton -- Rust lexer +* Paulo Moura -- Logtalk lexer +* Mher Movsisyan -- DTD lexer +* Dejan Muhamedagic -- Crmsh lexer +* Adrien Nayrat -- PostgreSQL Explain lexer +* Ana Nelson -- Ragel, ANTLR, R console lexers +* David Neto, Google LLC -- WebGPU Shading Language lexer +* Kurt Neufeld -- Markdown lexer +* Nam T. Nguyen -- Monokai style +* Jesper Noehr -- HTML formatter "anchorlinenos" +* Mike Nolta -- Julia lexer +* Avery Nortonsmith -- Pointless lexer +* Jonas Obrist -- BBCode lexer +* Edward O'Callaghan -- Cryptol lexer +* David Oliva -- Rebol lexer +* Pat Pannuto -- nesC lexer +* Jon Parise -- Protocol buffers and Thrift lexers +* Benjamin Peterson -- Test suite refactoring +* Ronny Pfannschmidt -- BBCode lexer +* Dominik Picheta -- Nimrod lexer +* Andrew Pinkham -- RTF Formatter Refactoring +* Clément Prévost -- UrbiScript lexer +* Tanner Prynn -- cmdline -x option and loading lexers from files +* Oleh Prypin -- Crystal lexer (based on Ruby lexer) +* Nick Psaris -- K and Q lexers +* Xidorn Quan -- Web IDL lexer +* Elias Rabel -- Fortran fixed form lexer +* raichoo -- Idris lexer +* Daniel Ramirez -- GDScript lexer +* Kashif Rasul -- CUDA lexer +* Nathan Reed -- HLSL lexer +* Justin Reidy -- MXML lexer +* Jonathon Reinhart, Google LLC -- Soong lexer +* Norman Richards -- JSON lexer +* Corey Richardson -- Rust lexer updates +* Fabrizio Riguzzi -- cplint leder +* Lubomir Rintel -- GoodData MAQL and CL lexers +* Andre Roberge -- Tango style +* Georg Rollinger -- HSAIL lexer +* Michiel Roos -- TypoScript lexer +* Konrad Rudolph -- LaTeX formatter enhancements +* Mario Ruggier -- Evoque lexers +* Miikka Salminen -- Lovelace style, Hexdump lexer, lexer enhancements +* Stou Sandalski -- NumPy, FORTRAN, tcsh and XSLT lexers +* Matteo Sasso -- Common Lisp lexer +* Joe Schafer -- Ada lexer +* Max Schillinger -- TiddlyWiki5 lexer +* Andrew Schmidt -- X++ lexer +* Ken Schutte -- Matlab lexers +* René Schwaiger -- Rainbow Dash style +* Sebastian Schweizer -- Whiley lexer +* Tassilo Schweyer -- Io, MOOCode lexers +* Pablo Seminario -- PromQL lexer +* Ted Shaw -- AutoIt lexer +* Joerg Sieker -- ABAP lexer +* Robert Simmons -- Standard ML lexer +* Kirill Simonov -- YAML lexer +* Corbin Simpson -- Monte lexer +* Ville Skyttä -- ASCII armored lexer +* Alexander Smishlajev -- Visual FoxPro lexer +* Steve Spigarelli -- XQuery lexer +* Jerome St-Louis -- eC lexer +* Camil Staps -- Clean and NuSMV lexers; Solarized style +* James Strachan -- Kotlin lexer +* Tom Stuart -- Treetop lexer +* Colin Sullivan -- SuperCollider lexer +* Ben Swift -- Extempore lexer +* tatt61880 -- Kuin lexer +* Edoardo Tenani -- Arduino lexer +* Tiberius Teng -- default style overhaul +* Jeremy Thurgood -- Erlang, Squid config lexers +* Brian Tiffin -- OpenCOBOL lexer +* Bob Tolbert -- Hy lexer +* Doug Torrance -- Macaulay2 lexer +* Matthias Trute -- Forth lexer +* Tuoa Spi T4 -- Bdd lexer +* Erick Tryzelaar -- Felix lexer +* Alexander Udalov -- Kotlin lexer improvements +* Thomas Van Doren -- Chapel lexer +* Dave Van Ee -- Uxntal lexer updates +* Daniele Varrazzo -- PostgreSQL lexers +* Abe Voelker -- OpenEdge ABL lexer +* Pepijn de Vos -- HTML formatter CTags support +* Matthias Vallentin -- Bro lexer +* Benoît Vinot -- AMPL lexer +* Linh Vu Hong -- RSL lexer +* Taavi Väänänen -- Debian control lexer +* Immanuel Washington -- Smithy lexer +* Nathan Weizenbaum -- Haml and Sass lexers +* Nathan Whetsell -- Csound lexers +* Dietmar Winkler -- Modelica lexer +* Nils Winter -- Smalltalk lexer +* Davy Wybiral -- Clojure lexer +* Whitney Young -- ObjectiveC lexer +* Diego Zamboni -- CFengine3 lexer +* Enrique Zamudio -- Ceylon lexer +* Alex Zimin -- Nemerle lexer +* Rob Zimmerman -- Kal lexer +* Evgenii Zheltonozhskii -- Maple lexer +* Vincent Zurczak -- Roboconf lexer +* Hubert Gruniaux -- C and C++ lexer improvements +* Thomas Symalla -- AMDGPU Lexer +* 15b3 -- Image Formatter improvements +* Fabian Neumann -- CDDL lexer +* Thomas Duboucher -- CDDL lexer +* Philipp Imhof -- Pango Markup formatter +* Thomas Voss -- Sed lexer +* Martin Fischer -- WCAG contrast testing +* Marc Auberer -- Spice lexer +* Amr Hesham -- Carbon lexer +* diskdance -- Wikitext lexer +* vanillajonathan -- PRQL lexer +* Nikolay Antipov -- OpenSCAD lexer +* Markus Meyer, Nextron Systems -- YARA lexer +* Hannes Römer -- Mojo lexer +* Jan Frederik Schaefer -- PDDL lexer + +Many thanks for all contributions! diff --git a/skills/measure-code-complexity/scripts/vendor/licenses/pygments/LICENSE b/skills/measure-code-complexity/scripts/vendor/licenses/pygments/LICENSE new file mode 100644 index 0000000..446a1a8 --- /dev/null +++ b/skills/measure-code-complexity/scripts/vendor/licenses/pygments/LICENSE @@ -0,0 +1,25 @@ +Copyright (c) 2006-2022 by the respective authors (see AUTHORS file). +All rights reserved. + +Redistribution and use in source and binary forms, with or without +modification, are permitted provided that the following conditions are +met: + +* Redistributions of source code must retain the above copyright + notice, this list of conditions and the following disclaimer. + +* Redistributions in binary form must reproduce the above copyright + notice, this list of conditions and the following disclaimer in the + documentation and/or other materials provided with the distribution. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/skills/measure-code-complexity/scripts/vendor/lizard-1.24.0-py2.py3-none-any.whl b/skills/measure-code-complexity/scripts/vendor/lizard-1.24.0-py2.py3-none-any.whl new file mode 100644 index 0000000000000000000000000000000000000000..7e8f60c834ede2b24462db74cca52a1b5325afb6 GIT binary patch literal 111418 zcmZ5{Lv$uwv}|mvW81cE+qRQ0>e#kz+qP}nHaqCocXHqQ@8C?%Y}cyVTD28rK*7*} zfPkQYT!2vZ=26qMs%e0L?yG=+p#S~aT6q~co6td0nSQ5u1P=mIU1=zqV}Nb~{6C*N%Mq>YUc2i#s(US9ss zyS3=s)M0zsHPgQ}JE_s+Z(dXDc(OgqKyrB0QfJqqRyWIBDeRdYUSQ4`AowsR{z+Yt z(@!}xYi1CpMH<%_ct?afEYOY+dnBB!M6V(T#q=aO^-PGk_D!YJNk}tRr%M&fO>UCi zKDl9QAt?}rI@%+PzHF4s_W(gpy|=Odg(fw{d34rkmlpQ>x4XN$lJL8@noo;{dE z^5Nlpr>m$5v5ewwWXt{V{br+T39-y!Zhy_j$@TT_a8G_tUrrR_E6}a+(<}xJD3j}v zSti2iu4!7xNS{f1hB@ox$s3|LMWgDN8dYP(PCZR9LDT$&*r8qlN03pBQLJ}-tGGl0 zuanr^s1c;00#Ph$3SiYzWq7q!FeyA^jt3T$t z7d)Pe`TN~ST@}{Sl{{97xAC*v2T8(AnfD76Nl+R?>Vq&908pcd0?+R)D!0lHGasQA zbB=wBHRwh%(l(RhvR3;QRZJNkenl(frAV0N)ck}!pnlKKP!IyXKx^+JxR**L0;pNdL*}8gG>sXZA-PvD>M6KljFfV24Jke1B)6U%H|vf(;bhO>Hk# z{!4JdJ9kBE0|SwVhsTQj7t<0pXpI~v|F)2x|K5mdw!wT|tMC5UGfb|=bU0LG$^(h7uSrkQ%F(Ij|@8xLqY1puBzL_?EYu8HJ;@{*_810_~n z3{fsHEAbJ$B*fBQw0l-Cwu9CY{bcBnTKK}m`SxDFrL1_mxguAT%QChT{*A(|hl!Ok z!+B~6Mv=5sg~^M{7?bv!cadJh6PUYQsmL&amVz^02r9$_;z<)2^d1R753~Gy^y6UA zn;pLP`M1%pzNzTMpRlogH1)?Y{?CA@u9=-~Qe$`k1*+2*dEdnDLJ;1@TAAv^r5$Op zRe>npfpk}W-} zjr|&Wr~NN6-I~plN<9rY@(}7;!A|lobgXunpCP{){qsX299X#8pw= zo_Ba$+}t6eKsS%)-=QrI#+b@60zb}g$H&v>@34q>QTU2q0{Gv2zTZz9PG6_T*Pn<2 z{rSA_GooPbYxC{d2%Dem>ES`&-`{wuJ-J^u^qEyuEx4pn9=JF+L#T5RrHN^t0OU_15iwCjWOvfTL-6sE09xW)LMiPQ|hb^ z=W|i|If3}KF=dA>)ivaXo^aDuSQtqqOx$zl-!H6{+LXsjvU~R>ti4uSOfZ4w``Bbb z8p(c;$*i!e29P1TC>RR^cl7ia9LaWnbS*}lVVI%JA-2jwffL?=`jWOt8OB=~9Uuw< z#QZ4T6wGD5U|09_5FkjVu(j;Rjckv#)bq7ef}Wi5;DxvNI;t*R)oT=q<_Q821>pm^ zGEMo$sSunh1TU-7 z$v8wMhF8&9M+Sa-)sVN7P;Ck*7Ux+hRLZH~4cvVQrFV{gWWt-!5x4_Z@g_^Z>UH?jyGu~VsMgw@TczW}R+KKbN`9e)M`?khZh zE^fiU1V$V#q69&ns1M8JozGFjE>Mc{J8IR%sOc91+b2Y|RGTiQz74PY+ll7$5h?=8 zf#YzRSJ1S*FED~T0-9^Y2wPE!l19%skQdwF#te@K->pS}%J8K|nZov4Xt-$Y-uFBd zf5%$>;8y74^AX5JgT)mtjF9VT_g~i*(w3hkg!6%5$E>n;bM*@wpV3IB{)kAU2y72_LN{ip;75X7(pNUT&L# zOei~U4|w#5nJ+-h9P{IsEuYnqcq=OuBKRB6!d(Ttx-~)S2Ab=6DZcic!U|7MOx;xt z*PFV2;C6yiw^-T0F(FDpbM<09ybglt)JdC`6vOI+HK>9i|Hd#X>=l4jU6!Qldkcjs zj5yVNq4F5inMtX+e+Nu}A4lCQlfZdHkc-O*BcNRzGYT_EV4Ud-*9akS@|LMGuWckK zG9V~s(or3L7QL(M4aiP&wGX<_4x!fM9fu3p{6y{dw<%q{->#_uk0jx`g}QZ?hFPM4|3NNt3rQ9iQ)=9$CugB78_K5 zi`^z0T+y2f)}m;%Nuavj#pych}@SJNJJeT zAOst3tIj!^+Ewelqgz@Z_0N>L$(bOwAG_hgJ7lf~~VfPF?vI$nJGy+!=#ln+7 zU7GY#&zoqHa$sR$yBuDsG5T;qcc4ZmYMY60e|rX z&+s-wsFpb763{)GG2h?kZCGAx|taepSbKf3xK(G zSjdlTch)<<-Kew|42%qu{DQ)rN(&)3OZUfu3!a+B^h@wiGtvigWzRgaBlBM&hU$;0 zRqsw_g_3!=*%~F{09?2sJlSjAT&^V0fgP)ru(%TIDMZMWbVd-^N-$pSaCLUzRJg4= zx8~AmBtBG|)JHWuf zd?zYrqP_@lrv-j4EVO_5OQGNm6a#yHvU^$nUbVDw#vA%O3kSJp1xxN%M;_Dn_dRs6 zFLlZp!aA!|*P|cUPyKKxm^`hDxT1syiwYc}vGS~#`^Y;}`HPJvVAH4u2rVG@gCYa9 zKiR?fY9@25jPdRLdoQ^z$DbO*E+-j2(=Lm%N>`xvYmPZQYs+CEnLzV)et(Z2<6#Ig zo(Ch=K|Uu1^MUSMhtVgBx@5x|7CSD<7y`7YB%Wf7qn8qw6WM)@3lpk4GcEj-K1|BT zA(fR`%BA2`*BcCkowu`+c_@Eh<#d9s1J8CKok4eRXio%G>8y;`iuY2x_)QiQPQ$mc zD)WghyhT~B;j-*YeiG7qviq~Rp`7V==J+eaS>1u|(BgJ~|JJCM*hx%N{v%fe$NZMq zWD}E53dYX;m$qtC1RIz!3s^sVrF|naIWobh#F&YxP(n+SWOogdyLl^!8-rn|Jel#{ zB8DhG@QwjpJwf^nH+MVyU6Z=GQpLOK<6q`8rq-`e^|esYU+1;bQG(E_{1T0-9%wEz zjmz;pavIuKhdA6#YTJ9_z&^gZYk34EfT<1(o3M@uxFiQxIqHyZ9bIcdn+S;Xz~O4# z*TXF(67hxHv={b4Z_kK7nS@N{rvT@w0nXp<&d--)Np#bjbi`L%_rq~_m)Qaq6MWii z>A-v=gh{&Xd}kiUE=03_+<)#t#r5>_dbtt=7(~&;s3GMCfr=?R(5Z4FIpwRga8-aa z@JCME+}t>~j^M(92H>iu3>vk%s8&SS$ZZN-gQ5eX=c9G}uXOWkG$z;DT?XaS*Q$E$ zLH6hVp|M1Az);()0j#a6-$FtREeYlCV6=LxV+p^@^ZTWI(ohGDOR6j;lQX7 zi&E~@kSMp^ku}W^)}*LLW1*AIsd%+Kp{rtQGKe~QU@9iI`if8Ji&Cl-Vercf6>^|H zJJSiF_JD8>EKhdE01LD4K%#Tg417MG->g*i15sWxK`b+tR&^@SeMd1Mo9P=>F$I7a zAT+Ic2FP4K?CXd?b+Zd>3tQgVQ3lpJ*RX!U%Tf84XpYM^qM0hVrY`Sb*`}L1VV%okKd@A|K%l8=+l9CZj2LBTe5bfP1=js z(87uB%czzZKq^;@U@~WJswncYl(yRs%P+_YosM5IufE{Qk%sr1%q|f!wE+78kS)2> zCvm$3#2_F_C^=;DvoYhr(nQa>BhshSV}eI44@C~vOJby13A1%^8DFv>4qL!~KFsCu zb@*<7Uqdxe7X;rkF)l^gO~V}o$dnHsjE8@E)Mv?I^=afnqd%2=k>M|$mudmV(`}fU zk1IE9)p#-lG0KZ6D|rwsS^$Z^nm0A*Fe1u1DPi$8u={ye+BXhS;+u^ruZSA;yQgX! zf8%kUmNy5EawF%aFw-He0ozKu12eb}WU^2sa;ZbgiZ3NtK2|MO1*wPR`+(A7%hSbk zl9<3S&GWE9VV2m2yik6*z8T~CT9+<7J>>1*5&e7{kSLxU@p}sMc**0Kc)4zdFwCDt zombYXPu&|%Uww3Q4}|wOH@9OZ!ss9Bb6mDe?%bST{9;tIl(h70P1zkWAN>o)A^AEx zz)<>bZJ8MF@D1S_puca8uv)@hk$-E0`+_XnJ@Uv6J!bj@(*NXhW#r!Nj*J(vpfWzM zM`e|deAZMh;y?&u$d$I!KsD;P8P*$!WLl7%f{c2v3eW7V@gZz#@>=$==IPP+Qx!f+!5^OjXE3{vy!uCE)!L| zpQ-sNE9wmU3Sw^P%<`?`zod^k#ckxD0@5+00PL)@aL-Du1KjHWejj!{&0t&cE87~C z1(HLEomZbu*wU~ciCnFVn$FDTCDr!!I%P*Y?-}ooK-22QX5k#$DNAu)c*iVBHPDX7 z4O^R(tr0&aKQh@S7f}sW=L?)_ zK0O)CB_bDcyj4xbY%OE>mU)@GT&CenSx$_*9H6b*7O?r#x78E7#0%%EboC3NdB|XK z_Hi~h#!?7!i0x$wu`Juh7k7Fj$o^-&s_P3dL05~Mkv%W`dmIGo!u%aCKYI^YKw*v9H<>wT0-p;%oxI>LHC!o zZFoj*s;TA`{Ix-qCf2p&-bq5s7O#dp&IFZqK3F)`e%I>ChvX(K-^kFx;avDV3?>~o z4Q+7e*w9!@T;1(3D^Y23tAw!%udUP-jd96yRR=?^5edyeo(YMFvPze1)mKtMPK{AuzZ1oC7`X8s zw-0@0+-qX&Gisjc_|8_`qJ?D=@XTuP`>w^}^gTuZgY1UOOcEs17)yi}#;={xVI#Ts@YcsFxwyvhZ zW6k;k2zr{C24~YjgO|32Nf|8oYy`HL-oWjtb6>_bTcghb+Lyz1^^0kvuwnUKD?1pi zk&Q1P>})rdID^kA9US=ChPXKF-@xOwjYE=4b)|!t)R>fy0a7$pgrruiVK?CdVy~gJ zY$0wE^}h#yi-@bbE7n66Q6q}N=2<2^7{SWu?w2NsE5bemj*4bpo#Ijh`k@8#!uwTW zJMB@@B4x3l*SLmBkI^ut$x^N;mB}|L9W)=}z*G8%D*01L0Q_L|iG(?gT1dFNnnq=w13{w|2|W}&HbHwKm2%K@?c$v0ciL;&OjWz)UjoT?99QA z@Hy;A)InB|%}#_Lx^{kE3Kd>{C`nj6cbuMj_(PyZ zEQO`{e0b#mt}_T!Pd3^raX#Z4MnjLYahFlgonv23z2xqNa-PW%(}yq(*RVL4NPV;m z5aP5TZ6-ya{pel5o~VcZ_OS3r)Xrjs=t`D1S!S&yE-J~$uIaQ`hJjE;vdG>m@%VhAdFxYf*<}lP<+Pt_6e*Q%e5V->TA2IbA-os3uD2)U69}SD-Y>?%FhDd8~1DWn%!VlRJ9GHdaC=J2x zkzknM0N;Vt?YdE^s3 z_Uv@=QTz%Q+bA?fn3n=sO>`f z3H_xYR*-A-XQEbRwcM{LMw?UGr>?uu+SyoULS+?`F;gsUr8WXtlgRFGRCg?m^&$|m zXn_&D8>t#K_DS3lV-NGky@y&l9!pZ*Uax=XsJ+6vNJ>K_>NuZwog;!zzZVKH5AdLt zZMhHd{NZ!`hc|vfh0>|;7Is=ZZAhL8)&hrnC!Tad-(B$=2LH(}mLWtEnpNF^k%a)H zgScPiWZ)6BB-c1i`S@)MTx|Qun5&NA>TID0AXzx<;s=o-uwav29EnU?-kf};%mAER zf1@kCEF2w6*%?b$Y?=B&T)J@*Hd7e)D`7!`@@%=uY+^~5S(TOK#qQ#LpSK?0a6zxG zrdhmk0UaB8>hn!TZ|6ck>L6BB;`KLK;M9*nb4@}N;|jRGR@g=xT^3a6>TEY5ywc~;LWM`=Ae};nZptRskZ8+vPxCpDb-HiZI(9`N&$|UFUPW)nd zCR{Zt18{)4HnAYS)|3(+XI%^Fzanv+eClzr%l&DJzh44&LAYlaH%Bn#2s%$%WAE<6 zu7vB;T?^#csZ6c@Npj%UgPgPC9_%%OHGn?xs0UTk+qe$I}W)#M3p1EeZ1?w-QpB<9;x z4K`2g4NJPv1AseUfzf6gu|kx+rlF@#OTzXk@wqbHiozMf@VN*f*eIjE+pk3XqHapqs0iGu?wUv*SJ;&%Z8&UKtE-WwWi(tdV zmC?`Jwde?$BTU;g^7ll+I@-qV(#Fu6gD_hjpwc8NT2|n|=zinhBI|u)L-TjJHznia z$*s|U<9>N&IZ_81jyQByPi*f19(rf$9mrb*nf)J2v=Yd;@<;bXr+2SLEw6G}n$kvIX+ zaM{qYXQkpeNP;SP8T0F)aTM056kqHSGw)I6yS$A5iSshUySf&j%yA^V=c5_^LM=2Wfd+C{~tRu~SsYbuP zJE(6+Fd1)*W<lpb7Zi@qQ3S zd#J+%qF=j+hpF_>9C9f4N9TvY8^>2V#vzF#L8qkxW&(Nf`iwcwFrn3j81|2i3yclb zKWc{v)1GUXvoKzGs3nWvv3s^;9ZQ^F{G%=y+fwkQq5*Qf4Tkc21F2 za9T>KWTa8XH7uj!sU&SDTp0F-EIVJT(6ey`#xB#C4)JS$hlP#u_Es_dRSeu@`ig`y5r8tsWIIlY5^f z!6JF!&1g~kCful{+W>yK)gt%R;PSa@y-L1e@o_E6>d|8vY-=dT~X`hv*|J+KgY>z ztCHTS1U$E}Yp-hETLoRs0ac1Vqr1Vn+I;Fc$(=Lr=zZ>%`Im|KGf>Mw>L~PCTzt5J z=&>V@ErJM$Atj2zyQMaPW97qGpMOU8j-KWD5S~@#FW?!K7vbctI^)&cNE?kfnNp%t z-At`4qw~i8-0OE0wW=~z@rA?TRdGjY1WEFl-Nek8`SgX?g#+WMdkhz8gltLb(3BCu zNw2c9GWcmOWU8}J(?vOc4Aa$WXHHKS@MA&5aZW8MuSXZ<^H5#aDCWekLJl&ps4wO~ zg932g!mn~C(T{sAPVJVboxK=d)cUTiW2ye%DDEBRt~onpA8M=3%nGiQW%{d27Jar2 zbcQ}D`lze5#12YjM}pHnR80EjsFhg`yf-o%XCB`}3iD-m1%!L0mbD&AFOJ-ljDYM} z@X=;UZbodQrXNYhl_=DwkHO(M&kS|<8Tvdcg%GVwbww;VWyWfTqzyS0K>^sifU_2k^LNFf!)y-qA`T=YoOn2y4{Z=CRFE& zM;dpZog%ak{r)oDxqEsg;r&F;`yK|5>0LK&@1#&2X%n=Pu2cfVw=u)n%DHUYg;{H+ zWSm!pxRRfjDWpSpLI(AxeCWGNCslIu6~+{1U$Zw(K;)lcheXxJm5hc`AI1urk0DL{ zfhbFp+BU*|l6gVm9+cTF61;X==VE7QzAhrx)~Do2X>7prA0ko88~X;E?WiAHg$vzn zD{;<@--Ue=#MHjtUH>Q z)1=n$0H&VA+522Gb^W+zi&5&pBkh{0QX{$ZO62l!`6aQf8vTp-X(RI|Md{sb)|dSZ zM4Mxlu`rR~61zi26mo924(yZO664iCy;Ns2A#v7Z;{anOQ8I&@Gy|vgrNqZ_&1_LI zP1jvMJt~U2fwjiR)>5kgIR_1YJ0jpT+65#}W^gCL2uKoFRhWqmW}l;bP$%t?-B%lW zJUMJDn7%u=8(GDl+s68Y^f!P%K;La zBd4THLv&b9$v7DY60b`qR;r&`PLvncR&*lt@dk4$Ke zO$<5rtGsFOmz77bA0TPf(dpwIZoeUikfF9-7Ce)qsu9>~>T|tF7fJGt#9~{EoS|P^SvydeMG-XA7~Mr`)|R2qUjR`+LPP z3h8yLo0PbQzM)LrZxmY)7-+=Hx6Vi_QP3ogg&U&HbX$J*JeR3We8T-L0&X0o>>RKx-q#isv#JI@ z>P*}Ax)AAzx9c5CdQjen`gIm#*Ksh1}}>As=%zSbVQmbKNX%SCi6v*wrq z-ij9pdd^8eot|*rhcVF#8AeYTg2*YUgA`lV!k`t@{W^IFuIIjl+`%v}RQZO!3gO1d zluv-bKd_6T;rQVp!@e%5lVcu3OTgN3z8;=&!~rve-f-n>qe}iNfBV_D@)R1gdk+Hi zSNrH;dn+J8085?-KpLKro*jWW(1qul7dRbiPse_d#wP9bcu=bAZVtML45re!3 z@Lc!tP{_(+So$N$OP0b(AuF*WOabH=U)3_`$>plWVR^d@{yVAh-%9bt!-NmK9fru* zkpp%-^cB3?oWqXHMc_xEXs>1-kO#^Z#@A>xt5_{SFMht>kH!wn4g|&wI+Z{Qs?LsZ zNH&7HQk#9@UKN|$vU{(j1OQldOd7bk_@trsjgBU+-QhE&QXqLa1Ee$zq%O=Ww*HCC zgpae6x(#-?v(3%t8^B44j9XgLfk&R-(XwF!-tA;W+a-yZ5t`YYE^l?ih7`X^_X$nvX z#IcK2yjH|7vH4jn#FO9W!nZ6GFv4 z%r-FZlyDL$Y8FB5*PBix3|QZf(*}q!hHc8&4BLf%Ao%tBo*qd$q8S&=(v{BVJEIkc z?SG`B#U-`WY{O=*nUe9jD4Fh%<|$frO!ZyKDa!4=^^2)jB5C1}f0uY#Z^_NnXGVIp zl~~_bo^)Sp^JU^kYHu(d+1E6nHkjeBeyAa!07F<#{vwlJ(VT!DTfB3#)j!*jgogb@ z&UD3}>Q{RUX;a_RP;n~#lsAZ7^Enaa_Db}!t)$g|%S@+d;eIku*qKcpl5Lef2blw- zw>^;p=8lc^={pRsWTK*g?_t$@%J)us{ub}nUrIoTE^&1PzFmUi6Re9bbb3AyAOh7 zC6~ZkVVaXh9)@JLhs%D{=z(}3(j=Q-Xw}rYgPGY%8T?-g23GuEOE2J`r5F6){&f`u z2|M-U7-XeH#N<`P=v_Tr%~WOL zHW`t6KQ;dK^wP*Zb<06%=_A&X@g}LxtPxteA)JkPnm;~c2%q_4lH;$_Z<)w0Ji0la zd=T-Q9W>Xnh?Nclp+#=<>-?Ne1VMmYBu`fSZoc3sXYO9lqaeuN{!KTd{FK*!&l%t0 zC|0c34IbM1P?p+GZ8wzwz7Viif_)X_+W;VD!ZkL~I<|Z;$3M1TX=z%?1qX&*v-+}s zVREco@7T(X&ZA$EXv=GZggT_O+fi%wu=&6gbDLYS6gm+*LwdxkC4_ws!{S@qD2`Am z+>Bwc;k&ufWmq^l(L(EIlsYtW$U1*ug52U%FsY zwh3gz5PAz79u@bv@p~x})vFL3*_o2wDB9BC2hV~@w%TMm`ifR2m~C1^-Wg8k&+bj1y)M1 zO^HO{ShtC#yVHvJc6bKD5sI!7l-nk1N5)Z(2x52wDG%%~+gWR*cwRZ_f;YFCGK zZtVmsl7cz?4_qqSDEy*-H?En&z1&9ukMKmc0eb*M4# zu~!Sd4!EWc0#-eI#|@`W8aE~TVT5Elf~p)&f`9R6*XY^&C?z`lmb_ zCOkjj6xKD=QDAxxQH0FE6Ye2KuwLoLBO^Xg;+`XZ>E1GHI%Wst|p6t0)3r z#b|14Q1L?KSg<;NhD3!x7prv zhB-fQjVV5)9EpE40#6NOpf^#DlF7O~3o3Q^snk4A+gTiTxYFcl?QrA9V4vat>*Iec z@9+LT228*}Ko+1tKp6k44+Ap~R|W$ED|;(f1A~A5_Y_4vn_)sE@3}fnZ)T)HzF)Tt z&%e+}<bk#BL>G z=o>YB2kAy3Sf|(wIjVh0UJZs3qae574O*(K1@(M$Xw3v8Q!9W4PCypL*!pvMs$AEa z5aa&T1iGH0f5=j~lUl()P^+(H>DE3jp{~BrqA<4H?Kejybb9iQHjS@=e4TWqJO%oI zZff%O{SG6Osz}FY6_9y2+spaeR3WB2=W7VncygFN7oFQ=Bg0D?WG%L(!x-)Su&tVX zhEgNOF+X!pITD8YqTfh6msD2Rnjr33eh>LykmE#n_wfEDAxW@*iRix|8v$G$3>+NI z?Ej&hqr&5`#fY@~LW7`1iN&te9EF@Ok|*eofxnqAf&`ZAnz$}mER$PE+1KlyoWp4q zR)}r8`KA9BuWa2W0U2|HOfG|$p73QjuI^v^*aE0grfwSAg#@{ZqacW*fG|RDqOEF> zy2hD{{Fn@8uQh7v%sE!gw&;(uoOruc9|BKa*Xb5vRNTa%KI-RS{9qJ4mkrJ{Asuz( zRT|c&xC}d{kR1H2owx<%j%jWlf;}!06O2yx?tqVtn%{=#JkcrdL{xh@mYK5ym&*+& zYZRAh(~7wkfyt*^4xCMq`mB34@7PmnD$rUV8(23M>E}951!Cc|oSr18FNI$W(tTxN z$Zle{d_TKl7(tUlS+d+WHt%TsdM(koOV%5lHC=rGn;40NT=1k)X+F4IyDGxP;IoNS z$(@_JPTD#I4f2D{q`bUgMYFCwp5p|Fo6XVy&FNIn%VGNh%F*0!Qt%k(^g*QUm`HVjMZJ|sNi@}{d`6;c~*wx z?Q!rE=JfG*wo{(HquISmbGmgFQdR5~=f8~7eImgl`^TuAe~iNaFQZH>%}i`uTs>{g z900D40M~!S0;mqeW-%i5JfV}k3dU1bF5_2LRyRj3scXT<~r?g5(CQ(-j@jG(~{N z1}O^=>JKl7ah#46U|=IqQ*Pzgvp=y}D|;fah6jxj_|DNM5AH94v|Jrlb`u9+)?Sa0+EEPBoONB2{t?IR z>B<2c6pO83BLTnEsHJRGSq~CWw9YIoxNc=61-9Bh!p|k}<>!?j2IDf^8M^navlIYm zQGVO+U1Fu3A>_hK@Cw-kcW$gJb6Ali=fBlob@D88?L=pZ_tBS!EqQBkSMz&E96o#s zrfBM;{PZPPxOc-ntW}f=UPp6}p{iY#D5s_Ju)))N*NTjQU4!r5R?mc0f>*^kL z&nzCG(i(CM$Zmc=YNTKRMOa2uGlY~Ix6q)esC0|moDQFpZx}A`5+9&YR4E$bQREM& zIM$F@Tw2LK=y+`6*|VH~E1OD%&sw>l=!D?hP+~3fqE&v9$MH9~Jz!!%OHXmbu62R4 zD^&q87;FIn_3zwL`{38`?Y(v(d<}>saBR}$YC-gnMOiBA;fByaX8YXL-Q1yql$D^% zJ+xf!Fk9f$zP7UEY3x&gF>7j=QmjP@)^UjwYKs)Dd|ip6lKU{A^KRisYzC!Dn0g^ zcDw0-YWGsBaMJeA#yy{2zxA}Y&n`$}bQ~;Ymx7#>a}Xk3cv(inYEd0WS#chis$o-`uoyIm<52S!>Hz#*?7U=Lf{eF+Qi-SbmtFE~FMYf88v(xCG~^8O-(aWxu(&(^p3fix?B%&1q+D4t@jo8JOWx3I zvzE;(b7XCye}jP!ATLKVmPopF3k-RZhq;|t+tE{TqeD2aj@$Z%a`D;%^r8$UJ&F8mb_XTF#Wy;x zYwVc}-^|`?u9R0ZIrmjTuV}u({*qstuxle^csy_EV<#^$OGC)1SIv=-Xf?a@L>dlXD;i|(+&ihQhFkIr%bmtZUavcG6iiTT6 zG|mtQtrqDYGlStr|E4y`9`|0$wE7TC0MsVJaS|@`q)(Y>#pmOI z)%XJpVuCW{y5l9}>iMITU#+W!z~Z#vU7|~m%qnW}ZFai7g}Nibynt2-0SM7PFV;Bo zK(MokUZZ`kp;N~G=nAP>QsLd)G4dG}4j%>CX0Q4lwi1_?3g7<^vzkCGRtF)0fX@CM z#}fTFXiFno7gsYQQv(xQBNrDdb1O6Ff037H%EWE4|Ie(9G&#V&56FBzB@QFpXpRD# zFm^`-PwX!&Nf^n*Ku>HuLi@PO!BR%7>&a@~8Hz-iaKO+0Fh3`i;$Jo_;drem5K4)X zl{r%_V7^GH5H+fP7qLDQIyHvft(=}oBE^bXZe?GMg2#OwJuO-YcXx|25W}rUZTbVt zET9(+_H(ey2|(Jl&36`7+kF<6T^@&4rl>qQCE4}Spw_($W)Rw+4oO|jtq_2nOT=iT4$PqDf$U?80@R)>L zayP7*6rwGDb4wUKnUb$j<+VF&etcJA&>2^ijW(+X$Jg$VJZivK7In{$qSD#uyc=AJ zUnqz|Pj)VoH95Pn0w42h@9p;Vd^-^DMv|%$T%7Cn+mTcCeE0o)y~Gyi}m5cM|?sd|FEQT3t^+a)8{L^$5w1##ZzD@{SZRMUDwzmO2W_6S$ zwJi(?4H;HVss>Sn)IhuGtcBH)buhdnjC~~&otmi$6XGaZsX2k4a=Tjc1>ep>Ri1^$ z#Pt(wf}1TJ=?cc9lc3-ywjo=h&s>C7y3jB-q-D_6^njf!N7(McxZaA~4k5wL=LA{<}@Hn5y7LcA5`f@>>G3);kjh?sxkrWt1tUa-lz>SwMoCK;Px^Sy# z=24pbX)7y~jJ(qXOA*11I|8DOtg7Te-8k9Fp&mEp%z%DH2Au~137!eA*a$x#NR)E0 zVsH_2)~i4u*^lc;7cPl!OaTDR4uscEagT^V2)A<2GsmCgj*g9YU?$gJv5cLS4hgI1 zoNH+#ie(f>8#_-}d~d0d3WH`^qrJDGrAR2Y=R_9EY~@WUZF;@=zG6j}bhDC94n7Jr zz$BCFNNL^@9(^l!a4ti^>~GbK#?5q^YSw2~j+F)gTmI>+Aqen^`Wd1ox6PVkl< zg`U`Fpd`_-IiXPb^L!yWOL|B&FpF9X2mHQq_;KdD$9Iw&o(7VHrV;?e#MZGm@}x7I zNM;zdBIrLY{ZPW>dfG}fbSHH1impjbbyHvOEXT!ILYNS!EHC1^79<~8=5&a_Dty_Q zJn+7|Y36K=8nRPZ;-?)8j`ETIC)#8H3O`m!?BdPj4f9i?GdF~{LRcUmqTyo+K{%)? zh~WaMB2ejQDbkDEQo?I!PMSxmh-|gXS&OI*9@U!seCkow8St&0#>%oBM5erC6j<9iM2e=bCSVsjvqif$r&2nB{#pVaT zd8KIpb?Cgde#Z~o#&c24Yl|uuyYRf`lrD7S2hIa3!0^9)#Cte~`}JJ#;dqw{^L#J2 zUTmT$_W!uOH&h8DAtC;tD4Y^X6KAp;p}#ERp>XpDZ8l-)dTR$3`V%S`7BeYlHsYNF zCuWr{t2(D|XBKt^8lFfm(Xr_Vd{tz?l2kVGEi=m{w9@Ls)hCgho%q*)EkG45MUR=+ z)%dfIVIbU`M?ml>v4tU_?|g&JtZ1*uH(;+W?wwqj`yZrT%RyvJU2g78vacj}DBEy$a9{Xe?CflaVw*|KbP zxy!a~ciFaW+qP}nwr$(!vhArmZ{GXnp1JcEGS`lbg@_$hZt-S#BHDjM!m5d=q+~j5 zas9`1c-qi@N!Q2wC8~$hasY&dEnB|9ZsZG^HDf!3OvMaalm3F~YKs%+nk~L#)wI(L z6E-{&!~=Z&{e>8RdYaWEs_IHnGjsH;z4-izF25nnd33Lk6{EV6?%nnBd<*jvgFis6 z#f4Sn+DAuwN3*-_?f&k`@5#(^gpuC*M2G)d`Jj6Q6OGGkKnT6Yby6qX`Lf3}r@cOy z*(K7C*}HYBO(%D~4UC&B?aKxOD_otT1%^g5Sd$I#Q|2g=qw|WB)ZqKg?)JiACw*?S zApOIT^+myHUvanq2g)&~DS30diSRieS`F9WIoQU2RHvBDVG9>|;r9{rcPT#li~mc* z2Ej=*GcL7Gmxr5&$K&6#yWo7*K}&hYbJ7U#^9job>?VA{mEDxCS;#v}Z$%5mW<8qI zl|g6jN+Dxd=?*{>#{nmB1(XW{Pj8lyNr?ppb2@~eCSclhPD4DFI77!~(2JJ|(b1e^ z5GL9DE`$T?d2qF0EO9cu`&4md=+;AMGYvcMto!z-m158Wnt27_Qj&=I5g5AfDTPmi3MrEdQ=rI|infy6$xF;@7NG_A$5-R5>c@MoQ%&$( zSuYhk+;zidG9cgnEsj1KKAnf}h{j~2`)u$Yg%3FLC>!g8lBUEz~oyw{G zx%Yj$c9dslKl0X-(Gro-=s=3~Ma*_kh@uh>uah=P8(z^`nUcpccLz6Xo4V+IO%~BS z4)6FXd+qyE!Req{K}ojPhJ1(T>6_(=QzT_2LRWniVbPkx6?Z>F=acMI`ol}%aUK1M z4S%O>8mpPgCc+CEHpnCmUTLFVTCH%ousbHgny$tq^-i2)nvD5Vf+K^yvj?3M)5xEd zZPuHZFMH<)pB{^GR(8S6L0MC!v|(GZ;IBaTqQ!3T>l9Qqf%q|G4z(wwUuZF2^XQSZ zrmRv-4Ce1W4A+)bwC$Zt3Pq7vk(i|6Bn+V?ou^JAW__cf0j)1-<9x9OLxF4BZUN@CU-9sKr9wKk)$GLHL%7%{f zb7Vj=8xk(+p^aa*pGvy4Xo#$k7&Q|=XQOK6IZu!syC<{4U>_I1mI_xH3!|QlwjEDF zMiiuS!3KSzaHN^-Ra0(81>su$l2d7&%}`;*C9q~Mc^&au0>{dslCHP99{j61*V>i; z0eGvs>2o*{zu87 zKLd__YiJs8Y6=Z)BL1sff-{wY3+dS6kaxhV1Y&t=JZke@Y(HW{kBdUlZnm;~6L-$< zbVsbSEgu=OQ#SU}r6!FOF5aj%6fx~tx59|r~e`~Px~ z{-ux?aDpfxKmY)Q{RRL)|3@qAXl407Qpl0oxx+^5ukYy|{eeYNa>DpK!Ug3xCI?c^ zwsSVL=Su10;ew{TV_Yp1L1)eyh!E~&;xv#ttfeoy>SKv7!w^xf|8+i#+#UFVG!E0gc-s9!+{& z9x5Cf?V=&DKG}HQRxz29N+B3&_vzyn^e@}q^YU($r73e9b|wX`k8%DI`sMB-V z3?ttyQPUi8U`Dq}2x`*E<-@>$K~GE^F1?n#?ypopE7X-CcoU=rUyfwE+%{OP#mS$p-ZyVlb#Obisd_}{aA}!Occwao`NaVYE3V}t20Y+i z8kxQ|XqFyE*p-yOotAx0lKaJahy}ScDsgyE`WUwDfIYN8pIQ9uq#jy!5cm~5#7Q-s zS{GgDALNmY1(dFn;Y66p$vbUI8XlaLP@I*5A8A(lV``X>CoA?X%YnbDp#e1ivM!I7 z{Pc|mjQEK~$9LWZ&|3601hN6f*eL`{Sp9`!>>`9<>$&?T0mB6;5)(oA8tJTYD`rI! zQP~n%UL*8-13a+vHIZ81_xZ1V6 z4}KS`=cy-CITc{)RQ{fxh4_%X6$be&H;4Z9_Eb7><6+6kygiP^qFzNEXiJfh?VtKy zC}co!cK&H>xSGMg3q&dij`ySx=AhEe4Ss2-f|E^Gm2XnK`?&V;{%j-S#0ffNs46m# zQ*7ZyeB$ZjvP(N&_`C`*9NB9j0C!)4$#n{`fX2FPh==1g{I6;KzF~<+aYAQ9+BIi}PWUvKSTggNMYK0#u5Vn~)=yeX z@j`;V2{K_7{n|2NdUfLv(wPeb8zoC|E$!5Wo|10==7L)p;i4#iSk+B=PNGXH^V&c4 zAT1g6vAFX4(W2_Jt1ckiqV2d}p?6>)x9a9@;e(GW%x!k6tAfB~8sKe; zP0RNg|3~S*W_aHIJ`ClxbHJ&|Y4%4=HR*Gb6h6)x=RwJX-4?!YbP$?sCupSpjmPm9K2Fk_T#LPD^N?_ymY92h_{WW$bb z?h!N`;Hw{{9?Ze{O$zeVoZR&$y4RChE<@hx2q##^*kOxU?D<&Tpqirz<}ZmhRv`rS zfTfe-_!LV7*)aPxHbY=5BZQ3of$I0=sC&VbQZ>A?23t`B`OB=vfeofO_d@uq{si`_ z@Pl}9n1XWm!loZccRv1^zh06iq0$R~<)D@dLDgTD-S2nX1~3ERAo2PHnTJnl-~){@NC6 z88adty+)qm75f7IcNLL9_0&u7M?v)Ze|>ffBR7{Hnf70OwAB2B^Z*?K=&VYq^9q;Y@HKS}4UNq_XgH^>E$yR*t4v=f$olV(Sb5C)QELd{`zAPE?*4SXKx zyRqu<(k9=RC~re?JiV)(zl3(9f9nBw0`BxO(}5|Yz5xm(&XqRBNM(3r2)c2YfioiUu zU6)J`)?^G(rb$D5hI36iykgO4q))++?4v|G9&Rz~m(FVob98|@_P#wJ*qHWwz^eDv zo;~2~6o68+(rlZ*7SXTm_W2AOv~ouotWX#|WnxFtigE{J7Ufw0aCaY!qY}W}2hMcs zKIM29bsAv$PwjlG8K*^8P$HelG8AJ-y_E&>Ezyp0gWM}fA0la((IVB=Jd9yj_Cc=u z3eM(ek&6Y2#>8%PV2AVVfR*H`CAw>b!0(XI&|OHnY=;bj%QV&8|L7#nHphz8f4E=q zHt{yu`UH6}n*Mak`pO>qR|MUhZu6s_&D;ui{dSsbp5MX;eQc5$MsoV9k;dp;QNH51 zNc>HZQDT!1^pU*s6aXk|7$Tc=zhzIXWdXN+&}`Z4b0CeN6An9jYSUY(v#*{9g$m5|de}M{}V!yX)ne&xM7M#Dqouh1zQC z^!|h^3NWq)m8E)JH&G|XQbhsDBGc}98?)hOx@_GB|%|RvbVKBD55d_V;3%Q6nb}-BKxlpz;e@{0y zDbI+zJ7uqfF)fQOt3%>!oi9+;zM6>75>M=JE@4!ct(5yrV%LUw7e z#pn+bK7Ww#4`%jXAGd+6{l8q9pZs&gPp+oxTm@#jk7X)7F&ksF>_PUV?4emJ&@#$^ zSpYE})#mM-Gk!(QWJS&WRJZof?S326tCdRP$w3zI&BeauphCAK|a7Cxqe#}a_c-+?$oX+kl> z?ncpy{%O%o=psVEn2Ug8afNYlu#{t@23e^1z0Vit=axMK+_u}0RqQoDXWkOUGJ{>9G5+RfUc*zXDaNmX z6~!w+$pp;3obRUk$?C<5GlIxlVTjN0DrwtqVpxf8&URBANit=@qMk5j;N|)Xi#Bl* zb;4{S66|SJDtXWoxNtKVZsHH*DwQc!@A%Y9I}AzsSc$_CT^p^)jn`(sX}j3Y%bgf0 zyEd`yKYJSMP03RFB{_>9Ab3C5w6eoH zn__G_(K1@-3_}IY^`B%5(3N|ZEtS>RB{r>{`0R{HV{gc>Ba#5o^K6~IcfU(V=aaRP z2ZFuPgT5Hbn#oa+so)LaDjze_fim_^G%=)S|5^?3YUZ@BlszhK<-_KJQh->x*fom2 z$ID;m?9ww_PAxZGoAt*(&T;25F3L{nB4%+fXq>RVs?ZQU&OX_wTHTcJK_MF^OS)Vpo zbhs<%wRIK(&P~(D8Xg+K&ljv2fF(O~zovxXtQZNpluRaD6^P(AVE179Q;OeZ6{btU zaV?QHqY7B2)r?LD2M)^;{@xFbc1D0b*vE#)`?}e09cC4A=aSln6dL}BT!O0`yA$lO zm3k181BBvc-m%qa1do?o)5tTN%WcJXdQzin~T7cc@PA_ZS$y`Ws8)Hvr zcgCLiHaUNJ*np^nXGaIHH)C`gW$9IW@WL26z_W+Q*tEBvp5H|N**`t;$h#=Wz}f`K zk>QxPD!8N;LE8D&?iB9~i9E^a46!uviP1C2G2s4#ZS0|CNXsGP*JbIok}mkTDze16 zU%jgBGv`Sh0-sV-^ZDKpa>46qp2=eaGTzBe{qm2%;7e}wk?BumtN6h;-haflla=ky zJ7;BN?f9=p&gjS1|M8uJuf3osI~tJ$EjeMs28j5u5E61ZMTnj7G}0Bt+WWG-r8R~D zaLXy4zTX?B&?WA=?tDXhhbb*`Ez8#h^Oys<{cDwiNL*#XD*0o@4f;iiv#NX@)&hs| zk8{h{+X0#sNMnpkzcB6GGdF4Ufl|Twt5U&c%reu~mu( zfsJrD<-+1m{vh8j79;~hJSAYj_v-PdTW&57#g4z;>@W#)nvRpeT&1oE`n1tprN$x-2hz{?_nbJXj>d5_b1&KlYkRnC>}<259T#2{3i*a+x=eK(N0ls`j=`rE zz<=i9D~;Jj!5OLW=tgrcA`GvMA@e@DQ%iDKLxHdCHfeoH|nijO@hI^k+^@9xw$<&EHoD%o_)58sv&2`qMKRY>fxXL z_yFkx&g2ISMASz9Juu}RjZVv{%@nxL-ru;~^HkUe3JPm3bPMrP1S(qR2r&5Z z00?#mK>xvGCuG>UfXD&lhz=t%cbJ%!!|AQQ$U7T(vGnTFUIrSzc@4vPV?z6_MZSm|v}@2~7*iA;0MmW= zYoD=CiaOQDsukWkA$|#?_x^Vmm0~vk!<^<-CPtUW7LYFOo+0%(3-HCN^drxj+=wbu^Q?w8U6E7J+VT6+M;nt956P$p+mLU2S&avtQlofSHj%4?sbftl0)$pDX&LuPlmayqw@E+4! zDS>r+FXX5<2pdV|{)~H2R0Saqc>v{J8pm_W1N%KXq7Z1P##c(vzt_=9pr$DlF_x4rCD6-H`d8JUK(}IrsZw?$PdLx!4ou3E+Wz z=l%GKb{Zgo02-S-7O8pJ1a@~1)Ws@r0c^sOvLapIWwq$-+!p>21s=Y13iMad5XDU5 zFsvl}6rRJ&YlZ{({Qw>&M|gS9emVP8TuM=1GP<{7dz%9K|@d=-(ee!`i#BR4j_hQR1;c&o@T^ zCs|_@Mwun%;Qz^RB9UNpq?%!l!&@HW_yzXKxP{c4uy)vKTmMG2E?b}34nw!|_evPrC-&(3 zs->aPEz&h+pUj@2VZ#A`SDHAHtbgt#SS0yUdcbTE0ff)R7n$U~=TO`?uQ*xp#4lm( z)E$Lb9YVRz*BgGI3Hc<;V{|iduXi#0Vr$K=bNm2jgNchyjeK|Q;%NrA5Zn<$}dWbyO zJT;*imRKGkE|NUFfB?U+aI|<`@uM9mII3DcY?WPteQA`Z6f zhiMZ@Wvq9mn4P{%Y)+mp4P?a_HQ)q(0m$6%0G#?UWVn@uu|+E$E|H~bRC~IB!Cn*M zG2+-vDjE_uhPD@0=;C|Sfb>)SbZLK@{pgW6 z7|oa@&<(S}OXZdzV5;0pOttIIp9dO;bI?IqgK6&ZDLmZ*4zzk$G104qxl_Vpijp|< zfIb$N0)kcDT{1-sqlzU7ApDxiVDkV+iWl`Mr`bfGg>pt|ja1;DpVM3tY|`_DLT#CU7)xP1faqyS9CwPBRbN9t71&3LEkpr0sBJ)&aBWjS-{lJmJlQVS2(+gNEEl^_?DU?%Pax}SCmuJ>Lzx7_uUM6qGjOf zB9_^xV%AI9`qxqUhdN<7Hw%Yt5hEMsUJU0-IsK(|S11eHc!M{w-nJjP!sM54WH4_o zV!6DY9*!*PNKv8E_s7v}FFcQLHxFCRycKUu-V`}Ah&;=5S`oqCmeTFMe@`&y4!D2n zmXvA?05YRF69TR1Edo?C)+_$v3Nzwyccl;|EfI~?C5kKDH>nN{^hlo3;k>YYFG;Md z!9W=l!qO#`Dqae?#07;Fk2JO6%4qe$aUN7-jXzudgw8OS?RB&Gej$M+hsy~HE7yOn z9O1jFR0hMA+Wz%|q~|lEAtEP|Mii1O8%GW0D$=YvWn?!Y(vY8nm!G{{8QN)|_cHX^ z2S??8?F9?^n?YMw7G7EtbHkwe!(te&p?hyGCpWaol7WbdWo)47abjjrCESA_yYK)H zE%0Kl+N|RJUM$;RJ(4FO(1grJS{qq49fM-*N~b_Zy;IeJm6{eagElsF+rhEjvxAqr z$)_6Cn1cR;i*rJxv`mOjV0=OqIP0IrwyMrGB;^mmK>EpflK*Ft@qgM^LmgW^dp#>1 z%b#({pH?>M|L7_2DiAvQf^=OE`l4p(TZz5Z3gY(TV0r5D<@ut*@hMglU(e#=A9-xn zp)|p1&wmf5B1sd(9+J>+WK}YgMiF2r3TNZc^8rlOPfYdk6~_q4h>r`IB%xqe>*B|K zgDQjmWPkj2V-zwy58;g^f&E)o=Uq2U7LJ3Ha(}r1##00JV)fmxcH;=enQ)KxU za6~}%nHgSouZAdRb?Vo*z$3GRv09OQ;%^4D34#)gmV2xi#CPQ1dycTuRA{3ajlk)-sLxG z{-q0N zQ;`*rj`d)Yzs|{YQ7*6Vzm=$P5AOmkA1p6zc{ayRt*!XJrXWVF%<_cMHWKs734lmp z*|4`NC5(9%aVY%Ha#0Z1*3lncn_EYma^p4buM5+QuK$ z{YS!N+uOxO$r;ikI3RIMjH zVsaU6B&R8dP3W9}C<{tITNYv`90@J>4&p8x3+IR>WRuj}fZijEM>;nYVCC!kom-bG z^^!ij4nHVwy-jiORsEq(0jn1h4?L~d)pJ^99bi$UvvirAz{<_;UP6a7ettE*^mtyi zih%(EN?iL%+n8oY;K&F9BP15J2G*E%GaxR6dKQx$mdSHJh)id9zaxDO=q)lubZT|bDi)3dc2I!Exk84~P1vRLO_QwRI_jU2GN4Y{$ zNm$Sm;yV2mo_aEv64+^Rp5g|euiujrjl`N0;1J+j@OWj{bF)B8dQW7oM!;WOPk4wP=Ng-h7bM)=6!La1gX<_n3-mRfQ426{;$!Al zKKmtPr;mwEJq)vwA=UkyAduj(=2X^lQKt1jl9ea7B_brJlnZ%aKWNl*G>AV6Y-G+} z&aP|{HWTMFfDvip1f!>v0VaA`+?0RVXa5nuk1`Ts9v`M)!8Kh>~G zO~Yn`4b^+429E(giMu}H)F2#SO7I^k6y)asmRvlpQ6HZjibTg=Kd4fdg?#z>eSUO zey&aF)CqPLMeWLH(xb{@@=b62N%{-x;*Ok@%k!r|oA2!o`3cYEs89S&0^!}uS+yC3 zwMB@h%1snVP!b+>r|0<#&DmNcW15x9kMI>qs-e~uLR%^p7tNUTv`yzs5Xb#2cU^9eB87uYZX@Y7RSLB_J75?i@9rxSjH{-#jK8BV zc3gg#2z+zw`Sz%D4zci(%NEN3=YmSDOc^Vt)ex{m%N9=7kl!8x*X}b#r}$Yb=f5?A&DK;9h;`gi{)%$y%tDjc&s> zKRpmE?lp!{Bn8rjivTx^Fj{*65nB(?AxI=l20Kf=<_{58{G4*(UmQvMX zAiKy-+Ew*0Rl3e|C-9$22aaz&%c7NAUISc`Fw zR+k{J85coS;>vxVAQS2NlD!DnlZybWML&wJ@I4wmDS#-5;i0lomrPzh<1Q%dr`8|uJb0dw&(CJ(Ko@FH=Qs@7? zYh>uPGQs(0?W*~5$^6m2$vvd&>lCc~e!#vhus}CH#zVD{>KAPQr2(4T68R*8wKgm6Yx^GWd~idtY212Lv;s>*zMJ{HU3h z1?=gvVEuB28aStoK6*2hMUxNB63O^6VL~+NR*g{|C>V2*vUoiTAO*&@TY_F#K3H)> z$6J$1IQ4*gU7svAxu}2^#?T+*5@1#Rey&&ldttc;2u={qP?beAgNtcDkqWXdU+k4F z0Qh84pS({x1!i3kYZ9?x)#o@^eWZtg@ z)4_J2MPf0GAHi;Ajdi`B4Q`nno>JKvCW6pcL@8}S^iy&X1^!irAsr+NXQ*Otuj~|} zGEQkLcyGjG_QjqT!QS6h_P+(DqEW!13>22VolmZgr7D)fiz)FPvUpt&Y+#VeoeI$7 z7GR;)DPy4NYk@zJW(s6Z%tECA95Gu9U<;y!8wiVQAcpM~OiYMmlfkG%@*9%DGC@Y~ zYwov23dEH_gXGERn20P#w)0X6iTLY5(k8qC{kTba0Eq4uBHDMU?t2nW*e>5v%dldz zv_p25!?so^^G>#;hd<7(h||$EUYz#W>?12Or`Rj}QJEPdWRuL1Sp|<|&E=6yAU~9qaIF`QUIRYr`C`@a`T)sPdAS zgMbzn)<}zq-6@P?8jl}LHDK9jW6x!ouud}w@w_^VpNU`-K+7Dr&LJPN-!71kvEjKz zh1ir7dV3Sp-#Cg36rh}!&EN<-piti*TLP(KaNO-4b1ly?b>oeI?;XlQy@O8K@td{qtJaJN zmce5$UwF43g1(K9?Haz#3Y~~W!AK7nCsh=mqR`a)strX~yJAUy4?xN~4mDzf zy0YmHIZPK3_U+R>ug=$bO3qMq5sjlhz70J^*{b1`-yZ^0fg=E9=i{TNP!aeRo5fVe zxdEI9)k!o)2xK9by>o=<0j`2p9az*ONXF^ZLbBf=1G3zThNJ5%AR$ZTw_S4t$rh!V zce{f?Z~ns2f~;8;V!guLBt(Y1dbcJfcRdK{OrMoCa}yv z%@}nEFy_TTAx3*r1FK?izuco0Hp6>y1TGpoUww@m{}kHh;s$fW=9VPsBM)Jt)P!OE zZ1>Sq_;_a|pNx3Dpf`z3g2bq5QzanLY`Ixexp7Y&5`l3;NFetdP*GZ>W1L0J+aoVj zsOh@^Xw9QOVjKtxu7Y+Y9S(}qDOD2QG22IhY^6ysQP^Z6jRVVGl9NeXf9#p_nWhPP zk!JD;aDI#*<}&Z9;jlo%Y|>!*#8~JDg56o3ufkWm+p1x;vDxuy9oyOIdah5vUJ(X7X)Yl=4Eb;K4*{`4I@dhDn6v#MtiF#5qQhPSwtba$rMVZ&TnLXk+*R5CmHQB*r+Mt0|eP42AV@+2B{YXYey`GauS^;F6QoJZG`uY_$? z>LAP-?!+I1IO{$1U}Ma-DanpjsO3${24;uEvdq)HMDhic^9{+1qr}^f2F1CV)lu&F z$qE0Bgvvk7{dDXzNQ*OSCQpH*{WdkKn!KzdkFf@QztMt$l;K#7p{UpbtzpLPD3#$r z2kT?~f=+VSVzQ!#Z3L%wE+MA@!d1u)Bj$#&B!&@GVLqsuKUr|~+4LyDiCm>>>9?=} zVKbNck(V*g#=hKyLvYGWG3|+U(^}5VMPmI<3KTWqULr7@m=Mdvx6U;ClQ5A+_AOxY zK#_v`_+eSp18|?h`xW!Z8T$9cP3sJ~kbS?5dw7un3>_x7c65$_uIwJ`-`^b3kOn7nGR-u{!S}~C!Qq=4UgeZx2!jNhSANp-r9`0rLP$dY>jVs6- zk#bL%n!O~n3v<`o!@sd`(#s`%-yz_zuLq(rVb2yt;wdrN(lbZVgniQY3wat_zYDlB z8l@K^YbH9mI=>a*UUH(o(O}R2eEnm;4lHw+wa5<<3;h2%xn?HTHugq8le9+m|2u;w zH(~yh0PH-cL`Bobluwvf0M0AQYhej#cF9wKSu%RsM!gvaNGNa|TvV=&jq3GB z8Lsac4u=M~N`Bjt8NsCIseD@>Ccb78RZXrquLs8DsCHEKL&rmsMC{5{^7@^3Yh_xv zdtKa&`c*@r#LG(|9QWVZa3E|_IN`q%fUy71lB55kb7nR_M6L;S!~8(_!R-@85GrYq z1E>chc^(#$3YoUiOegY}!RaUt!Az-x*hTTzb2ekU#IP7D2YmZv+l~wZNCP-cT5RM* zJuoOfjNIB2$|+hz5s`vP4A(J zHOplFX=hy|ZRD`nG||4mnNC|e6J>5cNhm$b#!PxH=1ecWb1A1d?Av4KJs<6-e1NA? zY{gZ_N**U(>^~#y!r8W#M2HWJ8bX5c;^vbRh=!;H(%>NmP(?;Qzh1kR#o>#rm~TjZ z3i>#xF||!y)NKt`ZT4}i85>=? znk3|yXw`ApQ(ByCPZRtjLT_N+2F0u7M>B;<388Ko$4`;h%3=1^CJw$-Q8k<8=}y2G z(r{oVA(A7tE9+Aq22~sK8FHl%7wCH`qyQKnicO*@46^ZNButLrG-d@mEIeo(1P%g@ zN@T7N{a2A#R%9**1@u%?#grfwxSjC;iHV|kz8h#d1gyZvWxZcxLU>hNVs^S~;%zEI zu2qSDCL%M&&sit7vwCFQ#Mpg(F_rgfGhPvb4D@fYiM$lQIwde3?mfxCU_CV8h%ZGz z+&;Ps;}&Wpu<&~{?gFnaPN$PP%I=9zfb7{dA5kjQbu6|9Mrre|xGZKWlf8%f3=aEW?~Fcvq*R0mEb zotBAWm$j7;nArs@OA13`L(*!K;>L41-`R{n$3c%Gj=|R&x(#J(neqiJKty3ee>!r= zgbtS>bTAB%1y>Qw8N~;hsOc*ugclQr=Ik2wpTM@uQG{G05EeqxW-t>~o0z^(8mF#u zXu!_1dvGO9nXu{kt!Ix*^y^EaY*2 zHol{4hS8y&Ns&EtuFYKfjI>Vg=25!u7Mpku_XGTX+cza1l=Y4TC|J_&yV<=R>(*{Sn?P11y8_!Fv1{UF{}1u+#-+GCGD zmr`;rS)O@=7_Ie+ulz@vU!n(_UQ{@b$QcU!SUi$uqA7}m>t+hY_Fb-hGJcJb$B2i_Vo<-;@Y zp`eavq@wV$(@uY zEyZ@>H^9G_6}97Sy7Hg84g>}OfbpLMbSnb`J^g=$KcW=>Cs}+Q3sQR7I8-~uM|18> zR(nYYvpN?`+>S~h;Wx-M)$N^_Ng}Q6$P6^7FWNmaQZi}WaG_EId?TXTODcfb*s#k% z_ZD+g9_RrS6Yp3;5AXmZAyJyUg-gLNdfGYv1UTB~;zp{T z2`am4RuRU32J>D*x-8}A@)t!|~L5;f9)#>d)n?KM_MWx&!8;=(_sp9@iF zE6Cj`Kx{xUPi`3p@%bAYAw~dm4g-Y8R^?JJ#A-Is$ywKp0$UC0PQ7}d_tS&AJmcuj zpsBMe8J9v11iOh`UYNf+v}s6H6QV+2RO3_*L|XqVeAU){gwiT|cyY(n^liI3Tycfz$?9f0EnX6Y;hw~NzAo?g-3e|3|M=0g#7HPlCLTVB*=9yD%Sk-LGos&Nh;2KhNfltUKbshz0R)y>%)ui(E);r}tXOOC%dT%P>(cV2DG z-r7yZL$v-6eL)0Uh1!Al$P-M&9h-@}o{c~P87us7Q~xAz`jz=bJ9>?brz2~yZj(Qk z^JGH24q{5&DC1xq2~V8{J5cYL?R;~bnSlsU$f!o`dFL0c@5u`g(>|gRhWC5wcYfXc z;53<=9Bfa8W7F}_<33fA#v=g)0dJ177C+AUMIn<;WQ=#s? z-}BkEyfY*>qa*Glbj^PNyoEBSD?v4{aSpKAb)4kK-rA2JpF_a|L6162 zCyw6_=2OVaAMqU(>4ribtp_`JzcT41gg%uNOZ6DiUy=e(A-iW&nkN9n#2wB5^|STv zeJYF+71%^d9XMlQs21A>Poc-RcSoIfqgLs`A|)Yim}F0wBvep+S$(7Oyam~0jD5T> zEge3qslk7eAG!Dctf<@ep?uK4_YwDs+z?mHuV`+V(>VWFRTi~IOQ_SPVtl<_X7c{| z)%5fZffi*~j#=khUAycZ9otsYWYocNi5L1(+OIC;F8JdlOvk9QzJTcKlmLYR6fk8p zIQRYSxh=9DNLE=D05gdbK|42y>3k6cmy_wiJhRlClGj$f@f~O%)5GgaMvGlC&F!ra zu#ZFO#xdI7#P7LiR%26llMOY>6x#x+p~Y3mp|g(Qt-K(^lz!_j={>>@lZRGKs~lev zM!B~~5>Azp9Qd;{!!DZXZlRJdWo{KyXpsVT^$eHs5Fe2Kr~$)b?l4@-RnaI`o({z zIH(4-h7tdQn=IlGt3iT48LkNRG2N$OhkxJ*yy8jQl8r>T13VK`OEWJ^aNlqD%Q?%W z0}*(Ci)Vs(a}d%etGYoaf-TIICnO_202TZA?Pw}uXJ-?v;(YgMsUziDr_0L$R^nyZ zE|X4Apcgafsa}45=IU!dFH;LJv+pZ)u&mRc4E2I$Z^8v_N5}J$%{M1wN2O=0+q(lw zyT*E(U{|C`-D?y39iec&5RV{XvjQ%R0D3K3aVUhQ_Y6bjuNOgZv@Rx2W0($?C93GP z_gCf8>mN~)1KW`!mWW01!4 zsyR8V-0F4Qvt2r-+E$IlK+dT`=2c~f1=KIDT&-)dRkI2DfUM~^OR*>sA>paW*^b80 z=K6E1{mFI_2yxQg(QH__N(RO7lUp;1YWK+7h3}v?efd!KplgSA8bxrPJ;U@^R0tbQ zPM;!aQFPXIO{6bh00NST_;;bRE9o~HJSMBz4>QYCeUfj6n>^S#h z{LP=y`;0akYAZEu5uN}kT;*8oJt+F_J2R{Cg4T48Cq-vs{Ajq@VK2}bCgNoc>&Y0t zL?P1Mv~8+vN@<~E>OC|A2H@tv+$$Ef!d|`L`a!hhIsr*jgTE6Gv#T^;WuDL=t&8*r zL))~B3=m6P=Y8;S*b)s^tJ;Ej`^pX3(2I0+q=9%0_K)`b5J--9-dh9g`G+AThAMB; z41y$~1^nj?;joRLWC8Dc--z26#x*3wY7nHAka;lS0(=%h>$Onv*u{Jy$+{%KwhCQ> zlvuC^rTvDc{LkjeCjmP>0(^u|2HNIg8kzO3Y@;RHenItwSWyUs=|8b=iZ$hd);&>~ zL#MC>MKJdTCaY(@Qb*7ldDVV`4M&TS#);f9En0ft6JIUl>lJoj{J+Rbu6)Lkd^Ftl zn2UDFr|PZB;j_Uj^ye;@G4=7i#2fN1pk{`$2GF&@ZihSck0X(ux6wozZDhEhR#iMG zq-m!f;f907@8+!6C82UH=yf!0^9Mw8XkL9v{)Id&Nl%3xI2}q-xTS8u+#bc{y9;D2 zviio0=b;2UgmF7o?#cJH9^>>&{Y5BV8Gl_ZUu1b$%bFt&<~XN=F4F`k>--AMa(B>N z|74J`NH$`&&7KY;EKV!<>T_-53n;(bdv92!Nx>}tz2^@FjVzJl_+jQ!0}*>L!))SV z54D%XyX<3zL)%Y8e5?!I87}q0Onew1R-qP`41p||H-^`D-`?QT_VKCVHLX?b_{(K) zhb!WTZtLgICenX?wgq{&sL#V8|05Sm3Sy?uH4s0xWCHo}={9;V?u&Pz9rMHq#zHXT zH(|%n%7x`or#rteE*|qQ{3hq{rR3xD?i)9|K-9JM&cQMuB~Z@(E9Zqn5Iskh-`p?yj}oCisUVcv6gxRe);CrQkr*Six+59!pn6ny+k80lTxt*oz%5z ztWjJ^LSj{x=>*dsx59&U4C@lM1>-+LN$M6k>i+5SV(O6tiCJPB8F2ie$qxNt~6O$kMU(-t}}>Fti5J(^pq+@R$g+t0WzU;foof69vH_HY*Q44!fSvYKMCzpzw0#=xCrJfEE4ssg{qf7W= zmE)!qR?)6bbe7_c1C$hVA-v8S&iOLSzZLM@eCX(Ixn_E?r5Y0s9&8kSmmZ0@yW;=8 z#>e207(x=SC_`GSp+1OsC3}L&5BzD-6{8gjFGBeHzyNiQ2!p#HdQsXR`)`^=F+CIj zIj1lw75^2+Y0Cv4CAE`g1!#>~>J_c97sZATn{#w|W)(+mEm*?%WZlq54Zu^8$q(p3 zNT{)q5NDr%x2i$CJmPS`W6%6SgrS8qet~kvpgd(AcU7q2+c21QX$qRmt1mioC<_1P z&1g!3UiF@_7q(bfoAp=itS`)8sjf;5dDDCdK@3rO$<$??-g|rSc%OH03D_<&*X20u z*k&@F#FI?;SH*t>vU&5@TsB8wkeCYU<)3_cG2=EbdbwCLM)Q-KzOeEPXoPz^ulpKg-dy}iP`))fJ){W=;N)@rx3qr>6 za`e2o!$b=&tuq*9@--LfIcjP2k7lU4w%zh}9>57&C&lh|5u&_$7NFTmv~ z6dI45h`s#K^fh0+H-@Um~1Q}8*YPVfnO^PV>*pJ9d-8d(UqZ$LVT087{r+V zSII&xHaRE(!rRp0(+I^M)i!xWP>R!6K&;+N`Pq)WS zK#=Uw`#=$Ou2GSb<9td=Y&*kTbT7gF5 zc>;&_zbg2ZcJHzhygJKAC|5ZJNf_Hm+2!B_cm!}jgH)qcZAucMJ*u63piRX8rUog_1F*UtrwNRAi43fVXwXSvB z!UztbJJwk+d2F~IYzPCrGu2Tek<_2lQW-$TNxI$2_XBSsYZn<26B?VgnI5FH`aCEH6Gqf^~XnbA>%|zK;JGYh$wAz$10_oqf@Ppgy-rXQjg!g zS3mj0G@`_@`4!)jv2J+jImdGg^N-d|(iEk@fp3dH#0%d*WRw*Pv`r{!(mKc#Rocif zP-!%ng9evzF01GHVuwYfPsM3-4viZDDGBLeE_&BU_R5)6;)4x>sX;aA1R+Y%4Bg2bEgQ*V+zeL z7G++#e{{%M`QZ6J>kvJh7EG>s7eu+!Mx>8b!=;|ot~2JxuA$vHC?Ci-j|q2z@Q`tq zo#@S6gFr9#6o+66wVg|(X`<$p?4Ce{{A;L7QI}(M<8r$4jZKaTe{Fav(SrgVS;1LLt!TY3SgsQDbk$EZy#Ccy&}T(T$H`wxSQ%lEm$}z=SO3#Vg6+Atz#NS zqP2SFbH2Vw zSzcssKTiZ-w=+d?QhyZOF`@q4SBje1r1rt?uFe!xV=m5z32ztagx58t&#$|FK*Ti0 z9!v2vk1Rc^wi61+Wr!~9L_@1L18mW+yMVM>g-=1^pjveO+>690Fo{ zq`ircx3^CtXlCF@{VG@Cl_$CFI^x1w29GZQ z+Q3A*3l_BfF@#BfmRoOerAjM-vce9B44?bVrmpiq4)y3~f^)D3sQ=yLdq5`AP5eEJ z?B4+y;s13Wj!wTDjQ`m*{eSeLRf>AQdQo_vnG*H`A(WHUSEHit&;UQ_*tbH(zwSB% z^+wDXoFm;GDJqsN&Y}AX@yu->Ut>Xz)1pBR{5kZ#wzyGB%f3?r6|M>W_E1SU<2Pc#I=VkQtidZ3|A5iP{NJhye^(i<@Ah`1TsaloH8-k=b42Y5O=P* z!f^)wuxlO!N14%-iqM2n#>ihw727pa@3bQsSpJ5?T0zJ;D-aQKWL(i1I*&y(J1BDC zm+N-rAl+(e_m(OTV{uc~uncqkIh|F93$p9~xIa}2XdG4o>Tu9a-rMeepEqZkjm+xc z-7bQ@1Cn!vG8wQFa=-duXpiKCFXN7MA~JBNNgS>X9oW@PN)dVdU&}L*zaC^%5Pr#f2Owo0nYqC*O{f^;kdzy_@kHeXCxz9JR;AC zjg8GzE6owRJL6dcwz#sQjD;E-OLVe?(@gj z_Ex;qpP#wN+N!^4hlE6uv57yT5yXTm;e&s(C?po+{K7!ipvVs(fAP}0ugffjPiHa0=C13EQVq+rVt5HsweOlZELgQz>@3yDVGa`=fDt4 z{$|B;BvBLe-%g8k%bNEs!LsxwX?92UVjpJ$cSuI31ERl8Y;9!p#HE_ZDT^Me!dH(5 zmr!6Pl(EPx;KV(KJw>)TWKlYW@JOI?8*A#sF%`^q!efm=630u`0eUOcz(rr(sn9~| zPah{2wvZ2ogJT>i(}ZmmJBHPm%&>cb`i0`I*${;bCp)e1^?3Z;oljjZe8kw=cy#Gz zDpHI^@!F|?PNO~0QLFZu*NE*3M49sG>}Yp&xOs>S4-SNy%?L|1N%F`P9t_yDD)ZNo zfxI9!<2PW;w5Csaj?IJEkv;dVjB=Vwd;eI-ekxIxyLJAxpg;@}gPsP#mE`tsCSXq? zt+Z>+TN%GW9~t6|KI5z4YZ>7kj}a&ItbDm#6$?Il*8O`6ECGNEUB)$%IMA z62OUCqajUj{Bg*j!{BHF2aXCNywZrsgVGYoAWkN(#E&Q#m!~5ix2sh-eja#EQe??# z-MgFYlD|Whp+%spD3L_@8UH6=iD2Xc9NxhjdJIWMr<-M+uzFSU9f$1B-ogGp^v)U8 zDHhM#*Uv{RRLb5t(Xn*e!UX-2tb3Y>+RvTH0vWS+rx*{O5&{(iCZTwE$)VAYbK3Lq znuE4`-snUM{HWr5kYf=+k}>pc)nnf(U}=_!lZ=VwnUz~%!oEM-ZBq9%r4|E>TsyOv zjyNzgDyBf8HD9Y+v*sKn1{%j+@d{qP1N+cm9nO%VOARKw*dumF%RW`8RNY=O`hkYI zFHz`5()`nHnQ$>YJH97I^sTw8!|2nXjI-r}(2;jBy*+&dWG1QN9Szd`TxH~so$IX? zx0Cn{t#`K0cIVGa###^Ux1+ts+ui%X*2BmqG2l_-s?hozx7~|m z*}WSw&BQ)*p}v?Ec5VFC8pU8ey_;&J9VTSmSr=yB*wStsu-)CFvL~H>Ou;2o;^yeL zZ3zc)Sl$7Gxl$ssjR0oPJzKZrP#GisFF#w2cLW`gVMB-eOp{mjEMg@zd5>4msJJ=1 zfNOoWCn}CVCLt<3qZ7UjCMMfGeODI^+kt*piZhmmVCYX@B+&m4_jQ4v!&hCPzd#Bfp*Nw9WMEK#L@o$N#xNaySB zFAMLwv*zkH9HCefGcb4Pajh(t=+!az6RnV|E|F!;j?)Q`7L?UtV6e zv)MLF@|OdG!4_Y<=t|Z08g-ui2M8>m%XdF6#vNxKvCM`~=wUlt*PSc3&6~?Cyl4?y z(~Zh<$5xq=E+#D(%r&(-*yYy=jX4Ka1Lkkmc=dI4v*2JPQYgtA+F;KOgqB8atQJ?| z#9wUXUQYIAfi_n=#Kq*`s8gz|0WflJxK!Gz?_H7Njm7s(B%bW@2P*U`xsm8W4m6FZ z%JZC8n&xtxhAtZ00vXnvyl95S55y`995;*Z5o`j%AZE(wftQM8RD zi3n5el6tFyoNt#E)tXhU%f_gWMXcDfoa-5rO8UT+ditQb5%ZXI2cFo5$}nCj<5FaS z9BLa{{p;3j?d-V%O$wEe5_@G@=~tQGFweJkf8i5#UNlnQ>TN7AE}8?54TBeTKWIIg zbVT$4oUgJ*H+itdv!8q8&wvuVeb{g55;&D6Q5J&?!j3>CvdjgZ%%laUc&1&&Hvp?= zRk+94IK2H!5|y+fl#qPvmUyiLw6OjC>y2g?-IuUjmt0oFhBWohrVIUFw+FPrI0+5) zF8ti71$JW{0sneN;E3<{_gAzOo(ct!O z7@AdxSospW%YaT>6kkM0D1B+4P|Ky2Pvu`_Gj2D=-uvs$(h^t{nvmHsZ~C0vgu6#c zHf=6r@Jj63sPzLe+VNcoB2mRK*o8OT$*#k*@LK-`3l;XEr=%ZSBvt545mipY=>()l z*}}T2d#JsxSCJ`4DNgv8Imfjn-yAE|hlm%?7Q{VZ8^RwcHm#0`d`E&A7NN{dPE8^=UJGzh z04nKww41WKo4-vCz+nQ|u+NP|6`U?pILwS3#q()RD(`9x$~W*om4>R)-a25c^&hKn zVX6YO;%Ue&GqO3ptJ2Z}z*(Vs@71r)R< zp4NZMG6UGVrtN2ji37Sg;9Ua4~1(C_HQoE0zdZ=}F&8*jGIB{Ah;u&LJPz#{#csFA=>K9m=>M* zfi8Q+7*zaX^KB5jWdnY?y~)IO-NfkTKtKE7Wks2ngj1#2TVxD7;Z>9NEZUm#Y6Ny<3!q^%Qw_A z1)wE>|CPbM8H3RNd5WiWxmm7g%Wf3A_FTO@l12VfRjX0Y+@MfA?#w;=82WP!WISZ+ zKKWjpC8yf@u!4B?aastKZF&^uc#o+2s?Qy+8+=5F&rGpt$BdIxb2Ojh zWZc7Jn^Ak1Hq+q~VTL#?_{Sikx}C1A*0wZXXBD7sWA_EerFRH5%Pryif8>BpzuaOe z69bo|-+aF5_d)sJdA*CV!+%^$ek&Z7C;?c1dU%l+o+04o`BHo1e*eA*q7JI1~QK zG9;(sWmo0F`?)b>U3828<30YvUvBV^vLyUzf`iT`qkV{LJPL}4)wAiRXaoZCP=leOBmzx@r~D|C zgUOTnZRITUVE)7kEDNCy>K-&0>k6X!Km+jzv{aR6pvk(Wyt(=1Zh6YAA5piakkPQ} zdErYQ<20l_w%;lGeH)+?TIH0%SQqwEce?wG8B+wgYgSS&!}hcmFsX`@>er9Sr%-WN zd87N@!Mg`lJh`&Oj(jc|$w{&6spHN0lak{UpVR9+py2^sR&tLYeBM6*8NsYDagf=x zmBXsYx%!Xwf{2x49tw-1eu=q*6wxneoW|yFfa22ZKH?#W<0Y;r`?zZ@ggl zjg?RbbC&FEKDZ*hPKaeotT`Nkw_WovMCbgW*>oK^Tgrz7bB8K5x+qe!*gCEmU=l8H zV*(eFTQ#bPJw_g*rd@vJu|d0-_$INJ+0&9p6dA$yClUVH=s}q8vn7a-r&HwnENafnR7s?#8VQ<{ z^thD(?WvDK+kz{>E|nP3FctK8d%83B#Kc@2I%3h+Ou)neT!Vd-1`SUJr)(TYRic5k zOt@?@w<9)DV^7g@aznk0tmk#e=WDf`ld@o^1ZVM+GwGSM(Lr}Z|2Zfm+awK8DEfE~ z0I$KD?U>u9*M?_=!!XTHMP$WptQS$D2SkeQuctpaNIEwu#GY6dAN3R^5oMt`$Uinz zx)juQaiXPZtQB$x#V5&5y1`E{9N&gsUdCEG)}yZ&BMCi0&)wrmb9II&Aiy~`JhG?t zjQR1wmsyCUPiI6Id_)>7z?zPd5{;V((a7Zvc_{Q5H!!Y}??yV?r-DPTM}oVDqL+UhDGhu|Ku1|LdSM>BbwZ9cjky zu{=TuN)>0ep$Yl0Kv&M>=|=d#PDj9oP;_d*p@4#hHmDW++2)^$y|?UVxSHAMt+}!E z3f!#aU73#c){(s`rmm4epdaPj?B1u*w0-y2`0wZ^d-Kl&0Y8hM{J&6AAX?0EASNBq z%t!NkAOIl2tmCo@C$%=y8Yb9Cg-|df7f*KiYh9|=xPF%(@XI9ezuM^y7m@^S_FalE zYLo(MJZX!%U#Y$+m~gc*EJ~PCFk(QqdUE*pJ|F~B^;Y(vBSG+1a-9)MCNI-!Wo-aW zFIm0EWfn~a!5*^Q@?CAv(^KZ`BTycprzU}?CXjP7x$Yhq`5R23`w6sS@zz zDcdxt0#GXt+GpH%m5-8+O)o4MWYABUr@391@12a1@173ml`V7a1=~LalN|_?9Tcgq z2e?1~wTm6q@>#s{+xFr5-_kBC0~=Fk1Jhq4m z(iN^YRLLj*l9!g5FW+?}iqognXTHhXP&yG^>LshPMVQqt;Ni7a^PC=SM?#4+U#ViA zYi_&fb<9;x783s|i|;qwX*MyOy`abas&`{R5m^m|5hwSQML)jnG!Yg)5Rb22D!Bf( z|C9&+in#f$Ht?qBwt{Wfd0Y>h1D;WYgkQXXT^}5UR%qISm;dtvjk)Qs*0+4yd0*?$ zJP-1&>$gk(=>tD(n$wxA+>2}f`=N2?Tk4Q%?;q$hz$n$E``NC);b;EWFshW+NppaxTd#`NJK~_HQ(X2I_kr8}&Vg^pS__ z+J2!afKABo>1i(LS&5Fy2)2u{7K^dVg?|I#i-6yRg>e-M!AaZGsCu>Av075uO`&l5 zk;cy^Vy%*WgWTV>^C0Q;>tn=mCHb}Ixd3Q1-0L2${$e6NQG93g#`oV`mEH_kx0fcN z(%23PYs7*+s=J}00mBMK8R@cZri^%Ha?v4<1X`uWdi}SW^|Ngd`&S?@%LxGRYjptd z{})5){l9+*Lo0L3|8zyWw5(k>no)kGQGW9$fkxB|_RU$&l9`nwSIagM>xLy8U{VKu z65_w!W!M74%bCwDH@d!suYVL1*Ja2?+zGCS0zRUl2Ai+OxG-yeM@htaWkwVl;fdNZ zV@qcg91*WPL<$W2?nqd}a<8Na=Aw$K-uoC*Oaz1+8pVHwI{5=96j$gedfe*zXtoSF>mi|qlxn1MNfXfU zzXO&r5~YH_aW=jg*`bC12Dx|L2GBX)zVK*^Z(m~cx(OM3w2BDjrjseSB>=W;G`7Qs zsDns8Flb>5PxQWS1AM-g_`Wh?R(8BzD}L@idVC)IwBr}wFFUi2^t`9}eqP_QKAzoe zrf^=5p7pNxA2yx0y%}1xU*0yew3dw+AqXi++S(=Kspr?)^+HM>u_K+JGC4-5hL|Lhq95spSlL9-oh_7zC~|=1?y@!B|U4&MyPG zuBv-d!KD*DwH+mP@hIzS2hvp**ol1#a!SP#IfxKECi&r7RP|-TP>93a7xZxP?yRf6 zdjovZpKVKJM77lR6TP>xF5X4>diPv{6d+0HId3DkrF3lMV0)8bZYXc&0UCZH-Yp@ z6b1?Lq6DgX%}XKl!Ydp!HSE)V0SfwzKzx3EeP7<;#BJYTQ4ai_RxmChGL3t3(a8TW z>G2At5@+cT(E(|#XZ(5t=&n^PdvZ*srl)vq0 zwG)dBA4otEx`Cf|!2+#2OkiIXQz1|8__uOj)h{;B{*)iirlYG(AK;G*nW%e9b(0N= zTh8>pW78$2(u`Upr$NJvivcp1ai=uU z>OeX}AFHEusF*#e#kCeFZ1q8(Av^2yc$l#GQCn+5n9G1}m={`Xw8q@D*yQ{~1%nC#WkX zL-+tLmJ`nVL(YdXCq^E|nrL_m`;VU~uAG4cMFyh{h*x?#^vkG1bL@NV8TqPZA23xg z%+DF67`?Ktwmd8r2}&ZB2{<2JI99dRa~fz=L;9R?*I+83DIGW{BmC3~BP?W{9{P`3 zmNi&>HyQknN}Z!5H+dNy08lPPem&+tx<3gxF?`slcOLe zd_g@Z3DlUt=mat;sp61A_JyBr(Jc#f3_P^vK)OURLakVASUUJK zPo1=SZ%SrRI!DkOe;?HiG2{WhEJu^Zrq)2{!2&vL2l_BNfkw#PHLS#~HQ>1SI z1-&`D@%qa|p0~`j7r`o4A02X97`7#V6dk`c?=1YS5Q)v_A5qHv=k9UluyB7AbG*$G zFZ(yx&n3FawKFMhcw*31V!|@})idltdC@(WN9(F08yptUDU#=mDY*7&5Um<6NQpO# zEQI#KZo#1=ZLtH7W-oYic747UBIGIn!M4=<4(^JJT%Dv#wcK|jL`4n1Xqn*ZhsDv5 zVDoei9l=BB58(CqjW$-juLf94=Qy^)Hnlm?x*L|&ES}ZomxSFqCFj!4$Ze)k!CIJ` z#Xy9u&7ef?>Cu99N?#@m@nJJH$swHNj_5t!;t-Eyi7LR0CoHGKVa+OczrUfZpnQM| z-X;4?#BKh6GTZ=AwKQNP$a2$&?6zrQ={}lKZbd${}rq-)c;eGrK%Ghtn4uJK-M&w zL!6zVlGkVt693#y5fK4|P1;cVaS>nKMUd4-T--u8SLp^QqZDH-f+1>&-)+TNzKQ_3 zd^b<5G`C)X>&eg9@a~YvBfX~XY+A52?%hC*L2_s^$;HPG6m3-)0}ZFt490v;^)(c) zyuxl2MG(_!83L+G2=pIYoz9K@7;GOjWAh*N$WELQq?glS%C91daPa{J+Dw;x92RtF zcAz>sZnhyL%3_tqmUNksV9UE&$q7CNBTtBi)jDQxlxxOGsOHR|1^Z!GL~Vn?Un=W<>w%nYaOnZ~Zq#XT-bIyC~A|vg3;E;L(TsJoyXu%vL_JYS_gv zPX$u`HbiN}AlV^Sg3D&AkiZJxa<#xE3N!m^K~IAzy#hQS7_)7&03s3+>|UDBMkMA#ME1Rj})=5P60+7?ToPS93xacbc6bGbnrV)iQ z`w^Q1Og6##2NRdeAHyOQ(@7)`)1T9_LPt?HGuMc@;A^EtJ_!%y1#L*>>F z9>g%=@^lAGav*GKmC_wOCujLwK;@x9rx9L51KR;cQKF&cgl$>n5CiHZgT4f7x~;_Q zNG(ob5~!2?#-e;|Ng1ayUE_fYkr;C3lDg1X9UBqzEpJpg69D-D6r5S5(x&z80}RU4qP8!lJ7H8>?Wyy#6_~tw_uR=aGv#wW`oc2>qZ#tMUrm1^ zb6n%I+xfzN>*ET|&MXhy`Q(-V<||Vg*#>vrC7%K&h(13;7GsIsgPumpRaoLdl6Nm< zbdnvD{`ZftPRq%jL%qow4~l{iUCpHX*&~1Sa&NZ0CY=rDLizr&vD{K4m_C$ISLkV* z?X0BlKm~IkjdR6;x7uG;fM;4k%g=7UyU3UHAHd>f7)*z`J+6$_sS&GWG1JYu?JM-o z4k(+$vrb1VUbPhf+Gt1@Qdzo~D361L40}K@5-uePNA~8DoDl<$w+;P@DMhw+hAa#> zy?dvHN|~1}cX?2+#{J0tvRPbjO}ELqk?-i<-7&F$*9uuXE781*S%6;nge~@>cqM@z zlNtL4;w;iQ5z3z~V(cscPsJ8LXQmBsRnN0qu*`v)7QO2ppXPL&U)T4QLBsFMtPp1T zvioz81OptLA1^j0lQ;*e5S=X-x_R`Kpodysx{)*x!`M5^B{W7APMeARcUR&9$ctX0 z^y->^s69hk_)B+v)pGf;eb>URS8C2&VL7o_okvkG5|q4!bqy7$Mpkx!1BQ{gUH>jA z_k_9tnMua!lu`UHgpl+g970M=((nhg*qES=8f+Wb3bXU#3-mSmfH(9Pi* z{Fk3_oPi+o6;F5!Zx6^hixhwxN$K1V%uUu8oQDk#+W*fe4mw`sP42;vHDX>UG;Klt*}-h z{*|EBAq^qZjan~}cumSc$^?ieA@jW7Jn9n+pi~8%^W9k?nWP7l>OlVM{jm&s5zDz$ zXrRRC>=%)}i`nStU#~*{)2_eg^s*BU8*$##?BOcdt%k-Kq0sBp#Ruf6-;B_vX9 z*;MCOsY{BhXpdG=njj|t52wD`fPcXS!_|09d->@nRZPct$8Ku2_!LI=1MTkM7^Xst9AlVT1Kq=eMrnLM4$(+}cVpBlOr} zzH%d#=&)Zui#)R%|A);pME-H(r-oDnLj<(!o7*=uo#>@2OVV6?%i#gRA>>;*$49 zotg-e6j9;RHQR64P9V^$m<;AeWnQ1=7dD^CdxFG6*J-y&6*4xC-`L0anK0~^{Ur&v z*6nZIzhW?74AO8cFwHlI9J_Ou*xbhfcgiF*#upTJ2|xrM;Ee87uNamQ_YSI3^;mjnM0LR zV6K6*h#JMMh)~#kgS9F1p|c6S`5uO<=pJjz_TuK4c`M7BrbcCUS6U)(vuMq|yLE|4 zlNn}(fiE*uLTE+Oz|d3CK~*@1GEOq^2fdheOH!=gzvfD9m8uAoH|Te6ZC={3rO4e2l+Njz zZa1y;4r3#oXCCuRg%v(>-`f5qB&=%YqS-@zAg5z|REubs5tW!`Z+$FUl&fR`SnjJLCt=CqIz*APF8 zhq|f8yO=A8h+Z>`r0jd~q7zTLOQY>;=kN?YTpy7NRedEtvyM@RvpOhvssgCwGsz>7_PvggIYVr}9isW5Tt zHRu6_0{!+1^$`)9wTY*FZ84dOJb)3pkNg3rvF-FtQd!;z0{lGDj?0mdtaNtC$>spx zlX!yZ0G>7ci=J~+K_TPF`bhWW`pDef_6bG%az?C}vf~((r98Q>)CKQI`WrEP<^~~l z7)w9>%)?ps-Z%(xV?e3=uREuHEpC`g2w{fu1`a@bIj<6FI5p5Jp^^tnJDNkB3JTMl z^dIoxflNDs%VX$3QDn-ob=PQYQWvmPq_-3d!gLd?yTvq%t!>WkG5Ry$OzgZ~pS?RrIOG!Dz}2mjRBwuI@qT6+L4Qa!~1%)XLQzVEv^N= zuhlzHfK4pXC2`X2))t^=^cWM|I#~&SUKe$=&!a!K z!_%v@3(6eW@-LZ%w+t7Df|_m^TSzFZ zgbp(Te1-Bkm$CgU>!~H+7gryPK5c&O$CIPoH8Az+^aQO8nRtuXc$3@nd?=mG*b#EF z6mc~MuLfoKDiH{UrG`5gzw-*E-oX&WSt&IPazBU(u|pe|ox*QE$bz9)ua2HLLYU;i ziNBri(%Adt>+bCSm-)`759k)uasdJRA=jPxk1ggDP-P?cJ^-m2Y_(~E3>YFwAyomb zoSd|5eA^~j29!B~8As6_9DuYdq!af(o^4L#lZIpe02jCXyYU6J_6!YMaMSz*=)6YK zxXG3AY}dLL#kZxtQR7*s$`M+Jc_Q9wBcB51Qj6G^bDdEXYSV&?-FImndOhr7O~)xl zIeq086s88KU^ak)Rj&^tS)OX1sX;Y&y`-OYhr{u59Uaz%W1sBSiL?iE-RNDH@ zddYpI!ViAvI~R)OcIt4-y32BqaHQ8t&p+_Bj4tZzJZ!mPXLoS3T6#4kf49Hj%Zk=r>RGi_Rb40|eD-rWq7Fx%0E(d*(E7$Uzj2rS>(n%y$bxapI2vUyX+@D&a16|iU7>yHS zM8x2~QxWPVgCT7b=D%2|Nk(bA2ubx=oX5U3z$w-jvmD8q(vNw17F!yy*z_#2-C*&X zy3(_d)GTDozW#oJl~_xrw!Fw~U!C{tVPi5%FK&>Nw$^M0oToSJi z=f`cCWNtt{Y8XFFr*1{&!q2y9v_ITK>h=Ut^-mi~5gWl?y*^@6CtC#Vxzjuq28nUR( zn*0#^cXEwTOFfxOtX-kVo`vZ<20Vdq$DOPrC_;`R$u7wRRiFNGj`O2t#pOYdk5U(tK+Q75X){%ro$yghor1xZJ6Za8%P1&vbOR5O|;xbmy7FeuMP zoJ|^OoJan|+rg|xU#>#cKE^VobmhvJf-=9+gUXJ9-?6=1p%dO*KeCUu;n@+PM}!)+ zlLa)9tBqpVsJp29w6-+% z;0&+5Rx7Yy<2Xc^fn{IQqJ=DGL7SR<#_32fQNl%g+%55@L3}ZsVKThCj!w~*Z97L< z;9td0Afe{=te~ofIQJ=fF1-t8om|QCiX2l)cRT}3J~{q>3{7KMP7a&PZR(kWYH_Zucn z+u2GOSw3I<@P3u+xX2^xfqD`2C3PLZJ&?&!sW3 zhw-ykm%n%+IGk1jx7XTld2fvxTaC`vcX5C+i`aK5r5Jq_$&?~V-$>5XZZFMB21xE( zkzmY+nLj#MOfEi2cc8GZP%oCf8(vB3gVJ}mqO*cKdjxDa2L8LmQb5#szx-9MC&B-J z#0O)C|HO*l_|T;G%jsoB{x$#F)kcm-(-LnkF>!3joAL#e;e7NrfcoQ~9BoDT8zGPp zU4Q+*_C)HbH^LSS+~tk#wE4VG3K(%?#BNfxS*;)nP1b!$i{^nFds_+^Xk*+|sI0?+ z70OgZfPea(1VkPgHL4+VG1dLL=bmA@-*xDsi7_NAlQf-(le$2Pw?l! z%a)SNq~(@_PF6aB4;k$Vl3!gbt{qxRPeWUgIos}^LpfAYXC@pH+H3qNrKKbKWvmNU zT98IrmQRT)jwY`qYHN4-e7tRu@M`OR-JQ|7e|=va5rckFgY*F1o0wc`0@US_EL>3R zu`*;EwUt}Y>YMxnVMs7X9d_7Y2R2OIF{Vu?1b{buOxNX~hR;6TMjtU$)sIm{5R=XU z))=_r#IM8Z&$IWFe|`ERDM*pBshExlf3#>_Cekdd*fzm@n)y7&Ml$UZ?_6EAFbuZB zG=fOLFMj~q??@qh2lSV-M!Zrl6Jv&|mP*Cxh^CJ)jgk#}QAQFyeR0P_~|E6oGjY^suos=10F zq9H!ocJ*?irlH4HZhzqH5~b_QoBK`lJygtH#ms9t#Q7U4y6SOtBOMiQ$Pr zGX5=-xWOnRv;%|p43rZI{`AV)1%8~8EbDhrDv2*}gf+a8K+PH^Y9d0isf4zsz#hE* z1TOQ}yh|@JNj;FxtYcNC!neGW>}T5y5n5Gh?nu6ont(hzV~6i5;wnJJ&xILUnDb18 z`4;*8t-cPEl&&u97oY_qntf|isqsfqS^ZbXCIN;Tu z$?SJ^KT@yxub+Bhf*}YoHg||Kkz#JbtGb41h+Z~I9O2S`u|7QJFr0ub3ct@RJ~khLkQJ~@9#_RD z(PebxbrMay5g2YtMzspdKiOg8t5C8y#KVuUACGqq>HI%8HqqC#IKm34A z8eVoA&1m2JIRRCS;yR!yxvEag7@9@0k~O$5bo$OjqBH_d&d!Iya3`4OH0xIsQg1G|l=>xNAbk8z z>e#K^>DwaN7w*3qvpJbc1rc$c)AS;_>$(1YHYF=$c_b{E#v) zkO_%5dO5zOQlutJ&~@Zh_f*QW6Yq4pAYb#`l8H>?<;dwmitm)9W`|wJ-}d=3p~v59 zykEq4fqbv72SggSTe^t2(RR%GZ=@mC&Ad3Gc8(AlCjJbAbE zzC7XI=vVfp(=QjhFlIsa>s2Q=w7e>+jt$MH>+0-n*x5--^=GIZTSwXbK*TvS1FN;-JY7 z8zuiFBL=fH&aD&zQfi?_PcWZ`5#Uo(cJL(Z3CN$tX`jsTIg~~HHW0*-)7(x2%RmQd zzHMv}88o-Bd(JKrFjoo&B2WKXI5TZ9$l3XBGr%9NY0Q+(q>Bf&jxe;!VLhm!(D%$w zi+F(bdwLn;PY3L-byD2IgQjnk=q=B-p`fN@VJC7r4)%&cnuQkbL8*#Fu2tMyc$Wjz z51K~HtLHt6XaI~As~N_GMYn9Kn{vyz;bR9A!{Vf?*uwzJP^J%-|7`PMg@DO)FlvPP z)6g85p+@aA5Oabag?T2*0Pf4iJYz@GAnK^i-EvS27f~SJ*9Oroy46p$dUc%Kr5siL z$VZ1mFKeDDOn0Q+mTjz5FMFiOB73tNDR=+VRk%c8!hxwy@YypmcL4&K63JIVMkp-v zuBv5>3zOx;_)vf4s9tEo1dCLAIt;>xggd6c%o{Z_NsqS=$2n!M^ki~0+ohqqrpXI- z2erfcO@EdVa^P^>a_|l&DU|^fF7b;75p#oNedPV34iFNQeN8>__xRStB+{$B@O@Aj zJ66APgkm78zOz!_1ORdE!={^m4%gauT!QN?p`**F(y97A2On{b7Nmnx+VXDZtdh44 z#&L^HDk%>kYY@(CNvE!WnDJP@7$V{)!5(E>P%Kn3@|!WprD(tzgoXiQV30g;KCeg; zXAgy!dN(RsQF%QyM+|OCK*FJJF-=EX#fX3Kc*i+wwxsiup-LH;5NZk@po$qzU}Gqk zUYZjb1Q))U$vWrLoWzl6V8ky>%)MeaK;X8}mKY(5vU_vR9&H-h1>Mx1nyaV#A0Bd; z;u-c2Oi)R|M%oWmW#rA7S4Aa3^YLPls-{o#c)l@qnW-e-IE}VuM<=c4;9}|Ckx!|> zu6WrQEGQ$;5ohm?HJ0G z?yBV&zlXNUscm0BNQH7X+)=9k2(z5#7oR#xC@3M#CFLAh`Jta@S+QmiGpKwPh>^?{ z&8$(IauythzEINsANoTPT?s{s{sC=8z=Sc4ny%gs$@#_1hAIck%0v_60?Y{$lUf(T zu)eIA#e~DCKq4ddP@jv|7a6^3>5Bv4@%S%YF?p#>+zMH zPf{!4fXmq$@$Jw(K4RAu3?l-THpqELMPj&7I)kKTPkqe7{|Yy3&QSS8dVz%adiu}3 zhX@$E@vXiB?*w(uWsI# znnVc~hqcm>hzcq>?_a;Cl-{6iD-n`-f?ZKnvy~8t?3dT%3cM~i%DsR1s|j~nbg5Y* zweGAjJsHIlAp`yl8M<&*prNU@S1JBpgt8I$pjEG6#A!Smz4W%^TsGn-*-yo6J7~Dw zyoaaI;J!k9!)*MmdImP!@S7Xap;_($@Fl<|_S}ZmQZ4DOvOXp1`o!~s63X759?yL- zczV67gt8@X&9uGvfHzXLPXB1jgg5t{+ZjRVLQ7G*Xp__UdQH{<9fR7sQ)L`&WvVt0 zi=u5H)W&MFH+2|1i-)|&i6gzV!770##~bcrN&%Dz-*Ny}#rtl|^w*J6735Nj9E=Zw z62hIlpfC*-VJ9Zw!xT(y2=Iv1@Y4hCE*-dzhV}p6c?uHG` z4N*GRm;l!z6A|^J=Kih?!FW~ zA9HiR=|k@l8*P6au3J9%N6a=BgXknbh!ba#?Jph;6 ztgGloD@f6wRjf+Yh>#(lxj(Onc>!+A6wPP2rlkp54!yHRMo@+@0ak@{r7)oL8bt-g z@uX`}59yXtB(ma<)3@ub$U8=Ej=H2>a>Z7vKtc(2SihnRDr+fX~7^X??ABdE*0ZOvSce6Z%{OigGOJOd~T zkYZ5TUzzTZVZ0f70VB0mD0>eq^hUF{vo-WnE{kK(xCdZeQ#qjUjT>?dT3q3q&is{LsOa_*v@hDippRaIb zOAY~aJQ&N98SHjwOwk9Ys`^`DSnAY~paWDktE$W*#}s6K%9vZVowKC&!1*Mh)|xgr z8?k3}>Y|bHVyT@O&0D9%fX>eq-tip%mR-U&PqaQ;+{MqW-nL+LOlOn_F6J;&6M;wNM{6j>mdn^Kj9udXPN_uWm+uobJUc0@# zJ38*S{%R+84>REQKzk|RNX*>5i)a6`EO&;c%0N_&j%WB%j4pdG%xun5MrH zuVV*0I81;h`heA^pC1|vnrq?sj-<34F_EBNx=D=1=>f5QhYaZg~+x{2KoHlFiVOoQ6w3 z{9)c4I&w;^*7eiok&}v>v-|f|v`$BVFZ3f6kNu2o`WG1QzpkSH>tC)?)v;S=K=GNW zX)S<{=Ef?0W>GGe*jf&8%~0l&QBrFob1)aKEm$qBlK=j=#1KS+3AFVMczxe=>WPj* zxUZ(nnLG;=5Oo@J3>AS(0aE;26Y|`NaAN#JfPzTOSrEr2VVX9Is(mU)EQ0zWIxrH* zJoC_>C|MSxN3PUC`dN&M2Y@5+98abVOCU*tTVn0L@V)xI>yf*%lCwiUx|Fih2q_qE8b*b-X@VX}Dq7Kc9*b zLxq@W&t;UcoVOd!TS1U9A=ba9WN||rDCLx_szLA+&)B%r#7N9F8AuK`7nE%O3mQE! zGln+Z?mIn)Il~#LIDi?S`wX*Jja(m<(9AO1315wBj-aCmqng2@Qr&h2om1C{%~cZx zLn3KLXo;w8*~%3Sy*D!ZnG~5f_#iV_shC6| zCh{X|$f^W73riAQ?W73Ng|(zUeo=bn`rGYhe8v}SXl@iSv---fN{-04;8$`KpTYmjQvI!Yey+BK(P=bfa%r>ZzE+{JjYf>4S%iqB@DC{9by z#jy?{+P{ptBB~n!V6bX+YXVTPYuQ$pN&i0?)G9y$_L45qo2;78HZqy!uQ=iHpyd^PYmBemBvJK&F z*6Vi{CWooczu?()pJnsB!Vs4jI?{;rbKm80e;S4>SG>(Q+2IE3lIBgelwj`ox+R+uT#nvf10-h(k zL{fd4A^EJU)Vz9kKX_I2Ief;^yiHc;xSq7V47hPQx16;8o`tNIlP_ zM++(?Oye>9piCXhOzBAAxyvN8DtPFWg(IUxh+OJ|dMmox%|@Yy^@*nxh(WQPKBM zyE(Td_uU!!+wbP^<(Iy8`@4oG^mvpCI$MG~QR zO^}>YSY+7uW2VOS2O|tBE?XpAbPXGt8JdbN3c#tZySmDo4>0k>3xa%OqEJ15WfR*^jokJ^P*@rV?{HpYYg>5 zA{y?6I}`!jYu8UuDC>Q|r_Gn|D=^ac@zLxXl*hobReRcFR=qSt^js_sRA^XbzGDtn zyTarNbD~bPPJ17#t5Zwn=}>8SUi}2saS!Fs10!`+yJ;ohT?8sMF{=k^d@?*Bq!SD2huYHdn>L*CegR zXPtPpw`$X(hGDP;eQ5$R9}x-vF|ugGFDnzqeA@_D^_bL^DMZ5v2p+oBFZ zt>>+4+K&-ynlP0pn;#Jz!86k_yndA{h_kih_4z}pbhSh0P@lydeJZb_>+>(eFE|F$Tf!$Q8!fWf~PNB-}X(~y{xZ7 zM%lCv6pLY7-9!4L5UA#n?rHLM*H8_og6c*tu5%MXYb<@q^7NmQM5 zqeu{}%5_n`t6Fd&ov#FtC9vSnTt=>Kn_EUn_RI))*P&Q;^UE>p9nFarDaL$aK%r@- zsmMz0m!5=t3K#mq7K4;8dW1AX#Ox% zqLBaN3kNd#4borTupsKia6o(>GP`-ErfaqqC(rP zhM>{Y=RQ4WRf*toOg|f%S#yqh;aam=psb{z7T)g`+GJCiX#(sEzingvCbg|Gs3vY& zzkwd<G-NsI(}LeRHwuS!1w4L%=MW8?2fDW*{#}%+)oj){rGt ze(xXRkyk`eR|iid@?muLsy2D>p&x37vkToqAf(B)+n07wKU9d2fZB~H`&ba!7ruE^;Xb_B)J1ja-C zc2`J^?ubKxW_oM?@NlXEAklQJ>tb`C=~$aEL0-wHs~sz1p%6I)SNIJw$Us*^%LXq8 z%4wRX?}_(KdNtMCV%0TH)pOX^FtHk^15M8iUe3q!t5~BeoVGvUeBN^NpFT7haNoF+ z=(=i;^%=LO5FJ?AHO2)KO6V(S29S&14_BvRZvAHf$LJfQyCozRbcjg~``6$kLyZD) zgka*&7B$T&+21xo1f>Jl$Egu&ShvIRphAieYhSk9h^}r>sKR87;DMAM3>5vp>emrc zXElQTJ+R-Rb~SiP{=oQCx&|uXKex78CG{j~z`LuLW=8hGZM4XC?KPvA!~Pb~yZGpT zz~hdJF)uvRzo!Gu?o>m`$r@3OuCi9equ-URSQ11B^I};a{W>yTSr)7b-%-sn-zH@V z{X6CUm1m0>+ul98@@)m6In9u|vK2xc`Mx`2(Dj~aYu|g@1>el(?y24_4y0t;eeKck zwF0{lXlSezj{^T$GK)@b@RozO_?_GO7ecY#ZU z=98?*=omMc`-7PS6>jK4Y8KkKTFjG{xn~Xi!~CMrOtRjjkY#Pj$H8p39a&G4;a$8# zS=pv#@$6Z>23Kk*FJ)=-x1!ZWkhYKOoTJjB3&~wshG{X2CQ|ng3V*4Auws+-DbmhT zx`QEy$)>-$U8942jF_K0bdMOYLc{w8x1Gpek$_Oz-`?FH7t!8-UERK^+rHmVQbNXz zk*-k+`B=+c1uJ6QdrIT*6C*YtcL!d9AP^-v1RW=i@&h%gM{=|w|3b>O^jD2;&ZZ*| zv)i(IF~Hg-eHvxJvU{xSKyDi`c&p^@?_ zSa9~jZZeW}Dp45_o*`uhQ$SZ$YAJ*evc|vlqQRE8p@zV+CTDab$HQQBkUCt1a4tISI|SIo?9u?@v8ILj-ZwH08-+L$j*XBEfC%x1mOYzlLHR zE3U*mj}Ez#7Wz>aZbrN_I3FPJoVH^H?js5XB)OYK?q89NdSe(?-XoCAnIrffP9rr| z-=959GzL;a=8{!v)VM9|ds2Hi^h^HI zeA3cC6yp8HxW+1I|Cwf$2R(_E2UQ`$>(0O_rV(L>{Uj$;qz*A1O%h{=lG!K6oQkC< z&vy0EHM@^DX*AJ*X^dZ=I5*1JaNt)6ywtYa!Nb+?;5m#a31l$1m^Aj7DR;VUS|fto zBHMiHhjER>sN$V5UL(jh7PH{*2=2Ofnpd+(O0YKXqw&qqc^Yb5LFz`kC<=Az|GRcK zY&SyH+q%Bvqg70A%AfS8!zyM2>x1j6-~kzF0geR8Df5MX`e<7~OwFRh>=O2av(aN? z`Vc|@8$`=aa_d8HC7Pl}2u0@sDM7~Y+a-aTGYia?4o3!D3?gtKr_X+U2f(C#2beFK z7j-R*=YESV{3T`9O^-GxI>1U)YTYcRv9NK#0z&>nzbpcc#LJ8P3)E0BN zl8XIL8^YgE^qAJ~kRGro*(f4!WLN6`MmyPG`Vy9jzsi#82O0kts!kV-7XRC# zK7Tlw>mD+p-(AalV0?XPpj|3e(j+&npCNCP8mzJ(ZzdE;og8=(0`&0pzISlWzAv0n z63KMOO~wwv6kq?5grJH*HSYX5%6NbD*q<=O|36&*k0;{)r^f0;_d)eDzy#lTg!_k3 z!65Qbz5?xAcp_6eOcc($6qV)M*{zg^wt?LHN$?e+T#oA4IvYHq1OmGX_0n!x?0_5) zmj?bq55_yWJJYyFk+}&$xQitVsz}p>1DI7udgGK2IOFHk0xnL^_0v4q$I>EpZtg5Y zuNXpd9E}OZ<|(fzg;G({%vq-#d8Vz#eWD4Hwu{&cHDxNGbt`RhAhYB}ZPp2z+BNjN zea4Wo zfc|DOf91MgCxBSQA@g>19*^h}fcaYG4{mpX-a${TvK#I>|1`qLjq$a+#p{%WR^HK4Ev*A2T@7RYv1_Bs>OFsl9|X{H>xrQTTt>Ced;X^H(E50$ zNZi(uC0yNv>EMoNohF}(V{KR?n*EXzBsODvZ^hUCiBr^n39O@fu&nB~4Q8H?yc87e zGo2#AvdOy%Un^hf=yc+-Up&vX`0IKVIyf#N+g3d|a~>MVg&Z5jKES%wAePQ|@kzoz zYO4uMxl6CgxuUu{ieTA@bnWzZPN(d(%>Acmv)2@gKlr1UPX0Gzt(Bd#wZ%U?Az43q z>5p;F=cPuUB{xxi>9n>4)1`sX>SeYOI38`F7LJBiXD!e|k)W(qApW~sN-~*JD>~$@ zuX*&t;kxZau`p4#x=oYK8?x|tIDjyT?Qo?%`Il`=z9VutTf+D$Y#N$kEc5L}+UzDAob<8uQR6t$B$}CE{t-vInzdbISePX(mfdSK2We2N07gpX-rp?T? zc$2pL1RcLM?}4h{fDVnBC`nq9y=P^imdK>QJ>d6_M-=MU0(Vcq6?cpf*olj3y$KC` zAKrgvF3a?jdAk|Qeid7ZIR)$_e<3fMh!HD!12?lms@wNOctVRaCM(Aa{?n98dFxRY z)yvg|IDl)`Xq>dD)g1g$#a0ya{7d8r+0gCpj4LV0EB~op<*CvI&66UEHj@tU@m#Ts z_gz0WAK!H)lr15gOUhKFh%&%GFYWZAse|Q_kPv}16<_(@fX3`Q6%4+ z`pscht#Qi=lf0k^2C^i_up)H*I4>d`=2hqm{zk4GRtqEfHe1GXHFrrD{Iwi!i(-H= z+vu=cZq)pE>C5^0T?9Mw5ZI<<3U4PVMGQvjf|Wd5Q#&IlL(N&{ZN1HUCQ)9WKj_>P zzXhlQB$&35a^d@#!Rj_kZj2)goAmwb)y4BsOJFCD_Q+993i&cU-7p*7$Q8X=NJR(L z$IzvID;M7Z=9H;f@$qMhoZ96;N6**agJbqfm?|uhnJ^-9K}Pz4Ty<{0nW7-#qRb^w zY~jAcKPM2vv#?r{UAy%{Nb9IYC@LL+jN+0h2-5(W_*iR?VcZ-G?RC_}`V+OM{mdE7 z1~I&v5C$Zl$Z+^5)ow=auVdSg#_*=mf!&@`Aqe(vg@5y=5S1@^FJ_9JLlI&>nXgBZ z$SBn`x9psa8c48Tw$*T>d@Cw^AHS(l^XT2Q-@7;O?3IrTrZ7v|K~T8h%H{TE?v$(x zCi|3M*K<#VUYbh+f)bv+GsPQ9wC-VofIZxxHY2`40LhQ_Vf1jmIc{^EPKqZQ?tek* z3O~<0FovU))7dP*BV8K0iGA*68J*XVuVPT&me(_Ms{d)2BSg7H+S%N)-TVFTUbBS)tV7E554wN`g?2>pJ0qZ zOXd^8S-m3)viM3D&anYM4B6j_svB_#q)<4MwK#AfS8DimJ`!idmCtT#xK1)+Z)@L4ff}apvLxw{_nR^HswMP_RpN^Hep36WR`h2}LRs_Z|h#>{*W}-`JKqZQ)FJpEzT(vo;(iz?& zs=VA@wzs4xisoR|TqdjtxNT`L6fJ~oHSrW8MUv)V$D}(v68G^!&0W_y>M^>}d!-l} z?mng+z<1p1P@(cEs7P9xE0zx8Ei^rG5qWCH;OLV&$bNP$F`=Heu1iY;@ zfSqW=S}34xoZ#pDc?ygGAu<`@@jBZI-h;t_O0ccp>IzS4n9~a!9|W6hVJ8n9{vgbE zvn8^Tkfb<9LRW&+AzmpF5K5@F{FYMQDGOe z-pGTDC2*77e_KCCyM4c&9zOPw(MnV*!fk^e{St$oW|qtziy!zm*q_nF$g&WD;dJ>b zNb`@vp5l1O9q{B*jx#7PZxWb;jgXp<%4K6NpQXdQfTuJ11N~-rNXJ~2kv!JGaA{rS#1 z@fvV0sdI%2HEV@()pPsj?6Z1PGO*A->?#EJM)ms%Hyb`73CUV{h@-f1>dW!lRAk?a zPwRnW+n!Wlyc!nus4}Qt_N1i5tMq?=6qF6F!tZkgrwM2 z(PP6Xok?8UYh0Rgbhx~XA@W7hQPM4`R}~*Rc&Q^YEh0I+BhR00-$+tw5z3)In)&H!R!K!muQPK|4}yY9(C~bdwM>9w}Y|$ zIHdWLRz=dlhW?CBDpkiq&GcTbrgOrlb!^%9Ao^RuD)HoSr*F9wj-);v{MhMsN+~RH zD(4DirYwpD4`u>d3LU^KeJTTqapEmN2wO|~D{ z!b-5dF{hoakX-Koz;kmpyYn}4RI~3ndPf^1%SJ}5Ls_+mbvGzo5D$!XJ9V}Bpu0)x zU=GcOI2ThQj400b5${ob)%m5oN{r<;yj{{gOA}Z#(bgDyqvNb{&x|D%$yazZrWNzk z;EjZP<#-8}@noaZB^L+C2-z3r_A-bomYilLHPZnnpBV7gHOwczrK2&D?8`C;nLVfL zIt&kP*)^!QzesgFYiD4xw*KvWcLO%($3B~;)}E^H%r`Q|KAl0^{}FUjb<5@9 z`Z4KB(*Kt|WdD<7_}N2Onj1fy%YTf{LAcAWpR25k;9IKl(rPsSeVxDBD0d|oOS*0< zlmH|_dQtrLG9B9<5r3rJd^Q9hCgi{|kj88ltQAMm4=m&9_26|-L+Nz zBfaZeK$Z4i!DAJOEImAONM0oIcFwxDz@(ViJHExzS#hg(wLrRk4rm5|*;avs>MpGKnJx$)Tf>3OJxo-b`Rb)L+*Xwd{4D{g z$Y2zN*??mq(QcX%3d~SLC};YYscPO)=I-QMJ_?O66{LH2Z>^)OM`+ugV2stO_hQjx z$2h{OPq{G+;G|5Q`U_-XS1NDISK3^3d_eg8QSs|1) z8qCw{@<0ZqG%))}e;0ItUQUnqih-2xn^t4gDV&c6YK`Z2eG1rK?K|}2+hNErfaclUp)3s zmKL>oZL#<+;m02`KK$2k-L@^$E$R?x79vD|3jt>T$uoLiAx5;(sKp^PadA_sV!~&P z6B0>n8%A8*|KF|C{g)uOHEpRmqKH#EQt!E-aQo?zqopqC@R~~&qwP;rp zZx{D?w}yaX7bbfZ4q2-P0enxNk3-#%vQ+=wrh?+q`dG$-knZVLG#B5ClGEXpmmk-C zhTOqJ_T?ksd4OC|n4RS@%$PlPxZ%sYNClm#eOkx2HcO%WJ47npXmO7b`*B4Pc#O-j z1hf-FNwZL`%1|W`Fzub(#8kUM`fB6kf32yy$^&*j6X2TE*=k%y`fhf*Z%7-^%xYTv zE6RsLJxUT+-QNauIaZgz?^x?p_BT!DlKdeg{BOZCl|P=M7zY6>a1u>Ri~?tS5!JxI z@q|+U0&cBrs}sxJG`CQ`3G@P{#K&Ld%~vH&M&d4#C~j14M82I1g=n*3o(%EL@X?p} zIV_(otb43pHJn@7OxtoHwM-nzaGpF{&$F`E%s>N+i3`+7olHsAdGsLs-DYUAgYrE7 zP}qOfS`Z)V&~2TJiFfL`b^ zUnh}U#MTjs+(5KWe?fx7a*sI{y-#&bYqobJb_ltH?3Cqu7MJ-FYNao;piLM6aDqj< z^&%*`wJn&!cC~ZszG*fVo*W*os zU2)I;nPS{YGqibl86VHQio{HTJE5kL_}DW%p9XsZ0`Dho_qsPLi0hkn3#N z1_*=lyvl9ViMpXAgesB5GLNIEDF(uNXZvjJyR3OY)Slm*H%;G{E^YTALw8g}3(1&* zwL`dy#jgD>%OjSAmSZoO6@SymwY`wGmIc20QU$2T?A!;dL4$VadVNNb<6N-}U0ufuGy8jW#_NmHVTF+FgtG>yb+?S%_`eziKsR(GSm&njqi1*y{6wk&5H!EhIb*luN@9R-dS(I;R)AvovG|#>}-KD;swM(-&8t>5xuiv=SN?jAvz=zw`JWVMa@E#l&0;Vs5-E1WAq!$jwf*`K6ZAeC%6e%giz+VHTHK`Y z`yI@a3R+Z}?!mN|zoz#*u0H~IsM z=t+W^uvLQ;ly(auUa@dexKk4qo%4F(qst_Bq;g`Fh;WS&-1Te;&JRUfYyyY?X-6yG zd9u*vw&9R*2P}Xn;?iYBiWEUQ8oyYMHqK86Pc#++KF-u$9r3e)4>F>TOzFG$@U?)f zRoZxZWm&j1)i!%U>d{l5>Zn{S-Y~@d$mGgc&!Oa2U5f;-vJ-ifJq*D_v<=L(4c=HY zXnq~qKw*BuIW|SMgcMcH^S04e<95Mv!3@>s*HvWpjN73U@h^ZB)qwylh_-QF@i2YT z%>Zo*D3(JcWiaMR1(VgJDcb9l3t}JB5K)H)-vp2KKCxT3&<;?yZQPL{Aj4jvD}QW| zBHZ5)StyT+Hb6y7v4?e{7gbvmrlJ|<>C?pxXC&8VTZbDW>=FkG9ZLK@>|vF{EwdiT zV|D(VR5c-!XQWiv8af#;X?6#Ux1C!v65>C6n`M^G!x+_7mjrxXc-i3e?nzbh#5dsK zw}&+;mM%WO%BBHmIk|4xA9^3q9d|^Cs@#{+Of6SM)Q+~Qh27d2+qK;JpT)I!l}!)A z8;W_zpFP!TPVhVsX?nt3&QSCmQ-Y|1y&xrAoUqd1cW3V13X(v*`Z{96N|rGJc^HB|hOm+%`> zhW}M=b(?c+!L52I|b%V+o%P(QL?m1J1`H|*?`*DW+g@-}$Q zzw00*+-n_(KgxVB@qg(`{Op1MX|Wz@{68bXv-SGgKxqo1HvoXa3hjUAna>$&DdcXI zR3n&&HHVEBiYm%P5`6o(6+TJMtIoR)u=&K2CcN}_gt(0?ICJrOeIN5`s1Y%1hnHvw zH2Jm{kZmB@lo}X|HPE74JFo?{nOTv#8OM!f7KP7cQuezQDF=53_V5w(+1Y-rt)(4_ z_C{Kocr35{PNiinRIWEJUrnGz_2^?+1LP-- zjXv#ScpR?ZBcd+PjDW54WZ)&&YdxmcwQmr^+MP)x>kXtZrP|ehCS8aS#r%kEV3km! zYn50KZZPB}PeTdbO8v9$eg?%N*_kvAi8P4@>m##<)c=~#=1X-N9Y+et{W4tj^|Z?k z*?a|yE%MvLIW=OD(@;WpLGYs>nUAMkk)^|sjCK`WwKvi1iy;Cw;3Uz&L$ijC29Slh zEKW@t1Z^AchgYxyd&6Q@D=Ko}*bCrI>?)06{B+h}1`7U0P=zbWT&|HmCW+CpjD~}j ze^wkzt%-wR=yJXLr)JC`sy0%x!Gb$*h4>!IJ`W<2qqiXYx{Q6s^5F0{Tvkit`RQbM zHpj)^-@ORLA@$LOMliIxq72ThIDNIx`DRHm2!_&Tneto0R7i$AK-l;yF6*UTxR@#% zYOe#R^pGl&y6m2o0S>?aHoo4CS<5=A2N4cePgt)_SCvw%pYcYMwvv!-=QI7J7E=cz zVE5NDND1Ugr*WeFw-Uqq*y0V*=EI=lrQbZc$d!Di6xiK0rCeSO7C)W3jz9-OCeQhy zX>y4%flQ~X$bWH75M&RUq%4*1V%F24egj=Q$IPp9RiJZVnob7TLX(EfO{;gOt_#mQ z^~z65Sy{7y2pkkFCvvDYHRJA8lS?@wDjnJe;??K}Xvf6jz(z#R%y>i-@G%I2vYG>( zIjEssg2ASpT7nu7Fdqg~s*5uA5)J`&9)@4l6pcfMuxRV^){(o0>98MwO2K68gL*9M z;p+Oc#pk=A8XF}CQqGJGO_JLr`;IO`r=QS5B&_kWFCdWW_PQHMrLvVaS&XN!&>#m> zA~99(n~0Cd-fPX)8xzTSNB2bqY3>9ENV7lTj~qbmKhQa>taR=F7|@+~8>;Q2TbVUM zJeUuLRDV^kYCx;m7Sv;4~ySJ;c3_eZ7l>#DcjElu-5&yoS`dF@Ab; zHh$>8Ba!osY|w6B-zM+c1h>4!3j89!q6|uDp+Esx3isW|7Q|R;@1K3 zh_MYX7Keo9sHk{Rza|`_W)=h zfrI?Oasa@AkPhrf0Bs!TYLo>t%=kd#x@O_lDPqJb^Z5+c_qtsIIO;W3o8cRoev3&ar&t-(;yH}A_L;30#{xvy=4Fu&gw?-4Ke*rdC z{9Z^p2)m%5l}5uDWa74xH2e<=H;_On;<~v$op=82Vc|_AmnotgjVMOREe5FwC!1}5 zTx>`#UlqcYKyKtOUi#goX)6S2(R-kps%Hk2>lOkYFE;Dp@8@C}-`_NU!XqtU;28P) z0aleyMb$TuMzYzj=j;n{6Fkxj-73@2et$uBfC2SegS_^XOyGr)7uZWH?Xq45Z1`F{ zl!uw@T8515HP6o0T?~}j%Y{c7%u0I1LrM2uw>RNw%HyigJ6*NBn?U1pZg=*Z(yK+ zYMZxEHtR0)nvKUpPO>{kX_W#u*j?|8p5xVL`W63KqnESMlPfq|7q09puGd$NJzQIW z#;2BnrceiT6Uhm-Xh^)1w-FcZV?AC{KN$F!f5?(7j+L#t{pI=s{#}s9yPvkV($h0X zFEmH^B7E`}zw_}wnhGo^`BkGojwKtR|H6;Aw{~*)r-G%e<#f=D>bp|&>j+#*LnmF$ zEwgN8*_P*^NoRxGf_%;KhjOT{R45qxTrTnDH3JIJmqBu4oo=HzMY0Ua!Lf3Kg9Eg2 z^WvrETTk=12nCy8i3vG^1-V6S;iE?(uK;OAp&JQurU7{9 z*EbY&by8$btRh*$0S+1+5=KNt7BIX97Ic!C$S!rdE#;?M=>0?#h! zu=)sSc|=XwymB+Y9&d!!%`kU$2~Tpoqi4=3!Qa8*``V@i=9-6w-6Chd)Z!P5V*hG` z`Q{nRg?5t$mBSGq7RIp5(sier`65;J_P$b2z#oS>-qQDE?vDI|%mP)tcS0;yjDuo6 zJ5hWcl@7Q$6vIR*42#OCSqw}S^8pl@{StgNDD_u9^G?7!KRLKdAJ7(lo%N>TV-y%C;Z8@FbUHVIa~&dl#|va=g-_Hhm1Nr`92H_fv(S%x*)W_ zCAd((RwJZk+^By??SutY>B>F#0cNRA$gQ6Emk8F#8MER*S#RBe7yx~nI68~$o7EWy z;Cp!;GZnh8pSpuIpd($bT{`rq(DD1MjQxM%@e`V62MTLZvhCjfSbX^dCTgulKdaK z-myEgs9m>>ZQHi(RBR_tY}>X|v29jt+pO3&D|WK;u6@pjwc6T$VYX|IImeCOqjJ^j zS_8Q5U0e#p{mvy|Rl{S@dp-5WL%zl9hd2#Rz8Gc6cAHkKUh)R+Y%Q2u+U42uE5@)~SPA9cdR`qyeQs_3JmFGL>75Oc$H zJ>^YTKIIJ~at~}&&p5IpB%DrXoJ3B;CMPE}$kHyKXQG9le#wz<_$V#50x_u04&rn| zKvuyb;n~EqdXG~#6uTigC))7PRMQ{KKED_J5oSj0@i#)5^d$--g!y5COt}exD>TQ= zi{?#`?DL$==kF{5H}aqH`(~L(oZz^jR-11()G@wvwqqdk;E}XWv;`h4<67$=cwjV? z)iX{ixGX2@Wu05$=3oI>Q=f^Kq(}0HA_S{HL{DB+i^}`Mjna2+)Z<+(2_ZJZgYWw( zle#C#qMub?R`;zn^$~eW%40bMdtmX3aeqsZ`Fq=~fx_i>n8;`X9ik{8l@K1u%EdK} zBq@2tMpg#wf zla~n`*4woPU%BSz=o#cy>YkSM8u6op*Es`93aUv{S^7~F7{P~s0Zi(c+})>AubXyO zz{4au(^P&H#VQ|EQ&_|;*-J!&Lmn}1+zBwN|H(^pr(dMj(#3SI%D@|7Erw4iPn=g_)c zo^PNM3+{3TB_6k1r}ceu`||Gp74P!`(-`y#ZdgXmeNE;eA)&(y7$dV;L@^iT;T}sw zXXRPqYpX9?d8jipCb6?zGrYzv&uOIkQ93FJcgPm{XJw+aNF3CcuR?z=Pr#{+S#9Pw z8Uu8}s0)TwqrD@5ui71{>@OP}xTuR#o_}v1k)4d{GX9I2hTO#q@I<}^3EPd>fCQI1 z3I@A~zy<5;A&*w__jPqf@Q&pYGBA9sDR$!0e%Icmjxt`&SuaI91)`J(pj*LYKxgO` zY(akBw&_(^yt5(%-EZEWh#`+Id%j4hLGRP(wB%Z@mW7eu{lW(U!jWf>6U$XAoO zjvj1nIn3{IK`?LGg>UyQy;C$pKz@RG742Z%eZW;;7fP;Gmb=pEb+DS|B5Y=e|6qR< z2%d6@a(94Q#?F}r92<`~JK<|~6)zCekF&S&I#yM1LI*D6=E4i7*beXpHfhAiP8eFu z0lY%p03T$HBcolF**mVJ0cD`t8rOj4qIl{L-EwWtTAireR6G|Yqs&- z7qzzb){btrzI?^E)+&SJpuA=0=lnwKd_4vEJMCTlopz*p?WW%i$XC8zMhD{Rd7vUc zB0XcC(DsbQa=$NW{dh!J#o2kN{4R{)sZQId@-aw?S%Bs$nHc++s_U4hH};R>unN3T zZ!GXHu>C%Y+%0wE0W{8tZEgU0L2iS%bfAJ+m@dr=TO%j36ZRk$O9LIt|u3ka)KXP#eJNCzY( z9_N`l1@=T!gh6rVnw`d0+P?0cl-Wl1OcpUhbRcc0-^f*}9S6$uRmjy3#TJhM25CB5 zDJ|JngJB(7x(MQY<{uJjjyp)D51lyor919WY=t=~py4W^!B?H)YY~H@)OXR54z^9& z*(T4(_m%o1#q3X&^-aqxutC<-4|?Iyd1mA4#9vXG=@!;UOQCK`gz{Wl&A(@o z&aQLooZm|=@$@!x>9hSqeD|k@Fk@ej#B`;5ZvMn~UQfrzb{zcP_e`C*k&_qiYg>a4 zpI5_cgEn5wotgYo;`_&d!!N#L@%f0Ef3q_)z3A?oJGYhv978XQFzqPa`g(zz=R0`- zLPDTB_j&%5jPV>gJh%%vN`c)1F2TX<$q(#=fD05jLlNhW%>jNClf{auK=$Qe@`s!} zQRWC1^2c|Eg*zD70?>G)ySGT=_k| ztd3LzwoR`dcuC*Fw`{VN${CI3&G1^q5#BVOCd6|JP1Q+LlsyilxUbcu9h?Ya!@=B) z{Igxv%>ezEC!p2tC5g?faD=gBa*{j0e9x;;#O>CtJEz!zcW4t%s)F%sx>Ynp`Fs9> zv_U)9soaGP2DaK;52sZpnmmLLJO`s!DEr7gQ>YZ=#EYbsr+q)(sa;RQD^waobhPU zO}f;GW3SA$ba1;J!7Dw5nhIaE&74yvJ_ZF!ax5#}!T-%vW7G8--1gjasiWPQJntPV z76_PV%3yXI73j~}rwuEJG_z*XG_5WPt}Or1Ru<>LK)#MPJEJLQnC{6S+%KA zv!drfmfUefS)kDN9E@+i4-m^hOvzMzEvx_m2Vtp5$-X5`nIhAj`dDQ)6xUH1PL#bB zx$dBBI_JFDidn8$JFA55YwQihAVI2tE!Q);o|6L7Y@2x~%-}MzsDCitiZ&ps2A1e3 ziaiIvs^t?vo08@tI#u;kQp>Huor;+d?F)*t=m|j1oiiy&t?Fp6Y9?bD z7tHK^@Fm!sw>;1}&Oc|ncg4^DV}6C4Z7b^!3J{R1(Eo$K;Naf7swb(*T~NwdZc+2=&InaAsqC&!-}fPqSJDv|9);?zUgJ1Ea5><#0Y zHzE*pd_0|f*+UFnLLcbA9>0GCSYW{K+|=rF4f{4(s)-slTI!ddHF{~_utB?rCR!=D z9CDK}qv#4O7Q`GI4V7!OZI6N1F=y|dIg209QXLrJE8Ql5NL&09GpMcmY`d&Lf5z(_ zG)hX!vP;sb2Z5lhY6s8flzeNdCirO~+B(S{a-X zepwMB6R%NJEU<4%k+G#Zu%1MXJdpW)pS&D|bKdy-;eU@3Zu|O7AJg#}I;a)R?%A8C zk$wy^rY+23`nTih`i|u3?`@qndO-*lo<>?>|Gu07{&LL$J;gKdZiCGM{RHJzkvYW! zW@YIK3FQ#OpR`SM!k$+x$|#(=D5cSyfpzr;+}a{k59~G|jL-%2QeNtl6ek@=CIfQq zqhc_dvzgr{S~_*m9%D>JD5A4x1^K>LPV`Y7@wj8fh|jqY$Y;}x(?U zVu)?7w;vO}#om3SpI5~t^q%;bI?F`Ba(dMce~Ot2fB$j3UIrO5iF0G-?ljst6QHAJ z!N4%BCa1nAt~|;qUm=zxZh(5E;?ACI^LH;oYYw;e~sS0gTvta>oa+` zH_;0YFb(@J6gL;zK(z;SB^A_sP6?bVFd(oBcFYi*t^zxq1`GK7Nizv_8fDDH!L(J1 zd@clg%4zzq7zMajo0Ar6_d~@sx-96^bMGf-JXmd;MV&Vc4z*T9k-U9U*PihaRasS~ zO-sTV*9v|KT9px~bae-i%vXgi>w+zeCfAv0?;~8PF0bxLrOZ~qsF@t_lVh-K{<5+9 zyw~xi%~TJEZr`hYKmGd9?|7$8fhGwU727>DLMiF=Y!Fgoo}q~lSxE-JwL>~v#Se1U$9u9*WMtWK#s6# zMj-6pN(08pLs*`?*SY*9;izvuB6+i?_k_o8Rp#f%s7w2VXZNhWU!elMpQ&HI&-kBN z89fO{r3A5f@O6SsISc8Vnryb6nesJRqKru=d~xJfn6D&ap$_3}-E7Tzd>~LH7aoI2 zASP`#cj^1!a~|43mIc1*`QBiRb>tU-%CY$(Gc>V{zBuSifkfbhx?#IKo@s;hFxvZ- zFb0!Z_2-yh;%o4a!w8-o*;sRGw_QR_`MJE3GEI7Q zZ^%Z_)&KFjbT9@OyQV~B>)f06HK3#-TDOb^&JcqK%l_S1BLAdg%1*pN2u&u6U9g|| z(z=IYY=$$C$IL1FWbZKO2h`mID4J*=TR0*OWF$|5@I3;mT(lc9XMEGUIKArsX+5w0%ITfyDC-IrY3p>{3v&5mqgolmafPm(~R&y)=Oq?LHIR7$>jv3v2NdDYT@E=xJ`35@pzOJ7 zf~9R^({736Q~`}6Vy(yNh%Ps+GiT50#ZuWl{Mj{3QQcCl3f@-TjSFD3o7AXnUl!$?~c^X_b@u?4d z)sU9|Acyq4%$uETMP8nt51BJ**3I>xws1DLZ zQYgiO$1MHU_IFmWo$yNfpFX#lx6k9sA+Ief1SQ-d1*6YL0b3U1j%}A-{f=0Ot!}4Q z2=U)h=Oj1( zWxUeSq|^-OABn6}zZ2wDz@be24Qd4ofRz3XL*o5cqS8fR9oUSWyGDA=3o>n*BPJt9 zqg8bziDpOcr#H+`HLpr2LE9)R`HN^F~Ggh3&+3oMvrn>2JF_ZU64OSAQhrsp) z*w9=fE-eacAhIqabh5E*<{ogImpH{mwU<6Tyxxm(pRgd&(|cgSJNIcusJVy?>odQh zjPueT+3_!G*q}34>L^9`*xl5`#Iz_4gJYf&+bnA`5jX1SnAvqXeAEeiwlf@+2@+lJ zu|WP5*{ha%Km#ks(i^S=oTW;_QC~hT6NoHWO4HH9+(Bzs=9p`w=j=A0R6%L50R4(; z;gHlPzPx+7j1&Rs+9l4P-AfqNS-HDk0#}*Aqt`o+G%L(7w43ojl|k2|f2PAXQYZ_q zVkgTn4gAKd8-vwt`$5y-m#O&Mh6vWGihK3AAFr|D-vb|AEcGlRLP5znaxR)_lRW}A zylXQ(ts79PC{<(GGuoJkaU03&QBvmRtv#d+bwh*Dq{(SjoCVQT1eOXaWq4hflXf+- z%ED6&VQ@sXhJ?m(HJ?Dh)Eq0#5BI~|Ek(1da-dbANut#4{r#S;rZD8--I8ypNdCOW z7MP$MvZD;^TBjbCR9-G-N4UigRvy_Pj}VhJ(GU=^sNvTeb>VlZIju3R%Do#gApN_+ zwyk#$;N^@kFK$!4)?z_Lj3dDyC*NLp@9tEiaEk|q{(TeUDPB*04$XN)}*$zFO}oMC_kDY-@U$o zaT66YBWeSJ!|E}59Jh>n$mlLv`_*bmA8IcdD45e{7NzKhDOV@GiqIC{sk6X3R+G5R zJ5_^b=RXUdWaqrZ?Y7!i&noFpAvxM8xk@7h_S(O7!a!DpfRYB)z>ZL2(nTiQapOh* zYEt8lI}mXzsBPA`wZ5vUoL^%c^1~UiTsEw=1kCDDOwr=uffE5?Fej0@$U`GhslZqW zSKl08a$Zjy*%It8m=Is>N14{!)rtMa+p7sa=xRTbciBna^Ed z41}dJ&NIRjBnnpOHhk*?#ZkEaP$%hj`(ahtllwpc^zS7Wa@`On$6zFI-N<=*wC?e= zn%nVH-89%|+TUGIe|igdfomjk53*7ApgZ{`#9oCp&&G1Y z;0h~OEEBa1@HiI5s9zTqh_)Wo2%Ad4HT>BYh$6uJa17Bv!V6l59_4~Itg654>#-C^ zEBhS4^YLPk6h5h%@}r_ya5|K1LovB<@OOjaCVARes6@RPh$8`~ysJ&EL)CXPpEly! zl71z`%%b5?#%j2$*%oPWF^%(6+eC&%y?Sm3acT|CM%llb*N86r=8|qRgQ*Auy#2*5 zspTqEc`Myo5q=w@_OCv}5FjbK%KvqVDCdGCY)6 z0Z57R-gV@wJIX*LX*x8iaIjY2M$hETk7-z&OGv~BpejEjx`qPr5ngcJon%o^x9xU- zoK<14C`!NrKgW6pttXr=wQ{0)uy1Ww1d3U;igYq+pjQ3+`? z=@HVFJB?RS);3YCJ>fb{T+UCRvFWEEw`*v2`s)M-q;Zn+Kur>jC99QlA3wd1^E+6! zVY4t@XhQ>@ju^+?VQw#B+VtcQInh=DTa1HNAA`61_bEa?H36XU@uV9Cz0ufI(pjV_ zhas_fr0pO?;)U_joR*(l+e%PUz?BJA=YWX`FFU;5=aI9EO2TKZQDAigznA{Ak`CLm zp#GMhV7~0~-K@R%cW#A4kAe1#{hWd)n@_s#SzQNbUI^R2Rkv-%3|=QLs^ES}ib2Kq zd+jmh5MQcV1|NK*I{SvlZFN`+$AeH*+h;e#GgpZ(Xdoo!;|c{W_{@a993>I(Wd@&y z;Yt3W2+41baWM`)A*d!vsodrhh45mpa5&<&a$KOh+cDx&T&j(-1vsOKb~OBF&HW+- zGK9!aB9jI!C2YrC)(sx* zqiVAEcmGdB@K*kzc@~XV=mEG;gb+C!1Qb?(l2IRq(mzXw2IGFYgazdu&FsRgWHlSH?rm9<$Rcxk z&gp;ZYY)1?^hUj40Mp9IcJ!7pYoDO zwCVf5Pr@U46Yn8t-tXSIXX#?{_}y@Rr8~XOK@-)5D)r??H9X2s_(3gu_!tVdI)BiF zkCO_L-)L>bevlp?3d(o~s1mBbT-gw^^i{EET+7zDTV;wPYQ11M!60`L96DXKf=UKh z#%>e(omhzyh1(r|lLdg**OOKv%^HX7QNDX{Ym!HCIJTRIuXpzzb`helTL#WJXRu+^ zVLqq7b<_PR187@6j#8(8UL&>zhfOdF6pa+JvXJJ{)zd4}E^= z=069N9H)3W)_m(_`&kx$Z(B7zAC1{}@TEVaW2W#s%~L#8F<5nx>bw5ES({;(;suxB zb}k9@o~Vf>43p<~vy3ywgy*R`gYhw0bk<6n>&2_B_!&#hXy(#;Zd1RsK@2Sc_FTs0h zeuCwXgl>NoLVSF5H3qQuMKYB%FO!eUhlswLmzS4BtUG@HF}{iZbNF_$G4An;L^Mw| z7d3`k$9{J?u^rZSN?sam^$7BK9%Z&OZ>gCv*Le!(L>&sw^SwH8^%Gv2k7uhs==Q=o zs&XLpeb;fa1NJ}EXJM@|Fzi&poDD~6Ej2H?ZZgmAUGxd-qaoD^c-3GQQd>c2FSA>z zYH0R=^*NpWoHpZzRSIV@r?<2*H~-wKgmZ##Lu(!4*O-rU?*Xv!abAF=!F{k9ffm_C z4TEBsUWZJ>G`N;yJGl|8xDMJMdL9a?w}-QApEouYo>KI!zW|17EJgX-;h2A6K>!b0 zFKK@X9H*TC9N6upL`lE2R&s7Ze;EVl_7d+plLckq7X<#{ftu#u0%h0v z#-X$QqK@K9deDSwW3trxvS;VB>JY%>) zpJ8K32765F;NB++e|@Dz3N)Pb&(b}OKxl@8v6hA6R#(2rgRG>VjP>3>p#Iw=?8ROR zSaeJFUeaDy5s@)y({6knz_x7*O)&%vm+Ca>pKIuBFN7}hCYndw>9w7BcXc5jJ5NiG ztvMVuDONdB^e$`IOvB_^DLpHFZka^A2wBsV#@4RV9F^8n1nrAAH4797b~2zxp-AJp^kLisl?Oz zX)imsePd#3_2Nv`qwb!QcWU64ty?Z**-++rxw!yt`LQejxU`t6o9$f?z;Pa*2b0W3 zS{?x_jbWwx#g{G?(4r8#Zdly0WwD$5NUml_LllchwCvmKO zM>afr9pX}5<$7URZeorf9$FW%qj8%-Vt{r5Y(BMD%PKzwsAUa7dLRH0u;)E&_|+~v z__29BO#k;LY%F*a&L=h1{s>H<8GTAr&A6S5be5j_F=$b>ydgG<$wa-&gcx7l+)6wn z!_fS21MecA82TvxbRC16EjT|Hs#{&ifr&9W+ORGZO8SsCQowrD@)oVCvnKIRKNqv% zrKVWPwT;THnqIpn_n7fHzfv$UsYdorEZZKJNc zG&{wTntoYtZ+61r3eRoh+bU+N=fYfvY%|@qfo5T$D^D&fY3+FQN!i$~AeAjtZ56)| z%NQ(UZnzJDX8WGp+olu{eD64)CjJ=E8C6Kw$$& zpB+Ty99dTUFmy#4!TYxXrOJ~pkM~*L$H4Bv4&qdcGxD+Kfl~r-5Kdeu`|7>u%oAW(ZP{17HV4 z7*u-vT=6=33B7NItHEW~3LjZYcDlm8{ad=TI31p3WjJt57~{k*wp}JPyIwFO-pxd` zU1%Q6Pl0Hh&JizbNH7U=z~w52iN-U3Gt5oxVZCvsV&$(vuur|~9+Pn}56M|>Q87CN z`&O7;HR;8fLYpx@xB*_k1%*KvK1~|!Ss}*J@DE;dMX64XSx4fJBnGL=v=$z-9P?H(;fTHMa9A!;7NhKcH zXD?aG?EF=qWBEOoHN9`?p*&2=F}@$ug~dNIFnt~yh7*dn1&0K>Xyt{B#N_5r+1fs| zEvJ#UHUeoD3*go@0m{>RZz(LO2-TTfvBh*b~hOU$JwsL}(s;bXSTi2PDr6B;H^ zj?HAtD9c!HkN}}@B5{(7ao0dD^ptfR1-*gw-GrOZkfiq94iajKjR2{XK8e{Mxe5=c z0`<86&87CV&X7?i1q|I|7+lu`%A#+2_?hS2|0LgzR+Q2?$E5DZe}{o*CY%B+0kTny zu3Ns5d~<0t+1^LngwPgyTq-aU0SNfGnX@J{OFDE+j9N{&MDi&TlM2b6@%@P1 zk||R3fv9UV!L<)EG>{3QAVKRP=}Mnu)fp7%{EJRNV7iwEJ$n>+%19xt!qWz8o@d66 z9IM-|T91%$G6iBkHs5 z5nLD0vA_|{t1p37 zg-$#q^b7!h#hjwjgKhkV>b_P617pbUaI50 zPn7{`IdO~Sh6{v=w5Un)3FmhTnD+Kf2U#?@sJbJtf;Q-@7IjF0Q z#;4&n3eHr%JntiDzNWL?+8muu*5;k@mNjBs2&mq%T-~)6!N-0v3brC|Hka$h{`Q%s zC^P2%`|Ax03XE~|Y8Ls01MXL6mS8JBzV&#?%hWK7d_IXpo-vIZWn91<+U$e}s?Er{ zIwlidT?N5eo_H+QMBm67kh5$8?mTe>~Q=G8= zUn_fm>3aoLlMtr4tjcI*IkxC==2-zLwYtH6!jw#j4*#YCziYh$a7tfKW%K;(B>bNI zZ*GURdcgWf-;2=D>jt5N%wDg7KOi947C_4s48DHBOKw%6z&J3_xc)NNvG2?5K$xlr z7jvR(cjt1TK@_cKRs30kdN9y)aZ{}aPJmadZmlko6wd`52~6CA03A?v=dAP8^rf0% zT6zh-z=yKl>Mt4~N!=C|beSrKDtWCp8fgYF2Pt>M7@*qR#h$)v7ZFLq9I~jWG*nQl ze0>t-sb|L4viv5--;$GBB2>;06J`$tuCdb+&1UWX!dewf;`Bv_ZR*N}e%tmhzLdQ(*q_j&|1N0nfVNYpjcdz4i zc@p@;XaOQ;tP`-Q$Tvrv&f@6o_l-AUdxQ_|ZMZ1Sx#>5#Nhz#fDS7N} z;?MQ9BZO#R)^)lCxjvWMMDG+|t9z8GX=_r6b1MZ{n>gY6`C}Kt*pD%)8~WgoWF+Hm zY;WPxn}3$t$xhKA0d{~SV=Z-yl=LdW?)HE5`mLVY*r0!SFIPXjivN}O;^b=N`G1#3 zo&O79r%_IZzZQ5q_n=xDqQko2Qfg_R?h(=*5S5@S#lRHuy`4gz#h4|I8UNH>_Bo%a zeC}gxl-tN1A0q{xIHfG5cXzg2VBuoMqKmA*FR&*=Xju# zWr&c4#E}73;z2{BPTL1lE_o*#E_D=N@<|s<%bE9jEX5UzN-j%%*Y%mgAlxP`W50`} z!#pL41)~{WfJUPS6fZ-$%zsIyk>SYno|pw?pb9jv>ojU1H7MRHqatRG1&~ z$IO^{=#_pdP6O_Y?0~Ywr;RpXD%yfG#-v5Q)bQ{P~-!#r^ zp^w-EDdX({Io)0CY1`!LK4;b;W}G!`-v}q(e^9g2xkJj0^*ac$h+UM?g;qc#A)c*dOlEm`JV-TgHa2pB9($K@9WpoC!_5(t{#O8dST6Al3 zh1ScqOvYZki(8INpBR{=AgXI{RlIhhMhnD;e za`fJ+4yqz+k{;RiX_SWbMx_y@Q{%wb^E+%T29nZ99qWBU@g}v*fv7my(!3OZaZztRM>RVqwJfsMpu$XG_4j9kX85c zrg8+h)j%ruUS%5K1dd!>gaO${;vG@L{JJ4Ud(0L^!m0&Fu@DkW%Pl(%=%!p)qc%ir zf~CPO*;6nxzHJGGR*$0JwV@^|L_ltu^zStelA#GaBOSsAu0Xq zhMV@!r!7vwX4QOo^XZ5l>7EY?U>r*JWYFsML)yEI4hR9ss8;GyKhE9hKj^L>*b90} z9?zuUQ|=TL;$xTe7lZr*@(CR}hT-h3>1~Rkvn0e+qH9cC88nkiiek!vVh+vOrk}RG z8UfQ_;|&S*+nhqU7af*X-R=vF+t++V`qFeU;e%K(Eq<5 z0F!4DvXDRQh3}t3;Q#UcbpDT(?03>e)DM6!?B)|~5JR1V-LOCdK7_6~JlalWOj-I2 zqhQro;@2{k6a{W(kGF(?Nuw^z<8Sh5!6JJE6cIA_iRGM1WWW(0Px)AX={+zUE=qq? zvD0GTpL~=OQMCq1PWpBrT7mTk%A}D~{@m)S7G0%h2AClSGkix@blBI*yKw-S&2ulozQ0gk~S&Uf0VSVLG$1 zrCP9G-L0<;e()CQN|2{C)mSQzMT6qM9*Qz-mMAjqAW~9VQONeugFqc@<4=?h6mBR6 z;}3kdbpxF`|N7`?GXS33sp1F00CuHoK}<2mtC^E%`NahlUk-&@0iT}UF>|pN*){I6 z=S(Rg;sm<@qzrH&Yz%|3mo*F*WIbB`)oulU2tEkU=(Ih4bEUfPGU^NxZGPhfK4hxn zJ?3qqlDBsxx@1K)f^sQ=T*{~sgx$z$f$T~CM8P<0|Isyg1L5FsCY|1{JP(9} zMwtsYCSv#s;>IP7`qC*(@NwXTXJ3oO$J(^vX?C{<%1$6%jyrr8m6(@0Moe#AkPp_x@J;zyMo$J_r764J2up|Al+8}DhAa0(27i5kc zua&4>J@g0Uk$BKPFP;yfKg4wK1M*1r8eFd1uu$=+lc162Awh$G63j}8-_@@Z)PGGA zSvUbAb|sHbKZA&u%bN4^y{gL1eNy9Yb&%a9eON<$3+9CsPgoCzRSQKR+&6O%dM5zx z^Z_O}l%?9klA#}SIyB{Js&rozY#BagTOW>inYZElX~Q7UrxSV~6oZZ1wA$a;L82@R zvhm~9cFvA)l4o@_tSboiPOqlL^u(!zhIzWyp^&J)Fm5!NKZK5oOEQr**8t%MsCM3+W^~xQ%I) zcyvrEcObu`$D-RfaF*Q!s5Cks&4Ynti@ZxLcKalboGfcf%Gk5}fxa zq)XgZTi*2N=t~XAtNOliu2#~W#NKYhGQ0;IViktyn{B7mCCcy3 z6e*qxr={G3X>;hdrcy-}BNO+Hw`d^$i!AVBTAxR!!XMh+J$~E28a7?z z*?~kUHjPx}S^}V%#gkr%Eyq+j=Zm*ZVLK7|7;L<#eS2A3KMM4K03>VCz| zh#vdY+{}@vTonwW@F`!>D+cyKT-}}qL&QLPCl49+eS{W6+YvFeiGXM2hFa3!lgZDk zxyu!?QsjlGK~S2=bBm-k-qj4G+EYg~O;aN2+%ey@oe!N#>@6W_C9x-a)r8#Vbt%H(x}t5$Apz!zA({jMcyu$YXA!7))-LB_r%(L-64aZ zkQvL8-Cpu!{+zWD~K<7w`=xpirmfX|ySd^(}{dVL@H z@?sAjW{Ms!y49IK{U7l3I1ngQn;&3&@lOu_SAnOqy^W=brOSV=nx+ZbGN8;zp*#1q z5MT=;)Zlq^5r~g>25J__$kfEwGc^TkQ;Ee*VxKqiPs&xtrsGrYGd44l)WbTRPDL^y zfcF3bn7cYMYbVra5f&mC1@stJ>t(3Y%5jPG>2oh}wNQpI??%>|a4Lx-r4yUQA{_u* zN-Xu+>RIZskGA;F6C-A+66Kn*cMT?t{`);XSb@|B>5j~>s|L~Fw4`eD%yQ%sT42vH zBFi=gO_#7UJ?K0(<~SK29CJhNIq08EG@NrfUO6zYUhUXRo=4e>XbHm&GM-Q}`?RH* ztY1B%Y{_DazBKl9sQvZ87MPdEpK=y38Vh`2u-Zc6RrdK6*H2EH&s-{g@XB-3g4{aaS?AgHXdbkL z>D<;n*hh0heaIS)9zOJOTAu3m^h&SjyhuA49Y&u`93C;X#8meAAGet>Vm$!W4@8O> z`TwEs{J+048aH+u97x~SWki>ViFgWE9SX1*HoE8e{qrW0IyOSt&<(#fY%G+C&PugZ z3>X{IHj*x5i@#{?a{3C#I}f&QzG$F={)*9Xyt{}>CD6#BYsrY&#jKl~)mq47JS0U* z03*e&*2GUBiAkXDuS_YL79V;wsXIyR%O0~Su}urdQ*wt^<0j#}F-*6}&Sh-1x3Crc zr8=lscG#=fyeemU!+#(8MWsP#VOD4?LQ(8-D4TAa$eL5b#ozAtdN4^SP70Ez&jw|- z7Moc;btNyqVj?P+$n6MSnS$l1AZF29WGoX#*h1k(3qd?_3Tf7|$CVvR3Ig?@px1@P znSNVDB}9gIc2#xI8OA=-RoCLO%>H*0#9-WLk}c4`v1G`Y^AOA^CqVXJbawnkJ;woW zXMZ;$YQlB-NTX0zU(VIM?0E5yvs1b+p>VfEx1S4s$)jO$!Qs^bZI3lq`~W+*msH`U zEppQp)k&0?t1py$ym&0bP&ppX&%W5mK|j4D({D7SRAg{Z0zi$X1-uJyn0z>%IT?QV zW2Ki&Day>F!J~hzqAbjf540&CE1VXU)#r`})`)p`WR#xxF3TxR2&$ zs5$AK)n^2e`7U8I4rFpq$w8)vM#bJaN&Zv3hBB8(Lq=LLHr_+^!Ki3emLsNPr9 zYe37YxM3MgDAt{KOJBdIAM*5ZxCn>^ungH2%vM%we@u?vL9G@?Kc1O6$4jbpBNm20 z^=xk~IkT;w%P>>RkfhI8Bk$i>{(@f>r>%9@sk|OFdFdy#$W*RAi?gYvje-6eP|Q6r zN0J>cQwfiXdl!abc0r;!F#uu5Ng^H5!%ifPjyN1F6O$s{=pB_|;)4ZK@uWyoXOl=S^_$+NkMR3>ws%P_5LcF+)&ii2A5h!-?W?oG zKt+7oAFI(wCA85QS|IwoFD4Y5=|$Lotg`DM3KwC|gWqG?mMt*I?RcCL6B43=GRmAG zj*qReq9AMzJ0W9#GYi}?StVwVSb6MI5jQi-Ln(pZ7Duz$p=5E?_NBqu{*qN#haRMq z;|5;bl!(2wPMM23Kh}Bc@qr;t*D7fYsnVBh=T-snaYLPvoj1LEardzQne`f@m^9p{ zT>m3)jxE>^X~pfZO=$@d?rC|W_+VLEjzgj+NGG8))lO`Vq)s=i!_DPN^e8&)f$EJj z+AtU>9|ak(Vo+qkYh-USd4UMn6nQ~EazJ~uAC6nvNlU;oP^hfocc6r^I`;x;Z8PH{ zNc|e4njQC0R>mi?%KV&|Q9B_2c+{$dvvVHm*qvH^LlJM?A>6@Vw4h!OGEC3+h2VvN zIz6N|_b-E*U=Be&u>{5^keNVxU_wdb*W-mTa0r&!3bAkxq34KTr4 zY-ovvh{<$rF$Qr$jAD^J>X2j|YxkCh)yPt3(%QZt<_~4t#ww-P)>P=jLg-fp3x8-1 z`MtK%T7SKHV;dHtV{mX4+d4GA?(->4n)^!Jz7pTN9`xzpyAR%kDD0i&Q3TB~Qc5s) z=TQ<)PSZ@~Z2_uCJJbBSEhFKY1N8W(OS6p7@_Nj>WZ0BNrw3&0P2K`O{4a2Dc2Y;; zZ`o5yGYEZa46B`xzVkGjHl32LNYOzKjb+=_=ewfk<)brzTql986?y5yX)PTbkY1Lk^9UKDo_JjT`_6s1y6zGUA+VIS02kc_b!thPj=Ks+34osqk zO|oFywr$(CZQI?aZTqxs+qP}nw(UNB`-|9}-I=)a6W+I;imJ-W%519oJxV1C41q^= zYHh`H-09oib-V^-e}7OYylj5|A3(EGXjWa#?~n`z^}o$JZkA^MnUs^&{$Ew(UwMtB zx(dWRc~*lWQBMyGS8WoC2qMU~D3+G?63Kv~OOVVjFLMdW6cp^?3(+MDy09>Fv4gDH znqtk{V=N8gZ17RFmHONaG2l+#GSxJPr}XNq?RpPWYXmrxXgjM_4fHkC47-BjsLM?P zVDGqPG0d}qrVXK6zhxG6*H&QUFYxCgnaidY;o99RR=!~!U$#tDEzjJ&ErCK)(A5XT zC!>+O3d3P6@aRTV_gt^J$^_6((8I_hL3O4&(#44|o*`YZ5}8+_>a_ZRkoJe3ejtzj zf>Foux9;+KZhL?xyUpX@MgWhVM`~a0dTc7WNKegE=Do&jX`y8xRK%s zWk)Fy$D4;4F<`7FR;Wp`S%2dHlxIdym1f?+U$LyZY33x`XgAoz^E$_-k0DrC)AZ2k za>}Xa$g{6MXk;%zzL!y!1k&$9wQqm4R(Y&U1cO$_CyT>QW=7)E^~hKK`&b>NzW=2E z`T8E6o^6ATVLG}8<;j;?$9+vo<`>Dd^U}g-2j=(OR_g$@aa$6Y{s1;a<*Th$UUNDk zxs$b5i%f0-uMas5a$zO?99WJ>jXLaKxUN&gvIw|e4n@=0T>fG;op-IKDi5)ha9$~f zx!s9BoRmIQm4rXUGTE;UewN1sA~@upsGZ^xjg#y+mE4h*S&IcFOi~mKccqx~jjIlr z`@#!)?4M6oEE?{Ly)ud+Br!UWvz{9mMHB@YC9yPm7Np(ePe`qBMPO^W>96h|qC=6C zXm9lj>nOt1X3{KW)^yx~V@5P>`LdX<)7PfO17*)G!FB1c8)=JfOqG8Pd(Bn2Q|8XN zrJr5vA-%9UOw$MzG{V&!3Xq=0t0seNwfnmezw(G zskl)?AoRnXkX+{@x(^|emA_jOK#fVe&#aU;ESAUjRSl}HIwAL$S#Yf}Z>?Tj`Ik%w z3ymKp=HZxFO1SEOAn4#G125%~X2R*}ilnx?PJG{#gj?)VvVfVKJ-{`!`||TNY9aTd zz+u=m@KY{MUP%i=KLFNyzMi^T#W8R$VS5==J`7p>YR1)g#ro+dBu?=2KN`s6^GIi| z-z4w?BxJ;ws=5P1=X@(MH5tG&f~G zZ=J5<;3c4{S9tudyc}=Z@!MN&Kk*8C>!Sp(TeJyTNrZ$@=tdTM%zyl;`fw6p;`c8pBKEJmg571Gc_By|W8o z3GG3wibt~^h|W~2wAzwNm`lY9TF^*T(dm}8B0(|{v}^-I8OP=H-`6J*EAf>sNGhZ> z`ee)2g)e1m!qadmhOh)m-=n^}J!w>A=*b251swC?fDGNL9_gnTMR1jA)W{M{Aj}L$ zn`ht>j~wn6WQx6zcnHKPy(=0WzusmvqikdQUTwuw#Nf%x{Io}rH8U(ds?(;5mNXL9 zh0`Ldg=yLw2FHlL@~`0B$KC%@$bT zH37;Z#OLH%982O!{?15$Ja->tuVep8pq~kNs%9tUBH!r})W%sE3O#KXk>n~LZo@%X zAtVR4MAp2V{{h&~TCjL(z_a@7ok+6!`lFuU!B~<3mE)BUb)1Py-Q?!pW&WM+Rg3er zOlTB%EYvBhPV+4DtX#Tmx7;jq9k5He@VtDRQtp%=l%ZE{X2t?Nnn**AwiODxZe+PN zR7^uN?MX^PdV?UNoL_7}VR(R9J$~63=fE}a^9H|5uKX$5WO?x8VnL?n{1+4HWGA>e z4$d;vAR=@6wT@J+NTwuwhAM<59%phDO4!gYbFz#K_z^k_V_U08IM5MDbQ(oxpwI!} z2I#%5aEc=z>rO5bcBNe4MJt0LIWSdB7G4VV4+b~zu1>x&1-t+H^+)q_na);2xA($T zWHRXf_y0&Z4}-q~*T1-qC(8d~|KjZA{-1Qy*!~^aPW)b?~}7+nlpv10h&8&5no;SQ62Zq%rPNc(w^32Zi9l*H>0OC$mHo zN??XXkiJnkdh1Lg7q|yFqXpGVT7&0jH6q7Rsu?4Z0}Yzv$0S5jS%i!xNqhx%-e|T> zD2Zj{!p`D(OtK&VJr5%BIq0so=dmFRSjY&;IYO_M=k}1Y`{%%Ul_s{r6~=4}j?|6_ ztNLBBx`9Xqe73U1Y*R<1bo<@V8xNL(;T50HIOLKH2$Pzj0cxYqaycOQz$rG#lE#ry zlmpNq)4e%-eZHy-$;X34ke!dE?)Tv0y@2^wJ&7oBv6;F62i;7X=3aqxqzYl~njJf~@cF{O*iA_S5sBUC0jMfX&e52;ydPnN^r|2^&GMAgcP_a4rlUOv3KKYe^4 zjGPA84)AAv1o0KU&#Mn|rS0}!WM}z3F45Qkmwe`v^D5^C4z}ikEmOE_-lRdOfxxK^ zJ$*?@wTnhlf-f=4v%uG!~QmO%BBrdq#|kqfSP3 z7>r}08k|EW(ZtH@6LPuA%XrO}B0sbQ`J!(rLY*|m_;A+LvSbnry3cFZE=CIXp09g7 zmcKQ3f5uL}RHkg7e%i1VImzODaUok=5duFGGF8YzkxXt#ZlOBi>8@33v?%+8G0r}rX7+&A*aq#?URe3vw}VADAD&V zWSyySQ#D&MpcrjL%qJ0oif@zFIORYbqKHK~BqxY9HJ6U1{vx6fHnu%kGfo}=urKg@o|_`J|Ht3;6@OgX1@}#;YCMAW5u|P7 z0D##3d=%hTC{F+-*Oa+4*N3xr3np4_G6=zC&~=ua{6g7c0|Oz?nV^2S__Hs3x!wJ%Nxhdy%e{jlD9S_|CKyyg!3U3dJO-SSk)3 zlPU;>0>b`=bt?IcB9N#HNPlRJQ5Rh6OOiL3LB32^sdbcvxT_T#z4t+z)U$E4(nzMG zzBDdYT51_nB> zpGdeLVTSvw-+MK0X+Qz!N%sjMus5qlUd{lw%B?Q}v8>BLl*Z$V6gOh4Xhno^iL|%t z)~Hl&4bw?8{bx5WC3POeYqo^crvz*t=gsan%wA+~7Dbx9u(6FdFVR_=nLoy#D{YwVPlHa+m6^Nx(>4HOvrifn3nGa~<^qK;aED0} ztx5LZ)sn5!geTM$Sc)s0X!BrGD()^4;Ep{;7*#}kLXZIj)vc4)qy0BkZNb#H(UKGe z2zQ0n-l?KfS;4GqlN{m5RDJ{mcId49dN7(5E>g@OZ3{J!5xcg0n++-@FWR+KA6le% z58Kz-D|F@$ecHsRcb0IX8JtM9&U#hiZj|z}6XL+;`2D_NQ7=>cS?hWXu?e1aJI~zY z5VoF(Ftc)hx5{D?GKR`}2p}E69#>*u2(v+R7((7M%dgna$DLkBr$66Frq0Ogdhdme z#K&HA!@M*{up{EYUTEKc*0@S}w45gX@if#NQCvKw0F1XjXB7Xd$z@Xcdf+dCAi~MM zDYynd^YV!0GQ3#0J;#T5`1|SL%J(tzOvnq!BG9ih;ba^~)`duQu?Da07UWXc3NhFC-+~_*y@nqa28Yl@F`ZLLyA7r}GGvp+2E;D*W3NHpsGiv*A_v{eR98!-E8Hms1 z1%u~9o)#>K>~%4z=f?qzEbdn*)TkQGq}QsVX@SnclTGu6OhOI6sZfI<)ptC~IxR66 zAnKsrZY^q~*q{DPR?Cors!uKX+hJ@zIO>8JOn!z6hoNotuLadDYTv_6Uz-gb_myg5 z+x%y>WOa@yja%l3Wb~3YW`hN|?*xk{*P0a{CdOEf*k=Sd za}yn!4Wn@vz07@_bGsPxwbgBGQuhQ;x_c1!!=zM-=4c~F z=I+0a@_fx4de^*tlF3b+@J_+S&J@$_(HcSZd1~ZarP&KM(qOyjq^@Mk@h&?Aa7%PS zW!V@i^4#Umuu#b=dx8)oIUpclUVqw}Mu-9S1Cb}Q%F3cmgclQU9VjbV8XHoZYs*wA znf0rx*DD6%Hk+*5#bfFPDZZ^CEY^$U3u!m3Ar3|)N2BGF!R_mIAEQyWKg$ey3X~?* z|Dzf?Ye5x=-a5bk+0H(sbCzE<$IFVW^Lus$=DGF-1kw)P=ND?I-52oI8a};D#etSz zuH~7QdyDGEvAfVbWX^3V8fA+2O#zNJIoE^eb{h?U!Mf~9ojAK-R=;=1Ga<gH zF8S&W@CX@5eGGT3rQ9bYFYhMr2ds#K`^`wJw&^S?nY5;8E ztq39<(;%z50$YAQ-$2K1!bFeGQf}GdfB&$q-S$k}%Y4#Xo6(fNTm0yv*3|fLmC|y! ztG2x9*qARm*fYsz5`y!@EUnQviTD-aD{&T9Q4(cpnokA>j)RLk{g}k)sJpX3^ck|^ z!1As(f+>%WiogrOv^ogXvE=JO%F$c=?!8MJ^ zyU;hY$dhU?8vI)z&8267wvD^uF7CojLd-2}fK^uM3g3ZFG3C1^CAV}jK3ikdWixGc zn_6w0D?74)(LrG7J8&y#mo=2t^u(HejW=4clBY$#ifRM^v;^O3E9Fo&^o5HIwfm#y zi+uQ)?zQJCG=V*9;Ra#f2L8mKC*knpWIoLU<@)> z86EL7y6-r~S&3Jxq8wa}hGNylwiOLSmb-eWv7DO3*E^>yp2yPry*@w-%6#MlPV9)l zM(*5Ww=c;{#@kUK&l;$Xs7H^8L?!vxQdlr5@IW)sM~bfG?W7IvRyNE~90nn|FD#kn z&^$iywHJ4OHs}#-g;8h|@j!`de*saB{%qGaW*xx$v*?S*0?FvrFmXIfLN+9@nbHb; zov?G>ke9^yE)>=!A0{9#`J!&I7#)4NDp@Qnx;5ky$W~kmy3>$P`N8#RgB# zqdZ5qkk{Z=;Iu5x$tDwXyaDf>2(?Cy@3xu>)1nT}(<^wtEBA=6Nh7xBU@(Gn0%+aMm_A=G3arZD@>&|Ac|V1@Dv4>RaqG`PaOcC6>!4 z0?;fNLe<}9jyeLGtI3hLh?c258}R~nZT@hijr}9PCMNfsow*4x(93dj7XLQw$=qH&}Xfm7EL1VRW0&&C)qOvkQ72x(ef+yNO_x zO2X~WUxy|GMNdvf$hr0GZ_&0#`M%{5Dl<^1cjz#dHQLB^9OgI~J4t=op6jF&^IV?F zmq;Wze{+k=MC}W#?=X2EE{}Ia6fJ#gJzmj z%KDc-WY;uAjbM7*JKIbLc=XR+i_(h-)u&u9#UNl=3%|Ui&zJSr$=6Zo=i=c#L|kPQy00UF#5s=AEE@)bXmdk0J?=6!Hvp z$+>0X>Yt^m#n3vJ_KK=!gg=;^WV%j^C~LJyJd27k~V|7i0a} zfcgX89?%WQY!9$|JVBPg=Y{>M*X+!aK0O;7rYyDgW;>a*_kI?Q(4&W85akc1Z5_0e zUcGe$Z|!g9JL5-bQ*x4V5yJxG$T{5W?1_(2*|0@V7LQxf3g;Z@Hy*Phv2;nSve>_5 zHHj$Z#QH`KO+>NtdqpU7e zr_gIG1LeN}hWz)wlCC3q&c8hD*=(-q;sr6|8Zsl`8yqK#)GVb89D+a+GY}R*rhrmV z0cxufI$3r<$^os2wH2Zm1(7mv5s43-Ax+5UJy;wGOPJ$soOei+G<4dmf`w!icx>DB zONY-0ny?9*#-jrD1a|SGGjLgQ<7%N=Kz-S)6~6LTB&ntMU<)Dpl{fB2^J=z^!bt;i zIM~bmWvWSxdmxSte4*>D-Lg1F#;NGd^GaX952}maSD1^E$cdjA)x2#4R5iqO-Us?6 zGD}jKMm8~^Hm*^3f-a!ji1kdaroOd!-|rrvTT>B;r~_V^(yMlyQk{H4YZ z?lQVK%uVhpxOj4k03 z4{{jlf-s}_Gz5b0A21DO0{>xxL;7-0F4@bvJMSv)&nL-^+cVJ0y@<+}nclU=Mz~uR z{UgFT`zu~b+;P$*a1E>>Tp1?;oB3|0&IK;O!2w-jlwCTc=+s)bRg~95C%#6 z<`-@6cb_$|otuvUKo1XyG4~dga(V5O6DQtO2($#riUN z_o!V-*kIk@n7g-eZ`wYVg6Rqs&Y)H#VP>BS2BIvB+H~`%_j$0b&~;)!r3C6ti}%us z0X~;l4vJ4^rZv#tm&j}HH0onmEFt}-#)gD7+snWK9*M?mMe&^3Ej5vr- zG@748J;Gn|PumspFijpe@-N=?N+!zVDhv(US>_XvBRVhPZsD6W38s8&(_dn9ZrTj9 zy9OR#h}w4vytls5!$H@pr=(!^t?}(>@%~8F6Vn`J-#zwaG-UC%_HSL78K~17RJpO- z2ybv`)*+d!g1)VQXE5lxJOdM1+I+fn<7UW9W@B#(s~ikJ%y}yLnea@Z4Wy)I`TN{N z3ls>m?(rRl&FmxN*uM;RnYNUB0>mwLgvRf=hBBf1LCGE zzWgKIx3yYT|B>xjp{n&Om@f%awWwWl1t!zsdf(zl!Nww+LL)b-<>=K_j+e4q$wS`j zRhgAYT$|7~-tZzy4BA2Js;n_Qmqb_o6@q%4-+##t4R=N~Ne1lRqAdjJTx=@@_J##O zlspA{T%tybCTgIdq%H_mV3VhgPfAe`-7T1GN+G{?To^?{<s4cNtpu0G$Q5KDk*Q%H-B^iNTUd0HEB5{ z3dZ;whbDdYm7)O(_zs_%fHkPtg5@PYmR~_4qZ!P!Rktg`_sPkyydAiszf4FbJLXTtq5B)pj`fB(WaXR+x9sLnTpzA0vfXlB@g+xY2L z`^sxwf##BVejU$WIq4ZJ>BJ6bD5Hw(sGEehf|cg5vBvE2!zVW94yDLTtn>I3z*E0a z)n>(|UAKCf6M37m-2ud$h5nZJZ)9AT@%s1!NHF!pbQE3izHS=h;2jx-uohWAM~6#3 zG9B>^h0nJ#K9GEjr}ns8hu4=ZvM`h#mQlaei(cT`vB5Ot(W%-KeW5n5ei(PUINpNw zg$&AfU%u3a*#(Fz#qXe&IQ8J@%ME1|a1d5WE45Rqd&@zk8GkaEdV@xN>2n$2i1i7e zy>w_NOAqT~bp^?|ALo9TZruOvkoRmP=xy`IO@0ah!?GOLZ{naFFi>3DV+HS#xzwW{ zx?@OIhHbm*p^Sv*(n zaUJsaY?{lm_%eK*HlAw>JUSWsgSCo2&$AEHE}Y+45PO*70w$22J4zD9oCN&SWE!m+ z*5Ev~hh*ZfWRyq~k^kldZ>|aiPzOxlJ1%1Pe%s;nyYb(=Uks#-i&Kl;a^e$&GhkHX zi<^QQ`Mnb)2c=E7R1WNk0%rAX zu&_pxXj*(DaxJ8wf-%UEivh36VPV}&4Za7CZMS%stK&%Q974Kc@{wy9K$YH0&I=XW z4@sfzRvCR(9DW^2)_|t&Uzi#yRQgBfVW7P#;VYc0IhI3}P4I1$Xt{Q5h7npu@eMp1 zRT3eKh-F6^F=z;~FFjtAB|+NJFfp@$ms8=Q0)B)?tJ1?KzG*D#EQk{TzWXJb50S-7 zuC=*O8s}Qm6^yIG7U>{!g8JSNFNJkp4O{$t_X-!p4&5W98@K@ zf~WXu6{}0kBik|iN&9#zTRZUy?1qa^EJJ)8p*08Q_An&fl9i@KNH$xjuj_Ieax|jO z?oq)jKK4s(__?`T&l<~KMV{5lx{se9>F5KhETAuWtm(T190A?vXchX96T=)b3OSF< zn-EJ-k`9~F+)eD#VA52(^6x9J634Je_yZ6&VvHdjwWDv^LGZ zHJ?6tXCG4t9JfIBsz;p`eu6>s&tFdt9l##i+t`fe8ZG}(18qT^nVcpwt*ACFwpMLN zoU7a|ULI~Wen^!Dk3r%6d=|_d>r3}pPj#`uC2p?8V0blY{QOMu6jjT|6e&P=bF_LE z*FJmGD_5XTf+dTK9RQNJLx%6xPSMwcGKSMy_CVj(4fAWJf)QNV|1G`DU#Fn|9{+UV z-O6GG8}cb;F(lI2n9Ek;Y<`4{hl_1Tvp;}kq#JeqZ8v>RK%P84c4s1%|o>FZOs9droQg z3~w?a+ezEdvU^+cj{|#bF9R5vtlCkiqD<8w0m1Z$qf53g5Vvc;3jRuWcy&zPR@T4Sqr3~4Hh+TGy^Brf zx4pSx%k?GZC7em#IOKg>iW}|I^_Y5pi3_zbOA5AZ1p7~xrH|Dwl0NSc`!A{SI*psh zh8Prb7JgXAJ%jUF5&_nR_-cjs%4^feFeHw|%3f;h1>^d_Uj#Rget_=za7$vLaHw>E%t6YrUM7ss$+{dR@i!O$ ze)^OUJql!NnY%o6Xw)!wvd{$rr734Xc6LfRBxaLODF-!3(ay(i^$_E?ILxYni8aGQ^ z4K-!5kyhnpx+Ai#k76I`q*>TXJpI0d_W~-tAYst!p`F%w1lcYFtE&hxnXOIW-$#OIJ&b63**qMK0YHr##NSh za4+7bXuOiz?sbC1yMUb?z-g*|_ClMjPUmgCkId+TmKTfOBr7rFg=;i@NUMDQ%bEAK z-OW)U7SO*qNfC^iKu}%bvG{ng;@L89pbAi82j2AG>Go}HOAhv;W%Lg3Ivhj9xM6V= zIAVYOfc`st8H9)SIq}P~h6DLuw!yA0roT{sv{Jv_f7tz()Zs4#UaVU{+_Yt>fSq>S zO`CL&1Qbl!WiBLwNvWV~qw)>7rIwBq^kikPc;n(uGwbMYEm+1?!cM~GQ8 z57*UCZ<$1c`LIZUd`5*bHFT>}B-DbAmCOrjE8lbM0U!iXFSA?U4HK;11tAQ88Imd| z#DO2Xcd6nNSiRyIhUFDOQ${-tRy1B<2I?$Uxtw8Er!QjPssA(_7xdKrNqxR4@ouf+ zKT&5MHCACU7pK{6+8Q@|pZ7hT_o$4s;Oe7zFtT0tiX{toh=fX@w(*tiy*>+OLw=Qz z&EKro85w*N;KpUc?8X0)&dD{RZ$3nR0tw$XUEgfIgtN}F>=DA<%!L+fsyECKLgp-I zm^6GLj<_BHkDX(~;BW-3ct=es9{fV@rqlyce#eBIN{3@8? zkzwyivI)9mqzzPl!rdx-5<Y5Y^NnB_61U2JSJM{n{6qB z7Q?deUrg+gB0Fc!6FJLsp+`C=kDChj9}j`}LyptVISR*_({R3g4@K=qJPWHnDnX2H z^spJ@3QFhUROz;-M21@S;u`H4cfiW;V=2j}d`7YN7(r?jT?ug{zdRC13Yj2=d9$eg z;YEJ6*Honm5R1BG&U}IO4HuzZyJj~{PklW>bu3b^?)7e<**NxKlbfdZox)a2PgfuO zd&b#5>xn@}EGZjK8qt@-)Xkvoe|OiHr=71tCL0D7O7oTi)Gf;qDZM||E3S=g*E~PY zJsRKTs1=lo!0?{7iPw^#*;|eFVPffNgl4S zFl=6sq3S^=r80*DtzIrEE@Vorw_J%vnR_)EwVr0ti0n$|uW#$m^Quve3D9h*Wy!Do zSGC5)!^E8TeRo$1Zq8~x{$QI7dJFMRHGtMUV(PNVGUA{qS4e;4I6YO%Ijca-khbY0 z3mXybL6uYz;yu-jGo3Bg;LL+)iJej>l4!C#KxcykO4ocl6uKi=X{L7PnM>#7G1Ebi zm^5&pjf5Gjf?I8m%c;BDIlj_7-du6(*UNQOUF&$xj-Z9jQX6L{nY^ z$FDq*L_! zd-(2YRDQ-YbEKHf)9QXStV!Z&!P_%FiavBY)rE$*8({yOP`CumtwT%&-{QWG>GwDKdD{qnMw-x|+q++5zo<)iAH_V}afy((Xy0kzi$ znfLnub+*TSc)vfDp7ZGgvj<2R)X{4LKMNi_zPrA^_@h}$+H@dsX)Q;i%8UZxWTvgj zsZqhfzJbalXC$xC1i|jd40t|j8bZo%3M6J72Kxm>@NO2~iYT;P^#{b3ocj6eg&d+lqKGqNXQXaJNw1VF zN|~v$l}5LT!ptuxFJU_O6ltjy2E^M_Err>XA9;+d3ORjBg{68zHOY$M1~0cg$9AO7 z6mY;UyDR#n#rT|Vd8n_drv5Ca!xBzUWY*cjG2PUCp)>84qjWDH4q2uT4`p=C@Rn37-DuTw1bU6+{jr4{9RQgY z2FE*J+WKX@d2-4kU@}O9UX38Rk0Z{sa_l>t7Yfs$R@AylI!Lk^fH3*`{L_9Sn7sy& zH=7uINV(q`CIeuohzu&wIL^t~QmDC4H=wj4b{#rUA^ab3{VMukI}Q5hQ>OeDl;qy) zzOW}7g#)9n7igUT%Khafvbpko=L_-}MQ87q&XtdVMbN%6*YN8gKIDmh)3g$hv)X*b z##N}#*SlQ6$^I_;?b)oR+uJJh$H)mRwKI{<-WPDO43yAJc#pg13PKiI`K|4NvxTHANFD zGHD8gVKH@mQb_dB?w&TxfqsCGFaoz>3o(nP6acXHxPK1o=Bofh5C94D3TlQ84`dAx zH&!ICK?qSmLG=n^)y#O$t(%0<$0re{M^aZsl0~!cNHB^Ejg0y>ET%HTT%JB$f z0Dpp_#0rn-dFUbmC#2Wf?(JO-XIjWHs|au`Y1Vym*eHrRdp%IgAW?WXR$Aygt1@V) zn7xT6THYJjc@0QTL^2dZ)BEp5r-1ojZ>3VCbA-V>O%g4Dpg>lmmbp;i9AKEbeHl$n z-#9@0spq+*xHut8XZnh|nR=g|La*VeXq|&mImC{gl`HWaBCNcJIcV;JaZZ^=pn-qi zbp}8nhrxJr;F`b^CtJ|YA|O>8J`E6*1${OMHl{E?y^ceCCN*g8oRI{?P$68ykShidCt+q$vUk3R$fmz&d_CQG zJcEGlFP?1neE-oDKtMQX)+d)e7g5a#1}!D0)z=wB@(O4Kya`Y>ZrpAQp@0NoF94fh zQy_@|S_%RPTK6+CN0}w>>h~N}HB%2{W|v7PcIsd6r6+|G1j(9*fcX{*Z8v7PkDP!M zW^7V_>zUQP2LfHqWEckeXShVL%#G(4-ARz`fl2luAZ_T)`5n(pSjkO5P4>GmBE@1M z^al=bxFsco%U4Fp2oW0r%q8;&(*f`+!lL|>@(@l~xIz*PN}*>F3Fz}sgJ?w(1p~Uo zNW#T!e$1YME62n^?TPngX*!=%9ITr&3d`IhsK@@8<+^Yh z0|*LxP(l*4Wf-RqtEK@HhLqbSuM@XlCWuK5$q+-soG`fR-tKW?c|?OMy;moX$>>M>6}tvHun|&cWT@p2256*1dfnpE^eoN00aS zzw1435it0Le-TVEW^Z^wzmCY+7RpTieW?%&qsPXMGI0<2j4we))m$+nj9O_oG%@{R zosc=#;iifuA+ow5MvJAAggWbi0AiZ1*~*e+^Y!OSK5x}D1hfQ_?g*!l;2isEeTyW? zd+jh3&5$1AN7gt44D72Tv)jI1_)5SMgs3$MDnH^e;jH;vB%(>5KrE^ks)%YOnZ=1p zLN=`(?G5te;<3b*XzXp*Dp|dLpWAy0Hh`Q*4&C~`4XC{0612Xn(28Nc!-xHp615T! z8RJ>3N@gN}&ulOjbcB@@^*GHR$m-fD<%-#zN-$yq!Wlh1rCw7OfP^qySu1J?ndS4_ zP>yW5&Yf^Qr+qNql618w>qV4`#-V@FJyb}3qA|YK_`UpuR_4V6>@29S`hDgLA`yx2 zQQ?0TtJ}$j=Wy>&Jy90(Z7~!?3`ZH=9VIg2e`|r$7;L{YZVj8ikHC~u%Yd!uSG+qk z8SQY?dRQ|bdR4R?i5x{D=tTZ$zX7G>(9De5TffevlH!}b1t$hq_qn>J%gK& zW1!J+Smb2y@3X>--O+)IkslK?3W1y9Z!uUlHL{amr}K-ZVFcMHuxWfl3`dbm+{s`L zu7-$9AqYtGT|*%f99GZ0-BH#VWeul9tvk2pcU`Nj?sva#z8q~eA2SSph`@O2Uqd|> zsM8|TGlXevL(V6;-YR}8(E!?p&H_(lMLu5s?9jGB#=2K820(- z=^?!!e-$ghl^{^A!`A^#+MEWuBFk|zwrr!KbgUiVFo^@zXbx~sGO+?kv0>rB&(X_* zU+;O(tHSJ^^>+l!*E6*{C74E*6B-cg>g(O%tdsh)iU4zyz3rLD#DTnfDFOhI#6)QI zgrjH=U?gL2T@hY>r9zU6W<$*wH?JDV5HLt4YwxGCf9TJ-h-{zaZQ6JAn7`+GiUcQj zZZQfL4x$a>bUcNA68KLiA&2PXF`r*7A7+PdsH3j9hrII%)iY@Fl!;jOHvUY-zJ}ed zh*p}Doe$wO8bePfrVoy2Q&Iivwi+Xh0}}E8m@t7JVVlOVWl{Z>uui6j7tqGUXX~nV zu=tK~K7Vnj#20{@aR?L*Km^94aHijc%nsPSC|)3u&)dmiB(xXuF!TPpRN1?&DkkCh z+kDe8v>M`HccY=2%Sux9jl-r5BChoUrQI$$jjUvDY}n8T0bdsb z*5$s}0`bH4W5<;&Bzpvhq&VzfH$N*v*PZANyDX}jmyw%@@t3`kb`Xiedc2&%4FNPs z2lgCr7I59yO+gd`L?xp8b2am7f1j&Klb(!qC|ge#PQ@9&`QfYHQDzRI)uj4b8=w7B zy8P|-zJCSq%12*U1+jkRM*XY8G>jZ)?H@?Bd3qf@dcl5YhnNoTeSWcTKeJlBQOU)) z=vJJ1CXU4usp7HvV!F-?A{`X240Rn>>u+m<$Nal@_(FzBxS1ZN_4I~3(??gwgJH7} zIl3gQG!9R7-B0(1(0F0^hdL$0lVP!VHMQS|OB_c1uF3oj^gnM@rl zwk1q3c&-Tn=&i3fL?;KUtq^)r1j3;Oi3EnzKS0;=u>5H=u}MWuiV$ZFi(O;kiGssH z%*7GYix`?mp8w!b@v#h zA%1;38~|OsnhtqWBZypx9imho>ycONfQ}$6x2jYYKDcPzC!e8GtE>kE_Y%i~1Hbot zqaUv|oqF`-Df1Y)Dl`S;`W{y`v>)CuJ|cLe-d#VEV^Nu zIl7o1AP?cBWTR0BH1+k2E0CuJ{mMI|y~5i(NC3qYX3d<&r2$^T`+9(T>_{VSa{d9E zbofZ2vFGMYD%f#_Ctc2uxhR?Dq71WkmzRN$4aQ%_{R(~+iphSgIHiVq7k?cLL$98n zeu7;n-fy)Uv~fh%WWTr;lws2|M6#|lgX5U5L-dL;^fc-V=Y4ca6y9@48gt_<2KikT zlZc7pN<>D)reV+Uss15Z3h=H%sZzh|y-L$|1Ne*I^A z{m<82+j`88>+Su8j~z?4&lw_I`$Y+68=XP+^eV$xLN($ za36{%>nx8H;5PKw_uM| zxsKx5B}0Kqtu)e01IbX=VEX=bnd1I6e8DWg-0$vXq|Nh>xvUMB5qa?nT6gPD?5SzLIyOU(9iUfkNzKvhfrAc&Krhfj zwYT*)u6H2TBM2KXu~`>4S;2=_7jKRW_jdpo7|PJ|y@z zD!=OHfG;uKRhR5Cg8DrqUD+%h_@J;L2X@SwD-RA@k+>V)GSSqgEP6LJR(X5V?f0Oz z&lz@GN7CqF0nC-@f)?e+`xULuS}|*tDTzvfK7a3Wy389sC?TEOGF>Y>EW*S>=?lv~ zYk^pRHJ>#W!=}=wA@04s8Q|09E#$@^v8i;5UZ^_xr6jACYIgM7n>?KMTsq{ z7%xQXnK<|Gxf(8=2W*}{W2*ALrNzQl5@OL>oBS=~+v`7gtF8gHU&j`HVJNZqjUt`W zoolNBx`QBLyPFs3Zqms*-!|{1-}JoKc!gV~VYM6D&f5rm9ycCa0%?U3U3RKR{Nq~c zy}S2aHmQ%TOpSC4o!9}JKC3rvZnGzSURrmx_pI*grw4oTEIh-hm{4|%AY6zjO#0PQ z-wAZ8!kE)wQ=%w?I%c8F;tPKo!i*F0u8Rf)5CvDdQTc zYcUse6T9>R?0wUt0#b00a|#{MT>PlG+5y#nKnFYW*{N}`?!JqIyC1gc#HM zTx9pXEQ^Q`1QM*f*3r^epzpdY*eO0=qKB56d6O+e(Mf1daWC)y80m}CKS@C0qYtpT z4u!IdEJBTpI7#J@x4_qf&Z{AE^$4Fdbihq1oO7_c9vu1i(RCFAlNgh?IfjO|Ot18X z5p@l%(LxB)7?F?& zF9MsB{M#l-;%mz3lAaN2-%g;qQ^r+>=3qpv3X}x|{u&0uRlm7vKnx? zj!E{6yAT1%^l@RYHZY>PIwbsYJ20(Qz|(+c8CVs787|5>Rw*)Km+NMxFe>vpdlcj- zJ$p?T>jb;CKBNx6HBj^xt!;vdXKTA>lT{#ivxO(3?*7|fcL4;!I;`LY%G$%xcSb-# zV=dwTEkKBB!RPz+awtv&IaZ7B86f=RTU%1Ye)_7|zD*{wt%@G+F%$}Dl+>oOHX&h@ z|BW4)OGCs`yW=1YPKl^6^QIm`X|HBs;d4muB55cy;#BRXk|v@&5{*llWslI!f@!8c zYlK_)D(b^8*Ilvq(Vj+Aem&~6?@w;SCAVn}^ z2upbQg2A9t`OYwTJiL%td|t0DpN}icSIO7C@%jG!eD-1Cdo*;EeQeF?lR&+FM27vp zZ+t{H|EI4r0f*}S9{AXeB?=MBmVK|3vTtRJknD|R#%^XZV=c-WNwP%Q%Vf>=B}$gi zLds6ED_ck*63PGd?dR`1)2-Vx&+X~ye&)RAocFx%y>sq4Mohu)>^apQn4Tn8rDe)o zZXmQ=vVkLGO*tB}>KiAy+a0e7N>GYXbrXGKllP6^E^lW9nk7<-dPnbmHDvwg` z&HAf1qDRoar7C*&(>GSW=05Rz6BXk?sa*$8bWCiJu zAB%f5!wwiFT$~P}3t*p)PqNtjS?uej$j#XLd{|stPhAF)5Ov@pJB>c|ajw`@MthIx zPyREy$ITMY-=F)EZr9Sib~bbumXJTMY=5dYaD4qcB8||y8TTyvLs7`bD!IFe5Z;hu z{*lt>zLrk$r=9iIh-J`Ans?c6FKJ56aW>5$UToF6=E0$PwcR?i8RSn78>lU-t~JkH zUgJSe_Z#O))in)9f1$QYvLQE+^Ho(SczWdmXT+0yhKIxYykR}I$6^dcb6)yXNPKoW zKjEJLUEY@Z`Q}F3=Hfuyj}zi??7ea--UdE5b+Dt?CmZ$^Cs;nZ&ZuG52VZve>}%)K zlvgV6Zrek(`JXpSwtlo(yzTPN7W`P_0Y3Zx@}rHJww9JYeVg{_Hyw<6l47l|+sce( zx(|NCvKvGmvq`9q4^jv$1OvmB&!|mgAuiB(LmL+-_MI?3KH0y)^ zBr0}#f=)-j(pV(-NavfDHoczHuM0vWIRrEx{ih1ymS~Z=lcooONW$PF@)yxiXuPK% z0*i4$<8e}WUwo87Ynym$n@+T}fz&Gr5z!kHubLm|O6VGum|l^2;K_1$iS(P&d*oKQ zz&8khkKkXv36H@d+)>^r_pPsK`4>9&ip*pok4;d<(ZK$^zl*)Ut1LUoI3AznEpcXq z!NNct&I5m;8*6D88fymJZjLMHi_@C=Y^5m_a5$HYS2eHT277!eu_#G!w@L#SekkYh+C`kG_|^4$hs=`e|r^JigqzAuG|S zVpxCP?_Rt+Jj^Md>P|QLftGWW=AfuKzjekSheiX1xAg=iAMfI2rVCl9)h@5H6I|Bc z&XnghzM+ryj6h}aH*=e}HJ>+c(&LnCtr`!i&>wouV!Jz4oBTv)^#$j_?C2SaRzzFT z*r_&rRZ1S&y4mxl@W_)Ds_&~4GlZE-YL?j8S0^chSOdqYTv%B*N)n9@@A@RqT`<_1 z6sk#E^x?db8#}dV+^zw4EX^&ghw%E$cFn#bhLXA?*X>72Hok`BM8w>taVVJ<_MbI> zg1aeCNXwk8Kg|EurH0i9JOV7tF_h5_{19d|7d$Xgb^itEuo4*EnmH_hD*~ z&f4I^9mbquG4~cmwu*=QPP!$`#`{eg99%5CyWXa}@*vM>cUhvJv_sC?I zMdPJ|_A!^T(cDyM+ z4bSo#fyQTBPG@RO>GT;92_c$W6i3MRBsD#cKXv`E-sKlxjn>G^c= z=;(6$F1SMdViGDum|G)4pd_N#FQcclcB)gFaf5>(ta>t%KP*&xh404-;?>gy>w#}p zV^zLp6Kj!#T8rsAo{enL5XT2Ga{ht$+Y_P{{WKc$C0odGrGy#daQ0hNxbP@i(|dxP zQR_Bq@;~A^kC_A%D1aQPXrZ=!OX9FAl;6~1V`QX`b$BTp}NJtdruHU_J} zE~Jbdn9mj3d?)J7O+Q&+-PkEcg?T9WnMpj>Bt`yNh6p0+mA!JzO`^GE2RxUu22s zn!!;l3^pH=Q`L>H$mRQ67F!sLdti?^5DeRgt-fs5hofh`&QBy>yvB22tkVPk4aa%r zbY+2NMO(JqUU?}pvS@WD@zQ>g_u2>isDvXzGcyb$65Us8-dLSf`Oa`S?$Tq|!-d6< zs>6!StoL@6D;~09vAUUq!X|!YF$lP6*~#$cbNjx%5^jMiCP#!!y0a7io=)8|)T;Iikc@4{#z2D$tyI)t{kGG?)QeT3rl`tjd20vYj z>Cc}y?!A~?;W~W@(@J)R_nn8>)j9MN11%iiesS%butggcH4gPa)uUr(=XnM!%!AZF z{^T4~vv-MGRx7mShVy2qkg1aK;%qBVf1q;dFVq|DGjtAk_Lz~kOQt|Dmv7{SQ0{_w zE=87Zy7bt_E2n4o13!N}o|si;Nb@t)ZRV%=MoiXmD?S?b7|y$)w~x@b-PVeB-7|aP zz{nBHe%HLH*A)9u8prQu_*?Kq#|);rIo0tikF54%sEq<-*)px~DdbJJnCz$9;H>hI z@Se80wMs+rDMu$G?B%;J>Qqe1OTo0T-abNS%V8P0+U6HkN%HE zp;K*cwd^!g^}&Alo9R3}#$mJ9ao#h8x;`qdX|$IABdwKk>KoP399z!2$!-*7(PKKY zM)-ibJh9buhTxocKFyjPj>8=+;UV24)tv?L9eOL0-x@6~S5y1o!OrIvPCO4OaTlHF zt0)SwQikoGEjL^`Iy4%q>GSOjFvc0j=aml(<4=9EEM~0e#&2GRb-;pO$KSsfcE}%Q z(7OAfM^Ls=DsMvY0#lRF^+Bbm28=@Y<6aKMa}Qc5Jt`l&4G4CG{Z!Mfz9IE#n&A_r zHbZ9D#wC93uO@j4d2TVY^M0=>Iq>QurvxYF&P5tF1o)q%?eY~U{#f{`P@CCs@!(9_jf zH?uAJGo9p%CvD}p-8fgAM|Xcm1%_Oa(%@RNM5LuGD}HwCE>@ahKMY^bZ0XNRbq>3K ziXuC~i{b}sh@OmDQp$S@XC{89O*ylvI|P}aS}|er@6QyeFCv(@Zo3WM&eHO0>7T5> z%8X7I6v;lVG@L~9(woAK@*h&97V+ zjZ9RsUfKU}Q!;lB=@G}P#a@J`KBxLPIaa8+T2U9-6m7+OEhv}kWQ4|h_&wS$Qzq9O zMal$>Ia}$h>krC{_DTsHjLv9`IutQ3AsHsUH#4Uld$eb+n(1@NmkssA4;Q?(HyndY z9V^_D402aj>^<-NSgxr)_w0$Znd)mP3$HOf%v<&q9fs%(8XDfR(^DrGSY{*7HPzm@ z8c=n-tSs-{=(9kT#l-iS*vS*d*kGMSjpT|a-SV&a798b+D= z{<+LEGRQf@M#D{CL%k}Mo^U#OT|VAv58}&g*JwX_NgsXdz3(u6*aFiF*SCBdLOKKq zx<`-7OwrA26uRw-HH1Xkom}V z6CpXOVyC#Aey-YT_h3E!Qh52an>_2HTFlUNF^&DAD_6_euJ114d1pt8p0f$BO6 z_{e5n&8q_$ZaG=qYbc{1*z{d`6cbxH$^U-DbS~iX6i5KNU*~A`AejE4sG@#CvCq?@ zDVhr2HF$kJli24Mn96&zo{~H)vdcFDh52TqufD34mnr5lNuHfhM{7u|n;i;B9)Fv* zLQ%q0z38mTTcZ>v0@i<+`yt_Ew_5YvtE`8j!?MInQ$&~F5+=R}$cQYeDQ~`cbonRt z#|qy9&*;XdZ@7y5K9O4*10AIYD<(axUXm$U@sF1t^`csF-8~Z$vbMntFB_B3b^XVj z5q66`o+VvTH`zB5!Gd`mpDx874Sq9iPzysrkp9d$ZWTImHcWIQ%s=sn2c}aeBsO?xr47?GM-pZXnnZ1Ovebvm(t#MO`i&V zrqtL*;YgB<+2+D=B`t}hll4HQWbf8{qUeXr97}r zH1qLeuN;`&r!)IH8&R=b5+e);Uk)QP-AYe+3Ug>(9(r*3Le4_&spg~m%7Pf3P5c)3 zz8sp=t!d2@v9vpi5s9`fp5VXJK`}VAH0%03suJ~b2#cK&Q!koNn@0GC=ecUGy(d^B zVxKKZx^&;Yt-(GMc&%;R>2rKcy5alkWH6D2V&D!q96dl|XNkLurCEL7##fT_CW~R`kT9p zznv!%B&q)QvaOp`svl}}y`YQ$O!hlawWqD0UAE2J{f9cqNF%`{BkXO49497 z%auWF#_%if!dzm5aVN%i$tFSLWTZjPZMtll8X}>K+!M|5?M?Q%j5%T$U%vHP9j4}UxHwGI5^nZLpel|Z~w_;z?lFX zVjbW|2((Q+s)-;75*mu(5?FtM6X>}ae2D#MJ`#$7 zLF2Y!CzO)=GcHY|!JtGzSSJ=oBMJ%yHSv07l%(GR<(>iJY&~n>kaS>m{%s046M~lGAV4@gn7v!k;7^_nsYM-W^aRsHoU4CB3Kt)Z7fBK5a0aNqDqlq!B}9ddL|g`;IJ5Ru{{b;u5-_(9LXhz! z6blE$0=Egk$DL5V;NAgjrNAy2G9ZjSjNo|&%<%d*Z=ainK)Ye^?k-Tyl`h&9#sx$a z1Tld)uuRWEaNND1tZ?o%dfzS}j1CZn*p~92As`rgR|g1Np7Y66JOL(*t~4p-(jo)} z3uYb!MK+c*y$uYg;I{$ItwB+op<=*6g?CSO1EW0!bC=i&zHC54{_IdkSnm}!0r&LK zAdQtuRAj$9I@TR`0cu==M}|@!?190&R7vNZA{!LO5AO`=IAYH3_j|zH4F)%<#ETKH z2m}Wz(vmhI-f?RYIVi-hmc*n}m$Vn_4#i3S50pQI64IuQJ0xdCND|}*o3y*%4&}Zw z1ZBI|AL+12+qdmt=+z)Gq|Mw&fuy~-c7QS9mh-nj`Jc{Qq$JXQQ9C3#EhrMyo{zLI z$POb`4~p@>E+M3%kaix}!FU@&V7B`ZkYY&dsqbK>jG-|9R#zv5ke1xtfxs=GAioOm zl2S-(k?v5AT0>F(TbY!UMOqkOhqa7_VErl+Knfv!y?h7q)e{2p>t*x*LZDt=-hm9@ zw;{hj$J>6DnUqGFRkK68dE?Kte{*a|QKVTgJ1Fs(Kch%x>4e?k6AdO4#<1y(T_Ob`5t2EjTh4Qxta{|9uo{wn|g literal 0 HcmV?d00001 diff --git a/skills/measure-code-complexity/scripts/vendor/pygments-2.19.2-py3-none-any.whl b/skills/measure-code-complexity/scripts/vendor/pygments-2.19.2-py3-none-any.whl new file mode 100644 index 0000000000000000000000000000000000000000..3ec3f105e2085a0466cd77f3205cd4100bee04e8 GIT binary patch literal 1225217 zcmZU)Q;;rf&?NY_ZQHhO+qP}nwr%sYd77tf+qTX9W@00D{@IJ?rY^%!6=qWZC-7PL;&7h^5Iv-a03dMUkwQieJJP^$au zR8|&?g_pZKJFzeSl{y=Z&0s=b_w1QvR4I{=c}1;fajR|}jc=VQjkQ=W5>0ts`braV zmE%}FPMUZq56*18W@Gxh1!irPt*b6N8rMI3VO2D{QfgQ>vaIqJog%J1tc*WWm;qzU zf-LDaDRb)fh~R>8dLYUzc4RI_W+kxetqrA_Fimi+*>!2h^ucWMr?Oruj?(RXWOK{B zF;9J9&vmq!EsA5vwh!g?IAUo2Jf!i{)rHz;aW;Ha+C$b@t8H;@x~`1n$GXHq`!z>i z?{amCL8=4lW6)NTX*TF17%wR(X)Qe3H-bf|$WqI3Ds7`NZ$+Nb?G(P=#FC++N`!6+ z+SlQ^sWlmuC=XV-c6rwdq`iXRur+wwGrKz!t_&w%I>H2}4#?hb&!?nM;wzA7{g#nv z6JPr`y%pfJ3dhUc5j@OaVcl9^BgQKuej?&q2|3tX5OaDi`hN zzXisAFRJ%OX)j6PfMZUTEcB!-?d)vKF)(NXyUe1G=r?YLeuU0^=?LD5g-Zj*U%>vT zYw$gB8IJ@u|F!bipx`dK{7IrW=>6LNdN@9Q)hAZ`_Sd9~;TD?C(*I`e^ZUIgietot zQV+Ly`q;1N@DK3td%W2vudGBd;^q!vNbRbUVz+cvTn|~EJKFvX)xV|9>3%JlH@+X+ z5FCpE)0=-|{=OM~a*pkbAx}t-gh@E^xbXn^j1CKc;P)(;$YyVaZ;a(bN_;PTd~u6SgTDvf{i7 z0~K+^P7jF^#R*1dL~mO!RHxG8sr$!CD)@q4!idH7CP6hSc^-XDoS!M%{1jQR+%1Ie zS33*u^^=4oaJ;3ReXHF&-!=c)%y5>Tmp(EmwvXk=aU0r0u&jxQBOh^H6w2H_7w5hghAGxFie7j^IIidv5%jXF=d!g}tnJ&R7RF%kV|hb_8-#bDAMqGEs- zjX84;ZE2kBZw03Tx7NNjkM%(g0L&+H4XgCIR=?Z_4u0Or+HpVF7saWevI|UB033ieml7f#=On`l z2HX!V+$_A1cP=P?SG5yEzDRNgM9P;mAT)IjmBknfap_O0q(2I=rkj;Q+KYKkIq}On z=A-5IkN4~hrrB3<%i{Fi&@b?#J?ZAb-@O_E^2Z~EPZv4@V?cxaOES;pc5?yrc{MKT z7VZ>HAFKM%V|}4&t1r9krw;8}a}Ytc;cZHTPF^kcJObSV=yn!2?5cEm`TPmkC;rX# z2|@9Ha3kb=G|$d|oTv-*|KmnGBP;v=&5e4BJPQ995qG|67&w26>m*Db3=m(dY} zxSUc*8c_vI4U~HBHl0v5h&b|kGj6&`r7(vYi&QGdAFGvxFRHQUci$V3_(j1)TVV|2 zmOyt$g!DEAd*&Nd+grL$Qc5gi;pE-<&>DK{6D`&cnP(p$IKd0?_rG@`LWQ@3Btk3x z%x1JFBr7n5*ed$a8mjRteEE0p{7uif4@H+68*oVDrNIuf{7CA2ALy zmW`*5w~&hI3^&+m?3f#L6R?`N+bSemGvdVF->4&Zy)m^)cPVHR_@Ditd8$($zyJX5 zbN~R9|5GbW>`ZN~?9Kj@ie=tS_stRa+mC3dZ-62Yk`C8+2sZho($QKSZB=^i5v^7X za1c=xBq$~F0P&$>Cy*3Hzio`aN+Z z%Kj}J;H$^8bK)V`OgI4L^5U|X_CWL(`r(~=3~V>&cpZ7gpYrC>a9Gu5h$PTTf!#Xz zY-Ic>{*-uTVSLC0rk`49>cY=YR|bJo8kk7F`9Ms)JZW@B9SoL#=8^VQGc~Mu?9i28 z>n`x@y3r%_?eCGWEHS=05^r;WoMRP~q7PPu&;ir8-JptPc;^$U!y@B2C~pVomww(i zJZ8P;P9W<(LSTHj+f0m9A}PxIQpxUv~=zar{A*Fd_sL#8DbDp3pfR-Q2LXBkdE6_>%h*w0RUI+pj{_UrWg-I(lNRz-%_ zctR8Dqz^?xihS_+C^p;2a#RN83X}s}h636AS$Fy4&W`?&g{`nP@3>@DXpT^YE-XS6 z*qvenlh1FE|1e9sxgl<4#8qHHE{0jYvVnKFde}#xn2lT8>7A03#!Z9uutH{;SZ?|j z?-aSA4I(U?W!g~KrRU7tQzDHpLnpEOpIIuL(T6|!Z7U^XtItHvYjg%acc<{AL$T?X z33v&7IPM5Vy?5;{Wa~1skO4eC_YsMIpkH(nQ*&wcAY29SXoAnN%Z1OYXA=XbZ*f5T z^)8>Q5~hc;bt%tXEZTG}rSO_-Y;;)7w~LFqaLzNW&MjXeta3$0Tb2e2sK7LYxth!N z8>Zn}D-+0h;&e6O93nr_-=-Lo(a zz>7_L)b&-YyG;q$*{_j15LM2YD1K~$w~2R5p*Du{#QCRn$N9_0%u!-O&Q;~Rq@m91 zMu)F~2LK}!LAFXv3~XvF1q&@R`a6;ELRsXh^fv)cDoB~VUnuQBZ^Gp#WUckIt(c(?LjF1YYq0_ot!MPTAYhH^kd1tXk&&&8Udj;5GKzKN1uh+r35mzruNq zYszB6aw~xy|IXZX<2M+$rbL%A2%|s-$i9=Z7?UDmKGE6(ee`m#EiCQc1lTNcR;m-~ z7M5-Is-_h~QI-DOE4)9OKRruMa;{l`6jDsf!(Or%hz8>)y)mAVf(=)1QaK6jU)zgU z#=tZ4-jIvXnEo7?T56w35t0W))4I9*V3@{%C6GG}WN^c)6!E_(21Qb)6r(rNn8Z_Z zoN(Gw|HUq)T4tUR58-W#&G@VNr(?~&%ISfI4+A{!R;TNdGS35kS@{~@-0nPPFTDp=}n&^ng4(dLeSl`r4Ak@oX$PksvEx92rJ(+1wWI4_2{sgvGi1F1I zmWURwHN@cNwY1#MTo+H-%|GB!XSu<| zQ)#V}+|%w(Dd0H;7h%5n?9g%jK{NZn^~n{%NqWC#gZMri@T&nOBoUH6Gf(`+0&Ut6 z^@)wTZ;>d)?dM2LXUsIdU|Q#_it(datG>O^+zi%jnk z*|FR7XC!lr#^NoM4$*za-hSyu=?h^-oPvYH%{UK0+p=VpuFb{7Di10l0P@R0OdMJY zkr-}}+1UN^6Pmq&P(0Sts@v|5KTHh!Fy-6oOjM2$5}@vpa+#I_7Y+} zL!s>7GPh=fiRk{u;%)d<6=jYM7qwG33{$B#FTdWVk>Zj#dBd5qI_;x z{4!qem_wb>EgkO|K7`TAQ~u_yM*=s7#7?rdRt6r)mVQFuPYhPT4wcr@e#uJACh1P* zh_crfyN<>9wAS*th!!oy{B-7!VBTYEy)wCyZYgq^moZb3{L#-_`%;itjSTSdgYH86 z+@5xReGgPd@O~L*<+1`|=Qgn(wV$oUmRpiHG?7s_@Cbk3UUqZIf85}uq@lsptQzUW z)i7G$z)EMj$Ge&BKVbc9q2f77ULQ7$CbzgTY6nxwC*JJ{EOuCx&_Eiz*ShQg+KPQC zlDo&PM$^A-&1PqP37z{~K3;ob*KD_tx>#PGs}4#yu*9Ul`<`*|dp?Km$c3D%wVkw{ zqAp<^2MFRmll-_ie zKD@Az!M7Q$`t2U-P#o$YEL&_l3(f~?wFn~5zS_Wwn6f4@my3~}E|}KE`XQ6`#(ev@ zc>20Nj?thPEnuj%WZGQuJbq9*kD_%OZKo|&b;Yah5`|5;zSViOONRvyq5*8RHiIQ_ z*mCMb^4Uxv0G(@SK@W8sHbx~l3)CAFM%EAtQ^GvH#aH0tfN3+TRw}Nx6zbTzn<(V5 zjRgIzHLWaN5JO8l@`>}j3(A$&iDn|UL?d~Ruc=%5>6AB>FGy8;VPzv zNaQm-wn`_n?ty4d-h)Db;iBSOh>lC0%ZrT6oGlFeQgy2thKeGuvZ{uB7deJ#EJI`?q&TkqjC{bTNmuEs=*j$_Wv z?Mp%TYCSDZLbutUk6M+lSO+HQ*(lfm9SbhPc-Mon5qeczY6RNUL?A<#s<+HwpEr{y z!IdMxS1upWH9OiKYzDbwBV+ZTcTMGAYUXMh3R&E|-E}akx#l0roi+rDQ1^HwQFO9W z-C!LF--lX0VQ=IaTBHJ{oGqp-eJ>+N9EWmtH(xI2%EbYOj{{|PF&>smoGHxd(zjQo zjk9mAs=*fHb>u2_Mo+)+@a@Dc6SvRi_!v3o)WAFB31H59+hnRl5jEvQtK0Evy{?aC!+UNsM zuJS=}p_4UHpuLHpigM3nts=us3PM_2FiDTel;^QoCC{mIr_D8}e#aiv=|`2q+CgCF zEifavS3#zZj-OujQCuDc&sft#b+t1x_l~qP1>(#$`XcEL z*C&N2s>W4VD>{D&wIZ}3zN(_s3^#FQ4^q_YJ?p7kzotfX4WNEHhLT(>y$-}c$i4>R zE4(=HRbUIiFH!8KT2Kd5UI-K0$`n_Pk9vgcuxbFKk59%!tkrRTup&$6I$ZK@1B0Y- zL?03*M?qYz0pnj@-k@dNZsQffoJ`P zlID};8;);bX7uh}mCJ3^-p;`2IWmeTwNgt>^@ALxZn7@#Y$G1voxI$r9d-vMpJe*T zDT)rZ2+LSX;zqN%MaPk{30J_u@F_ceZ2br`ld>e|tIDa1DG!$2zcDg?;_WIK z3xUbRp25>|y#buRyHsHRvsic*1&yY8P|FLjyIH16xAC$>5j)u2Vx>`F*AFH1-nPGsRDVC7Hb z;A*w|2z#7^9g-{LkP86_>#mCgHJBDCtma&)FJPLJ2d5i7zqn)h>cwAhdt{zNBu%vOE@U8@jD&1-L z$YyC!ND3%BgoL7EXB18|XrL^XF{O^Xqv!1qlsfy~@VibXARj%OH3L;LC4$IknVf^k zoQSee28;&a5uSmF!mLVS2#MpR4dzPah>ActP(g-Rf-C)E7pUYSHOK70?Z&gKyY7%W zuW3RWZg;Trqi(`FVytrDOT<;_Jt|~{*!BYQhs5y7Ya2akCpT!%SG*DE#vx~_9d5J- zhjp`2rO}(p+97V4WrEMfu^pbV(ke@@0GQ&Az2P@8fQ%0;1#U?k^Wd{jT7e<42eR&i zfF@t`@F_J#*FG_*!TxFLf}`*T z9)`jC#?FRI9rD1ZSh{2DALtoOR`ER9e%tRRha31BQC&O{swT?)W>y$G;f19Fc^5mV zKM5aG-OwsMtWc%Xu~XR zi}Y0Ed9>_3iRKLM&O9L_EbkJV+@a=fNr%^sP+3xKRI(Ln$pm_KvZISZP-$(n%{J+rWfq^#jv zlSAw)@2tV6e!DoGIIVX=XgOb#LGm5Qw^QW_r!qtAZ#X<5+lmdLp7{0XQ$S*YQV^Vo_yPpm?OrK$?4Hs!Xd{>8%5GUXJMqV z-8ILg9cKa4>#j`xxq}J`d#VpN@a^o%Mzp*-AAl#XZ(m zlM=xpL?v|IFjBR|qk?zoeer5eA&o?nK;KxJ?${Ghut$wJ~4-j8jcV z<{s3!uS!M|sUW;6xfJBh8-X8^&^ke2GU)7=wWM(rEJakF2~a831>z>a=E9Oq;~`Q% z7?}2&*ePN73%Sie`qG1}+6FuKOOuI-ifV}Cj{_u->C2EWFUn|!EtMcpR`M3~Y29gq zaM*7#WeAz&VuJ{gAAB0S;1n~AhCU1LRO9)_is3eG}uulVoIBby3q=hu-IA7OU zPLoSoN=d!1%6Il02rlv*7QoLBRI-haghz{ne~07JnA2MSB}Kn0TJxsm>O267!Nnca z$uIcS7oGVfFhn9Yk8#I>Kk-iLm(BZ!>jNi!(MZ^~K!M4Gc%wbXfoy%vmhabLBx6Ag zCE~+esk#ve8Haj;r@-|oddJvl<7dh|z)qCr33fn@UbXyb81@vqj>aqAbkJh7ACo?W zfP(|$hNu%iQ400|pJ4Rli*mkYQF(Gs_opkD{Z)(ZZz&GFX%6zs58u5{G zVZq>iDyFe}PVk;;k^)@`DvR~kA`)tS!V+sGDh>9YU^)QAesQ4W)I8n+rk`GQeB~5% zWleQoMq)nlLNK@ov~Oq=A9G+5SCQo9d6=%54;4|>zhPF*FizfbXQ9yeg?Qk>0d?eN z-dNB^x>14P-*?a1=-pT=Z$IjUTEoONDAW|4|K>VdV)a87rWSci-Zj*7B!=J5B!mjBU=zsm4W2GIou`XFZR9_NAOZiM=A?6Ur_g&p) zqp7v6l1)cp-^a4TtEx>hR)M%jC2$XM&#pa^sxd}WCV*}?&whR$-pybJP-1VNe15N1 zx2iCjje=slomMQDMBx{0h#~KEM2M|Kl%9j(^}|ays{3P7bBHZ!!d*C0 z@$Qdu6!&!cwji78=-tV~Gfk*|st0s>8a%a|cvl&64RMJCs zPgL*OJO z_;@EHcMh%tx>ObJcY@(Mw44m#M@h2uLZ{qmGASHIiusnOj~DAjy+CT4$m%$UJq-SZ z@z#s4r?{`yGly!A*w-kES!5E8Mc+n-q$Hy7eNHZd_;v{b5rtQApw`AGY6|cd8 zhf3Y3OooJm>QGs+fkwnKU*DMZfctrxp9ae4G|sP+94^jB=L?aO-#Gs;tLLzE6ajx; zr?kQDIvt_ zhw^9l(m)C~T*lHV9#mykW8bXlyPuBxd6-|$C`g0O=Nz^x6?}@WY^ftcKYT`MtdDJ( z82%#^meNk^vF+z2>Mz9qbo@t6e96;*0f08}|J(6*uy=8={a=oMnd+wfq9CH*t46GP z^%bjC!y>z~-ZPLiQW_#5v;vVd`wOX-@vZnv_K4Ws_0_QZ6{0HX%diWVg}J-=N+=@4 z(t3IgYUfl$(q5?x^N&ByAZoaZ)6#NhKs9bSK@>objNf?PP4-3UlbDc4*RV z!mRku88SNA7~9||HH~EHIlx)e^Zod76QKY zb(AJVsjcugp^dwY)O+|{g6q7_QeJhs^S9a8pWx9@Dmt7{FB5X#03Na(R zg6(Cix3U*?9Y%I)TNU=Q^(mALRBnHL7~_%v(axI2M!iKM&)KeE;YpxWcWMx{D#D*? z?csCUtP&|1Mjflm$LbH}RI!nw?l!b7z&eVzc&1$gZsll8%4j8=Y4yc2;WHO@>O(Hw zG|g*esul%tl|SaPfg%5D?>jDjmM%!^b0y#&=6PH!x{#znjED(AqynpPc>SMSczUw<7_Usrm zq;C6W*?_W6bF}SntCv-`h+WXbe6pw9-T#SV_j@g^um4c&4gmll|DSg&b1Pd{Gw1(6 zoTet@u*nG9d#d5MP9(pNEf#cbNZ8-S-vF$`L3^#GX3LFas1r`g0TX@Gp5XnK)*5l5 ztVJx4yerZ4`0fk-WCOnmZBvpWoC==S{mJt^y>*O`)lL???gUb7E$wRm6qjf+u(=Qz zA^F!Yd;Fwu$lXG@<0{_K-~HwN>FeqX#TI_%RU;As+zTy4JTxZa(?KBM4+?@xG^|Y= zm9L(``|E9c#;V7q4P|NZTzHH!MKjkjOO6tGe@;0sV?j)OixNDi@2KJdfoctp)*-VS z(n(Mhgk22^HZ?0xR!9;)v?pA!?|b+}hFFB??Mz$CMkGzNKwGlOaX784uPoTf1q)pA9N?!Jd>h~4i@Xf&*KRvB3%4x?*r zV-nTn_0k+M`P4O^SsQUYowpe+@q1ZikWnV5>0H@6B#>dwGSO_uakv^8B#5eJcfCB+ z3>%%x?Wsg&HlCXsyZX$>=Jv8QBeyofgz?VH{bES+lNw}mGvb6pEE>>9@&zsKz-!{M zvhUswdv8@&bNxI?Lfj}u%hsCh2xeKmC%+W$&21w0b8EaWN8l#&8ko3Z$w)Qd!%;~1!`q& zINV^X;lS0twOsfM{mP1#|7m=GwPZGBHmxZ@O|TS0;lW^u(+V_w8R(27|wEMvo7}legQA`Vs>&o;;4UB zX}38c`!;b(+{+&+R&xcY_scQPP1^--SRICq8#kTG+vo*7qcqT4mS&vVRC^ z0y0<}Zos6q0;zUgWZjkH9J$7xzs_ z3Q36eV6JuPVLGdbsq_);eeMz()V1W2&4Wom3vxl&^zOa~k{c{|iiy{Z=bbs{u02*6 zVNvU{`D4Et;^rUyQ)yt4?cc=8GGxo68s>nke(8Fxhz=0H()qwYCa(u?mbPG$xWdM+ z2_Q#X;R`r8RDNYknkBWzpXx3Jx z`{4_FSK;*?4HQ=p=V*7S5MXYdvVG|oTs{uG)zuA*V*X?EMe~IXbxXQ%s^cjZZi?0y znTiuiOC~A+47oeKyS;hjTLvvN=|RhBQ~C#?Y!G$3icpNP7H^O-Avc@5Dh$C1DTsiu z#Vm^a_yug-KQ@GvZr~Q$OkrNJT_+P!G5o*`->{nomqfIc8sJ$Vo=n7DkbNDulb`^K zp}$zAGV)afwngFMrF&|!GJ6Q=&K7)Ul4Zl+J$I>TGk#Vt6b{k6z+m?&lYZAK8JrodP zdY1Ook})S%ra-fq#SdG}+hk$DPm{?TI2+`cbc#4`+n2DS{eR$ds z((>EHFh$O%XkiM@ZmMQ1$ZbW@cT~vQ6@s>4Sy!#1e9}JfCV|e1pvph>kjl(6Prv3Z zE!?`(T>aOJFRzQ`6sjEsgD+j9they_2AA>uaZ$tkT@lx*@2k*=-a+MZX{#*&9>gHVXZwmiuyv?pS-6 zGbv<=Br-6FJ1vK!Ocn@&b`5aKKX@?!vIzuVgOq`fc3L85SqjgbcOjJBG(M-1Qn{l& zT2m{`KKu~hN6*2@4A9XOZWCK*5Q6ht{H(I2`TBKf%-4!BlG!+Lb+%j@7qU=w>}4yh zwOAi3ixfXVv?GDvP=?yZ8HP{otx&FHN|5tEF4P<@N%*oz`(#zlgt%{*%WsaC z9P9NdCnX44ZQb1K-6+6L&VYXCWSom}Cg6FqH9Os(cXz851TH=epUBSN1d}5v)ySgKcFS9%^`Kl!-$oKO3`#!wk zv~E>>NY@mYVyazwvM8mkT#R?O2l#zI=1_%O* zd%K;8)dlNggf6-mJ3|`)>{t46<2$J=?dv)Ju02ZN%qw?vy*}Co_j|P5UK;bNX>etBWX|XHH%p`qrF;vpNW{^cvht%B^~AAE$iaP=i@@6Ht&KgcdCpO5SO zbfBe02PA4jLwZQ$UO`z^wmI|G`De8xQLVpP+`Q4EV{4U_V9(O)I>Mkx&F%tyFXJ(`sGNG~e-7@!dp#sPi3 z(i$4nGxdjhU`N0!)63j62iktI&ER2iS)hJFVO8(EtE-P*Em{NQBXnF!GKk7|62vYe z3SH_X0rVu9!P4eH#3_#r<6wl43 ze-_vbX=5HI$$OWG{dWTkW1FiLn@csP(MgL?zLhM7(K+6Yf&Kmnk`h?Fkw}KH&R5ef zv0bt4`2`E*#qtL1PkOqc1%SS}>6K4{A7xlm=gq~cYlsEuq=bH!S4!5V54Xu4YV%aK ze2AiP%{9u#lq=I_>|Bv(C81cFTT{|b7pkm-LG-noO zBFKVj`W&>Y6{<+8xLSzM;TCEk5dD*L0W%@P^uRIX6%D6+097&kk;=Qd(jg;8Cd?!5) zU%_vUu@Bxc-(?oCy3%V2X9Uzy0~=;Y#31xW^@Y|S2cP zL*|y=&pilbDEcjs-GVxxgoa}`CqI{DKRpr~Ncd-QB9dbt#Ypl%!Z2m=Tq7LX?Qd;h z<<3#TpJlTTq*|t-S+gn=)^Ed7-8|XU-gV@SLJVeG0wQ_dsfM6A5Z_{UxUs)vj-K_h zJjuE$>Pau{m#C89*OJcB5C0O*q#d9oT~RB;yfZ2+>*W-E%Gin-83g;y8!|!PR(RVR zN-K2!88Z=gJG216`ny_%oZ!7lhri2qdcN=96fTmyl>$#zE|S2Wf$iE=TH2qXIp7l? zsc4f_VQYZ_f9j3r7})|xs(>|j{Jv_!QpzTo={(MLQhUfhGFq;Px<+{zS7qll){8hW*Kw%OxN=I3S}%<#ZJ3seI1zRy*1V_9~vM z7RMyCm0_&Q;r*5XIg%2r40>>*g~4!+qT{0I#%IFquD8tDK6y_^_QaZufO>GT_ifuy8de|W_t8{*trsHGM8 z8-8*_LFhF`o18(wvP6Qy)Nf!f`>TIGUXEdsjhQ^B8|;LVwH{{`#l_hi(PLf_Dm0$S z|9u}IEgaUp3sLv@{ybo1HKXNt2U`|AJ0(Ne03*C!MPbuWg8rsxa-*<3!bw)6nk7+! zoB-+C>#s2Zv7Q|#R5gmk_JB4h(4Kh+^fln@7E4?mV~)Q;f|_Jk(9Z6t$&v_A8xaP< zZSNs4arl+km;zAmbLe0%NL`cFDHS}24Ru<@#&Am(=w5kZGWG)cYZ&qQ5j!yQ_l4d`P$O_cxZyX56Ck z5eC7d^-$0;=|h;>v>6rLrw~{7K-IW)6qf!)7zcalXWcvAiMZUyawyIaD5|ARwuqw{ z({c7ML-dF!5a8>xR3DL+)o5!E$}2Mz(n6VMG0)eeP2NaM?}#uD8dgD*>ntvW^M>NN zOLlqvdWlR&U@t_JjEdxBScG{+dZxK*o<$-i3@1L84@k zEgR<-CMTcKQW-GtUeHN)4PF&H*+d{(B6cAk#(}P_WmuK%q!0cYd5)i^A=XBj#Qorc zuh*PYT(B0T)Mn2z*@6Q5@&Bf2J*OZ%#yqUvilf(q%fC&R0C>BZ9oGN0o83o%y{3>F zaUixg$8#*Cyt-JTv~&Y+%#}#dI7188i@U4BkG4aaR{E@GXeO$`dI70#Lcp0bIXi`5_zIKB#LTT8ej}}oQ?-Q3%Aord4(!)Zsqp0 z37Ky|3$T7yfMqyv@(fWKH>GI%n^|!oXTM&a4|Qt%nFzbiLbGk4*bFl6+l={HumP^+ z6~jmojNQeo-0b#93=>E}V}yaSgj>dq6V$UP+aqL-7!SpbsIYOA#9=Ye6{PrnXD1Ai zxq`)~m;Q)`3ADc;+aUSQWs1om0fp$K&)dEW-b1FnDyqmr@O0w*#zP!GgCjnyCv&rkHuNH(sJw7H;yjha=FvK6|@O}wN`+4xod{-E9ID-g4OQsv?KPy z`-%eR>E#hcFp1sau<6lH0PpyR6S6kQ<}qUJ;qHNBv5VPukHj%KU>A_7jv(~PhRcjT zd#R7mQbmYEn11WRdO)S%Uuag5msC392c@wmQiMLg_D^<>Ltw2n!0>R*w?aM~$Ppiq zBko^8&w6RzNM2LslaDEUY)3rLXpoPjjVmUbjET>!7CS<#7O$HBGTgwDQ_0O`GV7qQ zCmNfz1S4*kte0UU3x;L`E8wF0V zhPP@MD%!#d3nstDjm5xt1A{CEt5pqmXWOS?N@?01wSg8V9qkKBTgx?~FR33W$FI)@ zQBQWhPaZUYhA{?R!nSQ|m4-ZyKyJY({#h1>^=6kVrY+ZRxs4(Q`u%K5f5=fEcN)P~i-QhC4Hk2Q z4@XN!^e0Dt-#+gtEwOrbW$(8JM*|PzK=3{5s4KC6B-s0hk&Y@VJIB3Sy^BaaTvG)p zDu~8vsf5`s`JB!9e@6^JB=2AxBs>Xv^1FS?_R!goQ6ev_q{BguYltvDE5Qj>P#KWS z_tm8NQ%$kIz2QaZl$B1xVzpMx*=}za#mhApe;ga;7l-`pNXasa8hXFo1Es|xse2Su z(uhE^kr)=i&@q|ag<9@Edep^91q(z(DeDm%!Vfl0M_MOzm}Ur~xFd zHmdB^*t4igqlFqn&i33F0qEL1?$Rz*^!ae`72PFD2>1{x5o*oLWI-?x8yWbB8=D6w zKCiHAEIbvKzK9a$ScTLXmK9>9!vZ*+=4Ifxf96jP&-Z(yH`Tz?=S1eb)I6$Ko2}-O zvEeHsbR6SpA??9&Bs9*TNHZfX%+7}`{fK{6&&jFujq!ve&nT>A5=ut|f$Q13Z5R<_ zJbX3llJb_2PHl;N?9DHso)3tx79N4m6@XyJIRD+rVKm}HuA+s_;c-_pph?qmC>S^x zjej#TCdS;^&^!(SNwX4>@xNE^%`pPuxWSq6INEGk8yDKg0Ztw zKoJ*A!fNn>Jz7DHz`@`+9bC~za@ELgsEwM-ku!+IE~SCc0AWQ1iA81)(p8<((`Ne*3->1;0lGdA4`;Q(o$&~i9 zRrJg=*>3&$UKcT;oZ5jMru7u|b0})Q3t3{PpvqG5+IGt1Mw2{p7|5O@IaEy32&8_`2di2F zJB@)-jh3K!02Ogc8<8a`lPRmz(w3QyZL9aXUN$k=(6b(61e!eHo93c!P0q}vQXn4> zHUds*j}f<2erTkb8jUO685+8f%VQKus^-YK0w)F%)TZWsBSeg;i$X}erCGsr1MkzB zT}n_b!BUxmR%t?MHB_eKR3wD5n(}s+n-rt8V7TN$!SDot_B7l@f|o7weJ!a>&Lswb zr+88iW$dT=fJ~thXhCVBs3Bj5vtgb8C@bnN2ct6g-g6-%+1|MMo zXbihmlqo{F@>~e*c;g9d`!?YSnv&ahrgn(BAMTD2v;YP@4P_o83bkk4-C2&3-l=V3 zk;j<_4MkCg{n(&Da)LMY8oJBAko6&>jZzs7Dgm z$o51=4<-nj$5laN{ad7kR?5w$RtCb4YuJy3<>j>pktbZX#UQk403a%X3FqTzjnlT6 zEe|Fc)wXe1;0S{1UedZqJZqmi;wvb4>+ZKzs9v@u41qjYsh(~xVC_WC(j`PBnN=^b zK>J&4O{>o?ndNGDPJ~4N?exU)ayC)$pQp08T>}E1_7D$3YT0cv1ZEuY&T{j13Geum z%@$x+5@g{&qBkh9g_nPyf&KwWc#zNw+Bgc7Fou5}zcuEbyc6l(wj+46SG_py%IkTy zbs?kE9QL{5mL18C6cVn-1ErlK5_lA~+bokdDn?I551ZZmQ>8+`aC^Tj7aWZ~%>?-Y z{i|bfZ?=3aefFCoXDg+)Y?djyQPgTA^c#icO5IF$YhboV z6PYOL;J=UQ#AUY=Fgvte$959F6TqanYU|4mFBQma$)|dFWRYC_*Ww_$0g9#g z_~1dFEH_ozHVPG0_2H)`IJZGA{GTU-^mc6)c7)s8OEZ00dRrqWC^KbE(NK&=DFI{m zvXB}~hs^w1nLsu3r2h|B=M*GLuypIOZO)hpwlN zj*hD8)ft(&zWN+DA1K4soB-A$#*K1!LJ5Uc1I@meUddC6K6zUWGA_dNx6+omm;1Hm z_W1%&gMg&SiTt`ck{iowu+~Rs3a7XD98T|zS}9b~tPz)fFp?*@P}&gDwc=Fz7rqEP zv1KN+RrE60Ve{Ve?$n%EG$~5Gytj)mcNkezS_AWie%cI*v5BY``7T&Pt=A3`C1T*G==KFMyIBM&n43(WlfkMt58G2l8me{_ zo$KJ^WlaL#ubbj#1W-GJenDRY>u8w_m{Eud&Tt3(()m)OJ3X8*xFdEI<(6YwJU8hZ+=-#w z@2IaIfyXkEA1%qp`WeD(D>!zJLIF?(XB2{f^s82o(3N4r?@zTNQT;GklKY+Yw9nHG za)h-KQE41o)W(nm#*}xX@etb3LjLod8+AVMD z&dZXwJ?)*Hy6H%J-}5)A*^~TH`gWEaBdK(@q{`X>BQ}c7xg5R18>dpPewyBGCK0g7 zx+VqYR<%j}s*q5YZqtF3`YHkZsJSlh&MZqdGT8UY5e??`*ZuIRI#bZ!b&-(S z%A<8jW!+RrGi+hp-yf~iO#=Io@nUR$7)+xu2SYq*Km(ahx%hmDF%`?-i$BcVY3G73#gum!;yRUw2xkfYR_8~#OGyo>tUDqz?rJ(O;6FI zBxRyUEnD&i2l59a-MvEE>q`=zh?7S+FCgkZPj+ww$0%F6srwX0&5!=3vC*_65T^|K zY!Xga|2#{kY-k%?Gspg#Xc9{WfjbL<1Ch|?*i9Jn<~DuI*}$(oKP76j3xb8t~xEkSxVY^m<}O^5O#h9r!=@;zvifND%O@vQsgb=<;Ere z-7HG&Gg+NT>^UX}N+eu^^^s`WEvr;awrT4WXDAuUA^`SF@^bi(a&Ub0;MtePEG3?K z(f{qEa zTl!7rBS*c(=b!u&cYo<}Ms7wlGdnc~xU7X;CFOaurRGqvotKn@lcmk_-N*}#XkgiM zJp=0nW1p(?*PY4iLF7J13C=$v*xQmQ98XrU-cQJr*3nLOgiN=esilL#c*G4?#En61 z_V{N}UBlQ6NXzsV677KY?&r5@F(hc0EO?^zZe2Qu5or~)Wv2S&jRzgMF7vDbCfgyW zw&Ql2Jr=Kpvy(Q<#(`* zUj%Yv$4fi^p0{0cEN&1O@j-fwm~yLN%ai4r1|rt*NH1?-Egkb zORY>q&+}{)>2s_K;I!H{}Fr+)xCgmwP*Zs57suUQ%6;tP-phGtr-(Roxs)9 zeMyS)K4w?&I&!_I)2$a{An1XE$>#KGFfLmZdw})jKYL1h=iPgGGmS zn%1nBucZ|C@Yeo{w)gyuOtE4=!BBp>sp7- zbvrtxr6OnCePY=D?)pC5oTa7d;hbLrm%5eeVH=FDsKV3o!Xj+W0W@pL#UNG9k)2u` zWPSW+Eotm!ZNQU;UI}jf)2WSGe0-rLFDY;Y$RszAn?mp~7;fLG4L?9$wjfFs|Irm* zXCNlS6O||rQhp2p!akAlVbieHJrSne5rUyEIM!17mdiJ5?QxmBG4L+m%>r+@SJ6!? z*UxhKns{Sl$elO)c&bBNGn@?nJt>j`yy()}u~rzlHskN`lEtHDB6I?}*Bzxh5ad1S`S<2>k#a;2Hm83#&4F|idn(lgn>f!I>3@mkvS}F$lC|s1s+gg>> z6SwlcS3d1Mutg{6?X-io{+PiOM~D<6YVF&ZtM+qn%3(*{eC=5z9-u8S3Nc)gHh8~Q zx#bo<0R#s*$riJ@3;FmFem>58BM#519xJE2p4dzAnH&1#=`gx&cJ)37eQHSFFzkEQ zAbi7SV59591A&epa&G=-qP=w*vuR%aG4c^70CU5&As;mvxWPz&G2}HW=ikWoF5KFg znSS#Vu0?%zr@>!9eQ!i|_vHiF*HzI^d=$3H2?7B2i>hhku!y*`DWZY*^C~aA)!(g{ z2AVuuKUm;iq51gVcW+&ut^`#EA$eGU4-2#v2It1Pew^F9oiDjpwDKPh{PNyMmWm_Q z`09|acoWf_hr6en(BhH?{w|Do(fNCk32dEseQgQ?C79Hb>X8*T8M=@U^r-op7~K-1Zcy(rI3z zcV(OVkF6^=j4SkJ+X1xKC?sl#@VMh9iy%L+eqii(Og~cSo+sW4Kn>V2_5%hVrcvmR z0e3&ChcnwfQ|d@{~VnmwE6+M@VZt8ppGbhSFq_A`td3#z)z>tWm^t z4T;V9@c@<{a*VdQn}n;-)!sd9xhm29*ZwI}9c1|f?bX5OP%d}^V7;i)+`Wu>GRC?3 zQ2<+hsYE1VR&UVKuYQecCa0cWvrv3H=lMKvh`zo-_#{dNs^=gKDyfuCKFb!&Y`@=7u76BON=_wg#sI83>?VJt8PU!5ap!&ENcU{Fl(S%pz*MjrHeg4eOwPcAAN|WMRLE zd05m<{FPDD!iBA4>=w-hq$@}U;Y}<*>E!DSKdlt2WiTBAU-A%KSY?IRO{QRaaqOtZ z2HcPxiRYlQj-yBIMcsc-&Db#Q4`DYg?*`R}gD?N?9!p>zb4{-zKLiIhDc|aOj&EX( zZ&WF5U-$ic@575fi@5cdswqYV3koV}ddcxj5I7y5AWX{@D}%&)`~^V;b5`ObL!IX& zM9rloD1ZS5DU~aC)?C03tp!5tVnK23pdb|!)UHo^6lD2IB#Yu!sI?w=$65m1S0k(` zhn)8MkU+okM=5o<`(@7T1-tI=Cg2#qR;!SgN3mJ_r`qWNqmAWm&O%suSL{GonN@)u zU-EFqM3H7HDU3>3NGj>m(}=r9=cFt^O%8~gl%P8e$8hF}`?SK5 zNPK@>@Nce4r)5Wk-I)r2O^Q1D4ngiS4)C)1gq%TC8kf4X1H)h5oU+Hn2+s7Jrmln! z5Sdly!qPXn$!yy5>U2|yMb-F#bW`ZPb3A``8*ZBo9?Tec*>gYu_E3vvQbIPlCK*6@ zzfX0Zhxe6aQ8;U;R%l=HsB4FeOv~(OkA5pI&gj%YLlq(+A(5O6pZ#9#vG@sGhvVIf z0+eGgxt^0jz#7eXN%(oV5;dWX=;zExr1(bN| zjT8kAR=a!iox^*HS6RfjIW3W$*pbslyWVvtXE>P1#l%G9X9Sk8_Y%BvZ+Zm(t|2FX zSgid|pHZVF*#=st0TXWK6Yy@X$%AReh~;VN?X1jHLJ%Al_CqAxV#(ZQ-O)8`e<+r8 z-#eZTj^B$wQpw|Y=hs4HgH5E{Zj9C+y!B`k9Mr4dJVw+3cqGqr_-qmn`fsQreHKM{ z&BniAy6Ef<6m=1qS8YsTjfi=_PftaUd;Rx^iIP0*KQC7Leujq9l}kH9PA z{%V(7YqvPIm}D3R&uSEj$A#m`G zwMUA#_Vhk2Uw}Pa{h%lrt+G0hYeExC^h4vZjVll!wcm)F=`)(3C}*MS+vvf|0jEx| zG36+bZilT6mu!nv8*q_x*Tmv0tlt10wuI5c%)%>DbI@g!_ zA00Whiu{uh#D4Z4q#Uq3t9#|i3$SW86g*ozQm4ZJY96I)t3hANOTq&PuC)(m;WMQq zG+M?Nit26%Z<$S7HGm-xeaG%tv0ir9v(7HgDu$lbaW;)F@(;piz2)G^1;yy-)eR?> zQ3AD`f%i?{18;FMeV9pQS$`q*5Po-4H6m5qlOE{&X1N=JeyJ$wy+HnZhxoxoJ9_`K zaTxutjlI6Emc1UlY=}{`*&=EP__8c?C=ZCVx}su?7+l>c zYm7Ch>#6nDu#|vbYSqG*b2q#9^QP0OaQ@C;Hl-FRN%S*)t*Wi5Cv5Og<5VUcf)=rj zxh#cRaCaE=Mm|{p7i2~wtFt0lX8u)$SdMLi=Td`9;xM{zFJ($$-xXt0nPe(0wXd?N z5a{!?d$6D#?>r|Z$dOn0n#+(+k#e~8$8$kipnVVndgKW~XUx#J9A3hOm{?N!Tm zrB9p<&8Z!TCQUb0^FkY|g=@E#wY{%Ux4QPo)@IU6RlPVx`zpj+R4mi6DBo+PQYVXW zVN^IFXQ`)R22$7p`pWdIY zv6hw>OBZ&eF<_ZNbg*r%XEhI0w}Kfk$GDs^knStkkb$Yn88BJFBdxfpV&hRXC^yK0 z9b>k%b${BJJx7Uex^Uatb2ah&^~c+n)T!58#*VcY7dB_wnTT*||BO0MCH{ds))kyx zXMpp0MY1@PI*GbJs?X9GH-jH;f!6vjoxwCSqVi3)t&mb--x%}c>aYi}(bml`y#_Nz zqb!N`$1PW%F$_{G^W*N&6sI{Ai`x>M(0GuM!4B(7lo~g|KDT@oMPsNFQ2xGteMTpLYX!>d)!vSnb>r>3nGC zXOyGi)9&Hiclq44LS<6NnmQK|S3$wfxwG6$18YfEVYn4UkGRH`XT^;#m7db)xDB7$ zZu1XY@YX|b=+wB`2cPi7?K7o@EbmpX{5$Y}@73rM0AQJaSXBU+|4}nX6EhQcJA3DU zyER%>CU%n*q5Dz|rVCusN~;dt93CiZ*!8jiaDiFfPOiafjMb)b8C~2H8C<~6%S>Ey zMbfc>msr!0cc;q($B`bx-M{z2d(6E?h5S9hk3!N$_1Ef`yYVYdC;<=$Tt^tvJW`Do-bobX=GniGHvV~iXJ5Ud;cjFuW z7XK6<-3i!g)P+!Sii01EJ3cvNDSUHtl5p_1zuLZ8MzU42tWNr=eq#eoL-nZm$m@O> zOP-^HTu~(*$Jl|A&FoYcNiJ=}rohE@56g%9)}cO!4iBGq+8}gX2u3F|=`&$?(*6C_ zu^op_jS9`Zx7$yXyY>|S#BYj^xA2-Iln;`7cJrPG>4#3wE;d40`?GvqS-_Y*UOU$q zj@pzynA5CG7Mo9|VVX988mL8k!4(bb$(J5zWWWz9ezhNH%arQmA>FjL{@p6R5XF0S z8WA#7KVkc~%m40$3#I4Gmsh0%UuyCmd*71+^1 zA}rpu4+u=0)oA1iO!8@2J}9v($rm`k9?5GgYGW+W*0b@Ri?Pn}#B16zDpI4f0UZf( zwu1lOSS3WxLl+UH&l+s;hgc*h$Kc|NO9LwzE?dwiF#&7wPq9Yg`mLPlJm-mr>c!(L zldgq3ceH95>MyS$AoV50lTiCsqp!7a_~1?n5m&Kh5``G+TN;SYy@wrh451@stP^C_YIV#r$UW;2xKEx_S*`yJ3|i3LyZSh{ zf1FgrBIa;RZZk=ZTJ7<7YeVu5o}0Igfwl65u@4CWAYdyBCox0cK*jDcd_HQRyq+UR`*6Fe z-IFg-dOE8`z3Uh7zvU-t0HMYIpZuWwlb`=0bvhXt*xLSI^ZK%?l-(99LeFh&8Wz1P z`Wj_uz)M{jFnfIz3PeKRWmcr%fn+0+Q6!2u&!wor9S9(rly^Y2y*fWSSwM6-V$z%m zm83I4V1gV(E=W2#0vce*W)RyP#3e#Yb(o~Li2c=j@TZv(GzAkeVbpXRW&ufUw0puZ)D)pAYK;};MfA)0HNT`+1OiNUh4!5(Q%yS zL6)xdEs^if-JKo3wXMM!^@0WHa&5-WJXZahDl$CNMva9otr->A!B4-dEV5KT?L{D#HM9Q)>o8vBoP!olzhwhmiv!TuGKR< z@i{R4UMQL3Q~TwabM8DA@r_zgYK67Z0@+7An%9C}+2vJ42p!IKbEK>)wilw511wnW z_$fpwYwld)O6A3Jiy=bIS|P@2bxH-!Y-+d+N@@9EODhF%C70}c7!sZ}#3VPAixogE zK8(ZlM&Ug%r%|f!PVXR@jipR8MF_fmGVIUF@svx&wc>Jg&@CExSR8a!au#u8M;G!g zsu+E7ZwSir3IaS9ycp48k!v8Q%;m$(r>cb4xqi{f{D*>wa2m6CHc^)6h zmhc*8NwA&O%_$|=5HUyV5RE7fa5dKG<4l`$X^Vu~TbRzCvYBH|#H6PtK)JXhEJ8+n zZx0hlsF8#mQIX;>F!HWMSNtj%?+#qlrL;Oo{E487_{(?nGx;~KGmr2EtA-0QM+J#B zq}|QjT1|`m^_6Hsm}){UWGWA5U$2oURz#s)J!+Zq_Y{~5jrb8dKx8HyvmeaGYXUsf zI~7xwY4L!3uW!AUj5~#Zp6a|^KktP}dJ6F44oEga+62JfsBhRnLpMrA)cRx!s7WaQ z+A#0c@->bB0=0~g!<0ri?2GG+Wo7B`Wq;@Iqfi^x-57#r#H6=UCqSppd zL(6KkKvmQk^&I0(trJcnd19+-r7%4BSJ0Wi#bcx@7!iM4sB10>?1d z{H&BGE@ZT5WKpB_$c!iy_O>zC2J<-G2Ja7`$9rj_1>>*LaHka?2TdLUkb$VP_;Gc6o;M&}-1;s3W_eSv~s z^!+2!Z{Yop?bONM+``t~#Qi@3)3%0<-61Q|&yAh{hPu3D^9HjGe24>W-8OqSXu`H( zqAt8W3=PqIeMdtP@_E;iz2E0_T#~Rt!>S0hwt~T?`jNdUyZ1?YJTqRb7)Z^-yYI9U zg`h&s!dC>J1k0D!Emrxx6^RJ{-%~;wM?z@FStnhFamdXv!)arz#G-JvGBjKM67__? zeD{(dlA{e}{E`Da1|EVb7NNaDHu^~D$4<#RwMW5S<$ei`i;I3;+*~}Lp<6u2?FUC< zhph2g)xd=8WVL=B@2j8itNE&l#SHFaWJjNuPCS0^FVk1s&oI-8!BYQNxK1zz7{!+j zWyFZHhRm9(!2c_P0Dk-VvM_aKewaSz;{P&{1|urpcj!_$Sx<*xsRW^0j7%d~fYt8kPKSHbE}M7n z{8UQoYHxojB>H6A(eYJ2;a1_Qb1DYHTF*qSS;h3_lCRf~hp!Tsn;~pF#GBZQsi$C~ z^jZRbP57$;w$A_;gE|RpS~o$0^6=ksUSVi6AP$SbVUidap;IGkiVXjD1~f|yReVN^ zekOUBmqa2VN10@oD`o4H^WgE+D2|Ja8}7_kqY4)sG5oc2$X`PLuyR`pj7hCBTo|>MHzvl`@p^Lku{K_p5Q*&#iHcU zUhZhZ1%fr35P52Ycf|um#_=c0&m8Jc{FQBDedN-D`lsdzJquhdHn|}`O{NRnk%EKD zQc@h@5dSutiYkEEk;DkH&W~cg2(}4c&TS{*sgYi7@S$0tm8K!F)T5lZNj2=Sw1Esi z<%xP|hG}0LcD06vAorzZ(Z!fc=9}k|#~aj+X(#%rwq~4;^OA6a{4Zh ztvD^lRg6YyTM>S1o3q$YJ81j18$J;yR-90}0=NYw+J@x-u~J=-&Hd2gCTsxzYTOy% z`%rEvrq(XWUv8PP#ji=wFX+aDc(t6{wn{_9Va$bMY}LFo;tJ5dwlG(L;rl`_t|vUF@0=JL8Q`=mJNofdR~NU5_TC#aFaKE?nsq$l)N+dwER1gsy$qV`p| z&$z!PV4!1#WbA%kUV_#2;;{<3eH!v047hcM0NV`t-rQO+9K{2XS_N7pE*)zjnuLDtlFF;g&lL#F4loFtU)W;d92O$EPVo7lgSJ9rgA$tH{BKPS(oP_wRzyc5Nz3K z0QORSfI_c|bt;Z+TQBPpRijXS*a4ciLvhi-N_qa!E>{_yeXs~E9)Pu^@jCcVq`*%A zxbG}9-8V15jMiv$Yjn~tqtU}ywSSOJd#&>h%(H#@=)V)#^TLok4VnLjh6k+F1+x6e zqkt|U>AwRxcB2pGDilBxh%=IKph%}F_hl?3s^b^AgpyiE5r9`Wg5ah^t&n=_19yjPDt8V|y1m4rE-!{8>TY19(M#XR4oIV%e+$Q#76S){6#e{m-H#Y7+j^B_ z39I9aR>X-asHv|$c){V8R>{+vyR9dKm`*}RK=Y#tQ$&7TDGTluwc`9vRp1D+6R3Is zdGik`3`g?(sb-Fc8`Y7T25~n{cj}e^&FVGgSlTod0q7>?#)Ly?tZr(~;up*2@7@_Z zZe7Zk{~1~#5r!Ew9fsfrItibmULKu?PVRR9BfG|Dkrc`|pf~%6Lzu`72;GC_%m-hu zdC$rJM@V-YP8AgDGbJ!VO?@TVLAemj+RW2xGwccJ13g6;iF`DuG>R|)LWY5vuDAI6 zL-|mE4!=hEMrdp2q8Dd&rP#ZMhlKLP*@~;Du!NJyBo^Mhfu7AZX6enwv&taX#V&mE z@`_ul4N_B&rmQE8kZ$LM9V;}Lx##ZEaaGKC^8OpFYqbdf>ZWr+kLD*y zEVO2!7QS{Hqe3BL~M!2A6DZ3544ZVs|w8NB6oUoVc-;jcvBp49W&emdZ;K->q00Twat z_RkHC)^yYty5%zes7Kpw(@q%P6-TukbBR)y`|tk6^E7Nr)@SDMV76Z6f!+4^_pc|T zmq3)qHSotdyTN|NyCyJ#ZT@Gcxc44s%&K4Mg6XU8t<247pPlYHzG%;Tp3i}|) zsIih~-ewzQFTJ9|%v$X3W(B4CrBh(F8kJB5=iv5?HTW-G*dkS?<^#4?6h$GIQ5(r0aY? z5-vQQ&V&wUCXthspX~OWa7KASKzO$s|1Q&O*#ld~{O#x%>0w%Eri?MZ!FJ9_-!K z$?~wv_xp7|xlNV;C2^uW{P(nQ2C{8iqFCKyG}*lX)1b0BE^_NoU;O6Au4ikuu=XS4u% zpIL0gKp@AJ$auto5t|3S8Rx~z_h>MRWV_IKhDbtE8q2OGkk%JCEx0TjPJl#_7Gg6I zYIGHNkUiMz2%)X3if&8dzW3835N(jIC9CRaldh$zL!?X8Qw+1Nxb;Mce zM(WfEfl!;>9C6O)$7K@IqpGRuX-$J7dN`{G6WGL|=!FvGC8X4$2*uTQi}I*CVw{Yo z8be4#8K*b4MOwO&w$n=;x3FAbs*Xfx9l@U8_;B)bVZNWi4aXmJWv% zqoqQIsMDM5aC)GcZPzuFuXq5IdUnf348}RrDpnCLt37f29Y_z-n@^b4UQ8&Wr%4ZO zDVG~fWTJ;#>Fh3Hj5BOOaAWl_4-W`htb#8CRMcjH1cC=xMo`i%Y-1N@z~zg>ALn?N zfS(!YZ^v`S`r?$rFp%<3!BE_-0CPf-O14YWeDT^fTE|AH3hf|{@L0gc$@L;PfueiV zGt3d1FDApL$Ro(qwIa>x%3FYccph36HiiNt1*sg3)z#%%HXOd#QS(oZTsc$dIgx(W z5+skp%A2pT93ewYH`EZo9)PiW4^qOT(KoOfndeNI6AV3y8avC&FWJw#IH<0KH&_6h zxn*s2K(*ZD-O%2&eBJHfxPA^bzWT98_?puO>`?Sf*I!SNEb&Yuz4geN?zVUmv)g^y zXtl_&JQ{2=nO7js=tg;)FTq&Gh6BPTr-nlpXbyKh>e)sYml_YWnk_BEjh)3>5vi0n zYZqig_U;LjCT8zLBZ{e*C1sa5pqqDln*&1j@kJ(KYTFEc4{=^QuAxs zZ0x+~2w5Kpd#*8DlZb(tLcWHW26A0ro7T$CF%1teK{vLj)a9Z>oz4;MPcKTKejD__ zSIs{W*^h?m|E%Nl#aObhQBiX^h1H%T1Ktt=HWP9Prd`vKHQc0(rm~j57T&Y@MREEP|F9Qh52k-f~>C>pWDj1zXXadoKhnP#y6dR|y z9>U|Y>Nt+!GN;*}N0Dtqg3D0pOxAJFQ>Xc$DcrIp48u~fb1E4OS=r-WrauW}4q#Vm z4s!7WC9~rBk;iM=#_Z%GIN(ZVIJFm-%|S!byLC-}!pC>O!^au;gd zuKwEIzHf^o7EtVObYnq@Iep@7Ww&}?5b`9zNR~+=f0VQ+ydOfmQ)o36cLXF#-O6V1 zu82|xVVcDMp6ytl&lquklm#g+lC_sS?br#{vbB`ilJC~?W@9@TX#01S%YCe|Hi))e zL9#(3zbqzp(Tvgwne_ua%u7A<#@0`mx73}^L?!vpYXbnCWbZZw{MU9`qYH(G|EaQV zbp?F^Z1z6vVfdC%RoN6HXgSd2MmoC|%6(wYMTC^bL4yIDktYa@DTUYyzi=+H?i7Qk zPv_kG|6~MIh{i(#{#E)EG5+WG=xk?Y@_)rXbuGKYHdNo$+CPl&&YHFKTzxP7eK2vG z)?m}b0#~jIFko;r^NdyzD8iHx3QfJ;r*TQt6UrZod_$zq_jlSoP6{(KGe|(WAM(oD zSlL+3rwg{w@ZhZ(YbFZ*m}DcSj_7q3#0}Rado7m_TyC;k)I(B(q(GKo>%L z??fNh@8t(&HpRW&m;AcX6w{tJGU&A0L-Rl2J<|Ws67^R+t!*w!P#)u z`}O*bfOj3pb`v1&D$GmRW}vxf~5$e zk+4Z#U^s-P9%ZctMV%BKB!BqnrK*z!*8?l@#hFYq0eWMibQurt&fTM>9}vD9k-@C8 zOwva?kn!9QvvC@++`_k+I4&~p?fW9D*=M8LxK71sJJ6H^9!rrxGY?+<(UANZA;@&s zEIM+1{)U?(jctuR9S2sPFXZkN{H|42p}<_Ss!G*th9%Dw7?${%fUkv`aCKX$P4rU5 zZ5|PE1ZQR0`)j*cn@BPUx=c*L?tJADimq z*%l2vVc+Nk_l-{ZnIW;OV|f;pB=N(<+92o{^tBnhqUOJ`0yU+2$~8htiO^=|W+Dy} z%oW-|Y7 z_dDr)Ni@2zu;55(JvEE+@f{hM2uJt-;N{yqoSp5Vdt}Rxeq0PS`BXFC!rD?^MhJ}d zS9KScfVMa$A)M_3xEr&=BA;I#2RMtKU(ux>TQ|O*CnBmz3WhV2H#j9F2)MQG+ZZN1 zM}D15r86pLy4cFlt#nm#^rTJ(R9FOf?&-VBi<&Lc#_w?hje+M1x{|wA#9JSk2yOh@ zGb8Uj?t(6cVKg2!?tlEuZEP02N9ZP`>9w0`Y>8~{cOPE2QXe^_3G|gkJ#t^Q z+9ZiXBLO!*sehcEmZLmP04em}L9m$DX~?F3w8zn%c|inz!$nE8pjP&qelQdg`tkuR zfOH}jVwBW%zzLAq7hCF(aRK8g`>T7JrU&Epj<((&_YU3KpL_R45UpRAqW(Z>AM$Lk z4t~G1Pz@Yk{7yL(X+55q=K_gVB9DEe!m}PsK!#Sr zl*R7zt)Pi+^b1+Bp#q0=r1(H~MFG61otct|0K1I3vG02U-Vk4rO5!l!eKs&@zGfj~K(E%71b6ic z@??gZ3m^0#I}8}lq+jZA0mxO&ef}tm>Eg$m2yKSLAd6+pLVLR!L>NsoX-a*XKApEYntK5;Q}{*}|);rBdZD zNZ$~S_1lrHX~NTjdn7PsqQ;*LDHk#jT^sf0bryPqo#$bx1%jO;B9wGu{K>=Xg^vt% z(TRe7PytHcp_X_+b{>O|W6#RBf_rR`p-UIU3Cd9j?8z|7y&-*Mx_9oz83=7nw@C~} zzPNz!%O}c0AQwGrn}CiPJU$ZjZBLAv{fMoQ0eVuyhQPqpE&pijmP|ly1PHjg!pgUKn9Q5e{<~V{L&b526eN|RePUmHvla_;RWZQ5k#Ri$ zTQPCSIHl{gMtPrnck<}%wkjzMUaAi)m#6-IrYAi*02gx{fPo-o#Io#8x9 z=0v`K#Xy@PKnCiANTgnt~3E*GRd#Z0B}-F+Ga$eav&YUGJjzejhzPd_7mb_IUqXEnlY2 z=J*NT`F&3GxHX*SMaNfox*VQc zKR3ND?|VHdPoM5zU(fqtJ3Zddqs7(R-EV95+bo7nWV5Y}#A`c0dzGo}OC;Xk@$E{* zo1d4|->2N)Nj_il5Rc!+e`lfisqFNA^6`GjA^kq~{;uzS9}M;uhPO9Cy<6&rGT zg3^#XK&GfEU!Bw>?n#opj4VxUUd~_Aja}o>?z}C(n%|GewyH%^@|xP8K*n2o1JXej zunI+Lx}R`3h@eu|o$5w8-OqoAjQ#n3Y;u=95UP!2(mWFcDbHdEFI}BXC!b7bJ)_70 zsCH*$ct?Q0HRe_276@Ucn(D_W3loLBnCeex)c2YRyySq63K{6Ul4nFqfw)#d>e z4azxyl)oUX@^)}J;K=K0myBu2>8@K`+4oP6gkDn5s{m3#EvrC?NaloEQpj@zYDP-% zPSHhdNTNK9|A1H$D@(}VU_vQR7Oxf-@)EHmk*pTBBvGV_*Nj+_O!^345_@7+E+Q>h z68^tITkVVhZ%5;OLUe?X1<@pHC}>t5S>eOiICv;A+==Ln_>;^cTNqqa)fi{^j|LQ% zvZ6ssgrX0_fkcymv^SrEWcjc-jp)BYENWq%uo=T-Nf=C%k*D6+>j%jtxyW~60bJ#Q zDhZ-IzRB|pE*eQ#QIRMDOM**8lFM3IumB0En{XaK0fl@XABjXFiC98HF_h}|fB?}` zBtbWxT&@C4(uY_w`5ZO9!DJ$ZLcHXU2Z^1>HL^yEwwYtLN#tacQ3>wj`0C?$=DYxM z61S;oK~`|o)+x&H;-|z6ah4a)5#e~^wf3;mdy2Gg2(J#>4xN-=?--6MhH*%HE6#wo z^&B!~E&XCjP*yUssiziX7an)gE09jX91h8yE{wcKrXwm=aS<*hqH}G6MaRIBN|E-L zbo#|7_~l4=mDjH=34F4z7BX|p2rkGoX4!mn-dI$8xUg`-zsruS^NDtm95IR{JZA@w$89p_{)#907{*TC5)D_SkegVxv-1 zB7By>&OY!p8zTI)_X62QVJf})THvtIe9)5Tnp7#c9hfG&I5NluD+n*{8-F+|{J*zl zCgLHNw2YS?mXv8pub*$4L)_$ED0&cVLOqUh3qC z$;oE9w^{3-@RnZG*?7Wpa+|<%2_zN+N;G&otzxy4FK2kSo^CF_{ezCv?>nY8P0!TL zV!s(*Kh2{%7P*wTe1fzb52>3<&zK%l$OcQ9PsdvW&Ehxk+}{8#V{g_miVU>w zmVn`Juf7gKyIVYA!|Lh{;lTMm145kG)n`FUg6-)ATEuRcp|xsD?HQSx#mh&o_r@DS71H-O={yLs-a=)-zOpaPagJ;Ffs3R6%0|CL}ds2I8*1f981 z*u2VL`jp;GsogQk19a0~=cE?#+{u&~l>j4kWT) zlHCh2p=!g*Yb7nE;Z}M8EsWaO9K4)sydz6%vo`!mhqc4^TZ?(t@ENmP!E<`5%1t)& zBOumAUIf-YtVzpKdQz`cHdvFo(X{wl=7hM-O4DywrKMBSs2Mc4ucRm8l8rycZ2o;7 z6C;h^#ru{rPVFgWHH_NbaxkSmT{C?;z@MD1>|%K=MwOdCc7x6Cu-eRbb8a(5^QB;5 zl>rjKl_!H1yxO@8VM@S_JvR+ts^4*oICQd1t!)ZoQL7C!n$VNHz|WVH9G12cO!>?e z?&nI&$4hb_N>TePqr4!nFj#3UGPzUam+ncQ{&Xuv>bb^EPCxTnqam!c2a-AWwv~l7oZ^JU=ED$qOdw8 z51V?~r0GzKyUT%I{i?+0>nZcSRKi=B(uL2YUBk`M*sc9trQg=18-#faQaS&EQ6s{% zH47U`HPaW*1_u-BbVO6dB%eunW}v+JXr-UdL0M>;b`a)dKayVD_Sbu4o$P?~0rT0v zl7r>W8O#0R(y4AFFQAd8(3fZ0kklp+CuW8#q}$ z@dCTSX5}uGFK-xzR^En$ISF^Z6neI&DCnr3?zJp#F1u7%p=@PeEJ)OFJRc*m$V*+C zTDa8MRSnH+Rj;UUthQO`dba#f3Ht9}mIiZQsQGS5+z$}X$=#a`8upE_9P z%^**6d`ww9JuB(iMSYN6EcFR((%=+sZ%AqaHw))m*3Q3AmE2pcxwaI>B(6rvePlCR z(ORg9mzp-_9_rLE1hweY#XJ`2O&zATwTtGCe{k()m39<#h;|=ofNC_Rm`R!0oOjt= zETF(g^JI54sW!AoNbJuR9PF$_o?V;&7IM}r_Ow7iBD)m{#%(Pl8KVucC976dXd@yqcJ-Nr@?0Y ziMG=H1>FYJX?u=BunV^exTETS;}24cj1?xWj<7Wgm6-z;`+{}nGNQXVE}j6mj{MBp z#`Cy+4BB5O7d)4xp{D35Fp@2x>MaQ3e0?}uYFxn6TiplE0JlHr2jlodRWOZyN`8QyahIg?4sMK zy7~V&d&ekAx^7#z*k#+*Wmk9Ec9(72wr$(!vTfV8ZQEZy?>Xl^=id9{j{A+dV&#rC zMvjbFv3F$VjJ@ajh=HHY+-KaK>9LKbZwk7yJ(Xv$Sbwru+^s9f&R~R1aX3?7WDeJA zpid!=;-Al4w>{JXO}bzBG;jc1-|We>5DOeEIy2UOv`0a;c^ZRTg>c3*b-8}Gq({h@wfC){^?1DO2e>H0d*3f_sv`El=|hF>oWBx zEhH@#q9`sxm8om}U@iV({brY?RSuU>ui4qb!Ch%K)qC2&2iL4p43njP9N6c1X}r;+ zZl<~v*d+uU$05?6Go_DCSIPx0#eno@pN01q&bx94kfAbXm;eR_`!G?#r*ur(-!p7s z=|6+)?!^`nNz?>%ksQueFh0z3F3~;x$>e7pAPP__^)AO~s`{l@3o@2CmnwK_JdgL&92Ty3-#lI@%f&YKAry<1pLmR%{Lk^{3og~-2vOZ*E=pZpcFOm-VnSexG=Z{mtJ`{4zx4) z5KMe14Lbho-j#DD0q!S14w2XGD{Z*9Z#{?avy!fDr)_}kRdr#Ug9w;Hz?b#;f33iM z!FdA#I|umu7eaghAm|W;P;-(1ZmG0@k(228?Bv$vrl5W{z2q<3F}@IgfWjOEL8&Eg zM10XCi~>?1jOx+yTn7>mr@UjAMQ4#$lAGI)-X5aCmjFf_2GObs;|kC;g^{Ch0}0au zPH31!OAM?tB(Mwy&(MQur8z+U8$hclu?rd`iT=|1>!bj73?>PKsse2Zn)-8PHU{ti z?pp0E&VRj}4o7=mocym2J(k=m08&a;;^?F)+WMS}a#jJ4dvR7F3q!dF`wDDzHl~Q< zY)ZgyeYUy<=mbI5W#<&VYJO56f>sUDJq_cY3V&;Y@P*HUhG|K_fcF1zL{W`1k}#7E zzWwT8(g1~>dl>eeyJgql1QRy5|II;PP}TsV;g@~{Z+PBgTP`b6AoI6`NZg>I?Xvs8 z*ly+vjHw^u9J;Pt#ABcxcRfJd`(un%T>E37U3a<$T>WE=6Kb z#&^D2T%Z%=5RWr(?tGw$;V8Cn&&@;gzC8E9H-g#94-NUnXxx=2X2LSm?&`Di;r|_i z`OuB#k~%e$(AhQFa2v|{$8CE#=w#^T;4iUAjWLaf4<|xwehK(z;ovz zrRRMHfN#jw0y;XdCn`*L5P=YkN@Ek-3*HByr*{YBY@k8#x@Oe%bpH<^kQ84Ras+Mn zBTW1SDxt#nM&J!isO#$n0AZ@kx1NgngZhKee(cBhzrpKEYMr}&4fMeYfTH{A?lu4r z8}jfk2>$}GTN3COs5c?|Vc>UQPH6*o=npJKkXgSIK6}3dkWuh>m>W=hDxhJBK!Qvu zU~Xli52P{sW*)3)C+#AAf>O+0h%57vM|O9TU?2oLvYwB|fJl5Ffwp z9-STf8;lp251?-7k{tP6h6ag z6BQ>kgv~AN>AbB}`K0e@7Q-nS6Zk1AV zV48+ukR(j1M)a_}A@d|KDk#XY%BMEFB`^v?V&(p5ry7I9wPa*`cL8vaMTpG$o)%o{ zSrM3(lBtXm#uN^bp+X*AY%QoYmJ?vOnP^)HPRUP+dR{TZ?Ew@{?$GOk{8_X%L79>p zLyFKclrI3h5miqCIr|PIXi$W{fp`J^*eT|_VuAMYB@j{yzY%>Q{?M(E59IA5*oAOP zLjHjI0^?<)$?cBn%h`cQOJexI`hw+EtxVjV1|*K(rVzv@?u~juCr=;7YGybfK?f-1 zy<(kWrJ83xxNQ(Cch~|`}$9K&q%IAe?$NG<^ER-`1rr^GxM)!pU;0{ zd!7B_6QCjfuQM4irpPVp{9a)igMp!^>8QEs>jN}^j@C+BvclK@FODs?TCaG|nacX# z4&KqS;06@JG6obo7fOPuv0b5^^B?yPER$jY$;sT7qdV(d=2$G@Qs$UK#r=9Dp6`bN z67~a-H{dTqM6N#K-ME`BN2s_kqKEe6XeAVc!^n2meBTK+nNVX;ttQA3*bg+I4fSEa zqFWTf9ki0&dsy6fAu>{MzP}t-x2g7Y0iX~s)TkMXoN`#roi((@Zbv?g1|w3hq6Og14!twDBHwS}+l@Mm+>p_M0a zB3UVDpj-2r>#(>w+bLKlAi!gD>3YMpXVYhp39pDuN?T?OjwTdOOcHbGjp=gl(h zxtMkMoErEdS?n|mZl>EwSr-B@$A5%LP~Mi7u4sti(dYe)B-X$mCg;dte(^jL+-Wmr zW!Kg2s7&9KTRZov!c@E^=Dd@1Ut*Ck#u`8*w_LNK*FVT@;_rp+gfsGg%9{SxRah^O zQ7x)o^~V^X=WMA-li3na7m!(lY?aBPC(s$OODUvUo8JMXN0+zKD4 zcQ|s+u^(K~L?6`kw7IFOuj)YumMn2CU(Y!`HH8vc_pJ=4?S3N zvH1|;_n>BL^1;EhfJvWR`$b$}N#$$VVcPu|cXp0TK*(CoUu9 z)oKv$=+PufHmj46_#aL#oup8lNgX-`au+1YB|j81yhHaIUVJSKP*QmJ*AU9RKr)CS zbZj9|8i*k!-w^{*!tq7iiH5Mg?PhL+2-)$a2_iyqf*N$J%!+iF9^D8LYi8OY!|DV> zijgz~OfzQ@4H4|gK)|n@3AD_^RC~zv-Q3V8ymqjk_oAXZo&$*qTo_aQ5f4D{!je$& z(8W2K;h56G*l19^(Y$|jMW`dAhztS(kotke-t!V8?S|A0wRdLv2=#RjYI+C{7bv!? zy_zk~_<2hp^nL8t6Ga$kVaQoBg%*<^9WYk0nimX=591Oz*ca`x@i-EB9ID#P zA7t8vIx;ocq(d?xqgp3e>v5Hzx|U}P_SEu_0|-+gcL7Wmj*MunM)imYd;f0AmUYO0 zZDi}{>X`D4V%Vw4vvbPZ3LFeBF>)ZT{+0bzKbb6Hv=#_b%)-Ews7tABGJlYIC}EKd%Tp|4Zh%9%!rdzWeUyYk7A{S;eTU;y z*j9G@(Wv+qB~&JE7q~d#iuNKu16mtlebVb1!?QXPgiTB~Uxw1H<|g!pm(FpyosVPOM8fGzYFfkPyU9#KeS-ME@Q!*9S=KHfEHIX)F=nNcAm z<1d|D!tehk@4h?J`M9m^8?;|HIJUR=)Z4s5=x^%<)ok0iMUXYA5XAE=w;zFo#R}G7 zC}4W7jvyR8UN883z;0`Yq^!3NYeJ;Vr7W}?d$X3AS^gkg2OCm@w)#;2L5e0vjuh>E zwtYgBjn)=H&PQL%zDy}djQc6ciR(qYj*ldRh9FMxS<^U6!w)%(g%rI{bCfvwz3PIu z3wRkd8*~ee)Y1uWbZQxRxRfbqcU8@cAXyimn_CB^E<+><$wTNB>2yV`OSmQ+T|$4x zO$pReejC}mB+4GH1w2&suLxWNrK?H=g~>-m&IlM#i9N82(FRf@4x&4seTL!YVXy(8 z`i0FsIPXB$hO-ucNq&qk%@qPrwnUzPECa*xO;bqe5WRhn-g-`Vbw5-|6EqhcV`+k<}vh4%Nbhn}DAz)n&~g_kS=GF>M|P6*R`loh_@ zyPP?on+usBKrT^I1a1%&l0*QM#8xl(tNe!CrhI_8F4&}Izc&ADuOvN9UA(U(JxmR; z7xm-zMmjlFGBT9TnO%?lRZbW1?XH^*(h-SO({(h*NW%Ebb27;9d8(iL9Lv1oM+HE2MkOggGX;xHH#lEGEVDhlT1pSJV3^5x+jxg1Xq%~KS`SbT z@{e#>v~WQ&;#MAr%5eMw^Wo0J{_w|3ZMp46MXTT;N(=+&@J|DWS1h)`%K*=H@th5a-k>*)!45esLKV94Z8%7c|u_^ z7bNZHB-1!ze?)zc950iM(TOl&tv1{7A$Tb7TF}zJkVr17wEn3)f4F(WQE|; zpyFkg)6*2k&4QO}da~?k0#{(H+}UC}H7no{-;g0;Xtfs8ls3a6 zyHI!9nh6?UbL_bKL__Q(ysnF84pScP8)P82%mOM}32{m8FjS*8 zn$&W{XFY{g0++Q>lF#IQu$C>D!6hYEk|xVS#b{KNnIsWTWZk_watoG+@JlbR??{eK#z0*~v&2jryEZ68m z^y}zHagLaw-?Rr00nK{iT4?-UE{T65(TlwKXtZrveE??d8yd264iZALhhqbGb zoF{Mz^c1BsGp+J8ga6o=sc$I`nc6#B_8z8ZnteD9I!^TbZXBe%!k9M=V5nD&{yPW7 zSc945IipsQSc^?Ps0ODPma4x9VCpzu{!{Lw7q{rDi2leFDn-+9W*Z^I_nglo}+2Z?Ii0E}xj z0@RF~ENB!VFEE}~CrqB#2>5$T`>`RQjvDghx9N9x!}-#hb#eYzde;Te#QVTSZ{GHy z_|QxQQZi43-bGxu2RG)XutGRV-!~jztvHmP$hSbjOsLhQSxriu{WM{JZPHX#o4Te| zp-(PPdCGtd`VTHACxdo;wt90%R9I8z$O-xzzWYS6!Pkd&M+fF*d91bha04e+sUQPP zE>JM`YRA8F16Q@mqeAmp^$NPu)S%&{NDy4o2NG7t_Vi?;p2DbMBY1_d+Lp(fA3}ra zSgjSSwkKdcGBRY}4u)q%1j#1UlNx#`8+EqA1Q;w#1lbq6XaQ#Y{vLzj93;oAJop zyQNRhduAWqGv)ku$WV(*B__M*0nIx&x0tk3O#T;`O7PAYq9&FA!{iuqTAT-Izfm%Xi7z^|_)W0Ck#%f54h!2VIx8F&7 zmjeN=KB{k8$t>(9;Fj(m;s>v=2A^6p3%yGEk3Ydg^G=k^kEav+|~~#}#}KiQPr1m$w!pLdH+SKZJ$9@~e8Ds%?}RQ*kFEfBcu8#R$AN`$$D zmD~1CF%Y5$D+C!Dl z2+!rv#`@;f`0HQ@IoW$haVXZrQJ`;8QgY2e?>qfoMFwlm)nPcCWOsHqG}vuPqL6>8 zet-^q0`GkR+4%tVa=x7WD^!(9^?Q91IWiW=Umm7GOCYY=h&D<7{LA@ zdDEFG{EPoN?qVT*LHU64uCi7DP2PwE&c7 zwWNjHF#vxC2wdZ&RY_}C=RX0l*im^?K9P}Lu%=n9F4RtZ$5!2!p_r{De4dZS z=&QUzp*j4#P>}H%e9+o-({$C|LS_Cm*~9zIYis}J6#nW5Sj3%R5TFajPp!<-v2u=R zcy8nxK-eV6uZ@PWZ170mUjX}(LPJymYq!wMG8hYCo4nirV9|6LCLgD%) zrd`mJU*H+p5pEsv1T;%2b{RMoLUJg2H+h3Y=}-O1-?0GF9TNrpD{)c_S9IjRKJLXSV!v+wBjuC#8R|b z!3W3<5E7H;8$4qrCbU!JjIJ8I`7(wCC`jwv69Kn-*} zbG^j(X{9+V|NQ9$iw!R<%Q8XQB^-h&E7Uvt;7KzuO431mi@<9Y2e5?s4!6f77;cMi zs^7Wv=K7C`Wb;ZWF5{9aHd@w4C@oCckbhZioRox+FRMMY$?sTwezNsYMH3nf8hZZs zInMB06QLh0>^C8!VaYzXH*UVYEKm-}N8STnOoG-wcsoLhj1xfE*_;zs6*kP6kfNxuS=-pTBFgEh% zxqQK3(9(Vn5Uhi~MiN2Kz;Mh)>{Rv`JhsJeix;qF4n^k3Ih0szDP^#MK4X=zz6*vp zEkl=K+<9wno_l$9v<|3{*2&qlU8d$&W|*^PZ8^2Mm0mu`ua~l@v)Z6*$thJBYX@ey zA@MYp!as^T-zsF3ZIRL6dwblv?^mTSr8{h8$CO_fW#r(_;GZ9FwyWF(=C})WE_*nG zYfcQtJg$LIVJ&OgiNtW69c%BYD~+dU75tF4Dmh22u&BCyDr$*2ay$!X{8jpA8w$H< zJhe*OzGd$OqCMX&eZI^>h@gHH!ODJ+WHmQ5D0dCkN|pId5xf{0S}c7SRS~I7S$I1p zjGBOzb*U^;@h~x6fqgn|oq-wsIRQvhW3haoG8(7VQprWLNzg^p;q2#(IRf= z5OS^Hny`kNN$tX+#*I<utm>CtK){mDymC^Jryt93z|omT+LPd?*#64B)x2U?SJA*g_(UGO9fnCi zt`Rw*uQ0(S6uTIQxWN=mZv%E)_qe)v!er26vh&}l8<`=gf&0RxLkUWwqrG=(ff#tS zxsL1kk{6k`n>`o-o=6qe1^U`i(-5VW-oTQ4)k(Ul>w;XEVhQE7J&US@^!#_orh|mk ze|w-QV`~YqU<_$OU2%?R))MJO=2g>cJ;UO;z2kAv%t$UG#H9wL4q^?{h8f~)asjpK z(9P=bC1_gWXbp3|e&;WK# zg)jC|R3|1p_wOcq_ckd`X4Q%p8p4^$lNQ%W_-X-Wn?RXyOr5UqEi~J;2xU*U@XcRQ zT#*aaysYEcOuT~&zV&fR214=_Q1YjUwE*T{$KS;ZQ1A6kO`Aqx)aP3MY^@0Lyd$_x zgOz?9CsgHI1NJJ(J?*Yk4(vG)95%v#x9djhFew{Sc}@_Nn=c@0FbxVQsKXcjpt@m9 zz&9n;&^XCi65D%R+6f%uDXt4eHx*=ph%yZi5FsfgkI@7Pr?dY=OCVtkn~E3ehaQfM z8Y4;vb;NSGwBwU&8#r~1S2zk>!-^BARpU}3`8!CGB$kI6O#vst1v|GsrN~}7K!qZ} z2#>`{{^L;-ef7&}ou;Yz+a!azQf8*Hipj7E~q6qP^0h!bIkEukzC zFV!W#Y#aLPNKmL{|4(}Tos!Ee)v*GJ{Rg6azcQFOX>$Y;_WTRTKrCj&4`(nMpC)eK zP-R;%EOeyF9TNEgT~PfRE5RDRxOzTKu+v4txpLt0jHvL@!cZo$#7Y6f^4Qo@ijjK9 z290fQLQ6SA1iz?4qB$h320lXvsB~mhw5v0@IGBPdQ@jH#`eRPyZ%4b*u!EY$Mg+?m6Z)nJ1ZLrnuOeg&CVE|GH`zH;^@$4O;Q{&hTR$8{5rySedIE zG_N&20Z&_UBG69rEX1-d^MInEbH7jJWf@>An@bA|p7S;HOYsG&jo8)~wEHQ=G(m&* zta|I22$*jguHn|{`TiRGb#S-le&I#h4VG?ce)IYw>9cBDYhY?Ikafx`*-IRC>gOZH3};_0k_jZAWH>78 zRWgtbKSU}kY3bUAmGnX?;W=@BmxdLlPPN`E6hSJuAdPVL8vQ zzfReKQhzD3Ao@t%I9CP01r4M1<*D#avhz5W>BSx4qHjJn7dkGs#ca79W3&>0q>`8P zBcZ{RU@_M1m;t6x9>RJZKYVWptwIcgn;;>S$@mWt(ysZa8ZD1B$?Dfsn;ulm>gv>9 zt{D1=O2(%ONhV`HYy##VsuKWDcFiVw4}b zi9~%{$ks`)aqXx7n5c0Sd3t`%#;22Pir=;RE^&7Y99LA6Oe20> z>UM{ebRCKI5E^lwBo|e*ZnbjU%UKyDg z_SOoJ#yI8$7d+(WT7Q^Gp)(YUAJdHInC@XG7;U$ZkL8@!QLingkT4$~X``3#tBfb^ zm}c#$OdPVkV7CAUv?x$5jHr zwOHRwEjX9NjIqc^L?(ls8fV&2r1kU2qmOnj2trwA!>Qjt#6GGnbc~~e^MA6 zB#Q6M_4cxZ-36cAjsT5p)OGi?!v&^G#nFjHyg*pVAE;^O5X$#~US2Z`n~jEvsABeJ z*+QIaXbb%P2D8$nIGO_!9zpqH7R@hghutlIwlGOdZy z>DPeTQ;AVdQJLgIo#>!;YKf7I`A0vX*D@;MW2066R!W(vM2V3);<~e$s)hPprc`rY zhQ(OYRMcn?wn)aJC3BkaKSZ&9n8+HF+mLsYRbV{8ohtQI0fNKKQ6Y9(rB4G`o-eRSl zhCGvb8NPDWbi&aq-F!Mz^khcOZmX)o^Q)vom%H^3p;wR5WV=kx*W0tC5z?AWH%pKI z=Sl9gislv&ljd8sPr+xwABaqL$5f90vL>}L0Swb=ZisB_Yu46iH0gp4a7UYAqF1Pu zTd)gP5N9r6&iwbv*=z{KZsxPx8_<6ObtGmD#hy>2vw43v{=?8-tr!Gm|CqA5G2#wA zYGegr+a%^N!4wD0k2P}qC}+fG2C`EDT1M+}I#lfhX{i!+<+}7cwFE>;oWMD>{G?TK z=PS)rY4Th^2P`Xb%%PgAOz#5RKZk_f!MjNa)n3D#C_G&NmOFzdv4vN*uF=XFeH>tt zA}t`k=PPn*)X;V)h9W#mxPb@h5_0AWQMT!)WOJI9t^K|FIlImEdi(uyFzfR&>LK%0 z{G9VK{k@*>{b~07Ot$jW-27qvY0|1=XaU|3v@%_f;aM6c^}ZE{}GS)o&UAt&M)2iOp`6hMNt}` z;E@QXCWmYYFqv5=r+hp-MxoE%*=|TtFv2_jsnY=5$%w9!5pDsYYVs$}_gPN2KYab* zTb3voP$0f_rZY|X#2O1E>>^o=(nUPM6-Xp3y2RwV#KXcdD4Wuk{2-a??bcE)3jWIM zp%}DBBY2Gvs(Vzr7RpL#r0JQ+oE79uZDGg0471_RPJyi0+ynhBH3RF^+Kzc+sBm;0 zV!+|B)U(~urJx}I8*TYQ+*GOyzAA7#r7~*^FZ0g69<|r~xsv?OXTw6Q0!A_YgS0*D zt4QjBb#$}0cE49ylub6uJ+x^76mLrO5Rza<*29*-_s+Z(34Q~-c`!6LKvZU5I_t>UdD<`{=TtG!|t+9{Sr5li=h z?&m$&_i+!Ij%EBcn@j#J;p(}f(GV&3W2YZg z!v*GZ)-VOU)_N55Yxz`iYkL$!=t`G@?96SMWvP1E%=qro*zk*qoze6xCZBQ|w7Pg# zxAVmWpV5oPzWLyHiQrH;#?%V$V{ZwMaK=B;kM@F+q?)C-R5BnGm-3<1RUNPaQ$Uw; zDg+`b1l-DR(9zMIX=r?%EggfdWS0_Bi`Yi{!_2}_=26KrZvOE_lv6b2vdWc8%HkDy zu}lKt(?0>_w_uytwWFnF1x;0ZaJ)e(Xim9NJA=@rVaVOL@A{0$s2+ZoDFEy>uGA8a zO2o+=N!qbXPqA2Ey4YS(O|5W&C_*Yg5%ij^;d#aCX2z(M^df)d z4?>-qi57|R$%pV;6u09r!sAj99*{1!wD+z-j+jV-;0 zJPK=bT|j4|%6E!%P()>;0bvRMz{@vg2}{I#QmENE1{lq-OY{%BhzhtAZT4sb)@IoR z>)zCHe|(*gD7I$(d80w+@v-z`Zg;AT3q|49lWfCyUEo%Xqj0L+>7?=@9K$941=%j#)#VU6w8-n=${MUv-O<=F+%dQ{FL(ehKhG5F*(6x%4&l zq@wPFeCp1Z>VKw=RJ|?4%WSdO1V>yl;svjX)6$m(?@IIsf7L>G^+h;vf*48MQL%n< ztYjtjFk7#6`AJ1UWz`;-vff^KJh&!b|1ZGw8|uS*M2)md{27I^Bnnhu>BC7xgi)6> z{V4?T#)+Eg#wl5ugB^7CWS;o!>gzZ`kO$p2L^vup9u}IY9Z>p-un?#quIM@z& z>cZHMd%OL57zk#;RG^qKKm{{UdoxgTC7wVUA|SU`F?dt~OJk5O41KS?E(}e-i5dwJ zSB5wvfNq@N?_;f6?xT-Im&Pp*C>vA!b(Ll+?`F}EmXsb~9S&wn5?1p}Oi);wR-_Tf z;M=*aMM+!NYomYSn7NGjV#o~aC6^dl$G}Nc@pm&z{K;SJlC$?!j6*3U!YB`I4R54P zrWB@wg@#iwtn24XFTa}&_ba7xK)0W*bT)eJx%92QAj3O{wzY#H91}Sdv*5q}?d8UQ zK0y?2jK1cBRDcF^B!|I-)FhgdtF8d7(=UCp*JgP_Br%N9JgyWlAgWJ-H9Vcu@Ll&X zcJE((&$s7KP5@kxu!{)S22mD^aEA{s$$^*Z2;IIj@MbyVZKpPjs>WE8zI)uvoyuP> zSsDIPH$Hh8tSvS92ZGaBxYm_3bDwRKN*6eyImH4~koEow37mNu-8QEg!`%hh@efm^ z*t3yqI`y4djN#(o^9%4V6i(eCcb@4E*^dfn{pOJM75{WD`Ak>$RVn<&E|QB12N?e&&W72Uu+BCF+YO z>X*MZ6pPh6%$akVB=>Lj|K_#5w&8F#oqfRKHY+SvGTJuwH%|80{gpH6UoZ+yT5?-e zfB$)Q2vGi{5k6Bk*=bR!R$CX&VB zb#6y-%nk|xRxJIxbfn06cQ1InGPmroWWo!CUu7ohZR%oGm0M-D&r^c*&W3`$C*uUT z#TMHiVX+PIXFb1bp}A`Kex9}E+I;SwY|$^QS6}-KO21cCj=%YXDS_qZ*72?w-cuX3 zz*!?*Cb3T64WAC)F9TL=tI>T-Zu7h@4r7`Q2hO_V?S(WmkHeK;ZqGd|lI}XF^{fr_ zD<5Qcp5(1bH0kIlnLEkEEE?^k8Ew?mdlGXVAhg;bI&`o~_|ffaG^4wBN94{D{5o)y z0XwZbBgriqkDagCB+2%O7i;bleFn%G=l*#)Vc2p^vx-l}iO@@doHObZH;f2|A~%fa zN`oqf5RMg!Kfwq46Ua-Kc6~h_oUmfmHjsdijCVZu!|oy^iYqPy2G7~x;FPwPOep*- z(m6-S*j%j9hd8YiLQ+_(rRg0+(eu@#2}fACi!{_pX73BQU!KGPrU@qhK}xtGn@j2T z<9hWjaV4|OuoFN2sr#V&CUoXIz`0E94PWSnL7g<`>(5G6TO!H_uoqaOZxLc8*t1>~ z!%KJNA~je3=nl?UC$}jv%+D5&9*H=p_6MSmQ|Ed*b`H5x!tK^-d45=n%@Re|@iBC> zt$glnhE`I@U$fmeqL(&Wb#4e|#)!77Mo}=}aQ)aA;@bhKTdb2#<-@z2>!)phK|}CE zDe+u&a0X~w9<$9M8}CW?0{(i2A#C;ce34#wZYHRHlm*e(DmYUQxFVFk$AZ^*2t>1r zZH18F(`R8JsKT@x{=5$3^{yP)H`TG>O@iF5?2tE-s|Jg$_Pg@cW=(yi@}O{! z`u+DiH>dvd12Y{_WwS2k7pEE~4&So&O9*Y`&ntGC$XzV-q1c@dcNl}`zXJ3Q1}O{2 z@KVTN!Tb2|AK6?dv=pAH6zwKS7cpssICBNhiUm8Zgh$&$Y-vQZv$+g3Nk5G^Xp*$< z3SEOSDYL6VmIhfa13-%OmoD&gKY!ZE&-&P>F{9*7d&oYI4`CVG95xT=6R6W|9`Y)1P+Yw>m|;IWFYEAGR@pIb?(pf5HsLWY zp*@UO*)M_LhD=ZGrjIFOIlxUqE|s8}Kl>~t;?VrdQiMR&6X)q-P(9mJtj&f{hKq>H ziyE^aenF=;*+pDl=f|B;)|?lp+aABm+^D`~=z`|b-PP(q`$5ql@mr#GkX5i$hu#l_ zX|Dq3d>8Q*x~M-yRG&)qu2iC-Z#`$iB0wV-<|z;?Ema?ZbdAvO#!JTq6e zGVE2J7nDE*A(#pI3FAWYx_oSoUhq+w+v487hnC)~A3#6k+P$c{94+!?K);z#yb!{K zHs`s^%y=k_pK}G|tDVatRrs?WtBt=WbjB`n1h~6f%fhbp(QjfX3SaOme^BW^hSXud z#71k$b-J-PYF)2l;HLSAVLz^8;&sSP{syr@v*)5%axW+pbmHVP4ZG0;UcsToQ+OT< zkgxR?a4e4R*SHTtnc#FHoH{Wpz4=Lt_iSk09YES2qUiFW%L1o_HxeMw_v{#0LO0Z1 z^VpB06um8cy3dqWaz75W+)6uq_ZJN_beZJSb>oODgz?Vam z0q067ndj~B8-SVFM?Y#iNP$9_=joQu3-~2l!NEJbOn)S%| z^T*L(N7uQ`tEV!3e40;qk~f#+a62!$D1>I8r=sO_h;Tr4wzH7w$o_M&=1K?d9Q(4D zw?a=#6M+Ktd}c_cx_^8*#cZRQFKQbjS56Vk%Ybixp4YVeKBf=58!rC#TrUhz+5>Wz zkI0{E6|63|Wt1^y@3e2;Np^~4qnr;e7*PFU{|1{F1xQ^Abedk;fLVi zPqe$PGTrdHA>OZoit!(#x*WY@n*qadUtU5R1fadZNc+kcRx3EJ^}rsWjm`w8Pg=cJ z(9(>uH720>A|E;+Ng~>*2wnJzPe2cncb|xvlHM$3D}UN{^raVe3K)+-qVyI=Ap6FF zsDER!GUOj^>LxEj%0n?beP;haC|i|#f7nJixuA!pJ%6m6Vc-H)4?aH6Gr~xx*%D0P z{^an)%;pV7PWbVztokd+ClAlGr@JPL^r&!-z&q^eK;T@9Up6^?E={c{~&F_@l35qejgS|w{fb@I?k!rz|98bnQz zm_3M$EH%!DTlgnqa9wXPu! z(usnBksZB#0ddf5meytoADdBw1m|0eos+^-O!^@0&gI_qOJR;QD1h-Ej< zUAyu`!v)8D;q^TR?c?>m2km<7Wh}Y5sYKZQ zxXH=$ycN{<5jO7zy!~vpdfV-I*yY;2&|N60PL$ z_WDPS$RfHOAXDUej3-Xv`K0ESFoscmP5RjJaa7j-^>Xs5gR+)jDoG5b*Qw`oZS?vv z*qt~`d$(zmQb6q&R~o|$H+^q%xSs;hKYC>I!q)cwSglv7-uC*wx*7&(5y9qo;%{|( zeB~?kbtk5HQh2XMMR_vSr@5EC9Zj6pyFaVBeX(_VJkRHBtzvl(K02-yzuy;-KP?li z>bP`0oQ@CEYacx=p}b#xx>Uuu93}e=?{nXu=WGVleA0b1$OUEt6yM1GN8{Tuf6wmp8Wcv4II37-y^pwowat9)7=yt#D$syJo~kNa%! z+@s@J)yI*O(C5J#ho&B<$^Y5d4J+G69zTOsOeHOYXCW!^`Dzi%Wd2$1c78qFj=U$0k>5$!kh^x22hF_leM+CFa=i%}`*UY}-W zm#Xrok8aPW^)ftO)U_&rPV>5#kQ|ruP(7&bWB;5XGk)Lxk>8EDX8OH%=<&QIp1=1o zy9uZv%erZ0+@T~@j+WvPZV^aU5}4?+O*gaMDX|+d$MY!e=J4$1hWCc)^9a|)!MW*C zJ$AmNzdJsy&1;>m$$!*V|YJ`yFZ70^d(NZaz0aZKIXT-_fmY8 zi@NQf&m!WpgQl(9?^CMZGB?*g?{7G~*31vd8ckZAQcWJ!++NEt-UGI~z8(he+YdX} zk0h_TJ(yqdyOrKv^J1)jOjZ+&c>FoLE#L$1c(06K&~wQ+TZZov?RI;B6Vb(S>2`a) zaB`y|zLJPZeFeQKn?xn$&3<*GF8D^4C3O3Y-T3xOH9Fy?zTV1l21seJSj=``D@gyj zu<5(b;qiIftn6dVj$*S_J9Qun?@#D$yn@NVI!9U2bi|kxtd>cEv zy|z6z9f0L82GHUe1PQLYL^U)t;C|$++H;|m&gvQO1uB23d@`AyLOErC~aPXWp4?JNY zPOR~y4#Ot;;%oJE_x3fzfF(X0E@;;dgss{4NQ5g(+$a?}DlT3D{m+py-C$p)_D9)^ zC7EKu>@ck@Tk1V341d4utb+t%YN}`V(c6* zCr3Ni*K3tw*0)j_>WlQ|ulJ6}oUKZalZG$5cB|%(0J?yHhl%W8?1H7YfrFGDZm;*N zE-^s97S1l`ZV&?JGrrK^TxJq+qP}o zc=G+ex8}{8S@Xx~-ltJjU2E^^uI}n8I3nU6pb~1ymZb}XFP@YmVUyFDuORX?b`{C8 z94RF~u#TfzG$FK@)L&o;#B~hiopZda40uJ^`uXg*OjAW0!U^oSKLwHEaXA9f3$Xfj z@#%>K_@t~Jyg0&GkYC6$h9&yI{VAoOFgQZM$_-f+nUFMy>*CYhgzaJ#m|4L^y<#d3 z4;)#hW`rNPxTn< zO|Ef-7eGTTu2zk{kdw5ydX-ZN9yg|F_0m&XdUNz{OyGzT+_sf1sGo8 z?Yqs5``pGrB~i=*`Yoa@B~yPD(p1?6ZLd)zE@mhEf%lW@-Wg#uu*O%)uAif|YG|>= z7bFpW83b#-q;PdgZl-Nj#=aRjqPSv@>iQZfN@FdtH?_a!`%Q=gD5s(XlX~jgHp+Ze z{Pqc8of{3{g*z7{0g_V4UpVC-I{=DOL2WH-Sum;%yB4Xl@|Wc2x$*gur}Xy*#|)Ms zA!O2P7e%T-=l25?#>ofkT@&ff(fB(vyD$cL=;hl*SPly#2T?7`QTPtH1>)7g*D#ZZ zf-1t}r45*yKX7hX1=cn}jpa?kRU5={)O!UpWO@0^j#C_oA91rH+j)ycz*z=)JyAh$+5LJ}+UMK21-qLH{(arG*H7sG+=W@az;y-6fNO2_op**`(R=da!S4tfcU9QT2!zBts27Z zj`!L|gQZaF5nep*7*#h9rJo<%_P5~aWRp_1dxc#Ej`!YwiCnvh+}=0uds-nMyKWqf&lg5_QY#`+X_e&5FN`+L9Wg;0U0a>|rp ze50di0q3qv)jCOV{^HoVB}V%zC~YKHw==a|L3Y45KzY4QdLHv4V%T7}L@p<;KP>c0 zjum5mMnZu*o>U8nSdGL|sVqQ0!vLO+d4H^hb$*B=g68_qAn-d4{_&2fkI#1S)S4c} zpW2F8jrPW|;OB#Kmm*~kgktNsEfD@#{&3uGQ~Zz3j2jaS9JfNr@V#S2Uk!@(-DX`J z6Di|>!+p8&xwHA+rfz0623q3=;dTES@edigLts&iV5`S_kGVQ^RDaPt(ukQGD>45M zY&ib8+{}G(Vd1ph08ot%b^$}C6-$1ma~JIH;Q{^Uqa)?2-gidO*grYauaDu>Bh+%| z=10$OI+h0lKZUImmMx+R9<=sRViZdWpg^c=SdgiDvvQq|q~>7(ff8$%^~t`(ej<)^ zQ8l&k=~m~%k*1qyx=ge!R2w$De%?Q;X}uV)(9Br>QuAld4vWPzrHwO_clByRrBfBk%E z>e3>WPvlb-TK6t!QDu}aTH;{w^l-%EEqjUYl`Z#2@GD_DJRfY!3!n-E6hgOM--MBr z)*Qm&)Wi&KkUS0aWdxXdEs7a6u3Xem9(H!)b+@24{pUs$k;sko(@DQRn>|$DFb;o_ z-ZLFDfs=x6oHe`Syo2VAKQe=@K@pUP^W)ydL#%b<+NEwchgEY7UCc_i9my1dZao`w zibZqGbaO{K%r60S%Y;|0bmLZ_0a^Jn=ddbs_LKpEXbX*ih!dCpcJJ)+ia)KJ-W2W;o37KKm?$DV|@&-#Iu?C$(gdlcU6n!S*(Y1&Z?8M-&@? zw6v|E*DV1*17Ruu4ZEW7Zoo-z{FZU7o73hq1?hC6tXWgXPQt0e^5)VNEf2r79O)e* zhlii)IEm3kbzlQ{ zIgY{bUft5UtKp(+rhBkXcfTD$OwHbxHcPU##NcqF9ZRM-*$`;2fg;vbJ5D6Z@gLu& zxt4VCrD5}uB%p~w;F6a+gRws%8aYYzDm zE|H!dC3K;Unq4-6pokY^ycIpTMBd5pu21Du^a(@y*^+kR%@VB-En1g*bHYvVBg6}3 zzG~vhhLVf~U7Xz&Jod*l!8*$D?9F86lPR8Pn0L2ad14yyOE%AjGXNliHsy$Noe@#B z0-<${ey|^S#i|2eO0TV%@3Awd`yx)fC?woCYqK{pOZSs4^RW+gPDE<(DP^+AGAY%x zY+5{Z)s^;zn~}y;Kcd?(>A#d-BGgCo5sYT{Fj^7@c`jE&#KUnf{8+#!$y3jut*oCj zGF9@nN@uc0!YkDu_6YQS9|6xl*MC0|I7uC%l+_9g<-=zZF7Y^(MW?uEv1}}RWF~zS`G*p$`QeGzkfI~ zsR995?OO>t(VRDeQ)E%7f^LYwD&&l-Vu8JpR;usN4 z%(r^E=zY@sVeATs-O~dHjo+&8T;r(~-s|nqGhqNg?L1wsHxvD}vPrf~TV6DZGijuu z_PX^MIZp0+%}ZrVi;O3#7D+FDVPZ=Cwrc+|Nv6T-Q5PicOD;EFLMP zQ7<2xy-uHp269FnH^n7XN}+M;B3)n4P|(8{W)X=twnb5LyEIRV<8zh6=YE$btWFk| z0nClrUXL;a&pXdh7tREHhwjVJ+_ST|ilw35A5 zw$W!%m57gr$ftxgBxbr7KbwZQR33B-%ZYvVT_!<9)D3$Y?dc1mE6&Kq!ex#A%N1ANBznX7;3zG+dN4a!TPH>R@f2TVq}xbuy03!Xp3?KjN@ z4Fs<1T>Az%A%s8UFtyCNE>@opXb(J2oM~5X8albU8RI2rQn)-w^B@h{>T#(6U3qkA zq*kC~F6NZtGHB&~X3j{Z2Nls(BCa9)S?rzFyrrqGkpTy&ExNMo6bTe<-{a;i^F+bh zymE_1RDDjb`=d#xHW*zxnLnB7w$stu9c~s~-;UK^oXSMfWOTE2dKOh;M@e2mP2l@T zsN+Cx`}U@fcZ|hq8m!!gt;b)7^sFmY74@-YgyhgZ0N}=Y_(GMq0pmD>O#h6Er z>Wx|pY!Zm6_^{i+T#Dw)?B)YaTkzS6<@`ZRR9%zzX~Z~#F%$iem&7F0RJ!<;V_;`b zpfd)w(oZ&d(1O#M5%u+XKh?f!2HXW{!;JUwY<8y@ceLMJ@SHkAg^0}&{M)5rzYpD5 zb7%>BOX+tjpJH7AU>Qh~{xmhcqd=*MjaF|2pvy#s&S)>&LixGiTS4c4ScO{c`@ zyiLNFKrVs=^opx|(xjDv^S!l1Q7GuH>u?7R%&$%oQV}Jm^;xI4w&-R9eW}K>rX3>< zvfLDfaLjhD9tQ^+IxGTTC0{Py8h<03&L3IAoEKgOPW0`+u-(;ONDqn zw`$;*Nt!u);kc^{q4JTi)(G1TSliRGd)WuBRt*Z2WFo!F&h^b#Tgll4nFSSOZko;d zHN8$|Njf2VHT1}6M9~(rvs0tL)>Dz0lgx|=bS==KZZU!Xg1WFWvoVoC{)<)_nyT-t z*70RqZBch+!W9JuxijU$TJw#Z=WdL&{C&A)Tfm;m9<_fnpkPRvu?Ifar1~jm3%0TU6Mr+g=dl`MzEc>L80ktaM?)C_lG%60~ zs!0thr-cM|2<6Wc9{Mgc-$$q9Yj#NT(NNB-g=S$d%YLmE^b6Ye7mOH~Us zk%ZEALZ<~2G(AumVfmLlGJmJ3qGaer%lI5X5N|)il9j4@1bBAV!4TVIX)-@RTpaK;beA>FB2e#I8j~x{w}l-B>i(?kPJY;Gn14_ zuimcT19ZA?aysWH0ALl2>+sI#0}<`@>qQA2rVUpxnqWonby2_LtUMcoAizuELd393 z0Z!hkaFC5X`Ovn%T%6+(2=tB*XK_*qpxMkf%#n4Juk@s)6@Sn{cT9U}5dLW>vXhVk zEN{7(s06^SndWr4_TAsv;WdW{u|@_&d2$aS0QAjIEzDpT@|l#Bld3|jPc=Eal~Ou7 zz*_n#Q#trtr^xX(3{l+NVfgFD5~D{bq?msni(K~T^$1yD)l1gxgaCfPQ|KWB81@xW z4u*UAMT1yTJOO+5i()A#7&SWpIxygwI0La>6KIf=i|Ky~`4`ed`9mCTVnu8rz{PAO zbD`uBGJ%Knpc@ML+RXHFubH6XVedkCEU@skfA^t4N^P4%qXbYQpa-#P7P~9Lr4RAy zLv6V4V^a#wiPeuqs@Q_yi5HPVfR6JooP`>&25}WAW};7QS^;o1fWi27=*%I+R)_&e zBf`%0PfZWnc}t++J6pV0bmrzSwGy43CtiE=J~Udp7;hlbHR0u)k0fgT=|e?{VJUsh z{=s7;fK+?vjzb25U6gkKRe+pC-UK-AiEl$cr}F`s3bYu{l<}sDi!9cZzBsGYn-dk@ z#QJ`ohlk@F$RuvE5zq;&3X^LTdkyVui_(27J~sdH)%=_-)*DT^7G0sK8u%BYFR~wy z6O-wo^=Sae>p(Ru@JT3->ue9;6_)g$jPr(Ckq5NdM5AIC&{&A3H`$;E@WNXL3dT9X zLu~_D(FxCIyFDdZ0GV}%JMqK8*Z_%3rpv?&q`aA}0{Ld!JcS_qL+r5>JZusFM7!Mq z5G&Z8y+{&=s5b^bPt{I<;ksXmEEi&^e?gcH@(AeDv-y|YU$JnHKtY8Tg4FMKC!PKb z`w#3}JTS21Pl$pT>P$#j0+e{z_%||K9QciHS#q<(^N9Umk#87=LGb>rjBaZF(| z5fA$LIw84gp`lqU`0Y8&ovT=rR&&kp)6OvDHw<7E7Vi13jqZ@C(n5W#l33sOz?UfG z0(D{ixr_ZnBGWf%1dYxTtS1xXOqr4k>mUviphc5(AEZ^j2%Xhc1KS6s7C~iRM+>Y- z09Z}5iZl+A{e$6-hr zpUW;7)HsX0Hof;(c|Kd3;nKJ@RY&u|Sla)5*OYj`2smU%KnqNj)zWPZRYQ;<9NS=r z6bdyRipZntW1r#P@rg|_}6*N3$JC>9vHnpyHe3;s^9AKPW&3kUF!Omxe|3N)R1*L7O`g46-X6@q)TaGAi1=?1ptSr|Dmg!`QOa*zlJ#WFF4D8nP~GblU^?V(=$RxgPWV{OidG~ zd~h&(RetzvPmP}bNikP)2`2Ibq3poLixEs8Qu*$f41=|I>`PKH zosuT=xQ%vWF@y{889Ai&r_qyRlzhQ*QZjuA7Nb3?kb<=8^g>UTk#O7oyZ&7fSbR{y zFt7ero`?^UuL=^L-yz)F@Ho7YnWdgaj+7Aumq~_?1?mf?jrNxXunl84bim+(d5+vr znK9aciFx6$U_DH7UE=SBr511HL)T!)6Dl-^_Py334*SyLb_9qP)i8r=!ni0s>nI?j zlFjk*`jGx!?kQu35 zpF3Tdd4jfzZSzLzf)YLyzOQK9Try!a?Qu1wc(x}_;1$G+Y~z9ps1*sPcr{~y>T^Uk z-mFu?NbOz}Zquo|If$V9V50=fQC;5I%Kj=k?>4LhY~(T~$IO!Mz(>i8HKF;nvxq`; zJTcj&45@)fp!pl3{w0Wk5krqAgs>V|{#y)*Gd@hH3@NaZfN0ePqNE0m9#}(8Y=YWK z-#4>d`OQ#BZE?Jeqys!iO-Z~IeaMJ+3;UL8%2kKE7FZz+)K315ehf{8e`=(E|7_g# zjj0WB%F@>gCbZhKQGp)z#8+Xu)E#38WGX5l=_>LRS2ebv0)o8MNs&%BcFv|vea8?qMr?H~!WLL*xXdRAe58j@C{S7%A%fKRDpXIo zp2rxco8a`I55)R)HKYgWAy#{H{O;+_+tBswB{dBE(c*>elgjuh8n`y!Ql^OYQPVqf z^+*JpE3Mu>ss{<@^}Q#C(YF;USBP4uKV+s@_5tYV#8^ilaB)Hb6>J6#{T|Z!;6nUw z+cN%Uu+S7KjL4<<-*CY-BepK!H)>3qSOhje9m$Og+ zffaG~8iT=zfhOSc85Sc@jK z9kP!V0YD{_rB9Vudyf!kO@!du4hf45>v7$8LA-O1&RSUHigZJjW*NPoLXyxLBiWOeBjQ!noQtMEL6Qd6g`uLv={Mt+mEO4No%_P}P7ECpJ zf>n7%Ft~U=?eDr5D)>i`O$7paFJ9W@V^*+-(b{X!IxQpWnL#lj?e@*=Ah~RbK4t_`dv;G7xY*!f9lfTW8 zCpjQOc0SvM{#)=^{R~#S9@S$Z7 zz$x!y4D3eCl?gPM(}P6q-hS2(WqeX<*nH^~3gW41eDb!Ny016&G<$|9R=uMoFzEOS zDBnKL;P0tP=W#1g!Wp6`Ad{aSP+A#JSe77Lx`4td66_xw7F7xV3EZdj{aNACzH1|8 zV6%B>xLHSvj$<#&fU*XPVl);9#V~f&ifmk3Frp0qT|%o}UG5`?(Tt&4B65r@mv^xb zp>oG{bHCI3!*e(qKFxHjxXg}QXyT6Q`7jX-rCR714Uk2wIgM_!Lc0DN)R{*aU-Tp* zCC3z0@RHeW7fk4x5a^%h!p}r}rEl6`|AFyf&*M~r+L2ZxZ-QN!aGSpWh$s;x`y^bs9O5}@Fm#``K- zcswP=vR0n!@`1b<$lqqL9=ox)s+Z>+r}2+Z);tcX;iY->hyCCQx0o}&#Pw&#rO4Xd z7hacjb-W|an{~*;v3w`^lavPQ}W$*e}p2@&S?UGw3Z%-)i2y8lIll+7hFk z%>Exm1)O@!W8u{5WWMsIu&f! z-J0c`ee>Hk79)PJ{}Qf(hqRJqOH4QvHkZvZSi!e@g;uD$3RQ8wGV;Y24C(ge&mKjNH9xFLT`EEol)8h>skM7wRqxv9aRHP6Xlid*)l z6F5YM^MdLn>4(Z9?U}uZ^B0WPtR2i@R3=tSlAL=XCBIhPd~p%|8P-M>xT_*__Y=l~ z&vD885_Q)Arl!pLeaM&XR0(|Cb~YCdcZ!;DE2kdoe2rPfgpOB+XFYP7(0TRTZ?M9n zQv;yQ8kcW>#_~4OpQnv-)o1ukMOBN&Q^ULtOomYTF*CfsxinCMEL5Dw5(E@X33+54`){1pp_3t@wG<&&Ad6*QXVKYpiZ zCC3t|v?Z2JH@FZm@q48sAEs-{5POwATM94Mz_|;SvI}Gc+CP6Jf41#^RSxzbIBBBfMApvQT*z?J!59 z|7`W2?uKY)lFCYa)xACYS-xUr&hdkoa+FLH;`o`8T(qB|{g>~7cXcm5rFg#x8-+ZX z9&hU>L$s#(-#oP*-7TUFsCyXAa(Pg-O9tMa*TdKE+HvizfB$rzYUNe!3O8W`w-}Rt zg;uoo1ztnuv}7^u|x%g$bIV;Ft zPsde*faV847Nb)*Bat@;QJXY8I{w%Q#doq1(EJ>qo4a!5>;&=DE zC8zcId0^yJ^dBFe+c(wXVS4&#U0w)AT_1=!-rp}cpC9w*^ATd`!^QOZ=rL2?D}?T# zS3ETKc_+s5<3jVyno;~Np8}fPO1Z{CvjYwHJeQ1`q-p%sAB?KlP}fwYr~?aROm>AY z(hW#|NGLR@LN>;mIi8U1;Llt7E!hpHtrpmmO7-Snp8=EwPt5hKNlsVG_zj|DW#n1T+a<2(YshJ)s=>DQny7< zlxtka;cX8a`VDWOK87vlF}uJe%CR#RHo~?1tGvz+&xkDk$G8FykY6A-H9-f*+A%Om z+9#jOZCS2zRr7j%l9=r<@%+i9hg3#22GtsBH02gkdrAVvExR}f{pUvwhgB!d?kw4O?hVjudc8m z+^}u2w~la`(`bt(?Z&b*ZQiYA#GSIdN~XaSKB2GiF8m*}RC z!UwFV3I`21F>%teiS{+R2GqMy$V!-Fmt6(ALzYCEL*YxSSQ7`W=q+pH)~kFYE5p3> z2n=0iDk@F~nb_G4#UaRxd!*np4V?M#Ny(9kDlpIom3(HzDw1c_IrFgNq@%9Vv8AyT zlj<|WxhWC#iu`evgcNHE{C!Rr>4XbWFSj3qm}A#XX&wCzm2cYV{TgH~>zZp-)>1|* ziiH+Mm65E*KsC^lBtDJVi3w(VS4&KPiSJ$K#yG)+&XLA(Rin7eK+ixDewNv*97i+c zQr*i}o9a}Ws@-*@3WV@5pfR!$*3zdVwK-%(rxxKz$(N5-Ds`F%DpcaWzvoWw9vRm% zQl?{O$`o@PJX7#cqeX}dE_>xuFSlaLStP(|Z)$AhkYlcwLQX6TnmU*T3q~0ebdOsm zDEEmIb|yRiJZ}x|===4nsWc{gVO)#_G@ptVC=Lrcw!kW#lX=tJw3$=d{y0~pVSx&F zLhXqhVm(si`%Dt$T&XuQY&2&Zl*8-1lFH*R zUWr?u7z{=fEaOSoTUmFV=0r2@;!}Lp9pn|i)k{|k@cNjo9r5W~Ng?M(3V8aqsHuZC zzYR}TSy_=auy)$t4c$$qO)n$bXuOJ~;cK}3(a)}pRV)U>vtZ56g*uXFHZ}>Td&zrV-)BoMBp1B4Z&uH2 za&xnVzxPtJc|8cn59Mg)pQ%_{bw>+I7k7%S(LWlBuYzsO`nJP?+VJML#>#^jTfbU} zeX?FM&>PRjZCpA=NZXfLqMYxy)42Kv+}2kQyj>p#=ePL#kn*z#lrGxd#0^O?>Ma%V z4=R-tk|_uPhhGt`u#T?hOaFE7y1t(&S{711$fUgy3GYlBv)dP>hULh}u~Fg`ScOl2 zu5f`G)U@-YudJDN9%ihmBgrlRBB`>Dg24b}ZFG0#*73H+qrVYqJS0ab@pJWnqxYd@WS6s;v&#Kolwl-7MRpi{xi(68QaY+iO zrJ9l;ATjD1)aB8;6?;Qk79P2p%EI#LcV&`_E`2sB&wcPrbC+*iAi?t zM)F?>gk#vu0#B0nL=wYOM2DiXni^%psYID$gau{d09laNkEe zAdjZ?H!Bl@hMQ<4m=)KUZbsE|rESxrzl;R!BPLvHHbP03+pz|Q@kGO{TbNLriO3FB zJ8Oo_ufa%#;k2TAY^QLM<$e=KeHiRz(XC0wE<+f0f)Ndcs2MS;*2pRU%7292!lpe* zOLcV3Tln(pIS5Qog1~$edWb&=K?~YFk@S zEG)1nd4oGCPR!0dVz;(038YtWx}>gwqXFN*Y)302Zv5JJqL_)_q_A^{&gid-&j=kt z+j<=xBhbB;ndtsVS0B92u9wCuJ=gT0~Q*gMrT(ZhBEv(1GEA>JH0prIGzA25`R^K3O!=%G%${Z#K8q zfEuf*c^(QmFq-+5?VRT`QNFQ0^eYy#AEd}yQbd(x@g3G!cBs@w#17)i9@mO}pvfK# z`G2+g@j`*JwW!Si@<|c);bal0G~=Qv6HM`FLTfGr{(&Z>mk?Lf6j5@}c z^A@qf%Uqy*qaRZMw{OefeqFztfh~T9=&yzBu%cNzFDJZ`^J`R!!yghz4H15d?m)bo zRx9=n+=5}&Patc$y^}K}?Q8y*#x}jN2t`?6H=s_jrS3AOv(PB>o~JuZP6!88P6S0N zb?cekaZ6G_?lz2h?6uXwd74U~U$vT3m znPLmt5@~oGqf3rZ22fDIJH#+g%t$X6iw=|^MtsiTZaVO6X$O>61!Pnk(FI#ztn7nI zB&bdtZvbas9j<{uvEzoXtfFB8bq=wEyrzqfM-IUsJYX2_qIp_JTDhpR9=+OZ7f^2P z)E>RWOJN(aSDBx_I`m3@L||gbiy{$h$9B5{e_H*qWlWjhJb&(bHx4|7HCH!w;0Lr} z4m`fHYs=A5DL?t3YoT9kJe9STacDY7WsbK2b>bIza$z+2uPptxKO&tZh+@bpJz(I$ zYYLpVZ_L-ieG1s#)DAG|2NEYiLFFlRP->TC)qoOxGej z_~olqsfs$B2pyWzmJq)5Bl>h)_Ac6cqvs7lvsfHdd`es4Ue`SMn)^W>%G?}O_G1OQ zz7|pWGzYC|hy64wkR2r0?ShN%#OseHtH3z-)hFd>e5CLXc=>7MB?9rZbHze>n2riE(t}NPxJH!XFvR3nKRz* zsFTnTrP28DW{0}S&!(FnPVujQ2#S4Nw?99@{^tZ39sV&v00scaecwO&|4x9``gV5a zHm2W(Zj-9(Hv6myUel_uF8GoM=1`mOw;frBX-{Fhj&_noq%UMWzH+nswvb=!5t@cHDw^lj2GPADf(ZuRG z-a&mX0iRK12_OmMhWiLYu{IAp7tvPr6$S9#T2AxcnMIguQBRydE-s^&63jV&hy47@ z7j7xQu&m(GqO1>=0JZwUNTV_*WtO7+8;bi!pG1ili8jbl|KsJ2x3o%2S2`L3~gPbKma{4WRR`Scze4Py0CD*imUG%`IaInnNY%T5;zclOk2 zLh7JrPklvhWy}+sde&Af=Hai$q(aEd54S81&$B}#RSsfy1S?f#(6|pOLwQ)OnD}Oe0 zu1}Sc{LHf>L%U-;T9B2WkKaoRK@4Dv{0P=+NOUJjKoK;ZjFLi@0f7op_D=(dsg?JH zUto#cTF#)WO*9#@DrG_L2yTHfv{_-W8Nf1^+sEOtNh4m3bFQSHW5wbwaFUg0=5Otg zQx%;%v=2E2nH*`goUhr0n2fd<82a~h*=1d;+z--icu}Kj zQ^(00%}zZOS>p6SuJaw)1Y29jNl3$4K5|#GP0+;89Uo)8W;dxPgtTp+GOK(7(>A<@ z%LAetD@Aerx(UtLyMBW%ynFW{V0bF;rUtd}oX_DG4!K%qkaj67$Ns)6?DJzUP#3s7 zoV#hEVnZXvupLPqYNDhR16T+NrUP7orHF`Ja(raI&52eSEPkSdLrQ}=fq6{|W_3PFY=u^03*7R;&-6$0Z^}+FLPX#ip6okB*ZK0z=or7*ZbjXQOwZEUh2{j#IY^(; zkGH`tsU8v_Z_V7Lj!}FN15u||)Re?W(S|DRmkYWbG3t!(087Wp$>geOzR1K&P;gvLXD{Wa}q26yIdYdCWb@TtuB z%kDbFDKbqOFN$Drf`d7>&GiNNpTgA!O(4b<6ad(P1psjWKiY+XfuXID@waHrQkIJO z)-Af;RAE^mR>ShOX=d%G2|?B)l5< zWxjN!+2VDywWR|-$NH;c*H-N;%LLiUsjkeFUy&Wc=CQ+S>R89R=FZ>BG;)$wV2%sYtxJst(ch`MT$|(sM|Xv_1q@dG2O@ z?)K{N_&WXMVH70Gl52${S52KTs;m5EPlhiiVon8tzZO8;zgF7OxGI58I&=_6xoLzX zLNuKZnOp;}QqJ|Op&yb}JoTtggS{=yMYkS|-2O1hW~w($>=#sa;=0G9732>Ob1u}R zEucREB_sWPCry$i9>a2^XBrj@q^eHm?%q^hAWYng;` zBa2Ec%*mKZ`SD6~syF)8R=O%HFfePrl~PM#3$ntmE#J<(enzsgp$13%2`go~$mPfZ z*=NKe^}i~mgw9_tvNsq;F(c0`?qmzHLIUxqkwHT%V%i9jyBc4#08g9sJMX*cC;?L_ zglA{P&=@{SDx@*Z-${fTI*=B#gIjSKKSU;xv_j1^>7_y&nup;E zv{7d1vjn*cgi))f(lgy_bd58KxDq|LsO2&z6(+3a)rMAb9WWhT7o8yeg>)wDJi@Ui zV&prgnuAdX+o%t#OfUfyrEP?Bx< z^e#m6CJkk<+YE@=x9CpAS&P_^BDrVFOh*V4jWfVek>kl@rZxRZ*mrQ@$9?l$9DGcZt2k;{| zP#0+3SW+^*-NLl}f+>d%i|y#D>i}M-sIB^0uOTz0tI&41rA-bTlB1Jj(PdjTrJa&x zX&Bjk!eYnvd(D)&;_L(btAaH|0BdqDc8F|}g98?`2W@Ir+Z+MUwWMLwg$s(G(@{(O zCc4&3zD%*u7^*BECN25#5M-9csIB>*U3O^C;1hE_0_zR8qYLv%P07z6QZ!q7Nw1Ne zlXWn+C`FTy86k!^nBrQ@#dMGtoif)a%gY(zd?%r-;gjmk`And@+oGqZ3;FJ%!C8|s zt^&Gg^UP~+xdsXPrkT>m5c7X)Z@y%8brcP=FHcj3cVG9oZ!&_SUbEyqmmu%(|EI(X zWFA#V``!;(L;k-nJ^c56h^d3EiOD|#OHK28GX%x^xmu40G#*RU(O(AuLOD}(siJoo zT-1+eB~(N@c0{G72N8VeSPS)L(&xvs>x1PskyXCuk8IR(aH6L$2_^YeZ5@cj;R zbqlf8^nl(FG%ga)q`)AuIsq0$^tLi96Zk1GDslrU_pAHmj{wFV#W zXf)oMTMa*!G*=|FAUN?8sB|x*<7N^0DMLLIjapJ{n$6sOW?DI*kj*Ge-P`yb>Y2n} z2*-lH1!#lIWyvX~~kEirYNMp7eqVdhKF!AG;^gEi00Z_zxlGw+B@PHW@uo zzz6(KC(xJ&rk>*~;G*f+bZsm9JtBmukP!m!@JVfjb>PbPZ!znbqpMbs;sq&(Ztnpk zQYn{WrE1zRfKsh;diOfNHgCKKp5)i90Su~f!#W+0_VUVmUnaKOJwCSi)+YEwF)dVkj+oytOe%cEH zx?|*QzP?%zeX`{ac77DrPQEUzQZlYlQ?TTBUBnCV1`KWv>rKP zS37GCK_K|ceMt0>LDd`eZDpKexp@nk%hu^d)tb`F_@M2|$MqIi+Ut_oLG!zqFzInaA(v*miWe>=0m;hP%*oXGj_JM)6(4D{kGnG@fswE@Sa@9Z#+L zW1s`t*CkzYaLQ^H1ccwFINzcpIug7kj=bh+7sp=7E8i0$M7HPl>#LG%NY$5&%22tw z*&wv`nYsH`NvlLmS0MTBVq8>K@#VQdehP8u%J7pp}0bsyZtdmkB!kOd~_-po!J2`{NcY4qeWQ;lt zd%z8vlWl>ocnRho@BfRgb8HR;O15Zh+qP}nc5ZCjwv!v)*tTukwrx8v^Ko9){D)I@ zx_a-mI@C8X@xa+du@_wXhX-ezZ{2D?lobw*npnTV2qG>@Zltz~!v{gUX%>!LcCO5> zuanM*ZTou*xGDae1rJ}6NC(oXV=*i?nYsg-9?Uh-iRwvJ)2*+Dr2=y~_$X+?z2oFn zG-ChU-gWw+Tn|m_Ug{?lly*IKgHJ$kId&U{6K(`V{Ui9U#$6N^*xcFXKeCf6UMshaQJ0-JYO+NK zfpLOalmk(aRoTqC8(uVR(ntMZ3wPc#{v3jkd$9sNmK3>lcxfg_4b#J#1dj4>4 zo$xm~6=_OeC5&*X(2979B9xjR=Z~#8 ze!o;ac3rxvmvrGWRbB6dT`?m1@lD;aVFELn1UA3C?Vm2GMA_B%l>_0j!JnMCX| z753Ru=C2a^_e6LgJ+PtZM6WijT0Kx=&v&H!a@03xto%N=InQrrJ$8;a9{i^Z4+S;8 zad9~SX&Q9FK{L5Tj!0_ivLazT+PmS{o}|9qNdJf(X#16{P15&!NGlm-C?^dOPp~s_ z^9iK~CKJ1n;)T>Qh>$rB3#s}mWd@|++zCR~)ut1NY!QwubW8gZLvm{p>!G5`H_4V9 z5kw5IK!sC{CpzQ}5KFtKOBZAD#!-xWJNc08QkdSu2XYY;~Y_Mn1A(frB7cy;6npes+)^|V~oknvnUC}Ya?ci5yu))xrPP9n)}Zm{5n z8LE-n*IY9f0o&aDAjVZw6|N^zgtqrppHNt$8^f2y@%{v1jh;di$5->+a(v z91h;|C>QGm-)jasDja>jS$aQVqFOZ(m55xhdB4AR*zQ(}nfmCfUxXheixxCIgqHgq zQV;4i*?T8>Gn8BDS{A%S6J`R>aP_qP{c8HYHFeR#&Tq9z?*}~d7Xs)h^z}^M55zv7 zpO5qHLETPHZhO1iB=Py_@bU8S@R;{3kZ&5gxM|Cle51x(sf9m>z`sUbiUC`ZHKL-` z4|aCs;K!>GGY*h>5bY=4z#G=7y)uC+d4(SA@us$uXAk_<5)>CSpF8EJTar)bselJ~ zNGT@X=v~OCR(bsVd`Q4l5ickKH^m1?%!;VC~*=K=JD}($JWl*_valP;QK}M>4$f}Sn~OIUC?!o+%{g6 zq^d^)-PU^Eg8@*44jH>+H~%Aj`hGovZ-)!XrKlt4zPwl|E~V^#J=ptMv}m3uOWlKT z$dGKFoc3zS)y6TIrBBeCQ` zl2D#Rxq3xLMqdK|wu>f@Kr127WSU^hNkCj%d`MBme-N6+!;X3Lb=ZCGF+(^Bx4hotdj_KNgc{Z(SgP z=Tz(mMs`*m@F-kqzsn21d*iVN-J@C;?$|)_Ax8moCAVsieyI=Ni~sz#sx|cIwmvJ@ z`PLEM|D!ieH!NNx1r)D}qMPbo9%+O&!IE9Y@J>sggZLh_E*rJ9DLO$k(6oBdn_atD z+0xh;ytlePj--vWLkMQ@7X1-F7lx)1l{YIh>#Xl5x>GP0J%eRQUz|W%F*aI9vMO<( zXT36MdXEEL1-f<>?1`aIh{5_u-$6hFLFo*Ve1oYfwU;H4iHelNYL^^BWqy|zOP{%@ zRY&&juXlo0-lqqo!!ZmiNlSesh?kLd)B~LA>OcRM+q7YUtne6iQd*e4`RPcFzz~Yr z>?9n%TzCztRLMOV8=^tI`xp%y{X64O8UcTat`JQSK?oXTs=tPblqk)i>j#rEP&z0m zayUjA==kJ?Hkiwv>a-8mGF@hd$s^d0iHKD>uKYeTM!+=yG1Rl+lco&Rp)G$2TO00C z!BZHl>tv<0B68R@lPgI?hjS^BAs_*z5c@AQ=2B4jhJhMQm7U4J4esngsQ?h-{naK1 zvpkw@i@i^gR+5n*D{ZU3MJ5z`S)^DUENmiFr1LMsB{^hZ1Ob3e`!PRR7xMF$6(c66 z8IeG+Cv&2yG~oQP)S%#Q#R*rKl=`t2npb-ig~p=j4GFiqukmW5#?ii6aJDrvNRa<>V#Q$}Q6v zJX*xKwI(sD3Ch`00Csty5!AnweX#;RqW5ciz^*}DrGp9jkS1vbY^OQ^mhMW|{(F1} z7=02Tw{f0GoA9)L4Oq(18hK2T51+)7H^kDjgwV(Ep%ZLOHCM`o^`O0V~nhJUT9=>FjZRyGu12057#&r$76jQlb zy;TBPnt^JNAFF^X;z{^2-O?`&kCb>KL_$OA-A{tfJ6z{8@cG2kO-w~M{ z%ToYC_#TyuNzqkJ8(~l$26$absG=rtAV&BpTJh(7lPLA+1@-v~Yj;JzdFv&CJs6p* zTev<9eZ^0_TC^5;L1XAXFe@dylj|m|GcdTA;*2?iF1bn3Z2Yt%Av*MI5!aO|Ok-kY zM0`8}Nm>vvZxy!vVux|x8m1mqhMi9qNyBRD^gXkArlaAi%Gm;&oT_mNj5BQ{5x+zU zKQ6tVzX!pVS(ZV4xTbDAtqaTVfKixayqG~OD_T@o_zZd)z7$y*Sy2Y!?hVf{&lMj} zVit|U`s%WapHH748Ui!-O*u@P`198sjF73?9`htZ*5v9yqCXL}qG%c%YZ!uHsMpwn z(GZSW11`LX!lGP=%h4m;!);gR`t)p`5RQYAZ zK7RV0Yn1Do=05D`k9#f*>vx2yW=4747(M1+HY$^v7@DZ8hd~za9CaxY>adZq(U8Ht zO65h%UY?BNslR`~J7_h`C&U7%ri;ZYcO2h zJt%Q8ns1h^lpBcOxa3(skaP+BbN8zGRR+C|wiwZb2^IOE@c@)}|>UeK)81RcD zOBO@N^3mm_+~`UV*T}w@(>e}imere1m@*3dnMOaV;&@QOznlO#u!YEN;8^y#gI#n> zl#>Kjg0NtY6o?a(Y+D@cD8#kVM`$UyITm#wc>gtG#vJXX!mJ0>RTI)UwFWH;Vw(h| zXTT)DKe;!IbzlidK_c*v&&>#->hoGSSWv*^k-~(<2?mqaW%>H)AKyi9%XLcKu#cZ0 zMjMY<7(wPXgn50<@+MzWAsOV-5-zEhaHXY?lo=?7F#VA*E_U1FSgz0dZO4*m^G5S` zgs#*$!H=)Vcgx{w#CF(QZMgk95DQiTzv|j8BSy~?ZY$1kEwqz zZ4&;{Kd2l%1UDbTi7t0AR!(35a?z3N z7jaAFP#aJW6;a^LIYJ*eKZMVdZ5%BQD;+Byx0t4PG zn>9m`jsm)RN81hz-PKJUn6CmSx8Qlk9!ZH!Fi4Ll>nKq-$%Uq&fwQ8m4RQd&gRs+M zE+_Ph%IF1f3gW`xI6Zj4-euFkG|GatSCM?kTqDDcU;~f)YaUZ*X4EW@NZI4s|7C<{ zdho&6J#M1WuL1{ASKCfn1J?~hj1>U6(_bsyzx>QVk4S zk0aF~o5KqTO1Xf`ie{E8zn4->(k?)Xv?6dI6@1n( zD=)s~zm{IsnXOuMxIbm;!zzLgx{-EV#ymxn6E-&jK8}4kBlha>lj4XxZo5kjJzPfI zCqFQxOh|B(5N@x01i&0&{!h|C^L|#!MI*;^tbvwf#9ubm?S2nPFLkpB>$kLhSf|Vt z_Qy4ml~`#T{zcSVhv&7bp_-~E8=VeWOMK(-L2$jiYLEsaA}3O*Z>7SL<@6_E4=zg> zcrm{b{Wo(-A!$XJ@g<3vF5CvXT|&@SM8(m1VDLztW7nH&D5ci znk->x>^q(0-<1I?*|#fUgd~ha`dHGrq_mw0Gl3g>s|Y%RXm0KtV=IP0nE3k<| z`A|}HY*lrMa-b9#fWg6BrqKXAn&F3K6sM22Tf!hoiSUMk21#B&ho07jv8x3ut&jJb zK{XcYd9=qJJU1}EcheGUDB3fb$i1yhgqs!7W9ZGtz3;uIAqjQhNnqhzyDgLg1}lWU zo)|(si%R8f=9qmvCr5iOpG9aqa1LRzwUKnXfKxl; zlqAJcTdX*0EJ@q`Z_rih+GTTV0;3}kV1o|PLhIqwV#3tkZgdEgAPlFB=ug7ICo@gx zxZxqV<~dHYN*_9^ltoJvdU1(JqB)9J{^>uHHy1rArnOD&Bu*Ew?X>dC!OBE4^2%wV zam*7X*yjQ|VT`sy5-h5BLZn-nk>l5<_5q&?tv3o38S+(I0XD2%5rgV*u!99cz(OQn879(rdzc#`8 z;{_T_*db)5#LPr`hq+ZBo~VGp?!fe4k+L@%xBafY@LBG?9)vxyivYw9l^1!3C}c~e zoXkr{m81y0D0Nu&KL$=PEiueW(8l}N(c{2uw1Neo>&GC$m`g-K7D9dSt5+b@U(@S) z2PY7z#P-S2s&pLgW<64!oj@RjnO6B`2O?}xm1)z_@IS6W z2lr2)6+>RB9F$S2umR>6&f#J8a1;wWYj_I#?FCR-i~QjgHM8MTZymeb$m)I_B zsI!osA0$OnmufUjcs7oggml7NMQhTvzRk_>TDQ&>!!Ssr?hcbwm@b;r{8xE(R+H}R zZ~xo_7=A7;t^{;~io{s@qT)5Jqx@`8HViRp%LJc6u!MFLz|h8Oq#k7ADg+;BqSex` zgl!ULD5@Q3V~_RW@Uiks!_@h5ne#)iZ?&RwvYYdwq@T?~&4I{-2`t2D6Q1dpWazew zf~PUWrj{F1Jc=X-Mc=(1vRehyYyq>SugBYb`|i;v*v;(H+GZ&E?--WWA*i*A)&~J} zrSmo2M7d!2!2J5W(wv(6)pnb)Th_^Ub@w(Gt?PJaSD#((u1RjCA3YeKIMv?W?vK&F zxo=G`5QIY3U0@6F*jH#YSLt$Jm*}=ck)feqH;e6*H@0pZ7n}XuJw)#|@n+57PTw(y zwH6k1H=sAY1ODxG;&6ZjM-0^>{@Xx=_o&3IxuJ6Tq|6V;}tH{0NKs?i)-d_p6k92w`l zmoD;LMD3D%DQsg7yoT+)xwe{a<&53lRo=1!n%)6e_US!<@wrp$IQWUwUYc%mc?l9- z<@MuJuEeW6Tuf<&35^{Hc>s_uo_?q&KgvEIg6Ht&uW;HfpBa zJHAJNRp@rjoNsdyRIN#Pk`->#4eI@Gz};zCFLRmVdM?p3D53=^88*~V5O_SxS!wJ8 z#Rc-zgD2!W=L+L8oTGMs8C{I>QP|wM^zZ_UQlK)iva$$bv2eyYk0mAqE@`_Eu>6K* z93>aKhGK&g-OqC;6%m;bRs@WqMAE2&at=avN$WrCyMhrYZ*_1VDvT4l+^*-{Zc#)f z@kyZ42dFst8j)N&G|WmD}`# zCI(Vp0gCHz;G2M*&JDcB6Nw}ZmJoCoiZ9S_7W@%pmUKCAO4Ne+F1~ORqNv*N6mG`R zTA?18DBLl4i?7;fq4Dwgx~3hLnaMzwZ9>OzlASJlH%7i5e2cq=QBKn~wc;r#Wck&+ z?0c8uH*HG!0>-YOl=f0l?J}46k&IC?pM2RpN{&S6TwE@Yu#8XlYTIR3Uy&74&XTjs zV%%W<`cmh}b$(0z>7^#Q2+7>7x{eRsjSNUNd^4)#0b9&BH{Gm{+uUq5K1*_J_{$kZ zvjVGu9e^;8`xTxsFu2Vf~--NO~35G-BIY%HI zkMM2AbvSV0n>*@I(s?lgz(3QVpqS3_Tk9DI;i%ok)pQ_&2L3RN|fd&l$!-0WZM%Ze?T z{4!5hBFpCTnR}msY~gfT>DZ9!RT(PXFGytmfK{07X;b>hI~a|N9wlfwP#~u3SA@H~ z`k^s6{3W7KUjC!RRS(-bxZ}8izYTouCp$GmNmE=_bC}?b-)GXR!J+K^a>+*%p@I;bvZJemjPJQC>sBg<<$7BHv6VIed!~VqJ^SD z#S=)&TrW3<`BcP@iM_*{buC&HvLS5yj z%o1O0hhsT2tvb7Qgm`M%L)nZGg`K+q&qFq&9htwX%>0vsVCfHeRYApC9=&1v?jGw? zOZzQTE6jS>-dMGCTZGfOO1Y3D9<~>q;U?TA1$ZC4-C)-;Yh)6loz11P6g`B6`WLAj zlsE6}RcOJq0XN_FI2%{RVpp6sx5eq@6hZES@xE*-VT6(9n4u^t>mPRs?uqbI^P3dI z-egLuR&)mG$Z0nXEv}J+mup&XFw#Vsm6{I>< zAAUZ{yVqDQ7-^(@yErGn=K3Z><*r3nag6H!VJnSEay#`HP$bV2&(SzyK%MK z%VfnnYBD$G?ri_LKu;nbe%5Vp|ophrKc1E(MKEr2*4_68T1x73{o>puOc;qT49#9 zYQ^ct@1k1$zVbVHAZs8i8_0+&#Z0Ol@)|$1vDiLuFj`9(S+(NT7LF3iaP6|dv_U>k zRQ>F(b9bCE)*bj@T(Kuz-St?~1oV2#;y`||Dq#`iV4i}&wzyqW79DAUxK8rx$5!HA zUnRFdw1SGp**W%W&XDBjQZsupu101M@#^j4RQ5nW-?d`mSLe_C?0si%3NCUVV}hM6 z1D@-JD#zRa?u%Tg!#QuCBZiX^k?*nwwQ~R(|BxnoO_1q^NayQn&PW-tRm}q0H=8+M z7GS>yKX$M>B>?K0c#WM@{4G@fdg#_=*lO9?{{i0yt&)?@0%1$?jv(E$Y)V& zk2o!7tbCD^`gu0-XWMU}JXS+E`8-AJ0ZNTJG#JDoVAnOd*12e-aU=il0C zr|8|BNtD?JP^T zRj|?evLZ>!Mf$VKXWo5F!{aW6vIDP0nCa{ffdTu1wv(&=;?XK&(zD=1!oWP=+85{n zYW#Ow>BtA^BC+#aBEOw;uU`rtkX<`J3E=4Nn@tWJ#=Z3b_=oFMpAWg?gKqXG8$HO< zn#&Qw)LVSmEp8@o;T5SywCM-?jTYeQ$0zbfGDpa6^HbIvy6T~Ov~8Da>Fayw;Z;N8 zRo&=QP5$4~1}2980IrVyXgr>^#=`dZNaPA3jJlova1p#InYy0zM{TM9uqVkdpuVMAZFyt%xKd{15twfX|LbsSmUdm*eZA zp0LtaSY@sq@%MJ4n}ir&X%W^K9CUTj(!Vi5aF1T@GoolE(dliHZ<=H~_?H!H7KJrs z`Tb?QMIhhn-25ET-6g5ZfY-wtI4kfZE1T!&U3MHc2&7IVTOU>8E*GhFc!;R&tKJ8% zh(A%e1C-a`SLbmfdELM(m6XD4h3#j>z#a5c#h*&ySQjvWzW3;iAhLa#1zxe0#Rg6m zF!H?M1GWNwW9y3rlGO-KsGD|J;&EF~pQws?=&y54?%qRCLh~)I0fes@>%EgHkANOPFJv zJh~@^*USgtuqSu}-$8F(W0G3d#fliTNBXF~!`hWK({l1*Xw(nDfy$lpb(gD>*DoIu zD~vz&go9GvZp01~T(T6BK#enpDTz$Bf@Lk})_yg$N9suAAXdF*88ngGya?wwz&d~Q z_wtDOW~T_}-n>!?*n#GPqwv!5k7pa|8? zO_$5+^Pt^;?~p=q1Uj7;G#TeU*cC%;mqi}WabGoN^V8Tice$c6jmH_K*DxckX%!A& z3<@tcw4Usgfy1pViWnrl$eTna3tBs;#`EAe^c@C4=C05bQ?A#}ZUv^Ryuj8w9SHITBGDN^y zAeke-#^_6rKzDEU7ClB^g#sh8&kc^5I##LjKbo7E=J-oQR)W* zu2>>f{Q%5jn)d+*0oLS zHt$rfwyC98e?o^bfdB34Nle&M8z37NB_@u{#8WO=qC*W`hGt02uAB55^;}oo((iI$ zRT{Z8j*K2(*efAGQI$3tEhbP}PW@sK>K(BbPEQ^2Jcf30t`@Ju>{+-Xz%eg-1eRbZ zWr=}9s=|&8gBRplaofrl-0kdSt?ZyJ#9ufLzVgGwtLl3-{s#3HB0nz07IAbwnmEr{ z4;}xWtNC7Die3$?hv6-_PJ~uGG-FmoW5ETV&J_8k!$a)$pzN6PI*b{Go^N0?o5F0Q zu^Z0-(bWw_w9BK`GW8!Nq@43 zwEUSx9y1h_j`YzR1o?>u_ ztUu)7eBl@Xnh=cBiod$_bJq0Bl3Up6C@v8&oSd$#hp#SY3`xFiil%JhwZ;uaLQuUm z3%tlrz$1F#Q8@X*Bk|v>s9fY9Kq_*d;#N^4aV93wNghN|#r4W0S)J3nqsGaTu0%a& ziicl^YGJ!=BGy*Lgo`>C>5(w&sZDCf7N3@6hn8IXdv*S5*VGDA@cvc{4Gvy?n9;?XNw9#-pd3(`$TmNY~_LvWmQ&Pywp4+SnJr?{SAUpC*)%-ELPNo$w;-W=6ajuR%HV1?2hu za284zwwZ#Mxfc7HBkF2~QmhH9wjiH~-v8a5QRk%2GXy*V7duriy!8$Rbg~p&h-uYG z6Lu2p<;pAZwJWZynrox|<}n=XIk6)DE^B z?_l;X%HI4E|Bp_@!}@jz|3z~`_IqO!(_FJBeQ#*VJWy4r>xTse`L}JGU4aSO_H-*& z6EuTOIUbyv&2lrqlUHrBhpBZF9j&d@#;tF}TLT6Mcd6HGuaRyT>MCg&cfzD%-t1_mD2`C!p;}9f~!va z_-J(2gxpoToc0lk-2T?;^iWz_mg65L5+Nj`$ioB`V>_ysjhc?%`wM~f31gd6VM{VY zL~EzEEbovkO&uK_1G~zbGyjMxD$G;(r!Ah?Mo9G_7Fi+9{j(Anq5O?Bw++(B#m$Z^IbprzE)!`b7;((OiCdkYvQG+o3jl>jvu*g z_j7-1?Ji?Cyi^x-biKoSjQR7PX2$OBhV%aKJTaqQQ6`x^*}icTOf~67-`R(ef@9E$ zVFKCDA4m|MDo%`GKyTN(;}9U>UNa9Y?im=Ow`k4O({{lZ&?ARQs%%Jxa!kDEO4rQZ4mW?2u_fGEkChj$HX8}qVxfpa8xe?2rJ;5a3E zeOP1-`#)7IX-PP0$dfb^R;G)1xS?XPY9On6#}zbi@)GKaJXU+4cB)WdDNydK)Tn!g5 z%)p5Dc=Ni0n9j`^AVv;ZNE`?7?V!#GS zm(@k+nu3lGQSY;EkCOy_JzKjr!}&Z42aDrd$f57hCoqRuU$c8xae;|g8!dKxe+Fp- zoY9S)Rl+8e+9e#gjoO`K(W6uUy&kisRSz~cuUs~)|3uJ%$ZZdrNtPzSPUs+}8c)W$ z1r3hd(Enks`qKdDJ6htwCVgx@H35S&qo@u31m1g2VwJ{B3s{L*D=}&m zfD8#PP$UEOzz$H!7iwRclb!?x=pL2Rf_w(z(8 zOuhei=sx7#C>G*SNfJS0Z)Y0_+c_Yro|;DEM^1ga0{fu#P=aLVm?7|U45FVVrOtUAii z1|>|XT>nF^y^S@aS!_=0k6Y-wq+O6gh4}=(@Z#yKS$il_^Ncd}#F9SZE~M~$IKBqJ z&)08^u@Zl5r5G9{xH8RX?sEAj-$j|~cP2(6Cq1j04&pmT$k+o)MOZ7be|Z{O8@mQ* ze2tND!9QqMj|=JxT}9xr1su2?`ImIqlH)#4lh9Wl&Pd^Gj=?v!7uCU)30nNZ;;3_*+NcCu$}yOtx*qdFvX`v(1uUMf3OJ4Ab+ z89lcL-i}2D_XYg{N$n@ozGSQMxI?YD{WH0c%+Urdi56My*9oNnu%JWM_TuWWw(i!z zpY^$_Et7ZIanPj;NWW%A`600UkVh7 zW90c8I2^vMSL-k`M618%$M^o?^-ca}J4>MKW#MS7gOECC!dp$4k&rKwbb+a6_k4g} z67A$7GOOq3)?0PIduJxUGJ)qq0I;5Y(w91KN9bs3=;44}gQbU$m;X)Y2+hOHZi)yB97F;$UFNuS-=X!#Zmn#FI?red>tX;dG##13KXw_ zaWfHrf5@4zxCMwTrI;)~*R7NSirziC>tYx)^6FGOTB$sQgUsi7II)+>fK`6x$-5*+h{NSX|0 z%wVGGSYxX7z2|OQX@wxdZ!rbo^{$kp&Yx{*S?rG(*&Qf z0ML5{K}4RDwJ?BVu1V zTB;9%1z{kK5hh-?w??my5YqLwNW4XvJzkB^=~l+9oX2=bP1q=t0Y_!G*22w3R-DF3G8sV0+!tJ7b2jb zR?+hJuelu?N$G9qx@*jv(W1mjvJDS5QWZ?^7?8J+8vBS?o4UYH7ZjZCE0CaJ@U$EZ z;3Q*gbC{hw~pgxy{+V4&u*Nk=nT5HvwQNt0SPzk z%e__Q2yGAR!dJbbuPdO+CD*z|roc+d*92^`>bb~+QAS9|g_Nxz<`B8V;ouJ&4mjlOAq6P!g-l7kBl`h4ARUB1V~^wpW8#M$iphoG{TY{vLUt%7 z2dt$Sr%p5jeO(>nWW|Mz>x7!piYqj5Sm%8PT?IKhXjFKKfIr4|wzE6eo9uYFy)ORr zaQT#JVc6ad+Wfg~HF)Bc02LA+>nndnDZ&g~58UkKg&`2ESOHQ3>MMcv&KS6lC0#0MIIbA_DZ}eD+mo4FX2*RGgQS$j4QX zDkST#nsLl*&WH6jm=?L>i0#>JER*dF3Yn<*VrdJR z7!zkKI%1KFNI^7HuRkicysV>mke8%M7b9zeH0_|z(QvDp&eDMyqvFS|*EwI#^D^^> zLgN;xbDjv=a?L=1eY@RLOxaws3XKRho1!ohDjG$$#AK;BIe5M;Gs48x3Bq2jWHJ_; zz;P)pDvxlt-SniS>Wxm9r05cXxBi+|n}DQ{6Mp#@BA}W8Rhg~z4zRPlhG;{^qsHrj z3$>U$pT}67=-?KJhS*Rq3F!iy<@%BOvX2@Pg?1x>Sye|{PLWud5~DVi**wIT*3nAt z?a57FW}S9SX^#EoztuHB{`2-I&wout&o;YIXU4HddYCNhBRHfM6NW6r^#^K}7N|&l z>LobbjdyceYf}kL-kb+UmF{9*0`=r41nGF2A5|yI^oJG@%Ps)zHqRwR}kBIjy#hHYGBgC#qEdt20kBhzoq>ZFZ za)?yVj0~zNIWEYmL;wIDhKza;xwi$=b;OSLp6LzUU3e5U3HkwSoGr?+PN!#6N+eu5 zhwHQSRjbg*a@GQ)y->qRe!kqYDs(-OF1J{=uG)pinf;e=%XTk~G*R3^-VWu_)c(hl z{A3)S`ifP*POG{jutnfNi=$)<(ZIFGen#xGW^^;iuP#aXKoPvWbx81)uaHs#AD3{% zlZ%}@(3wzz?2NqOJ)^sqLzGd$Bq{C#`9zT*e~Rd8*0*2QXls1YE;T>wY8zhY;A}^I zRv=C_bXn#@7vfyX)D-(b`OM>=t9ZX(E#V=y1Oz2RcuqTC_wL|1=W{Z3HN7LJhWJRb zm*i0PByeo5H~sjdJtA9-7w#!Khv8B&$0}jayk_8Ekt3wK>N!V%;~#8%FVHT6dkBU} z1rwKXd#I!+xdw6HShmR*-;&AJdF;A4$Yw$2IWdPy1$*k>FY;{Kpjn$oc7xp-0Bn`7 z5@-=(f3h|&_i_d4*0%IYs3p$~A#)VzjCJit1M-zczBt9GiFSo2nSa(0O}MnxiqGX3 z=R`x-cR+=v*TLq|pn;1aKk@d+YR6xC!dnIn7WNv(D5rw$b zY4cQak+-;vl9~^BgbouLnhid;&XujnfsGX3L2EU08k6kja|#oV-qQ^G=9w z6mWzmG#*}K#&1siwHjA;o)H!Vjy`w**cox~f+VBO7TSx}sV~?Zujab0eIf1hHD7O4RQ7Q&v zWJXyo7>3+yz49~t*hUl_Jv;!V)Ia)6xI(2yje(-TQ zS;J;Tb?36bgZ2}WxB?x}mRz-Hr_bE6!d={V9IFC0P_J1jT*pK%J$9SHtG?BcEmL$P z7ZZ+Zt59C@_Gb%f*PFPJKZ?gj@1YO2)h!%P>BRXTY_2T?N8fygBoA2Vr8jr%TsceQ z2zpS{TuMdonrqgJOVk4%H&(v3CYU$x`vYe?SMLt$A3 z-;~<2Yoyj|HneDA`z(Vr@3}WtN+-Z+7p>Pu5kB>!MepDR#qGyQio~^S>}$i2^x_ z+CA(|oZJ?9>f9FmOZ>(jsHNSArKVC+8*>!tzqP%=`~1B>G=w;LhL*Wh4ubaz^3vxK zY6fF1c%jbNH-Qp2z2tx+K{`e>EO&v%wnY55#=Ip}PSMa(I{6@>vUjC{Gplxx(-mVZ z#^l{0QqGIKbsozI2u;u}{y`zqndC=o&>p5zeNVZ?dU2^bxIuUSy1DaaY(Ss0@Q6(R zSXlWepE-XS!jUt%M6?TSLa67TPWekMzwtME9ir=imxRxC9rUU3L^e;`jt?R9m;9A& zm7H-&YBwaaW*hUS96<{v-yMt{Qw=P}=hQRDTAigft{b~ghN}k1V?pgidVI(%^rkDi zJ#KRXH1)QZA_X}}BY-8aUUQF<@!^v~-Ype@MVUup(&J)HoLBTnUF?gqt{5%Z;$xu= za8;)D6=C06^*r-R_Xt-VW}Ptmq5I(GA2)#xy6vbKfkX6Py3!4jed)dnam z8s1u}SvN!A5wt6HxS>tPlvwXRG6%+wY(PC+nfJ^eh~8lMd!WaM`}Wg26w!oyWN>Bi zN?)EbgjT-R7WFpKBu-^V32WXapHVa!j-!6ztV438K|K0;%1dPpE7EY59%@fMSM*gL zg1J#Dkh09SK6(~V=L%n!duDlEo3}Hf|G8^p6pk5j|1k9=M1Akq5FbJa9Wkv?&vg3Z zfmRvbQATa-fK6&52l_`4kr@Kqo8p~M*E4*YCcjPsQoBIs9H%3ZAFeJQWdq3*MFy|D zAp_fLatii2M_?YP&yQ%M$j(*>=&wh1aFK6dpX|c1;8=rqqM44NSWwIorm+A!dt^ig zcEwOoIxTY79d!p{F;7-C!$oziO%(}B`CaT|Bv0L>L(ws|b-mL+PfA=Cxw`UouQLs_ zHIYmtlD-|;4V5kzB^PeV7gf*LgT_abvKZp=!SyMZK)0G~&}<`ny(7FPL#qiu&)P@4 zkCwF;E~*WGe6~B@XDxgJXBtub;MQS>my8&^U3l$Si}iO^Kr{Fb;ErbR`94;KwhQBt z7%%psC_k>(wwrmiju@nYZXYVMiRC*DH(}Q}-F>*@5}u*1_x3R-%<35CxiG=&cr?_? zq_IcUO4DMo7jC-%iqUF_XN+$Ya~ddp^{$eZ(u;g9&(GFe_AH`LKu!h$YlX;!&$>}O zVFfUvIO%|znE<*VRQ}?BdHJGUGB1C}eEt;p)b4(HvsH+6#qqzVj%@28&edoSJvwCq z)=SBv^N!_=Mb#C6uL!~CvKNb`6pFpdy8(Xe5Hu-r(>ow4c}G=z^qXtr>bD87TihI+ zD|~_KeB}w+^19mM>t4*}yAiebMU2-pi$xb|u?SyD1V8ulM<{T<8nmcWXc@@kevQI? zyPDuEK*vwPccR^|Sd=xCDA{r6if?b_)Oj#%EQ$L+940!Rcs{7Gc>*h)+ZXGaUDhU5 z=`<3S>C&8iVp*K(K6mNgpFbYq?|~)?KjGz`={I^JkE{xcvR{3Rbe`QvSZ6=F;35gp zwoS`7+fs6Rj~v3&x5tH)G$GragR%BNAR0D6Sk%TFoS50PB<+!uIduCuQySjribuL5 zf4rMR@j4b3_ry4oEdD_M_eR|Mya;&Y*FB>6>mK=^8*xXY-;H>(hRuI8tiR`09qbhp zB2`!9-s>S{16X924Q;eo*4_zWjkWS^@|cR^Rg~>keLrI96DYMUl>Nf14tQVgZ{2xJ z7%(mYeaoK=hguT~JSf#vD9H{f(91MGYJM8b&}OiG1V}pH+@VaW%C**(IA|-HM$x5F z2P__bF^Yo{c=(YpdHrUvCga;a5AdN=Dxfip%uk1>_pgzOcvj^@%7cWV+xz=u9lwv` z$IJcuN%yJLd=El#FW^a2JGyaIdA`3jdc3<7r4l@0-Z zG=>fVheOl{L0M^(C?4xiP$OY9bY|-~ZZ>W6!c=FSvaFvVPGHVj_lk#v)*I9S`Wr_x zP#%$=dlyRlaF)y_1X3(rad0;cIr1-RnJD{cD?XXM!kM1ps*NGH-6@^=CMG8qx?bX| znRJJHdS8Y3MfH9S7LduTzu>C@G7==I5iP-Pa6{n(GIW(FAMH*xX9KkTeYBm{b-Ix) zw@$GJd@EIC6Pu`OT&%Ftnj0fq>JMOtm}m<+bi)m6a1MjwwA3(9FfZo&bvgtu%lg3X z(`GA0XmNqSwS#A+)Yp?9EmiaJWbKbTNgUw2I!l2Vb!PTq}Ptssc@O1v7P8K911uV(6QN5$$ZeVS$XDM z^jyw}T(VyI)x3@YBR7o@FN+S!@}RA;cLj0nx24UWk%1^aQoSw)49>N|b#UIO+wVqK znGWo-n6Op`uWq3TwO;bolFOu*;;GdsXQS2+YutBl;p??g(}1J6oJ=4mkkNKzZzTh3 zWv7m0=&eu$T%qA8m$7;ks9R)M7}E?}w(7HYv;;62#2^kfw5RAulPJUMCCmlsiD&WR2`J3}w8SXwHtxas@PAAu7oxY_Z_})q z&&B77+D|4J`7TG5N^SVK1Oxw5g3DB>hDy;JpTy4^h$uG>Z`2NSBu(IalA@2}k|KZOzPrsDV zKPBU_H@q-X8VgTr49Bhd`<%aPCHHV#X=^_%<~hkg|9A&Qo&?}@p(Q$Pieo|C;@K53 z({0VDSTH|77*S}r2J}QpPAm5z>K82e%hT;*h*@3%V-o6;xYtBqvmPKsebSk0>G^eL)THi!)D9IMWao7i zr2TKFL?3$Q9Kx`l-L5eR^*JR}Pd93NQ+YLE^TL`^CJJ>*$J#h|%@`z_JHr-q2H9qO zKD0T{b=VNov*MQZMsA5d=F4tD-FZZ0M3&Bt&L@g-2LRVeZ=V!@GawOj*gU;{YS_qA zNj);@8PqJ=*}**NMT2D9B8ewC-n*Z~ zlwr;JYG5JaEAoA@Ix_oQJaBLuz$O06KclEe5s7>M!s|xr6nviJN~I5acCV(rhJ4Vy z?H})dAAjYvGdfbg17Qxy|6RRhrSD|y_Im(U|7RpDlJMJ=azbn=zQH7Or0wEZHD+9U zt)y@ntsQ?8kAi{-0*`LpfQWvZ4@X_aIZ0+CEMBAYUapYD7ulcPof%nXKIfV$(E~jaDBg(VRZhTj5B(e)p`-}2#A0^>nn*5*6<(he;Ew>T zK8Xr2Jt{oQJlR)6VwH0)AfE?938ZB|+6B&Z^jFGHdc;)KTc7v4n^U;QO5Gl>E-U;b zVPve=W~!4(rsz|1m8dh-B;?n^)ywHqAMDxF0|Bjny!q7>F(0gST?zSd6nBj5q4j{D zcw`{wCWn73sZ#Zy*s>B>9Dl9(gz$Zn#;70u*iWX7-(pXuv8ShB`F~A7%zP>Oszb1N zfm;{t5>5BaCgR5izpXCqIzKiQtXn&`A72k)Kb_pMzl(r**}#5QX^eg#j@-c=+@p?s z&-h+gMe%)C{jt$R@q5qtbnpVbHvFq>V)=M3`DOx$;&h$xYY%=uuK2q=+_7E@yytOm ztwbg~=LR3>N4uPHkzw%!c3!WaK0eR)wYogro$nC+B)9)Qy4NXSB%o#auPB?jvHEP^ zIg~ngs+1TW137)ETS;#^WdoUH98c$;*ZD~KI|9_Wlvp>({H1|#uRKhF^~QIp4B3^q zUyCD3)dZ*%UmaVLF3Hw0Op9dfa&2L%l)JV&q> zNHH-_K_pXAwkc6Eq9jHsoB_vG_KVCg3KwY1O<|ML0q%=1l!mSI^344|1r(qp#A>8= z#Gde$0x|!%r$EjAO#b^M@wMUsi2i~gGf?9k;MjnpXOO8&G(zo#uKUC=DTWEg}6?!G-6|AuTlnHcOpn% zZp0?Qgm1;Uya^-tV0QqI&ObGa#7=UuiC>O&|+={-P~G00>tY*Px8foJykahrw+-j2ku+1fz(tOY7Ak{ z0yqeC=5v{%030BEMl9T4z{2;DfVR)~kOX0*Hm>#ls+C7@30WcsP z$^I~~VlCaJz_+(&d>iUyu|&VOyumf-9q}N9eXIiO77pp=>m?aoMF_}E4i54sE2x99 z1GtIA6tWHqJOs=)n7ag$?Eaf1^0W2OgjXlDr~sN3=NWg9@bF~M2wf)zb0pLcS>%;j@Y%uJNK^+y8+9bR?VfcP5sw`@% z5m{eomyht%JaPmXmGK^tk}}&-!fNS)To{_ZA8@>p1KzQ@Ko67CP~RS!m~u5aGQffk z4xStf%xg->Tz<45JQh5WoN8TAP&Ps30HS4U3>uma8TYHxf~{0|6ICKyHcHS3MlPXSNWOmTQQucrQsSe1lEttd)<%1Fca|NZ!BddB*NRC zGZ5jCn1aTnJ?cwTpI1BN{d^sxnD#{tqg5Z7pLK-6Y!G4^n=@3H?)Kb`6KbvNd9bzA zPzwL`4L5=Vr+4GG*f(CB@QRc&$Am#;!A?ogR(7-#!w6eXj28rFJ;+4=NI|!FOqOJ# zCB;p%(r7b{n8G5-PBSu!@o_Zg(8-HqorA?9Y($~ELVY7g#-)>jFPam;ITV-e zcco+w*wpj}0vX!lD3}vq*bmojJMD;-an3fb0!(HLP;BnG?>3G&SDwqBbtkl}EqI9{ zmS~O3X{AJq8fCK;=9#K>^(N7)y?Z_VILiQXNS*N+>155Hj4nOnFado|bz20&{9cR6 z0}EHrj4mMHRl;vkM>|9A!#`(1P!}mc{r>OObe%b)bC@+RzglfP4Ar#VmxmXT&V3_f zfCzNZsJK3}8}=+qD9V>kY(b+PCvD(Grg=@)87vl@T?oX!*7*UTVDo6W3n9tro`F6~ zKj`omXtFF6VId?m{@a9AvxMu)H(FGzea`$Oo6Z@uol-I=sjTk)RU11E>JQN)3EP}W z|4_BHvr)pQn6x|CED{CvBOhV+xssh{okB6qH1zOx^N#Yo>GH4@%jS#=EAvyTZi{Ot z_4^WR@88-4yH=P*C)T{&&g+F)*THhdF^eMTrJhQ1{b1GUie&#clpAZnC3Z=&X2Rg` z2Z>v^A|kOXfC)2kO9V&G3I_aC`e7_$5=#E$KD~MXRMbcUbx=yNA|cD6qky{21dR1~ zz|hTHQeoY__xwBd(F7O3XVNrlK=YV0Hx1>hjV~O-!qlDNCar|bbIsexVLrEl2LL<8 z!CG4DR@C{AMqoWeKZgPD44yfyMeB>*l!5h%cBp~WNN|Z%mGUd1%6wKm?7e6b@D09 zs(}jSwp4~|@)tQ=cUSz+&DkAqt3{1~A@!sF63Mg@sVFl8TIN(QS)BLeqRGQcdO9=` z_qsS-gXpIx24XNQH()70(anqQR%hDnwPa8Dd5Dvib94nX3F23YO-v&u!~~Ws3qi&0 zI$K2o8M7RD$k#AgXP&eL=U8c<_vELAu^Mn4xIq2tI^x8`)r^0gr}JLpNl|G5tjbj_ zr8`OP>~2L}v84Fw3jNXd-jHK@m^R-fH_6QnQPnGyf?NU_I)aWmHtA$CN0$r4$({O` zEBXHcgBBy)jZqz9!bAg@1@I&ys5t+8ahUQQBQeH^sq6 z+qzfO#v|0%5fCwFrl3y{g#wU+sb??nxc+w*z+�rId>iX_0>WJ8_bLcd z$jCcouOGwjEUVx<&G)f`6qgVQ-3sB;P1ly;<{!CJLvjMC?u$i0|EX!IW*_`^Rruo{0(# zoeLG9ojFMQkHRox;WUE;^z<@Q=J_bA^O|{Nwdzcgej{VOz-fyBB!E2T9Oc25-eqhe zkscidQg*-IXjthWQI7f#Xx@&%6};!}-cosnwwqk{#5&rUApleg#EzIcXHng~-Fq0S zd0yDz-pFy*UANOg?f_3b4xRpOCp(^jqCMVFR+^O_ZR?Q{qF3GCltc@79z7X3ldVO0 zY9p}JC>2*v%O!tkmOqwT54hFL*ljnoi*C!k>!%7tJp+xX95ZhB0Wp$yb9NXm@R5*f zTLFgS`{~dLkrA*CBqw#~g0aP6rYy_~obz0ue>qe1%1CBdah8(c5Z)=n%W0mi7l^-7 z5elCj(l!9b^s#WqQ2Ni$-8xAUg4|&wZ%iW5oLn}q$?;B~o}-8Po3XRYLk9IzV~;7_ zxo9O%K->JI5*=+`acLGsNf?bxOB}aXpMKlR|8(JK)beo>lLBrbMEwwQlgY?V=PRdX ziNx%m>DieXRC7K7H}ehj6RHCimVetuTn zR%l#XllW2^BaT(1N4a93o%q41sLGU3B3=MvbQubAm|?NY<-d-)>clZ=oY!muS3VMf!PKNye7qwRZ@}D3?rC-66b;$q;%_N@ zyrjm_T0E#Yzdg&*J@%-YuMzz5c89HHeSb=4?_#8rd`hW>i*{{(UI%z&$dB}MD88Q6 zM&U3o03%w_OSvjNxOfbmRqUEmLWwIAru%6!ct`kG&W^0jZf^UN-2fvjvNJZzmKkg z51cqq3nM`{610J0q|6V{qNY`p(J5|CkH@ys+Msf%B*^@S{qxBj3^|%_zi?c<$ZM_LtJ~>ZQu`g3K(XAa?4%m znk!2bWgOs*|H(}sgZKTwiKTj>mVrIU4fthG5e?yo<0K-?%>UXv)dG) zTr3)Mebw(K#7sp16MAihnQjL}raau6E?$sQH26(LaI)nrKsQ)xVbHB2#}e&d5P*dc zy6Ey2g;|v{DY8Nx{o{j(ET@cu^m$LA^%N4z$4xT!kL#2 z%L)4^>xlh=RuJ6WY}ixPpmU0y46vdEJ`LKj7oD@tt5-KjkHnjlsO_Y-litLw_K-|a zv`~$cTZ1!bu$(x5DLck_EEZVN>BJ{WAn#_Cnt`1I#aorG8q@CG zz)V$q%>VQPJ1&nj*AgFXA2I0j{E58{)YPVgMt_Kh3ggiv8@iRyO7P;;jpBV3MmDNw zBl+*XFgwwe2Ud8id^59hEO%qr(ebC_+d_VxNGB4(H(nO+k!!k>un2YA9sD>febZCt zLkqhrP&IxS>Zqq&&DnYdnkD{3jJlxd$1WRV)6s;%MGQ+OLjty~CKE0AZ=AOv0+@5^enKT3G0?8ehN>9jzU?6fD7XqIJ9h$tv|(=Qi}10+ z@zzN$nbF;hjwQxnN#b^Ks=I5|#YyIF+Tx(qZU@#Nxz$7ZUMl;U4XKH=pRJ(#X2#7=)mnm`y*zDm&_jyfY!o9(7~wF zuGqEQf1!mD@~br}^%d89U8k-sqf-=+5f{w7`pDL}PELRoibBA)xK4eK)HI)IkT`fw z0IH~cua&*7ek!&wm>tub(Kv2SN40PC8$e19uYfG8S(d$c`2|CbCLT~;e7?>*Chr1H z2ywgUw4a<`3Xnf9%FjGYV!bCXX4^gy`DtWd%%S_mLiLqUUC&qwPT#@2uU3;<_G51f z@$Cr$;-NmOxirf2BxD}o5Uu!r5{RcRRq0!GIT23bs;?eZn^~yQvel@Lp-0--mz*(f zrD9+t+E{e16pmia47!KghTgdXy;;Gx5C=s|!IkRu?Wzg7QLMG*1oR!w%JcQDRvktA z{#lCfK`uhoyc})mq^ZX?N-}v#B<(c_6IQmVWdb zyz>1r+BK$NWL7)AclD1yyN5wpcX#VL0-0{|t^Uy?IJ`N$w`N6j*#7Y^Xt0;?ny){EQpvu0;$o?3~ zLV`Ru(NWGy-Cw&QmfN^0x5W!gj?*Xbq0iCb#c8us-ctIqUYW zTWvgeV7K75Nq->d5t3}NO!)EJGrP$-9*Q$Nfi4EnMK}*#5|sM03epg6*W*Y7iiqwT zOa^1_?kd~c!_&&H?4iC5)Nfy2bE?0png4KLqD#d_;G+oMD*xc@mtk59MFGYb$`*>W z;P=F5=Fe+WehV9bqjL(bZ2LA4UDel3?gd{h8w_o#d=-G}XaRHX@t9&IspjFNtGET9 zc1<40*O!`R+0I3=cvLYam~mI(Mw0?GpxDLsb4wFijs`)@Ja-ym{lRSxCS-+d4qv8_ zF)YVzm+0NqEpAg4{>1CRkO%7rMq@+K{70a$P)`XTUqjI8A^D@m`zLA7)q{1$8hQT~ zD?rWDnSa})xA<*$CFA|?hCH~Y9O`IY(i^NQONf|lsYziBSQ62neRIr7eQB!)<}W-^ z{|wM8-{0M3bRcKxu>Rru-zxOoWa+ih?-17G{{w)qbuu${_*J2sRXzVRhW**i4S19j zlh7uBsnci901`j3>W^>}>w-~3T5A$o-!fXjKkty4^M2WmUz$ugJwky>B$h1KTvM0^oRp!_E4!BgD#>9T!2Aj+AHmYySp_4OM3bT zDFbHw?K<}6^J{td?A;|uB0dRY!V%}+yPg%TqQkO2&3Ae4mVGcG@2a0)x67-G8QFrHd1YxV+HuFN?38$1&>I<(lF!>C~-~C7x>xjpCootAeP(zVHze2;Y3Hg3lgf@Z(4XS;h z<{S#JGe4YOGT|=h3;i@ZS#dQ%SVRS+9C*{WCE!7g$EZC5T z;$*bMnx7xX*rjw|WH<<498-v9dO*lAM#xUuHYPgnf)_9>nsXZ@2yG65$|Hu3j!~EL zR^8kIn>bZaEtv&fomS%y2tu;FMQ}3;xiZ@(jvEq0@Q=_uwY*hzTF6vEh@^h?3K)X~ z6*l>&f*i|6>J|v$HJ4NyX=8dZKW+DKZ*H$(-JCcb1j2UsI#r9*g0X}E=y>^kwEhgf z@33eIv{6)io`QDCG(o0jV&Fi2;7uJ_R0Jad8d6ldiU^CCLmoP{~kk; zUg_~cBbS%4G%kAf@bW{TssfO`r5{1rHmFpCOx6S(i9Y= zt%qIBMYS30jG6Hz%|hC(!&0gtK1DX?Xv%GI=n0eb!kB2A&=!n&IyNr~yWx-}`6+Le zx(?F#;OEIQO7;Q8~w#`%`KF${&x1TJJ6_U3HDD|JOktgY_upjxOmz-qa@OUh}|b4 zq%|{#l@P$=Rv^s)56$pDcJrf&5GSh;uv$zv?9fxIs6&-{I?_DB2EU>*_2(^L!)oCH zae@@-y}~d<vL zTu_UbW-4^K&e4T_w!@+HZualhCcKJ1CJ zT{10LrP{|ZHRpZrnBQoqWFI|o=Inw&LMOrzVFR8Rr0XkEHAX8}j@P3aZf7|SsKEJ? zGgZ^q3Zq0}R?43|TFSEja?-tfg!|etDJJE^7U3HZEBT7E@w7d&|0sQ2r4zXP4vN%r zQwk=WV*At}5`UziicEE@gi~3h=*nisMbl;_FsSeKeL61uM&;F5N^=qjMN(rnpZbyl z*{w87@6xBiIe5zF5p#wg`nQbT(s$XT=i~ZzM^Sk<>8>gmk=E)3J@OmM*>q3#X&ef1 zg9l(<56T zc1vRtVb0d>`udufhv%-~1if=-@pJs^r^PoJRHa6HS*rO`>G26`4vpe@8IFIIt$g~| z8AVMdPZjIQIOfk3;| z(W#wUGA!ejY{UN>n$0GkPJ9Of0Lb~JHxvGk(5#)ljj64*zJsN+-R~Kdt*mdmNC*G5 zqiavGuw_FamCRSrl&lAx24Yyyms%dNb+n=3cunTQQepn-G9$A=Y}pC)wlL=7bGx(c zxM9<($tT;9K`XD0G~?8Av|?Z~kW|^O^KeBf%YE{1QFbzrr)n0MzCXFpF96udW^pS7 zvLGm!;3P#Q39gF6IFi8wyx**FFX|H=+ z7I+qWd?tEX`UWqG_`H8T#MvY`E<+vOdZJ$GJ^K93Tx2Q5&XvC;u03c8)IAv6Up#{(fsEeh| zh7&;&eegW?FQYSj<-N`~CNC(aZtck1_zy>q4kv*0nzt6xu&qdmIS}Epkc(;_<%yH( ztKmU=Az=JY?-zXUypCOqfN7o1wnB;NX=L>Ub1 z4pd&27?v5>X9?w}+CW8uicN|A>EgSU#6A%nq%9!iBCvI(2xv_8b7Ahz$b z{DEC7y-*9o9fBvcD~h6Y>~EcNhSA$ zWnAsPYvoq3c3|G{^!{8w;(cAi>_-v?`~E*D!u)6Ck}?yq^I;&irEpE@pN;Q&O;2%|rAF*AoB*R7oZ;(kR` zofA@0@sZoZJv?d_OqkhpugD+MF{Tpl;yO11#ct$Z3M5ZgzJs|)9MOx{cz(LyY5SSD}@7&yTu0J>Tcs%d%yVk?oMekPM=A&51=*u2d6?Num-* z=|%>sN5uxcDfG1NpM_ybCFfXqq~40VXYdluvhLEzLFveyb!T%*5=G48A(eXBr{-~J z781KcB^$KmaVMEo6AZh4eT4_{&qUnl8u?JXx>80!iQSICs4XL0$`)}BE% zrbl%6`+O_^L`sVd-r(=7`&2^V`DjB~a4=RY$EFbmDfmZKt0eC)nm{3Kk5V7`Shf(; zfnpIU0(KcG^$4KZNEfLNi%YFe!cJwZpe%mS_^(T<59f$F3YeHU=2o!0uONM5z4sSY z!XUf!IvDholS=<>O`PH9k^3O$kaAoXVPM;Qpq?~JxLKt_?9&XiT;1yjcfLB!#->qdriw zz+De)-5?9!&wj`N2?ZMkWY*$1-E+r+ znLeW?flE;w-TBBrEp4ADC5Y1>Ou+SlcE1LIpmG^bz~5?2?*%If*ah^hUgPabbQ({3 zu`cl~=CXl+aQeZ8vVfX{Uqv{+ka%s4<#Q@O$IIWgl~6@flb8J4^4t}4S1m`5nZT#4 zs@9fhF+h|Mp-knr|vkDF1hP@i~!-$78`?qm(F$$y@-ipGJz~s zvsB-}d_^qn^~$^hmdnk-AUl79J?TPtzmJz0wod9sheWaba3dzaa8Hz<}})grvUh{4Er|A`_564?lL z+qF6v$?D1*rgGAc9K2ax4!l~+rO!|l7e0-$%1v6fG$zzLvo*sq8WEIi9mRqP-4 zv=bOk%H44zr4p|5G8+!j0T5%&P|v{#jqz%~fY|PoYGyCBQvIj%dd=T1G{}~Q+lUuX z>M;GE;X5KACA5oERG}iWkWt00HcP~_SqLWs02*s(g}WA(H_nd?Z$Z*?koX&Z64~

WACl{VL!OG>)hJcH6%w73Rt$DMZySnT{*Aq&L# z4As>eO}={#QZ90{lU3vSm_98SqCSyKCm?1-E)61q&j-ma_h&E#p^pM&dbIhoq_BUl z{vujL`Uq(z5#@em0sAZpqahUZQ4>Ng11+lnEI7J`I>TC{)Bt)jR0rH@A{cNM^Iv}R zyS6NY;=VS!;*&!EYOJC4s=NP4A^vgI_C|306nj6RW5Ft(EQ9PMO>-bhUu70Q|5nIa zLg;=`o@Wy7XZJNDVY&^A3fb`AvE`_`%TP z4hN$NN^K3et})YSU*_YoZU=MHEANIn=d4-)tOJJ9xSkJ8Pr$0a%N+0AF1`1Vb`0|k zcbelayAe;;J8X^~f2+_#*V>ZS@KkKkdh8pq=^Vs6R`p7G2oYj1xaFn~&S_=+J3NV+ z_LR(y{?M1dqrzPZ_3gKq_JLZA_)$vHE<7ytE8HH}dxhgz^Lg=0v}RBxm0Oxcn2hq*#cyku>}*-|05?n$4Hz33$~u zXB=k(hLE*@0+$tf7#_^<1TpADGTcjsrE3>_s@ip#+68cVU*D=_Kt{a}+-=65L{5F3 zP1?;V<6cXC+udUWIDw~fhjG{`a6OF=G7+?esI!h=7l`ypBc-TvmjK4AE(F}o@FRuO zlLpePx zHk<3qjRC2_yv@G-ge?bDWI-KbM5HD;EK}y9176PVJG=t)H|z`4Z-{V5Mi(PS_t?;>Ny*uy z)lhEB!48S)8?SwAq|eOYSgK?YYbMFr??;kwB+5n0`mybgI~k?_1TY)qSrX16Oh*G3 zJ`>J%PtLX;0xmaJyNU$_ev-z)+d)a~y@UG0E0YclE6FK_PE|ff?#S*z$|$%#W{a|9 z^zbkYK)Pz;VB3tgUA@~|6DTcO?Nb21N2k>oA!t$c-+lfZP>kN<6U|k4557a<$ubs; zUIDOYGH#$X3?P@#0t@u@MEwlY+dM^m$~j3)7OtQ~h#|glk|iu|rK0DPwBl1hH|VE* z>|w(iIF0GFI{hMgq^g2`VDUwdBThzoybVy8hNYf8#YNltwIQwxZi56ETvfCnX))nN!xmu_P8a9iq6X&#&o)@Npp7fR7)y$o2?CJ{`AsiQHgHR+$?_fKS zG)^omSXI|sxa+t6?o5?EJ4S2BJ467oW?SHOH+jG(2?Bvm`YSMDPW0E!V$UeN5e?Tl z8zgtv*_25#f8J6%Z{OQ=wAMp{88~Ug_c1qoL_7n!0N>eAFc{!!JZIjrCtc-(G$3Hu zje|5lQ$T^P7{{mzOcr;i*{;9?_$n)Ri2dzk7P}SS5x=bIFWWSogN2PX zsI!wv65JgCKjFq>#2}K!Z}oVdHh@FG7&kZ2B@;FXG~u_ubP{#uf=9D5$E|yr`Ye8~ zTjba$MWe8CgQruVUo!@b<&FFYjGdTxI~)l8Dm~4xSJLb04SD3UF_T_^!D(X@ZTpUh*7Jb;<7L>iaQUG>sm0-#NA=Jx#&=QL?$4F$Gu?q+!+js3>hKR{#t zWqoB319820l5ga?o~?$F&@I{CGmaNRmVTomd6XLpc9;xE(iW8ZyQQ{v;8MS5lRTH{_;fqS>)^!}#^ZvK-_R_GI;Xv42>&%=y@Z5yXu?SjS)3#0~&WE6!&S%h;k4P!@{5}OQfTtgwj?FD zioz;4y?8>x|9Ws-Xe3T!ovWfzpJ2qKYD$>xU|e`?*M|SFFYH|>m~SF4@pW`KkoYU6 zu}g@6ZXBerhtPxjIr+B$mbHrHI_n)!C<2ew(1}GYy4^5PUq4iEt%~09=&>e#_9Ss} zXA>gq85qBTB44N??^ip&gln4x2ivNqv@=Y5XVY+Sllgpzl~^%!L!$jU!{t$ZevK80 zHDo;sfOQLgS!lR3WKUXXTDNdPNT0EN}>WDnD%a z`-jiw9KuoMIV9p0LHrhZL3J$05pYLMGrI*Cn6t{pgGl@zgp-nzgVYi||0P#+?4-J9 z=n51{2_eS}KlYIvK~`TK`%IApoCCWT;gyvmp(TF+01kdBoBu03<>>OCI%RH?wN}6UW4V^@76AVz#MWxEq=j6J$; zcsI`SNFv+(KuD0I%RO~|#kGUcoKI@mL{c}%=)IQ^Ioqi~f-w+*j3dL=J zcC6~cfc!lSX7ke$esQt zUP`Fsn)<%En%;hLgw_3sSjp@+!5pJyDJTo>XDx-D;}?#41(miGmT5rl^_CH`S}@O(;>2A0Q=>4p-{nD2;0bGF+LqeM zA2uwYY`UJAwkY~1AA^Ux2{=@$aK#y+As_??;+J(NkB}XnTp0{mHikOoDl`aya4buF zwLy##8;)(M(jXKFf(NMK=N0~uyF=rZ_u`<&Zbo?q8{&QypNTib<+W1) zxSow@s=IbX;o?s#1R7P4wZxcpfFTWnA;~Hh11(bN%PyfX8vUt;$mbG`K2%-bSOA#( z^Hjl30JL}xGo;6HtvFYrprix^AHe87<1gd@A^+=moPvse@V~O%!Q3QpEpogA&_U)sCi?3Zt$cA@e_6>np;So2s(;XjkBSu>uj7d7%fx4l0i*TZUj@Wb*x#ZujG1VBxNk|gZ zLODy)(h*)geo>kC8q^NrDMu0ABPu1DF@gGK6j)s<4@zz$*}+7euwL36b#$?q;{$Lv zmPoWhkn3<7dkD=Y$OS?)z%~wh$B~D4A-o1HHF^NFWu_&B+!_3sJA$8<#t=YOeLcD2 zfQ)eP(-0c8i3r(;V$_|`2ZzYXg!)kR6@GlMLWt%3c%WobAC3{sc>PkHx+$A&;T4}HbEcBK+jaPD*YMIVRTUuH18rN9(c5LL2NKT$(!2+!Yu~LgaRF$ zu30N9*8i!};X6u8|K9&&%F)AA$%$(FLFJFqFRh!+v{$Uo~M3 zdk7sxWWX{rOJB*S9>=b9j&i<080)&MqO%Qc&*~C9ue>btxH)*Uf08x0p?KNGauFM) zCs4~+7I|rSwYR*!yoU93>SlX;zwW*R*hK^Tdu3?B0w*&iGtT^&(Y3x#=voVfzU;Gu zSJutw)T&i>E|!oSc-r0=*f3yY!F6kBVgSw{e4%%tbqOB!I`#5q?&qMK&$z6CD_~nn zExLq3;n}ePh7;M~OW7N5%#4jL?4Qo+E-s964{k?+c6E3Zy2fKUze=SU$U7t|XRQbr zYn7$Y`bY(2J%%T1H&}+!^bTtOT~)Q*bCPRUe9We(7_xYY1T3<2g5_4hZGORKh13Y0 zdRwYJh<#ke=Yl~IjKV_8oOTttabe9>2CDQy4#JLZ*Lax0l!M8dqtoH6-mF>s$LpJG zU)LIzZZY9#(s$%#$f#LemTx;`?rfq60-0e-)+Wa#sntbyoG2$ojw}ocp5%14-F4ce zqF_UAfeqWiO=VVCV~af6A1eRl-rU{zs1V?$Bx? z%;NVhi_kO{hsM$_wVOW{6JNzW@=LyWy(#Bnu3=-!3QbOY;k3%680d-| z*Z1)W5gfcYcM7k7TG6O3a}sT-x}YaD^E9`VpxOCu4Y#2T>)e>abkT~etDX5%J55aW z*qQ>;k?9Fs~qVnY*egRkVsFf&Zbf*3^=y=Av&Fw$tkB~b}3TsZqyi#JCw zX7muDztlz*Yfr;il7LmoGLB-97i+}a81&oo2tz%*6(=DNcELZ~%xT-K!PrB3#U5r+NiQQCob~lue-Cy`=)HIjP932M9LX+9NdoJk`ybs1)#(C>DBi2 zsQJl~QWZ#zo`8<*ZrkmuDq1S*qU7Z)5R);=2Daf)VKH(h5j|2%dGJ!F+zUu^8~Dzj zFTiV+F<}XtGcW}wZ^Ix3hUJ8NWeJ*MWCjx5Dp%I&5&*BxYJLUl6<=`Jgk^Vp$of=RRVzg%5Xc2drN{ZQWj5q9QFY|Y@ zLe3;NdV*VXX|1r1YCuM62$GKS`8Kr#1*No6a1G|y@uH@N-~umxRHYp(P*CbyRRtz5 z|L`1bA1_~LN1trnEgi$;xhGy06qcPo7+m(q#rvMfB;|&qG}MWF9wg?L2=e;p3QA_{ zdQ+yl;p20Cj`-`~f0~hxhxjK_9H7?u| zy*Jh)3t_bLSDOSqv|2CnLvq$HLzFY_Z|ltTZLxGLNZoAhNdVo_L|WkurgHQ2?31vX z-EWP08k6Dt&`^{qsvK;Rv5deUpa?D=EdM;ujGl_TfJx{xZuASNv>m9A1f}(s88wyV z&4vU8w;t`A=))u4XiIz2@BrJ_)O#aGhulJ$il`Dm9mctT?8U{@sKdA#VmHev8*Bo1 z{+r3WHvUMnZP+pJY8`A+&C{pA8q5LuX*m^Yr* z%6)FGSj7n-@nL<$H$Sdti=)S@M_J>n+E_}pK0oJpFixdr6}E27eG3Z5lBO;A)VsDN z+1NbBe;E6rJ3fRNZ|I9_hquKf*MH|Cn1{DtjsxN|#Aq`D>a{!+TQ&7>1u#hN&-OV$ zrT83VsP%B$dmhgSJH9DXv+cy5xXPuT*2 zKX5q@QfI^q{ZVZD;`&l|_>{D1m*StMjE$7{ZdrT)3_&1UuSKrOON=l@I1RZtGcsD6 z6lX}QjLN7Tl&{uo4utZ5NKn>QG#M)BZJk6fCIcEr8!_c1L@mcD?cW93xzGta_#d(y zN*=jYrl#M1^Qvq^@`qThLk24v!DcYwFOaIteOfzxs#S%=_$YmeL3SA7`sq6)pv zESU&2cgE9Xkq3h}g=#60Hlhm7&a7o!dovQOmDw(H`^uaj;(li|#7#?ZG~nx=B0&q? zedijTeoKJSLwIgnr$!CXy#A{Z9Ta{U(C4=PDD0$|8}#kfD!0|uACY1R742AclOqkc zf&Az;K1g?hX`8-n!`70vq&ZoXFO8cyoj@`kfY-6i(i{$RpAo4Cv4o+!x53=$jOGE>~`kD zA$Bi*!H_WMUwb(~*lA)!wi-5~UP%$yu2sUP-fB$vZ9j}1meY`We|%r|gnjz;dVZeE zNi%vmID?~sIt(H0!4u^$j=Hf1ep;swS-Hj@6gQm36os+FrMnOhnQ~%KBm4TuV~dZmJtyBVpCmeXLD=ai${;U(z1 z5SLs<`YP<p9ma<-6SFxb2QbFB38@j$-* z`}oOb$EV#xI{Tr`E&##HL=cz_?D~9P8mshX>ht12+=qvN)7ac-Q0@-Vp3cYy_ubF> z;DHnf3?5+5KiC(IY;HiGL}K5%hny5}V`R~F!YH^V;WITL>VWLeHn5k%q$J5PJCvbg zF<2QkYz!u;NXRyZ6h+6HuzEjA0C3_}h+(|nMrelsk`!Q*D-94y)Msgf$>AgJN$7^+IQW+IxFwPRUSK4AJzftjg*uDJqQ6gnTo6xV+0+|}wgs~2{`51Ri)js!$(R00?|@&)XmaCBc|2O13DhoU zz&J*=s5dXbSf9Bzvc=r5f5>0i?Ch;UK_sIrj?jeCTYrrpC&9Wz{VLdwpt9YBsBoQW z`o<+1o;i&VM$8Hs)jzJ|$g+)o+fo=jYBzG|9z>XTlik~P6=EE*rE&dktd)P)nwzKn zCS{DBL}eRh96~d0o-szR4~B*8?Vg){&FQ+-w)I5c4IaEx&aJL9H%N;cxc&Uonz;xT zRQTEzg(MkOn8L3law(~etq1nKD+h%GudqVbNVyl9&wPMOf z<)s&ICAd`Edsn5!dhUKQ)Zt1~XyS@I`l`)&K)CkGw!;LT4)*miVdFVd5{$}A36BZo zC+=DveOqu^3jo})eu6+SHic3}?!Yje7)ImS>`xFs@Cq-h5ucF{*b=mjqUmr!g@2Jg zW*G_L06-_v4`_C>;<8xs;fRL8e0Xw~D6@lOW?&b|ehvga3RW9XV3#)X42_)2`ac{w z@uibAmHeKSb6ws_NQFSST^wSaCRD*mceo@^jw|#FVnAs-Qcc19M_jwZRb?P$hEm$A2OU!lsOzx|c zVzZ2Z>pY-E{@1c^;27kwNQ*ko9ln{^rKXe@cE||C=tyLH< zRYwEv9n#zi^Zw-d0jl|81D%O9J?Ls5Wa#+V0D%=8Ma+kASYFhD1kLq5f>Ief$dfF#C7}fYAlKli8-3E-%9Db5RJpvk~g5^dM z8;n~1VBxpOZwQTDF}um4i^Q?WmHR~MXdc|zLx|bZ2908;7Ef&x3oycULzY_(!~>En zxVDM|5LXP2tYR&-nm;LlR_{d3>S-0z+SJafwDgUI?9LnV_c>@8*VSf$CsF;9;5MKa1PUBQb4Rku2I`U8b%fbqzIZ6lnBsgr)P50YC zQeK$bi~AOsz~gsd zoqrznj%6ciH%Ls}K=q5YRUS=I4l^aZwDJJ`G>Pd18TZEFaq*0%{ITrEU$Lap&2DUi z=G2n0r@<;CO_*5*aTfk6BcyyFCUbZHnGEtY=uA(PQlyZiIbMTDR-y|sZQs9=rHRdU`Z|xRh1u zA^|lS7wFFFRe!XwXZNAdR~6-C%w|nE0h#5g+p+ zRd^uJv{HH$&#F~_lPW%3Z(5Zo>%$u>Yvbk3owLy6<_Z6MK5TvYko7YI*~;DT2&Nh{ z)dV^2=Mb6;*ZpZGX*3g2Ea*Q;U9WXGukkztEqwA|rrHMv?CQN?GP`c^j$^beye@o} z0vW_9()pNxr}dn+dClr>qq?ZU{oV)NW2z`FupZ^DGX9kq$cQidS7Jd*Iw`tuc896Q7d z`hLT8y)p^x7Rbz3OR9L#UbM9GWS_XzI?beFiCyMJ{O#g%b1D(~CNmf7A1pD-6r|XX zNtCE*$v|-A6$tRy{u+o#;0pPIbe&STI9X~p)>XEo@dWUZXBv1WXHAJ>vIU%t@~=}< zWmA_Uz_xRgwRTB|P)fyQd7d9x6z^o9(F08LHaStO!-0odq+fX{KNd8}<~5)DT9V$pIEor>A7J<-|v8;qZbd(=#55 zb$v`PBHH=T!K~OE#Sq`kJC>sdwnW`Dm6a6UZ`8MSVKD!(Q+c;io=w^uK{_9N_;4ZO zJttdwZil@MsG}#%91V3SFacz%l3~w$5rb9I;d8Im?Zm5SuROA(e-XE27bIuZ$~SI3 z)wT)aa;%FkeLrd1c06wq1#3dhYF<=>wUy`IK#r~&Zs!M={v>Vk&{svq%EU@4GrWoY z$}dh*dN_{aN@c2zFB3Kp3Ya9WFKwP}YId--{Vm`PHftu_+u7~`N_w6{`Ez$GT|XiC z%fWb;Q|&mf`iqTId9Er5cwJ;nyhdHkkYe_&+7Ws{O+^&p(4HF&zNFzZ)cgy@#2N ziLJ8}owbR(iK7#po}PuRg|nXCKW%WVX%qWT8~k&a_TTu_Wj&T>;?(@nwvN=!XaYZ? z6=?X2rcQ0VUa~SYQ9!A1s%xqtgn}g?jn@vtO*PC3`%b>ME z&)TFX6Pmc#{p0anPp{h}lasMnb8=!VLVB^@eI~yf>2hwJNvsG(IF1TTw+_xy&-LT`YaI-`>+o z1(S@A%E`!g;U&bja#~xvc_Gc3cvat+(c|S}bX9J4IQ^}UF}r7Hv%03oM?$zX(*D-= zG>GC&TxmqLWBM6r_@jnRcfhm>29;{Zo#bt-GBXJnI zz-S;Ol$_w3k1)HV=Yftqs2L9Y9xoK2!y z$yBv;+Pj8WmBW^oAjRwmC0jo{-X>YbdC1abKZ$)W z5=?d~$tfCC<#QD>bk4j6li7y`no@3LuJiJ@AH%aMTemZKUfMp&v}7Nddgg~jxA1Cl zKODVyRwLoVqqz!Q<(esN9a*sf5_?TI%2-)=TWXM#nUm{I(?g%3<5Z1tq85EQw83Ua zQZ_-~yDW8TIG3?5WiZ9z7r|O|dSv#7LmWfnae{dqe=U^{Q`9sc%0c|;QM`Ou_ns(8 z6SVrw=EBU$z$j5`4V!+j%}|c1o>d6K(9(mbA%0=E%a{!15S8LVG^4(X+Ockxrix}< zkSYwF`sGAfv^71Kc)n6xgQM7}YfVPaq!hb5iFEn`>IuzaL*A&dvKTwow6pUY;O34B3jZCesWeHkFWBMyTUxO=sY|>dF7c?14<^2;_^`JMrEtZ^E zApu@8;N%cBa;5Ct!Dh9wkVe3e67=Khwz-y+TIxdyYP8{6p!CPbvlO;7qbt=eBl4J; zLOAUvFh)MY?&UQCKLj*%r*j>s*sz+kkRi(nn1;06;k^_?8DxKFBSXAqn zD(YKVskeUUYts7R_gq=TR{d*m-okO&$^648_p^djPV~Xsa;gom=Swf zzufjkJV&c}r=i60jUbG{iT5%5_7#wEcE4gfCGT>9a&}mhO7B_Z^fPGfQndK_F z!nXfWvgUC`Tx@K7x}E3h2EY2B4#>`EseoV*=LU+nwHR9P6G|}*EfF*oANF^45o|_x zDO4-o!rbj3s1S}0dDwG?&po?-r7Cr6ucUd(UjfNQ#Ohv1<0O9J)N>*nmaE(TM`?cO zxiNY+{x+ul82ke*x7rY4?Y2=>!n7dI;7neSgJMDaC=+ODDeaDSA2YS^=bcti*>Jsm zXaSRB^upHV>R@{cVQ@q%dh3hTEL3+{)cb7A=cx)sHaz(a0q)yj!wElMozhE8OV4?U zw(ya+TKoc4a`78LzM|~n-j|P7;h@42r8o}2#Adj%bv84U2x%(c;zmmSjyD0zf`W5b z(7SKLURut*&RLi(=QYc(ZxvW0Zt7+lZ9I2?#3Kl~kjr{@lBQW`i%N>|LR4Y)HtGZ4 zXXz`l$!Ct_{HeR7hPX}$LCJ+e2=Bl*#){{yJ)eRZ9fdJv5{GjOJ6+>_hiu8dZsV1> zM)U1oCLf=N#YIp+K?$U6|JOwb3tl0}#j4E&`ECS}1gXQs>|j7cNdUKYuDZDJl?IRG z6oUhRURdRGdo3Wv4P4bn=>~y4s?*=#-UrT5*BtI8nq-f2CmW8RqlH0cpNKAX*QS*_jnsu^{Ei*4m zWQ*b!GQ4nF-+G?j%%%sgVD+4RcUD=&93nXIia{Zj;u9C3P8M8)jyMXxHrxA-@le^pj;8lV+;(zy!pN#gq;W)Km~yHI zaDl-A-lBY{XbBj|Uc;c6!EO1~1|n$|EPDvH{~A!-J~Iz!I~xtaOou`vkgo%r?Lg$*QB@3bta_j{qLRDF}@& z0F;-Z(a+amq#d4~Zk8?{N~;^m-@>zx9qxIT3iyeg8p@lC-7f-%Cx@(st%kA`JzbFA z_pFaqgT5_S9I!d;*DPLQazkMjF~&kJ_Cc$|sLy=N1*#EL!>n(-FaT4B_nk4h%I(fY zAAoKLsi{mn^AU`zwDXV|_W`xnS5-Vi6)B=+7S2I9>663`%3fu$;TRdqK1DmldyiP5 zBx+9|Cp)ho(}&B-x&!npD{le0!^8~aeHav3JC0Fw7g8V@bp{>07g9#hRhOyy_VG{M zx%l}IbmuX_D7Fb+FQTe}#&r{6r3 zuNjvqEw}Zn=C**T`J>KQ+`WE~ek4Q=BwaiJ+$@Fy@;3ERV+8j9+*n^=4gjDn923V~ zUiYx!)LhmVfiy$6AGI2Md*6$zh}#jQZe-TJG*sZ)1D)gSjNBwR7-mVC0u}XRE~}yT z&=E3Z&7c)jKFzgMUUxk=DIH*#2B~bHJ*0?8<*OiIQNk_fQ?*{t9z#>N03U^|S^uJ| zUT}-vjS_DUbxHT7PPo9HA+m7p@Ug_-Kc(zG@u@wad&mP}Yi6c9gF}LtYXI2+F$2nk-xPu&r0a3v?0^bmO_ z;WV9Ogmw}kp|@eP?Yb?vKThpL-pLCr$E_Doi2YDtsx?Yo2@d3rgU%hb8b8&L1Tn-4 z;6s`J@e~s8mkKNmEg|yr9ma) zKqa1SXBYOC1&&1j-UdeoR6db+H4;nvv54Z}Q>kYZN-DfC^SI=e;6k>#nVswLoOC4n zMPZ0eVI_oxq9wf*^xoux)|yfPDb}s&fF)`N(~g*w1~~t9|8hSh74{K^eFdc)QG-+|b9+T`(?2i5V}vPs`!4R~kvHs51AD>EC9tvU75W>2!yrs$S)lYly?lQ8^)n{8jy}^$a9*0krZ_G{g*F7y?pd(4fHqQ8|-m z2m$~oDZ)7>b%JGwousV*cP!Ha%MK0~)5E*T#INGyFjnNljBni0Q?iOGQV8#l$_k2L zou*vSy}aY<=0zLI%u#2Ek%SYm>qlNOfHWRVQpz4PT(XEz7xFKv)VA2FQq-cd_Nj-wI&b}uLGZKGPU%L4L4D70 zyop!8m}KRRcYe(m^IVHVUNO3P+56U~@5!D)Tbx9&~{eio>8y!>~zT+X(gk-a%t z_zC2-zCZ7Lzb_Z3YQK9xjy$%dqI}^iOMb--O3OcOzt<=YERSn<;Z@0qVKhdzp|KL$4?&zbbi?hI)LMgHIH%}aDru}7Yg`3 zHU)`OFD2o((rO#PPOc}pr%$6?2|>+AtjM*dBhGPIe9oq=HRDp8@f<-K^gVa?6V6HU zSBMuYC)OqzqnPf$>ewD9#LGH`Ub|59Y{e+OjiQVfUxhZz9Z#FU4z@Zhr8;WLs>BsL zpJW@hc9fLYh1pbIH$fzyGyRD(f9=1-&|Qr_D0`l9pN&Kn^BsLRHMsna2P9ZMuxEt< z$Tl2Bvpoi!k2BhH>@J$S-byGQedUry=1zmriV1B6s69WcD=DRo5O-B~X7%QxA6im# zyP13-WP?feW0(g6&7+xjI%kq%LV5QHQA9a^*tD)`q8W*W)jI#oN-+C1n*L0Do|MGC z*0{qA{CWE%bg)=yLE)b z8cN_Q-4>)q0)6{9@s-^ub+X*+!B!of`HUES_?ij|?Re<`3HCW>rBA8pu6P2`gZ?q( zs*SHXXhxUlZO>F%%KhLmWxvHV)A3|N|1f1QvXv3pXuJ2PiROIn)SBmsB&0`vbHx!R zvh_jdIy@2Fi9gD)Y-TFBV?y;a%c$mV)b8olM=1kQR=D+MCOUSjQiRuku5CY8WmiUe zE1EM4>6seCBE722Mz+wC$gDC)Lg*0@X)a)Rch|11X-m&@tjyQ_^eEbnjA(1k-7>x5 zSMj0Ic&(wlzK`AJYGA$Lo|65Cwsh-W?*0-BsY*lceH=kpog`7YR52JmC|qxoDrmAj~>OPJVE!vClOG-Wuw zJ5DTGg02wMlkVgoS8A-=_TLJzYOz8-z2!r(!7m>*)gBOj)Gv24+p{DNI?XpU+0JOo zOmBJg+c5T>Q4X_~FKaFE5w-4Rx<|@qhaW`1^r$Jr?5Ge2B=HEod{z%0w2bk zn6oYqhe8hotQa`0b}JB;7W&58cPM!_yw1Qrbr$8Tyud8fD0f(79Gt#S%A zJ%dV=c6$wsg*MSgQi#wHG+%tr$ciy{ay*RW1yEZz{Z|iyTCS9lL(TD3ZW~$y@y(eq zO|XLJ7zuxvXK}JdY2rzGWCz;<2=_*h(04h?s0;Q_S)K{0^ltjcW$*7JYWZ-Wc zfP@2P{X}oQ)pA=!M49ne~4~4yYm3k&iXtEDrkT_}%^~Nx5 z58AcfN*aOZ5sGaxOJ7*Ek)Bq{{s;q{zHsMX73HrA#sIh&vbzbFRP5p4Lbp&a%H?Jg zF>3R7Wr#C+^TIT2OKrxEYw;%3NIOgjlr9knkElID>ICa3O!2`m;T`zF3KJ*FxicDuJ!te^smwEjhj!Q zpE69I`RWnVhWeXig>u>m+tGmvbRddb)0GOCrK#5BqupHpPv9PEpocK3)x6B`mL=#A zUegEeJkL90-dp-rS5WS4|2vJV^+)eQ9RXiQ5fZdj4d3d^yKsv;eHaij4W#-Hh<|I~tkE?n35Ptg8n?UY#K5{9aT6#ya zM_f)AoM3aGMH6}?l$AVuYv*GUBsNsqa!BId-@pnSB*gnwVQk7$4zw8JUED73~6#d4mu&cKvN9 zOpoZ+yXEqRpa^2vZjeiyDCqf^c~F6U-WwVgm^njd{{^kNYvi~FOrakE-`tGYe0b@Y z1t}{It1^EqsHrPigiKtTcL&I4@7Ze1&V2z>xtP;pdKvNcvXu^>L}0@sS{mhY{ygb=-Z|^Ql+5faz~1G&wLSl#}f*Auxgv>K{;B8==tnsiWeU{gf0kqDlp+ME&j>|7aXLYXUe$ z&c^3xY1LpWx<*h^O6)O?rxDaZty@{)5l1RCNUiQ~M0_Wp3{ynuj|pH(#20}~YjdI^iU zV+!4Ym?vp9EweNvY@yd*08v5C-;n?|aqFjHI>Ud;V&|A?kXOM3wGYwzY1Y-1+p!^$ z>|A@SiXJxX%t~<=+3Rfyz9uY`IioDl4$`O$y$PwJGR)}G4v!p)*mzxqZc_6TKZWUWR@ch8pRZkqW)DWaJ5dgPJPzhgST%z0&OQ z-P31R;c;yCe7hOQ(s3Ff#i6B~uc(k&@T8yhQ zdX}0EVs!x-Oj;0(iMfd0I|7H1kDjXJTKmH=o0*oux*SaSZWucaN>e%97_YC#{_M>O zRYN%7x~`dZnxuK79}zPrbg-oRLm!M)#pkHhrAY*f<3+EBepR)jPbr`4R{Ju1Xol_r;(OGy97zPAbei{pIv@~wHZ}V5zUYuBKyjRS z=xFys1(ibTKd~S-YwBxLE0YTT0?~8`b6Y)W_6rby--!^rmyxlhkiCV$g&~w6X%s_@ zL&on*_6g~drct0qFJO$JTn^P%ESZKO8jzorse($)qS9ik#h~OFLrKbvT4A7;h_uBi zlKno(?pN)GLB`6O1cibflB*fEiy%M1wR7n3#ZA+AhB2f^^Tng;cAJVU!utk&nd0+}e z>#G-um|)YQ3EAO`802pRZ{}0(VvVKrpTV#Z8$p0BgEbEMo+89L9-IzVePF6C@Cr{( z9@exG(LXAYff_3p*+AhlVbDVITdgqe8@;1kjJs8-uS>Uu&l6PZk%?;kX7_nG>qgorRl0Ga1t*`KC7onCd8nmr-KlR3q~9=!+y9*0v-gZB>MVrQhx7z&U!Mzi?Jpci=woda*{iK5EKy# zq;+|aD@hC_6H@ePyFH=%e9qP1e48c<10zlZ{TNrU|#a*Ha^con8l@zRmN?@jm7v8g;s0_h9 ze>Wo$=G3rL%aIOxh`1s;6ZFkf-6?m%Bh`^ykU%KJbvuS!O4)R~n2D{BjA9swR~z}% z#E3_c?FHk%o`<*^&a;{H_)z(j(qx^Nve-eBc9w0FQF=niRVJxA?%A=nH$)bKRQUdW?JsO_t-^YzT@(l!jG@t#WDXRp~JKviz4eU@X~Z0*2c z+x|AJHeXF{uNfvHl=dXmn`rNgDPhOZyAQS&4H3Suw6CIFCC**LjVV z0JgA((d;c@lZ`ze9Y~M2l#>fWfRtQElPt{8j1a*ulU6&!_?Hl2fab@rcEQ7m;qOC5 zi^MsNBINcT=URlC#Y6%}eoJC3?zq`M4$+AD?;u`VK$j&8&V!|(j~YyX1gxz)jmT9# zL(En7Al<*jQ3{%WP^9v_MOri!9BDZ>XUzP6{V@TCvgHkOskrU6e$pLTyi~E>yeLb zT%~996{i2fc}xhgbGY{Rh;eCym;VC;ZJ!B)jb?SdS2u7EllGztZ?x-oP7y+8sL5Sct9|>8^R4`u*i3^I%;i%x41B@|IiUILEx1F)(}QC3@URN?SUOc9+i|3 zc*#Oqt{N5UH;scKA+9%?-%UIsNT~pyd90; zp*sOI5H9YI7`-$*Ywo>8mv|+z(j{KK0w_B+fX8fJ;qy5rM0sUit0WPcbmbQ7tJTNB z9tCWpPq5XI#!vzbhUi5oqa^zgA@IrpAE&u}!>Jb<+iy&^y3k1fa>Et8yQo2Rz> z>@CX}z`iwt(5PJugCxXZ2bXL47fZNP`*zem9Tn=k4Tbkp*0n)|Evii`;Ry(WNbl** zPqV8lZk!kPq=gTF;ykZ`F?W6dDm_5GOQpA=|e|RcWwhsKGyPC;or_8(7AbpI00j5ITc3diyhC!7@{LH`py)JOr8MELNL-FhkryS!>(iIrklw5n3V%+pxsuT8 zdL$3ncH7Z;ALSvD>mPS#!)=6E7NK5z25U&gsgXC^b#GtyD1Ut)e?IoVn*-VWruwUQ zy;^itT$b27yFA<8+&y2>?v~Hr?d)s=)W0upeD76>cDwnyd%z!Jmoqy*GbhvMSBy*?z(uj1Z@?Jq(>- z`4o7Jn|mF7hv?#;D>M#5bT^L{1X;xbzm`za z5l*F;x`CRfs8b2L#OJ(OAFthv#bnc%vgcGNpK#EQk z(0Fq5rI(cXr`WuEOXvjNPdFnTng))zg$Yb18X7JZHBW2DLRUm4U6A!E8iT@aCRz+v z&swzGE#Y5}*>&lKCW$je(LQV$imR~FSz^lihm9LAHd}(+e*md)i|j7?AGQZ;ykw15 zyk7}}r4~ha5wbc?r5;VG%o*UGzq#>uwyJk_r#fzdM?&80QFAa?f!Z)8o#gybVBZx! zeh+6Rz%YxMkkyxe6fo41y+`L07v^$#slL+*S!!d&pXbES6`p_FXGuFthWYLjQO-Z){l?z&tzeK@1Tc6sbCB;jF>s|>oT&CWF{FY#9$CBQ95 z?=${l!eZ;O3w$!W?+>*a&?&36s1M0KU_&&#KAOZD#iD=4iE4_VLAT2fq*?+q8=3Qm zwoztxe4yp3bqO3{p>vVhcDw-2F5q`R81PQ#s)WwA&FXWn)UdsmKc20}ju!4lD!mX_ zzwF`5hR0%`J*d@Fkd2@s9z;A6K3)k%gzFA1ywANi4JB+}&6l!(Pyn!dld2$rvKXL< zkUik<8Pp&KG3G!#3s!fNm2hizNb}Rzk!a;93*ed}h$NAO{*m}=^l+{mNG8yly3$f{ z609NuSs6hVg^COLEU2q6X#@thTxkT3%)QF>PUf0ypz8b4#@pDAfd}`wST_i(wc81- z-;Vh+b__6MjXm3me92dn0t z3SHAd-vS%AH37yX2&b))0rLmmho~)>m?`rfhA+uuYo(_oXMFZekyz~65Z$nF<`pU6 zQ}N9%(MIow|kALCn+6#^N&P#*u@TQQlt1*A6f(O%i<1zZP8Z>CHm zCvND`e~Hsxoql_mH9HA*UBYD5nqd5t`q zjd|4s4BXACv@YdQ`Oi&aLE_A}95aX?m;vp0K=-6QD^^RX$9+Ss83wKjY3aj0JMDJ3 zRu*Gde5kpe-W5Lmh|!JgBOfzsmv5vr8PI+{r@FF*oQZ3cxmJaPTp3G%tIn-+hV2(w zH)%U({dD~t?ZfGKty4H-!0h;~t)!(PWAQa`X;jR-gOTZ?p!hyvZ`;LNxaQuB8?Jw9 z+2&T2Z+P(26I@)voLx34>QgUsdS#}aEJj(;6spD$b0w_gn~r$Xc-WNET}(1gQjH~T zUIrbP?$;0}!rL57oWc;JHo~qcW^MBls6uCam_&h54D*b^(*Q0T$(hSPk(QlEf={7| zAwwKpod98?FxJdOBV4sP!QZJ?KnL$7y^P1l&Q)aC>Zqgr>c(+@=f(kH@V1LSCIE^B zflk8UzBQn1*VH_s{DNIaoEHtBScS*<4hWajHXnmJ$zh-@hWocvK{+sj2)u0>Gluex zM{N%LuMmKH`*>5Ntj+3QDP&-YP8yF?{9E&39q9nk+^!VEK}lwhThA5$Lp0a26LEb= z4?9m@Ye8yUAZ{7{l2o+mp`u6zdf-hw+zJo*$K5NJ+1_BTo#)1;Y>gh|M*ZjBb;Yb> z5#*Gj@$!GA+B>XOyl#69DLDAXbOr0+o6oV&9Qao4M+kdR(z;|8IQGmJ#FVXr%T4Q@ z+`)3sBO4AxF%cVC|L>b(2)0h_R?Gc%ui?Ljia1q^s{#>35BG#bWt)wls zPJCP6q8Bx~++KH|M?*IVhtNcgV+J@3z9{Qc9;uSmY#L-;_MpLgURf*+HgccCATKp8mzGO}yl zA=1PRp8+xOFU+}9$bcR@(bgWd3E;HymxEvKd@y!@n42ZIP5t`4wO z_Y1Pwi~ED#mYFw8qIhZd7PvZ3xQHj~GCV@S44Fo0k0|M&`O|yIKC;f>o*%~`i!VBd zhA{Pzjgi+&sR3$SMKpkIKWd&_cIFR>-05+G zrTmBR6T$bUP)`bVUJpg_QW%d;x9GmiG}q`s+sON*!KDiqygQI*LMZl(hvY&t$bX^) zr=n0*bKGjQfbdJA_~`#4fblqg)@5bX$i#FWCzJi@cuDb}dzmw`SO98@$L~1&bivf* zwEcWMID*_v@4_J<85?k;`4S2E_!Gq`bOXIaMFPO*1*}2K*}^7Y8wciICXQapX)$DnV&U&6MJ_FATdWn=+Wa)4tkQ5_ z;tapLJw-Cwha?jkhyXj*1n8)A6-PDGbcBb8W!#!>W15yM@x098#e)GXV6NOeoxz3bu(vsdE80()pyv>kab0kB?M*haO zI3MIFHDjPrmO?joRyZMZnjYC7xG(lP#mG{nwV-+GN-6cl+BTs` zsp7~SWYO<@!K!KfgDuzA>-bS=gQq%+A}-Ir6WHn<`Q(BZ1-qFZ2s^>w;Pi}Z!}tcl zI4^* z9SYDCSc;JZb$uB&*>1n{4k0O;56;DW_JAV#!%KYNZbj^RT-~7(Ak?|T(AX{+c{)QY z7?jH^yii7o8-dP_%(7Fg`O;etF`gQ@8H@p>ZQIp#&$k|o)B$H9P*bm6l|*1F|>JuM~4Xc;1?XY zbvfk&5J=~nX7St&32cuZ`%_5EMg@AXD2rcE8Tqk2&MoO7KaxhGfv9f4i2rJX;Y>Cw zITIz#XlG>H;*Q|#Y^|u!i|$~`Y)9sOgEii~An&+B$+1S3QJ`tPXoE(5Y@ml*&`NH< z9*@qFHj1I9Hf9#&=J@oH)`x`%=Oj_p(5a~!`aM$mJ#RpU{Q=ci`Dk#2=EjTHho+Nb zZsu~OJ)b!~h2m7@=a&MYCS~=P3?p#$v`aMu(qOPNz4ivU8Nv}fmqD!scv5iw2b@Hr z5l4P->`PDyof&`1Z4lV$k?Xa5=M+IIxQo( z_q_QCvHzurja5b$$=^ai*PeEW)GN>*ruVLkN~GPcauqiO<6vKi%Bh!WKXqyAp*X83YU?$vI*NH6>A=N zw9}a(b^j;u7;zE^W(A=qcuMFn9S$za=|qxy=!P|Avq*Ys;`7(FvtQ6e7Y@{7Q#(SF zv`VI|`+?Fb8RVrw$b`j85T~06epwf{LHUr>6nj!+8JAN>NwaAs4|Q7extbZwX>I;Q zaJ6U;LSI56@YIL~&&oxp9XuqlcF+K*(iPGuszPz!-Rco$;=CP{PA#l2>Z209NaLD7 z@$S!>f^j}V29;w+nW6%joDs7ivc#N><{CpaEQd}! zDNI0JaW13J6aWXqI3CiA++xg2-k2qJYWFK=iRM=^?-%0QO_y6z)7_ zY{`xlh%6fIT-35i+}1_nwlG44Bg#z(%n^j3#M2e$FvE@14-k?4RRq{KIjOD0-VJ{A z*$c6!9-AO;wLuutXd@WOv8an^@$XN-1Gc+V;;@G`;F>XPM87FTzGNtXar~p}S7Gsq zXeYXnA7WPl40mQS8QbLim`Dd&5#0bH)HVc&IS!?Elqfk8jGE*|bXdM4@Z;*EnN8wX zP$(mU3}&X3C)MTv82Chy!@*3nN+$)%nBUhbgj^EW(Goc7Cdf;4Fo;tEaU;{6vdN=y zlr7jAIvMc$`GQMZSjK&#`;|0vCWAw^v=4L=>nYzkwpFC%3xv+#C^pd8{gbtRBFP?LJ=4P}F1JC2xcO=kQ>r<{h5#rx!L>;cSXq^&=4 zl~szY$v^NC46aze?0Yw*1c@kSPPz7rgxY^4l26I_)=1m-fe4-|2PqsD@ajq+rG!l+ zl>)D=?8odn*a$9->xXgV&V~PE(i>UE`!@C_51wG0GjkzlqArZfU}{3wqImB@e&7sM zN#juqzr}_cgne5YsNzSk0lzn(A?M1UWmLI@O0?b}ld4i5(`;!@6TJuc^61BkmP$p4 zX;rPy(5|pF^*+nOcEM{b+t+T;x=2_Y+U$Z-q8=g2Kd6&ieB1(CKehI7FC5$}{w)&o z*?<>)k?v4sGOq`*8RRBbaXph%MIbtxmW{(IRBH^5Mu3&a@Ey(rd`oKCLW1AE2*RJviC+@<9bDXmJ42Xt|A}H9OQxu!N_gehpL1plonKXFXC%;6es$>7J^Jm@qnn>!Jb(G@)%(A{ zdHWMynWNtOA#`|?rxW+2cd)OKikYy8-~l_XACc>1wa5|EJH^`{+P$N2l$Y)mSOSeF67RQ)Ri`5a9xia83ihKWdL)$AJO%TZT zd247+QjjB>4Dh+yz?KF1Ld<`lrw!JO9EPAeBEt<;ph@a(ojI8s>5`S%kgNQ8-6Vnq zCcQ{G<<;gg2@Z|Ghet06t3(cbmNhMO>q7D-Z17c8G2OC40vSK_w5tTiQJ!B4pvnSk zF|$2DvrJxWKz>dR12Kq_Ii{xgR#rxE>k*Q!!Pby7B?s0x^gQ_kLkpRGspoTI_SZt0VyQol>7!4VJSun8zpqJx3;j!lydzdIZ-o|{G&}8 z1zRvNDyf4)sa+|elLJ**Gi;EAWqC6G(4WeAhd$_Ea#NkhMS_R|kPK70S_ z<;}bI|9tW9f4zD8qtVxo^WE$FCV;+=`Mv`3d)%&Am3S=H;IW(Td<>IZnR^U9@fl{x zYcV@XhCIGM(Un@5!H+MVzk2renO<9G!9H8>I%=jvP&?GV_z9t2pe` zQgR+Am)9UXC>_}s`>oucfA{a7Us!V`_mlGlVDioP*&h>!1(C8RMLs@9+*# z&iiMV50@SMlWnU1FH zXA?1mgZv$@l)BS$u1D6KCF%p$#LGw8NU&G|3BIg|DO)SwplNPyDhU?>xK_Nc&prN| zPyXkM>V|lIxm#Q8rQxf^f4RzA74rEA@ufcQS+2g>Amhqa$jkb_DPQRwl`ENoYo4r7 zwOCly2F@6aWAK2mk;8LQp^@q!8a= z007mR0RSQZ003}#XKiI}baO9kWq4(Bb1z?GZ)0y^Ut)D>Y;3FitCiKOf>hhv&zyHl|p>cjQz8l|-K2MjM z$@+RUo@5t)z4))+jYgxcYl^47gR=eZA1{74x?HNSD`>g8dMMP_ab0yyU6!iR#FeU{ zPFKbDi)xnnYw?m}#F%R+{(beOs=ro)4CB#uTlZBb0^>&Qpc;z26mSMLs|2!vjzPfH zWnMwu;|FwjTo%ftdNsv%Z@U5$9PLurs!=VlQ`r)koQ&F5wJkOcP@apIzR6wvF=7V) zTX>Fe$9}HWk$r^`TIgDx`I13ynz}!7U|u{md2=FYHQIK?E2Xl^Pi39!BCSW&c3(H< z_wfida2(e=#o^0(clzoE+qCO5z9e7gs>yI0ke`N`Ef5Xus| zy^eR4caA?w?n)iIz2~?p8ns1gUe0l7OrU8Fd8Zq9J*i%crmnDxS55sSigXR*ISk(@ z7|xxkE<(-}>>4K#{S)Hxq#$@E9;0T^rhV&R`u3z$2ipT3nALr=H6Obzx{Yd~TeN*? z*rtd4MHGgBGkHkV)gtdYr~oq&8BDc%o-BrC9o;?s2h>03(MYz1ah3k|T~pZ3d?tsh zyy=QH>ZKfW~i&V7ATXHe+s#fHBvvBYYVA$Q#rzw)Z*5S4Ixb=h~_IeM^)CV5~weTD9k z`mo`U#o-KRvwe3|)w#&Fw$5ZAU7>=*M0VXsP>16UQ!?oJrNYj1SDmdt&UmH_oK1Yy zCqAYByOWo0hKRV#L>y5ah@lkor8=DF%2e|TBBuWAb72{Lc#&5yF&fymP*_tjZp)$R zkLpQ|2tE@IQbqj@F#&$r)L&cG7=UMZV<$Hb-jDiSQQyNX8U!!Rk}i-@)8wZBCh0MZ z{qW;uQ|!zd!eCwf!PI&W%Opr4Rtoe%F9>>U2|`@}Z5m~uMDsq3US1jng1aJF@(RRp z6dE^q`^C_mq^f_8LAp~cY-3<2glf~3N5WVQ`;pUooK1ou__&4d*K^(#G;#8FteY;z zX^EDS&V!M(^PP$TYJ)}bcGq{uzKh6b;RWohb|9^ecKI>&u`EB(s$(UEbXwH?257WO zkL$zHNI40%mZZmd9hJ@#^ui4@K{u_QX~_=;PBIG0;qNkil%Y=LZ4J%zH80~dGDSOH zSka-4hj3lHtVA&*Rc(|Cp<&pMDF&Kk82s22ov3FGOe@(e3=I=zq~ZRQth@W zjzJ+gQ}L$C%l;X9T9DWY#Ay&j>b@OMX>90yW!!Y0G3<1b(To=vGC4I9Ps@m6gzD|$ zI4^PIDt3cFI(B`dD9(6~`}Q&>AUhudoQ6j+Lo_4yC&^unQ%MWIFG)=96iie_SyE+X zs_UStS`p`?d@$xJy?+(3FGObQy!h32Da&H_!HFDSDO`&a&{ffHYuJ0Af+}Y$Ox~9S z)_-H1?N2Rrpt9p+bjNVx&!Dgb^m+NM^=^kt{1r<{WN85c&QNw2u)`~$Po^p?!M@tp zT^%DP`^hX5!%$Mm7&teUuWC2Uk!@jJui{EbmBQC(A``6p?V^Dg67dix(U}FCIK4@6 zxOSpF^I29syXz?}!PvqM9wn@{=Mu1`(+oB+eu!o!gn52)BhSxw5x^FV5Y;1!&Cg~R z>X(f$hFDX~vOH5KG0AecJ=RsLLR>U9lvJ(r>cx!%w&nzl0+R&f?=+^RqcGq#h-H{3 zwPQkG04|Ek2^ay1Rgb68xJ!cRl!zo#IzXXJ!8Eo|yCTQ37)g?$nG(HIVyrqO;qJ(z z=(@ubPj;r`@NwP^k*?~OArc-3rErlY4c+#Kyc$Z!omdHkoe>NCAuoGz+To}RI-^x6 zUFJ`v!lU5S=Tcaf)9JY&`9==gZ_#)7fsbz$Yy=@+1YlCkFg(!!j`28q=SGfUNp>$w ziW zS`}Np+?yZaZgW-osaF5?mFB~v?H1;>dHpg-n7q2V(^Dr+Q#XlN7EeZ+tC!+QiUpqN zW*W2X)lt^nwD2B7`U7c{br26y`^Eh8utMi`eUymox^m9f9I1l=l-8QO=q9lAZo8JQ z4m9PLUiFH%ju*?#nS!~q zE4F5F)d#9R<6N*{DR0zUmvtxDqW7iY92$`KYj$<_kRF#II<8@VQ5AY~x~c0j%@z&C zj}6#(mxfit&-n&eZVNLQK8<_Wm>oaK-JJn#vb3eQ8rk0{vXXtQCR z3BpF(LFu6BA0og`DFNGq%(2DUvxP}UV^-#`4bS{Xs4#YNF}@T@SZO4Yl9fe*1PyEs z-!-;UHa*{4!?5_Qh<$%QrJKU>?JewuitW{@$`8d>95Y+mq3f{P*hW z`sW}1IhFlY*}qIpmHx6$acESOV+|$F@vsCihsurY6}*byQA`VBtWvO@_2@AnM0@5iOjGypK`r*_Mv6nxocKQ z2@o!>Nd%8CWR!4+S`fILkvupB%KOtza)dXG~Co*R29%sE}&~fSYd?x}r#@p~ z?2GKD?Y9hKJ=LJmx7bv1iV@Uc>W{9e=9enpnwaLd<%Gs+aT900WT>3J(QILP zX*hOW)~2cEH>$?9)r4*p-R0@P40jYIuRaL{14R;b?Ij2s1;5QX?f}VVIXmQKxrY4s zHS>cJHg>810Do4{_2RdDLPz~x+7I)8q*f*O zhm-zTeTMuWPyQn{8L+>TIPA@i`(k^kepRJygO5eA+#ly0mvyUlH+lc^ZWRs3{I<+} z<9ASZ>+i{*i$~|I-SRj(SteG|UgWMm?7LT&QW1$c8mgUEiCh6j1 z@&b?X5s4w>|Ayp0CN;z{zU31i>G@HFcRq066=l)oO_5jB4%cP=L3aAYH?n`E5mIqJ zUI<l7K1`eQ zeOpFvzwDOL2kjLrK$k@YA7cPN_?%}}slqgF<}djn03~=qRU55HiOQt;F@M6YVL$jw zpVPLJrJuouTKVup(P~xd@d6L?O4ZbK3~V)104}@2A0$(->vxJVbQ;kit)I52U9Byy zPVE=ygLn&gI>7ex;Ggh}7uV2z^#LyM`;%;m^ZH)goSNEy9CyISeO;L^3eI``;tIN0 zlEgHEB&&=>0P1cGk;9$(f!y&%gwoLBhRO1+aU&2&DUrbb$G>hSgC5Z4fH4pIH( zmL#vCT3?ke`8720*_>!*Tl{y~?~l9u-3ri5Q93%32qa2k2&7|YyNX4ln=K zr2zHHVWtcV`kBun4`UF?GSrqR*^4Vr9{y^J2XNVE)Bhq1_H=VojRoPtdz|bl<5Fr}D&V*I+D3gRE*uxfCg`aE4p<*jWsa;Neiv zG+F+tM9H#xt6v%66MlkLTDu_IE7gfJztsq;lIb28K=NGXjn_tiAp6Y=Z3APJeq$c@ zuDDGQO-&~#GAmbkyML-ta%Y?l40^}vrYQs?hGlwI@7aX>vj(@ea1f?ll=)JmAX#hP zAj}a?(Dco9{ne*Ks0d|>H3_SL>vxpfwFqW+4FQ%06ca zE+az5O_PI!lH~Lj3#f;MAXBXcxc-U?PpG-D+ihwiwDDYfjxXo3uiMw zNagFs#@vkAb%=%C?bl`liyw;Qr={x9+>4`pVvES9QPqA`4VHkTu56!DlwhhPOs^Rz zh$WuX1}t1QmlUxXB+}he>#xKJ(GrsmPpVy*LESy{ZyeHBo5gj=BR^w9aj*UajZK0< zm;m6ns*Ju3FbEGS#=E@fLNHL?x*DU>8psN`q&TIdtFeF2G#P0%-&%KN7GxEQDt4^q zF-VuIFIK*pjMxM|89f%)#eKCkMc11h_K88Vc`r;{vk~;ygZ>GnRCr(~LogrKj@yEE zs7+>0@-;svyCNY&8lsa%O*qBq_ve#YaIHMS%j~|P8|$n5boczI)R&L|{VIg44lH5^ z_e5hd^RK_)H>sdvJ@EsJ$fL$+Z*Vq^dM_nzc^K6%>qe+L_JKWu3rwPa=p5r z>YVqG@H#0rEIdw|}_Jw-&P09h-#zPgF$5(<9JVzF^ME%c3FFt+zsqNR8 zz)vP{nYZk10ff~aMDp5(in6S~F6*LIyg)%7U68s06Qq!&@VtOpDNd$nwxxol7`Az( z#n8qO&(OF=>v2^V$Hv`E<%;Nn`jsLB+XAF{Q7qEa$x~^bN}7Y)h;pU2Br;ip4<51h z9&Tn6r$BTwyGDyl)?A9&c%>Q%yN?#~hNx^^6#%ZkAi$-g^}v!>SFkJN_T_+5Yt3c& zg*R^Or!SZb+79D@pBzx^M0^wUX-uovqBFneWK+QpTMoFZ3xU!pc8sbn!cwK57Zu3Z zfW9=nws;Z`N&WPIb+ZLt5U`(2E;h&^|00+vocjI(`j4tAwfPFwubTMdWb(^b-=e4v z`5P?qAb-9T?whzWfyiIhA!V7to@s}hFDE?!h#i*wG=`*1h<}}{;b}{K8kghHhFCuD zww>$h`nPZIRo)pc)RotVgkinF{s~4`lpo#-tGYbB)RoN2FMYoAU$!Ct?X@3)x~}7H zAGg68?ZL!TlCMzw#ix>Nyu?oivraZkfw1YN1E0^0K0hN_N2Eh9W3MXlT$!rgGOS5W5clp;`7Z+Z|V1W z7-YK^MA2485tedJ%b0#SX8;}GhtUyut)DJNxR-=N3Y3iwr;|iH!7ys|G0i@pF;tuZ z*iZe+E&@id_-^e}gd4nneJ<^l*-?}>`vQW{n}qtOga3-joE7Xn;w{(!W}h>epR)08 zXYSGr?>58}W=~t2B}t-eml1*P5GA5mi=sdv>x62jzsO_YbF7ypKj>EuNsuMu2#Ve= z;kTSvb_>!D$o4NN>SgMU7w)pG>m4S;!SU7Ts^0cih!X(M%@`+d*xIHrpEdbHiQzD1 zVvf8Dw#OJ(4ZBHy#UcP!TMT7!^@Eqf8AGcDlB9lsxi)54CO8 zUl-l4M|{q-{gQnts<7MKs=Vy>?zDk`rh8H}(?P{v_C>j~BQBt^!UPmvS#xNt$GZ7a z_nke{L4{N;Uti# z=C}?u!XmNC(!co^j9D#xsA@1g&1mf=n}?-Imo$YZ?#K+Z7V>Uqk9(y`SfLAi@5GzY zw6q9_LK`ePJaQCweC80kgou$;vP57xHZYTL9Oi0yp*PBA+2N2mIAlR~P{>d_uLkhl zSCDeseQ|UzM;MXq=ot!|jsD0?IwQ%|koj#kBoN&B6HK;oxLw#YNne!bMABD-~d5x=huF+{XrRQx zN4PA&WnSXV-soT*hB#=rt?@SDO1nkz0LWNM1;OFyXA-Y-%p&TAyAeU4RX4#s1R;#+O4&e1H?O4VysX|0=1`l)vU|7zo-6g- z)O4CZ#DS{nZV5#oQ~AJvcGKDU`UR=GQ{lJ_(Vzy1YNwMpo}>hEwMN2;4{3>|^LYj1 z`Gng&GH_U`S8OJ(3WF@2KO=9wfRZ8%eMB%0@|?+lQJ5R-t44bmRgi_GwA9C2(exg{ zvN{?0{PZ3$4lvdIQ#p|FNwrIimJpS>n}C${GBD+!d^tc7GR4vf&h#rC<+`+|>GpRt zC*G)T^a`CQ*CmW+UYwHAGP+Uk=B51m;>x=CXc~Xlhljoh-3<0=U$nQPAmIcp6@OKK zRn1(zs&XimG)I&c%e7gagQSKSQYo35xIw`&P9~%wUXRPj9+VR_> zMPGd=+CDGQhEtGkNRTnz!$I>Cr*vi$5K#C%DB(>5#-YhM`DN5L@EU^C+kknP!o<18WC~_oHWW z;B*>JINUR&(KB(~Ny74-4TUixisfVgbyc=!TbfOo2soYMv*Ja_oaVw_>3@m-$$;Uw zhGDoyk!hIn1VfLW6%5ArN8a|gJYToyx!9Rt@4J^8UkT{?8UC^`!H2=VRf%;t#$}g> z4G(T?WE2KP+;|AX95sg6P-ZpHA&z!Nv@hPN)O#+t}#El_OP z@!xWtqpi}x4jd~XBc?BVo%OAusG7}OL<};eCUFCI*}iVNZQrF)cR`Azs)KbjZ-QVn zy_@4I)7d_I?@ib;DfgeD#Oa44fbYRb6NmF4%{maN!(?YRXkpyt-mS z_&RSCsu=pE+zT~6jQUNdtoSt|gvmdseX+H}4>iZk<;t{zF3yXQrpc0MfdNlu%sXI_ zK$@6J;ks_ja?-{rM8tssktM}bLpL9ic$=> z5J6gLa+J#8e3v3cu}Q|sy66}`Ns3};q(C)X9Ru_z9Xz-js)hvZeEg5!^GFdYtLpmK z_?$9=gRgt?@bP1I%u$a%)SO5XhEh=Z7t%CvFKQ!jFy@d@O%EWoDorl8P$^1KZu@xh zBeyVu*<5;1%9kXkpO0cqV4UTOGJiJTv>4cZLKHPYO{AgSRHQx~P!e$yQdW$I@rIk& z8Y|botPL9<8jb~9p1K)qw&4y7<0O%P-^xu1@9 zJ5v6)lbg3EkfJ8kDu9b#`o7Dyo~Mg=f85pwIs!sL(yM(T1s{ov!$MxVL2^*WnC%^L zH9Gh6XUoI;`-gg)KjEAIlMS;9NkHD#01`^rGo{ltgB+id;f3dL1`MM;?HFXd%X+ev zf`Pwr6KQ|TbnSa%JAcdg&!Xvi^t|Dow9H2=_I!mdVz_>xMy%T6d}V3;fMPKo?O?yx z6;S2EAJ<}RRK4Q9v=}czOn*8VeKwy!g2?p67Gr!Yr46Mo#!NMKaW&T-Tj)7WQ@s8B z_H%&$(&Jl{Dvsa@&OJ7x;uC6pkP?9@c30UPBby38CefeLXfkOc$ztKbK!H;mdrpPK zM&<5N4u-As3HW{4;~}B*Z(N$KI#9Igo(-^fsO+H*Qt&-@_7z_(Ndo^O{f1uhWQqJG z5DW76MhRHDrcE%o*LDe-eekA9x>N9qwQq{dPwB%m0>#@RK{YQ21%^tL<_gu&=n0mr zuqB#8D#`c|hy4k?3kcI(fUeM$>+Wy(5=VF{B2u?cwJymx2ZhfpcthE@+T{u4pT0#` zbMKI$^LN@aGk0DDy!MPU_4UC>*%!RcXTj6Be+ud;L;F5;yRDDKR&-MKWXp#(^$}ks z*;bZGnUL4;@fdUaf*^l@#>2WR-5o>xt|70J5l&_r)DZq@i14pNgnx?(Wac3Ocg9N} z2d&8m&kw7*tfL1QJnBvOo{CX6PBcaL`+0scZzvo=`dL7@ep96%h}Vm}CdhVK%BNW8T9qz6%N`FVd{PF6ip*PPQf@KxhUUUBQ)rd27tava zR!fAHy~S*@8A-kyRsSu!TV5&Wvcq9av74}+f@udIH(RyL#mu^(v#2b*exHsLV(Tq} ziQE&1zaSQXIm{M`35u&dMjA?878<5^HfQYehf@ImTU{UAjTGVb*qj3VwTb|k zLnDE^G6Ckh208`eJ%dru*SLG)>B98NJVs<2IErsH0ZHyV7=d`FlA|WY0Kw`cXiy68sG}RRWG84FU;%1X1;uA-d9KoVVL2Dt=cUD1Wa}9QI z-V}3M2`*8b6uXM~*eJ6Xp{pK2at;;|ZHjJxK+DP;rKGi}Wo)E!Dr5!iLR2ZjtJ!%) zpJy~4#XnfV07#*@mt*wF;Ysawq+Ayy=qQ{WO^{m&n7Nw-3QUDY4hc{r3*+y)KMoce zmPvh$$)i}?Q-iVQ*=7;t96`Yp{vJ-QZreCH067zs5~h%dO~Zl_Cz`#=cWAn;O14_e z!V_5KQ3I^%x=gi}!8z+%?X=Td3AVzyC_q_>mZLaOKX->*v}1NG3(;>jWSLq7ocQH{ zFq0a_W#NG4yma4;q6wWGMDbV(%Emi}ao*aBeBU2il9xI5uXppiWUE>NW4fXJkTu+I?5VVz=Ay5(x$Zf7JIPd7H7{Djz zf}@7^Faa724ozSVB@B|Xeq>2dy`{O{2q1c(0ZH@oO3Px+Jqb*o3Z-c7**y0K>)+#wN}nHE07^&ygZ-Aqb%XV}to z*X~3PRsqhqBJti#GwnO%_Y!-LPKt3!CT9G)danIxen^pX&ifqhc{XR`+xrz_)&f2% zhq7~s0UDn+SK6w&7{dZ$9SnUX1QVBlc&P<}rIC4Q!2D(c@Q+t_$FO2G)0H}Q`>#c- zKFIBaXM;@M)HWr58?x}kYC$8K4S$B#$ep#IVNoVw3B<`u27N#$Lr?CoZIvmNm^HI6 z#nIjgikeRlSTwv!+{xKY-zvF17hd(?e$W>?WBTkazz*RYRJ^8O`)(^iYGGDM%sZLA zvhfPli21&qks%O?bTq+>0J|C*!lN&D&XooM8)p^F{Pb4A1RziDH{8${gd($Efca0% zjnf(Zio%)|rn4K|*Vb>aGsi>qQ_$N6B7ThP+tS>i->aAMOswX`8Qo&R1RuzdZe&1M zoB*?M95beP(RDM|PMt>0c8|WW!pN7f+gat&uE7(mF(hhBr=|CIvx}UuL+d!fB#}oS z=f`BL8k|q-o4FbA=0+)sQXh_ZBEtl14MMFZYlqQ3SV7+qLln|PD~{wlZ7^9w_cr$& zo%|>9Pey6bQjx)Ow=KfW{$_4(jI0sb0ho4vZ9t3?yH$C(GHm8k7|jh`vz==qdqDEB zkxlb_t7~E~Hz_7^n;3E8J-pgk9$QRX=>s_)BJ<{@Q%Kj=y7_Kj2YV@SY+F@nE%O2HDHE1v0oE>47<^mm221u9tECdwz&Bajzi8?WQ2u|aa zC4kmfjcwVmqHZ&@9M3-MI~yKvIRuq6KK9(pTir~6rX>6#cj6WzA`jy zT0~|_kYMh@djd|m3Y0}To?s8ZDRQo%rTU6oiO8fB*Rg{#Nl0bm92Q;GScbJa;6W$d zhR3 z04*LkG0zmH_{Mxt_d5Q2(?iVCH!??^5*RNQXOLYr3cZoKbO7+QyfMfwo^e8AGB+WU z%)^;tqzWK0nHfQn)10KSQ^w`wVYLjK2B7I|^7>r`g>P|MaJS`AIuev`6jHIQ5KcmG9^x7k3sVOa|?H+_LmU`{zIZ z@Tbx0>fRvBtI0Jx9ySPD@(q$oEro6Jmop^+YAVvhC_57gr~{G2tTBz_dL>S0qIx;i z?*_d7v_xx2-&=Q3UdJpXpf1M3Y})vW+-K(k)^G;iA)OIXOG}%g@umUEgrCI z<$;*p;*1?fo488?thd{1``aD@$P`bw8uXR){JqSm&e+eLemLd(bH=?hLde>&Hulj0 zvC4k<)83)$ZKF5iAJeQ(Y1SW)+5sMOK^U{6BCam>86Md5ISY0%c>=O}cFeb~!9xaj z7dk-FGvx61hbioSK0!1ZJ_*XX+zwXU0oJ{uMMK_9dq5Mg&Lh}tSA!rvV0RE5TJy2u zHD_>YD6QZJcteU?_Ve-O0h-d6O*MFRhNWa|kA05=LENvgUa*0TjktrRhrINT+yh9=Z&wP$D*sL!wkeqocG z@1&GSLX0Nwtk+m5w~eB$O$ODa^>CtcYj}|#kFF;fsCY5f=WFOpCrb*xggPnar@e}X zpY{yI29FI}D_J-DeL4P!>!(6H)zzVNd^#R6uj0>Bgk~EkWa}uBlR^s&t zbGf5|#|FBLOH6h(Ji!>FEg=<#Q7EtiCBf0!cAK|*I-k@*kW2skr$7FXm@B5$R(lG{ zOD21SsFyvcM_9jkCg^57I{)^rI1qScO6)uwEcVzX#oi>t2PqjnB{cU;8lI^W%s=`T zyJ66Ml?yOlr!^uqBP|+C2^tg)t)WloA~#O-Un7`oQwCK%hpD00UNXFhnc)pX$cWEa zl>2Q~D||4#vZosMi7^wiB_u|^py-#o0qEu844*>!o@1soL2tt~C~!C>fQYr~Hz0&Q zd6(A+Hl!rX)?0%3oM|>utOksd4o^lSPLnWz+_FWt)SLe33dyK@a0YBA8`dkfm&@ z$qS)cqj!La;)T9z6AZ5A?i%PTRK2E!^lc!Y6H^Uvu_I5yvqC;LKFh!)hWEHc z)4k8^_h)B#nX-)F8_K`2;>K%vfH;s_!;{XQA^PKp1TB;YJqKF>1CVds(_hvb^-vfI z=#kVD&!CoOn2?hw>)0$se?SxHN1`#SQh}uav*ki$xtz|O(;+Ft= zMc_hdVb41)ZM(!G?>(0?d76$p&Cebd2RT8~)O~|j_1bEO?*L>l6fac$Qo2?zc%d1J zsB7b85`C^kS+)=BQHH`x*q~C^J!~v%vtpppL*0j4(W1stkO^SEl#^SooBDvoagW43 zpN+lGcjnYOGj6oaT&s2Wqh&SC)Nl8ff~!F7h%uk+^+kwp^$dQ=i*CT!lqTZR@2e*N zN~+LI3PEiJY2Z&|WaUCzS#lkma5;q_Bm`dsKvFS*J!EHRenazE2U_7=a!MGzx9DEZ5!$pxZ?*t?Y0~9>NHk;#PzLmH;0!M)Qx+sL` zWzl8;0Dj8K!twDZBT{cPFMkAX&Sz6J-78X*Axz&{wEgd9@NhPOc>2A8NCpXShpP-agQO*&uD`GM9=-4to&&;W+%@SlJg2cbe8w*CJ8{e zqM86E)$V952`nEk>*_!1-X2v*nhc?F42ig1|Kn!73NrP$nPQU`An@#f2=nj_LD+*>qX!UL znE?YoGsW~bA-vhchN{Yo(!RX_%>BF_?f#>0J30z9Vi#ccRNMTFBzKF8M~D$P7|lE{ zj4AdmC;{0>Y-bX1u<1LXm=B4lW3trtLa&rXOuvV*zePT2t2Cv>{<&~*f{I|5ABBr3 zuvE(WRT|n0q_=O!ewdIfkzmtXR828Z>Hv9lOGmUKq2Lx%06O2ybKdGUu0ldmS;V(e zWGyQQnhc73IpW!bd`I>t1on3#ubMz6O7_mR>gtw~Vi(bX*oU)HlClJEvBV<0zElq>nn0o}kk!*fxo4x@Zk*%uaji5T<__sAVS&g&s!ila6O9q;mTzkYNJF1= zd;=fR+ah!4Fr4>@Z74a^*w*J>$)sQ_ghpNH#6yrp2|1TNEc^h27s4fnJ4+l)gK_Xg z#(DLcx2`A8jFny;0Qv6;cDhs}xS6T#3yrW^IHSk`Cp_Rr(Z$f)IUSfZd8>DuI(%KC z6H7N0&ZgBD_X8q%MpEU_s(m=O@Og&aGM?qbR&+wdJDjekeJp}^aj=@r^Cwsuy&*|N z+P3yZVN=L#Wn@hKi_Z-+lL-+x+cYU2HeV#^sh)56_~>>P>YemVJx$znIj)6jwtKu9 z_W~TFS;smK;#F{px|PI&q_r|gicx%Sm@V5EbV_8#!uyQxC77UO{o%lr5ghK0nE>hA zz2Don9%Q;+8d$3rSk`S9i6$}w^LdNEM@AAahryurzbO9OO6x9c1$hAaVw+>S=A5li z=A*uUbJ7#}eC!&S)I*i9J3K_gs z$MO_o4f=KW()D_E*+6q_^R}~F51GLPLp;J5Q?#f^S~?P&?-;sR{X8yG#a%bZuKe~|jq(rpk!vME;%ERt5f`i%wghkQ3eT-yI znn&cdE_Qd@ZP{xTaG4YI!$Kx;V@myIetBzuE+A`&;A3kvlQCot4E0Z$?idZi?Oa4_ zYBA!2m9Vx<_9b`_Gy^?AMd>ZLfeBKTr5yl3xy2)imfg$%Hl0P;Am5pH&sv%c6?tpg)mH^;rJ3ac0ZKhbv(`UGNTgJpeN2Hg!UNVaX;OMG@j=W&0l zeU2IEgfvq~yWtLX|rS-NMAG67iiCKL(XJ2_<-$bEF7- z+$&W|%46O5T$H)%hIE&byNWu&@51uACx5H~xCKE9YJ)U&Jc6>V0GB{$zj9xp-=hEV zqGlk$RW3mJWZm?cvH#lb$4XtHaw5hnFhq1h9XobB1{I?JE)rikas=PeeL#u4| z)0OM1Gq#w(CXKyAfnc-EcZw+$0K}h*vY->s0mIt#>FXJvh4CoHq#3lDppn^>MAL?I z>}(8^oKK!>tY@Se0fX zreB6|yJO`mZ7|m-hE>_|u&EO)Mrx=UWr=v_VMxzSZEIN!6%!bBWo`mroN)@aaA~$`_&%A+=w|NQ)&7gmqyT9fgwJsCzik? z-X-bx2e#7b0T^1-HFpr%M(-a{cOstVhlqDit$Y5MZQ!FuM9KbzU*WWAX` z_@>xs+xXiqK((4m2p-7iSQH#>)-mEPP2Koq^wo5!HF^R(E9(@*L`6n%$3Y?Hy|IQA zCXO-^-Xc$>!~0<>y4jpAg{tFOW_{uWQT+YY$M_#$ap7ZiGYgu7m7jj1z zq}tt39f*e428`!UFMBf zPH>R)I}ku;`Q2O`;I!dR#AM%4zZPg#LLYNOxa=Ff^)g|1_i;Z(6E?cPnB^#dNCnY) zq6lf!LwySfqJAiwB7itw)~H?N`cr~SKb%bvX!n=E>3MofY8wfvQS|WNPVlUiI}z3t z(a7$c18B4_)+(f7rwpHLlAc1`Sv=cX^(}5rSv5M^5Qdo>!p$8@@>_bW;Y3J6bzXYI zsvlrGTa45&melzD;m5d4vp`*Q1MKg|3!YP+joj47K%XJWf<80aJX!TRZ0m9u*V}$TZ=6$h=MBSqn(= z1dG>Vz?iCN%ff19hy_b@E(u)=?vDO_)+9%hzeFuWpzB;1GPj8l5VejcZBU2URxZpN ziLZnc%GIgL55-nr4%(h;K-semh?yKDF!8}%J>E~PF9mUd3^1O%%A{xpau=1s^BqIJ zQct~ouW6v0zSt=iUevGDD{l*z1%&-gmjM!U-F02DU!nYp7&}bRz;H_sQ4^#026hgl zU&QYL0LGVT@HL~D=g;m2NTMuB4!^DT46Cc_blQxNOqtPB>82Aso8mxnD74G2a*Kv; z25OB8%ziqxSQAdvKjA5-8-|pf4o`J#pT+aSo@m!+_uub6mId#%U2X%YzJ7z zFgL!q4cZdM-V5QEd6R&`&7miTwc#l~m&@YG{9qXv7)5tSRq6L#e)?GXqf8H^M$Rg) z23mcERTHeoUx91WJ0YVg$%9g$Uz`l{r00`Lhn9}HL9{ZN+l^H1Ir40Kz( z?8)X$iiR^|0*|w-@pbdf$JLpZ&l0j(Y6c<9zT9=eL5Y>c0R0u`{vZx&m@jMqH;H(DAS zp3pSUqTVSU#$ySl&OvZm7y$ybtM3o`K-kUx))%mO6`bk|o+U<&<>)jqLkYi?ON6H2 z0_^m}9B3~(5Fou5T=Q86bT6w0mK1ZFHL$pAJuCCO%8znTkZa0h=}GB;oapT0$|YiY zzfA_eS-q1cJk&e84{J}HpjXd^rL4VBTp*RnkFm$Ns8B#}k zk}5Xgjx@9@bqu$3qkLW09hV^-dHAF2TC8oG{5a5BdYuC4H#~x{1PNM$0WLZKEUH|c zG%+f@9MTlj1CWlogEIgm)4m8cb)@*qWL-z|lb(a@1t4mMv;j>UNf6dWEj1XnG(|*} zj9|@$JVcxlj}E#iXzc_ogjpjNU8$bj;Lx@aL?G(R_$J8wZ1N&_)pau#U#AFQOS$$@`7eh@&o@n2waT>t~@NwvqG($<681zIW8Kbr2Y`-T*a(L@ru9JH56zxSdc`ZjpNsilFp$A zSe%(T3y6nQypuTWbR=4~}Z$J@Je^t0bSOQiMTEJKt32bn=Fl4NTqFlskTcF)TIGRIh22^(!5 zrSN$AeOU7P_glCW{^M0lX%i6Vi}LPVn?u$kq7m4Y^HYDBw`!L`*Fh5{@kp@zMBynX zxpZx!i=PiBixdAgC%**&T`PRvcjo2#!}7L1XcKhB)}-ezBSl=(Tce_jzM7`mo*Gg0 z@poCgRM^yMG8I|mgkYBMcM=#Z5wAb?QqBS`Ek!r zW)hlk`$gB3WJcVTASRJ{F$wDF?v(n3?In46zPmj*6W;I}oD5w@UmdY%f$ztWM(`Dvjcpt&*tcC;~!#Plh0Gw^W+HR6=rGnn_T93 z$oM;k7~zfJ1)aw+JIrzc)Ig5_>rzV)w6^PWEr>HbB8DBn*Sy(Tjpzupo507J5nOSf z=eYyWnHF*s_Z-9^T9LwB+Sp^+$Jup}P{Y%(ZwUxv2lBC`sc7?Ie+?3by>QA=-a3yn zJWX8X==Vh&9zD)hSJ#OQ!gkyo-L?+E{drVC(`T(?ipN<-EzZogJf{J(!Qz8l|-KCjky8~8ELF8+G)U%wk&E@3N5Mn0tk#eETsE?25Pmgagnbukul zzn9*9^`)x6R_`)ijS_lt<9$)?gvqirU%+xn#xm&uOMF86>xFtN@^Z8jvU@PH3UBmbXW4 zsur(TVS^+#%!wxWmWbi)mA(dNnaT#Nl>}c*Gkt67!dPZNnWimGRk0l z%G!B9r%@Esh9-GgA4AN58?`Y5P_FDZ?&+BnnM2IC_T9|rLHI-x-+2(G%jn#{NidK@ z&4}Rb8Cx5dn1Id8gV`%nmT(?3cC3Z}aGs#`!tPnk2k?9|&5E9xJ%5r1*X<9H_3Vp= zXH$Gu7SX&~d~B1PWZuhVaEJlFtLJ=fT|_I2+w)u(gCw&LRiR$8pgEv4#j!AlVVsbf zVjf`arz@xhtS_$pVCwHk`P}>%7`hMI5(f%1a}cneY<=PZ*ASzvO>_ENHy>esKUvFB zmhW+-(>Sv#JGk#+PFp^(%7+1u1DD zND1QTuH?-Jj;G#<1f3w`)>K?81IMqO+M@N?05@l*G1}5SYFKZV#TJ#bGjv1Z9xNLY zUO7A#-F{&VN+49%mRf&NdwhsvgKsR*AC6b_Ae4PbVBVndcsgGE>+Ci!pF3lg2M~Sh zM|Z|SLG)2}b}Y#kwfqFq(Ac|IX7Ct_+6IBz`QgVur|U`kL4RuN5?x*i>&J%3x|j~@ zd8p-3Q{yhDJXNqpXc5O!j3pBm#0sW>vjonF?E(!K$Tl!U5Ha~Tr(`Lb0gL=7LeQVy zH~{hhF4_nF$z+~ab&J-4e&vj4g1yCS$`p3jwY|gpLnTJt7(E^OJ2V1l)>PsRB;wU8 zwgLl=rv(Y{Q6~;=6aZ*WEzE9Gx-tF!9w+HBxYL8Q_Lp$ELuNrI`{hQQvP(p!z zaer*x!+1@3G&Was$n={D>i9iFgRRwbNhkY$ww$~j=~bCF1gx>^yljUP(?r(e1&=Ab<H_arcV1W@ zPw$R?G_J0$fBxa0)3N~RefLcn)``V&MbM)z<|xraF$vyvvE!gykx>4H-AhR$lx0XY z?NO0-x9`glNzZ-}oJ=O@-zf@~gpO29>uU4cY>_5K@^ID&W};Quzf7y}IpkPtXBTWO z{$-sITi4LL@=|uPHSC@0L&#X*Zc*%#|9;d!xioG-(uh7e8 zl6_gDQZ3$3SUSHu7_X?C+&xo<)X+Dr>Ux@nT^$7nevLU^rCdVEm|D|0ovKB7&4QS( z)xNbZ<2&pzSEh^MJX{Lm##|0R1ue1JKnVNW$(WVIt46``X^frKd3_FCUBs-m{w2kE z5N7CPBn*Ri_-4*e-zKBS;<~u6wx+7p&P6lV10wd!>Uoa$bsZ745%-;b5s)B-@Xgpi zn!lNiVC$^(+vVXQ1Z*$3WfhQfkVaoI*DC`o+&R(B?7D#^%9dFpDRWnUe}Vjf8XY=?$sQv5ddK9iikU z=#WS6yd~F4#c!`9hZd&vZ**F?!y?PC)398F2!ciktyU?(@lC`HTAD}oUZVtYGu!ly zR+~%Iilcq3EM!p@I#f3M(OXlVwZKI34Sc>Z#!j zC2ha7kng_Lp_pvT&p*hV#bA|)!yYp0*H&{8kUZ9G^Cndmu6_p z1N(v08=OLtBy@|Vh6<;tST`gUBsjFGtHIivRJ$|TzER!iH7`nB0+V%!l}p#R+I8Ol zB#oiSP&f98aesz~oCJ~y{lAaXs!egRjbivb-{8tOkS_L&vmXu_2ERvYP}-a0-&0-s z9SiQtQw$6<{&aSZMwS!uCYdnOumnbVF~wpHqNX}AcHbw+cZ#;k`&OOlgRA_6t|JCW zC;}xvN#`DE3XHVDnFh};>QpT0-6k1Ta}3Q9sP}g`S5S=DYJnw{0bJnN8U+({?6z!&{m)_{M!2X_$(2-@kaOvr z8P!%rc~Ms?IUMh&Al;ChY>l$$qYNQpCAfw~X5^-=<;U64ZLAXe9F{*%%!;N6Rt>!` zqd9=oG{_WdTuQ$zrRFwMe~Ib(Zy8=+>f7+ec(lWt@dAy&*;l>1B#UvQ;)({s-Vgis(fYH>rnl^Y^i~^-R_yZ=q z2GV*^5ilyt+L$ae1@2LK!M=X;iO zdNmzS&CKMCfC!G6biBKm-KhF++1=86_gm;~xwR{S7wBD>$Ra<68_NZEx3QS5K}Kv^ zyw%1VB^iBwiThuE%DR|qjj*{rys&^f;&5*%Wi(XuwA0#RMI-6l_TRFYSxZCy_%YQl zS4hK87;UwnV?|^US6iG~^lh=##~e~Qf04bbldIXR5CHo%?}3JMYqqSTva^94;Pbyw*nwKR$qfaEkhUaE( zR7o>?n}LvyB-uD-lCv)@JXwCnfX?NLhK_xgE}5aNilahP!|pvjk$U?q$A`GiWkAu! zi2xuKnXDr-S3E&~P>BI{fQmtF1Cn{$1kjmRg5TFxMGJp3ms2dLcP{~Ge2hzkOhDS} z5<|XK`Z5rHS~Gj8HeobYsb`6@#x0N5SWU+lv@klpmN3R>Oo=wEB4rIiS=~%-AbTf$ z<{Rr=YlCTpdtAj`)trk(q*PDh&b?WXSdvNUUe@q5$bsgL_Xy_g1wwei1ho)X$tP`Z z(;TPyTLvPY=qdc)^4HuxQ$in{+Ts29vDm#R`EX!4fmtflSYBR_uc6H2e3QSJ@HD&o z;h+BWFFy(7pa1;BpFFTAuAxLeyu6xRYkwb!ncC1jx8O`+L2Mrrew|}{7-hrcX$s8H zp}Gp7+SOPV8W$HGGA2)|*P^K_Yi=TDFfDCwx{$%W|5;{NYK!l6m;C|S*cw`D4XtRm zd9!0Hn#SFo5r16bQoM%IFsb?j(_93QIgjxV_YAJgTfR2EP={Q`=C_^MflMxkqT^pp zFe{_2J`HKHL0`M_gCj|2rFfsIPSAXvbJRLpb7Op&anB0I7fnn+`Nk1u>ZEpNEe%Rd zr28sQ5_O+Q))wxvUHYh1x`O?JWU zX5oWJUmM8CIPrO#EWvO>qOOxWyqD{u*H)SZ@BE?0WWx9Bxs=)5jbElJ7hRjqG`8+g z28$A5T10^QcstQ~zU>u{oAJt5q--)C{SJn}%LMJ~qMoVI*^~s- z3gaHFudc0DGEWHLP_|y(nkf_y6e5=rt6RK@(~qn02}pp&i4kmM7jr8>uSxhkCAs3s z4(JJ6+2T91(Qb!^yj1%xlXHar4rk=>`yBr`uU_*uD*g;srl%gd#Tke4GQt_h_ld`p zrsKTlvtEAUec$v+e&&!qv7Abm=`sQDq+GPBerRs=n@utaXXJ7>m||cbb7Z@sIUwrF zXB~|5G~abJKEOQ0kn{#FS)^BRPIDfoD2kBm-8YU`F;KL6G(jG}q zqy`dhx}!(<2|wP!cThpWn*R$((m7~I)P9x&l=L3-kztAwI9nE2$iL5btP-7rWb_zf z{r4q(2a9ylE`86i1n+L&b#yD4&H0rCxqH?mg}G@J&H!g;8X)vjDv3!jQWZPr6;$$t zQ>k{F>6(Yzy2+N@`J|D(b+OE>BHb?XWYZ1+2RH(kdy-2>Pe6$jxnns_34Fc9B=3&R z3Wqp5x;^o`Ox~Dj#0b3ttqWVbf7(72X}Ow84B|}527olLFBr`BfSU{CuHC}IdH-zO za_X?(Hn#h~s}g*7`l`}azJpI$M{NNd*wFN0A&3iKZ8C*R)HKkV&MspTll*DNT>+1D z`*hw}m7|jVla9<~BpJ_r!qCopkU~SQXH#kFV`QK;W_d3O=~QXZLzb|J8@XSKK9csm z6tWFCTO0IZ!3^23&XO2e!Jrz4t}B?8AvW2VC0)ddnDET4=sv_67T3P$a}Qa`yOZfp zbfGb7+j?8D5_`Hc#bv*F*7=hB^jRlNX;v&sni6ZCYd1(il~DJEJ|qmR2i}+BL;`N% zMJePC)0WPZ0&ednNduv?Bi6fxqa*)D&k%wmHiHHUV=(BY4%e35?Kzs@NINW&C5bFY z2V`^@V|1vQ6?@6_O@4TSsKZ`#LPq00NBhAx+8>q{y53tf zj?7Sk)is))g_#hl{lXYESPck=D>CW#tV-Qat(iX5A7T|AJ4y+cp9F_UICh+3O5rLj za+wk&;Ch^pEQXk({3Y##q}-ka^Fz!KCXnz@!mbz+j!KdVd_-Ik&PIH%h<%NO(T3cF zB>2v&pEK7SuoB_}haH{`jid{TK+;cmm59ja2dVcmai;M-Q-E8S6e)P9n8N==ie&Z{ zqyghPOX!$_qt1eTDm6!>al-sr#C{`!=xP5Eq4$X-lUPj#c(|4H0hVwelZd3J5Ql{h z8)P9gHNuQ1gh2-i!!oYt8AvmdS%u?;*q;EWo{1=YPbA$uh&m@%2Uc9c6JQ3%IW(y7 zp=7}qzwNXB6H>A3V?^BPV(iS!$vuuhO}=@xl;rV|gNhj9(xG!zJk`L8eKQpiH7KWIHqjmzY@A}K_Yxx3e3xPE|9jKd-RWCo}7IJfA^Q&9cJ6se3M zJVP)-%8f$c`XKEpfTwcaKD*=$tf}KVdbD|W=A!MDrI2MEb*=P8gro+xhWS??Q;L$2 zrCWaL$ZgrmNJ-fzusq8QMy4rv6Sh{z@}ybN88QC`-uZ3!>L3$xC?4K26R*?7bTfUJ z7#pwSNz-7r(mrKI^YK>qu+Q3%jsbV19wo5Fr3j_IiyW+6#C z!6Bucc3SEEChA4wb}94axknvNr^2Iy!^}(bRE6!bG-z~?OcaCn2a`7)_Lk@RuGSl5 z_F+d&VVJ8$emoY|jwgxV9skXunah`5DoL49UN%k9Pi=jB-W$EA%e?xcKjd4)khE38 zgVEI_64}6Y#^hPho%fEaFO3i2UiLCCvF)$(9mAZO8w*+s*cv)@YL(-4H#4g4sHW%H zxA_T@KPc2F8CgoPLnCh&`Af0YQ-SNujtJY4LkJsnr;Je&WwbN47^h(e^XN%Ql@)=FSdt5V$4-&*p z0%#a}>2Mq3yL>lY3^C~;i-v7UC);%IHbK`aFHSV`WO^vLP#sfK^u)w$K#zmye)KYL z(QIbiO*Pq8O-#pngpX;{gwDF~xZc*Zyo+LKiq#Pju=O$-6q+m*H`?LD$-Y9oEM#12 zaVa_c{TXD3lTB zj?P0--k7Gy-7$xrv{uhZxopcyPe^}6+g%d+A%=d8p`SS97PIjPExf3_v(w>tbggY% z%i#m?c;8oF{9+3T))=8CjI*m?z&@bcuXTuBetAT{Jh%WSo2Y4ptaM?m-;>9M$0G%ezlj@*H2Ie5iHy zSpOMcd!@((@$=<+dUZ3o`n0n=8Wkb(&v92KaHEqLKEL*)TEbjDo$QKy)?s=F53F>z^7| zVVW?uzO~V#$1aR^1>DYxdPGDkjwLx!kp$Wa!v-BD`#4P~ET$>)D*}(juER$gUG6s1 z5_bOCXHfZMo{EODpY9OHe+BwmE*y0C{M;)4ErL)PPmxz4_ptz%$!hP>iSLnQDbgS$ zVCtVbxnv-^i6OH={$m;lTsM6o!N`Fh&R#55lElhoRgU|V4r|Xh*ut4 zd|KVDH-yo9SpB9hRtUULwIVkLvs0_yGH(dY;+uWb`+d)(+8ymlO-%KBWTJ&FL^8L1inq#i(=)GeXh*)W z+h3Me-GH>3>x~=*v;@}YrR~L*VxA$po{AVOimEa5ykMT=SuDO{4e+(fVNgM>lG~<$ z7=^#)?LGuIEbzv&{x+EPx1nsl4P^7}TsGgru+LH7T9+5tC0Q+dLe}jPH0sWHZtr>x zqU0oMUZawbdHv;Lnt|#znR-Vh&$EVbE6GH3bIcXs?zM1;dGCZ9-S0vM6sz#Lr>B9S zg5oFg_ZgcbLp(_HY0gYfQVm) zvhh;8hWOcpwAlZX2x;2&a-%;Ltc#wYAH8eMgcO}+JOY#au1q18;&x?bP*WTZ_;hd} z!@78k!gmANhPtIsc0+Y|SMkv1Yt2IcrWzAuX>LBrc%Nj%PeDv}iG$ZiyS}tr3~@{> zX<2Dr&u3~oQ>QJ?i_UD$5N@ZH6wR(?&|kl}8gKzSO*z2zu0Vksp@)qY>Ko%8Ph9Hw z?bO2toZG7Y{276lw*7oJz52XU?Ur4LS{OE5 zDN+vo=5xV5iQp3*45Zi%WgxMw7M?#pX)kPTNQONHeCu;MZud4wE9%!XdHT{k{nO{| z^V{by>SSH;I=t++kgB~qe*NJb(cgsz9ADH)pw094hkyCg&*29v=c$|Xl0gTv=sL`tfq-=7CqPS#k%-Nk{SI(z{5b;LoQ0^7Ze?5p z+EaLfYW^vX8fJFFWhv{vGDM4V5e&ol1^H=2{M?9$I|{=2B?3+C|NhNyisy^Zx`{qt z{PnLmfY3Y^jGez+T>P(#BRUYw|8h~)7h6;x{mljZ!EP5M>x;VOp+^0Ng4gTOV)A&m zo@Jj`qs{FFB>4ySex?1xC$!G-eSSRt!5nF}f4IW?FO<3d0J2es`;_(%#R?6>-(}w) z4aB&zs~KZxcVGUX4G8_gyyyJ`T}`5O+x8E?qg*{Z05Vj!-(k2Ur}lTfKeTP5A$*X zeXaUVe&io7l$nhF`n$e+{@;K8-S02*_To8C!R8t|IOgp82GTLT!TzZ?RF1m1uL>Lz zDD~_Voqoirp-OD`-ScmW3BG=zkWbe#WbH2(|0VPJ-@pC+$LcVuj)vN~QQtqqY6V^T zyXJS_{onuBU;mFA4;R1x_pko|6N)LE{0?TQ!Des5rp7-xuD;KAyYG0PIrn)x=KhCY zcSHAEKHi;HEX~`1^K#_@6aWAK z2mk;8LQs?kqxkh1008bs001KZ003}#XKiI}baO9kWq4(Bb1z?Gb8mHSWM5)+X>4?9 zZgVbhdBt2^liap)edn*(vQs6gV#`{tEGw?M50)%lJGLT~R=Iw#Z>IPG&OiVOEP9TU%CG1^>#OWfAKQWi~PxxXJ6+}@^3H7mKXzyAF4)qh;Sx@ergsMSzt8y;t0;0?0(uim`< z$EzPVZ{NLq`{S$Yi@zcL@bJNd2ie=}mqB&)4_T>-Ivl$J#Dyv=$7*MbY^w(E_EvY9 z9g7X#y<(yZ*H?wLE6>I?jRcrwtT~TeMxMpFzE(}v_E!PHn`);jGyAM7>Xj#-eZ4J? zD^H&FQ`q9Yj#+S(p*P%^Ea+`l*v7j3)))OQ3uF7CD6M*pF;qSa$1vztWkXTMcb>|k zzii#4GIZ7B&|5t`eDo;GzPxzbSFTp!LU$K0c_v&udwlu)@so=&=x+NxQqY_kcCNd+ z_`16If&+BhJ3m}_#lFV7_9BdB?_H%rNU`r+uez@;-u?XU6(@7!(2oN~j3XZY{y=8l zgRd?geEqf6VfflfTb?}lC#JzrEARhCxxd29IShaM;79f7apl@}>~vKO>fxi`FaIe; z$Cvn9tmJpE9tTzUYWL7T_^&_y1$9WoS7`I&H;*2p8y`Np`2J#@uKw3l(mh$iPo>Kl!Z&TvjfMS2(mvh|tz8tLKcfJBZvkB@;meDM$mXh%1fHrNEfSf}sKwk@OZOQC zboUlu{<@i%Kf2>e{<;=UdsYd;T>t6$U4NE;e#cM9f6qIfhVrko-#3)de}n(j{FL@A zzxj;%?;B+6T)){W8Kb|*AJg7j!tfa-4EL2_Kcfxy-Zt3JXoEd(gRJyVo9vSItN^cc zr-x4~K+@hRU>JS(sRdBl^8)l-#9tSn|D=ddU-JIG={$T!0mJ*Oi+l7x z9QVh{8v2sMeth-r{i~Nhe7Iv|{z-{ZMbQ~ygX=&l>wDkHU@=^U)pQ zV$gr4a!uz-fGtX831K70dN9WVrSFT%M3CBu#mf4k<7|wtPC+S}W{jB$y_|Q}DtcW% zQuU#jRID#l6gLwoO?WI0!}o)QJuXyU59DR+z(jN5uO0Uy*gF^Wpk2&&t#4IisQB-( zH{tyNGx5Y0rIPoQi+mrBee|JMXR^^VjV!&|&g}`Tu3zv}h%8@FjL_P*NcVR&{8V(c zRi~g;@+r z{k-)_zyRzIl%}*r&a8LTpmA>+Jf&FOV6vs(<@CbEQ7)&|%E66Mf+Z zoA^)8O1Rsrnu#wUy+B^t0;>+w1pG%owk@Ux=<7HCa2lIO=5`!7_fKj%zEM)@8`SJ* zCrjMbg?&+Y;ciR#PWg&5GPh%8q~7-`xUP`m-tQoskHoHvgKJzTd{-Fdx#DYVT&3OD zdXVC;2i^Qq_@4NOu~aXV<>pXpUpwCt|80l9lYGC8-lB>kqp4n00}T{Jm(Tb_B8%q| z0(n+6gBz<|h8Vzu&8pmIRZ$k90AOT}5p-6S=rJrj3&)bv z4eWS_zOPgt*wNjTRXcRZ;4PCaW?{sfiDuWqRpl}4)P!L>W_&zEin-c#n`?Fpb`7}| zivkez9mxwY=-miKL>)$H z4uY&{BI&ZK?dh=?V}}>-hL^OXSImycABg>^tdcNckWt_~SsA%$-GE6-sa#vK!(Hpz zURza#I#K7VM6nrTiYXjsf=-Wu7>L2Lgp*Ya&Q3)7W05ydMNFH#@V>x^I4UIU$dRf; zP}>M)M5>VFDZBIdU{hb z0ds+-O{f#;+Y^g{DM$jwjtU(t#xI{k`Rm~ zy%?-BPwTWU+KhEMIU$1_t05x#j6@k@RnnwMn1f(Q;12Xqm)D9dT+BLD0g-}HVfRwN z2zY@#0U5snR}po@F5)2}rSC|7*BbK|ra-MW7G^g(yT)d$8k|RUI+(35pmd1MkQ4Sz zGcaDLT z&g!x4jkvkMmB3R#>i3Zo2B;VGOCMBS$;yj9J{MM41Lz9@pQ(OV1elfT25UQ0u)Sz8 z!G6G~vx~XSn2I7U1sy6)@1hQR=wMAkLeV3-iNsY$E=&H1ng}f=fg)m=Fa`Lmfi^l? z*d&Q)Fua&_6Vtf#qX&iplRET5wwxJ+2Z;s@4^6E`^JWKVRf!%Zbpdk&;BsG+)}oOC9fHJ0SRCk(p*W^O}g9vr}!Ga>J&>mRHo zOOg4@mpTy>Nfcu2dDU%UNC{_4mD#U!x4?|7PGNJ2@4Ul&Zm(snau6)VpLrt*Lksp% z)?hbJFWVLeIOP>Nr12nU@G3CZ29D*FXnfU>&y!^ZoXCa)?qLRL5ZBO?lOq=s4UDzM z*_s@fi4iD4s4(>e@Yar63`eN;4Cn;ev4u#0zNpwitFXPLA;tSCY?~XZ4c6%dwVyR< z(hPBM2NpJfo;;YnfeITa1AylZxsFU6W&@UPGw>W4`1(e`5cXIKz}9)Z?%;@v#B2sg zO4^|%g4Y>H3SidJ=9@f|iojk)nS$kB6yz*E+maL_C+?=UnASmDEiib!&9-dDpuI9L z6T$kGDsP);ieQ#Jp^FybyZKR*cr^`KEQ+R1((a$8n3{nG0Kl+(|RtNt4PUZA_b!Imszt zPBCFM&DA9DI(Aye&UxK%%Eyr0$MaU^lIJS37S_Hd|_SqXfz68PpG zfp6{+_~ssgZ|)KJ=3L;LbAfNp1->~K_~u;Tn{$D0&IPwQ7u@EK;5K&zx4BPnp9jX2 zE5U8ng4?VGw^<8rb1sC4sdoNfIckYHXR$#A0NaNRfOFpS5yW+NxR?T?B;kOkp<{D0>vrlv%6mxVxl7YP#&= z`z35%8L@d~{7zt3bREiX(3Ns~C83?598k%XhlJ2o2wzdDrhvd^&PQq3v5$ehS~7QT zP;{Fx^Nx;ZPClcK3m+Y$pL~66i0dNcg5!{o`WdX>wu0D>?@F57>~r>b_Da2+l`UR{ zmk}}9fw$c3j>m@~4h#59WLO{|FgsiAJ*FXDkCD?BgVz{m*XeIr_d*wGoPv^ zAqDURZu_fJL-avVfKqe%D>|YA#8XT#?gIIsm=Q(yqYcR7hc=+jkVELSgKNk^X40y3 zlHSrSY&%tSH4UJ}6)}%^6%i9^Tmbr@+?qrA=*mR!C+638<-v`Zr| zV;P|y{fSY^p+m2B@zAI9JreSfdM8F5q30w(I>7GeojN*@!=%IrXWe1w>9i43hfYfX zG{FFTTNANNr0a;$#B#QXCyMCz5^G7ipWVv{I*Y<29$^ZYVJ7{|g0M~?`VJWl6rQV` zT>7jWP_eTkmmOVj8eNHtB}&-G(7qJlSR;W(JTa+>NsS~ml9Z361e=H^p{^sF2pu^$ z>|CImuv7_asxY@BybWrkOJZNyv_Qo|8|@1=0NKLP3R%6@Z@ly49OjGIxvsl zHU%NZK$^~sl-%qIE6=g3vObE`9LQ!Y-@ou+|0szPpAYz#aCIA@8830YAe!eyA z<>Ife`5v^dcZF>tpj^5(fSgHHZ(9i`TL2QJjKWm?|O;R z{Vz~H(;z~Y1tI4OC6xC}d?I^YV_A|%bEeXc5+}k!SgI^i{yhg|v0|hfHpV_MHUhEL zp;mNpPekq?g>gevH(@x4MZrrf)kP=Tw@R36=eKMINfz_Tu{-^|q>!@ZjsRq*8*Z0w zxHyUB?+zI?N3yL!*i)YdOm|Z-ktRb~N$0fA_e|)=I0PeSkCS>L;VpeVsE#Kf7@6}# zWW2&Y4#TmdBKK(2F)gl0^vqEBk$;sVaG6jik`m+zQ%s&>t`er0s}%DzVTyU0V!ln7 zV!ll=&l0AXXDQ~pgem5`6!Sb`ig})5Qb3n51yBOIn2ew!axneaNkB)WU>V4xox9SW z`qG{{)5hMksXOi5pLXt0n|jpFU25k(wW(9>+^cr(R-5|O&K+y#p0%lK?cBF^?p&LC z*UsH*=l->+gYDeIcJ5-E`q<8$Z0BCKshe%;XFGSaO+9U6SKHLrHg&d*y=_x>+tlBd zJ}wo|5h++gV<3m!V5wy{9kij(tBiMQE%>c1RT+JXs z5W5Ul0#5}3L#VJxiwwpkRw-$ZoGd7>tb+Q=*aB9pJUsuT2OvrZnYN+u2EWY}>YN+qP{dJGO1xy1AT#JudA2YZz3vij_<_akMBjEZ}xH+E~(7&Dx36&i~w>^TvlpQ_J75pl%M3b)Q zv~J4~W}l4SlF_8F?eB=(l1XnMp>Khd@5SL zqzeJYRWxV~QcFC96z$zdX6~sg88atoDGcAnY2ZHCWP5O*n&|%Rrl81 zD6$!k_R*+2qmqYj&s47$TcU}Hy$OvsC6Z>4h@9ZW<6J6HE^qcYrCB=PyW-(PkYgXy z-GlPFZiKi0%c;e=h0yX6VmWdBBmrMZdeEC|6>~lT{70@96z*B5vHIyagdexo5~7m` z%Ek(h%xdO6J&hZw7&kM~vKal+rhrUvCL*+f{6`PNAp#;#X(^B}^hJ2D6$CI1`eZJY zCkfzS+kr?Fk{@rWpWBBa$n$+C9MPMfkzX`Rb3L=7R!!P#IEMa)S>%8)w;N#0LEwA3 z-E4W7D5juP>3o~JJ`J(m8S{(U6gtA4#~oh-xb$!`t}r2=Y!Ah76BTIPrnfF>a?SVqLL`kgg?EB&?bpZH{#> zI?~sBl6daEqJIXAstKhL+-XZd4?w!LB@AIXx5rt*IQ(D%vfHo#jB6$3OWrNg(%D6lW;mG4-ce>M@P0M}veW zMoV$Gq4Qn(LW4FYXldcO9OSVln^lU3iS};>I4V50r6Fb&x3U!7je6-2A=48n;1i@V z#XEv74%GDxy5s19BEC!71?MOoM>J!C|1!gPNh1?uDS8uEXt@`SbsFtSEhDK92ww^o z{JRjbRxJQxdJ2^)!ro-cDDqW>Zl{s)^9IWfutscd^6y4-;HySf2@7i*%gbI9`L;U} z0tX<-?AOEt3?umA(x5a>A02&MTJ{?t#U}qfN1fuhP}jVkJvlOQ3j%^1aOC>7<9*_T5p56`1!Gs-+FQ*z>v z7RqsiOXqhCIR#y;Q8HlHAJy(AW@b=Mhf?Yg${6{`-U+xAAj{jAQw&f7y}c!ljDyN# zhR@iv5wLlaa}g7U&IA5#GgRueNgpZG8>!GcfFOIQ_^(@T&d0{oH&y!2hrk7R#Cv}XU!XH)W=q^t zA}HHP;6M{yc-~0Ug(0PSh~JvUfcJ7@vK-vJgE*W)2kP9V!2}ypnnZWSEG8}T_a7aJ z)AU^1RN4H~af)X2;v!&hxlyK5lMH^r+|D zWAUE;m+EZ%X|0djud~ssle5m5+@D#mm$%NH(as>9-0yamXOt$aUS9WVdEd_$df&Iv zUZ3wITjn$sc^zNh_k|hX>Gz%Q_XBy~pSRM5;SQzJK&HT-$30@+&(I3`Qhd2yN%prA z_l1*8$SFOtt)@V%6WgAt2#d4leDf>4t!XgFYzSF@k|46vIpTYdHh}P%g)CHii)xC& zZ%3v=s1Z5vStpM_@POE*8|11q*W=x0_aP=t*?dxuC~s6AW5K-pI{J!*c55=(mRWISz7&WmT) z5?H7K79a7lSjaPhCBAQN4E4B^H2OBEQ47Ml#T#l&4$P1PR^5Ap10<#9K}B$$56R8( z)#Z2H?LoxNk<`Rqcvha4j{-A?(c)0-QzWNv5*bG0K;-B`k zpX`d>>HaXZH9BXzs97jlB2|*bQD-=}{1V!NZyvC(zuwHa%tkfyj&=Ur6!aQX?+-4G z-xWuusu3%z2$vK>?Xl`ukEAu{(=<9Yacq96tV2E~lGhaKSq`yKO8DTcAMav{D+fWt zy#}TCTBGr(#*ctetS*o`E6mx&7WIzfqtS9IcFvBHSZK_5-r#v05S30^2%@k2E8)na zhbP@A=%A@O&7WTC0>P=({M{-YR2gl4Lwsa{9ll6KoslXjXwgGMVvR)28K~vz{FNy# z*BK@Go*is0`Mw=oMSLLK_~EZO@%%633f_A%1)oLk_5f5N(>K{d4a+tC^K0hL2N^fY zraOLVwG3$@WXgEl+j$D9eJyoC(-{Hk(NQiMN&(3#Q|0huT6fIleue3X&JJ{`j}h0M zsbEmkQu#vycdpEpOmvWK#V}NCDA3X8Jv4j-%qYU0YjBf4O2I)Emw@C_@+B1*;KRdf)|rNXT&flu<7f-g+7d-?=nSPn+RSLG@(;jpt&Nks^lOm^w7$md9H46YrrQFm@tvo zqHn>l_}tQ3Gz$SI!g?P*GpjVZQ-$A`?F_fZ2#2-kprefIvOG!am9HrF&Qlg{ULY}7 z0dDqR($`@B(|$e)Y>j5E%Xo~<>@JhP;`Q#&=JR(hhw|G;W$%qPll|67vWwbuk%UP~ zVO)Y-3a*<1pQ|2H!@XlQbN`O{ihV&_2#jSlA$6TMbZe-v2})OR`u1^uSulG%@k$DXPtTWi&o^2%Wg1~k;6 z^{?JWXFixC23aBBgfiw&s(WMgz65Zg=SH!q62Z< zy5OXx>0gDIV5Mg4zS+5^!ZBv+bA*TYQFym34{)uT@1K~R)1dx zwOtlCP}Z?MXhR*zZ4wVKHyQKDMgKlBCl90v6y(wcq2Qi>o2@K>>O^PUkXXb7FH*m% zILwCv2nU?8=-ow_3T$2QALSsLwD#mOW$SL}+Hf*wTdohbJ0|u!HL%7Qy|t66oZg`q zZFFQRq(VN@b9WUeKA%7a(@+Jii6=EH44yZ+g*!BlZ*5`IEc_OX`sboa^EPQ(}!%+Wz&vR%TfDzw<(_ZUQF_QLjX+o36Kq2z-c zSm`;XeETHLB*i%<-B z4z|x31NDLHN7ig;Zct5bQrO8*rEqG%Q5HlT#eV~y>q=78n;5g1!%=ahLpGfJ3@;Kv zNFPii;X|aTja+_Dvu(!vq1-w57TQn+)M|_!M1E~dpyBlRk|C^rmHnC{uxivKJ-6$c zJnF%82g~{n^M{rnsQymFlyf-9YiE{8%M);SNTG<|v~MI*cUfp-2+X0YtlL8v<^)Uo zL4S{f@{mn_R#muOgX*H^yrIKkdwN(m_Z-9VR^M;KX_E##OjT)ch7Z4Hs}{|YwOOL1jCZzu%0gd2Gv|`b|^Pmo)lB zxL|slmUk{aK8)2Zj6^jro&IhzA8JO|g7LAW`;LQDd3Fo6c9xh_`7#<>K;O=_yoKd2 ze4u8MaOnYhc%z<($$46;M485BIHc)Fx@!BcQLwD2m}88^T_+vc-9iO!BL_jfKzu_f z1EM9}-J-^r0JQ>+bPiC1_uc9&Sf8G2m`#CQR%lo~G|7W90gyt@+Y zcaBlIj3xIvnoMgiec>k=-Xd&Lx0RV^anQ5lC(*7Atn86G>}MU(;5#SNuxoX?U_cTg zi3BzjYJWF#qo-H*;Q8IWro5(LOAl^(O!qw?|5u96{Yw|I?XvR+dK8O4;7)E~)y@eA zck^WHfe2Njd=$3%XY`s9Q5rAlP^vKyh9#+pLb`U%5n;AuE%fW!+R7nO!PL*Q?|-Gp z7@M&kiU}3j(@oM3Xe=F9o03KyYeJL!hO6^pv;6$r#}jKe?u%tC{)_NK@Z1Mrzgu@U`t7oD zrbgxQWnuJ%Kj%6(w}wCK+sDJJ<9dgmr|ZYg<2@Jt%awn}+pVY5SL*5eZNgWzyx{%q zZSU*s`|u_>BV(%_xvMtbm;-UbcxLwK}V%>@#yJxjolXC=l#{~CG!@$#L3zJ1^A zjap&m;Q9D?d1`cfySa9h9P|D7es*@ef1iA`&bY$Ayx98i^z$rnMKvk}BR|?0G*Q(` zYrh&UJYN`A4jM?Bb=PYu6j@XKtLv`OD*TwBovxr^^R7vJQBd|aXT!c~{&f!7sQgm7 z4Zdi=ax9w5fs?hQk_}d0n~*C`Aefy^Dc*jGwj|kH3fQ5hzrca4cr`08xw}r`f*qOX zN}Hw9oRl{2DxX~~n1YgCc@|09suFP(HI}NrEOt3z#=c!=jbV)Y!%P)zdvz)FO`pEeMG2W#>u{*@Z+cahnPobqc*hMp5}bu5pnVc zArPWn9*5RGE>*&U@Wg&&m)wUaL(GNhsF}WI+ zy-Vo3JqkUn6&dIY*H>zidsg@AQC_wG973u)_d`&}LBx@HfQ4Ig^bh6NfabbVgYitN zU$Ae^dm4{hX2=soo$D`qXjJ1Bi7t$; zrOls_IoWqRHXU?3SgvGLoQf^Ew84cJfM`P)$~Kv0fVYMUlaH&;yBO7gtS4siVW9I{ zw0Yg=79}XU%r#C9>NMGv?{Zl8jdz51*kmFed`&zRf4tWobDezH{;kY4*f!gm&0)&Z z0VbbTz`%$Vl6}^|Qr4+xnq*u2yb>7}W6AHX`2Jf}$JGfm8&RW&%i?GS$j8aqgN~ae zmBd>%lF_>)1`&>)m6PuoP|>_fgxFstZ|c$#L2By$%LaQb4nxvM3PRb3dd_Uq`72dlKtlL@N>7p?8QJn(v;2em}-Q5jQTU&Nf| ze~^sPoR=E_JZ=vdR{tlvDGZb@t)rX; zpvn1i&K6n{yM?+w@4xT5v=Qg)sy<6$B&`|K5CB~ur4_CdK(D&peaV7lM zAD%#XKD=s&^3@BqHnsirvU<8yEwkjuDW}q&oXPYi?1|){OL|qw?bljM>$3N7HgM3V z@)E^p`6dRlMR_*Hzi-2AOx0a#N2#STsG_M$sfRmv-VGjv>dnhED#`0im<7gsP&yE5 z#WtQ@uhCs7tzjnHpaSUU1E-i7E6EQy{_ebIsZ0q=ixnUl(VE5V0)<*C$6~bvAisfm z&NzB8p;2Bs)=207to=Xlq4i+Qt-8u$5i+a);c`&?Wo4kucx<|B?wDg+j`Ilcu9mfJvZ}pt%(R=Cm~?KYsQsmbmBCvvi~=X?kMJbeSAiCN*p# z<}{H(*)X5LBqS}aQdXXnW&PJ08d&;vz6>oxV!vTAjDlbxo_?#{YkTT&VEITDIb7pV z)mkl zjFM-ZcFW0muJpCPTYwVFCM3;^>YgO*SwU=*yBIA3NgyW)aD1s9OkDM>y3;&ru}-SvKH3in7bD?;v^RniBA* z8zy}FfYUaTNW(@W?IN4u3D^gfXtz8Hp^Z-JZX-x z$2?u9(6SCXVzQ}S0yN>45*4;(jMb8gF$lpjJhYp9Si?@a=@&D)`ojcSPDg^E-jG7R{NF%0Ee|I(tZX#}S< z2ulu4L<1(X2{E#J9crV&43{Z<(55B1%RfUl=Tf({lGs4CGFy9AA4Q->6OaXjwOh0W z+A_W~er;~ml}+M5$x?-2FK}!jgzX&O3Y=V=)hLA!t}Gt9HNyFJG3_;&dEGr*f@3uP zBM5WIF9cIr=*wy*zyg11FdTSj-|XpxOkC6aO$)m2N9T7zot=q?y{;=P`w4# ziw$6a+Z^;Wc6iQN#zHaTo|LXG0q7TClpu%2Lkk*GW6DciYXsN2kmJu%?-Ga% zDy2Et=*&599V7eyP|>BpSjj!ZR;JZb1Q$Q;sl}kn3N!Pphf&VRn(9DR5)-6;5+${{ z2~}NOF2@^kTGvita+gR?B1P&@*U6Lz2upA2IW-~Fmzr!Yf|n(w_;l-*ormS33M0>( z8y%3*-(Qj8oT!|F?mS)MuAUDXLy_>Lh1th^PHG4M}x=o3o-8qB3GRXn(7y_Uu z>Zi*)n=qHs^Gfxj!=JkU6O8XU^n#*w#Q~^evhA)bhW2c0GQSVlRJRAc$Or*-uI&f(-j}GyidCD@OFM$Ss3l*<>^7Dhs6D`te_YhHiA>ON#PFIQo;6WY2NbKBR@!_ zy6dUr68I++KjC%w3bJ{2yQdrhB%5_|z6rPE5WYA*WdJ~I0=%+NOajv?HAt0y!)3YD z@YB7lsUl*zg~w8wb(W%aKCpQCRy#hJb!~r1idshi8m({_-tih`RvKAeTu|_Jl%LRU z^x)`nM#kp5s^r|9jKgzUZs~G)=44~m7~eV%gp35MLhV)?B@l#<>2Ao4AvYK)v>gN& zjr^fpRK?^N6KgV8{gBFuEIo>nn?O;tQa2C>OE7P3#%>E@x0oe8`JA!}HLY9D9$^#J z7a-4IH;Xh9j|@JcaUQ~(^2!2i?w%DIifu{z_8JEq$JjlI&}ER`N0X-^m)|_!XcJzm zMuwH6E}lW`f*?BQa&QVZ-sPqsM4p?i;H_!qBg<^ZU;(sGEo0(5*W71dCAq_f8YCMI z3YyyWiNp!-`t(b&@DgoVC;}ewf-&73q-8)^APU3ItRrrj%>1jfSL0N(FoYOCq$l|ZLE#aAm z)t5Y$>%FG^{b$2HOa(wZKljfqg_w~)-3G*=RnQzyWg?gFz#O=a|Lz++OG!T0! zgO@Z&;uM%I7dXA|5RFTozxJ<5=UGpT`;kfD8Lvrjxa}APfv>>zcrF26vJzZnXGR`w zF9J%JUJO-HU3}uigo%osD#W493Opu~2-QIf>BXW*ncdpaWt1p&dSM07QDL)6c}75I>lwh8M&S(6 zh|ZHxtJ8t%@Erl75bwt11EX;^{<_^k+Flv}2WxjYhukA}gZzBqcQh5_x}0PEMUSIH z-9Zp^*BxVGUa86_jv$Y4;*sXYprCN|kx>r5h z65oBqzrMu3dXan^l6*T7rDbSjg#BbdgpghKitpniGb;SxjC_3Fk>4R}+-26RVT0dc zGu&y$%YPk?I1|L)iK8KB)8%-XED-m_hreaHE$xdxpY1z;b99Pg9j zfQjy14Xt_Ou6-(if8&VoRHqpm?Tym)GR16Tyt1lGyZlqRwXBwtsYZ6>?sf3+e)|8a z5|-~%Fmdnz0G4zB07U;?Rl>sA%3S|HD-uT<7Y-XEsNc1verZEVNZMAdPfk23Tpe6i zBnt5qrWzY_0*47gpg@yAv!n;(hf+ua8AY;=mJ3YgAdvvYMY0PelItF1?tVRix&*be zptpdlphuIBs$5D_YNj)Qb*3|}J&rDb?SCkuBrZ#x7tN5SSW5mnNR%-)e?dztA70@~ zE69p67kK7;T<6B0mgI)aNRB%y)!$ht+u|w|*IVi&qb1}!z3mTEC#KTte8)FAlB9Qv zXp9n(B#uo%juVJvY*g^*Y_DcTlqw`s&A~V*XY6Q2)z&7Vr7W|h7InxKV=)?w8uLNZ zl*NaVbc>c*oiFMcQg(Q*uxO4fot=dkTx)PEDR*tIpV(pyg|uQIi&N@b#&i*;bqrdIojiWb`2 zJG5{BT1F3|z=dPp1KMP%u1d zz=##M03rptFQM%mhx?yPqW#& zLxmWe$GJW_SFBlHb-4?D0g>Gl>SixbpT zu0LJL?)d)>neC2YvN(ZFr*#5E_S>b~KYcltfJcjdr0&-AKstLlMsT>Apy+-BeWE+8d!1+`oJ zeX%iG4Wq~hX+y^?6BEO;rfi{PTk0vnT)>w99AFpjG)lrh85kCf3}vSKOHAU6Ij|`3 z`(?0ReXyl$iDq;_bVts1LTmwwz+F4Wd4MBuNm!zvosn^A2(L_dxhUXi3oL<(XwTL~ zIynpSW1lq$GGq5xzqZT)Syj!6VO`OolkYdw(_!DOh0Tm`SvP@WS5JstG|UlMH9=t0 z4v0n8&9z*n&Gppheki^7kaPe)kfp2nc=c&gnp5pAiFmz*s$H3lD`Rm*!}T=yFVky zg}Xl?u!VCtA`pjbK@PQCgS+1+pogn-_KOch+XVmzzvB+u5&VSPcSYz8e8nBQvAbFT z71y#G9LCbJGNt3#3qbyb9&kv$w%x!;*3R8PXudYyY4R5C7h#FTO6Z!lRR=A2q)q#N zT=Y$Q|48WTw%r)$tWD-=@;ZC}XlN@fy#pX@^i8{d*k4eXp>x{{K>y$1+a;)bM)<;e z_(bQG<;P9FI}Lfu0SHIlw+F)U+Sd;b$G897ZapA=$!9SFuIjZI1wZ||8wIcBwYv`t zkMH)XRteAJwY%roybp|wpZ3}}1Wv|hK`PJ%xDPw~0JJFMw~e0X0`%Aez=~qmO%G54 z=%MS^0X6#pbju}h1O3Mq&<)S;1keE~4;+kR8g$q9mhA+Y6(J7{Xi>lq3*rwjpap>d zEJ&F@052oJ5AY%P{{&u+V=tl?KnHw+m;XJzEA+dzFA#v|KY;JZD*zhMf&_pZ$d2g$ z!lDEKAIPro{{fhs885iLa6k@50CKS1A%H9lKV)D#13xTmKX~9d~zEHqd2Dj)3 zNI97Qf8FEI?-T$&*zO>}D~#XFF7z}xgF~DnIOol2H~_}XV><2=!w=Rxyn@yGP+ zo0M#<%1bV|vW|oXgW?U*5}o3W&%NJ^<1-TO7GmykPfTfN;1duqI0naV3$o)!{p=jw3eFJlW18&{8Bp2-^U4tW#h`gOHe$*-O9v{NBqwC$?}<~bDUig4Q5=27x&Em;0mnSiP^ICjfVR!r zCQ@A5ZVpNd;K~3F-LDg16Lgk0=}Q~omdFbpSW_{dP@U_DS%TuVsl( z!yLc5DOP2DRLi?5g3s#(8Oad#%ZU0&qd0c?6c`x)?#;}e4{AtUi;v-1p@VNdId=1P zwZ*+j?Nz=CZ{IGU0ld>QKs|U{*I^+1zdqsE``Og4SkBzLn6ImDfP0qUp?M<@ zmh#pp$Z#AL<;oH>SPbLpky+WA(bpI;WUpwvWzqLe3G0`y{jsZC%V+0;;M_(@S*|~m zTc@7x+N#Q51X@RxbkGkB+UjXqG3AV!k<%oBiER0b{mm3mW>`@CqVbYm zd{iT99Vl{p2e%7_>@*V-bw;^`VX)?&>%OoHxn)hTeF}s@l(-*X{bOV4KkUfy(wlmY zU~VQ-*>00^5E6>(xR8QkD98F#3gxw`>k=ttm8HM>9O+pGR2j)~Y72>o`j#i7QoKt| z8eBXuN?B&kVw}pnQp(vS!qSx?kW!B!6*psN5e?;r|B5s`_zT6DjMijBJyg z9f~ukS0+V0!25M;3p(P)T>m<%1^qbpPunJ|P!%_8^-o{bl@O^;aY`K}nTFfzt1hz? z{KlQSZxkr9hFXhB!ChF}+IlvEDpGjyguC3DgO`>|Es-uO!=Ijn&0!UoPFfbFHY7Zd7@JSL2o z8{DVS6%4{?)yW{EbJSi~u(dR05OV0oX;xv!ug~izH@$+j6DTEOL{ud=VR7=zRQwH( zU$4Y5N3-Z?PGs#3=EB#GapGA+OQN~7+f0IQWd`;fwt|SzGb`a95PD_GpwtiLm@J=2 ztvRISZ_8@cdLy)*gH_a=z9OvGW*ld58_#W25}qj9gPR&QymY{x3drY?evNe3-n|YW z?X$jXQy1=22l}&&;FQvq;H^Dg#{+?D?}!@)t)3*yc^OxZ_gxF>UlT7%VrBGCRn_1y zZQ>jN2Nrr1-v}%udL7s4r3kp0LI}@^I86m#yTCJhB}=C zl$dlf{?{({{8^ z$YO;in;9?XW6%lR8c)!ysIH~wMi3;wKpI$Nqc{jP$AoQcT~C$=n>i+l@legloY`gSgCEvhCK-0o=pxd(EO#v! znO%Xh0Odcf9X@@YZjx3+IjJ2V&Fk|C(DJv<1S7_=bQx+xt%34QWACNej)p5%7aGYq z($Y!OpmYOCV=HC$q9w?rb=b9&HpH0sva~8}RlyJHnS5U!dLZJr+)z9y)-s(HA4slv z+~E3(0x{GTj|!EkR$@MY!JqOMX+EnFLIerZDR9k zk(MDifkE2I8Wj}dFb((EObSVlX&Wf= z@)>T|=9J@O43?KDIE|MixE_RXor zwLSoV4J^=qf1bC}cXYJ<&-46nw`T3-KYOy*8r}Lm^fkYY*)?%4TFBE-`&fTQy(gIU zdq;m!6XheFE^D%OizV~N&*$(knT9l7k%nc3N>-rD#)}VwyM#$;BO|Tv>znUuRozqH zgM;;UAGn{z4tVxpU#@0lh{3b2SC?mRg!G@6SKnTweBa~VSNWcwds{Dm`UgX&J^bED zr6bdK+mqvs*;)RIo_zeR^c&Ql+t?=tJ~H3#Dc{d~-4$P-xA)AQ&&S-J@8gS`o-BOb zo1;>J zB^Kpa-Jb)vw3+Mk4_+s{kY!~~l#uU3Q+C3%FP}-etc-k}*J8Poe0K^D_(f%J$I{&r zSN3eJi`QByub9{ergj($-kk=EjPYPS2A|2keC*JiJCDAa*L?Ff^$Eg{jeZVu4%{u9_YCu zMDM`cb*Q8_qg}zUqUyW9!mItN`-rD(8h-jsaGzKiTF$8F5gPYY5owG2J%ZP#rrJFx zaBz(qVqDSK+s=iRx0q@^eIV}R2)}}JE3KQ9_p2?FI8_c@Eh?)YY#-O~DzlJyTTe-c z=a7T$i-+aiRC@jX*nixoYHR15BdGYK4|Z5put+#3h(+sd%urqKYODhm*h|H3P~1KQ zg>_jU3qyP=^LNZo$zjiqAGeeZzuHb+U%{x;{j8l`I&d|_y(QVzFFx2o69V%1Pwc*r zExsTI!xH2Bl$=%7%E`&L%?M=FW4Zq$7jRTeG)9d*-|r5WJGyd2lD^ zFl!;$WdqS-Rl~cL;w78&02h@e{ygN^G~u*>Xm9LkHSaz6DkR7-Z(Yj+SdhnhQgo8S+?{5Kn;6#9gEnhkV}4Ga*NGpjaYtDD=z(sEtVN|2$usI0Lbo@2QABLhUzvEoX5zCtooSAurr8{)l4^acl z(l#XxMq~ZbL)fwAx8h8gIAKxQLoAAzz!c*R^Zm8>0Y_|z2R|-Y^IeA- zlPrPVyNDCrwq{B7(2SL0lc?+HIh&k$p{1ek&j|&u!=VoE-PbW$mm8m#geSf25ZU}B z-v|z-b`4>0205Saoq8;M3>iB=zYH;gt$pS456uXyN?}KRK*a#j5wliV((}hwp*Btv(81SjZ!*@KoXwG zV8a?F${esM*qux5??y-A9g`OnVlOz3(nnH21;_{j+FS`5=N^2ufo42(fo*2lGw(51 z>aO9$?DO)Bez`uB9R=P)vCAf(bEd~y4ybbkGxQq^OpNIslRf!F&|UiX4O`zkZGM|X zYqg#4*`?Ngk`n#2c(s?+)dSb8kV4+}TlV!;n%BDD%#Xpojx=MiJYK>~no+{m6mN`? z{@A59^rjXnoSS^h4o)n9U>q(Se0*+X9mw9ebBCeAasmfj_B0yIe;m_5BMwA$Rk&XMDqc-kx%gV34Mt9iGBq8yBk{KTFgzG&gh0N3oxGUX#mD<}%~dD5Tn4 z`FeO#?)X1@RFUMR#aUHM<*B3TR{`K)Au~KGOi2n?iucbcF2*Hkts&XyP}7;Mm)t52 z_Ad1Gy`^p`4V3cu#4;fz`B}|*Ev%yz2ZemK-RL;cp(fV! zhEl*zvZzH@I%X^0Ptl4EN#X-thxO|Ux^pAlxNaCv(LrK045_4irM*mwaF7jhZl-1UorhfJZaISQk@=et-046Ox^t=Hyr0(cDJ{Zto3lT zkSwE1OI=*a_96XfEe)+C)GGn~18U)lv*;Ti<3nQMz@oAEM@Mp2_PdZDUV6vtP3bWf zi1dg@rmLa7%U8T@=4BDcM1UtzX&cIu2P5e9X4kYLWBR4kB0n@MpZKLy?cYL^xgU9Ps9$!;ZbJHC4D`&3mN|_>9ztm&xkM zO)c_xRWonLHx@*NqYaWemNFs9qwD)bYyY?L=Uw+xxe2egz0yYJz2yMkPYbD^*QKOy z?-1i|ruN*qj=yt$85{Z-%Hpg>Zgy{ct?(HJ&~C~0jmL(a&j-=j##+3*Vj5!->7qX@ zaRkQnE9ee@k$0-As^kGF@#wBwBvP)jKwypUXLZkdv#m7^HR1!+7qxE^5;#;?kR?ww z?0On{n$Tpt>wf0Yd-QLPd|wE(Z7eG!-#oCcr|034%#C*`n6aMk(YV`P>mMdXN6hAw zFyZ^%#ynN*5*dpXXBOJ=rQBJa{te$WT<&Q$*>k=6czwN&&hI2bo&qE;Mdm0oPpT2< z+#jfERNaFk>b`2PY^Tf_e+-Aq9_WaV&k~V52KxJ!gei#GFO>gJH_$o2IUwvM$H|XE-Y%!XcNtkcOog_BvWr?G*uacwC7)8wAB`un`B z8#-HKJ>Mp6yyZha@bmdDwL41e-wu28zEMTv9Np8y`j$eQh_G`-?((#T`%E!U^X@pmuoLID~IcuF-Psisk$t`8?~% z`CuFtn9IU(x9oxYveJ4PH|>|EJhZQms5A}D0fsV?#ie^0xZ!U~#I9)Sj1h}BL+~67 zdNYG&x30Ocs8)HqGOYj#J7=1rYg887d8(|H=2zDKbna#sw0-?A{K?R={M^2d&^qinB&p8*7{CY{1NCdk^ zc5MV~XS)Wfrt;(3pbZHUcsBX0)lyemP#BK3VJ6m3T>o1i4)$ zk{=ZmuMnOWovMgp}@#24@i6 z^a`D0}6G+fte8h+*8qmQA%zQ+V1%%pUeuiKlf5r)+ib z@8avY$rRscHj7fq)1fawy3vIT8gfK$9KhKDu8(2l4)PIrX{K_Ktd~>Njm%*wueH5z zuvefJd8TQ~1#Y_gH7>43=dbl*hJn5IU2>Dq&X^Cxb6NFRI77QGG=`1l#KsAg#VbTR zB@ecmFE;!bu*odNtvw51$y!A$ha{i_##l+yJ@HD_*4Oc$6tuhHK3g{2`9QRlweEpt zp>$I<0~teka{6o`9s|m_mn1DLYEr!a19?D(zl-sh5a5_SY@bZ5Bp|jKljLQf$QcH@ zjVO&lA&V^`AF9OG?GT;?2~1)h^~!tx7IBxUc-uPvXW`^KQK~q-=Fu7`?|6 zIJlBN!Q-pC8m9p(Sr~lOw&%fCsKgFn4Hy~FvVPhOE06l@YR6?MOSYU%G&)n~i?LNW z4mQE{A0}zLhHW%@1Z>3jV8Y_ncglahl8kQm+aG;WoL~2IqDNBk|LfIf6I*7^*u;nX zxbKM>uI6EkF0Hz$wp3SVno8desFD3jPN1=wGB_KWLeXFErC9jH{gzQ>0o%mH2J^w1 zQw_;av`ldM9!yKTS5x5{oR=I*|DkqfP^^` z*@L8Ld9JS~Lwa%Nd-X9SWm;O|fiYZfRu8^2X}cF6piwWpo*5bZHU_p&ygxRUKZc`C zD?<&}=9pEin)&-maF3*Iz?CgZovS(T1QCws+x3l;jes7r7qEi#B2S%zBQrEX@7&10 zg2jET*#Xs%R1yXmQdDP+nGqxJyN+~u_|}Osr|ZadZn9dS`_NxwODrdjm*qy-B$Rdu zd|1xf4S$!2PQ%(RmXgd>7bpKRDCmG&R=RmHT3IDaOfah=Ej6x8b2Q-4J`|&pPT%!5 zd9~%7d>DTJ```WH2a`m9-_?QD8gwgHPgU17zq2UxQeANr&kg?Wd-f~#Hh!(-rBPrT z|L$qddHn3Ap&-Kh2Z9+4@>+PBG0MwkAs)HzrHGZcCPW)s!rA+>#eEBoHr-Y&6i{r=K`%tkzmD!1155p&=GJz}} zK#raL#en)emR(Q?ak;>>$-g-wKn2LG2|1$ym$PkJ$u^TvbX~3UFPf86KHsdQy;|k0 zWVF!tW`P*SQ)H9>yHkv4FvY)Pqn4xxLQ9db;>FaBA7U9-EN<@vWN(4e&1itbpcR-k zPuVO7V;uwuMZGyU+KOGK*s58r5Tp;VvBDP8IcRq&WhE~Sda)z{RumiKlToA5%woY_e~8$^ z3SM0orlHlh-uxOs{}pocu2#&4@d&(~Rr}E=8WfxZ;#@dB>?1;7ZM<5C`C zm4_rcRZ!oML~S~wX{2=WRKag^b+(yYyMN*`6Ihv3g$*N|LdATtvEN6>TE)ubWP}$z z$=%FVOc%>JB~NHPvZR~poqYexcJ=X<;k>t;U<&2+5{W!Z6>YrHhIr86Zz<; zmOS(=roQRdNko^&Q$7iO_qeEg6(zV*7Y<>&n3ucAs&zoMMM{0SL`_Yr8ZS|ye6LWx zS18|0sJqS0pwaJ`R)5|^EeBLze6Radq8d)z<33z%vVM@btpZ2;zK;UcsN~*Hjy_GH zixAnLJr0tMI=EktDHV5IDBh2rq`sSf))u=lCePn+`YsofU<=}@Lcfok?!W{D?C~F? z4!yG7TVe2aZ^m4dn83=cMGYtI;fnNfUdohSGdHR;dmA%)J>4i{MP0e+I+-Jasljeqrj}58uj&3#dJdvW zuT0cZpu*LgV1VuZlriFg5$uITe>Qk*IRnyP8=}9 zy5xvYC&cCCd)4$i<|mt4dXK%dcRI{1!G_0iVmN7s)0rxsEYAQ1){he71C`l;gq-(@ zPu*ge)8vOsG7P`YILGiMu8nAK}?J4^odpnRZoh|;}sVVv1txHcZENt zuWjs3H+rhy&kL%MPmZ)2rA9qjG5r)+_R+^IKTkUJP4TwG9o#m=a(#u{GDTCLQYKeg zZx9kgPs_=rFv5QmsNYV_by9QLMxvc&Pc_BA3huiuWswP}RJl90@MrA-$w79dxsplc zD6`|rvs{cEG6yI#rGC9~+RI`m!2#2+rCNV}Eo5xhc2r1VD~Fc|?6}44rwU%?LaT55 zu~c;BT_ax57P1K*Ojkrx&)r3-*0(B#Bb236Ua~j$IFJ+1&pj8bV!e0ieAT3`oI^7& zn#J{0?2`z|&B){J6xu2`r|l>-Sqmj;e|0M29cPuqViJw;3GKSdKzniMF5D)u#T!Lh zcczS1{ydSL6r@&_gqjI!V^J>-|8V+B97Z@A_@yAvA;PDN%a1cx=3a5$1oQLdM-&3} zgI;0N^O#DBPY;XmcY zR|O+Qndr=;lV?VjYp@n5QpCS6F5gW?#s~>`=1h-o#KxmT-Z(dT)`c5RcR$|_WeQKt z6#H6kQG__rKBZ!iKJD>N=#{+5h|ljX3B3BgyApNT;nk5$fmzrO>M4a|1f}J74{)Yw%I_3lU!rFC@N=dZGY{SB)yr3mHg>}a))m_5<)sxOIzui2)l=yh{F8AT`>LMT2%P#Wq^2>ajUyJ1STT9~QgZ%96yRdzl z^slrsmFHAa(|OqoI3YgR=Ry4y&C`Atrb8RxBU*G-J0m-aM!%awiDhh6q8b$9wK{xn zk3@Lk0)+=fRCz!p0GSpv4@v^?9Y;pE@Tz&0fUR?kBl7ICC{;+cnWBcL`?aMp#uNtv`)r$SgTfAz={?%*thrOElSI;mA~lWpy!KBr(aIyYp(4H9@{LKZ>OdTm~|^xbHjbui!TiHE`XFo*}} zghlQ)x~=-_LTaZ;&zx1rx+62z{it7?@`Bra>y5q2OSpBdF;790P})zF`%d7xhW_GR zQ|{w)u;Rg6k47`>XpXw4#K!kIVwxp7Ur)I+d*@qP*m9(Y zpQy8SwBigjJUFU#`=lp69Kxg8OtKeee~+pb$n*xuWdI+i>Mn&fA-HER;XZEFk}FAj zx|g#G(!{N#J-J}{DQGh~Tx0o^M@c5-XD-p9t==&?^q!xSN|c^fi$y2M!eb_q4-QO! z$tV@I%T(z@>%W=M;fLE~tY=B&G6qO0)h!&2oMLFA&vQO4D*>(TZz9H-ME9VHEss%? zEft@XrR?cli{3k*1T ztxVw$3~3&8==reX#xPnb0b_nt zZxZ%qV54KwG+o6M1ePoIDijD1C>vbsYS=`u?}!7b7$=K$Llf>t1Jo@gFfM;9Y}5-_ zDbR}(%rzaW86u)rZn(_%3aL1=tDVC}G|hV+CBH2OJTyT&d2)R6!=6ohto^+*XrsFc$y>)Z0 zI<$-?BR-9d^$eJcKEWy`eZ_GHmGDbDRMwrf*=;4ja+qMlj=f_hDA~8C3a{=52n9(v z2bekQ({?^|n7HmqL1y+u^h0|ICf?=sVm5rP%&%dkYg#G0!E#MJfg>EgtgtmlMoMWP zD{h>NtyHf1B{VK`7C{Ruk21ZX2LWXyO*|%rTw#s?xN2hEspKD@i=WrBoS(i8uokHh zposK&d&xF0?l<}91(XQtAQ;aX9zhaBpi0AdeM7B9P>g=uK)N_iB-F$w`yqNX1yQ-UT$JEZ`E4rA% z_9d71b?%Kz?+vLr#mzgh_y83 zUG+8Mxu>^h0Y0;3g+fmr!my3TB9{Kl%T%=DYKehkr`2hzQ1*Lv6>lO-AE&ebsJCIB z9IPgj^vF6n7E@EDr)=%*!6@(A^(})rmV-HHd6_d&qaV{2r8-Zl(=*&o`q}*Tyr_>s zXnkv~OS5I%fG?wJA3|@rtt+8MFJ#obuGM7;#sK^ko^o&C1b<)jL{4yUSKa&sLvH33 zeps!99xE$6Rn|3V#xmhq&z%Afivcgw)k4GcgAIQ*;P<)E7e4CIy649w{H$PW00et@ zg~GrvN|Xy8EBaZ(nsc?7Bh}J97PX%f&>negrC=YG=(}RwTkIIT=FsCAv=My4upIDM zPt1g|5Z>3rC(Yd8JVC&IxK+LC5_fhSsOtP8(Hw(~GFMk4XoFED84vYnpO29QI|j}g z_YBxnBE`%FV#uG2c^@WXHQ=etW(0MPtd~iUncmW~TM*c?mQ#B^Rqc4I&g_4tsg8CM z%aG5Z!bsxjQ*{AL=Z-r|TD}qq*%)?Icy#BX)pz7giYHYx6|&X5!~u?6Zg)`!=TSn8 z;k;Sts$A1u>9KVmR=WMfbJyj18@d>R2ymX6spMg96D%USjF^3bfGoi3i_=uMhZ2-rD8U3#f`WA z`EJ~W`lI2d8F5lQW`*L>S_G{Z4ri{IkT0f!%3xEdwrvTlq9X+@-HIA=4bM;o`PK`ozzA4;>n11OLY&O^3 zYL>ccY2qF`1vb@VxO6m&-E8U(*+6o}_XmIafZW_G6L=;fy~4h9EdJPPg4EQ@1(rEs zXsf6(`7W+8pF!?ke!JVS=)YlbxHS0zmoVQJc006~mP?WM`hjKBNDK7O15{ZRHk$q< z1N0FWym`-#6e*Zk1U5fY;V#@NrA`q!vssJaebCRz?{0lhPEiZT-wCuelE{VH`_qlq zX+61#z}YvR=BFxz5~oW`@7T_>(hkcQ(o%S*rFGemLX%B|>xyPw_C4HbA%%ymxn$F2 zXff#0=VAdzz9;p4@%7uOzBy~c>Mu?CZ<|+t z`*v{Sx0iQ*JGk}Rt9uVU|F^1nFdj`{NuetAjFnZp=tz$h>ux>g-kfvqshlsQ`$$4G zkA2W$hw9#rJhzoK8$?n@W{Dgp^Q|NKQJ5sQ-HqNM3(o-uN@0^s^CzWXM_dB?4BTC% zSemuHO9DMtW+;v}37l{**2*OGi2;`+XFs>@Q(bRi2`Pj*#tCY1M9F7~=As70^M`=Cl zty+RhK>c%(zgNDC2_GK{Tx(Dm$0Iq9Hr$L=gb&KULGZDH%1eSSz2L@9d(Z0a?vCB# zF?s@{7oRA7n@%w{F>!=PKui|}x>uYi#wdl!L|1*BmwASaX^Dy1ku2smOh1qQ!?hNP zzKIEbr)9_Z8+T4cq6%fESb%F1N8_Q?FXhh0NtR+LR8ZGD633J*ujIe|qN_uejZOSA zOlbOFxl(nRC*QLHglRLy&r>eBPNzrfA=m9wE;Vv0&4n_XkE;mNg>uzhL^$#0WO|YA z=;CX|l!|Ls`41E*UYFSE37!WUt~ttB$X$ExypRC)(WMHJ zY?K0G|MO=*vWq6xl1MN@qHonO28BRdpn%=!O#l6A;6^m$lWAcx%$c8lYjC_ELpMsy z_{J`)2LiOuUr5LJp-fpXkfEzbox+QEk^%x|(=wwmyDzMt?gSad zaxOk;iO+EA)Zl#b5Ai?0{x{Cp_T~3q{#;$dumAYv@1K8kI-%R?NvfngvxT2d87pBp ztn?@=;*(+r8H3rv^bF_ff{02Ll*?HfYc8S6UFLM@B@%rb^1VyW&gd49)5TWcrS7;9 zWHCH?w^(Sh^lHqO-;CMfi!oV#FJ?=xwWioB-FRJq58jLS#aqKzpM6b8B#!9?UYeyu z>L9iq$&FItq)kes3T#W<JMAk?-LxQ8qe61YO5*vfikM8hHwNz;JC{XH6 zku81H$QF82Bul+1vZ^;pN~+7pi8C(0C)<~bj^CuB+kD$PnhQ-a&ek#I(Yz7uuuLiG zvtIHdH_e3l_MW4qyF~Ny1kBp*x)AR!MHlq~c6&zFOjaBT?L!gq4gx$Ps;pMeywoQ% zF?FG&V8H}%l1%m~_dkUHMoy(@Utx1Dikw6OieVPoZAW7+&B%3}ea2+Qi4WH1hs|;r zqKZF^xpq&gzBC_Voo=oLJE0hDNL$nLLU7h>O-(%dCCX8MiM9Hx{;MszHl8d>6sR(e zKcz11LQ@gvlnAzf4_n^$#+HbD=|<=WEaSFje>DzUXyLBQ*%+@MT|136tKvh?=0#}R zM9UtpOn-3AB~wC0UdFWUJgl)tJee5xX>jU`?D1d_T(NSi=W6ruX5P9TC)*w7sxAIC z*V904HK}A=fzLHOs5xXTXo;vz;$)mSv9A=kI6>)UHKX}&;hm611wIOJ&=0MU2EjL^ zf$6Wk{a_+suFX3>Zk+84$gnT0y+`^)wycvz#0K8Hm%m*2R{>rVo{NMf!H;h@P$;z= zzpb{OJh{SKQgj);#@vABQ<5yGq_c^8#mFD?*RIr{(-yiFGYrAN~ zRnT^VRac4l#X*>Jow5~aD_dkZ8`wv#yN4!Vy{`s9g$wa_g$vfZyDM zHRXI*B~HCd%bp|X-M`a=3oslTuMCa5Ve&sKwIT6mK4Wy&HPh*r@!T1Yro@dM-n=v3 z?MYgXOD#K~pG2Zq^Nyj521)A`?lzu`)(aAMb5#S2)V858L9QKWBs_q+ZW3Jd7XD`1QK;jld z4{i`pcBm&WQNPN|7(O(&R>JM(8n`jgCvjx7*~7#hyhGJ|SYamb}S9du$~Mig&P!7VOM- z*2mZ^dpgG*<{A|nUD4pTW3|<9N22L(M=ATCpy0S7nE7a&(j{u)np(6a=;%?|UK?N0 zyc6dot8n^QVI7yWk{Ft_;u12oh7?%>IxJ{oFwAEqba2qhpx5GxV*pUZyLlc|NS9FY zsRxo!R9AiYZoytE*(Yt>>JwsyGQW-#M5ywOSaUwbf==cT*^>d7!XH?oQbh&5l414ahnp#T#&?IMNjC5eoC$Wq#Z1Wb=@yePXf8GIjOwd8ZNa(Q`D@A zE^iEk1|?jQ+*Q4GG09+t+En7^>T08-1sASCFo>IlP-9wE4z$|-Wp@2dtko;X#vT5p zN%p~VD`Z$_C1!N>AI)%M+IMX(LyNL+X{3OiD8ex`&5)1Aj_pjZBw#&M1bRWv6VN;-R`Y`IkY#I7w(X z8+X+nx)E2Um=ny7v+Qds0R~|n>M5^mEE^tcwp5sGXbe5fn_-FkXL|6nyw+%UtWpEw)Uty7%z zk;4ZNvYx80Ykp@@XvJJ{6wflzN54GMZT0vRrUv)sCmIv8=7!!q8!*rRjy&gHsypUc z!cyE!ed+S`u(S=^T-f6fH{F(;Y-R^O&{j3=4e`}w=(V5^JTq(GS3{QLw?8GwzJ+YG z{KWm2{?k_#r_^~kx+qJ~5=_FKD}B4?ZtbAUJlASWKAXvew?q2fH#E$DcGFN0am9Xg zp&1wQP48J&cPAHzGcL?~v0_s(Rki$~9{S7&FgNjV;Hpr~46`uxF?ns2Wk&N@rjOASW@nrj&y&-jOvQ19kD zB!S(tzD+43O9Gf$I^#G8vk$oW+=eJ3CKVq@1S@uv7Vu@5k;8-XoLwxZLB-amNsc(nV z#htec$@n`t-aa(S7>(lJ7cOuhVlpmr<8v-;gdgSt(ebqS!faDIUP+??vv7fR@_p)k zOTX0tEsGbQ`tb9-UiOdIqn8WD%Srp?O#G_nbiw$R1eUdNAq&blbEYoN#3xkoDE3kl z&%|CT;(O^MJv@^;R>SwANBZ(!?pO)mOCIasd&xr;Tm)Ze;4Je}0T-cT{X2@3)bGe1 zY2PeiD&I^Lb#EbGj=?BZYN1ak-m6$q?_LEC`}HTa?oG6)bU!6*oqOd=%JRw|P6{s+ zZkBkhZ?n{A)a{+zr?hR6e>HdRgvz@1PUu+G-U%LP+L=^o&d!8NdiGA}U~*@nl9tWn zp_0u4N0UrZ#TN2F!xr9Fup@P(U$exadc6r8Y1blDRIZu*lx{78pHQt??ohK%0tLmI z*_V3tQ|L&o7O5kxngtG(Y8E)wsafJsr51rtXw)qCNrie9E$Y)r;80&nGM~_iUMtf{ z=##qiCjA*zItdqb(!?Js(tT-KI39iMvnH;&g{CK7YhqTT7^@KhnW#DWF^BMof5~uPTBnvB(pq&Wj!_ zkt}Quh19Jle2eyw6HvF%rRj*c4^ON%U-yrRFl(c}v8* z23b|a6ayKdQcP^@3#J;eYF5sZ+Sg+y0g)J|;q08z8&cw>W#kAP{b9`4V9;z0otz+M zh`=y|gxa4quDb=;?7XFiw@MywgfZi9oc$FHv~Nyfs5nzYes`Lubj+Mg^C#WM;g{*f zT6SP9_^#sSP}j>uyR0u8 z1>jfr>-y?VIIck$wYbs)Y-v|~s_qtyPpABxjaU<-P?;4B_^o=c@b?L0@7VFeir0$x zSl1An_~sskc#$WgQq>Y=>m?hZ{v^hQ8-p$p@gr2q;t=Vl%`2~GsoCE)$2^biu6oNq zz=7C*{-oR=J}DQfW5I7wv|dXH(mlX)Maz9cWO#WqRtwNxV@OR^-M)~q>)mTP5}eii zXoyMW(U&r@$n=p|+Fe3NeoCSHVq$#Y6N*8c4>lvAuW_P9AZ`9FuT6+I| zF<&nwe8G&~D~V3%5jRSb2|=o}vCG-0X6f`>F~J!o*#XzKdTL-A)AsL( zpxc*;QvYCX;hZEt!waxax|++|CNJs5ziHP?GlHsEu%k$6Xi1T23$*ymJNVKc~`QVX+1v?nmxWllsULXaGXc&m_5El zl(BBXwUB)YE?hK{*43m4kKN)=s&Lh(K=I0jw%&0^*Dkj+jmE-LZ%pCB8(+F?;gO%z z!G+5Y%$3ya2i};%bvM3r*}?-qse=obAGq*Gh7{Fbk;C!xa{s9v@(U4HIN}1PUP#GjvM(tsQh*kmfYAFk4f#ZeQ6)!sv z)}u={r40I3eVo7qCfKc#fd1SrA&jx96kMw~j2aig$Ta~hyGbOp*<{6oabYWbP>8&< z4OjSis@i+?K%rACu!PThE!1U-44afvKKh{#1MCGr#5w~;88pE}UM(GjindXF=*X5I zedWtn#tq9x+^b3H%tT5o{PrS>)mO8x!!I+8#y)zb;vldDK(`(^sUN2D64WYZkq8{m z@GJY1uDWva6J@S}lCT{YVlc*T$1zW-G=|NwHx8z6eKIQlL8J4xbXIAFT233bWqyLSGo$__k#tQuWz2+vR16asidr zPUu#*1WZRO_QP)!YTd6h8XOL=A_nna0*<+lfnd7E$_lGaL?~ppir~;0AAG`dSwXG- z624FRqO@I7nCe#_$=s*sf_U<~ngvceT_dOUQ~{b29-U96oe%ZN__z$mrt6pPjMrvT zwF;x+C&OE5Qcr5IKa1Jw;E{Qo+US!JISXay5-TadlWd}`MPD+~+Wzq5OdMFqwxw|K=J&?#?&vALYi zm2%l&xE|X4c^HHoL=0Vqe7G~Fv=Ihz&%+q@9mdpYGlMEFRY1u2#25oNx| ztv}&|aRZ!!z6$MrT$S6a1cyzSU|h9RfnprBQ2qC;#HylDF^MsZWdXIeDKJC@M=Z4J zb};ZbB?HV4>WktMU0R*t_>!}C7^ikov(Wp2h!^1Ug@OT{R65j^fKJTX0(;}2FIifuW zI^>tkG=i9lOP#xfnlnH8OI{f{TGUl`{wKWf!Mer8Zh+k^klPP>&PA4 z?)Yw{n-%WW!ZaC?-26k>CI4+?pJC8JtMZAmO8vFL4+G5;lO|Ao_Xy+uV1>aiH^If@ z+a0g0lo>Zz&xYrs>vl}pkv5`fq7q0G1njPMBJb4!<59ob_JGkxob%}87c6ZOR|d?y zE9^})T3rjzxRJ%j)Y)cM#;laEk^1Z4Pt$D1_`!4*Zo|FT^H+qG9u;BKc({XDNu>mb zHe^WkXa)@*f$-+XqRk@E`jdA__Kr--CG8+)?y|J82udcJ(;+vNFqno?z($&3_$x0v$PiWjr5B~CTT$DuhJ)WMKj1!nnz zop0ORR|^DYHL;@VDgi?|Fw}1`4^R%{sZT_t>Rc2Bt zdsCqFus+U`r>)Y-m>nIoG)jJ{bnAFB#s+rIHM-cEVG_CG2imwf~86o0yc*v*a1dl#HMp)r=TPuwsiroD5cEm$;XriZk;fVsWpf6DTY+KwTt5hfZ4xVw+5` zgi@fToY)=kqSPPowCD`1#_`S=&nj2ZVPu)(%Z1fTe5pbNq5=|CSNmDVbWl$vGb4qq`{yBTWPth|ylpELSAZD)^4#EbiJ`7wBU zIp|CrIx)jW4UFjY)9UgJPh)LB7zrcr9T9aD#Gs^dWv447jAx?%)8U9V7#cgCjEQqw z^>1QmMAxZ}HHpQy4fYNuRJ|(!*-xGq9N*Kzk*|!xCx;tELXhPzHN&n0+w{6j-j=Ygnhp|;VYf1+5rw?j4U^EpSV$7I z|Hat99eZ?b?KHC|wQuf@wVQJ#i(dk2K{B)x`-fls&tLuNU;gyZ|MK7e{mXy(@;?Oo z^`9Icv3~iJ6gz{m@w$>~?rs>10iTR5b!0U1OwMLpXg78mqt&BOKY+VVghwbZ9CKg6 z9D|;b{5vl`spgeZY>gDIcBk3rJ2-H&l**p4wH9oqWlUFbZ48GJLaBX6tZ~aIEqdqS z1L0^>#X`4Cbmzg6SD~#D9f$q^4@d+=ww#B!4N9C{QpYs!I zu83W0=R~OI-HrM0xD&`5tq)+|uOJJRPEQjNT{QU#s{_g%XI_lU0~;t-l!!29^I1{f z8dqkQMdML;ai}zmF~1w?@BltBXCsMUqO_Gi5orga8JYU#L5-J4>XU&;^uv9bYXb-W zw}=kjXe2qTuwXgLupd?*@XX@y#Q7ojPy}KJ`Ch7p>$cP z4x1y`h1Xq}GFER)Kr{!80`YuYgvq|v;?oGy3efWUYOF)=^zHKTbB3dP)^s^_f zsP&4Ywd@>j z{xY}jV&W6#dLaT$y##&0ecHT+C3>XuZ#CqV^T!A7r4f%&*<2GwKkR56V8&%nv zq&wRT4;q_^BgE!hqsovIs{GFcY=_DQ?MuT5Opq;0O4-$B`k{aK)4Wv2?XnXqw$M|i z+*VJSqgH!L9lZZjsQB7Xkff zrMOoDx70$QuUP_Vc`OX0j4V7}6e)I?$d@}yiZS=Z=m2VH7}pB4T-`_|7MC+#uQW9oChtIyRM3Ms)T00{+v9kg^8WYNc2d&Ti8?CU6 zsm2&ua3_j8Ryt?LCZFr9bcGWl3Rhy&FFd&UMQUZAK*M+(Y{`Cset{15u z=hE$r$3|nq+X+b<4om0gBX^a@s|nLqXPEO6Vc14iYC`kXN*FWn(9d*dd}BU0KE>KS z$|-uKqpt;RpQNIyaa|Eo-AkQGCrCT?Cc=6)>$1Mrj>dbOM(UeWcivc|r);EFLW~40 zl9L?*H6%Ks=Sx&A1oRk@y!9}=d8d(JK+1ly;ed{hdB_pCT!2PbGZX=OP%Id0D-bVA z;rvkc>^e8R>0FHwF%P_fg1#zbEr({tkc_N808D`th6OffE>m?qmQX!WSmSUvBdnL; zelkDIpxG#(%oGvcj@z$=+CWiEos%S(mIbx&reaq)2HlrE3-nk9k5(h)M}pKQV42=s z-cz$h2&c4XeL7yoRKbOuHYr%nR~VPi=XvbGt@Z*1Qr`}4^MIejrK1MEDfI5#n+f{f z{(}c7TMgtED_B1AFr(%y9Vn=7BuZ-zkj6KG%&KPP8cW2%x|={KZ649@qsE}iMesxa zH7gS8sRU&G^xETqO>Xh2mrtx?2FqcrabR_A-1X~}=dNIH5BxnFi0b8n#=JG)`VuLL z?ADC$a$tmR;kIeUIJcrL1kC_vr!ke4q~Hw3Y1?YJBPU2OHbaeX3`e2HUyNx`MJItf z1S>3U^)MEv#I#j^T|j9{lp6LC*4_q*NFM4rlO_6U`f##B74tm(&5+_a1EP5FKr4>1 zk`apAB!X~NvR1kkGuEn=LgQsrQ_-Z02JdD_!HDqqd|G^cBur&bYNqOrTbKsc zB>t@y$tI+!D+k%R#+b|k`V}>~Lf2S-GY^*wcI&yxx4ccR5Yf-Bb`UUi$rCI;#&qibJ@VqEz zZ?DP`hgCU&uqr=MbyyRruKAq?&61A2Uf&}InK?J`4_`G5x*k`MeW4~;u$y((;5B8I#bw^bdcrs{s+!#yg z*+^P>dDqeadBY@>Tsy+|9%|nTNKf0g3SaOI1Y+k31MXo8(N%Uoa`lLRtFM#?Bgf<& z8tHtGehZNZHq-q8z%b1}{B%}L!T66~ZTzR-+W7BnxH2$xJ{#dnbJ1YZeC><#hQKwY zH;8QMJtA9ro5&X4NgL=GhpN0Bl+w5cw{=^vifLZmSh%Z-%SzA%ELl4k)V;(ge_cC= z98_7%-R-Tyj5PR;46B!_^KD)>j`mQkE?ih&srP=*A=HB|y4s>T1i9qAt7e|0bBq)V z+{uYYsZke_P|L^~^IIrgQ(#<@1XPQhRdUj*;AhsVRN3pKinU#eFTs_n#mF|Keq3gq z?(8CAY*-%0zniLB?yKmR2R#Whe2Kx*k4sM%2DDXakpwC zo8b9FgRoCw5cB*@rz z&@M`s77|9VW77?R_Z)zP*q0V^ZC!>|f8k-se@^hTW~ZEczB@BARVGNs`6L2&X9+ep zXd8!39&xb!B~)p230DLA39IoPDcC6`R8{$W!fLybIJIR;xat8B#*Nbk&-TtV@vv1& zXc)mae8ux!R6@;769xtwy;P7~vxu0O0Y5BbiNzZi(Ee!`sJBGFbLEUfA@R5R#3}wB zvB_OxW<6cFc+YAt3!>ru&~5oLkJ0#O6f!N_l7wa3k% z{Y(Ic9dfy@m$As4tLf&@;WOf#s^5>TclZLtS&ha91)N(L9`bG|koaiDa1dkQfXL*o zu;4ufR4*5bJrGo`>WwiUo@VvK=UI9CVIDjv^pnVbXK5>^9f zm23l*BD67hul2<2{Gf4Cm06Du{-kg=+mRUkxTBO>jNI`PC6VG|36GIpi8qJ+aZINv zP`JGv_?eLhe`mF3Jj9|+8wLU}DXv9F9djpPcFo9Joikm~c+d4#G+3&u@yLo-9nUh= zajrc*Ra#FLGcRgXbt&=QJZhb%fY~@gn@=#P5vElU25E(4S)-AZhsBDi7t_(W8;I}2 z{jSi1*~DwCVJocd7F?A~Z-)duY}1uCTh#Yvd-^aKCu33-O`!Eym_0=UB1kgl$5=ze3`fBEhBIDT(d5Mc)gN#gNNq{ zDs6F?GrAW?g=_V*>?Y5vRzUWfCzF@Y>CI^sBqFp33HqU~F38gg=R&8HcpF%TDd zo%j(Y!to7a0dFml3?b&@B~F=~pRDD?8(B#yDJD6d#|Si7Pl4vH@qox0n3*tO7t$w)+6<-RJKQN1UVkk`AV^UiZ~zPnr}`D zC7c{@xYfM_?I`Znb2 z18$3NQmxDM$pz}G%@zJci`zSYX5^3)!F*W~Bs6U6$oC>lV9P{EJpeA1j9FeZf`c_! zJl#~bHAbZyS(|AgsT0jV{A9vEss|J}cOzn$7is{Fg8AjvFnqvtjHAiMJKCgC%Bp3M31?V%V6QU<-Htxn+?x@11L=So;BvQi4CDtTA z4iV0t>V?735=Ob|)H{M|n9Z`i!m3>xY~cLadHa@M=P~Bqr1>!M4Qqvctc;oy`CR84 z7&QLkx-p-){RI?8f9Z$W5F->Nue5ABJXQ4VaLvw{e0Lr5$y#z2nkhwNnndkP5^?QD z5^?Q75^?1^65-0g)YKheaEFN?8Ui9}xQ9nVTYU2M+%WzmNIz7>4LeKDV9CwI7GUcZ z%GHb=3LY)5Mwrz_ozJuKh`+VxJEKPJ8s!w`X`oK@ z&Fgp4m{gPC%5n$td*$0)3(Fb0! z^4=5FniHg#(M*GlgBpQRDaVI+gbMaJANo9b1pxz$yI`(bjNWke4a1K{;0wc+nn5@e zO7UUD^69C5PYwlA=X81r!W{ylFypz3QS*nO09|LZa^>*sXS^C0?og9p(8wZvopfxS zV~j4*x8~cnZQDNmwr$(CZTD%rPuu2c+qP}n)|~&nGs(>4CQk~f54EeZR#LV0UcYBV z2a>X%4$lOQh-9>?OPDKWO_OlxgGiO%sf?^bni1J;f)MEPeaA29Yn;#FN~U}#`mtkY z@$HtAFZaE~hqou+^Gtg=t$PRo1QlJYvD+~8pQoV2Sqsx+qg1|EKa&_Kli0zGDyE1k z+81Y2=yw>zyuI{AC4n1}ePh_6I>=LkAC5Kqu>8UeuUeJo!&`ev4z@ena_jNGHQxxl z)Qp@uMsMyiq%kF(l)+xetmARA0u$E;IZe#zjlizxpOew$vpx_fhI@}%+@B5nmzKB= z6$RIK*ADjscoyq;J2d8GY<3SfSS?sUeiH{7;DOLEBd9;I*sZzE;dx-|Zog`pS~XLa z3`riGnASFx_XI@fU`@3YXOBZ7inYf?A3u8+fR3Nd7tax=AkUTd-2+tRALexO!KAt2 zS1xuQ&%?1|v0CduiRSD_>$iWFnS9N6lhB8=?cZfHM0_c)mgO&fT3exW07CMz}?LCV}u)75Y&}+l?d7ZOX=pl<-tYo&2 z99#;)Hrom=b#Aiv1VafwE0j3}$p1X>wo>cgSU3#q5*Lz{kmhK3_1bBpXeph|y_Va7yG@jV45PlRSQa^t9ErrPnB6uR50@ zAhk3~b)Q@oBEXB&A}CtXScWEV;;|VKyaAcYNh%Jn@`?#P$3bG=yrpjzJvfZWvCBPX zP1Rj$8`i3TuStcPdWi>{vI|gq8|T^&ou2UBDkW<_aB{UA+dJ4z-E86eUF$k0XmyzY z?|kY2@3diqJh;?@_D@C%zMcahZjWQZTervT^2onk76pZ-27lR57GdfdNIZ)Dn?kXm z-NxRX-!M2>_rbXw@;-PCm4q+Ss?pL@OZk)3-s9Fq7mbw(p8}|@>xNKv$6W1Pq z5JCV~!M*{TQk~DgflyL^+JTKmAcPJBeL)YU*@qtfpe}`&W*b50bAofq-FuGmzj#N9 zE{F_ZrcDoVg!_vL_W-clojoxuL_u+A&O-x-GKS8y{UZ)rF9&q?&^jhHMiNpl7t=^R z%U%K9M5B!Q*^n8A*M5Ha2qz~dDKI?uAbrG$8GuVcg@*jZ5FJ3w6~u)an?-*ba?MX$ z@l-d8B^id|n6<)$$unsORlpP_g5la7Q9zJ7YoMd|)+XjraeZ8ju3(8>IRl-KZwt(H4`Yucnrk2IDs?t2ZFK{9ggF-a$Hf8 z5Wd4CsV1yyo68xHnAT&06n89q8L-XtOca#o$W9K1v#?+?X`Y5J@=4KHk9y*wuB9+- zDDtN_p6m>5DaRu%Gq=*sTW?HQHUDCBQKR^UveY?dUmJ7Xxr_Hfu?B`#gtOYxxW<`5 z2!hS#s@H4b-sLJro?5b{$II@#!0L#wFaIiWbYty3`=q7x#_!y?`xoOd`?z8XY(_?K z8yVK`faP*f4W&p(QLC9q{HmKQ+>BRFIU+D@)=_uerpu0LAmU=7#lc$JL5=-D@tTg%v5@Eweh2$1i9F+Ba+rfK>2s zZnNOoy}r9N)feRwHIJAY1C<=Gu-zW+y0(Sn?xP(rhVtyA;Ym%ii9^oU53Npiw!ryiuEgW!q_TFu@OS>f{!C?q z*?+GHrjA$p%w;0QGI-dA^mHkJf(TVWcPeE596E+=r7lPP`H%(cT{n3;`|6EUck z%rg}MgLx5zw6@ZE(ClX)FD#uBv@W`W@oS+6A+ zd_6;7M3seTX$PGBJ#Y6^Uqh)6#g2|Ufys%~)-&U2Wfl3cwm4MVcf?aH83A>%R3>uP zOOV81W1=j#ZqAf56p4>q%Zy8-RF~{>ENln;lfmxN?PBf*sDxdfd56oi1dF*_hv<*@nQ_ikSh_^Y+^gRz^CLtG4g%v`Dx$q zeNiaF`l`c?as)*re`ALZ=T$0{t}_Vqn@QCkd;S0b8A}T6x{3}65q0NOK?Us@3K;?r zWz(40a&i?QSg7ne$KIhblMmZ;TNZmv_TV-WyzhLcE}HrA477MA140N8dd#?LR3-y3 zkZW~8+3>Wh|8~=1=z7KNFh?rt`RV60Xv(^w11f6y@#8djB^5BCDyo8|#V}%873j*> zxIzZ6IKl3nvfLX9@tL|ne+ucu^V5&0R{ul>Ric{YQ{UN1VjSPTlE<8@}btrL2KA>FX_6tfGr{t=K($f7PdXu9f+ksB>Igy3G zsCnNt1zx(&Tetx$Rlg7t%TbsdMb$f6g&vaO>6+;r>d0ckiJNPN6`e$n7AHtFvz|~5N3d8i+>Fo}^aqMUVIT+*>es_O#}YO>Mxg3ix)G*f7Q}I!e|(oI z&e{$OcQpi41|_kH}3@C@4(DpBeX~3i6%%{p(L$U0YBNs4ot;s^McZ z^(2`CsTFh;-_p6I93G`@G7mfQ0DNzTMYWaGi$axmt&L>~W~?Ms!4I0p@`FCDB>RO> z9kE7FVNK+BNyK|&1i^_DQMj|xcC+2Vhx8#mQG91I!=w%zYUv0du70cudOyCryB0lG zH6-7hbSNk~h=NB>!MzBLBbUDPNg#Il#nVIiGF!N+>9cIW*|jmZY1`~j9a)NoG*DCJ zG^_N2Jl_;^bT>3ifT(tbSL9LEIim<%H7V1?(Cz5$Uk=??^}@f`VUWJ*44gJ*aZKvw zM?3}7uwYb0`fH9jCq+C~N+>1i;ld4AZnP+MTWD~qEcnlC!pJb+*%vL&uA#Z_752Bd z+YI`U7I?vsaesxTGC8GY%XQjD4Ah8lxBN&nU7ujVUhK;E1xlGa5QQe*gG2^mO^c8MW z2s?l-ajH{aM$^mqCARv+{ewP0^6v?#;`i`Tv#*vu|2a;giedd?-7Yos&A6lnFu?*% zRbo;9j237UYL0WwLdYe zLVj=m0O*ndhiL$3(`+sBFid9q2P)4+_?F-}*jYD5S}BdXnDEK#!BAaYO!SaJ26 z=c@wKD<~7kze!D`CFm}2RXwYaZN4+|_0a*iC-}*Zeh!vNbHVg%=3uDkXp+%r7L!ec z4ibn5ENC`W8eG}I7_NDV-QX&2plWwQd(tHb6Xq`-=sr8iz`{XF+`hmr`g z;PUegp=kr|GN&kEbegIm<#H9y@+>w1PPM-W>Bl1>@l; z!=@8RN~QE*&!8lrdN9-C*llkU(S)W$OYYnaeRM>Rz%P~$kC|9{rr>JSOL-9fwHf&7 zz}S@y_Nv+IpS1OXx+sLys>#?oL@}cbVGdxUG^|Iw1MBjx{KiyCe}OFKP!*{4TM^bz zQaeKg38gBvfZ#}JP{^7Rx!+>4_6mSbv5ih)&IO04V3A~5eFZLwSgw_s9g8hmLn=RH zcdIQ!rj`|Uy7VX};Q}z&H$eqq!r3?e^#iDLpdq;q3@;^Rv>Vem;$c7NnV@6Fu145y zaV?M*IkGNNMoY*{!VqY=%-guTs zyu|(*qbvC2q~fg4J*j8E_f8Nl^yrwOqeQGcMd>k#k z`YE#WR$|OBmZ3fO=p{`sRJf|lL-z*vOZRk&p}gkbz>pAsu9SJ2pppEwhVP=&f52QU z>O+PA*O|njg(N4V`>IzF=%YzTaym_FLh#cZ{?Z8#Co6P=c<2#Dabj&~O`+MGEsP#q zabvi!q7X$|kop9tt^i~@uFH|z9*3z}=DAiJ&yX1|W2XVq^xCpW*CeDq%fWiZO4j8m z_r89}kL-YRvz?|GibjXXn%EY-6jWTtSNJvx63mul{#39QcFPG@B*s=WX7n(rl97e+ zjb2!o*?ejrbEu5{NJ)OGH1lPDO^<86z%I;w$vRJImhT@P_lc->x9X>YT-Ur;rc77G z8_Jk2_1nnPsLnG^+N(&InNZNrYg+Ip6+r!6E2Cplb3>+$66c-?GyOvcHvh99;K@3U zztXD|z$*uRB;oAatp|r4e%lrG?iadx6RdIBBev*DEViz|h%k`+6Vz+WZjT2JdolW? zj}fN*yRvD({Ey(r6s57Uzb0A5?>v?R9%;o)F^7`aVYE@LPgjkS*z(C>LRGGU-)w9o zr!EqB-yHvDP;n`qI3twziEFwpe&#($U4H%O$G{pr)H-c^UW6QdHU(Kk&yb?yr!ZP@ zXAL39ys5m7{cJ}dO%y0Vp}koTx;Rjyscg78C@4zSlKgAwReglL8s1{+mBuoJYLzzU3W|~h=ef&HwF>YGsma(^-=v#aC zpy$dz_J6}JT`&WJR2Vp0yoXk0$OF%v`g{TXMfj7k9lv}Z92-0Q{lVypx5yAkJXBy# z_5MXY^J@VcL0`W2a|5~vxBnK)_(TW#bwqs?ANqTs_WPvv8}R?SMeq#coYI*B0r8gr z1Cjqv?vFN>Hl7alb|(L~VMJHgMdG1Y;=k@AdBNCki4g=f)~qd3%?+ED`b?X|b-TZ5 zO9R>U#+58&NPi-!$>L+ARZTN5`J%qZiu}F8e3?T#!u9uqs`f{VIR2vCwShtnE)5K9 z9k};=Y+>rZ`=3ng9fp~Hhj0ICxBL#@=)Pb)akQ{^w_ah~Vj|t}_rBl39DSMI`Rd#G zBH#H!4C%2LelR^p7&#hBvwyvxsP+H(_)+rx`l$Umo7g&XzpVz$Hd}jZ&Gk){_ zloy1{fFEdjEXEQ-0JS zq;)uUE>t@IqEEJp{WdJEccs9#vm7?OJ{DDunjs_CYN=iEvy9cS42p0rp_cVlizAs1w1m!N;BHJvW!g4G3u2tz5P?20z8Da@kJkga4$;tMZ?GY%S!f+KM!H~oWdtQMrXXL>}@%P&dD z_4W3^6?alj@CxdyoDH!<--;clIZSeS0rETo@*D?FXEOg1VV2!gp1Z|TExRxVi~h<3 z{+$l;^H0e@A8gtD&bj|v5Ah`n(S-fd1qC@P+0T^JjsQS(QrqXYj`hmp!F$b#qmY6> z`imuEQL(o^WA_n!2{`V54@?4<18Qd}#+P%Jh@I2*#--gcN9m!JU(V^unSAAIl`6ld z(aE>sXPDgS9HPqykACHiOBW&e{w8;GZE1oYt|Sqb3L5Yh*6$8k6Amu6&rh~&2)KWA z0X$2~&S;0*B|x%M=12;u3B6Cw=$(IcXUw=B{4Xb+UkA?ZbzHI!I9etrOY~v zIvaLADV@;HNwqWdn(LqhxQG6!D^F`bN$ef;|GeR=u9=t=F2zOg<#g^hcfj(N{T%t3 zD(|u{`nmI8)V0Xm0$~BhT-p7E+S4Cq?XFGWmViES)LI@GS9Xy29{FK+4kna0EXca~ zqH@3#yZPv@adbLiT#oxBpVZzfU4MBpsEo)mN7q~%llAygrKFt_30X2+6r^`FLCuK#mC9RuF-3w-tI*bBQ-_Vxvw ztr+n(uCb+2^)4N#gpm&XwCh+{cgmhb!t&niLp>`iO7^CtjCD!bDo)tNvn!T z>xz2zB@^}~2gGO0ZsIWOVXbfU0OXpaFzcc@FRR8Jobu9Pm*ru%WwV(cr#|py32=^0 z&95QJ8nN#4BcTW#X1_mdub*sb3v$0hth@KT>>JtH)^TvohGWilVy<@J0e@?6jxs+S z-LC3+rDCEHc~@8|)$ImcH)Oia#ov_krjZzMx;3@bNvkUpHYG2tid@;}+i@&)r^ZoB zd-k0UFxy;Lub z1&Qjno$_M(T3U?Lw|uHe{6jfXca^2kojyF>%|~hZ%6rJhyJT5zGW?U><}`<08oM2) z-pzl#UAxlQ%=(4lt42p&%YkV&J9FHYjsJ{wjmvV=$($0iK?iFx?|7Wag4DgsI+Z#R z9Y3x9fF$*1w)3Jz5}&7B_>1PD>Djp0DqE#GslrvMi9yAd=u&p+I;mo7M75(uwPU6t z6`j#LC+zD#519B7(36F(i)IIQ6kBhu>CMQ}-1DD@oZj-#zXeu{ZCx{{NhaSYA)#FDC}a3?XqO^{@w@)$-7Cn{R#5ipYl> zAbB^yc_Fsk3~zNLv^n56iTOa%^S5|Y2NG-bLDcz#tWFMA8y6h^HlR5&pgpQP(Qn)R z%Pmoi_di|j1^3)_;v22))Sx*cYLP6)D^fyGqJ}I_9#|g}WmI3FN}I%_iA!mB?IE18 zW12syC(vTd3FG00FuWcO^;iS}SOpTwG85V|6JjSr;%mZjVt)Ru;g5zC>)7J5D7b@O z=5Y@{=q4Po#Fqb^KB&lWSlF(-f)lIBz{6NUuj^detH_u#rGp)|hLnmhXge|%?r7uw zHHv69)_D@xkrvpo@_)n$>dyFIn)$D7EvUQx#X?*n?*9~QE^1;)KcWDGIYy1gl&Eg! zxwJ61jkuD?$3jbz?y9E9$M5+@e+mBAcdb^x&qUgqSHI}0^v&{(|BCj+sB!)2-wZIEK^*s&Y7CE7snWhYE6?Q%u)$+Xlx+t|d`C&0 zvWy|J`W*kL{ibaHf3yRax;px;jY&&OhzWxOm(T+tHKBo^#DXX9Le}_Wp-xF#M6ksX z_41Kmix6XXB?TOYy=I%QEO&WA+^B~c8B)7Dfm}YGm!;{QFVZrZwobc$qSocX4z=z-wm5`GK(vV3ia)+L(3mBhoa$ z(KI0ZA7QJQ{kI7IbC4+Lyv%HN{b(vq{C}r~Ud?}}#lgR{n9+Rn#`wzn-?>kX`Ptjk zq~q|D>6v72_EpofdY010mtj4Y(USw;jU`mrdZ_6(v^%ad{f|pXI>h&UJYL}SCTags zg!oa!$e})NwauGZH50bxH*eH!$xPUk-okkN6y*s}A8wOJ%eiOSR?a*T&^%v|s~vVd zZ*Bg3^h>5aCXoLq)T61$GlD>KIX~=&JLGpwq}{%R6Ls3TD)GX&Tc;do;k%6WLC*Ph zNQT{gdtvP6s_u5*zFx@(cE(?;`Y<#-20ByhpfeK%1Qta1B_h#P)W>iDAQgd_kf z<8-&+wcd6%HFqb`{$T%ljSN?qXM~Nb^{^EGDJTXiu|@aP*y;Qt_xxv=_kB2bkU*+F z)3xX1&X%*yi~BFO=9L+B=&%o_UG7g1@B3fXzVmJ~n~Rqxmyz-XY`DX1-Z2INrobNm zpD5*yM$4>|Ownh}%%!e5ai#s4flCa*?ob9FiNv2}dhxDWSJ#guwlNHxDsNr+o1%*r zd^loECP_9~l00sw1w4FAecl*z-)7u%-5dVsucIPD7<9*)@IKk9&yJU`JUy<@n0jsN(|=1$eZWm}`3U?^F8=>gWjf>Fj*JV&|Q< z+tZGko1KV+V0|>_4*4laYVaw!st>Hr@|!JAVQcT3a(=k2?dITsbT_Yx8=noG6HyxM zc>A%sd$T}}H%I)U{G1}=-X$-;nSrmdmyGX`RwkqmuB8DV0`qB96R~R?F)VTY4}NcR zV0?`cv%4MBuq4NxGM*7f1QjrrDBTo2-XXI9B6=ks9u*X~#b=J$Huw#@4QsbB5~ zsYCuwRPoMPL(a*4&ReNr*Cz_E0CClUQ$_?Y`RJ`yu-Df-Rb(R8x}23ai{lx2roxQ` zw~CInGl{*GAnOAg|7;!m5(}2{7%zJ4pt~mjJI1}(-1=%3!#`qx!KTDz#=;9)cm(

yH|r1S(-xyZ>v{nk8{~lz7M*X>&tNW z71l9KY-u*4~c#_skuJ2hHZn)Q0g!bGy`$?Qzi_d1Er7K z@evW9U~6gn`dv)wK4abqxdSFcNo4ZshSAV~`%27X8XdXMI?w-nP@;$rGuX#>b1@;=7ujzL!D>NzQ5UW@>}d5cKkiW4q%ZcPM~h6# zYHu7M0qLEC&&9XcE94~pEmaM5-A6X>`@O`N(iZ(*h-x{57=7&A5dJ(tmAG}d(qm*` z|Jw%3P*Qy59o&GE_VXCsW33_@cb@6_1>?4(hG(z>_1o z?cCwEK#Y@=$uIh8*RnwJ#Xhf8gtnZ*;kq1I>DCung)mtCyR@n{kaYFCuh(4*MT6t^ zpJ(1e@cWAkRIK0!4@^W`SN9VLQZ$SAu8e!q*Sl5I=paoqg>5Y0+{s5= zfYH5`Shev0ZLW%-*YFfIW7*Bo!xIk!Ct55cmVfvqnIE^#I(4=?ywr9CljZ!%I?g5; z2qJocZijyh7v0RwsE7FKTA1cY3$xOaZBcF%>H&bx|W+8d1dn% zZ=so{HYny9JS09~vDOu&slWtkMF{e-Y=w@nCyKtgJM!-%N$gi4;{!+2@w0tcz-6|3 zAhp*n>+Y(W8nb7b53C-@E{BoKKt2Wn-8HaRf=4$;P*KrX$mryL!I!DM8u3a?RERWv zR2WHV;@9}e-O8rYjk(W%fOqMd)&q0MpCEUg=x0x) zhR~M!j}5)pB( zN3a>3n~tkcGW(ae*SM%VC{S0GS1;XBc?IqLh^tlK5FTTYW*IkfT=a=&r!1{e^efLB zB)l!!$a=Du(-`orA?Ex#qc=#PDI@JF!3QeJmFf(Qxi+5%@VN83kQ8`mJBKMMYtin9 z=&=|Sxd6&I+|rOYQ=mf840sadLjwzJdmhmG2+qxy9g!DU_%h$n`wS}n;KxPmT*Y&# zsS1DN`o1`)6IWh3liDG*xaU6be6Z#l z^OKRFpy?~#?$Z6(<{Vou>&n877GwA$4#!4?a$gK?)?2m-H$8;;=(Q<9Y^lXgaa5iy zz2Oodlk_CE|_lqb7S2^jz_bjKDWUVB@$obO`K$N60 zef7xFNu5q780Ip?A~^q^mMtm7>3y6T(RlZb7TNiU@$;yT_>f;Ss^6Yk^25W(4qoCS zDiZ9i75gEKMgtBm)&>XW7h0dGQSH=kd$?%L`>t8P?&X!|ool{yf%HzEvS@gWSsyOebsvU-dEly_ zACy+Gu ze{SOr4HUM~syeuF@UKNl@A;bSHOMjV?&>wBF{4#=B*ys$Hv=KxFJ@oQnxaeNK)Y(q zu3iEhx2Gmxr^Awy20XP{?M+@V#A_oWZtlsYX7-Y`sIbNX?kYPiPCVA;*0eEx?+Cg| zL}=*?zVq}&*8-&hoKdmhUS^XJdp^S`gd+!v6*hhQw9Z#q?^4~!s}$EDy*XQOXFjg* zvB1w~moiM&LoQ3$zS4urxZg$n<;1w_KnZkE9wdrioR?bnqZsAkeF?pCP+qU%(r0Ic zx)UUL!OFcH0?p#jfc#eYw-bVp!Qr})bp}j^9(!#E3Ken$NY%2`T98Brolfx^mRjm9 z-FdhbKuP^WulbVOy3TWFWlYX5>mT&JF0^Usq`mqaI!sh{10}RE7p!PSNQ(jSRmT`< zH%e9AZteTqDVet6ga0`#SfmX5Boxj-h{uKFXVWa(Y%Pwzwwt-xztUA7lCyE3nD=F# zAwTDdhW_l#Lvz&^uO-tKm9n?d5yXacBn4UD>n;X6pYKY8bLDUTW&~`|+E1Ufwz!*j zZ*8&*Q%Le3kx!&6aZl;6MpLG=tuwlt=~9P|x&Uc>C+6~dyUJiIiO*N-WEmxZ(>%N^ zWg>4h*BmAzYUZ`%3yZE*xv5$2gw>|HEYv&3u7(kU`sOqy9vyKzR%s}9@@>~*HW_VI z#d6+a33ddW5Tp4ElPe2_;FCNLh>_CYzX)BcU9~5Ym>_q0AhHO=d*;MFTwJO=;yamR z`+KMQ^u$XJ=TFti#gZ~D_#xrH3v#H{u7(ImC?-XjOjnqJlK4sA&muzBhlMsj@@Lnw z-n`t0E{DJ}Q(TP()*NjJ+w)^6-6BR`nCczTOD|QZG&mLVdcZiYl$dxWDe`=y)5iMZ z<;m{sQQEW@>^nW+=P9)HsE$CR^(if_rCK!+A?s#NoZiD^%nL=Hg@Tr}Sb(Z{5e*s{ zzj>N$VUgJx_>aCn)+U_Jr%6;uBSKEG8rj>p-rGhcB&^gYTv<0A z!vU4iGXu=jl?0Q#*~-DQGbWMjN9&%4CpGNrR@ynJ3*o}5#NGth(=(nDr=CkyJ5gNL zfYon=M`^`ootXXfLf-BzT06HzQo96~BE9RLd3~7WjemeAH&6yASL~}iYvp|)!yGc6 zX-j4~daeXo5vwmY96gOG6mv}&bVz;k9E^IfKpQ}SRarmlMJPdNj6v{{Bi0#$s>Rjb zfma~x10wEt8=Ihv*|1A|cMd2*4M}v?8_BNJd#=`4E^HgafS)d`6?&BJKUPVCITj>; zrX^M=Y$1uhOeFqTMRQd7VU=D#KpBfJZm?)fh~h&&`T{UygMQMUT$<_6^*4`Xf3f*3 zS-#gmCgBJA_gy8sZ}l16RCPuCgVUDLZ@Jg}LgQKE?U8z`IeLQcA$xyzCnCM2s0{SG zod9Vj73^gb4&FYkingTZu2edtgoG0(D4$DAz3aLdc`kr(y%DY&Y0cUWZaZSi$#T1S zsXK?3{}eUZLu(RNm1=NJpMSjYgTb}&yAP76YGyDfxI61!c>5)Q95pQ(KferKUN-a- z-8`jOYC0|wUr=o=8~^SG_QIk&W+^P;u8M)W9cLqH~(sauWR zDg`s42Bf>84gD5st7^xhy_&g-@eE-bwJ4K(Mfa#}ziM>)k9%Q6ccI=%`VWA9!6Uax zN4H)`^2xg>Ac+9fUqci}Wieq{Iou2>#6Sr>avZqlY`gIoj=f9 z(|H-}EetN!<_wFBF>|8Ie#F}4Vy=KWy@nlCMJubX+6Ae(u@TWXy2@=~?N{br5M9of$4X{};mA48us_s?ypYS;Z_>MeJ_p(Fn$?_AaR>cN4g9 zVgKcC_yR8~Uf1N^A*qa;Kb$p_Asc6sLIUuyT_$FZ4m?2k7$K*)Ul#DD#ICaWD{%u5 z;xPn1x*41lZTJS>q-+pO8DDN5H9lGnnrHubuzQMpr^`(3B_+bb{V)(^_(&~%A;*Fy zUCUBDmhK58d`#_g_5oyVFkD_2(d`0c7VC`5{wez2@9>xU^QT9QWW)U~QqQTKHqjc|MDF!2-5rl)3 zwBVv4s|yH|g7y-8PLcTqCOQGp#z!B2O%5tav`O>$Vj$w}&9fIRkfg_xhy=@Yn4CA# z_6vQy#JXdHLIwwIikOy0VPP|@`g*?TG0D4i%N~v$@V0j;Qjw*d*f1fCKH9@Yn2HmJ zv>-1ml>hLllFo!g8;4c|x?aWv6Ye;XIRcID!)DiaK{@ct{Y^RMMp7e~)NpX)**nRP%=&V~i*s=?gcfqvaCq`Akq&Tcr1AiRBb@isc%N5W2z1ZKX7cKkVT^1gmf zb}V7}_EMO0jRBuwuf6gi>cmg~7Rn|$UYvju@!%F42$yXY_gG`F;Yben%G_O0Qx!UH4qITIc`Zkc3*_GGlgCXq<)Q$>Z)>I&1m$8_v$tR6%frC zx@Dt?AWMe4e5Vn)e3$=P-JL%MgfC>qe{7soq32-mnsodCN@6L84=X{hmm}J_RdmkMTZp}Kn#AI&V3(-6 zygc!1gGlwn#1}HV^p(3k_Pp(sm*&fJOorst&*l=5FS`77KWIor+HCq@ykKUyy*0h= z{6~hHDs$|X_S3A-fi8Avnn@r&O^ntY6Jccg`hKb4mi;1O?Gp@eIn=9eUq(1Q3EtAAkn%=`vu^M zg(J7J=!gyd`-C6&P;pH=khi&de?oL#G_>;vhL?{8T>Km^;8|k`&|{IfKVB6Kv(nG8 z!g&Tu)Vl_w?VE7-GJS? z2g?~Jb=0+l>wi%3i@30=S=Ozu()GRH;)V~s=(8w#=#9c8BhH(8K@f&g-^~VSV*n!5 z_vrqS+|~;GgX((xAZ#YGPT5hqM&b~v^*b8%7z}E0jk1Z=>Ie(3b-4AVhcb=_`|&L6 zWM|9fpksv%5#o;BiC$S|HJPZz^XB6bv`3T^{GqW`IE+`zdaHMdTA`?WirEH{j==V;jCSN|cYzj%y&yHs&Acyb zGo30rOzcB1F1I?2#*3)hVgn%db$-B?WpM5PmD}_cza3P? zM>gXUB0*0>bIRe$6|PSqNd!##y0tyaq^K`4TXbq%26SWl4Tw}3EYj|vfqXc>9Aj}S zecq8cxViJ)9PLFs>YJ zLmk-p+M`xIupHoInb1(>;rpx;K1Hm9TMDr7)bRSj-Dg#_74B=Y96T?qQ`ye4 zXcD8AS{J_Sjt0zHf&5L_(erRYW1V zc89Tx;74JOgqFwU(kHPE(CY+Cv@f3O!P4vn*{`xt#W4T_I_h{9GIVIMRR&_8{bZ72 zaMTdqdqiLbMAoiefeEiYb`(C+;eQ*z~ z=S6Z&Qx?6B%|Xg*GHUE3GWTalSQMdT{FC@>Dv6xVMh-B~iHOrYxv+E`-7@beaCJtM zhE8p%*Skqkxs+e+BrrwX2#qyW7O_shY=fyZikjk~^92O<`{Z{EWhH!b8C5mFDky&yR9%Mu zHTH-}Wq_yC#jBRsG+wC1i4BspQN_+LqGhMW2IfEiP)&iw0ai4@`r;OY(lQJ2$d@kk{k`cqvS5HEpqRSOPNX;f1;$ZSy) zkJ2SK8O*CZWIk)5|Bl2$zM8k@$S_>-(4JH$Akz*f{ndI$Rbaal+Q9^VuS96Egq81L zA@p1TDgjd@6o`d@kyDwb438{dJbeuDsN5~f{_IhU0da_`R(GS~3^R>1KKCdqoL`S= z%(VukZ0j!dY|8cyZngS~zZJ42Ob!xrkM3IRkeWOr;$II+dH_zPQeW&5IUPuMA=q;!<%O55mst zt)g#=M2gH4ORGJ~M1t*mNZi+Aw6V2_-AOCj8mI}6({T|-1imyCHO{m(T8w$w=-!J* zJy)=7->@yl%vH`1_`MxXnvz&!DJtkMbc>|Kk|K1J<`j)7U-ARfH7&es-l<5;t7_1W z`vVQB5!|r8FawUGCpeu1t{`rA^}O?!ozHCw7PKI7k?y0xu)*e@0(G<~{EA?&^#-q@ z!EbiD+Vovf2>?$2C;!`iP560&Q22ZW&o%+v7=L7O$SpLt3j8P4nK%Q zQhovhJ)UJQU7(!nxMPIQgDTidt0N5xS3Z-+YLdDhb^*(K5SD?r7#*#VH1tv3F&gwt zp}iM6J7#@v9?@Zf4u`)giD5YuxO-$_T^s?ZsHqv&pob&icVsg=7*tq8vj`ArKsu2j zTA>%>9GdYYn5qmEakEkEI#LvN%JTxMhlTL9;PW4NsdnxL79lipk@Spe1NkJY&=Su% z0~Dw++_51d)E`M60ow|28QS|K6*L5=Bt1dxTW52&KuzgfNnXWNkHOa%YCt0dpje%}XNhs_K+ z!`2F13Kja&CKOP^=vi=M5al z`Ew=cKvN53>J%K<`|%f6@v=%g5=X9v51Myz@fA#jimfKA0@e^LzUTK(q^){ObN zj+qA(xdks+iI$Qq!5@AIE*)YBm3py*C|(Fi>;rqRZE0e`Vtdj;1qt|&8$>OSamtMu zq>LbuJRV;)z)DR+5o!OG-Ewi?_Wl<*<1}{J_MA;>v~S@z0-j;w=M`SJU3gDS7@{sg zg<33j+$1|3q)DPt(6qEz5#p9lM3`x84?7hF&1NS#-KMj_fJW1FtP!{PjNxziFDWdr z8~9~8?;f8sMzshOj5V^h;R67X9g0xF+39%KfnBGRxNg4ps>i1eAT}?k^&M(mU9wp< zq}>G+XNBJPjd=EUC4Y*}e|9&#S>-897A&{`6+^ZHP?x-x<>n_~8iyGvE2Gl*;Y0yR z_SWAP@%;qo_dj&krow-MhwZ_e86DCC7r6}^I zRIg5r%flH9h_kEuF{Ug{v>Kn^a3lBV;?hRxV%mjS0AuG?j+E-+JRABuj?AH;|BH~6 z9Y_$QHps(-VB`)|WA{aV@JMkW@k*je1nveWhTY_boIDJ~^2n6Oi zzU$`F-+78f8%v*-bQgjD#U~O=04G!Vie*lNSJQ*R0!U=V0@{b4ueuFI)yn`F5a#_d zZdBrJz|)z?qjD#jPbd8rhFZbc=USBf?*|K3{W91*(&1E|eMi^Azi~so6}%v#+{i#b zoORN{-;~v5(qU5|;YLtMBvD^jc0^e-cz9bS)b zNxI?uS1KIA(`pD-nDID;g8Jz)fnmclyb}(~K9I-FAiHPsEK#2aqnl&CFwQ<_Z8+BH zzAh*K?`b&ne2On97>Iz_sVAyme=&|y1dqI)2~ke(lj~~P3p>Fuu5sLY_)d$;um|3p zEyv}UaFcZespxa-)uCfymf$9HB6HIWsqX$@(2-S2V9d%BGWaH~QL5P_C?A47u=+it zxbPXm=>q3t;*QYgb7zf7dJ+s10v?w-Ji*GZQix1ghb{RgNNbQ8^}0L+g@L?v5+pMi zqf7MXvOJcttj3UuJb#H%28h)Dg$#Uke?+loB}u-fGBg~SN)({7_=KVihX4|gdQgF_ zL{ptAMw=bT<=XG;Cz3gjE_hv~d+HKYWRbA})3A|jVWA_?bYblyTT-3B9Kj~+^>Dsu z*<_xef9HYXMx3s5cbwYArsrlMpB<7eJ4J!87Iywt%&UXJowHp=I@AXaz1r3Q|v3AdX8J4CO2jrtfsMF|4lVWiXDLRJSMB%z8!aYUP z2pysm^37w0(~e(kVJQwK71%@S1;)+0U|Q&p#2V2-MRylPXNeZqkqsGf5lH%T zIM^UxPAP}9t%#J{N zpuD*Pxjemb8gtk0m~L4e`CQed2R!|};J3F#i6!U28H3*8k|+XLAiS zB^n2zVmNE^QRACvQjE_cFcOO@IHn6toe;}u0|0njwg3m{r~ThsEpaW;%~NXZ=U?R$ z#zwvecpu7~f2+WAg`bAjd45~y$rtpnLB26xb?dIz;hKGfc+0f+zds+I-mn`Ux2nLg zPs~nuZ8$|r42l0Zj=wvbg97mNXxc`IdfhO2(dMS+L!_PRVw zmS2?W?TPc@=>(C{g_YBv8$8v(+E(s8AHj`;ahTP~OvgzcMiY{E#HWpkefXmbgv(mm zE3=^8{-uU7>F;Hy1sph{77Py}mz9P5<3VSJ|J9W#k@Lb)y4*O8^)XqpOh!QsjRIx@ zDcTL$1E+UP;;bod|7FBogh~D^9hDD|!jGV2mJ>-JebRqQT$1=G9PqA~=Qe_=9B=Km zGHJZhd4MLx%6#J7c6oInqE$xQJ>9&{jzJV+{C#&JbL@vNw}YmfE&U@?8+1H?WUIip zw=iu$m>{jD6>yx4J>D_SREU~u(4;!KkBC7At9F?8BN{5MUO}3xcH(*U_Ec(RO`vPr z_}&#jTCRa*uSb*S7`szG$HOQ7(&%)?p zH)T=tlW1;)zCp`4J=;!cOGN>OCBsc52MK9d`ZFa9<4Q3+N62ukSb{1HL`0$|h(V1| z%R_SmVH_Iu($$7l*e~Ns28|=SjfcEWJLIlThGXu^0`{jmt#)C5ChLd zIs;^dJ#dfEUHEhAX(ep*#sUOvLwUFBR~a9SZdn5xk&uHAkFC~_ewI5;xwyD!Bq0F? zxg)QiiY;w~^p8CS!W>K_q!AaZdy9CMj46|usYsJ)uUsbV09Dk`m$I0{&3xZ=RyQj= zG!)2IXy?vm*eEUdLo%GE>hnR9)IFiAk0mW@o zSM<$jTIUd!Bs=s?m!KBcnxJ&9O^k?%bQE(PDjAzSw9FsgWTAj_!yW-iBuG17Gprc$ z8>wZ|Az7Q_TW|F1Gx@vgYcuYQHMe1{;!ED&k-BYl^CN$w%k)*gv)j*<-czO;UqIx? z@x9^5-rGUe8>-fDF@{K$cLsxY?qS0X_B1@_4DC-RYCFh(164nN(bTO4*Xt3BqZ-lQ zuR4SN&gu={9{;xr7z zt&GSqU0^7w8eu&(6hmvV1jV#=lCwXe;Vvba!HgY-+(sEnIV*g+%-Ww%{bhDRc=(!0UNc3HaQ|B7LaLJ7+ec zHdokA4k;u`@#XfewpjBk%!bt4X7CorVFUV^D3LEwHC`;BulYgvN{j817LFxPs3=#nT%e8lVL}Fo)V&wp z)7E#8SXXZlEBmeDS9OUTW2O!Mh-6s>GPz|7-G>lnlm^Oq>?#G)2Ox)`bj)4Obaz4B z>$!vMzYcTuJ2EzarU)`WQJ=9mHS}b(4yledsd^J1{kQn_6mE}X`Ow#2J&8hX3S~G& ziIZeWM$m+HL_+N*w@p<4G|^i09`?`5vb}xO*-WIgU}lW%ibRPPa$;?o=Y#0 z-`p-mxM+*WyM-Tg2ZHs-#i15D$gDX^R58H3+8~ba`5t2UOk3P2o$IbfWb7!k;6}+oQ5!UGR@t8ZUUQLOG%2~ zF?3t{$?;M)d0o;$*)kJS%4RGKp7Zo{O<6}2oCo;yGNQUbQaz5 zA{F8=h#R{Tau>x$IBCR!-`Sn7WFNmoK-72IF=?@<9JM|K(cJICmY-BrZ%KB*AI|Gd#I?dN8QZ{I5d@*Yi{fG+2p~m@FhYVB>2NIFb*A69gZeiYlsk&JDZt^ck-{Tkcx>qbKle8xCTBzpxsc) zisF8$O_s%WlNT6jbV?yKYDuev7ZvER2+l-=_VXKomJn+W5=d$Y1PX2i6qI;Dt5@23zwOW2r`|C0 z87N#AV}XU7c*`nyoHTIbC<;CRl|4z9Bqg^O$v1_3q5^4NzLm%cs(%uAsyvo$_<*(H z02PADfgdknKENLgieFXQhJzoR+D#vmLM3$wS*&c?;ZXsez;r`AHUE7~Htu1M@6EE> zP}IbnC#F)IxrRXsr4k0y>()?P8*ZD_I>xy)BMmN$+-mXm?B6@V(rGCW#pJdXv+kF~@nVOJ0>x?+-}` z5F6dl5Zux~Qd~#w5joZOM9dyK9wq@rYhoWbHX#Lr>OP$N?ilbA<$uj5_1KS9Q|P z_Dan-e%a79wtk1Sg0gufs_glUM92Oet1=oOc7T4?HQuRt>8t6K@PtV^ zCq)qiOSZ&Vt+2*2?gFljC1=tQi&7#w2G)J~0EEJv3ixY0&!nk7R$bdZ;CivK63J`V z6EQbLfkoJ)AH8n8F~Vd6O?ReA4*Jt2+wB4#Y=@~}b2b+Jt?^0btD3 z@%_QLfvAdr3UdP4j&I)5Y@VRPFXo#k6RuB}M3e4mMK78Ek>b8Jto3moU!VBJ z=A3myESreV9EP&1vKFrmxjwnTpJXW@GjX3i1|!YeS|>LP6#brbnZ%vC^Pzo)%&=Hw z9AJgde9D|DwB(11^t}MUyC$MGA|QCrpV-!XQMRwj4Z-iOF;kSGY*h%6diMm*&HDoa z5WXd%-eVsM6-aSL_N8JT1N3FAr(*Udo}mXh#Cs_$OhQYu>j~~ua}mPW-mOCa<>_FgnULD-4~z*%Vz^UJ|6J@U zX2MGvlW|M{gPbM|qRr-JU0+&q%kfy_yRd~LC745vc7|G{<`g?IGPuxsie+?MglNZ% zltZuvyMa?%icbLL#x1d7{$5LlUlX^gg>65X*z*&MaZGnXQ z2GX!_qFMtDWHV=!eA3H+tzYlxM=zMA=LEI<3T73ogj;TN&zj4hE5H376Ju9uGtoZX zsNA#u89Fv9652{&87;Mux_lB7QQlwkni>)3@%GwDUVmIR#P1J4Q^KXt**t?OrheK6 zIVPLLB}yh}hCyp=lVnb6(oq2~*$BG=DcZwXgkbvafi&x};wT3eE!4r~0N)rQmK@1wFy;O9^aO0haj=0nh2flD_bkJH zt@|K!xnKaXgf(hcK{WA*IiIWJ%=CdOPD4&BVU4pDYep+WbH9n9quX2~o_ysm_3?#9W zT8!b1GF`>CY>)~PqfTt0kCe|1!TR8Ato*CS)#HVA76M*IJLq32leh3g>gc4+>O&Bw8}lh z1u_wV-Tp$YM{Y(0D?ky5g8OSFp)u4X%3>H@jUNN(S0kZBw?_UfPaWz!DDe$m8B{q# z&^YIa89|Ci`Cra}QZ!j)1jamgHJmX#H<|%I!Ju3Rfxbz*(JN>5-l9QY=1!5P#bFFk zCYBIA_BJ~10eU|$#g4^1XuHZAyceSTldQ5>_0+*`p!fqo{<09mhyRvx^UI}FxZ!NX z#6&*p;gqcig-B=o2Y5|$$Ag%LW~h^nG-+i*=5a`frY7A#)Ngn6)5e&I_vt?b{Wtk( zm?-4_8JycE1Bfrh|5&u@t|fK()9&_4vO!GcEm*ingKoy9iWqPcE2uGts2ivYUl!MJ z;I0=l9#j6$s+fX90}hWz2O8|OyhKMd#6hXkk=aK58fErpoh~%NF?I@$c{bGteh|k_ zL-DGd(peB%KU6wt-g#MW|^!%F(VBJc*=ezm*G>}xEil-V~}ZLBjwr1(eEMzVq6OxBy+ERh)) zx|C2QG0P+SUg^k9ded*SX&q%5h9Tqrj7Et#@I%2)H$c4E{7sSh3;C|Xn<(_GaJj^q zzRWoL=kNAmuEF62_OuuWa(FK$ZIa}^B1ev`bh!7Zh)}2Aggoci;P)Px!CH0Qx`<_-23~+ZpAp0iTiag#df7;D3Xqc`#s8Emg{nF!6V!|sS$2P zM!1VN;EPbgbToI1-cBwgs3iy2dHFLCr&uxO#h^0bEzRP98G7C5CX{)o`@p8Zwt>va zD+D9$)EcI79*6VOdAj3Jisi*4NbT$5?Cwh}Q(D${aY+zlew%`9H49L;v^CqT#i~0X zk)$De-AK`#{k-jW5Ipl-rR9AK|3==vJrl{tg^o=1A^(76L7e`B4P;)_ko?@&Df;Pl zsk2U>0252IFwCK!E|i976`fK{1Z%l#=oTSJVK4fG^(yDwhPScoBR^k6-GZKXIFnz0 zq1~aP+Yirhf1X$Sky-Ibr>?mVT3d}lnO?0&-^l0qg~{*nxf~y($+I$pU$LrmGv9jx zk@Ef$6ZL2KUh42~dSeIU80sB_HJu?!1X*(fCp``u5+X0!NdA;fEi573b~qt&b9Fxb z6$ItVy;!)4u=n7JEK+P8BArnF;-#;V0-bx`a%&Um9;zaR3}gNLR_dYhcM2$Q2)w13 z@B5=lP!>$3@Bm+6GQM0>FKtz>ES0bv!i-8ix8MQ?q68rGjZ>s{NYUqP!Dvx~b{Z1k zt%$ZZH<(HA$vxUIB1|d=hXescnIBbEY73;U&dQK@Jt*zVYBw_heGq_@RbK^m)Y5 z7-ug`4WUv#C=Zn2>mhz2G_J$W%u^gfUtsc`d`w{sA+x*Ca2JSzGpe!In-%z}wX^^z z%C;fM6m)JvMJE)mT=nd-7Rz&iS{PLgY-#EE!Xy)gxsO+Pimn{}DnOoqyw&vB_+Cy` zDR4Vl^DhEVTkcQIp#-F1_@yJv!&WZ}30!{Nm{ z`~>3TyP^b@90qet3X~8~z6r(>Rc^ShSfOS71YGI$U}acRJyoM*V4oV;#@1Y7Pt7R| zyP@a_nf>`WVovS7N`_>x;VVPYaoST|rpy(}1kLPSXZY%3S7&zSIJD{ zIkb*O3fY2I*2M>2DjibcIO$+bO!gdGqigt#)*}|svHX}!4vDUAfOGWkW17*iNB;B% z{CwtWi~DS`x&jIzB^F~CDbX{WVIirxKEf5R>?x=EHR&K2Oflujjy|F+T9`=!$O~;& zw`Zr!3%kLuoD#s3R9{1JM#`~3l2ts2P*$MTx`YIV7vql!ens)3`v3{ELs}mhQtF}t zp-Lbwps)o?A3tRJ$F1&q^IBl`L@VXxNQs^NW$n58dY?RK(Py?E6gSMyf<^Y^2R?F63B5F0OA~_heKw)X%m_lB0SrN< z%NMSr@%*&RdtR5yS&XSm0Ysb=WdVi&#EL%)!5pz@H!Qem6_HVBJu?rLP~H&N1Y~u1 zL8mY+tUa6@{PuFt!NP+zDW!{?eJ3K@EbJ0I!E{4m;^Nl7#H!9be&2*3k;K?{7{Q{m zyl3H#hE#_=*e2Zi;X9!n3bxr?lqITU8_dTQ3?=z!_4!b%!r}g;;-I7}pC%Me5_@%+ z0btBmqgQ5op+RY{*Q*(fvHxKXX~8EZG@f}#&$mEtBL3r~_^D2*^*)cfu}$aU0iXn8 z5X}X+WwC2OQ`dD2}*jSC(UJM^?)*j%j9Uf6ft>YJ_XscoxL2_Qt_tPTm?hggZ9L4d2 zGoo5Wj6YY@M6fhLo)fAulxFAG)mCk(uQ2p_C))t5?T1ePh-A*xTRffkItzzo&;`c^ ziSR=H#IJ%Q-V0{{^|UcR7`LG7Y}kcVpSZBrkg$ppiC_VD>!L{9Z?WX52N~Qj;2$lp zhg*~h7Hl3%g!#3xRZ_y5Vk{WJ=iu(Cf+i%(LF8Qjl}byg<29E7L0m(7#cc)jY2Z0p zvQF*gZv6zJD7Nh97%Yf-yZTVD)?r78(s|7>0QS;XN#%G-ge+q~GrtFhZIL;nuLbG3 z)u)4J8=6{4MTiPXr^L0zjB+cRqz`X(b0+zOcxk5ymWq z68zNV8>)@aYt$krrQxn=I?MOvsms?}tzXt0p6AMO(hKv#(@w2QytM0u_a2{r$K}c5 zS0K_d`8zsD(M0p%R3Wtcd}=H^ooR|o$V@5B6P+rGvdG}FGy}`eddPWD2FQt(Vh5zp zdS#)1@5DzfkyrnAjPDi^4g~X>N{wM0o?u<`9wIVx`Y)F@0o=;PFh@pxfFOW08?tBT z3jRy9k!DKb9(X!m8Z4~PZG=RXH5SJaqk|>RDGLVB8Z7-5ubGjzlhh&|AdhE_8b z5W?^Y^(z2t8ypZI>&SRHB>gFy@GY~fX6nJ@pD?@KXwwQ+#1DXLf_*;54Yfc1+eaCJ zh@AT=`RPF6w-xrQ3!vmr-)AyDgK!GbPIfL6nhm?ZXEya`;BWsCMI-A2ZsFyjM^B$q zc2&g!4VU!;c{Fb-sQT zWpZJhyi`gGImR4vNl=j=rZ)R`%<#<4gi`Xz=Bl#rGtg&w>Aw%KpO?e?95FrUJb|#1 zN99+B6x~^%)A)Lx?Gjg>5Q20^BrEE9+7&~wS$FQXL{oT5xmub?h9yKGy7w%?w{b(1 zIdC-_W+x@-etCl-GacCXZii_I{*XcE6wan-28OQBAc2pCspsg+U{i2iW~Ud$+r5)C z0v2B3WR6LgL~9{D+i-G~C=v3J4!YK|-0S^}MY_tf*`le_e3kmynyk_?G;$7de%jUe z*lw);e5^1Et=97EUnnMRQc=^68-4^BUtuI0X$^h*>49U-WqC_F^u&DAw&O1b(LXVU zhX=N7l6-oUQ!CK>Fi6)h2=6JfR4n<^uvkE+B5ZGp%pu?>Gy<0xlXDnvXnK_=0#)f5 zY#NLa|ctwIrKWSzCP!gx{HJ<)?Z8`Vr(BQY{<;C)E5JNz~iDyjWE8~R+RBB zj$JLAE!v_HRmP6UCk@Yz5JO~jB4c59`z`+3x)JG5=O=iwE zBNzvhDDz-wDyFDYD9)IWVq}PciNF^93u@xLz|8YqiP7wpFDO}%e4yX}u!tR8hQ;a$ z?J$=UDH$bEpcKmtzb4zz#psS=@j%D8K3p_e2~R%#`{C}2+vv?#OObo&HHmpY_DX`5 zW-2UuZh#TX6i^tNBcPS*gqT82sVt+QQHt1b#0=}AiIOZ2-%!KLm-sSUp?oeCOwi|n0UW?x2jG5Y1@ zfNkU+VUz?Q^wDxeSgD2tG{%KG435u&q8OMJ`xRg~q6V-$Hx09uUYUQIuR9RzwuQS~ zgHPW8ro7)|O-PGFVBIGjS#s*+na`whapyOny%r|CwHjumR)?OKjXA?MP3Ka}|0di? z$s%?7lRJN4(<-I!CVDa$Y`YMe=FS{V0BrBfDLKBEgYHrs9yHk)bR3TsFHkd&-_5TwX%GCi9ZCoNwOIYK7_-U+B3Hmn=8f{9}o!L8WG` zM)+DCHK29PpCYMVulDrP7Z0{|Vf#dlV|T*ke2v4j#%Zr>h~|Nk^Du)!cd`oxT-SHB zIvVsh+VZr{>G`Z?xV@OT`+1gt(zNgG$8=v68VRq=+{+q|v&1&jtg^C6=6Joraq{*+ zhB$$>a=!=6!$5&F;_n+#Qo4-*gDB^ceP9Y{(neu?mxZdwtierW!0{EgRf(;U%910% zwqSKLy~0y>fZ4i(@*X3&52N2nu=QEt`I(&d%JlZ0wtev&U%wU=>dlYdR(S2A&i0tG zKl2>VsyQDs*?Op|JbIt@So!*}UlbG%d4L3?G7ozCG?jiKK*KL5HJsWHau{3xzQ2X? zjio)_>Ukp?)^1gVyiET6dFTZ)`o6y4op8=eSbwEAnbd~4AZPq_*O`5sHk<)hTmVnI zql!`+Gmx z=W$R$Q*XV1{_g`r82rbgAgr+LD zAJpAT12~!MK03K`u(vTG<-9$ZmyF(Ae1OJDZ0t~o_}qVBINxrB9nsGZpkCnLA+QL< zu4kkJnJPrGEh$g_Mhuq>Y?vN{J-rc!`l{~DzqOw>abF1Q5daSiV`r;-MJ~q9i}C!n zejVkXC4JuEwNY=b(LcQrTH}+?uwBGKQ;80Ufk;yN2Yof1o zHxG{h3@+vWn%^PHIY84th}Z5y4GIAlcdJ!5GKynCB*)VF)GO<7RXEq$wyD>^;H zJ;<2w8M;O!+h&`I!|!~D&EOBwWm2I*$#d;}(B z#%#8C`%V~+h$w?rk}rq$GG6Nhb8Ayc!W}nz5r*+&?j0wF0PCT4q^$%BR<~SD|G|iXXTgG~Xw(4ZG zX9Dj7OW)1t5(kUep?R12;-$z430diA5AIu_w47 z4kV^OGT$(;2cPowPOo7Ap23=Xui2=q-z|!7c0YJQU@2)}NL+Idx8P_i-~k>q1yPnF zw3gf8Ym=4$gA;SVmHDXep}z&OESX3~ccg?xslsWLz0tQ^uC3Kn16Z-g(OcgyIjY{gQX_v^ZCElq&Dw-gw9Ck z#dV~I4F%bWnPm5RT$v`ZgZ2~S7AWAfU|7JsrGty^>UNpGazaQr0V!5aNZ2)>e-Uw= z2}2Q0BOM_1{ct_~u^w)rs#R!p*c_u>EyfzKjhsV;1Os{uE6+P54en+A=Rk z$^nHnFF&wbB&OLSJ?tlid2`w(#F2 z6P{yjArybCr|2^$`UU=S2(l)-vf4E}JZhtU_Q7psA9xa#=@nBEJX)4mMc{pg4OaDm zX5!E6r`MR~gXb&9#}Lt8c8xZevlEr~dTX{K2`wgxHwx!J{g*n`7o7S%&G$WoZkw@_ zkJ#BCwO8lvc+pyb$8>;YJ8lyHyI`~*e=u6bHuuNVOBGJhp8hX7C=&mdSrfWc?js?k z2w5~CQO;kdwZb1(PIt_27tP&T1Ey4t9 zF;FQ z_g3NeqxdT7K}`}K8~{Hg^w_D- zEQG1D#om?RmO{n7PM@I3dMloj*}j0mULl|FczvXb(eb({;NnS{yQHdAOE^vno~HG8 zhwi-5_dNIxQVN*gjQ)7K?o%CBonCyV{3RelLb2BFIjc<054j`zVvKzrR@OqMYZXt+ zJT0kqI*S^cDh2*qkcox99In_n*S*Ju@~xD7Z|fJ$g7S5x+nW7w)-!kT6@nNg0FQrT&S#)hmO) z(ReSbXaD<{Ua{^;MdaN;Da&*_`mu>y7IBE5Q}>rsIa7)T0#~g!XCf=ET^PGxGV4)o zwIo&t@{&XhgR@vz zcJs%#RrRwGj)*pjUS}Mh$Oo$C`~jMQ?845fgsTDuI12)2f8p^;Y=D3q)esPeZlU!# zGYvOT9rU%fL=z%_1VzM`jv@S}kaNE}(RW6nR`_WUdHaLdH?rBXoh`j=-YEudJ9B}Q z$J~IGf{FOT`UEAIpV^}Lz$`eG#1L!=8J#int{lKjEB-VBJE)4i_@&9Zmn@k))!-t@ zh+=|Db`!#l@UO$9_ANf+r$KLa|mMHh0vEcom#xZU+E9 z8fhBgdRPd8F)YilomNMmV7rk>gW=*!cZIKkGI{?5U^D5T?|OohJ55AC`iQ+yD#VVp z%Q^xI8L%y#pO(2xZaWygj!<^n9-0`c#6{~AoDd`euCpo54tay6b54)zuo;e8=b&Y&134^UD@+qzj&L4uvtJ921Z)Eho+khJW}mJ~34$av&#C4H=h6mXWDFANMD2d8t{io{ zk;M=8hbv=sDijb-D`YOUqZO7EkHsCt**;v&XsF~cD1FxN8mXJ;7Q$i$;`;k&(g+3+4xPBse+ea_Do51j)%O~t0$$Y|lF^b~ z7kAH0S5eaJwp>Bm(XyUZgf2$!uzt`+6j54_DZ>& z04iX#7E}1G{a3Dx6<7XKQFXH6i16`6{v%?LxvXb8_qi6m$Uk?GBao45#E6GH66tAJ zHWwo8oWNAXf=dlT+ju06ICug0kDAV=g4EWzZ$WNw=PZKF)`}6pyog3-w#ayN((9z= zkIP*&?pduce;#go1@Co-Ni;(|E)Z7BImt072*f3_v{tES94x=4cB#m){Syt0Xg4bB zPv1iZxzJ)vm$w7h{{EuoBtQ-n53Vpqfcd$3fCt^Jpj-r&>#1}YJ;=8bhbB=W_jbRc zOs%oG-3a9eJBi?ypvKCR>F%mrC-wa+oOO%9_1UO!x{+tvJlTZ@VgRx1f*~*^=DZSf zc)Xz5zuR2-fEthuW<20borEZCd&}uggr!oXqaMsahIjT=B~`=5UQR!9SbpoTt9S&- zH&Ci%#ySc^`Gzmj&H8LjUnV$4Gkd8IDF%d>VH+jSB|NwxXU<0YcGswy)Yq}MIb8v zs$rKa?&njA-OW2P!T~LK!t2+dur8w6D(a^>>;Xbm^c3UfQAZ*NsIi|lpS=DFcwRaZ zroI#Yg;ObME8bMLW-^G_p4pKehYlRf;_slk!fFm_^mLANGzN4(kL0D%VPQdEIQ2Y8 zo=SwaFmh#{yJfyrZGF`s3t0q2?P0XAmmO(Fv3!a(SOf!m%YbbfgGy>cxGD9s;Bwvx zp-kWw2$C$8sUJB1ycWDQypDk|p?3cea7}K^X-Mf3LT;h%SBGh91Z8BX!vW;YH)o)j zkkyn+G>Hyx69ZgNeBC?gHA!S86NM`hB8w+@s0aX^@s2lWva!Kd%$@s|XhQUK2!Lq& zBtGKa6J%oJB-@IL>6;#nETZFU9qjT9N4}NPRpZr8p>!E?|2llqUQ*#Ax|;??w(3l^ zxGA>$q<}Xn4sJDsm#CMqi|q6AL9tMI<0CWeq9rfB=&?ICA-}?sZa_e&3fmkc1>{Ct_TVE`hepo2&>d!wr&0guId`$PkiBpJIH(UfHS4 z{0V6A+{m0y{c{&W*B9gfG>qZLaM5@15X>iW1+HZ98p1|JJ=Z$Ou#hM~Vwr*IsQ-XQ zIfBj$fb86y&GbR;0c`QTP*n{diM@40n1c2OrdC)U#Go#=1zT{`=EIzY(u0tSq6CqF z(|53PROHSPGe$B?9yi&s^`5YN9%5F@EmCp?ZYvb1?;J!t5edS(=k}y z0B34_elAQx!qzJ!-o*DhCFoU)z%M{P$|0#GnxZapTvWSVSBnk$2g%#GtOLueY9YxAcO5yu=9);p%;( z_cz2@2JbuW*Ed~f(lx6{NO5rSeu)~KIC$=rCp)lqVtNvZ-S~6(cWK~}w%$26_;*nB z4+ONAx&i_mW6%6Cn+X~b3;g(q(L!Y+k?i@w86UATV&S(YSB9e+WIT>pF6zRVfS*gS zn@ItVCf1u>>^y0yC3dUhYUEa9LLClB0uQ*svjU=ppiC;JXh)}Gk|SUrqXa3VFbu=e z^TB<_l)SJY`_Zn9%b~!c(IJ-61Z6ic6bu>7c|#d$f8thL4E~d}GOS@^x5bL|)6*N! zDoP01nxcg7I7VZMZgHp!Tr^H7dlPwGp&h<)989E=@Zfp*;rj_&?iR6u_TvZyW0mW% zKV{F3qd0l&=)Xq^zZqa(Z77NtrtC04m1Rve!F)X7`1QKhSVx`fs+^z27@*kC0wz)9 z2I8RoFPt?iPuLLA-#LbOCgw4$BqPGqrxneQW5=%zO)`F-Voy|07G^1;-XoYPS*R@@ z)EMRS7q?6?$A`~8R|<3?+MN+xBH{$HP1OuU!$aeN3wnM zgypl_Q}BpR_v7pNbMRHHU={MA#)K=5tX&ny@cMiz$LIUzj=O4!Ap@yq9Np9DtF*Y= z{rz+C{hSOdFG}zsJse3YeOu7{3B&`2U1g3{rGP>53!|#4_6RcGTg$84`+c|LU#}=^ z2`&w4M6#?!&E_tJV@M_q$4(HhC=R}fq~lk-G1V?qKw6^6L4eHu$e%OM?qM(ULP~i=8TnumC;CBq zYCcVzT&jt9hNdD0P?9A$*g36G?U=`Oxyl_>#W6;qrGgJ)WKjibac;Sp?7qu3Qr)I-Ap+$fHFqF2TWJhQ*w!tFB-R+=?4K z*Xu4x{v@zl2&m(+4|Cm9{z2YzE(@~Mn~l9mi&y<)`HHHvx6B~reLek7LwygN4@!&2 z*_@XN+HMMvhlr?Q3pjNNJfa&sc?}&4(yN-n6mstCz358_+RMU2HK9~mum>dZ97SJ{nnya^e1 z4z-_g8?kjW%ibM4e$HHR9R*fFOqfDRPKyAZ5((R4g?!> z{}qlxg78MiKw_$X25_@_YM6&i0Ja&bDpv;B%DY&e%YCCyG zPor^qlKIdRW2KI4ro19QA}VQsN9u+7xPQB-Tp|ASioKR;q3_2;`bHTDe1QL%};;L3V0!Sr6;675O+`mz660po7Cq%HA2WFKG zo%|#Ow78DgTv^YO-j&}}gYYbF&NtM>DKX$96ZapHcD!|BE7JU9++#n=Um%)LD6OGmu_Z2Ws$>rv8?TRnu6jS$TE*#}H2JnbCQZ z(7!s;HDDn&Idt1RKsrWy5_Irho!iojBWBV*M43f_q02c|*%mK#;X6zYL**}w#++>O z`{+6>0lL)t#_*_|44KTz^#(i>^CDXHgZiZ|?Yvx@x33J9_S8o&RefCpS1>0vnVfNP zrqHyK>|G&@Z||9_XQQ!deA?85n^&wo(*=#kCT}$Z6dx}ymv%F{dT>S4%?%uBnFEQj zjivMvmn)>q)b`TM*f_?mNTs#9-Cv{Us{$zPDpmBY-zUD`-FuJ8C1I^`Dx^sVndokZ zYiV>S+o}kxgc;%Tz9ho2(PxYQl7BR4x6b{t6V!aLz2DI0U*FrOs|xOqj~_R+^AC<# zJvUs#pdIa>y&}y_#3OS>_tZGUY>ufOy^AqKy8}$9!|b3*sFG|WwdXZhN-`1iHyi4? zw%*GYX>deZFmHkE^u!qH#Guv2%AtLhN@d4sH+Ip|kKskHZNDMN8Cv*-PRqXISn8I1 zzf0pQ4v26{V5pY^%#NkC0T2ek502b5@2aA0A?GtvUTUat7uRsU$ca&~tJ57G_23!w z7`~g~As(bp;?^Y1S$4#!^B=bL>=(*)@IOUnd$HYs&{lLrd<^)U3T{~r8oA>GtW$EW zUn$^DP*sid`Xdzz<<=`~Wv|J~w@AUhTsW&eUWw)gcZ|H?x23})+$+xZr%2t1UbZz5 zQ)&}6puuB_;FXg}Yonj!tGOfmr8Ocs4EkZU4gUS@W$aaKeSMuSemxmg_(6iErVN^d zkGpTFVrTz?OdQQcYCm4MG2U^>h4TFZ{>BEiWoA}ajvVchXG)OxEfmMosx zif=8}6kVQ4C8ijq5_)=;tXcEXe}Vs7=CGvQnCt@r0CWNa01*B^GRNgVb&gMhT$gcLR^pk!Vr8?!^m&_KP?@4BtB4WC!HtG>{mW%S>#ljG^N z34QvoFfAFa6^bZ4M{m*MP;lFRJjcm413|CJ+uxcoFC>9>8hKVu!ILc0LO&nK2kaA1UU-jGMPTwhPu6hU|qD z%(CO$T0tLF;kT>n-M2b{&i!{ldHKge~BVA ztHIzIPMDnI1q0Sj!3-!^MnhTZN29|HA=}~h7=-z&j)L;5(qphGP8L;5-M1Ixf(b1l z2VX@oZ{1Fl{${8_6tl3J!r!@`q8$)k?dd546>+tIln>38 z+FibLuVmhiGyZLNyh?uky3S9~j=YG2$NYET3OWS;wi`P23Tj^osKMfy)58h;f5wNQ zq#A}T9{>OuG9Uok~F^z%M-6_TjRq)P$C-SxKhzdf68TjJ^f9s)orf#n~#28;<3$7uou3LGT< zc4T8dYN(27BZ|5FGOHFDY>H`QNhd?aFvIjldM&|ny)shr8eB(&=Qq<{PIO6%JRRDHDr1? zg8GFN$ad{Iop-%ka|*XxeFTdq8-G8_#u^HVG}d|qnG*d&WJIPmAd=V8i&cyb< zfRZl=HeeM))frV7RhUqjQvg$VL_At@J56b(R*qb&v&tz7PvaUdmz8RzC}ykN)U!{d zP{_gCQDDOY3~TfaEJv;cW@9jv@q)+_Wp#mT5f-v7mV0v(F=??7SE#lc7&W&J$qBK= zKj_?8z<+hYz)pi+<`{kux5Q=9Q^^(k_eq#sGx0-U1z(=t*d@9@fBv#t8#NAFrZkI) z+-?=2LY(se>Zc+wpl5WHQnVVhC&cP429Kb@zCYvsb#Yuyc2^_1`?%uT5ZBA((r4T zP9Dsn=!0!B6dMQ7@UWheX%DK0@qL?Q;Ht&VF^iC)ysi*VDeNDG_hs!i|5-L0r0;Ru zIa7SP6jlRicCA<}?3FDg;osKQmC==>C=uK!Kc=60p%ydr^mitREZVhnVs-U!Qt$dB zE?f1FkzSP*bv0q9(U!q6b0oW5tz>Q7d?<(!!ke}FNzWD)AGHNp$>V9!$Z)d2PE&>_2$v|Ca( zHxnewIW5_;JsNgwoXZyQW{aa%hDo#<>4k78+*UDZW|XN+TpsENaqH3t!?JW#&NYM3 zq^SK@jAe>^H=OIw-oz5svQm_abt{-N+SqIn+$e%_4?GbJwtpO(4$9cNPG~r>+AV5u zPOHv2YU!%Sr?L_l>>t(I0Zlo^l3sZ!S+q9~-rNzZL_X>&fsU^Pv$pv@07U@8%+50z zrBn5OBGwh3a_XW&KrjecLyAdZC;s}Lt1SPEgb45)35$|pWfVWxEWwUu2LT6Yo#c^= zj}H>d$Vy3M1vo?(&2>7;SRxjAM2$!BMKOtz2Jzp8-9HQ-FQ+r`es02SofZvq6keS# zow+-u@am8{zM7*m!aJDU;0ZRZbi^t&7X(bSrVg9V8&Xb-8r>0%f1IPX9>WwUQ}_iK zb#TA^M~XPLnNScuEmtyszS<{mW>JVH3*=<&YBp=s5IJ`Xc~D%+JTnIjw0b5C&0!^Q z4y+}>ubq7F9W>^N-mX|I;jMzADFlclCB8T5TYIZ!g^8)Mjr$ME62xY7QX*6f{ccVI z4oMeFFO^BTQD_fQqQ?bSn)*&2x3&-Sjh(XOeTXR{SdaPUk=G!XNP=qLd0hu)?wn)# zKz~v{(`4qW8xD?ouD$ZIvvWmk(9CgDxiqQO3X_mwzC1d$M=^rjBM*2bQu3vMRvN=I zw|d1^q!OR85y}K6HVc!K6sTcv*FUaQ2>!aA4p3vxJDGKX$!N(& zW%-v{S3DXf)7lDXSm9NED)zu3OJ-{I0rE#=3P{osn_tVt*6jtF4iVVkAY5)fGj+v5>&%yOM}A+mvmdV+kw_scy?;aV?L~qwsp6u|N{<95(!<$7 z66DgwDVgG%LM}eyv2@5J*s)tAvoYIKVwkf(1S*J*9aVOg#yrKR4~JYt<5}XWtu~rbXzqfNuzU; z0gQo&NvB(SUtnvP5kkjTl}^qsccB7-A42)_r)9cxPAb1+(M{i1o5V=0s+n2Fx<)Fo zpF1_((j`$~ZyW^CD#g%k$f7>im{y5SirBYWS{dc&N^OM^poB&>*F_KhC$@n>;F3U` zPA{vhyZR}&Of+@<4@sqRKeVg41HII-wFd+&LAx09w1>@3{opQyA$~a&hp|S=7 zhj}`orIW}jq}c@eu}q5O%fF3AW2{?r>7NRdgVHsslHR=L zSdo_MkH<~fETRWYtdz4FEvwJ{T1W!45(2$(1e32vbL!#n$EKOl@<5V;s=HY2Ibhdc zqD9H0^Yhi#tG(>gi(Qu_0;kygU96Y1+HzDlUwK-#w8k@^x4|wPJ^3tH!oV}flUeh^ zfo(-pVpPH|g(`fVa2`$Z;}S_kqyD^@S$YwKlvTdtqD}SK z+~@vS!k%@yN;7r7AJmhf_G{4oYyvE3;)ABi-Pa6jml`K~seK z6`qivijaK3FYiOC3xS0UHpRT`7$d{07L%$1O?-(Y;FUOJoNr{<$({>9rMgYb&efKa zi$y;xOPZ?}LZy!?NiQ?ltdtjuW0BLZd3}zC;&YhH0XF8fSNHgg4}*d%n-ze0CYFR? zlE4@?N>= zxjCIraN&O zK)1>582pQrgbCBsjm-=^Xp`eo1ylNmCFpRcgMz=p#L;u_PY>I>GOGRc3S>SMhpL@M z9E<)I7a=wAQ&DiWM>2j}V6@d(CTciNoTqYIvt_wA0yo;fc5FZiRn#B#%jBp^I~_qj z6`P$HmvsD@lOix*&h}E@2dc~68f&kz%osj#_5gfVL&H^;@1xb7vMHh<6MSOG5^$8h zd^QnS#6}E~IG`tlLRy( z^V`%u5dv^7$Tj1%?D3Jom=)gGT|JJ}cPcE=_cb1tqAM+uzS920eB>Xks9>Ky!g}7z z-2>t!nyMdNWU^{1N+JaxkOwIqpAW}JA#xzDmZ;RoL`SLb!E*9+-Q>k4d4Pt1)sIkR z?#TdrMW2A2PkTFNA&|75i9cB9A(tvq-0y){k|N}%xPJV141R7MdcajHR6d^xPRx+% zY_LR@q3*Y2JWrU~s01WbK~(}6Y2;P9^0&FY&@lBors+#E%0$;P5sYt>dif2IxepWQ z6rWdd3?crM_lk9q-pwgyphX|N^U>}ru90+qvfY&q5=4`hD#IW9eDEv9qyQ@KPaa0D z?e>n8L~UU%J|0ljGX(m>sFGhS(|!60w#Vh@2+Z%5-NhE>%`*|$3H_}+zBa>?7gIZ_ zP#?8db|~(Rt4ucc)58;G2w*orr!(D`);E=xsCPce1Pd`skX&~_@$SfKuYJIFHn>NQH`HOgZ2;-u{CB2i^w2baulK6IgAgF{y#iIjppBPn`t2s4|-Kp-no+wvZAarvu$%_!9YIysXP;C2r8JqTmI)QgV|73$kw@wx%C}g#GI(P$= zlmV*zSA)z21q2cK15A{9z_1Gdhos)#dO3Am9q9#&B*5wjwyc0G1+*W^E=?1s?vS;v z(KY-_ft3;Gd0Ksw*`GK8kh28g9-z4q@>)6Y)*@ed0r5#xE2MJ-*qD%(VVgiW*bRfL z7KGDgwpFeTUV6HX*KhcMgyj9^q-smqbu%^00NCo&!Rp{j^=|$bN>5exYFzD5$7P8b zIt4;$k8R3q-2)acUmw3C$dXIcPGJowKHmOrsc{FSR1+*bUf;1undyG>o=YEcHXIb= zU>%_nGFV#O@p0$EEHht$YY#;8R+xQ)#O-!_<43Wc`3{@32b{ZMm|4)DG(r(gphY$ z4f_3d>OFahwNo*P1PGzYcYR3Ss3$kREZj#(qewU!sCJcs$7goY1TCdZWrMYVE@rm?>x9$uN4OBOG_#&uDUr@-f8ZQ;+|Ga| z;0y6tbcNjk%iK(SfidOZw@#e){w6aA2Ku-g&PV+^9-0IZaKgf7+qX=$lF3VkLB`HE zxaw7Mhhxr(fr(psYpWItR{7c!c5>1iJJe@fZv6@Vhy%K3>af{rQ6P@o0Fyx=n58>f zhAY>7iNv_iss~K|unykBCW%$Q#)fWf7rG-=O6ABwUQ&4SQqLQPoTA>4nW z1+lfW>YeRbC-r45Tu3!SWK!7z*uiv2Aw*!&1B`|V_+~MiZvU|yy(OBE{e9qim^{hC zF2n$g5oSJY6@BC|5Df%{pj68_ZnsrMnt9u28&ulR(V3$VXx-Wrw=S>IEZG|Tg*O@b)e9KVxH z+AYBI<{e`l#5@!h=~~S>pN`0_Wir78IPgp?yyu^~lb9uiuHd4doW&(aXs_~4qMT4t zwi}Q@N{~R301s${Ni6rK>2%_i&_+2VREPzVOrz6HGlHh1w@8G!@qAu>5byaQJ)W1S zY@aFTwnIGOP=do}&L8f!@4W#+(h^^wy{-c@1KqFpPl`T16e#<~oDWBpOgx z^ZrdF$zQizvxrwbOXaIgt-CE4o$k_H$yKxT4gq=I5QaM)T=EoauKMGMp%dJGco`Nh=@}sA@QKn$|=TjDEnr zK5|1Ev6p1@IjXvAiGUMCDUk~Q8iXINjjqZIpr;XPT$^X&0%u@RENojWF=@CEQf=gv zyG9=@4B1+NNP9}aA?i^m3}evsdt0t?Eo3ZOcn6YT%C5eqyy?roYw1gUN2B|0!ye*) zo_Qn#7)u`qtX-9*-O|MZvRu6e`bTew$}&% z0E91NNFWQ{)v&7r273DO_&F+&%fTod%Go*Kt~gGLgP#P;v#(eE@;M`wK#Krm<2s*d z92rky{RPq-9^vq+q|_j9?I5u5M@F>D?b9H)Wt5_P`#!_4nhv=HAeFCWJA@>W!ZLMT z^|0zv?@HFnAuIeNL4Ah147pwR_E&WL&2cACsU#IhWGElUslO5$*DR1)=`IfaeRcW|#{9|Nlrbl=p*76Vd};&{v-*Sc>i5p!;h*x((G!~4 zz0=7ygUi{<>?3-5oPB`I{)4Q-s?e3gJvC!hRIrnrhshH(0keZ)URZ!0%t~^N*I=|) z3l&$>Aeg*N5j~(UC$TR2GqZ1Se13SH-3gWpCrKMFq$Vb-xnm~Kwv=|}w15T$X=$lF z?ysR_6<}PpQvV|wr+iWWnMq9%EEUqHEE|f7JmMicx;l*FtBY=kCjExKKLY`x>`AWE zz}f9~K#K8v{}|S4!tRhb7SPqC46#CK_xJwQzqOl6o_svq#})*lW7oP4SO^;x$I?qJ z^IUDN)}${O@|iWN?DP6`9zF7gJWxuK{FNom?GUcR_3FUmI+zL}j2G6M!qv}KZU-Xw zVWw3QRA%7`Kqs^puZK`&8~r^fbdRkf9~=5j-{#^S*DVi$z> z;Nl19Wk}*q!JJErOIbLa-K#Xh_kty$cLf2o{%~1;rfbVR6TLizIudpMc?f?*$&o9k zE8Jw#CJH~H?Rr-e!xZSC%K5@(UbabyLP7yqX`;w*_~wr@_@=&Jt4J%%c7<19zS}hy z##LRn;?Zn%-sG)IR-^X1A7PYCwPI2;T2Rd*A>#qCar=GslM<>K zljY{QoWd4L8&kHxF*FE9T-`}o&>&@GD{{sgXgfAk>JxfQMSW8)uth((|qb$<)AxHZEksJ z>#m!q7Q$Y4e)%-!83(|pKLKcF?A_)~4{Wd2VsdF);YBBWJTzG}_chKuIszCC0mkbk zaZ!c9)?z=1V@-{JjZ#LO@3F?^`;&NrhQXf5L#e+@D2ClVliwg#i<5|ZugSR&jW&lq zubdcvs2-oxlJffg_)29n&Hh~OggWYm${9A;mYu1cmG78Gs##&|wqy>3gi>Hcjc{b( zT_iTG29x!c3!NQLw0?YjIwz`GdG}ylxw_|-K9F5bFBb-_4fh+hvuReMvmQem zwvi<<^gH;Y(SDJsNg!yS?>oi-fTwO1O)~#JLwHDi%%$0(KT$N(bkP3kAswlfNn{}i z52w9r+4&r^{O)dt!0GK?>B;ExM=@uG`C|%f%>A!C;rvg<%#VNM5WZz@!_TUvxj0Ww z&}k}ef|U+lQdfp3+!Jaza;{N0UhJ&P5&I~AxL_TyhgP6r>Mbt-11o@3!8`ZY`db{S zTP%e2j%b;>Rtgbtc9zaD7lLzz&*~qlT@?$4e^^r3&?;d2mXgF>yYa;}-u4D;v8K!K zc>y_rroV^?hFr=h-V@`D%!_A%U_ycu|JIZOD&wsM3dHevV>73gX!V^eQ^M@(s)$sW zDm25*(ECRMERQuD+2c8_qV-2?(}p_KZ4g0zGLD1bN%e-|ukLpErvW0!uIgCC&grgK zd%VtHy7?&oqp3iS2iM`YVU;KKQgJ8l#F>UKTs{`mxiqlyLJ)`Kj*y*Ud4BR_d_EG) zg9hy-*03SL{fp{#<)LkNhoSTY9$cn5>k^Mvk1M=`PF&7`motv$ zP|~q*{-eeXRhl%7%*P9wr0?z8JP(xDxv00h`R5lYj9I`l6^82pX@6}B9bAN0Wt07N z)t-n;SoMd8OV~^Ub!+f2oSe@?$ab`AhZ9f41ZS-e${D_QDte);*x<^I5ln0-H!gF_ zRB|N)C%)78cyfA|?5B32BCCkWco+z7RZ@x;OO=Qu6BD6p57XO9-bw05UKji;`@fyT z7(5PD>+($LG0dKu&$Bhi1G^_^ATp~N!Q0$9GSsE)p0teTyz6Jp)k&h8*Ndi1arc=) zHd@f_540UxXV>L|wP8+;kYMY~A`L zia-UBI$^k4QW9aV3HY|I7#PMwe;(UoKCNpn=trrp=-sbc9QuM|p&gw*pmWiju{Y|1 zOJZEw7WP|~X`;Z8iNR-vzJ#(|SV@+YNO!m5v0PvOX29;1%4VE0bN&W;PVcbN%fb32 zr;wG)s6Hyn9kfPGl2|5YU@g3U2HsBN{@I0lh5t3Yr>hi5Jtlz&mmiP0Y&ADV&~A&VGT7`r`2a9mS04i0+OmIsV7NwpFiFrI!dTwF@{ z{o305Wp$DLuFEF2nL*oeY&Db)d3px$UxNAjYr9M^I5S-^Q_GsYf-Q85uIv(d7$3Zt z}qLZ8r_+Jnemw(zmgu#iVOt> zd07sL%m6_1%4@&InCx=1VH+1pcWKCi=vYR8h?yfIzYc6-6-QVc%zhuF()v6OBr( zl2|dokTgOQHiRmtL?^QG^6qnf0evvkUfQhvme~;r#@b_6FxxtoCm{)NcpNrS(x@(hl2R>1nyBB2G+w2YE*cTpwsN{62nzJHtex(~t-_^sHcVRJ_ zg}TwSl_h(i9jgVudP8wiLx1zKW>qCWu^&g?G*%l>LjeZGazrwMx3=zX;@?ORI8CyI zfSV%Tx||t&l$hcmSSH?7rl|+R;E;%c*bT63HaE<9gT>X2{`5T7wXPD}AATlbgJ2NH z{9$#eeJM-@@b>#r>Qo=?m=aBT2O}2ft}sLxvuJ1tP(8tXDHRA%enHgXh-HN z{ap})VIF%}{e%@-3`V_tL>KDaVXU+ineM*|{Jt=ILO08($zb7?gWqnM$7UCLFcXGo zjt;L54)35nV%r@B?21O%mDIMTg&dbp=Y;kJYJx;!EfD1}8we(+DZfxS-sH_rxGap& zQ`u;y@r~BV$ddkua{VZ6xk%)+4{_7g%YUonaW0a1%Hm)7olXVH_9KD@K|-Iv#?vU~ za#;0om{}RXNQg-m{Y>WP$_;7Y5KPDI#>>RNk_|cb|GhK4Nm0NyW$$#imzC43e5EZJnP<~vP~R))q3T@MbXVYPCX%uM>Ej%J)es|Q z^oaKFeuK&f`0U6v*B7?QR(#H}C$O6(&}t6)oA+6}f!sQss|+!z~!3BMdB_69ze&>OtB757x?ZNTMTNIaVpyQDMH6wCPca}pt* z^=}w~h45fg4lE2j;g=HYw&)({ia7>|IVMeH=giIX%4EeZc6{DGvC7IhJbTv@_$FQHwa#4kqur=8JdA1QFQ`8VSvN_d&b<$#7SQ zO;(kX@v+?zM%=igUFjTu0_&9tO6itx*!W;Nfs&&UE=MCTN^zgaf9J2%4zKaP}4z+4HcSgmNJyxPK4q6DK4R^qAI_J^tbMXYTi> zueGXg1Qm4Z9`^ES5H*h%9=;g zsw^rvI9zM%rJ}H+PYq!LE;j^P;03fXlV8H>p-;F_8@Gl)iARwLY5p9RuNcNj{6WwN zQm(5mUfU64`ic!h*;8}G(mbrBoT*j(>*JLZu*T0~SS5t8&xiAYnd9D>QHRoX!ga{! z(QEdF^kuX#epmZnMBm+h`PhN4e;#Opf(UkzcJgmbmDN}_-7>;6HlB94<9;+ERug{f zpN5RI?=Oyfi&P&OTF6Q-rz4yToaqKoHv%zJE*aHuH`I?%di%BAY<;Klcv<6i}SxnUib4N9IvWc+QMM6~_n#F`_CDFKxRvaZ-M1 zWS;0U03)4L>JC6kf}7Avk4(lFZ*p~ue}~{a#RZprZih`?{Rre6{dK!C%n zqvKzjU9dpON;ZXhDB6!l4~@~l-bp=9(3zzf_PMCvz`cSoF2=S6rEnwsL-8Jd8NByC zj;N?JsrCgFGu?yalZEt43$I;Gyv-O_17@?`SfQ1xnS#(shBlKm?p#VkO8u|gA^-KY z4=|^x)CHqbZ)#?|>Y}rQ((0Dck1Y1=%}W1}EBc&Iylj1denfCDiIpI&w~oqs&s#Yh zRwL>xgw@6**>+Gw|06K01|WW*PNGqnKa9LPN(`6x5}I^DV#1Q4@z#MR{6_ z>PgZRJze!HtpM3;geH6gtG&Sf!iN)JdQRe;ptMR5XmH_bJWbO_ex<+fA)^*hiBI#tTii9)Y;)?ZmYgnGSc(?<~e`gt5y)uh_>lw^Rw}{Pz}|Sf>1tVB&*fHVjVVAedR^IM;AN7r&-q0MnUxMe4UH*rG1Ww#WLR|n z#4m9v&1QqQX&b&Vw8S}}htQ$?v%wHV>Vg0eI&@$0qg_26=PQ1WO}!;Kt(aJ;DrU>gE~=FBVSj3QAf{a9x>Bl)c#^?84>alO z;;dvkY8J;8c(qV7@8p{Z)TCRMe_6Ez3|L;xPcte9W0gi#ZO)=-Wj3!S6tU9okzMsS zrCvpm@N)8X2oyKVAZu4@U=d%IYO-tM!RBWfJGt=(CoNu>IYb*B6U74cz~z+291MC9 z`$Q%ou?6ypu`vXXNQ~M5WocboCn9BOC*Db{q(>;eDCI*(5Y2q?poS+PptdD+S@ON= z?1Y)>U9rsSS0&sE$Wsx+k4wPuB|K5kP2aeMywED7rfkX1&UsKILl|Wqjv}-du@|u zrZj+`Rh!KvDFCaG&ks(|-E&atBcOm+cd{sGjJcA|?e&6>c+X@nUg>dy!$d6{2Oy_g zb{|cI7ue>PvVO5F(CV4V>LReh0^gPcJjCwhsUMOPImS{^M1}(u&BrVdFsdj=z@8s9 z*-%t;mx+*hfzM&KDj|iTf}{bX3>kHlHC4Cd+s(GK4yMbRt1H>*Ycucd_k&@Wi+A?D z<-U(%E`rR-_A|qrr-Q&OT6j4(;$;_xZS2kq?=Js(e;;M}@jg~^u(!+!|0Qz}-^F>8 zUdYvgRQ_h0&LkyYI6ub%BbNPY{|j> zqaVJ9(urG#$4M_*>Ri2`Q%7z_P%h-(U5Ui=aDS#b;sZTepd6!l=jeeB<&eQ%F3zTL zvqm27(N*_E!2*Sm`SCjSb@={enOIJTkjrd9D=C;~zG3Wco*MH;>lG{c`J2vZdjJ|a zHn7L4xG!p7xi~hrhS&qg{lIpOXQ3El3x3UK1J~m=1w~v4Oaw zXEa&s?zX5S63RF~ z1#(E)uX;)dHtegviDSboXsHHbb5I}9B0M)flQUs3zn1F)c!(D-+G3)8l3dC&h0Rz& zjrK1p-095&0Kz@KRQZ`A`+ITS2Pe0D2JmcK4d#<1n>O?oUvn0DpVKn04L%zFSI5gl zr6t+6!^-8T*sM;u-i~0~CzxC&qu}kK`xHpOev8jOENPzsDTwF4nKJAROc3LXCM0Ax zBal&hSjl0U$?m$V`2+!N@@O1GV9EBMiL9rv58-f=O9*j+c%*$jSpHeI%{D+HZ{+4T!_Mf#dAe>yEwXI58qXDt zFk?)|8z$Yj^y#MK3$;OZBkPF@-+tmqfzJm%Jdr9y7MKcNBe*sp9u)r-KP_iKaK2~? z{OLKSPpDqs&)!X(I;d)VCaswI6W?}nAT$3~4B^ff@BAN~^T4Z2Jh($9!rQby4ozQH zbMibnzkod_Cq7IBaqAg7{5n64={Whl!*ISwPyOt^1}Rcqt%L){i8bGsziyU!dQY9C zsdjD6CF`^m@98&u9*T%QSEAWww}fLq0VMbzYWy7HKP&J<#J~R%CApRW3+u-YP{E?t z*bIUEPh_XLJm1@d>?SdTsCSNZPQOdFe*LOD?+zj#;>MyF2@%g?knjZ2_*(tUY|0d2a+pwd>uAQE(wr|$vKQ4K6Q82qL zdJ9c=aJ^1$(zh4eifhpxw%>i&9pQ$qYZeb_ZjW@4SL9oQOKrrAMsmcxDzYn1p>-{0+Yn4^)c{)9L`au3vFQ9QKxoo6rJ*CfPZnbA=QTsErwcKct z*ccvOQwb3JQJc#nFdZtO zZn&(pCc05_jgF379||qTdzyKXqbKw4Ls0Yhnk9Im>2u&1F5e(h{Y9=YN7LdGx%0oR z>ZOuMolzS7@%O_ds1%2V`wr>%2fId`jNts984hsYc;H+(!5q&yeR&mi!EC(K&lJ8<)i%!`is z9=PR}^MirEYDQMKEpA)(MO|@!Ot5x?AoBp?2jL}+fDiy#&=Fi8GnaXVhe{FJr{;ql zii%(#x02#Ox=@%$I5g(F$%ae9+N-tO6&4sVbdz@b1)(m_xLN<)EDoB+XRfhW@b-&1 zPb!;>QHDSsa7Qe)UM`Xi-Kh_%A+d`TAwKlz!7jD*!W`#s#FOb1ExMF??z`&tJ){xc zH8EA(Ze_%y_y^P`ODPEAt)d#JAdPLwgyB?m*L;$}xsWLU3Hb9}ytL{D1Ft5kPh-jn z>z4iPPY!B&lZq5^*>La=`@qJE z-4X8^UeR^+18TNk^`}v$&w|0Bu4S%_&nS-}66l!97UrsLqPzdFWa2>h#PV%)9;H3@Hn9T<2!u zEvt55-(i=|o}4^RAxi!?(rJnoe_iVh>4QPGuYf+&>=|KwZ;w=>u`66Rh39I7qY$rI z2D;m2Vcbw5r*QMPIbS4?cm*MOXI}=Vu^(HOlRa1atw3t^SpmA%>(Ym)s38sl{5Qr; zfJL&AwsDpw7Pu>7bG%#|XNZmnS)?9xW~ez~t8IXVEVwz#D#`yK@4Se}KIcsg`B{o$ z$UMLmGY?P4PI;`z5|2QE9hEn!=&E_!M3ywln2|UJA$f$k=^NA#@^cdrHGP`cJZo&K zKN8zJ1>Jq{4TOKK<{rh?;;>}K-o#*~>#Sk^W8a9GfoK-X*#)URX^L|~CMr<$WU-bt z9Dv6RqUnd?*e`@nFzU!^+G;s`dJ|x+jLgj$-@dyFtq^MqYUgL8cD$j}Eg%R{6BT;W z4U=jOgu2XM1zo%VRgnsu7WC7fD1=Dllt>x?$8qMGhz$LC)j=?gFlryB8_fa9G2QkE zM0F|R{Ykm86ebv$F#H+HA_*-39%9Ba07r`P}CfuD(PG?edWSZ zK43xOHSg+GZ<>5txzmK6&F(EgBO6SPp6QQ&r{tT09?F1wX6g-eFs+z5T8Py&Sldcl zXniGb9q~TQW!S0aFb%9thq+7hfQV24kZAcv?56#U>z+T&UP*wGyozzF&=7#|6fs16 zbf8|O&dxi=N#`6fD-qOl)^NYxIB`4#F;*@Yv8c{8k9)jWRZ$EO2?~@#GPdPU@F-0^ zxJI?1x!`g-s*r)IdmNuzD-+#gFQwd|YLa+Tf+$RW=?&AR;jX&u5M)(P_+$SWX~ZKg z%{zwRH5f)DHxL_09Viot5FVf1>c(>?0wBOq`D<<1PQ4PhkVVfFhbWMR&{B_DQIDI7 zQm}XH{UnV%9t`4}J>v}#j%U1oyMe6kO^uf6X5&TQ+Yjqr6wwtMVE5`UFvcG&nkKi7 zUXDb;s`l!@rp%fM_Q;6Eoq8!bYc4@gnqg#tw<(HaDTyP)L=p-Zgmp)Sk}4kQKpUwj zB(biN6j!+TSIhDqD&Sk{f92#JfpT#TGT9VTdj`BOP|6?&mcwfc}osv=khBrJ4|PzH1@r;RL(viF1$d$*2`ir4wq^R=k7SJSOmy0R*L$OO6%MUXu; zD|4@Zz;j1WXI0Cl+l!n>haKxBO>d-5<_t><5k=?dJ~?b2I>#@xW zzFhUY2bxZIO)B{hu53rx8Vs-$jbg#&$MK~4LU;w vgU%Z6Xpa8Q16R29r`XEh- z9~!WlfxWlb>w=J1x!mz!0&&bXZrYnH;hP=g;O>$Lc^dpygiw$>Tvz+B2rPEC>qqe>U_Hak~fNN zbx2K02>V%DtQu+-Ydk!TVbo-%&*m9rcfwB~IjQ|muM${MGh)s>;ENyZP&*(x(_GK^ zU4(s1`u9uW;3KG{Hv2-g)z?4jCf+e4GY=bvk65Tx!KzOJHs%c&>{UmwnHbk^C8g`i zb#49&jpgthSl6MkU{{0UK`uWF`Z-=K>$2ys=WuLe+r#Pj4yBsne@!!Ux;jGJ zdL=f$`^`)1kMb>mXhU23t;Mb5bcs{>5}cV&|5zO24iC9$TUeOYrHEno^&XPN*w*Jc z=*x?g0OVBMbZTZov86bzP)W{P5e}9Ks4$|ob!_@8tWgwY3G@VEXX=}ZNy|6C7l$8~ z3W0=^&nlChb7}o{ZO^48DP$@SE9NtkXYs_73j6=8Ldu3O_e@pfeGqLEzT~%yZ#wL2 zv2z%BqSQvHs+8C3@Whh|z{zqiLGBH=T0d-CntIT0kLz5i7Sz-4rx>B1*)7tWCu;Pq zq-2n1b2|=2c#?449Rnu!Baoq?@tmp0Gx-rn50e|*Iq!Vk)Uji~IM~g}0io_nkeJ2< z76Xn}OWCs=d`zjZXOXRgmPiuC1J6h#N?|_ZVT&?qrjmRF@hN}J5W?3j9j@cw4eR)& zvXzwD5i%6Q?{*xqd+6JgO%=>PV$xitGXNscDu;B_)JM~yC3z&ZQc+>$su{nnb9+`5 zUkbfQ;K7CsuGrywZOH9|A5gK-f4F6&M%I-JFM^w*Z=fF>W;)BYZHJ` zNW}W7Q@GCW1MVGG9;`aS6ZdwEAyY^5>yd{!6Y2B*aJfmiA>*)g@&XGrdX8xV%lA`X+H5uLgM>-xa=vyzPRJxjY@vO1hSdO+ zEyI884HYNjmHRf!7xLlH2sf}!nQa8tf{htbxD<3;$%zwNPleY2vk5r3^5Qi%#JDz3 z+*l9ISmzDKZun#rsAJN zHn3}cbwyQQUyN%ysoF@7dvSq^@2OV?GjNof>X2%6V=1=EoScQ?aevIIiHR*xncv{XdS6PSO9__8Yq0qGp-lxN(FKzyTn^#qJOz>}@{ zLpw>+{jdWn_>kJKiH$Z8KwlA_c;0DOa8~s9egmQTL4iyF*crZ9bTq%y&ljJ9=TerM zaJ~giTCb(={G0P!(zMKmKZ$+=4&p`YY1Zn{qSEzk`&xUfac-)rncMbC0Ddyp7mu=I z1I}|+=IKjL7PyMdkjm6@NLg^1@1Zik^#p;NFApb$a>bqmHAF<0b^x=5sF2h!t;MJK z^XIuyV12`-TbTc#!2_AgGf`pxZB8Qo7BblTrM+KU}OmE88CV`Y8Biwd)Y}*$bVz2muyKj1?wzWN1SBwOux8bR_x5~E*k~4o#?G3j*>sOW4gQ-F_X|HhfS~q>$8wB1cu^_?lB=7KE z3t5w_+fJYKz6~r5JV4N76j1f;Bj{&E|MCi4aAo(g)qE&X6bNS3*TZ34 zLDkR9f}^VMpYynCa7ghcO|OMm2<77&JLh`^`3}-t8$I39*gNbKp^n0nmm=RaF`)x2lT^a1IfsaMp(89lz7z} z?Tg5*))`b@00-^YhV}=JXDHY#*!!~I5xj({B6S+ z)VW6PIF__Ku~o<-u?LF7e$-b96s~9A*wIa2e;trwu(QL{psLNeKPq?sXCS-ZwbfGR z-AxSM-+==sJ>?XR{C{#PeK*Wyg2mO-AWxMzT7=dQ^^*CDZmELIwSE#9+hL|Z`mhSd zu@H($(KOKcJ7W0a`LvW?#AXc6_+8~yWT+gzdrxoW?x}7UcVlUbJ_#1zwP zXBQ6Hf$k2)q#XfX@y^Ct+}MGb%qBQRb98w zD;kKAmn445l;Tt~V`@=*=*T{5A_=atX!428Gpi>gX}A~V*7)|sj)PIjij+97ziW9v z{QtS8jv^=h5rzQ(5TXA+XiYW_*8iVk5R$#)w8fFwzxSlperjS{^;()ZeZ9M?x&tiw z2WjjJN>Ia9SGgHvOIxW}SW2=P{;f?9m>|Ly;E_xP^1;fxXu;ec#rNYR4W9k}?B>z? z_{QSB=78UW|ErbT*UB~Stdn!)hQR1+&&!W*oFaN`4xj($yN}=R>#HqmEJ`LN&ZUF%;qrYgE8q#+PN}vj@Vm6p;*!@6 z{AVr;6grtF_KU%!fabklUS$WS&*vR<)3vkv;o#KI?nk=xYh~-M%gIYqS$5_2qt1_X z`7YK@#f#Bm4*q2#_^kKrW;Yt!mwSyIna*oAsnV}j>o??1uTCoIR)19j8us7q?IP&r zYdwu3gwR6%^N(s9ob%b~!v@0mJNx%?V|qfvRugx|{o>=@95wZACuzkZ`faZ!{d&2_ z#Y`h&l$Pjw@m|;L|JXXm7*C=oi??mtwr$%s|82W_+O};^+vc=w8`HMky|c+~vb)(& z^`VlNO6u0FllSiVoz0hJbLQL1-997=$i@>0pDwiJNJNZK>tzHvA*Emm1h-X`&g$$i z5G+OzjnBXEaUeyl0?X>EKiLxTVS5W2quvfmk_ggIO#*rGDa$p(QI)=Bp=iIDCe1Y~ z$Ezc9ul;^X&&lcFdr*R^Zf$OrmYPkueYcFd~mVuKScltK&CUnAs&y)D=xc$lZfR``=-x3A$ z@-dYn;`9Yq%R8*C7qan#PD2Ksf3X#Z?iQ{NNdM8-0ukw`UV!*(j`d{>Y z94Dwh-ym3<_eQTjwt$)Bxv&YL=g?6112{j$Oz1A2jmS|QwOa}gfWnoRDS+|5W>-#e^BLq9+|6O9EKy{o zeD*JzEUQwWaAnimYcH-rj>&q6RN)tS?4}H5E;(k2fs24j?gqh(HguItr$ z2ducp8--W)))ufv{0NB$8t_*DIlIB$swj{%l{c+~pCQl?U~!h>c`g;ifFEa%wk6Ba;#PCbhD)EVK-gt=_zp5*0~Mn5c~t~vo2Mgs`x6zuL| z3R!=*nnV_6X^EjL%NiJ^4iB;djI$mr1B9*L2v*TL@7D^I7UAYnMw8D|eckpn?2|;M6^$A&Pxgz*J)RtEaV~Gj z4Nbx&c~v%W_v(6`&yK8}z3&JhQ7-?qu#FdNZzbl?i2O3(B$I0j_#P6jSMtt+1%e0F zoUN}GLGqa8f?~jjUa%H!+cT>D!bS$;jE{hwGXlYyj`dxP%Yl25fFM5y5rC@~)N<|g zcJo;|rVnh$v9Q>9k~~{mz)s}D%atdVT-ON!EyEtNj_1ir#Y3ieV-0@MkRzvBv>uAbS3rD=$m} z`GZ&IV**4T$J>J^m7q5JlsDr#+$H)Ywn>^J(#0L5Qxj+@F?|N1O9IidlMN-03EdAE z-4?o-@>6p_(~FD8$4_YVK8j-r9UOoU8PyLkc8$1FE(6UU{BGi9>X^r($`OrS3nH1M z(ErU%!WbBEOy7>Z?Z5cP^6k7-A}`B7uy;+ zubsw?K#^)-CB}87@RWQDlWQP*aJU({ z(Ikzc8Q;xJ%xurfr;n4Pmv_fSOfS_gI1>y`-C#8MR^3ZP6C2DNIhBV3H_MV*>cOQ{ zz(c>1%^4c|Pzv43q|hx6Qw;W+Vo6e?jPx%i$p{?D=gHci1br0+W;hXrT%!l}dqvHs z*!$9&c}gA5$7afT?ahe`dw#-RfS?iVKx#B(B}3X8F?)!aD+Oc8+EnH;kf95{y%vDs z>l^d7I9?XYRW5l}2;z}~5r)3_DaJg`8 z_+AS>dnshT7ONCeOBSW9E3X-m^fKC6gVT6AlaN=bh?UOM(7oVSvL+Yt1S3ayN%I_h zELchyhJXP!F++)mMxaZbDYzBnej77GZJh0Eqx4MsFTQrt1)S-C4<%`W1L@?u{Vk z3qQI_zkn7U0tS`H4ZDMV4LsaGk&lDW&jIXH1)&T&(5v-=Rxxyc;~M$-1H)}Aj|M>igx zOdaF%o=nRKCq)D~3r-G@Oemu58e?3c!i>jsOf1mF3iH}QL5Qd^vx49~d$>4b1Ahs! zzyg=#HxyQ!8N1^=Dewsj0XJnumJ`k^k>#$<=EQ3}Ce9M+Zk<`UqUc+IPIMY+ zq=PQZdc0?P=@_q_A>G~pWW<&0H~Z8E;%1>3^Vxj^RUBLre+&hF0ikl{(A?2z_5FGJ zvQ19w3dQuPAXKoTxU`VIE=9y(7$Bmd@1%}mda7C5PnTEIan%knK0vRcVkdX<#KhQ{ zsFcGaSMHdMBBIM6pTxuNXwp@sWk-b?pqvrmN>e~F< zLGLK$t}fen(U9vk*pCx3q%#KHpgUq3%$)rPY^Y|zCt@ch223>e4aDosSKGAXe3zH*M_1YAfo{AxuDkUFG-uao5jG?m(B zqVGSIRl|7b(xKybFjcr__AhkN?T~4}CxfkTBTtwkU-|-KqLhl~zsFVJoGe=pw%-~) z6mfb#6lr>OCBSJtJdmR?OfIJq^(9`6jV|sY;l==-#{3m)_j+%9>UT!m7= z_h6403NL?b+`0wwOSGnSM1$0JQpWyecVKkdrL#oqS@*@N zFr~oJ`C6rwigmTHwJIP_Wi~h|+=(JJpnQCZSX=WcKeexwV&j}uONpirhJX+0V}m0| zMo3v^{W^K$hWUdKAJW7`Z2X)+*N7G93{Q%V)nPkv^fR|7SY}io;UU~S@Q2Alxtcyx zC=wE4IU zc1@<==$Yh$etp&QNu5{7IO$?K7w+=9+mqD-6=TmUW=7z5He@H3(R$ICeQ3{SJ!H(V zN|vxqp&R(IGE9f$pOjKrGWZZ#A`I7K6sqDxpE<#L8jPEv!m8ecV1?>qH-xm`Y<^5_ zS|mCL!;|tfyo=H505!d>umPuOEctZ60O}Y=75nxJHVd~67J3xEw#=vlnpO`?HPku< z#aAYMdKY7MHJo&>Hr}OC_(Y7}@-@lKr>>5_c}?n|0+=|}!-&+CFTFZdlUxqQuXU&& z@$XqnRdKM4Xhi!$JkAd89>%Ng(xuY$2R|+vzWLDlu+l`@2_86&!nNtTVsW)jlK@c^ zdcs;jLGm>QB9l*gCVqKL5>o36!k;B~Zw3$!A~Y8=Gr8Kcl=62nnV?V(0hVeCLh*&- z=7;-rd}z`-_0StqHX@WiMP|!DSZV{b1ajxUqY5M{0zv{`=0rH1m@g?dF_f)_(MRSg zzu8;?P4c4}BFf;P5lNOZ5VDn|l>>37V_nGwY*|gxq=ekvN%OWQjZ%>X4QbG3wdPUV zxWcCLS-&HTmY_Np=38u`drLXvzb9l>I|RY#B}GZ@D9Vu*=vjOL=-|D#mL_On9fYOc z;!y>e)V9K|T+@#(?p68G^-DO=3OaXtHGJFcRGxFK$b6M!FESl(v0w6Zv0<__9`Q&B zH+ICQnI{GFG&K0y6<8ttynyl85vBQ ztf+AymGv+eIF!-3YGxI#YNEm=DL7Cy7qfHO+W&&!~`u02AiAZ2M|Lr7d^ zUu!m#1Ae*7o%o)iK?I{$JKIj{x-=>2D2&`>d=m-I|aX?=wXf( zkytOXb1Qb3$r8g2jG+<Y!$Kn#OLB((EoDwUeIcw~8IkhVQ7@uqhNlNClv*k;j@ zsDdt(smp~n4dsk7T`p=uH8-g|91)(GDJ=!~Uq@k-&Kl5g)HjmxbS-)uupwsFHm@MY zVT>d?(x911V0*S(x^Nju!!tQZ0S3v?KSSTS-7-UbFLOB(TXd@3#l_sbhOjVy^oKa= zwaIJNeK(GD;yv#XG73VNbs~vw&q!g+t7eIz{oW%6xCZ$)d?|=~EXM}~jT?Ym zn8Yt!T3=D$DR;*)BQR1N)@?f%{D$C$_hH@pC<8q!Od8Pd#sN8hP3q~Z-IuP(WWvJT zQDQBQ0<)7<7q$P=h~$oFt6hWK$z_v0|GA@sz9A{vZz{NLHCY^%6UK^mo1C!3re`@X zvX10$zPE#5R&4^~qq0?H;#}u}37V_yxH3Y8cB-owl5Vh(2OAnghIXo=gyw)pRh2x$ zKrd&>&v|T0{=P0zI9&-6!Bq=6mA%&nnoQu5WbI%zg+NGDbbnAo(TNQguW2Uj8i~mx zRa|1%Gm+8yqApm@1<~oDF3{tl19zHkVVOK&nV`R9f>hw|;EeyW5N_3LoU`8Qv1!%1 zK>!5uCPQjc7?u)?s?muS&#(mDQNO^sM^6XCf4ND;+(g`60OXHwVN^3EGriHB#Gy9O zd8sxkEOlx*P4bVb1Cs3UDyVeD%n!K77@jePMy{3s2&hh6ii zwBvipptZTTh>&u6%*lP+raWgig!*i5=3S*o92hoh?#YqLP?J`#>^K|xqz0SeQ+N&< z12|qMSO$Yim8M<8!RjR3Gpa<5B-@FBlfz5tgkUOS$A$`*!Fot0tg%KCBz#Syy?A~5 zOuscCWRWRu*TMR-_YhG(x50kuWtd&$r<_q~!?f*9(yCe*??SB97qLQb=fg=(>rFmaK;?>CIimLQ zdr9g)C3<&UXxmdIn~5vKHd5zr=qP#e-LpWBF_i@*A%7ywE8bvSOBUwK3ICPDb)fW~ z(E*zInfPhx8dd#nk^+eOsv$I(X%ad2#?BO;{~xvi7P)(@H^4vlDELNX6H_<3YGO#@ zS%UFjZ>G7@gmFX$_mV1!z%7zJO1NmNxZaYIO^}IJ2hyOeJ5r%3h`wgyNJ)RstMdUq`>v;*_~dO7Y#ZGVY-4PG0FL=LXO~ zY0_>J5?%B;Mc#RB1i3!=bP#4@Qbj!RagTU03PW;} zhHG3Hx3ffIg_)d-n&Nm(QMYSZo`G~oVOqs4N}R|g5S8>K#mweuw8QP#n(F|&pCsZX z?BFzwJEsb)oweP<%%(!1qfvlzd5p#&llp7ya+dg@OE=orC#>hUj>;1?!Fh10U^771 zVE}Xa*LZu1f48>aBF6_K44Mn79w^zt;*m%5Skt_|wVIEl9WoNIp3|8liK4ZhXk_P22hLbsrP%fC(!b z$(SST6A^hD1$iosQUyzepOeKSDUQQJZqV8-Q?eJ)D~)|CJRR}G?PK0`cc5b&y@Ref zJk&m|NUs81Hq@VhgHVnrrYM^|4wtF^yPM{t>07*8@xva!rLhk&;mytpAnYx z;|duVPk%+Uf`qeVE~H?g^Uq#&<^ML$Yfp6xymq?XQ2n}^YGXV_VZYhz;L-5iMq%M+ z^sHf=(^?wj%v~?4cy!pnJa}Hg^t;V6E%z6ro9M0TlK6SVSq02{2p$^>FwFS6th>LP zAouayqQ)`!k8M4$cB}OEhD79b{_uH(%y7D5hp6@;~~y{v;#hw`jOxyYPW^TP6Kztn=)9UhBXlNqSUS!@HF75;ti-Iue+N zP&I>Me^zNip$^PKq2`nty|`iyr-mMnLPNWMV49QIoi4pprNvGj4gMvs{&ooA`NMpx zVbmcHl?Y{;$ypfYdK}wQudNM!Kv%3m|oq^gfprm9TW5baDRsk7Fm= zDdw|u+pgwmT6goi!B-V&7=A+Ub|v@FX%@CsR8X%snGu3ia!93=9op4Mw{?4IJjwwB za{oCJaLx5tRol6Wu;?*th_k_!;r+v6GQ`&>u3z=MSFyiwzK0;f63%?01i46?!I+Ag zYkJggrYeN+uw*XRgMR};b9{36LASR-_B!V54x2LJIIi>I%qI)mlB>~Iuysm8Tif232HeD|LfD`x!|MV8m=6<-f}P(1 zpBbILs2<}1U98*B(D8@-rne#2FsQ$}^WX@E;bZXSEcjP=oi;ZbEAmQzb+SMbESu9T zveWE=!q*)Lx+#X;{%STXV(6S}6N;u)7^X)x>J&pY3L1mOKhB<8nx=PYlYgQ4JR0<~ z&?jD=$6=o5P70;P3wf1q#UYxFr?Y71YSsXO(ZF@al zrF64l6mk|^VA|GvNJsqnKdfD>QD}Poe~{g4Gaw+6|ChDvWMF$r2a$|zAw zfps^E0G6M@*5`8SdrojBPhtGBSTbI~>vq04{yV`gzW2#t#LKKl>Vna* zw{s%|K%;s(ayGgysPBt>-{<}PxE@_);TKSPom#PVyn{V|w9`}9SKM%{XxQ8C{q+{T zs=1zgInd_ZtD3ra@&Y;sr_NsbVV`j)>axAI8P+dreGz>5kQZiYX<4@N(li3l;a0^b zY_4?a>{iUZku&vnDQ=o*7ij7d>ib@^mV82-FxRcY#W|Zx$+QZ{n?cYoY#K=8IrLgH zG2)q{Spy{NcMDir?^)<9gm+gZ|714E!G*aF<(B9Z>LpVpHV z#7om=Ff~GNO<^0meAn$DlJ$4ZtKYYI7*#2uR=dl7)PluYQvRrwqx1kH zooD-?p4P=!)C%7>*nLW>)*>;gpK5-xz~Z&HJR!YwB_nPALpE%HHS)_^T!GEX9ZY@R zW~gBb7^am?Mg8A?^D5cV``R-YW^1MEJZL-;rsc`1F$Ltft_#Sv&fn31inx!p94e%T zk^S$7jL_==4AfHxKtnXX4&zin`WEN>$f96Uz`uUJeh&Oy*)w z+{h?@f-ZqmCcH|m4)Kvpc|)#Aw0=`i+6eJVPOFJ-SDP_y+N!g85=FBc_XXX!?nVCE zT`$z96|#=?v>^|Fuhh6RyAO;B!&Aw%ImIn@Al(=?wpKO9amR7^2=Q(rle*LJ2whis z`o?l%w*omK*+Nr}2dC;AzNCl<;j3jNc3`UE)!B0|K=G>Ca?2Xdsv<3WuoZF5K(_kx zZ8y{pW#rl~Ca(GE2T*D>bF_&GBd9S7<=SG<<142pGit0mmx7CT9>g$idcj|$f-gxqc>`zARaJPJTN2^ZTjSoYjGJL$Z&C3qpKAf<~AObsk zdAlW9RQ*Gy$^X|!Zc#ThLp-qrjQSgDPJ%*~%3@nKr7X3wg~?;|GMnE;PUMq2J3?0v zmLgf3VZ+%4LX~TG{#bjKdmhzgQN?d&y<3fp$P1w)&UxF-om+Ff#qXvjZ65wX{47*S zAd7iY1@gi>g@hz|ri=~uKldE3U1_H_Mr6-M8s&gPDJB|T`PXr2i$68o;H|P37ZE?K za(oLrlcvyH16pwL`Dakb)+ zJ0nLG_R^tNMSSv+{BtjQe$t}ins*uQu~5>+PyCHJbSOI3b+2m7iw->sK|Q0V#i9lp z!g@753O#mE9!edzmgOr|I^+jJqaU^BY>OhJ#_*_|rFP3ULYW*r)Znq$Cqhzz^Hi(p>M=0! zL>rbF&zT0W=$fNT%LtXSwhb|)S|9!XJ>Huh;>WsX;4aLwyi@YKl>`H=eO`?CVmO78 zwfiKfYBR)0p-Gfjm|R>6qpAl(M8Rwi;TgLrRE70ZRisem)ZDRuQ?()ei=3!&vg_1s~m+96;2MBdz zY{mmTW=;~-5ehX2$OG=7dLQ0}PKF;Q%EQV5lUG$V8XSeDoCw^`jc9hKC)lxO77^a2jQN>ifgQ7a7 zOdEI|2CjLR3C?LqATx+BQxH60?(ldK9a*`qv_4=)qj;&Lz7ZB4W?TWgL#}cfIi`+%67!SDd3<#cdrCI(mi9*Db*eKm{~QyD){doD?p^36K0Pe8 z3dsk#-HAeMwe?Qo=D7^uGOKp}Xjm5Ktth^{#sNrWDqmc?A)r))kD`$+rS`@lR*x|9 z6wM{jre9_{lWqCbrgi3B|5*A1Ij773im#0T{Y;Xh5*IU8GxPXm$}*k0*_P))Ls~%# zIvO<_IsU`sgL?z(nuEq68C1$}Y8Fk}_lxR9^%Gm|Oog$-8CH=Jj&FAld24cR=~D3Q z5(u2aU)U`aFUQynUWaJr3P{7nxPe;wvrk8S^3%y7kGh+ej(uGUm$q4UtL0r&+m+7> zYCJ1d=X4F$*rF7frukrdVG=M^cWa^W(xw?m9tmNIO7u68bkhV|!_)%lkAZJ#rM-zr z$aV5e2-mw-ruCWR;hX46Mr%=W#@rB4+HN9XnnVSm#F)8V*Qa6+k}MI{#`}N3$=edS zLinO6_DuL7i|2C6^}JN~PjGX=3tR0bB7k{a`-4yBN<6HP2Y0Xval#<&x%<*9U&$y$Yb$^v*+ zS5J{(?B)N=(NVgzT2qFM!P`6T+mZmAJ>f&|p28PxbRaTLSd|`B`gB{)n5LOWT>S17 z;^u)9Q!^juaTebaWFOVVHc4uOxuSK?2}d_afqns1D7-cqf^E<*;SxMGpqGF$iq)nbdu-|u<5ujkhjI7HSf6&UCrxR5`x?&P)(VMUZE zpA?l~C$YPGto0|L=C%%fH;)9tT9rLlQiLnW%EM$FYTH2>8nKme^P5lxD9cmMVxA4^ z)fX^QIHN63Y>G9E95O-Ha6@){N{?ZQdmzfoj2mW^7Hc_XV;>V~wad!pz zGm*p&w7~Y-5;0iyb1*vKux^9-(;)^~4wmtlwr;Fr<2XP*W*l5}_B` zRs^w9!m{x}TG=?-iNeW3{78v)l`vD~#VvTKxe_?*dzJ*HSjtt{U*!SRv9_L3G9WRe zp)|A&gj3bV_lb2H&SY)?*LJZ$xP2;9tZGLHc~&D{PDfGo=E0oD1|sA}qzrajV)sZm zU4a_$5VE;b2ZQK95%f4NV;==<5;?om9Qgu02)<#WbhdB%AbgTr74S=bnPW&5Z>ZI< z$bEXJNP7r427zN~Z21Erz15VgfBvF91$gEOYp}PXA0V#9Q-y$_4;lRmBpO%QN?F+E zu);Y&fEBOr(-n+d^2i*Dy?HrVp5^-x7;{+{oAs{Nh!{_HoO@X8%rUbVh!DaC z&@IlwP^*)uq9Z0Fj|^)*l`L<0r)LBeBvc;Zxdz>v(1}X}2c96NMX@?t6Z^03fiurO z(qG=hVPql;kX$3ZAn5zgI)m7XoMhCn)$QT^vp`3%H3YW}$Z+X*LQ^DJk|){$CKg!v z=fIFkdQ!fcr&wdM2}+J&ntN@yh~~&cix9$C7QaK1^N`AdqwK)XIYLMzBBs$Y##y?} z%QLV^&1eXJ3x$|^-|LJZ34a<{L8f>5MUR-^D3}UNFaxkZHHsHH+5mq&;c!yBoAB9I zxV`R^8v?ql0|rV0uIN2^K|ls6iLu>yvdtYK$=M$f90u10*~|Aq!WMtqxr>sj8AAou zkwIDebSBS9*AOH{9wC{`4_rOem&(Vi{>tIrPNZk0?Fwy30J;Q6GNrz2M~w^qrJtEG zB>)<6y}t&*OZ19c1ue{5QhnfDEc#GjUyvL8Px~SQ{P%DaNvXByJ zE@gl0I7a>%Ur?ro+X?m{NC)~i2LcJkk>y67V-ks4t`RXrHpf!_FlnAGDFjcizo9Ie zs-Y2Gvd1u_HQU6|4ul1-fZ*rovliv1=0uj}q*8P{*$4Fc^x07&DgF?*L$L3nNu!3M zUkbOXr6ncQpH>e?RM?=7Ygb@hi@XDrl#kS~Jd0JV&iYQG&FTX1a>8n%V_>g=F$?jB zp1JyWhZb=qc+?;i=zguE=8~b)T)@OwA-qPOfs9N6ZOpQYV9k1(9XF4xm;eo9VL=m_ zXkJi5FoCAmToB`)2PUi0%4Sz3~DQ z@+1R-)$o*mWl6FCb+(*2Qmnc~K@hhp{X=6qPCYFq?h?gv÷Uj7$($asx^rbR-xZS?EIjP;bsS8g*n*r!C+|UaqDxz8Rpp<)D(4SNlWfG?S1^yy|2(C)uLpzfKFee;#NX7xYiiARRM2+t z{O2IBTSg9lWXSS5FqpULGl;*BAgMMAg?aq3nuG#?Sq=+)D?E>R)f^-Wr2|;9#Z>kM zdBHdpCrLQ0{~)84r#u72HMQdOyWkacv*ivA1qRXlW&Cu*=CloP%EK4DD*#3Vz5?4F zYOWwQar)+P_#~1&jW`PTv9qW+0q$x9&u}1(kiaTTK$k$W86t{x2P*so6f8#vXV&&e z^b6Hj#^-}mwFbh5xp(!|kCfg+?LV2q4UOH~uqpKjrBOH*EcXR=mgbsr7z?8MS;}zH zcWp!4(V-yaI#5>*{&LEr2rB zun0A2R2hhP-fD-i`J@Yjbb=itve^+lr3*1;A#xvPde*Fgoa=B`TQNNZq3GcEKB$u|7>Vg~@A=Ka;QNt_( zRxX!-j6FX>#!3nKhZ*_L8^9*wBk;4p1ox=dt&nH)I+I3TVvDp7%U&_pwD-BtiJe%_ z5DNIWiEB%E;rJz*$F>KMz$hl6{@S1h^U>S}x5H#PsL5^*S2QF86KRRiVx(-3BM_}M z7LWe>+?b9FQH{lVHY1rZb(~1X_zd&Aj&XsP5>1)+ZLplmf5EGAyy8fxRqg=6zuVpG zQ&Q}Hu1ltK_zbKv9oN3%jNsj}i^09(fH{{QEv*Jv?%VXo*taO2*#!K%0rcElXH4fh zu?n;vS!X6>d%TaW(bL0aTbG!?ZrO#{okwKs1yRKe9|)tR#=p<&1zmU&UgsC- zsx0{0Wi+3R=+PfN`3*F`3kh%KPbZsu%>OzF$ma*6pW#9#B*5pr%ZgY<&Nm7=?TbKR z6pSSTOO$LwDaDNm>c6S^9JD}XV$=O0H;MVM$6(gVncb&}sLwFqHDn=pI+(+PV<%NI zsEBmHf%0s64mgE6Fdv%>aiJ1x;9S7EpWkPlGx=PYyXK8R!Na>`tv{yDyg9(c`mNZ_ z>B&`9^UdSjmhci`rKLDMuCwsTVRfHq_$?x^MXUxf*-s4V5R&Krye( zEu(-eK)7gdAl2avn|*{K_`91Q2y-i`1Jh37TMB$}Z@|>>I_DbAuaroS)WhHNag>{1 zm)k>mX0Q$6b_@3crJ`)d`UXA=whvxFCC-+ki|!?ea<++5vGIZLq+I{u^OL!cYAwmJ zcG=W#Y{kdq*OMc3p4*1#cV>TK6;0Fj?Y!25PkAAj@%n`~V_J4TQ5mT|J321m#QkdY zqH0()CnW#(`B7V4p1@?E?~}TODF37H1%cqNcFJ>vO?bPj@Ev@@Z(pRCn(*9C-`BI_ z`3{G@DvKY1m9R2_shPLot?&{z&Tyscw+joy6GcP*kI$#3#UprsJyZ}RPH`tc#+clp z@z-18zPjqF?cpQSd}NB}X(jbcL=W3{A=KR~0E0+DTSvugq95krZ$?hKGDZf@Qb3g6 z-0_C5x1S(vp^9P|R@h`m#o3vb^~8s#i@lGCfgJwuy(&>mHIu2PwbV6fT317h^-CwCz2?%>({HRy-y4bTOmH?o zbfp9+WN>9N0ym>K%|p38>aAvGW}jcO_yUxkd->P=Up)EDPcCZ%=?*uxy5d}7=DAr zEypj~WN;d}EX~f+DYfi`)r987oEYHv?waTPx&L7|U-bH62a78*PHh%Wln-U`4~};J zaVWdZH;~)nb-Qmv^@@x5>{804fT=uMUGUV0`;pP1)H!k>@fNwrpYi*=Wi5I@ylykV z@&8!_C-!T9?JMV%d#@|#3nbC+j(v|KxLxZkLw~Q~rRvJ1W3Un)|)C*d*)M;ZG^O?{<@H%c%qWJGbv3y!Pwi z!*$93_t#Hd|6Rfoi4y6{Jl7If4UW(J@0)0QkMBi865#J>j1TCqul)ng>|MWS)N%%Y zQCJ1`-kI-Ye|>R*N?+B&WqD7E%OFUv%v#&8JofEe!4^^V_r|G&Oluax^lL+}{JzH9 znCRf)Vf#lrD04HE!0n#SPM4a*vd@c;(WjBcn}d7F#t>5(s4?g6Ac5Hn_{9%5FY1ZC@!^g_dr)oD8utVAF;w zi_*y|#VZkfP5;&(vduNXdmtk1{%s3~RKNQ|(8&yix9yoNFQSS%xFdb4GW`~A7f*!TId$|wMCrJ7*cJJj|xwEH#rZ-utL@7GcPpCb>$_tpLM z-G58(y7j$1Irx1X<^O2-bzaVY+)u9hzsL9e+&`?=8BSr;`N8k582#gQ8}@#moz(Sx zIKzmx#MidD)E}Myyg#M;zu%Srd%R=KGxxr`=-eyE2Eu2rmkXrc&t1l!s*rn+W3=z< zo854Ncu5kDN|Iy@S3UXNOJ4`cuB=l{p-r(NiRH19eIA&(+FePHrtS4QR8bF3#P9F0 zL7?z$E8hI8s1i=o-KVpivdldZ5f6i}kgc6dJkhr5-nPM3o#xzJe8{)tCS6gn1QJWa zC)=Uo!t&VbK^kUSt#7VH6sCpsoudluC{|JbmgF;u-Rx7o4x=@UE^7}yx8TUDLU>Nh zEOzs-wcKA_W2bVH!G}=c-@Us4%xdkAs%*xk&JA*NZ=43^0T;ZRV{9Hlk9kH1uH3EA zh*+9gGEN8j&L-@9L8awuSNjTH$aE8C0o?wZS@S5H zZcnU$j^{}%d2jCU6lZ&UeWAgqQGG4A5;2mv2vl|4Wcn>T#|FMPbeeFGD_ezPe5@() z1KM(>#RF9XtH@NER@d`9pUx&HvTD7H4noVBjUq*2#b}^nG7Vv6Q-x52ayAC)Xx9q& zqnKh?_^D~oRgC~`6r7J`h?ijVh}S!k965F&Ymp{7LFb!LRnkatf)PmaBgH`lw?o8gR$>9Nm@TGS{|pRs@R+t z%eIW_z;^!$HD^<_CRp5V%_xmxUJjwG?Ez9ofu|OfyRlw1oW^lobPm`?_Q%7yY!KgK z&hQ+1?I1<;Xtt2{v^HC^s@B|BFFyvVlhoUi?l%~C zp{=Y0*@sZdOYpV89**l1E8gT*7Sd3EAG6#u2DeX}>__&h9;MakH1=^Tz5w>15itnJ z^9&BeL*wK~`Gi>y#7EkRGyPIEr!wk)UeS}?tS;U*ji8ysnklkqH7_Uwg%VF!z+yz_uEN0AR{`PxruS7 zbm!_^zki(q9F&*7ifdLz=z~J4Cre=l&cxbzV-nA@xh6pqJzx+tUn5 z2`SBCU8!EQ@{7pk`6{Q4r)90Wj*1-@b21J`$kZZwwHq z8sD=WJ%W*%cS&48nKxpbD-g6tJz?O`ad` z&;!i)Mkp!C^kF(bo=C++4^PvQ!)KHPO(raaQD8a=-`XENCLGOex-5EyFmFm9e7B8k z;om{d>K`nWxW;mLAy|=fv5mVS6PDKk9LbAF-Q*ELZ_KGuA~GxEp0wO~EN`pxyq{gn zpx*z+YJtbb$bXOAkSQmXVAEO9lXUzh11)pNi~2>rHdl8B{fB5<;2%9;h-Of3;R1L* zQ-MHI95urT%A>A|So2WgayDomAKhK@bU%4Q1y6*VT^Sejl^ zYeG6Zrm7v0H>6*%stQCy89IzCOm9U?xgE>O4t3=eA~jVHZp9aiBxGXE3ioi~UsIOK zZrnxK#@L3*;(7gd_GbS|{$pYLq>EO!(7P%9V~&SkftvfFRuyfkyt`AD+oYVVtLD_SMhs~sU2a;*9VfK=zPqkutBr& z7^C}6h{e+184tV`FS|r};=2|F;WAn{sbm)#C{0SE9$h_Bx}@OdlgKe?@Z(<5m7MwF zdQ>~{_nsnVhw58to=ao@HGy)XI%e;Vy|$K~-2eksqSovUCBcoiMKN9PjW`_37~O2A zbl6og8A>W!8{vElI8gCK=@pdzoaFc})_+|s^+Fnm6k4ok`ykQ-6h&IUeW^v7zsDVu za>sAl%xv~D#4dV21&6lA1nVExTdY!p0)Mp!hz|^M8%fB(v~P&`w2v#hoa<>0&C9*sq^j1|RZ^&qZo64jynG*$3c(L^_lg@ViG zRwPi?p$zGSw6Yizt5YmUu);Pl#=O51^7|!W@kk8Ju)Z3NeaKQeqtwNuSfO)OrKi?q zlTF2}sgwlo;n)O_Tf#6U{|0O%;#B>ra0}yrVgy2~x;Txgkc4RuOtC@>Y$9_Jro%J} z09(2f-F0ni`^xQe7(vF;GDwf3b39QhlpvlSnN=M$InKsU;4vMQ8d)Y)J>MBL`p&*s zdArar-sG=UeK#v_NpEPMQMpq;O_&-(vyL&kOlsJLdcdlVJA zK0^#BXWR~)OrzqRz zVHN2qH)5?lv0|f+RErKW^m<$K6RrTSzk_hi@jy>fSj%+o(P7-T5oGhuG~0`&Y9#LV zo3tm}1$b1)+;W5T#yFP9%-?Rc?W~i9Te%n5=`b4Atxcod=-jr<1{PU!h)r|u3M1sonZ?&GSFx=t9>yrAT*2!A$RrxZ>`26h* z$eUp7eNIrkvhH06F1$o-nj!Ec0WR!x4=_WfW2E$>zYZAePr9@vuaC|;I+PV}O#5F5 zWR5JjDd9CLESsYyGMiPjsWe)eujW3JmoGP@m#Ap-f2R8X6^|TD7Im<7V-S-3&%AlL z7cHD}lDsf|a>!0|(t^aQ%)v z;4;MBf@CoJqiG3nuh=J=&xA*NMRG-s=qqFJ5Z@9Quxmh|Xok}7sP32eOJR8Kr$E&y z_3`!Ag<&kB7Ylsl{fLKkp>@|*rVpKvjVxZy)ehG7ncR7|YdO7>wM!Hy`GQiMIPi29 zGeW8nVeG#d#jvA)MmV*|Gttc0Fw{?7EA_ld23bNC{Sr7+3fpmQ-_iDFEoy^-4k($9 z)sfPEgYRvo19x3VjOGq@1wN~(Amema6hpnyMRWLu_aeWw^YWi583lIs39-+61r3b7 zXNoWs3np^S4YPtzd2vj>u=nQ~m}_5MjY*Urz-|Cuz9eQl>i}nPcURL}B5`Di3Pn}+ z-*pXc))#dMhPgpfr7GCddzr6`H`1=KysEj;uF};(5V3YNdTtXUHJ~jir4-61XlU#J zb*MvcTdDmUG(m6c#OopwKP~HQkvS1vG zG*}99qUm)gI^0*5X8Q*NF%kr5`#{R%O%X)LIUoZjV5j6xFj?&Gg|`3+U=C?5y7kzW z^wJb%7sP%=&#kBbE^V3CR2%4OvZf#sV_5aLamarvQaK6Zjb1%RGn!y7fal-a?>=9` z``3YOn1k7bPF>%*`ky1>!y#}ohleic{wbRhW}LvK-w~SsB=!@|qTl4=hx{ri9$K&E zRK;zf>XAfh2n51>0jVV8HuUHkUyQ%QVaR^~yFf(0yL<=KhLA3H&VCB1Av^q5u;oP7 z91)IV5O2ep`a?pUeTZ%VN@mE1hUA042C>IQc;7fM`+$0g9 zV3~_-Qz5-*@g`?&mSWK6*NWj8iOz|U7mt;87w@Ud6Y%XR){7(KpKq3%03y2CA|d+2 z7O+oVGq}%RGg`dlWSF%4&WH~E7P9(fu>+~S3#S9D8EK<4jYA!sArDV<)ZUQv#VCyw zuQZ8Ci!+1E3QO1Cpp)dLAfvHSg_};vP1IcpQitt5eJt+&eaiRJ62JRL0=RcmhUvBE zQQXm{F>(d$;1%st7SGYh@gc3WYHcl~tl9iWMl*Y3=50E)u1Y5R!UP$Y+<&CHoERwc zi(TUr+4qO>g<;KIc3j^CFp`W4+B9Yq@T5Jw7+#T`$&~QDrm`_aWXSQ%; zTn-KH*z)+qWIwZwpBiN(Nu!TanXs(!$N5n>TB^#1pkg|Ru9Bp@Q78;DL1p7mG0EiA z8i~c2+EHpj^fDY*0j8F0 z{D(us(G`+q{=^hNvzbqgqh`ISP)#Wqm1J0rROV{B6hnV(QSV|m(R&DUb3-X$o1Lgk z81D`Z{jufxiOGCsgA$Hg4`Ys{oHRW?v)NCL4;eQ?8JP--eFHHcz6 zinW%$q7o4nC3Tf*M6oGpt5zi0dUL^H4-GvAE-m9i=brX%t6G~3sX5wv7%IM+8tQ;ID#s0c5sUnonB(=eb_wb(zDJe68hXH4o#$u50Rn(d}U}pE zxWNh*@1#hl#W*Xvd2ua@Xjv>@eW;@rU0Kfm#&H@uwj8&x>;U4=WNaNddc_xBN36@X zFbao?y3m?wl+$+uKR>!SYN1z-XW!eg1AjmFo`Q1|M7;C!}dwM@Fu&uCEfsRiGGhKl|P328$!>nbGNi_&ZQn2j=C#9UO#jgLG?9 zqj|M!D)?bTB|u(dqYCv?@m?xE%L|>A3!SwKoz)AS)eE)!F}sMzjtE25$Yck;#V=GH zB^b?F`sH&+hOetVf}Z&qp#YGbND{&z*43vv3c)N;f!|{xt zxHbaxl6)wLU$HPHv=7+yWv4j6yCWFu+6f)!*&)BExFES0jD(W%AleJ&j+9l-;0yR# zGh7xJIDFs{9xyrpyLL|Ka5KnVc3kuGyXH~eh*40lF^C793`qL-0Sc1Pjd__r_Olbq zdliE$>)31_ZH1OW<#8XG1gd#qB&?vv)ZSbetjotKa7$(Ze(zglkdN~iLim1cz6BpA zvID}0wF_>Ufwxp_wK4L8#yS;xovsIXIO6uuK!e(c#RGxC_S>30YspSL#gxZ9Stw69 z?647`uL;TaypyLd@Bvu?JAq@uJ+>c_bW(- z@wg8hY}k_*_Q6i)h}n9-gm73t%Q>%4a{C6rc^yEAW?JYOd8~{QGDnpV^x^X48jjzL z-hY^@#eD5QAd|g+IGOxq1xwrNdF!j3L{9))dTv*DxsEur3VVP|sX8n772A64ozdY~ z0y~}d2l#r#^-Gc=Eolk5iQkeK#^wBRXt2kY@e>n|7@4C$#*>fDPEe|cAk$7xN(h4* zL6LBLE`nj6Y;qPt7oeuHlZ78aY#wY$Kql<2AH{{RyQJ!c zP@dHvN{)VL+4-U7<%gb;9|yVkp=9BQmVX~==6&cn_i>PIA4hriLG}vB8V9Vi2A8B= zZd{Y|*$eOxx&I-v{zJ+44=vL_)ExiNv-?Bl^@oztA6hPds9F4>=kLcs=6>K+TEnpj zn#yk4sG69hyUb6+Z@a-=pp!9KtZ1jaemU>JTsMKE&&bXj9;dF!N;EdM3;3YQGWjH$ z{!jyiCpf4ot|_d2=fcvWoK4vG)6k7F9R8$oou|eFDFc`T{h+MBwE-c zvbTYFmKK)lCnocmZT-|34z0_v#eHIWd}gzs8pqGAB&w|NRT>#u3qx&Q=xvLGhQ(2< zLTgf}?FqdxanO=DYDOHkA#%$RO9n?gf_J%bP0pv~uWWrNO%JWzp*A}77RNzzLuqSh z4Gp!Gp*JxO+80NS3pe9cp#05ZqpBI<-Dlh5CB5PCNJJ~?QVG-2krZgOtH^taqJt>= zhDx_k>k+D*LAK9t-sGV(XL`a;It4|I85N#%s)mZgLY(WDBtt~fG7(exM2#d?L7m2; z3XM)r2UrhT+&LN!t}E}In&{2Li0Vsk0*If$Qj3C8E8--8mn14kW!i6}pE$K`;PQn; za560T?8el_3?^K_KW$@l9orYWz(s1(z{#v^e^}Pprc8&0^Oi^A9$2K& zE9n~FnqSXsaOi18@ zrg9rG3hFgRZY=MJmeR-JP8NyFZ(Yq~AN4tZGj9%lw3MJnZ6k6s@NCnz15TXmqnR~) z)GbNe{3^?``4Yb^ve!@2XFPU9j8qYcy96YSj(J&vr`_y3=@e9;@T;NJAK;P1U-6ou zR11bo%W;1A19Dw;UOaL9TEH$xOsUbIsBGRHMY2K6#9+iRD;ucd(r|qd03)~Ab^#GoK%+zGZdK=+6ypCLITASqsooz zNq|`r7U<&W?F=U>VUaH}Fz=|0=uT6G&`*VCM5Zoxu+o*unUb-b0?MoqPlZOe13Vq^ z`OrYF%RO^4d;3TcDmN0ZbQulI%U71Su_dQ*k11rG3UR=f816cC2Ux!(uJ9BFdu)lH zm`HR}*-3TzBCc`dvFvLal;>h$zw})~Q{n}@B&&iZjTsfb@6V?jiyd_yXgwX+M$CV- zqpp5Z>xA-hqXqNi>+|m_SRg~{!K{x!RCGe6=1i@yTFMSfgNA%HLlR~=1W&nU#NG1g7OJd`el|x^ZUZQ>~}h#hAj(sX}Jyp zAPKF$tI9IjbB(dQ;W2+o8L=^{!pIm^BT+$CW9<>IZUS$>eR~lmyVtkZ=%S^U^*Zz% zkymeLxJ2s3(gSM_L!{p5MAiyYzB?!uxDh=1- zgth^5BQ&PC<TpcHB3n`F72%rQ9YUX?lD!(DFm5x%i#r*A zzqQi+`^`fkz0$2qRx>!yhgT|u>ap&Q*VG*j^Um>efTE>*C6YMSHlm!=dYj0M3ZqWl z0oEcw7n7FnyqcSFBhDcgY9FSL*kvdb*Xk93E6T)IV+(+lUN2n-`H(^zJW2RL-`=pg z{Q&D@uh2eN@hkI1%v4>N9Jz*j%byW}4&r+>huz>g6_Hmacdz8drxJpnO4B)eJLlF| z@pf;ReBOPvlaEB*>|UtIbsD=lY)hR;%>ik!Nj&3 z2FnC1!vIH*ZERt{>_FDzXrIQ+4M;d4f4i8f6O$v?mH<@pTbud2dWJ4u1%wfgl==j%O-(idPa@_e z0K^*Q+_jmM zBC#IHUw^LOG?gSOh@4gib_AeXd^U`(5O2EDk-p5+U1IYhqukzjZ5%M5;l8FWBHfdu zVJ*rp%Zvji&A>kT0|p+;irp|O!AEluu>F|SzXwxpJ5Jdsl}#g~*klhq?v`3r*BT!? zATu~SyJ5IuPgFn_7<|#Ys_g0<7#i#t4P$g8*d0c%eR8lFor07wVB~I|4e{9W$&Z}>%*LJ(mSA*%H$WRZAIlYRiLT9sqN<#A&kWIz@CE89@!qk3 zHid(=AMF@N`1WsF!fNc}&tmSDm#-SbW&BQJLC8DDCSAoxG+Wacx~&bOLs+AwRU6CA%r9#t}VeKP5H1RDZLsSxitmgY*B4f$K_B22b* zd`9Fe$sL_%4t(eKH|Qj}DadGSRNFk=*Lz9ZgwOqk-Z$* zE{=TvMzM3F+_RDH*2wm0ynDY4`=+gL)9Ao|NA4F|j9 zvG46x$~)f$;)iqh@jNmb_#WevJkPD>0lpC#)M#Y(vMRF36%bkD8sgAnD;a!n@-bZj zR!AP|z~>UPBPUe@$Xt-)WIrIQG3~c7i%RKwfFa~`!xL4!fT$2ULoS5>Tfs_e1a)?(jI9S(O@xpHj;xJW`mAv*_Uz!s4g1nwk7 z!zfgh{1?$<-+Mdd&%mGXx*-zbsA9&raV<`WH1L%w$K=d4rV?pg<84hH>>sg9$2C76 zmW!4N6Czrs48WH%BaX|y%3MB)<$ls#5~HA1W3vjKP7hAjqEvaiq34v^m23Ij{rel` z!evZ7%iTUT!+c(|=XhX+QMUy$qt#ib!jn$@0e&2DduRyEGNf*9+jVG%xWY*6Zc@=970AOieLp?Wwh$7Ec6C~#qpP_WO-UXc@GmSFD*`J-dBx81$m9h;CtM{s8D*_ zZ^9?}Q|&e3u_I!p$}l-H9U<69#N*r_4H)md%F$l(m)Y+O(xIVvZN1ER&D*Au9r_BA zPvKf7z=Jsx1WsE?`t5Q>c?a<>u2K`C%DH1YwR(05#nIp5qC!NMC$p3snr()3vjLHzv%;y9 zbt>d9u?!KSc97^)(Z?U&FTD4#!S6P()&shz}jK2v3xT)7q}w6~uw z)MKmk_!2#_J`We?ZNvj;qrN{I^ZI#d39KX#S$LZClj&7K#kGS8w>3=+5^~J#2s#j) zAZ*{Lx`!pm8=B3Nbz5*PVO4U+@(mJv9oeTgGNGT^$b@_KWk$z!z;i!utud6&!K1y0 zTc}gj*t^^U`}%YStG(S)&)Ory$|h@>5rW;9LZ#M1E3j~6Dj#E3(-{K8JX7Z5xMt3_cP?79WP!j!P+x zR*SxoH@C*JBP7{!iPG)>_ecDCXfVf?^%GM#uq{X29U30T7W;{5{LChzN0ufLDAJN` zx5V&4AWA~Yt(T-lkEcvUz6}%SA`RtsOwuA!P}`P?eQxUJTQP-8!^==S_`$NC>fN6B z^xECp4`@o=q2Wz-y|`b&rHHOQ-OE_Z^(=t+3GBWEXk`rUwStjr9?IB;_!_w>i2o2P z3X)lACS{qEXG)P7WhPWIAIo$+vx!V5GnbO7w9Ir*^=s?+R7FQ{kU16ppi_H*2hMS8 zF=XOu1&XUxDz3I;el=`gyinQYfI(Ow(@0bh53LSSGzj@7`&5(mjM%#&fsZ-_t(Kk5 ztUk-P^&!!dAlPc_rAGlQ5_BUhE^b70J+5ooQOeM%Hv`?0x1$~o{8v~!N7+SlTv+h`Flt}U`TrT4#Zmiw zjEV5hich%Pr)>pdK4w0I&(;D>hh5vSs;ZaM_Ba3goBsw-O9KQH000080000&P-f1* zTUSj00FtEv03ZMW0C0I{ZDnqBb1!UVcx7^PFJEwIa9?6|X>4?9ZgVbhd8K{nk|ej0 z=6^m#8pp<}IVLs9>Y;{1le9*%DT%whhE2|Bby#j>1Tr&NadjMZv`qU#`w07QpKSd= zTyS?pG}SgEX8HF80zlw!01o$e-+lL;1N>{AE~Tsb?s?%JUE4iBHrM8O8uFsgtBx4| zJNEA#qWo~pyTdsz+@YwGtUL5qcQ_AK+UIq3sFKokhos6-oVSOv&V~Xc=e90U#M!tu zhX;3Rl8Yni#Z|6NdZ6Wf=Zf=le$n?sb$H18>*0SjPyH1V|K*qe>(JMS){PH%;wNd{ zJhl1d+JAEReEfO1{osRN{qpvM+m8;XC(7h^zVwIQCFL&;opXmzfBfMO-+uqFnCFB$*p%g?{~=G7O6bKM>?*C%<=J%8t2sLgl2{`TdMUw!fF@SDTmZa)3) z^Wo-`!?T;DNrz*SWjDW|1oyoR-UHy~(;`ng->5Rzv@V4&pB=t_ z>RuN)ReMQ4k7-?1F6~#Tng7Ozpd+M1P-B}9U0e@eCx@DYj3z-P?RvV%tGn*!Jlwjb zZhK!d4Ydt!dDOnrYhGl>-nC_3C8AU#3p3z5W`uDMBLsA5o8hW^11#$LZfK4}n;V3x zHZ#|{*MsZ&Ae(UAf4i4I384Q-N4m)Ck;nifVXqdJ?RyQ)A7 zlZ6gANKDT*jI2eHZWY_J3=GFi9qV8&PxZ`p*Zi!PuS1PQM;aR%N>x5=3=_lFP%xOi zec00nhh1H4h8r~5(Q0_G8P$<%-e$b_QB3+t(fucC`b%kbRYYt#ufQVfUx-PU{wth z*hXb^8~=;)amkdo1Ep9E>8nhWC0lRAitwL!%D=>gJYjV;rb-T z>+3Dny0=Vn_ga^q^G8*VUkU3`1#{~yxsj{9((5GXfU|tvk0*KMF1ozd(_iryeKq!}vE=98-V%?E5cY za^D+HZJr%rF%^zpBSfPQbOzTXtyqMvIDRQ=aq0paLIlN0!-iogc%<#JI0Y-#rj%uF z=YY0Dj<|&eOYv~*4`>$33dpg}t?JD3E}lfzhMN5Z(LfD3$e@oWXu4QJ;>A+OWdiVW`pFe(=+b1PUXFl)SX?(SFry2Xg%P)i#r zY}8}#iXxr@GGaJuWELU4i&q2Su>{H*R5M&Gh8pI?z(q@Zyr_l}M|W#fN4TvYEV$V@ zus88xIX(e4=2+0yMR7_}s}*$V)s=|_y4SJk0AU^~A_O~UsD{YJ(+b@hq`Q6NCedQ$ z*f*?>JM65(z`Cl+l9siYqyUxn35P{!aaAAAO_&ByNHP9NEYlcNMrx0Mo~%9OKxu~& zK9dIrOEBC@L=dLPD!ul#mg=Q-mzq>N75*uMBcd&lBUye4$9|7o+N24$ffy<32XQ3D z2yLk72--uaNY}{I14f3XG2ZjY*{1q=`PrvbJJ~6bDkVv!2>^-Zhaj<7`LGlZnu5Jw z0|S59t7-aS>RuUo%= zM|%sTYlF60HidgEp*b-HCdc=GfLWcJ=+&in>U;@F>x{c|0%|U)IdAI9DJv|S8%tLlG|gDN++TnZ5J{QkL-ql-GSJqI-aJOSZ>+>0v$VhcIxRbD_B@od9gH5oJDQ{-h6HVd6#sJ;9<(uBTF;f`9X_Z+ z&@j&im>cQAVm2v^XUaTnYYVe-5y=BUjScnIt0H3hBFX90y8Apq<3i1LV$IGo(3A3W z4inTS7b^!-H5$gm!^H>pGtA2-b0f8A$DUc@^l6QQ+EjHp*U3gwevL^qp26}Kbk4(< z-||$K*Xqs`rh(WQbSZIa=rJRg2q`-(Bc$vgO;E27V*vFjR9n)A=XW#Sg2rXcWB@M? z>A3uL`{ZZvEZ5RZuWemxnW4(kpV6gI@#%wB8CZS#BUpPD@{m^s)};xhN2UBCuc>?v z3V#~9wMK=)6gOK7DZYlMPAsh-Dk&A}CrO8{uS+4$hUQiX`5?5Y*E=Z$onf6>hhLNt zJ&$KSlQ0JN38{Dj37gw2y+3yfpeCX?k+B zDMCdaYwg-UZFmODpA@Rmz92s9UU6Fb1$&$c7B!*Yays{T1mF z5K99THgzGZEgZupVRE$)t>h?Is}F#pN!D9?9tbzQSX^q!hr_8CRvB98D**0MJi#R9NMbnTXdG1M=+&DA9u3I%lzHSf55+Ei_crB6a@}T5TTo71N)1L1+>~!XSF5W; zT4`zI&UK1;zU0OUw1W|+Y_>^aZGUW{;r3&exs6=WdGNjxz8>=QZcj1jM#;DdR0^X? zK_4y`wEf^7iv>IU0r$S}Tax1JPQ&Gh2VrPtY26?3G`76jBzq$Vy7nWDxkKKlD8yMs zt!!p234EbM-Sc2t>$vKe?l%B8$&aaPXnD2qKsUCVUBtbq_H7QIuhm=RNE*6{!C(0? zKl7jrKaP2sKn(|0O$WwsPRH${M8W#xBF?KkavjQ3mE`OGP1^pE2$Z7FN;4h+Ir@3z z{}gEbzl9&#S zq;WN4p>5_U4YXR6k_Anpi~BwSjX&tH6AYcU7V6>%3LR(yisT}5Oh+nI8d>x$TlMtWtwCw^rjWeWa~-#y9QkZ%WW`f&~+8t;Vv@CoMInR{LC?**3aG z+u#oT9iKd$VLVSSIn-H|ch}hfkx4&%$D|&XfVetdr;F0235=LXoWkN;3d5ZbFAdk$ zB$Z=Ba;=Dl1p%d9V@P^?%{fa)ZPZStVW)VC`ORT7JZ zDoQ*1`*pd}TtYiB7IuZ%?7F9prfcpiNmcnCRaIz?p2CQWr_B`(8*=JaY2nB<(nvwW zix4p6v8_<2(dy<&6EDyGj2AGayTaI(4!v??N99Erd|I@Xjyx(_<9&h#O4(*Qo@X<- zhN%kBo8WpAEfXCpoX)CTSco_a;sDuTz;X zZ9Oz^^~e^Pd;D8|GXK!qet`#v&PMpoGj#f3NxIRA2W=vFJ%w17Tj} zhBm^X`I}3&F+Vf4>mjRqnAmrc0xgvLtJcYIQ{0zX%Ut2AOMi{)7wpC%N|}1uHq8%^ zQ66t}%3U```Pv@n^^tPXZrDOUG*o1|&25f@6RMAd4C|v)J7`~xHcKo$Fx)K}#j|!A zue7)tyHsNLq+};oT|Hj&EOT%301ex3!MBUFTY(OklLoju^Yarsb3Z?@ zH~TTIzT2a;ciukVSWk6tuX`Jh#1-h}LvN!%>(9jt`*RfN<&)bAbjXfsxhwLf0v)n< zC%sdFUOuv0CsrPM+e+P|6VK;9;R)4@^Vqq$(bDZyTUR_oZKYi*U+O*Y-PR%3ELDf> z-a?05a%~+_u*?B{hYmT@wspwv?Q}?kzC(vxJ{i{`m&{!{Bt5X_WR{?=m?swnOEl+_S;3+cx)2Q`5Polp6JH;45l%V zG!hHUE9u2!0?Syk&~jCi4+3EZD9s`&P6sAT>_Lqf`T|gas!XK25QcP{cmSHD>tI?N zbZ{`+4nTUTbQ603&<aXGV!$&9U8oh?Xp;^zq9H?^{(8E7vsveEYLyVbikf@KXh zn!)#sSyL!q`B0JYoP7jFs~bz*L5lJ56@%&#LM^*WN);*_t9JW((1*&QOgi;F-mH35 zh0qHJD{%PpcumMzK=9y*N)JOLKubx*_2rsjV9H&x5rVu#lf}IHqO|Umx7qt&;vLe@ z(0eo$3uR=HhvtY%eas0m6TWELyG-*CqtJ0HNS%s99=bSEZ!4=zdDA3DHoB|IHSQy; zWCsJ{PO9E>HKyuh$t3GTltwj%rhvM$uGonUOjSEFsye(;d0X!DJu%x`ZZ-tl*|0O( zhtRZufC%et#SffLl>Xo}-ew2@Pb6OnnulYXES-49*sb)%s!+*5bl(@Cr59f}i(5>G z&yf3~om|z|l|sI_N)FHpuY?tRPkRnKS_9vs-`U3KhVI_%;wCHA^^Rc{xyn|N$|c^; z?`Qb%(wM3{FNL`C%X&lp9YpUBY~NsHcn?#Uw!cAxoTttki|^-SOqYH9{%yMv-O=8r zJ&*&Od*wdLlAGfu&@L^(GxurN=XbGOUPfPIRZ7qR`JG6v}xS=I@-nN#& z#GP8glDw@YEZG}cf~W4%5oO5 zfoTgZfj1Mzczf67om8~_pf*ltjOw<;GU+uvM)X$}w1dkv^IW$NctZfUm-3X1Q8u%o zJQHJM$;ceOjSSi=TBLrNhy8K~j}0JV{tS`H9wX0rg?pr`Y0aA7x=T*3RlDoZXZ3?& z#$-IyEj zyT{&D+O2PmOy{S@@y$rSeK^U$DGMAlWf6BxA(ROe;St8!RNXh#$d)qcndrUpR7hVZ zA+PFl+qbLdw0iIL=zXsH@?davp`FQC9spI4ca{*yx@ z*#m5`Xzpa+xKV0#Ul)TpK8`HawqXe?0q=Cq#hUMC9#m|Cjzx7ncg2GtG|IA8=hn*;rH7E?lk|t|;=x*cxGm3_U7f z$V_xXWzN1=xvZ&+r%QC~VIF!VB=6xH%)UO6Q=vFb#x7V9L@d=cj86BZ>7XP%Vg#E@go68?+dZX83I{D_ zW3u)v&>?n|KR)hnXK7EjB{gbG18G6!iB9Br&1g##Q=>C3ao3+(-L!E*+2{}5{{8^e zCVw=}047S;KW*i@zrw0j+>Mbm-3n!5Y)#%wh^}b5*m%~(+U3+7``S0r7`Z=hjGe&P z7(BJmaZ7wd@z_u|Y$c=VQ&*l$wH08)DLKM8AVn7K&|ewljf=0@CC;U6MC?kE1)fq# zR!P|@uj(>iFF6QEi}nU4Wf)vJM%O;=q;mBy61|qxtd&s`NuCh3QObYF&P* zr0})U%uuLiqHndv8pbR|8QyZyGBL$5{k^Nd^rMuka~OX*+;2WTDN!ChgI!Ef$R7-i z8K*)6X`kMX_+l56j=^F?yjek~6LPC~WSox4eds}E0iw~Uo&1S6@Cz-J*~qE!7%-Yk z3ByTMTvELiQ=Fr)@?j5xJ6&bOXDfyw~pL;-FXolpue6ytm5l8XGaM)aXJItWJ*jm#=-ygtGfbyqPB8EWJ2&`3Fd4LMC87 zV#&=17J25fSHpK4hxk$r%>|VTyr=opgrSP?Wh28?TMQhfESN51%`XW9pD#%Bdm<$N zDxKk96T@JN7Qt#f!@oaf-;v1GiQ?f0!A# zCTeQs`r?WadfE_-Lc1clBy7H8M>e(-*~sT`LITy-9}-w6V_jY02i?XE?g5irdif)< z1C_4O`10DS7l(eoS$@6C&YHMPQZq{g8t*|&5sNm>;^qXezg2*t;aFVyOFSwAT&!>j zaJ_%|+1H1cpWXT>cei%X)&s#b8d83axJ}epq0X2Q_-b7NiWMAT-tbK`_gLbZVcY^g)rx2R z8P-kZ76ipQj!`m*gJo<};N5@4c40x&E@Lg>FeK#)Z9v<~RnPS^N8PAf;O4huSlxJM zylx%L21hyFj|AHoZ^pa6)KT3F?*voe8Y|knr@)d_ySqxD9Xs3OrnTJ!YZlK0>emK6 z6zNTd&78;bG68kjtZh&-QkM;_iw+bMONS#BVp+-&7%`6Xc=Th!x+l%albW4#GTQ3; zJhkA`3JbA@tznJ>UT%!RMDp=tzkK-(%^Z1Fp@~q^rFo7{^ZnmLfBx7Lzy6e0N`~jovFf9BTM8hF9VZ%n(D- zV5~E0lMEWUN^$8OF84gCyM#^;Sq@{{zfEJD#Bp*AhuIkcoh7uRVo=;jVQ~|xjx2sH zV`4L)TyS3k_E&G6k%RV}etdiD1K}oFYqY5q&2FxqMy?rfccvA7_=tvwu*LvmG))xS zrR&gm^Qr>&5Ybsf006PUzYDiCyjX<67=f`zDHzAs(*TTo8@f9rWXLuCP?=N6P{ndU z&E4!Jgi>4*1O1pRh>>RCUQM|8iJMdf{|9PTXqAg=h{G(!kjJB(AFk>i_h0B){!N$o z9yA-xiJ^u=8^PCu8{ARZOqm#}ay@~ie>vsvzrXsHwWXgr^th$eB>`GB5-^YPa%2ez z{KBtZdZnw#Q0e;kqkqS*(EsYg^|ORxMTt3eL=-7)1n3W(LJt}&R+J%@!-NoO{39aQ z2@}$TG7jU+#bsV&$Nw~SWxPZ3RoZ|Kb%C3Qv-$#W-Pz$zuuAdw>-3t@`W^$-?QpIX z0KVp}aH~5a;lwn`LnVOav<;yFk2Qrh=^nJNMktgRT?AV2!?jMz+(1=vQ{V|&rj_`d zw;jcyf`KU9`&csYD?OR$^o=xzu_kBKN5SH^YhReIoM<`c09jVsG1l@k$br4YU=tD) zP|n39F6%7PtvzZW&{8AMWr2;+jQEIuk!L+aM{&?RiQ#$)D*$a{#2O3{toGtJLG0$j zx{T438Qgx{Y;JO1_hLG9)T>d`?g;3TuCRt*P51FYq+yabQpdZ#ZLpK*8A8orF)XS2 zV;}Cy69zL7PDD5%Dd-kfQkU_J3bf8zwYmS8E+;dzGpa-6`|xHh1` zes9X6Jkq{5uxZIK5!gNoS>^Z_-a>S3nd3PvI^BFcWAs8L@S4y0{2)xt=8gG)=1+sL ziO9fhn@{W}!|+_Vw|I#bnpTi-K%yjt=g=5?7g1mQYi$0tt5Ajw+=3OXaA-}U8T2h4 z;WHOaV+)zX!1^8m;rtmgD=Opq)QF9wsu%Aie~DG-pYeu5@ftuCez+pN$cTp^t}h>cAVuNq z75V`er4wGPONu+J)&xLp5>7iH?4%UFrl_c-2|qf6>eJ0T0%Cv>2yowS9- z{C*GX&Z)>+O?+PUH^9c!;R zk?(kMm22}9zP7T*5R$ejCc5L>llS7hxq=cC`BvMlV@J!N(D8HAjj^cl7zDtELifEB zP{yvZT?^p@8zouL_qSgE(Aza&^P!#m?W7e?M*+ib>wdJF+93~mu4LEcH0=35jue31 z2hjccI-2y9-gYMp8+hK53gZkD05-{hP8#~?%}1wnN>AZ8`K7|(^W@WnDlxztaZ-NMTpPAPsPfmLkqHQuk34Yd z+~Y7R&FQIkhgaOLXgoOak$Pl7J*Gue+p`bhyiUP*IJRq~C2`+dJ^3!F^eccN@jhvD zwv33e{VEWT9+e|tgIx_J4!qFb^4|9C!w>X$#bs9sJ=eb8JQ8_%yEo6Cy+6-mW{>1} zWuyLF>n=yz>BfpS?nSCWx?)dk6L~X@?3xu5*a#i|+RxXHrcCnNWaO=K(Y;k6H+h?c zeD`O_qYTn!W1h;8bnW{_bw;kxl*n=+k8bK#NwKd77D;s(62r~OM9RClL&rZu-RhS> zC!9lVZMjQB>a@QVXVJ!Y3u9C}2bwBGyK)})N1O${FZ)&&ZW4b%Ai?ru$J2*Rv-q%HV zCse0_2DVMx@zwBVL*5S@i4k|D7OH#TwO9_qvGrA@&S!pRf18 zi6fTWBcj>uju>xwh3-K;iiZS884zbgPwKL6#1!wCBdGbO^_V;#Um)UYkVkA&ea|qh z?$@}qx?kaBR2{)N+_3%$ZNvN(K^W$bC|R^CA<|a37%En%`LVpUz4e{FvAvANv&Mxn z{s>#P^d6^8F6#oyf^)6A4F+9Si&IQ%Z_kMUs=dBPX|~T!`nMraGss`!WZf=Lu9lQu zS*Lj1GWlP?=HQ~+^^c&{qwL;|5qx%Z$w-Un9EMJ_MhIG(^__8 zw?Fi?RM&UmwksoT@?e3Gp3jfA6Ff7zaT{B^=eOW!HoHW5>aaxA!N2`<@a6$^0>Aj} zC$e;5!GP*KeIVw4{^iF^Pjj}-BppTdZZLC9REZ_Qa-Lpc<`XyWJT7cJNi45sdbwiY zi2(1I+xb(B97(L)mn6WeWucw+0+Mlk7obq*7NDYuEznR6n;}KaK?FLOWz~%8y4M2J zuR<;e@*d<849{K`82NyPE)v5P3>CJ_p-p)s;z~~h>{PB4*C7z6F`$IicAAeD6~oDz zcod~XTpSq@3|@&B#3fp}2fS}rI3+~%j2mr~K)U;@W3hr8YSx4a9g7fY!{o(X_ki9` zINZEKFv~EzDNoM&B;;w*dXePXeMmyQdTb69I1VO8t#feuqR=llOboB-XSl_D7Gdv0 znz#Zq5;6({(@NiZ6T@Y>4f@kly<2v)mT-Pi7X)1u3&0zvk6Kpzgrc`0NLA_fh@##}foRp*u&%{>)Ec@+P#L=sG6rFSlsoSuLO0%x zDwr;^+ug~y>&~FRH>3A%vjU7kA0ZgcG)r>E_EC`g+oseF zo2GcH2-RA6BKauq=ln7#GeXETmIEUtukMo~7tU8!lwuZ3G0gKX^!l{i7%0bu8>;(Hu{tp;dwogsgbB zN0#3l+hBD=C;Sb}tjtlEBGoc&P0@_5JOh{JXigHT$%!Jpy^R!}+v3&eIq~}OBs(*n z&?JLd=#f&$=*(+|8e==)Y?ir`I}j6tJt0k<<<+aXw{%@G8bapxexF_Gbe>pthN3kaH9xFw})y^A-zTVsrJAvH>&o$%oVtFNQ*oTpH`L3b9kb=uQwgX6XJNnY5^ZX)`& zw}VKnFV2e%J^n@8U)u9@onw6nH9ih$HXbsYQ`c!^z9_{@Mp1UQhdO65M7s1x_#EQKKsWlnGLVNYZ!^|EI5%&$3 zfX#VC*8o2JI+bfcK|7iBmLadvF>YHOsGv9}{(#L`!|D4k)FRSYqJ>oLFr7aI0mXe@ zVS@FvKtxc2(c*<4r)r9uRYI@~iVYfjT#EL%9^QY&z&I9}33YH0anRslLyI9G%?r9f z1?1az?fMVMqF?8wa2UYoZ0yy7MjPMw00jXh!FF$TTDyFDq!xu!myHVxd)nG-k#3-X z8dho6w)6YAzopK2?fxQdqHYJ@jx$&-ZKZj#<)Xb?42OF2j`7{bGKeDgY|76OfekBj z=x+Ql5j`M0CdHk`jXhDu>R|y=#UL!2(e1Tck(1_DEv9s%V$LbLta9DqwmkVnPD2w% zvcDcQ9H=Yijzzm;+voLxg6Z!NoT_TUVA|P_TzR2S$Dj5NV`R&w;Hv++SEOJjc@vDj$v4}&!Clc+?nYG90H(mDMN4`FNt>2!C`oobgpSQ z4(kwM6bN@GC{B_7g|ZK)^HL|SbHa!(3fqymS5#w@bOiFr?_K37rjb-jFd!0;|7U#h zahuq7EOLSV07Jg@IN0=kT1FV`!VTq$Zr0ORSuxy=dILju4a2P7i$xue z8fdHQ{UWYXaZTgLOIE>K^PnL8l@SUjmeGYkKvKUHm-@&ri#E>1s6|W1FfyZg_*)Vi zq>5o}2)Qa;qz=Cg*c%cUU{)X5-PF;%f4qko&4V_CM>JxAM})dYb?q+3{V7N-+pUUO z!7<GkFG?;sa`(MQ`j{&-S-D` zTaHr&w4|v7o`K+Z5rG$j%^$8Pq7D^wNQkH{b<08%>WC+sB5pV+qA0>Hm~gc0t<~Jh zaQsn5=^4NU{s)cPPh6hNg)AAwq=Lc?O+&;qZBP?tSvh+SzVCV#0x_J>R0&l~R+O1# z8!r|Fq=1XNRHbp54t=;(b`BLyA3x(L0Ix(6!no92*2&l^i1fo3(uiG!`l~-#u-ujBv^@=R#Z#V7MHN*a&Z@&>Bv2FP7GOzQd;kKWYe^N@t|z&W`P_- zg8x!EH57l6@L*pRX1L{G{PutmmYL6xFbKXdjDe~}eaK8$gYeb3D^H-}&HJh{zzLTD z#>=yFZ6I1nFNnATGYBa!Fl;756k4HL4r{on;8_uEN7Ev(k^%u0db(I!K$v8%Yop*d zP-$JTo-#3`BQBu*o$0m(he3YMOdms&);iV+I~I@FM8HHZLKw0diy!dW7y>wadw+|oC`AA;zj5N0_XRn z>mxfy6D1ah2RU6hcS}a)p*k9Iq%GdcEC5Z>W_)UQq>jQfvzAR9I2POLrb2ZhiXD!j zYa7gJUBfk>MxpKAj>^~Zgvaz4s8nN@UL}b(!1+Ql8_R!$!Kv`QgUQOjt&g){@$JsC z7VyE{0jSBzWCe7t^@7gCa2r8$b+Hipdw0xx(78PpW8Gkz)^3K*Uep|J_Np;guXrA9 z82Lh%>a;u9$ffmC0HrnrC>Q<>2J<0r!~svf1;^J;AW~}giG#Qx$r>;3q*gwN*eZTC zF3v856L&VhR5=$QiMoHKoe^hA{WA%U&rNM7p<863Sw8*E@uk3*->z+Y7ulHfiJxpO+ zvu&p$k%gv%+_YF$jyjeOrG^tHRiOSZQ7olY1Vt|?uY3DN8+qFaW)8GtJCfJDzP>gF z#%DlF6AcGgg^in1PJ+qn9B2qKQ*(OGm+ClSI#U3h_W(m}a(sqt7Co|Kdx$f0`;gh7 z7?>L>#*7Vf7Ru_yiD(@g5W#KK@bs1P;%u?kOk{JUweb`O(Sy6T6LpywDLj_5BPX%- zsecy+t-G9$_VN8nU*Xy6F{9fIf)rg5ReWH^D$OYICqc=;ix9SrK_bGcgzC>Bq2Qz^ z@5E<8k`OWMae2m(5TGcfCT#2mw7Ox^F@u~U?V>!7?)8ea6>+ZV4^f|}fILYe$TGJR zaVgs%qKUneMPs1{3BKg3Nr;sxoS~&fkH`w?tHiiYhDzEX?o{bQ%%pHGpuP`B1H=|N zdW=*!&9h&rVCXL53GW$GzORVhwXH{y!i2c52AE`BOq=N~eUbD0gY2xw#zmL&<);($B~47icH=P~B}tC^ z;4klv)7T1#;y3rs$(}Ji;QJTcFQ!J*5X~6sW!AYw#k|R z9GA!Y9m6l@E0m&s{?$LpzIK~`=Itz&Irs3_BdKS989V(;`OzU*E}Nfr_Vj%pS;0xJ zE1hC`avoQ5=D+8L^OqRTP;BW2i5Vg*o2TWWM5-9J$m{h4enKB@p456L9@8*qUA4*3 z==F;2#1rj8zjtYBMI;2m#x)QAS>7%49WsZFa(;yTKcPRb>5c6D3;Mo81=V+Ldog#b z%8Ghb$_i23Avz-cE{#tL|Gk5YLdlxzA`M}2q&;-`^~UzfdJm0DP*nQkY7Z04y0lQe z5MC)=Odzt|cIF!d>v?ZWaSdW)rf5>;TcXb&IcLVYSBX`vY&mX6Ia^G#-(4Rm_zcrEJd~r5VEd> zH^=ydmYf>jh}Zo%O!bs^?L1^MbU^s~F-yehB+E6y>ep9xGBw<)(`>x!=>tK%{v zHfk_QD}1~YF|zyII*SCxQupE2Y=(o#QY|jCG6Hc2D?;pq12@k|Z=XuOxn_{iKzDn% zY!Cnog11MAAR(DRTm6wpl6Y7b2Xpg((F1!MKig+!#6Dp#4_I&!pZC!+X zc0ck}dKf|~^ISX%vZBNt%G=&{f0QzZO`iy7}PWfnU7laieb1E ztkm5p5V?NYInvMvI9`AX6Pd^>%or50xYV*Gn-`F|f=q>JZ>l}f&7@A?aA=ufdaoGX zlT=c$x{%v!%^NVC8kL;8@JG&aY=Mu(0d6ncZ(a<*gmrI1N3e_8q7Ft;DN?SK$%Rh( zYk%`d09Dn-XDNjz`Hf#aIyf2A?JKXe5jKe1?$%#YqFvj=!I+lt>mU{F|9Ha;)6^d7 zsUfK@-c+ZP!J^KSN6?1vTZ4198?(PI2#xz^Nqi78I1v7!QcUjluA164%D50xl$G!U zt^kpyz~kjEIXz@HB|m%tR})y(-V|wZaJ(!G|Eata5^Aqy*!NHbI!lb8Q;JD&2Uoe8 z8DgX*CU()SJ@TXOU&Ij`K5xgVt7un?L!lB{%rgtCRJ^1T7%3x$nh=gGwN*#2pAVt! zDVUjwW7r5}b&G7a{dK^80R8-rRggtP6e2jxzt(ijiJoI$uZhSY46_+=*4S>d2c~V^ zaK0-1=D$$H$_+HR3S4-rwKcOb+buydQgaUR%*d?zY|<*Q=X?7FQ7PeVd!F3kpM$7s zx9j`XosdJkKxjm=)9J&^GSSBwMp@w=L55Hq$S}ey8U0}1Iw%_>Ux#bm4eJZ~@uvHz zA;zT{w)7U=+~qspaGF17fxDBd3QtmT=>;^gjF9kVn$g&ZD6>Y<#@BJ6EEs7~B;6d; zmY~gxg6#y))e>%RFw#v`S(qCy`C}GX$%e`1Le2j`NGfi~AZ3!pq0>=@0XycIjf70j zbjljySp?gRH2tH7>9UyPer9`h5JPPT7IdgX&j-+E&Bi?|13_<^H8kN75_kKIJS2Do2c7Px9XKvvK0{M+#NaEIab zeeM>tQDk{K`$7&Dww_P@k5-<(_&~$HTEE%&<+`V4cgR`UBWJWm!C#KvZXCC! z@Y76*bfR9SYwg|Jx_Ec4?Reeompme_WPxANYCZGo))w+~m0TFIeYXbkUl!^peBN^l z%!y<3zVCX$KiS+?HvP_QrCRF{9N|7PB{q<5IdYC(@+5cm6qz54AIvWE);EFohqr|v z!vH2jAdKOjpbWu3nv~Y7_dTdSt;b~!@aZP_bt2RvYY|EvLR*G!7=4WF=O3hx-P2LC z5F}mKuf_3w-)_fF;AE%9E8rn_4i;{%@xRAbvP_w{)44wo7yWX>n*3bAUu_q+`D+kz z#HZRU0I^~0cDu03Z=p6ibsk{gsx^tHc5qbup4}41q}2;G8zFlgYZ^}lEsAWf5SUaV z-dTMXK2UuYYUQ^6GW@E-LtB&W)lEP~?ACC2+<&CURR!Z0V)4LsQ?6UwSFYl4EtY=6 zxps6PP@#9Pf;1~P?P9>+q9%0P?}}uhlPdSE)G)3b1}FTz_VYAJlR{$Ok+qYp9-Sd^ zBl{Px&qe-4%Pq$IYejRM?{p+WL_tHDtrAn9lc+J6Y8*fsF5}|pPy2%BPd!?Ag!>7` zVt5Yj1T>*3Hk+7Ly?0IQ&c*Zi&S=S?v?c!CAE#oR=MFMMaU?C}gRBu$_d=X`6w_T?(YUlM&;nxAf zvEJQm3JdETQ~bC1pY^b|F}6v!%$zz;XsJKOKdoK})JQV5R1eUi;^84SAsD@J@k-hM zU96%AvOh-gBnbJ-zQX9+Xa^D(W2g`dg4S*Y>L9Gnd?uAep}J8@Kj>i2et}mKoCLLQ ztW0ssot$#diK2X!&`U!`ZqH$6i5F>CdxMf5GP5|Jus^EMcP8uM*yz7#OC2E?qT^>f zJSxzjVV_8n+Ke#vXnNexZ%sH6rLqM|-q7W)>VBQ`03wG4$KNZEj~RKRZ-#Gqx<`Rd zl>VM+ZSfihFrM42Z62UH(jZY8?Y3P^WhY-Kts?9h|43vRs!3-!KgWCw zAX^s=r@Jj}xtrvaa#Z=x+kDx1C|p|!a@5eBOgI)AyGh}ib~ z@=TRXR!3878NQNec`dl0i+Tc{{ATuJ;yuC&Hl4Am2oX!Fh0%VVyoGS2m$>p~M*-&s zwt=zUaHAa5tLL_$7P^)E=BzA!H&r#ql&htLm*RIRPR^#`eMEY13aSenOqW6)+M0r&uE%)>@8=$jHFE$~1{ z=t$eKnKb<&jC|NMpyPd!|e^c-%Mi`+-?In zse9op6XG;KA;TjZ7NNpT*?`#P91UyPbOk1JkqitBtKq+lM7Ff<%xM}VnozE)Y^s2^aGV)$7XI)az?Mtf>SpO z;6YYvoK14IXLT*zc6|o@Jbr0B5^X@tpfd#pbE%Tt*%#~uyQ8EHVK}6qX5HQieo<;N zXFnd71?sJ~=31%E?sqSQ$gjOXdf&>$tb|d!6oVGY&@i4O$~=VXx?u;^HY&6f45~ne zpDDQ*prlb9dmd!(Kc+%niC+hJS#jm3Cp||Knd&bOl{cAS}qk{nM6dA6$7;Wj`MAk6` z%(*epK_wo{2&m7{?&t0T*1I3yQq?Suk3*R$Ybe_yHG4K5l#Dj)RW&aU3YDHvyOD>M%7bdrh}KMI?qf zv^J?w&8-1!nr_EC+Q}`{8!>0~x;I*j`20y2cB-hJ_ZW zo~UB>Hhb0DxyCNQQ!%<^9>SVf)__3XypU#48W@@osWw}RM(#pnNRbh!-g8T^-)r%Py)*4pw9giOLL75cy4(IoqOFMt(Ds_V3 zBYN1%s^P#XU4!J*oDLOFU|;LCk1Hdj|C3JkqvCI|cEi7U1Slfai=oh1vSdz|)@8S& zFaZ80I9sM#H;osjx);nNZ-6j!6_iX2=I{r?ekNW5;XHOEgW9M=VT+MLKu(bvlmmSq zvp#K;zJr60VFk?zq#i+H*#GZxd5@{j~PsSM4+zw=lL?7H;?$3_b0QB zYuXBwQx!|@RMtf=fIeU8%bLy!o4SWhh+8L^y;Yy~3V1rK=x9_7hK%K?E=%Ie1=;jA zE{$<;vnEor4LfwA8H|G$P=~jzd~aw--^BcD5Wt2*PqU-IK1fMN@rHT`!@|gB1VN|S zUehCc2#gOg=Q1L>W|Z+I70!y3Qf5vdw`m2Hpy<9fKzQRNTE$-)8V4+>#u0{c zM-@Ca`}g_b7~itqXS`8-k#E>yP~qeWhiQXkr=F)t@Z2hDpHdfbQ`@+Cr(oCrc;(cC zIhr|7{R;8=>Ngo(4qgY5Nm|_wAw{^GROi;uxaT#thFciF1Mjiv1dN%4{pizS?335| z^Becq9N(I|Bk-44{uxy%b38*PX};PbE0q#)2PNr3Q|1n)E;gvlQxJvZ%^BWOzUHRZ z&-O?ofy(M+x#{QaSv~yL0dzPF=Rb>~14x7ABNw?5ijDY~k(Z3xz$SqqO#p1v zxhV}vru!!u(Qdlb2FXaY4vN{l19!X6;4aOUuxoeihG-2(C97xPkeT+C*9Btya?-gS zSWLe+Ue!e|eaw|_Jnp-T5hV&wTPm|+AaMYhnqnkwFoM?jT0}E zqkF(xdk|F8vp3sVblY}Fih<1FaEF|O-Ifs*Zcl;6?z&pdCD18Xw^-|plO^r-m%PBu zDO9cdj4bHy;d#W36sq_KR}tVVq%E~|iGgGvp+Awo@8@|G;PQ~Rm604N6!OY(FqJ?? z^JXM2{1Zy&>Pbg`DTH>(K{z^;hac|p-Tr_La7%7b?yK<-iF`Q7&$Hf=;;ZGF{tJKh zwhGd>O0ABtu}K?3Xy|7VG3q@wIK`(Eb%SZ2(Y!dWg#4wkMdxxL-&bC#i(inCF$d+K*KRE_MiaRo==-v`Ym?S4#lS#}gw-TtJP>4k$)D{yMwd&|jq%&1K zyCnES?TjT!`L0kcqelTt+@!NJ9_Hv7jJ}4FF^)JHf-uTWD?I26J-(d2OM!pXH9Uiy z#2W(VD@X;j8(fW^s7z-cHd$%uS++l#Cgrz-*dFuxFr=+d9p+ZZaL^cN;I)D71sp4< z?1(YX#|vn~nOqRYE}!9FdHK7qbC{-MfAJCbGiX^SNx^`zuy~&Uv&s~D0JT1~5DsST z07YdRNakt4e|U_dr3WlBPLAZ=@d6%l@dXzX3JoevQgxv zh%xLpp35r}^{RyZ5{pW##D%YDP!uo;o5f0l16XHpZ3l~-X`)~)*I8xG*(bT)r9X%j z@UuH}J6!&pgTZQ*I1u|PMnwPwo3j3+jvm@Lr!Z4Q34zLIRj>3q5E)@G68#=QQP~&n zQAd?}h=Ohz+Ow@~kzS8|g!s<+j-i1^#?Dae!od`IR<(pj#;YI4m304N8A+WhOV zyk-X;8Ut8N)PeVG6iA(DjK)yo-mc&2)oqk9)dWP%eSw^%PETBc6=57ESPlE54q^|Y z(ynmsj0(ip*|rfGo9J$iSoc&Z$Eq6AOW}o8aywpASw!J1BR?l!MniXe^_Fo=vra>_`A>X%k?J~7T)8c<|de=qLm#? zIa-PJe$*b#_gVx=O`Xb`*%|tzcbf7=L^>Dik73w1on>`ME)s3tsZFZgf#?)lLm_FF zv9@-MUBK~bNx8fv;j5oG5+>9!s%@bY@>M4cniXF!{ZGiG-{Z-Ii+#h9N7_kfs_Vqj z>;642g-|!BxH==G&?n$AQsjGo z+|^P&$Aj`bb(t>kP`89PgWT|$My~kG1>?E*PG!EnQ}2rt;+qFZv)(l!<5--N+s0}4 z#jk{-DJAqUa2{x@^u_L`rwp0CkT&z!Pq^<&A z)NW-BDYaC+^8u!|Yis_r^pT(t$5P{1@_5-)TblX&O7bVduY|K0ldPPQH6-C%W@S zeZE5O!B_L2SN*S>=-Ypwq(}O=zTb<@xrE)$$4CDR$iC0}tU3R;hfVu{3CLFVJ~=pb zgYT1V#pL&?$J5`3-Jb`XWWU$8xxH8KK0p5VYa->{-mm*AdLO7i75YO`H4f9fUv@FD zpO>$iK10tSSQCNMLv=Ga{4eX?c{0}~1(JrNTh7_1`_VkmOg}qMA@PEcU%@^`lV;<$ zKCQnmukI)AYuA8%KZJFqGOcHszVNJ|^ItS(PW8(Yb~nw;6MS57SM_dZ^hG zbN@x3=lGhJa_}#HrhmVwdA?8kdW-n>p~lB%I$8Pp2DApYHZ_Mnil$yw0-c$?;J6&Z z%zAY`Jl1DX)v!w&>on3=eqa6V&zknEtAv$I2|oA>&gd6R_I78;^CC=s9@5(6z&w5; z^G6SlNBswY=hoj}o1C9_K7<7`YS7-~!|`kL3GoFO`Avj1`EWfnxLNYO-1_pXNiE}( zDqf>SN}nw0-f(jR_*{Fa?VpH3*3Cx1;i)EW?s(TH)wV5)f(TL)+VL1}ww4N8TL=q3 znH}m!UWI{&7L_>I>3Xdh`3S+KeQ|`~r`y@VBI!$u%`ew2{2*av{i3|(gqPF%^Z4JW zW2FIVqXUu?y{{G?Xegu1owelK8E2eH&xqa;H{wjT!bT%O6H`4L{RiMSDAbr0&Y3t+de%eK6;gN1rF6Xq4U1sKPV*1O$ z-W54=aroGJrQLSnGJofrqTt&7_zU>IP_{jcFr^^`004ZJ|4&<|gT1qhxs$2$|4n9V z(Y|)x6r26M=`#}Vz?4iAv19JZ!j{j**3y}%rc~OG#$|VPZEjD}R4$GyHtKrid2QqW zTOwgV$!f{ZlPMX5{{q`@1b8KN{_I5#!?&6CEM}t0DK*kWF}YxrmA2A}aD3pGR;-$o zQn~5h_s*&K?UPu$bJOYwLc9j^6Zw})QglMibKvB8`A0mfVnys3!#=H{tk}!occWJY z-g9(E>(@wic9|uJR24G$<};V}@me~u8o!d6%}(N9?)diS=Vjad_Gk1)U*nnJuTLFn zl}NMakQ8A$B3N4OY5axdq6uZAQg#FMZuO?E^)zlCwV<)0sH9e$oR>jKvikI2)psqE zLIBSL=2lrz6)3WU;-o5DjYcE-_qdXYA_IBF3Kwv(fl3o%fd^_o??bQBx!HArnq77} z9s09A%YfSr!+vcZJm+0;6ph*?q9qY0GeUnPhdR)uim@`VjD^<3C*9(6g3rM;H)}6n-%z8ynD7*4d-attFHvU+atM_TZCydvj(y4>Y-8+d%H3 zuYL@Dq$5rA<=RQm>O7dvolwT!PR4GvJZL^g-Z+aWw%)w~*i#5v3upw+F!yK(Y@M|e zusEk+wL$23Y_h=XJI3+>#JwBI%&ds72Ezxq;+bYz0WJm`_l-emU{{bmMXc_4d?p(H z`cb~d_ctkE=QIs78|w;Aq zP%fy5^R4x*VBzwblNjN-V96`(!7PoX`OAo78G_7&3Ec2_Zr-nc=JHrAxWsu#7SQZ| zCm|!)(@gKq0B0hMb-OzzBhfk1u6_MB6OTY&D_(Agh372Z&Elq8PrNsuw9{bg=@>fp zdNXid(mD8;_3llCk6~@0Id{bEI7iWh@%$%HP6)CvisT_nr!f(q7P)L<*x$!klI2&730T3E$-hcs&<0zZ~8Y((R5j83ZwbQG9;6Utx6_{l1R{DG!EBFt zZ)E~xd^!7^XL}m%+SKKA>bARk)q5T9)AqA(@V%x!xasTy3*Om5u~=U z3vWT>krFV1OzA>gWC&(bkmH>jvrk*G86g?XWrf6}{b228*pD5MU>duWH+uv&Wq5(x z6ag2NN7(XMy7qN|MI=KvRp|>N8Wud#Mp@6jAR{xHJ6a=M{yxwm<4i+s^fZvsnrf@p z7}7}mOArjgFboEKbg^a$kM{e3q%Vh`X*!@)Y_29mz{12Y8C!Ca{T2h>5O;^j;JFZ3 zh6%v}kyxGnMQB7wyffvrlOW4fc2IbbL{o*60vHmJ&;sHsu^xXL8`J7?kM#rkEBdj~ zLu9N-y-QeYDY##Ae1osc!lrT!Yl)4%CAKFcY&?nRjgrqic`zdUo)1;*4SHvF({U5R zFs{Kz_{6H2!;vZl1<{{SbLX%jTRbKL1!=Kpi>WoL>Y0cEc86NRnc>SN%EEl|G(H1^ zRAaV?mkOm={7e2^i>=QzTsu>b)!j7{U z$?Xj`DNiezJDD+wEJ;1YH`E86l1Du5a;CX_n`yj1AR%qI8=z)_R&cok_jOd@bzUoA zRD%>z3$&4|!?FXNKM6U!_d<^1-1*ALP*JFmj4SEVK*aHs>BjiN@pSpvDx}GfJ$~2> zfa=;11q?Y+C=!H>V4~eRrxjKPmbJHJzUge;)gEUH4|1d8#6dWSI#ag3u4OX4Qn2zt zXUESG6Y_XSFxu8wCW{cMg=tQAoS-ej$JZ!6o^wTjiAMjHksR~}FB2pfZ><=&(wK%A z@Zblaz11dSn;_diz(9^>uEe?`4Hg}2iHoK79*cxg?BS!EVHQ~V!-VydS zEMXkZ(Qmi?WMe4Eub?_H{vUpV$j&1@B0tLE9%z3`_8(v`Tvt%}{mYoby(ikSz*nw_ zyCg0uT%rr+s=zISR2!URR1=_mo|H51hZJi%oT-1*0H_8e1?IbF9A2in5g-BKT9}QXBqanR*abfof)$$H zV2GC1Ujpw*u>W+Fyi0X$KxM+U4qrH*jx zRI3mE^vN+}arD02xH3{7eKYf%64JxkPP_YLEEA0upLI=yNE2Pn!zlm2P~> z>eTXXV-QGfu|1;mH~(}8!5w&m7*;L6>j}1)Ek=+XEp<0(6qpW?b6;Vu8S`K?coRtL z8LZHO`@1YdcKn5Qmt5wZC^{Dk24ailK_c!&4jAS=e}4(@LXP0Mut#M@tU(lDYyAvF z_N(l>#ldmP5#8P~Bkux&i-$8&{?CQV-mCEI1lL4mDRsUQisT=n;AYh;X_ao0^-j}c zPbo`8Q;0n6hY}=3;8b`;!E+1ENMSMe0#hX8vSN+WLcxcuu2i%sli-2jwTn{%aRF8e}L zl%fsqzjSRA%=e1{Hf@2n+H4Nks_?Wudc~$<3K^oC4}|uvnE}EiQ*MLPIv5gQ^zdgt!(1Y|wn#VP`Ri`9; z^#gWk!?3Wm;@x-%uA0x!4Lb4PsfVqScD+?_*tzVMz7HpS@Urd;oIZ-HQIIAoD2O9d zKET)+TSZOQ)4R95>m{m6BI_S-{WNoa{xM)%zjY~el})|gz%P}bLlEoU#lxABKAfE#xCfH^)H=(LqY z^@EA5T#Ks5g7La9>BsWd>+ea_d03OT&%6y2eGu6pHBml8m1#8;UA2nWtrZx5JQX87 z6r!vIHp8!Ile=Y`aRVU5Eq_R&*sI?+{2}n~SiSi|3%wkV#sFGLZ)rd60auLMzAi@6 zxo&#x)OJ(G(X&pxVNy0%))0yO!%jnCz>-(OKb=>~!xZxLMNv zQC<19pQyC-5mt`lt4zvkK`J}bSeAj=?5{YcrlO9i-;E=b>c#ILK)BT#%Mn^c-yhjx zlcpz(4XzqZK@IW4NHM9SkI#LZeNZ%gI(L5`RF=2Jr;?Mu*{R23TJ_M4WEW!^wxOPK zql<&K4bp)A{eRK{UVWrHuM)wExPv!87C81c0`6<81=uH<_B9FxlDiQmxR$(RUN)?! zDUU!UYMB-H6stXHsUEh~fWa^H#g2hL5tG176+R3?QfQ%3MGODD@BJN5YjtdT5>0-X zH&FwdD8w}*$BvnE>-BkhC~F*G)=)o5Icjrz$Hd`-4SJE;9nKU_S+rr1&K32^PeryB znFjLIGYswrtr&;9%U_v&T+MzY`bAYrj7JS4Pwgs{Az%u@q*UNh6jHEFfbXDNA%7j| z<8F7_G%XMP-u~tx{dRk$Vz&wg#=QHh=nt8c?#X0_??4H*Zn%nOYboi5Tf(hqwQWxj zbt-Q#llrfxu?6AN&S6`n4P;=8 z0y1e9ZTT!zK|&OG7tlrAry`+%Km^VwUjK~Yt9-rw=V7yMO0jI4s-8`u)j#o|@#YgM zTa_UK_Zw+-CLdvldQm8;DexXr;m*@T9AN{Umz6iLm$yztT!OHz&8Y!07LG zJONS;TVJqnVMMMRyCNmDF{>zY@=J08rhd~;MU0JPtBiXC$F>+~_MfG(G8W3J1!3Oq znveMS_{WmWA|3lh<#%DOj8rkCXX_f{hK*kly9Uoh{zPn;IhEkA08FZS9GR}R3ju9A zh}QUn0SAy4PLN|kfAyLjA>DBi#UoMAo$z*0@iKlJ@;^v4NAm9mQWYZxeS_s&rIICV z#th3{RXkAJDKF8f%ob&XH=9aV2RPifjMpEiD`#R~o--RRD>OR{Sf~>l-rm&Y)Zoyh z8%Fn~KPeMGhQO})e%?4WL|jWXC) zlhV5>A9XJK0w+J8NJvAkzUx%p=eS3wc-hzY4j3G=_YeC$DEq?t6+d0f68O@MIUEaG zdJP>aA)?i_?l~R#8grM6 zY0oBgcp-D?&v0Nbok?h}h_LG7&r6$BT9-ytpSjYnRPL|Axn!58`a8QltH;+phP-dY z;;?;ObEo=zTRp2??j+37K2;__b>%8nt5Yp;k>P+|M0_5^}oY9fH5k z{vP$g8z4>qG1ecu&*6*2Ucq{Ypex%2@Hp7-rTjcs6#R>kgcm{DHsE8S>v^pgZyxWK zIX4={IS&!7kCR~gZMZKNSRO#0Xljr z&$Uxi3?lpyvixcOzI8XjG?8=OGlu@!@)bI7HC8viXUdXj7aDk3F}dofmpge>AK#YS zi#cR`Y!ek)kHD~+)MsS}=NIu|*<#0te>kSX1D&ijC&yu3AMC>k5>pSf1zZ1h?<};? z9wUJZzEUsR{Xw)UQgZB~vhAhob7`}KI?q1ZpqoGbU7~VQ--r4%k7?q!aC|PK^>gXu zcky}(5=aGZ#PU~O>!GcfMEfJf8(*OW2_6A+K^zf!`UzzTabD6ja`9`Q-L%qMDx|2- zsLcgx0&`qbw&)Sr2|F0`R9y~xnVqWooA%p$@I#N2YE24 z)kcLZAN`4P)4s@c^l!D@2&*Lv^L@;pH^)Y7+A(kkSLvI5J$r8uX%|Sf|MQhTTm3A!&c{1nL z_e+t9cA<%Tow@ev|o;&Ik3U%@M zLmvN^x8X-Ocx8D>XRTu$vYs9YH_pnX#fG0)onG(P{>93a`mT7Fba5O@r#(po3hBFh zVlfLAzs+t9`@QQ7nK2|Si7&3~f#!Cvj?8J52i|o@gig3@>x1}rNK`79H zz2WAyoY9d;1G0%{t9moh?kasvJf%-#caHs#cKo+pXL$2$>`dPbFPIs<0;fCf$57y# zJ%&t%FjD)=U`%CTWse(PIEtm;0g)h=XPQCl0a#jMm0hHN#vBy}uW@@mhGI0@Z?+^K zQnq3k=Y?nLP18%qh*Tif7snvhV(Izy!<^us+G4I>(c5N!|JQ!Ur$jDnp>ndqKoWnh z&oEUTqkW)X{}*1g=f3z0QFsSr$4eIP?&Mm%@`+w9?{{^RB-ak@b(3NzuI;hw-3t5n zeDgRIw`v>xSqiQQq#R_9r~<-au{U4uv4^z_kaVKq+f4LRb-rv>4xG}yZ&p0D<%GE$ zYJ>^gM5R@XW1NG2(_T=Olt$mvY%(?t!E(LCMZR(xZt_K~f7-QB^a87Y1hss?kB@Om z6bttmjxWZ5{dyhQR)tB9He9l0SGkH{-Un!ddM{5Dl1{LMyDUbVtQhx{T+2e+e!v~Z zMvP=k8l15JlOE2=8nv4Oiv12qO($EzO*~=J@5C8A$^QHm;)<0h#By3gS|BfaTwFevi`o1=@Q?B1V_iJH0#lSt`nLXO)?DC$J*kZ zt<)ZI67O;tM2)3|1q?K|A^~LzL1X&#G{NV)2MH7S8T4YXOUbLsYV%}E*~-4cbFY?6 z`L+eOYl_vWcazm8%~|v09l51&kFMS-@0H!2?bK+=ZE!v>?r94kYugzUFzutGm-xMo z&3;l6GCSm^t#EUg!ThKl^<`S;NdZoz1HtHV%0yJGnRy3s24klCXIA8Hn;DW}3Hak{ zpYVp96M2@3t((pnD$kI=7KU_NVp0;2@8YAmvMfV!jFpp%nk@qY`S1*!91PUANY`tB zBqLz$+`Nko)J&@Dkszovp;AM=`MkUc>c`+V$_Uc{?+4W1-X(H6;e*skiwkIB-Z4=j z$QC=nCCsm0lc?Cy;RqQwj8ud5o9wD58yl8?Khxn!s%(B=Q7PQ)!MEqPe9=hCeyohM zVG?$tjI*yXl&=?Z!jh|{5*v9Kz^a5Zi6w(@NZ#z~-R7{2G6}h3bCM2VL<8Qwk$mMm z)s90&(rf7H3Ft_JP!Vskx=OZ?tlPB%iTZ^TyN*35_&z)RDE^dSi#-boEL(qd9T+^! ze{@Zx$;8CHs2Vma<)i6*E|ARg?XzTy3a->AJr7SEaH-QlY)-5l{k zE!wgYA~|Y0atNR>OP!J>Gz#SmC=p4N5MLVVgOpc;_ivNjMRlB6dT1H!9;^X6Rv1DD++4WlfUo%Sv=Hw#?+bOb<^ZU1-H z|FcE?5lT#;t9|t}v1j8V5&5lN$}Ulk6-`xZb(1;VRN4}%3T)1FEIDjP$ z2n?hQV_jLYoMp&qLT4ghti3TZ?i>@HaNFZ?shHFgebWct5H|hc^lHtEj&FMnvN1N4=(600Vwow6njQ1>+~oALx-Tj2N)vL z1Z1`{%DuxNZbFdRvJ)SZduxY+u97IVVsUi^E7+x=&Hxl#8(pM42VnSZl4%>Vlg@Nt z%H3WPXG^YT>Va6=aVS&&dG#F?0h}|*1Bci1>ra^a?(aMsO~CD>S^*o2AOOF%0mgae zGN6for_=CjB9L%hmo!vvKV>%R8G+AHiUXM8+Jj6*jw_ty=R!>|ck;g$#48VhJ5b2t zAOeKSvU+OBF*ufsB;m!v3Yuu{hsdyhktHRUs}%>{=6L1w1+3P-dAPD6yX;bTudrI5 zl+wO|YLeVWG-XA5v$^iGJ0;j&((iM-O+P-HGKo1ebO|RRIq20hkx-@j7D4FTJBMzT zDNH!>ibbP+Rso7`yA3H9Pb!0cC>d?x9ThQ=hWrqJ&)^%+G2>jqBOQBj(Bo@=%;hJQ?frH$v=^mj%GHZ4am>4uk$Q*)*6-( z`i`n=Wu3CGI7GvAm-QEyb<|*;^|j=*vflC=Ak9lqa%ZALLTv=SVZ!!PUa#^3vx7tp zwjCp(0?|E>R^wWO)t3vXB>vneN10UJc_dbokclp#)2fGNG6oKKXsWDC#U13c&z2l! zb9&rSmSU7V(iQz8xANyI3Jl6y$O&SQ9xVMeZ(jbFsv_7VM-V5A+lI2fj-hSoRe+#U zT`jKJ2Y9+X$Ey-L1?IuR`!$?_<+-rjbW>MM#}02|+U{Yk2h~KawM%-{Ees(h>U^5( zof`m`V<+fp7|v7R&eWM%*8+vO%8;#}f7XL0bFZ@L3ERcwZ44pfL~n9!CLeLBFM?xt z2prHRsYO3pC5$??)D5^gVZzH)M}?DoXA_a-ZEwA=J?RYXy_`7N|3xNYW!SDYGeSDG z4}3?r6@@K;ie>4f3;z#e?*Q9d(6xQnw%t?PZvVAyo!Yi-+qP}nwr!vG)I4q9-XC&5 zZ}NW0m;AD_XYWjQCfV7USu^XpJfp#XkJ{G<=G!A+5TnAjY#OMC+iH zc?~?(-WWWdq{7Qq1wXU)&t4MlzV`fa+#gBCpiB!O&(Ys?3iUDIVU^19hj6_5T3Ny1 zg`06d-?&4d3y_0as%}-xWL#Iv^`7+l#Kf9R1I;;WZXJ8?XvOKwi_eogJG(mH_IOwA;-UE zLBe@ue$cnSvSBu%*yc|6nT+t6ZQ+0`1Eq}cODf1( zT^}uS=Pz^8;WjV`bX{~TV|OkE8qkytE3nFzqx3fukNFe{UIRRO;4ij6wc3IymTBXBFXsqLK| z+c2j1vpHQ(ny?=pdeHS`tlOcf%jhKE=OZ%^PLOrSCv`w;me<+~vir0_Ql{w>LF!*I zGqRRrBx;x8QX$wuKyG=+C9p#nQx8gg*iQOQYjhmW7!B8<8mtd*mAejJm9SC zc8=6{%^op#vToa5n?&1}`KA7^6yJ1zj__1FpAB7xKhR2D@AwIF6sSnD$e-_)uVuSk zS4Vd#z8B6WV{06@+@S4?o!x~Mz8+TeyJc6Z-S}>Ltn8?-^UUVN)}vF45IMuOMz;tg zkOhRG+X({&1KnW5Sen12(`Hc6 zRvt5*#Y4gL9|JJ1k9`svMmr#;BL?fhcXxGuNhSna9rtJ6=w`%_kkFO6=5OA&%Vh_2 z0PZWHGwmdER+eFL8c61(nKc1gSGb+rpNWzV{>9XJchQ{MS@S{1#E~2PD8V%fUc?q& zp^Qox^8*(23EmoLS<#q^!=%%@f$)Ph?9womfeVZl(FLo=%0u zt(?vcERGkCFb=1gmn6mmENy(K1N;j@Bnw+5J@nw@Xf7_stLp9wT8CSOn zjlT4~fw|-TgkwxWH9Jrg=3f1KDS`n`jXmFwYYw-^Hf{N$O!fy`1qMsDzrJicnlYDa zK`TS06i29^b9!$dqh;?SeI9LnZ}jFxzi&SBZV=YG zuB&gpcvp{pDgj=%c#4*H2*0lQzc2Q_j?&zSvIN~MnJ<@iev++-0pASY|Hk+~uNeF~ zd1kM4>RQzhRZiQMYwf;HZxCIx43Ut|9KHilIm-e;XjdMsjc z<=nm{tX*D)y8iBdKHdx=?y>Q=5H9~M7j1B(|54kl>*@ILe%b5w-0N*g(o`vn`~76s z_h|#P*W=a3@B7ur|Ec3CA>Vg{XtT$$wGHRA#Ub65gUa{xaU_`YI2>N8K2^!Rwut}t zchPMJ_U7Gu8#?7mOO_M%O@|@_EZWwFOUd6Qyb8TiR}~bFCLNwh=TgseF4cNFjL{=Pd2|a%?ktQ9?jBV|*8lQ^5#{~6D8S3s?-6=6ZOOPsrz=pRxrVHNYwhdJa z&OM`IvcrS7+QrhsOWs zZdL=@E&x1TQ!hLQiV0HmbJKQBMAVOkRx$ofE_uwj+fisf ziT&AYdNyag&DYo%6m$Rf zDH-Kk?PWtqaUzNBtXw^^?Vd~8%}7;c6D-Zrr_Yg*&>_bs?RjolWYSq<0Q1f{;jr3` zJ_}jhRa@kR4!oNdP`6Q3YIn8J>hTArx#%u;o|uMe4s%p-R53r?9;>~p9+AI}mbgr+^|f!sIIj<&S>~V(pOKz4VL>-e4R|Y0H#65a<^P8f1`9{+7e*$zOJ+? z`ohXbssFb2ZQ^IWIG8fX4U@SxG8Nzbi%7WGMYuTCCy$u-QuY`hT+0A$iC9}yUb(D( z)VSlnXS&^g1kku-rvPP>gYuN9p0#n8zy_6B_YBvc(;U7Ug6e|w_x$Gcb3_DRZ&?Bq z{7U5F_hF!aMoiz}$_kwqJ>t#!fJ^&NQ1HCK!3kG+YP?$H#=1dK)}P98@L+ArAHPSrFW_=LHC_HG(AtFArz3tOP|YG+M1_yKCMXzzM~M%_7h`jeipW)I zGs$2F_O6+%ubUMR8W+wH-%#iLiV0nHOdCxe83PmP8?EZ4e1|dkmYXn&!s#bP>{Ao% zH8$y9oJas2;t%57#eV<25v%hzd%U7MnICjQwKB)QG^BH4>~1bY7_t(UM5$Fr+K)zw zwJt*K$H33$~>ZH_869{F~blJJc`n3SAB&I#26mj?9*!M2N@d)oSj)f zmVs~S=cdW<-9)TqimXFs2~C*(mE1dAtE8!lObvb+$#S-vC-7d{t1}&=TPwkGxL&hy zie|@CG+!Il?_mTJpv~XFu51sUptj^Fk}w4P`_S#xrgd>+d+xm5KgpuC-7KohjClcw z71`V0NCkDh=%GB_MPw6W`veS>~BWlDEtLHuXPt*6jIpMJ}&?a z7+H}(J@Nx5H{CQ)YIi+Y58zkKQzpgFtUzBD^qio6=l*;+mSV@~Uh>2U@iKjz1&3Ld zRh&4Op>qgjf>tb#yfvG920AGVuU<*#biqQ5V|HB)97qMilwC_^jYdL|HA7q{UHyY5 zN5F+bDoVyKV4slOky?+#G1QfN<5}exJ$XC@vz^&f+-iulkzMC^P>#?#G);Q19S0A) zp5+d<=Ay!-%vcZmHqsvfwiJ!}pVgs8%imU4dogG;X$NjGH&QoMKIKmc)3~05_)xPl z&n{-8YI)q_6H_bs{)PN*eKAXee>a58F}UUH9H1FhCbrDSG@SjQkHATQLl&cYWBJ!< ztaR&4ao(1EqkPd}LWQ~;sKU)VV61X42UAj3O~^(EPgObq9C4Mx*&jI5wyQH6WnMl0 zj}3xQs|Oiz-9Eadu%d;OjVk5+SM1aRkAbsMo|fcs-k@~+*j_{cyPo~XGZ1=s(wLwv zij)k{-jEYs9j&Og?zW_ip*W8VN)h-Mal_lK*?TY|imoaD1nHkMkA!u(mh3#ifKrLy zB#L7h0}_`3e`~bIO2JpD_RfryC7^=y)HF6D_~uNH#*4xB;fX;!Pn-E4Nes=we}2`B zUJ^3$o*|aXbfR+=;_f1!U7p^PbDW+j*ASb?TNRvJCUOq#($&tLPPGKDw9CMx@Ma;% zs4*+jTQ;`l?~t*hLxx5KYusl61p=aa3FPTB`7Ieura$Es44m31o%39)i2o8#Sno_a z{L5|igAJDxsbYMYZqxUV5cix;EJ0p7n1-}R(jAZvS;8ZSgG)p4{Js>?dO-_s!J-CH zV?!w7(R+(JZf!S}7&=vD=}mx9tu%q~4VhUPHX;tn;}itV{7Z_qw}d85CtD;2O8`pvc9(9>Zu_-k$M=qG+-932*%lSNMgsLEcSZT|| zdj{7;Tz(1`Hy8L746kJ0R&fiNMU_@{KM(lewlKi#? z)!zCIDlR`Td#%pYn-vPy+{}BiffFId)hq~1eXG`@d6ubIMaWMqI^DIs3^V5=SC(4| zd~ojEAF-~I$|zU!d`MtxQbx}&NwoK4AuB|R`3kLZJ&Zt5_RAt&rVP2h3z#Fzkv{ag zvZ+`U;DpbD-a@N<_2>SPkS5T<*@e4O~c^f1nP?Wo{m6Q=E;Y zjN+%7^}e_)1-Tx<{Wp2;X2L>~PW}=*g*5cNy=aRkh=Yt+eA+cqf<#yDNOdT-cWy-P z@@Pe&eO!cfN4t3cB98TP;WD+ZrrcE^=~annd{wUc*IhS&>Y`AW$8rqRswG(r|8W{> z;#kkd;}p~o|FQfIoMH;5v0NL-p(^$l!`6>G(s%rHSiZ-AI=cc6Ph#(z_aCi4{MFy% zrCGS~dG_<3`YI%l8f1XOUXNU*@{lKSemlF~Go(v<5Fsnj8VW(>e$=B?RN&q%Mn=;C zcZxbZdi^tmpH4YC1dRO{VBd$}ndgFAte1 zR?z?OL5DKdd77n=;-Yv)`BMR$UKd5zg-$i<4EgWB=d@tJeg#+tA(zH9yiK*g6otCV zBY|dSBdm73NV(5ro|XUQ_3IN+D3&zQaZde5KPLw_>cP_NA<|`rqql;wf5Zb_V}?8x(SHs@VXgn zx)$ks=^}2cE{9fKR?XUs+LhWsXYK_uirOP%W)ZKw8rS0p=x$Lm5%UqT^5&mEKo4Aj znNKTbqy+KfFi>`k&^Ol`WjjHl{So3({lz=TinUn;FJHk3vy}KDxCFFe>Gexs7|){! z+Y;x`bg>P!PEat~qya-mVSoMoe+d?vm+}Hl1XECq4(+GQFGY9DVw37IUU)aauCuz* z$RadR`_P7}_rP3ke+16-K!Z&u?a`jPLQg01C96FOYRp4g@DiqblE`)A3yf)P58V7e^29u$8Y8C23LVK1&i+nv0urX&MR80=VlNb* zs$6!r#9c`?Yr1XSbR0ObM6v|^JIkOc_b=&mGf`MIbA50YPuuz*l*IXfy01}Y#2SIb zQW}x64i@7C!(}mathQ0lc9goVqpYTksX=v=k;d9R;SmDD`O7i}*H?ev>Ht!|@faS) z?4}d)7LNp%#{7(Ya6!cvmX|wT2zhEt~bB3hm4b< z-uy=|1BImEiT1#n-*sRv$n#tN8kX(z>Zk?J>vu62J`1JqTkX$`s$1Tw z%(9t;BwTf8%08RWegzjT*n#bu%XJuJ(mI4R z)!z^_<%!6D$)*QPxsfl)Uhm`8Ks3?*0`KLg(u{U9?9+_3S2Hj-`Xd+Z>l!)gcJAwdfqaBuHW96PaMtI3S|>3 z!>TJD{vr?~X(;2e9$e$)r?ZN0<{@&rtEWP|FBQf$x4TUfsA;DuLLv_AG(lWo~$w!dD%M_oDTBoH1Jufk{}Fzw|Pwo}H>7aQuH0So2A;*+s*?uo&)yMvZp7ctbHCNpm^Q)*8%Ge zDD!APwrcv?+UqHIB~9s&CGl7STo8ESuu+g=oU`>-OjuY&6tG1ogax}^aBAkB29FXC zG@dx7Sf3ogHKiq|+EQo?DTkDV*%7KYd>L)jsTom*`(FUyV?F?KQe+ztte!7KYNMtS zEV{tbT+YjYr3W@omRyWyv(ZxFp4d(vb+E$J$CThn4xwf+)P=;1i_+kX9GzW{r4CL#_nLZR>^dJI4u`nR2aQoXtzJD!~#wXXnPL?GU4 za5~^2q~=ugfKiudMZ+m}0~CCJfcMDFqHWLn>$G_6HTf}K9mOL`P493-rm7z9C?X?T zgDYVu#dd_wuD*c_C8lpz67drSFaCq7+`UB4BbX*LX*^=g2pTi@BwxsvjAO1eq)F?^ zF$k9~{5&}lygL;ZEoBN|Z8nP{_je~bT@CzOJiS10D9mEv4_ezdokE%s-$cJC@t8Orh>yC8oJ=02Q8 zZWf&%%V^g$>b~ai<;3f-C7d~IM9bHjN9`y*X=Ilx7qw;W)pR92U1<{Z>AKR2R1m3M zoGmJIEB9q6Mzw-pULjLmpel%L-Shm(5w2t+6BWLHiajV74pHRhD)OxCiaQoSPc+pU zugYZo+9cXRzr<#Rx^5l-#3FANj$_^llr5suPV&RQndpZ@S2V;Hu#iQKCfl)s9GJw0N3gJ84BfY*s0uaV zqD+}Y&tcqnm}#X#%FBX63eK5lQGhs~+occHY?qC!apw2ar#0MVqq(+f!laEQBno{&IH>OlE-k=-3}L$B{vgB0emO}8MvD+KTygWZejhl<#fhRA|NJ_3&6nd|`~ zfmz!zDZW^5ncxZ9IiBf%F~~sc9$Jam{DNVej}S9mqxUq*Tn;4K+PuJxIe+{see4A8 zN!h?arcz4%P-(1s`dH%qSNeJcBvZU>p{nRsnKnXk`&QNQ`17Jod6}P_v^%4J&($5Y zR@E!PC{Cv`ml4W%-48QBXTCY;a8wF-%KMI*$iLl@nR#7@Hhjelrt6B+NQ;SL0f3;; zg0Kjn)+Z=L&A>8&Rd}PbWC+tbwo)C?kAs?^n+n<|ibvjH8qhex(=0Q3fsuDbyQi<~ zai`i9nTrpntC10vXt{&E9)1t$tRE+4tmGYVH+g|)?Z)FQ95 zp)oBw6?c|zndUAjVfd`^w{*PR74;}(2W_2M7~4E`UmCuWJ->ZH>9EnQ^{uef@o(Yx z7|ikCk(z^+&RZHmYz_H76>0(jO{*?^{#Bl@lE(W;URB%j5nJjurH;7Bmy@eQn+`V& z+_G)91nuB_RS0E~c=?9#ZnrxK2J0XvJSIukb1_Q-CWT7P%52EkpSq;r_I6OiI!Q(r z2eMb*Ehki*=%FPAgNT}4`mf$RrGwsaf3}C=vS|rlPFJL3wU3mHR z;-7P$(N)b$!{~Sp636wL>0<+x z?YOXQW6R(@dZ5G<-j=@Y^}UE>l&7(3v&f5SmmMz zZM&Uz&Gt7{06=)OLP022Mwr(Eq*`B(dIvzd!Oa_;Oq;umVq2L^n`$u&t&C!r2~MrH z(XAZO3$3Qnt*S=NbkPf<{<$@{Mc42~%{ACXHY->*utv?%pv^xIe{S21VA)uWn$!IB zX1UqCNTOzALRXtLvzE=CX&JGC^-&as8O->9&sibC4s`c(DN&+*R_nU&4zy*vRe z>NQ(NjDAm0Oe*80PS+BSbMU&(-`p<-U%u}B=Y~X=Y)g2Cp8@!w+y|4u#mVD@fY3=vt zuh*UY(Tt^mAjW^r3NuqgoS$BrW=8|a*+Hjpvjl9mUAaXD=SlU4^uTVX$C%YIvqCW`%6D?^wm1de{cS2hx!ky13MUcJurt1 zd)^P<|NH)?PxeC)B?}Nx*YyAA{-%Ysf#Lt$-;D9GNy6nwJazkw)SaUu9r0+vAd|O# zHRM}Q94+FUSnW8`{5vbBqJpkS+IdwIK+V3vC( zTT$2J?MUpqtG|%pe@m;`z&+E$=xw#pL|?Jqnt0|Teee68*-MJQ3_D4xT7AE$#NJ+M zW>a4a*bz7Ra+G}5;f;ng1p_wD@2cih9SY@zjiKcRjvocHhM^1}m% zy_BOid-{DM$@9FkQSamM`Rfu+5-QlvcboWy;bBAh5Yy9_1sEKJ-v*aWwphc&jxnh zSPty>W1Bo-Umm&oKfa>$XhT@&9y{5zk>=7pnXTw-v!iR!J0Fha>=u+pG@_%p#kM7= zNdf86X^;`)mleypU_U&9K>tl*eL3Nm6H>B`ZPPB2U!WRQ-!{>_Pg_uAwG z7XL=>p>opovW~~728x5b)1_vxN<#iQZEUiU((LxxNXL=!E6Rus)D?$e?vcK#B7e=5 ztBO^yM9xt16;#tZ>X~e-W51}5`lLAc`6PT?MgNzqsO>%D|I#OW`xjMPce-cVk)dgO ziuQ$h%IBvC_fhX`dmYQoRd&~tMH>tBt~PByIn950CVsIQ^~rMh*=B3QcJoU3m?2qa zoy(&Af7puJwqv>GmTx zm(VrU-WC0ShraP0v+bV)L;ndrQnz1*wjZ=@KB-%LQ#N@guI=NW@ecgP+lbgN!lR!D z#4kT@`MDrE`tsk7*#WW)yYCio|7~I^k|O)5%C4infu?1$q?;nEs><%BlF?m9UmmGS z-c}`Js-BTu!@Qwka$h}ZtdKGhLyfzo%1%&a4X3KvSJ&99Y4TYy`7V&tBL9D4&S9f) zDmOXQD(*e-#>Zkr*-TdJ{|{%fR`ew253-(M)F%c+H@x zjxwe8j)i=S=6^tkX^H>0o_YYsc)h}Y^&ecj=MAq(SihQ|zJxk&61UI)X|OtNp&U5= zBokzcwWVA<#&rtcteyXN>R+v$f6%UE$XltJ6ASw<@AY&4p=OS|^8Y1oj{Aqb5xM;_ zX{iyiqUEhdw$Dz$oVA@6&lh9z4d9Y5WoaEj8ALDV_N*14jp)?e=j-v&;BNMtB?Ix( zcw6G};B2>hFT+>p@@*K@#QK=0#D7#$$$+Dtf5+|d^{#ca2Te3wQKo5@|}z+(kioQ_NJD^XtI zcDwbqpkaK7ucb;qA%eJ5ke~GPJMAUbK>u+}S8R$z#;YYf&`xAb+}ZMCG&1?aH1 z3}{b`F@?sCBbX@!OFby>2Yfn@oZ z$)o+%d&3d(UT~pNrF2CZP6cps+%67wJ?73>PJb*`tLT=qt@0bPHb#gCuoiC%lQ~HT z>=h|NO$7j{^sAXD-YQeZ@pVl9hImAX97s8p&xfg$1qvclo3;M6;_r)$?RPqYpbZtt z{8D~^Hq7Y=vbP=^h1ZGMFWt@&H#jnBBx3h9FWR^Hr!pFWsD6$^Fx4JXTPA4Sn*kKp zFtjj^n9mEl5`*2jb`?$z_4Pg_E@0V!v+2PcvT7EvVP$M4KD(j;e%Rvb7=us<)c5~@SVym0d$YzE>?C`}0LQ7>;z zOSRA2Qh&N@x`HNpy2 z2S1`4XCnKr7f)=^&-Uo8`MQm@7u*m{MVO)_)Vfm#4w2;EDZiE>Tx-;Y2|%I!S5c2& z=)H40!z%-|aIg^aDRIM`hUUo$3$3k)^&CToyFf_~C%?0-)+g*!n2}yW4l7zosxTyM zB_t)HXDze5?>Dkzh=|&}jtwT83{4adRMlzz1JVUTYY=$y^%x%xmNg>sQq>-4_-3sC z8VwlMLmg76>R@1anEtyX=z*dNpFirO1a8Rt-J_A$>O>7gD>1wA;O4~z6Y11n28DVp zz1AYqSXj1#{0m9)Hco4>by<}iv`(n$9lVfrWHr~sH{?bVvPDmN*eOB9T9tkwOX|{S zWC>5b-x-DMg$)wQVoi%NGMYm^oCq&biaN53!<=Y5947#HJ(2CuAD%XG+VvDua zDMXf_Qh>`~k=iLvO55dApuQKeALWsHokvO2op#4wO*kM9#i!nZr=6H;cjv@}L9qxI z5+^M=k^=7{%Ww?5v-&I*GEUTUo@EK!!Q^W2SjkjD41@DTXxde?sZCkC8~$^&4PguX zntnU)-}M<;1_AUphh^S8^t~PF1eClpi?2W6Dzj+Cdlf}aF zoBP-Z%AS{mfG_J-j!uvt8;7Ye1m{UsK-)v?FYtkdc z5KG6ax#?(>WSRf|)1GWZS!{c`?wEzXHo3$UQKG>H2o}G|18e3uCIU152x~CEz zx%n4K`B*Hu_M;q`x&oKCA7g0oKsM3=aVcUK$;NQ@{Oi^@Hq(RD?G)HwqD|bbJk4C8 z12gavX2P0whBSrS#-v!Br5uUHDsKLdCj{$}U8boU0r~7|0&1iK2b|)F^H6aga!Bfs zOBZfY=czq}o0IB=-aPRIcJqXRr z62Xj+LVb=tJ81BtO?=QTjX5Z!1u>!luV3`QP0HpQc$7XzJn$v;t}_NHD3%SW z2G|v81@KoRnes%2ISR4$_^Y^0QtotY1@&JB9Bdz3jz>)RVBIxB2iVD?)Uyk`fIWnl z2=@d8L<-aHo$BRA@b4RJy`AtFli8Q@-u(MZn|KiFaHfk z40DU@iB8TUQHRzeMltEOey#~|8w0c8)u0Y5nBPPUV1FQC!953r`uXY=$|0ek!B@AB z2u8QZVi{i|ZgW4vh-SGS%C`(Q4r(+_9~<{l2%tjcU}o~wr^2dqlG8!vf&|BD_N&lU ziJU(Xxp}|rU!R6}iUYMf_rwg``q&f$5Ps!ix9{cB$dLacoT(RN(sHezCDnXUhoT(q zKHyw7cwxqX)cBKUpI-Hw@#62(F+juOSu_nvO5JE3bpDxh=(mVwb^$noTCxf*bWIGe z0b*owXrEqP-+1|RUOp`4=k+|pRKgxNcTc6mU+a#9l)W?qj}G7>zO-wt3r5@p(qJc) zhuGMb?C_{-)`H%@3`l+ezS-M?*s2kyF{3yK z2rtfqkct!h?)mY)jtlAWD?0h23~dU&Y57GQM+sjB!eIaLFV*(V`vqD^5cxj9uBkW5 zV^G?3d|!wc=EulDtv#&{KOjmmCfQbQT86p=%?sa2RRT6cQ2%Y2tuL(q_`;W!@jfrV zv|lOCDI*pAs|-xNqgi4L7nA8|z;a`;1&XnX#`TJff{a2VlI{x%HU(8=eO!Vq%|O3R(tT9=1^$KsO)!%UolrXN6b?2|(u_bYT&}(2~rj zyIgrr7aF;%Q69-HN>VDz(4%@$(iTaeTK+ny1s+GGA}S1>MOp?wV!Nop+7x2doQqu2 z3SP8{w9R}8(UbB?f6pJJ3wL(QUspq zaJFRa@2jDm4U{eRUC7?5*Pypm`q81!N>%q6A~gxsUFb%d$Nihc^DS z4KcVS_b0&xA6Z87JmCI19p(HYf3v@d8N{;ESsBSZQ6kGAuOHgs_Y%aEX`>_AkpN{s z{4O@si@Y(!5rNQDPT7@-jlw>+a5!a09A;s`c!3uNbCIhAjM+J59i3n6b#2s^&^P>C z>jsWtr24uY=rfeMqc(*SjprXJ}EMr~uflbK_uC_lh%Q<5b(YHMTd~?by)CEfHdVLt#%kyl1UNMD+Rr$pIBo$Yi=sDk^k_K%CGYo+;*~hCuRFHO3N95 z$ZBi9MZ9l;fBg*`U`kL@f4)I6vgM?D-QnK62#aX;>o@XG&KrqK zw!U;fa`7Ht%f|qe!9DkuhLo0r&}kCJNk{o@HQ9_hh7QU|tN%0W*83+lzMKvjr7Atw zq?wwLDpG0^VIisiJ25;Jys+eDopypwx(Z&9%`~C=#{A-qSz1X-!Eurr@RXqFEp{0( z)en_rT|6x=HGZZu8?26P0-nrcco5!m#EJ8jn!De~+DnIMmlZ1~jcBQfzr4$6cd zejy+OX~>PF{2TFthK8(W3S%-|KTTgw&kXM}SUAVN92FsSA;2N&&JN{EcCb(KH7Q&Z zib^`Gz8P9Nt3l=xIra{6+f=hz62m4@G|PCw1g|q#IL-bX1;J;uK%VmvI*jiGhyn@J zU|a(MdWy(^i}`gy(@65W$!`|p@&3S<`C7HG^);Wwv%$eZ76V6sdDZ8(!3#rY-9&C=>3Ya-o z1|Sn<1f(pJ?agm#Nu9f-nn(e}aBt&F^oP&fV@WToq$0`iQ75<(v++Uwm#cSY#vx<@69SnM`jhVQV*p&ncICHkBq!X1mZxKK1=AGya4 zCmZ}7Bm5mn^YNf5#X+bW7V!@J0fKMcT#0C`Tk^kGm_`h1tLgI&$3(Xfw2PY^F zClXxFq}_N_p|LAMY2(b|)CD9HrxN(i)o|SJ*2Te+{5?g*ZplOVKIe&z4&<@D&i}|} zWARQpWW5u^$^d|qgAC-z$ziYXsT1fp**_)CE+_6+lVr@R3DGX3M)(sN>iZcy0bW1C z@q{-QL!c!9g-Os7@BC!MWR#P-vl?vqUK)Viw7Vnkfrr*;(izlT7IQ8hzx#UPowG)N zlfaAd1*|z~ID9@j{RoSJ zq}e|nQuoF^GB!xo0s0@HT7-bSp4*9ysA1NF38YJpvC$>4QwtDfk&R{*Sh@!bswDO| z3!8yxJR-RXkJWTZakBGvF#MGKqk}av@XN%5HN}L)h{cE9__7)zd1?RU<3R|FxhaP^ zT;bg35C9#fQWvb2CV=6JH#44>EBDq|0Rzo$DR|7Ig3~SeF?RDdwz2O>%_+Hvgq{8k zzjmJipGZzJ!+mntX@U3TqJ!nN7%G?fNs^sYuq>6J^`Sl`%m$Df`{}OR>E0OUm8r%n zTbxU-B&xC=7MMU1EP0fji{*^Li6SP*wU{VGl#8nneFgz?dOVRhWR{9JcNzn(6kuOc zfRu(~?59*;CXahMO#$LGX2H6( zfYF34D81vK?Aem8I1`<%xf$+;=aZH^;?OKzWKlQqAT|R+)pM4(m>PfUA0}rhCl%j!r;2oj`4&@uszI7a99#t_=mSprTKw&Q!hPns zm6Wvf?sh17@@o}^)?Q<5O`8&20dm3~&i%eSxsHS=c!Ub|Y?q+1Y5VJgG&aN0C#4PY zp2YipUtjGMyRO*Bg$_Jo<)^m}soj;IZP-PoE8%0|c%H9*W^YYjcZ~PWpu=lppT7$ zd!0*8^d$Uie%*mD=FaXm!BDp5*03{^h_k7(O_67{Q@W2t-dCDX*XzHyhhvpvZMr>F zI&J=tUSj+T?Qi7XT`h|!ndR}~5B#qDvPU%4k_g#KHUU64!n+d&V3wVi^j+Y+u0 zdD!)?KCjH*Fy9CYWIqP=poi_rX2X>k=X@AUjGrI?5no?KIq<_M5~}VG->99V({qaM zs+*~Yg0CNu<7(SB2fMBZ_t?`_3d=RCu_hgp8zw(&Slpeu?75Oq|9{5TKkxA*5aVVT z4)sD6@kOsIMatdVicKts$?^bIGcNB8%s#)Dp~zsUXLoeZJnzT4*b^`C$&W ztIZ?oTa1~Nky*m;XqefTeFy0m;PQOMUyA(-=~!~(B-*Nv}u2E}MGG=fPJ z+n4G6HMB#aYK>V^n|LXqm*hgJ$~eXiUzhfEo}$lzhyin%@-Rwr(vwJ*88!6*qRugu zMVFJ#tW#bmg^oa@Mn=!>ZlYpxFbW5x6zgWfBl6lr2~rG0yx^Dnvk=mDOU>0N zSGd)I7;m3f!Rjn}wIb*wf={a^o3RTbS9&2NQ+qyfFW?#3y8}?W*(#R&sl1G<|$s zfagRiad_-`lhqtI4>%(s;RTEsQG^T5(aECHSekEmNc$tyM~`7UrYiPEM9QQWdfh(> zB|SN$(m(2U!CeXwvPL*KXoq=hC?|*05YyISjc3jt9=%~TR0VhzVO|4Y@r>Xb{3+bb zIAUC@!o^~b@&oe}yp`CsC;w$n&k`*QE$Tcg@H4|Hbd{gvy?id8mB!k1+BC%C z^Un*`^&7-FZCheNehl10`4JNZ5rm|T2ICNzp^E2P)j7|`YLV;tvy24xm)pUz+6S1*^?31_+x2n0OQeh%A<-9kFLcU4z4HGja= zt5-d*`*mx7aQjJHPd%xIpQDupKB49%&z1QxPC~;8fSOcQ#d$=EoGdYq9R%CFx47V_ zc${UCBU3=gcTUb$9~8x#1w+kjvei>h=7=#$fS3ZEptW^uMg(r`^7x;Yi}yS3;yf~% zJ(o-aR;k1=c=@NK6dE(W4n=V^hj;E8SrnE~dhcWe^XKS6C94rya+N^-I4;Q2Kz>w% z5%9cU<~_%cN_MNt_{4y^L@w0!{Or-aHN1V(6%uTJ*$4L!QjSu@AV0p*C?p?zL`Dd} z$DBjbiZG(_%PiJ(i|+Ub@q1+FiAJcU@lQk zL~6<3fczMIZ$)xl6vE;d!AEg4FG_8#n4o0NkBSNggeoW=1g=yA6GUv_5^O^=toWRc z#wlVSP>tVTASbkR37$KxgW2ttj6B}1`?F7#Yfy1g{#9BD*mDdV)FnbXPf*P1@-_2U zzCFIllW_qqMrZ;&3b-m8&Oa|JV@fb*?g%~e@;e!9PU1?Y`*=cmPGg0<|> z`L?Z1ab0Ky;xL}RXq=%%o`8YUn3=mVj#Y@oJNb-h)Qwx%r`a>*HHpH>**>wDMNXuz z>Vh|8noK}4abaqamX-Ae^ngR7KEYC;SoRz|ww3IIfbg`hM+d3x-4{voVvcx6%Z=P~=s+u)IXOVhgzM^whNLjEA< zY{Bmw<ufbp5&Oa~W!2~k2}W_kq!VbKH&E4lKm?!vrPlM`J=(OmX^4fCM9V`@)38tKzj27ZgvvvIQ3bq(n(j_?oqk zt+~iCJfxKu=Ri?qK^*%wh}!xb4t(PE-fkc_#yfZ~bU$A3TbMA;mMvl1{jFllZXsBn z@N9a0a2O?C!&GQZd9~Lh-$BE6lac`J4NqcLzANlDHtu=OKauCd>G<4%5(gd@Wow*k z3-D5bao%)!vIQnC@ZJMQ&YZgjiGC0u_!E8S_mzR$O~a;b++Ujxn;2R9J|EP}AQ90v z-`PE@im1>Nd4l|r$A&QG~NTwpBw2G#gMRo&95R8JxEd| z4CPH~9+cjNapA@>36cAW_ja8UJ~Rel@P0{tD0Jjib97;9q3hS6zYu{QEnl*`4IQ!$ z6|V}F)U%2DVKoEFBm-D8QD?uqWKa}LKz-Q3;$pSFrF0oO^1DD$9!PrmBXVh zx}q*PTW)m%hqPvu8Nf`w7&DP;dARXx=Ult}MJAWkn!8=341R(hsYJ|=qn0LCQh~+{ zYeA+ZHfN_pI{o4r(nROCTB}vceX-uivHruMqf|rEkagt**&17jRxU**Br1oHFxTKn;f|R&bRZ;sae+%zj~;*2KL7dn8+b{o#b7@O zbK5u|e25PUaIGwnd*I-YkIQE#_%uNep@rb`j$j_HD4vpg!GE}2@#KK?zB5( zO`lY#;DLc9tw-3259$cmf}(=OKHltog`IPkEj;Ym@_gTjXWM|xfeFw z+ZHYssIW~1soT#~NO;rENO^J6NO)7;NP5)JcItP%uuQTSLbk4DAFINhuTB|TkEoP+ zTXv4uZa1pohqkMaJJB#l=_5M2Hm3QpG6vB;p zV1g$VA~oy(UAW4OStLr-rKN`9zcBO%g^3C9$>aqsZolp@RB~0wEWrcksTa*u$IFNE zA}R&rnye3&hjmm=he^g|b}BC5Qo6>6&8HAIG+yXO{o{yh!71g!pF8b{zS4q91}h{b zT`0LHwZAAFe!Weloe4?Hf>DV779(?bL{PrHjQ@3h&oCQq^;KszO<7W~l!d-qoi_@0A)j8n_s;a;sepwzr_ z1g(0fG(yPTCY{oP(lS{Nc^YC8rkIx~?Eid2*q<+F;^GhBe|UtSVcUh_}6VDw)L*fOItMKn($Z{LWt@x();$Nf()Y~FTU7afI- zMK57VFIBiU`XUXDm4ZgQ1egtrLB|0;w||3(#TS<%A;v8WMV4i^SZg#0`Ewu6lQBq= z8$D2-B%3xWD2~lhWb)>Kl~~a4*Nd<~T>2GuPn0JQNWn4%-US!Zk0A;p>288eTgHRZ z$uG{mFF?UAe9g!aj;`#GcZCPt5>Ad{+%nn1I+VHxYS{rOpYh9`Lxcb4W~zdMQALjp zEz!D9VCA`~*R3dsR4peG`Fp6>A;O3iTPOJMJ-X1l3(RXpRp@)5edc+0kj+CvDp8)) zUj{+Zl}^4WN^3f*IVp@SjS{?DBkCAjlKedqtBm&~*i~PW>Jt88;K05>q~H4v(E%OZ zv{!JiU2brb3ENx4ZjY*39&I}ZYkoI(IGKW-Wy+Ou=H_RqG9D8|j=cYp$R504StFhJ zI*TA!NdQGWHoxjwlO$h2xV!Q!h$Uq=I^!RloBO|;qqdC0oI3;tAMlW4#kf2=$4L=6 z$ib9#7MO5^4l{W-sl`C|_0P%26*{XUG)m8gSpn&}e-0|Dh&O8M|uB_UbX?(w@`bPbuOSE*hE<7Wa%U_cNHw;!Hm%B?$Bub0aY3L%h|~Dk;R#Y_ZC$ zqkA+#B;Jb%`0lMqkDW47vYB}BMCrv7hrJvX=eLV&zh!06LCn3JiLTs+VHFnf$du6}DR>{;l|nii9ISxp75um<&-GTLQ;vRtrrlD6ZtB4 zuY7}B6#Z{NvU_JYXut(;pvw32PfmVY!BSZyS?|7DaE`1Gel!L0G?_t~^*qOtBWzLQ zOcOh|h*^!~+(lOW=D82xH!Zd1)7G|u^b~#x?u~d?6cT>!jHQ=V$(n%5^ACO(ktnN} z(Bj*y56&x#orU0Qg*=fo=^g}u2VV2#${M_mX%G3U`r_a%J>Zc}?V-ZJ(<>4nzYUgG zxx0;x*!fb3M1(wbg^&2Xxf}Uax|!mTZQ(f6{ko6VJT$b@f}kl9`}8bIn3n$A1b_0_$Dw0i2}p<8~n9GiVz5GXZl z=~v0oApUdEiyln_<@EJ0j*5jF%T;+LVNhUPT7-Z9Scprg?W_l{JI2;W{J=;QPKn3g z?>__OGjyV=_WDGcVs$pgT^{$Ou;zR^a`Ht#BUDEV0=Xjk(*>`ZXdcUc*j=)0Q>X+l z!RLO^1#N^rRx{FXDp**iSuP=Er_>I?VRO?3RH6GHU2CF)m$8)P*1fu0@oZSb(1T z-Rdo-nLQHh+mvyykC2tjP7D`O5|HAlPdEQjNv=)Z=*lTkI!$*2GxqMCOA${Y<1Pn1%C zAQ!fowC@#F(e(TcU($ymajaEYJUnYh!E)5N>Sod+0K=34Zl1i}iboK4Qbv@q89rF% zOHO@l9(7NNzxX_ZnBglvF*GYyF#SB$JHV<1=%hwz35Jpzf=;pc0cArMEo9r5KC)8a zpR=|W(bmwj*xlvd%E}$Rn3_$OBX^B22Q_ONaQF$e92!3O7FE^!^RBweJ4682JwtPY z0#qD^ax=k4ZSC?y`TK8GpL$IB6@~s6J+5;!jSS-Ni#5;bcxB!Nv}w}HvmS;slK>~z z4(9I8_`;<(U4E#ML4Xg2@>w1Rpgx{K;G*jJlVL`FdA`#8XLIM}c}>5k`#$M$O2A+p zA#cau*UaTkp*!I_*SFZ^Kv&w|(X0g)nQmRCOP@lzQGNK~c&*X==dD;vMic}3eW@qh z;VUg-+3?RC%;;I8Y{aIlbugC=J&0oFuyYH zg{8j;D|rHbKQ|h#RDbGvGF#NYv1+KD5An4owB>2`v(IF%who-Bb4oq2o?p7CYiIJh zRvOoL^rpJp5KXI^T}$k_n~wJWMa06;e=wnZ-7r)yUi=^g_2hEVl0O?$x#+#>R$dIv zqj{5Dm}W3G&wTSKb|R_IH<@{T3txy9AMc$nr{7WQHuuaA&)qqw0YY#}HN1>nvG|2y zh!9N3yo>t#)`gvV#)L&{SGG*pbqO6i*nQ7G5xdb#y z_Lz`BDZ;E>MGfq=} zt`sZ*p-%@7C8rF`UfRb$m(3bOCi9D}`ubN*&8{Sx+Q+J63`Y;s0{#h)dnm@q#WVJ` z-({aQI&`vxV^dpulmFhe zDbcfaK44G$=I;q))6i+5iAc)o+Um&MxJ;!T!@gQgCAK10OB=JaL{4lgg$!kQrqpro z`Re!;e+u5arh4TgIa!87I1AK)69fMA{&b;%^jSF_mPEO6+<}-UwxxT6y|z@rIcD>+ zdL)*W7PJ4+K)!ww60WS(0+k%MUKBolNSY~mU8K3>&(LVGlVX`Q(a^q3 zX3k7)iW5O=ua>ZZfUt1o^n@rOw%s&$z{O2&YL?ESht+j>#q_S@Y8TfK-L3UaXJx7y z6+|hP)6~z{EobUl^Dt2FizD|TP{gt0`!e#+HU8xoL6ffA#`X}gnV8ELk4;!K31P#% zQqHHklImVHImQELYmh~*M@Rky;f{BGf_og)rTKSmCwO~z1KkgQhw2c6`cbFN!!vI+ zSQFjf*USNyxu#}bm44wS7({HvdG1<8#Z)(nT(NrEhnBGYwGVe?ey+HO8PHcneBW5J zR=M0kS}rn(y#%x7DnC>y)1A(gr7p_9?u_su<=(0yadFbM3>QhH?B@dA#CO7VGn$^@p8BsDm$c9o*a6nE zq+c|MWzgi z2NDS_4Q`x#iuLm2B1D&!c*_Q628ST*z%qbb8`8V(f>hh8O1Epg0w3Ya$FZ-}Y zXEr3;0t2dhNg6_;#OH4V4K_co1h7W)ileINEep_lP98s|SX+GfZ0hFOi;rdhLnMM?i?E*sXS4^50^lCrw|XPj&R!=iEZs z2bMVs5CgcDmgQ&}Ydr8i>Vsa^08ceckwHpEw8e**rUgQH1HPSk^cdgr6Qgx9_y9f% z4K+u^E6=k`+;z<$kpyzM>Q@lhfi4Sd2f%6M5$a!A&S^j(_{=WRWb8!eByRV|vEUR1 zZy>udI@%98!?6f1;VcJ=vFM%}mjlBDsQN1lEGLID&M$ZmsLH$n!r+g?#(Qq($^gat zb%B(eBQ`gE6`gM- zy0gYQS1)pq5(-)SniIbJ^*HHH7t6utE2!rb!}vy?z{*bs6MPIHoe#{tScw&JI^XfK zzMJ6N?bD}_*1Lhf!i6{5)=*M9@~X70_OVbvPVZ;4kQY;%ei z_ttwGAUIxn#OUxWsPFp&^1*4ugZRQh-}`a6QYajNs<5--pV!60o-|l4k)sG`%P8^W5Q!2!*7Ae1Z9v9)Y zz?Bi3iavag6U0arEfx2Xf%o&p3C$A6TIOrRA4Lxl9u!OES-AkaJh3mr zI`W}VCyTB$*MlV8*XIbIvc|4)Jq&l=X7AP zsspJtUzc(|T|&Wu+26*XtkC#1fAMahIHxC=QJQkKa=RvUrlEQZQ}vD@8ziOg2rpEG zLy89fp09(0pG#|`?r^7L3l}`V}%0uX&svPMF5-#Rp5Esl@M8Q9#Vi3~!O))l=LL_*^al(MC z!BAq*y>ZfX%3Otr!9(;|JC2KN6Z0#hKNrSL*NWjusl@~XjBgQ+l{WdEJ|HEp`dLs6 zQx+XlxKf1D7O)Q<4jN!f;Zg=UhN=`(-v(87Lb0*);;F=FpmPGDi3uyhl2E2VERy{D z7G3muup2J_ZL9{sc)zxBeU*#r^KF!c+5S(*lO z`_PW)bKf!F1D6&bB8uRlD^!595<)?o^FpTZh5i)cp0B`brorS=Qqe9hDC^e}J%at7 zxayE64VxukWmzi_PA1Db8;57LB6{9MKzyw7o}9|lMMP*qmNCu5R2@G)wEY+z7WDx- zBShK-Xzs(l6e3qzc4mI^k97|1GbU15*|9?AALWHaS?A^ zqC_m*KnOdDH?4R{di*{JKldrnS)E6MAi1^O3h5yED=Det1l$ru>?P>*{rHLl72bay z!~NH)Zo{p7l~$*CRlxoFJ@UW_tARt&P@4BT zFFTkIR*W(G?_w+3xI#BKcD~l@R0b*$N1S?af-f;jf^PJDcUERk~<;MH+n05v0m$x`_jXM8n@;U1p8HS^8!!2-iK_=frs@N3T^jogn3t0%Y*7{$~c7Q zo}v*UzQA5*8&lT0hwAK)r(URdc;g6sjG}ES9^n2B5{eAr@Uem+eU&Q|RJx!CN8b6t zBt@$1==665z@P|0TvItbrjc#(_;%0QHM?vvEI#=(1G56Ni82UtFepLGxb)bQ(4FH; zcTW$gpca9B)#$U_kxNaT)V0&PKB9{G<2f%x_uNkeI<#h;H$g)O6hh~N@6Q<$_8x8o>_q;m8ySQz zpe}3%ok|nKT$n%4PPYbCT6;b1{ewG9XsS0+bs#4O25eP|WEU~!hoFbJCf>~oFU7;i z-Zk87QyGyi<#jL&bW6aE_TjR(QEZuVe$)wHR|A$y=lxo3v2f5;1x9Peh*vdrlPF$P zzan>vFQB<_YfB$FNkaKpz++1`L4-{S1YE)6d9}X!U^l4Bg4cczI$U5PC=2Oqy;<~6 zu4HPAExR8j(nZP`OlX(4T!*V5j|Xzy<(U5}=Yp#ld~NcMn)yO7PAt3Dq2sw#gvs&d z<%x6@*<@=#WY=XmA9kPEAk|4Nc+dvck~h(2Os*Z&Dg|(9hJ55!d})f0nQeanrN*;h z?c>e0lq?>V7hDNh@u_N9k@xZx~TWK*dibk@4N-sMo84gE4MhM`P z3i|+t)2YZJta;!N{>6c81xU@mMPOZ!T#-j;yS0CB`OYvMAUU0sm6d@|tyFWbpf(OY z%pl~K_lNGGO=u#|N{AUfKNM5h0-xJ+OO3~q_3xTI^Q`3hqG--^(Lz{}O0=HDcO~!i zisLy$9W`5pWHW&Rv_X|(<7&X}@A?oznC_2mY8hf{VCx@qswcuaK)B%S;Iu<&2aRs zX_DnmIBe&wkre#!bp!BLBSQPv^+Tu(7@@vh3xN=hlA9aum>jc^)Y5)f$Kcs6(KVM| z#@&+eq()~6TEkm{c=tc(>~4oBH)H6+waUB??MV<;Vvf>PQP;mI$UalmGwzJRV4KuL z=yjHqFdmJSPa@pKhVuce3%k5dR)QR|Rt9u4J2O7eoZ|PGrS#xZF6B z5a@tn$E`lx-7+dZC{zjTnki*XkNIv|+(lHn&}!_H;CzY$!;C7So>$wdC#3?XF+O5e z!8}jyGdB1Pv}O8+V++pLnxK}fw#*a1zhTALzK};g3@iA-e2>hCv=LiRk$2PrwC+B8 zPRDme==GmnrT5}%b{go#G3dT|WeNf=g%6V72DCWd)F)0C50zIE7#iT48$8wcIP5#` zh!@3#^%^u1e(n-M=}Pjd(9#2Z1}}_^OyTvmkj5dLY$C+LWDeO=p zQ(XZ+LM?CKxVh=%8e>VU+YRJlh1UVGfjbe!2G2X`gcN@PMZ7rM%oKTFTK)>1nmon? zI5SNM$!a+SPCag}LdzoNvqwexQ?_{0Wh0aup{HGb%C^pkM}|vpl?zgoY+7$nc2(1U zFnI|2#AD>MTPFjLlGnH@fzCwsy(?1hq}n|-OSud-x_Lgtn!&|-R3YcJ(FNMGH-}cp z35?r=1*BQ>?ImH@Fs>|lh!zu2yo@okhHkzI*T<%h89Ahu1T%2`bflD#TMwal+*(>B zs3{*%iclZ|QJ(Qy{^cxp{qK0`1v*{@ya_ntr;9J3PAgXh3=9_@}^ zJ_1`$On);sizIZ37X2ma1XP@GZ*y~Uu8IBxY7+$tj|mZ&y6r;TA@i8S{O(8_LCiXc z(T-yo-oTR*udM;CaUP*DACHEy7Z8a z^_&E?bH7PYD|xt%BSD!Yh`Biiy^9zSf;#pw|B%kls*8m=ECgYa(Eq z2%%_s=5Om&%r8D4RB&F`uZ?H2Dqs&W@OL?Cow%SB6}Zl`1yayfCp%vipKMd0c_#>K z`o8CPGEvw)4#?=yz}=N@+=xwuixuF^8$^L^F>mI*tpSe<)pxA%dbn6q7nrGu04z7n z++rY#3Tf6#g%4QnR`KAc9Zt0|p;&oK^4KnxSLAx!cJ&JEvnBG`9+cMYmk+?VbEbV%?7#cWZQ%9{bkn?2??K zEYIt_{$_fHgYV<@Ry}v3hl+>E>3!`fhq?lSjF0lQRqAHo&vc;DK@N4pmGk|Hr29St z5y1hQ(6GnJ9F#QOHmK|ZYSkKKOJr?(A$-JAGo5`d->ybq`D$e2SmFQD>CCwkfda={ zFq!AEYI@kjkW68BGcwQdo%1{?njOh(Kp6Wmyueb9dL{cxx(iPI3aQceBl!P9+=t(Z zH7apOK!F4gX{rwWYrHMJ-puVERKS!YCm9^5)I+6+x)-=D^LX&!iL%EMs0dKKR=?M5 z;FQvYNTx-77LPO9_OXq`bwzkdrI@^vJifw-#FZ*;7~ix+FGe(*WKPyB*$F$BwnvUo z51Y8y&7oi&b$A7v`tq?fqZX@?mWc&Dq8KR_Efb`}Oz8TPl(!4_iiv^d(9g66i$`T6 z(Yqe*Na6lz;tQoi_w;pu38a1O#*Pj`!pd*7EYgO$=z(-^3~YOg-RRe}0>MlHa)vjr z!yFsvJ+V@lxGCh;R9ySAYsDNO@Y+33*<9kwcl;@pmgzY0yz_^F0f%NJ2Izq$v)z=8+1Yr7uwowQEhK0w9LiXL5NYYc zdPv^@Wu&g!e<3D~_p@ZQdin$eppblrp;O-3Ucq-5fLRfa6DkCNCh%Vv>>KqXxf5~? z8V~G?(UWLgGAiK3QmPAX1y3~{r3;-m9Isu3VuNY6-8-4>QVHxxkl7zXy5B`=MM2h~ zk|rIi#iL1~jlnuLVckG^(Zwwz&j#}pbif4mb7aDd=izbI>~)!3zY{AUGl~2!7di00 zUOt`4aYSsftrjdj^NK-RQ?xao8Nn#|f~D|v+m4Y%@+ZTnariYxj4sBf7o5@2@ErvL zn@)vZ)+1hQOH70EG<(%GO~@N~hp77EE>#cVp3(ZD+>EUQ;~Vv~qWx;t6zxnCBG&m% z=OZ)*7a!l4Wa&lbqsPn&yOk7F2tFM9z_xYr+j0J^=Z+Txmw>cY^cKO=6-L4Hx~FI! zGQVu=D~|?I+{6`@>R(?n6yimk7!svQ2JajR1uU|n^agAJ-_%j3bwOwqB5S9`MIlTG zn(h6ikD4)s);UME;cqq+?2GiUN{LFEaT?&MBNxHn4qSPKlh?vyPYG#VpVowt!@e!{ zWR}R_Nc8D!P^0*z!^Lox;`UXn=%)OeiqOk?;tfLVjtPUo`6crT&(AQYL=v9c-0zv4 z@Y$Hqx}ir!C%d^Z3|Fw_8e(zEN1C}c5a0O7o6&(Trp2OwVC#6q&|{4-QtiPwF6h9A zI^^Wa<&_>H9}0#Mq?eeha^6@d1SCHx6;rS|4v(;W601&olYx;*RB5R3r%gA2z_nme#aj9P50;g&_Q}zyd+u~(XHBQQo*B^ z@=u7@x=2}J4SnS2UB@y3%_)~%5;LFw>gE`<+8xaV|ECMc>RaK-b|vF>4bafK#Ww1V zrx=aed!mNhle*(>k9l=S(G63}n>7G+tku8mo4y0r4}<91AA`7qsRsthlEu%pV5g`u z7<#g@hBZ6F7gYXsx^p+N(-~p1W0CFq=6-**Wog5{XUb$<7okVQQ7`g&qjEyH0@N?(^(jArNv55AHX|mD44G{`vz(1v=`pQbrrfZ`-*Xhy<5ey6@=oKQBmq)+evo>%MteSZG5Qd`^O^>udM zQ(pB!oqwqy@bgT>xfeG3-HcS%!=9*0n0%QS19sMiUlFhsx%u5RdbBNi-4iR_%wQk> zxKx`+*{(w$(8CV2eDo1jv*rJghoPY~FZvPL%Dl9L84Q0zqX;au%s}+%u38o$$f7II z#pbVjW2mPK_{UPTw-q*|(9w*Jcj2$-*4B2vy|fDScs&ip>UHXV4sQ$P0cMr;&vSJ{ zB2;}7+S{@Ug&I{ zR+dj_cL`WtpC_jTenRV1FGF9-_c95oiKdMkdy{Kd-Db_Y8C<2{=Sx2z|I6zgGXx*- zK>-2TV*>#Z{$F0t(dNGyJzp)EINVmG?weZ7(?ZaAFd;!;5kdG%yJ^@in}(s0$yj>0 zHLMMoDpQHSzY3%Ly(2#(;{a4^1B)6ys!2m$2_O4IQCWCjUw21oJ2iQ^`1%66Kh9sD zuU@{ND+BKbJvRio*8+UUNdDhFJz2aa@2`)JFArY>Pygg2fctmXwrN5w)8Upv z&(Gb_!xuvjf7eC~*!6a|=flg*zx$(gD^CBuDI~4YyTkAEYU%pUYbydXcKK}u@xkF| z;c4OEY0K^w<-HlZ50gzMiY>s%~EfCHPHx(S-=2Le-%>K0w-Wj zlf@kt%qI}Ch@vh>arZn6YzW12Wo%?E_(v0aruI?E4*3P;WXzna5{RNXaFa5J-#ycA zsusgK+1Dou;phJr3T;gOO`RQV++<}5zn9K6u_(nvHrjtbeg7iCgii*PoKl|sd$U-Q zjoN_I79gu8GkGsL!|VeDC%hp7p+RB>%N5z=mFkEpAMHXo4#DPiEuqpHwR~)}39uQM0swZ?3e?k$B{>L5ehhBkfQ>Zos46i_D^V*7gc2ZVOS`O(iZd!_KZ`C4Tc$teHn$2ykF3LIIo%Tbi z6h2cVs*Akrv{(zykTQ&M#2>$oNie3MXtV`(-Xe}ckVh@in07~Qv#iWre{U}SlMfKa z4WgrgWgW+;lBIo6n15?g-S654v%yeG1CotL%#+{ zB22J;U#OBY^@JY}mrQAOVWBL3qTK|gr*%6v#-H+9lNObZPC(8z!77~+X4aP-k=@ap zX?G;eE=6`wjtA$>IJd%!0;0|dYuA|scV(1 zn+!E$McWr zkRsmD!$1I8LchsxkA0fH@k8gdN<<>V1Si}a@{RvhqW|)m?bYK(=NOMZ+0-UOIqDbO zHg#+fI0=a%YlnC+zFmrMoum&3&Pc3oP$)k~HwnW<&zLN%p3yW+8P1 zBNn5;EQsw&TB1ey{{1#Ro$-&z<^Lz!At1$~PK6nj<{!m&hflWg2p+NrDp?AK_Q-5> ztjJecP9W|My&@QjVDnRb6ym^tV8W};6)b>eQ)=2pCe#8ujAw4cRnp!9R$L||LC_J` zPmlS9pM@!nOEs{VNzq)cv{bzhJ9e0*hInu|*{75twaug$KnxKplhM^-%^a($(aSgr z7gQ(=9@d9jJLwRtI4q9dJ!nUkrOZ$tE1@JxHnb>*q`-!508M#0F}O;8X=49ot(nGQ zgBS_DQyLo&N7zR|>mtn#kAK(qj$(9J zOfo04n|+4r7!o6{Xv?H5X?rwZ7<^O8221GrTfK_4$` zi$5chpeuCyCaGFu^~ejtb964GS>U;VQHr${q;c;{uSrVb*0NFWU@-hZ$8;5IDF{P8AL02*l# z6cZW~l`nnOD}6aHvz1axB5*902)%@(4-wWVP%-Xnf2T8|#Kdf(AT67ovRZKdsl@y4 zDvbkeM;pnKwxl9)<=kKJDI)x{AS2pGNnKu=4_p-iyrE-aNsSJ8HFKZN&3Gk0>0IVX zj?xUkplDyt@?chE)esWI2`ba}+EcQe!x~fF1uFSVU3ALH=+QwScoBTQXYVASWl}>l zE@cK|D}PB9R~N&wsayYA9zjToJUad&(cV_I2$sr~b$RL^q&^Lj+T+bo3dQl(!CPXK zKkmc?rKwI4dlpwzht60F*&r`;ZjSmKSYcNq|z4uE_5C9a% z$b6qL6zC&R{(0*tql&0mXN*k3G_GQo7o(RI2vT-EnqL8F_j)N{_(jlbcKcQsp3VCBPO49&|!YtSyl#hqSWg4kpcd*Hn$MiNYt=; zdD+b#E%292AO=c#N&khA3(|7~pL$QEi{kp)Sa!0=)cit2uG+fWpzu=QgkeI=vp|1i zTv5rFi=0@KO4mV_A1&FSeiiYYoS3zX;*{oP3cu6pxkK_fD;I!zum8W$>x0x>7Xk(d zD1;RVi0J=9FBe0*|HfW1I{$^eFdfkE3E>fcaSbkmBN?qu*Btgzbd+WBiP*X)5oa~> za38jot+L3eCk$!P)d5|6j2f{K4DA~Oy1ZN)0|cDczL)@LBJ6747mb@OTDE8Z1kJa?~6Od&M{9Zx;GeY9?Hw3>Twx2$JZtrHx#W)Z9$ zntdEk>Hl>4Hh$rs>s@|4DPNeSPQ6|YOSQj^j3E5HAARpVSD7hRXJ`BTecRjH%J@0^ zigOnb*rW?}LFRXLXyvHQXS4tIefjYBZ})$xd;2(6kU3l7p~)Iq*~tx=Wwg3?N;%p! z!cMN1S{hkwDvefK=!91HOQcIQvo{_t19ABUxwEIVYP$6RD=eOsXA{^37kcJuwt$^UsFt= zf@>l{N?U9yrg@4A%TFCu`y`LR|NCy>L*GxFZt+r=cG)VCU$o#tv@Fv%Ijn1QO`Maj zm+?@EX{=0Zgx)GTL4_jR6r*ZMBvjD4SV5|}gOyd7I|Bj*q^wmlyxnbHejOnQ*>)vu zcUIj1eaZVY_KPvHFM7rEM;M&v%KDLQOXKOD`wh4gzp#&sD;%K35hcOqinT{7dq;14 z3|=-QFXyE?C76YeqV=HiXB=?@>Yx(!jg`>ps)Q`$SJ2q6Jd$CnBHP=qe?g7Q?VHN9 zl@;}6vsKk>W0ryHGqOjnz{w&3z`ThD7g8;q*mTg z^$ehObsJ<9@|3YHrDaSZQ4SlC418U}sza!-|AhZ;jJIq-bcdV?kbw8;-{smzYz&lS ziGl&Xu``rk1)yO;$ z!8{v-TTq7j<6^ZFZJabXSHe0OJ3wg8?1Ff%g$b7}-;NF}jwy@Cn`D^HWI~CdnX*JI7KsQ}=kl`OR%xv+&Iw$IosHwg`|PF}r09 z-m)ydWiF<&^G>!geqYN!L@yaC=V7wsnQs}xoTmrVD3i^(cZF{qXvFFf0bYq)g0Fhg zt)BSft9Dk5e+BanF5GYFzeZ8PYVDm~#>97e82HYbLiCnBPZfAGa){tfvjH-x^@v9! z%Fi7DYlb<=Lij#9bh4{G161K00zHTt&Oj(KG^0wI>T#mV5&>_7*5#b>o=wU&Cbe#+ zsBc1H;F`={?c@Eb%U0^U;b%!fyT^CB9mgXjz5`ZpRc}OQB0ZXh`F< zXBE5#bl1LXCBvC*e&>By~s(#i=ydhk)vx6FA|xO`~E zbCY>k-RP^rJ?uuL6=iO9d&2uOSO;-AgYl(QxGAu()1&im&*TD%#VmV+yz-TvW(z{x z3FmNV(WhCfCu6;lagQZJs1f>+)Ar)>QV)2MUVgQ)VJ+*IC{2hbs*PUs4N0tHdtOF~ zbCULW2Zba+0(0E`^J-=2C5;<{wpG$*y!!;M?4F5?I4>lSoQ?*ie`jPN*JuyNW~Mxy zJ)ct^HBG5pzHt#KYBLJ(+15o=38Il`6fhc8Axs8=gYbeM5+StCiezinrly`yWDj_w zVERQX3>W2vF3uhe_j*l5JJq1J_e_KYE*!zmgD~=y5O2ibQ4sAXZ9Z1+JR*;0xP=yU zX@0o{36E`X$#=Br^wWpNjp9j_@ktV?s*hje34bi<@@7%b(b;OGIWuR2MSj7peC~?I z1yB!!0G?J<1FT?yD|T&WRHM^hIy53UW~5`KgwvprGp}gzotS9)r_@;1p#8C&c3zQ!)m5;I!)W2Kc=pK%hgT>_#i(PK|hiudnfzOCQn%kJ7h-Gu=JH5aVEsV?vS$G zDa)RViNsM=R#}GC456(*3Df>wpwX~PBlSXLnBOMBCs6bj=0JtCs9-QLZ1S&<=uSWE zcCAn+@z%NEyP&_<7$Bk4iee$v$_8`@2N^$c11$aGecEf$nQ+HW*z)@HCmfq!p9(D; zs&qL?MsZ!_9jO^`)n7>|Fc~W?40_nDc_ru}v96J5h*X5T2jhQXnDApsbup8*y_2Y! zj%R9wGMx((ZOlHSCl)fz`Z4jcB(P|6l~EXdf)$!c#P_8y747iI-NuH)+L<2x+LoyCUE^nrh; zCmrmmEK2S_bbVutXTkdP*tV@5+qP}nwr$(CwPWmf$NY_L+n#stxBpGvoAl|S2!7DB z996<-%{IgL7W^r&NImUAFw>1dHbfE(a@&whI?~7Trhr!AYS=@m&~woQDJzM4i8YQtK<0sV3yRucO&KXv>Psvm{f~4*wqIzFZ3T|`M^qj4T{gzkE{LO;S@8-AUR50nQ@idU?`|eQ^nz8e6XY{1*Pk6K~tOPBu|uy z18^Z;5W`Q^j!Dun`b5gfon1x^c9}Fe3@8xX@X9laO;S8VmW|;n|5E1#d^-YuhDI;J z9ei6mF=p~Z)OV3&D68)cS%T5CmIDcXGbyq;Sk1n9X{Q>0Hj>$NuAa>Nwb11VQV8M zI(4Z=DVy=!Tl<{{WuiK?kuwpx_8r~GZ@sW0qUaL#K~=wFnVSzq#+z1Qj!)3pa%o)n z6E)`_#B-%H?7YBcyXRzwcT%phH9WVpMwW760aC9T1)@~kv?YZkaO3I4#_(ghEKdVj zDt^76%Lm`rd)MFX-5%rUgX;3^mF3uCB#>Fc$T|KA3mm7AzeZaUJQ9D|tq5OuFuy;~ zVOfc@^FO8g=9|46!OXAZLXL27{6PLUi;l|%%Pl1Y0@ASt0wVbzEZWuR|I%nbJ*|X8 zPG|oD1Eo_+hDLqWm0bOP#O*tTo^T^vf)6c4=_y(HbHgN6g0J13YpB0fwFH=R;|#%Y z(finN5K9I8e(%TN-wOw<5oPHDa(e{MZz>wr1|QdrBUOLPT3CNRet!G_C|_aZ`tdgY zZ|E$<-vtqDe9pXi7X=3X7_axw@9#r9eZ7DDew;mf7F|5w6zE4~;J-dcGf)O61-@nj zekKDxR~iC}ZOb`B%R&YEK95!c%3lLMHf969UJQQL2{^;oN2@;1auja$cYH}lZlqyE zE-NwueBZj$rz<=^A0=M(Zarw6xQNGy<-U-R2nb$%U|=DMU)X%#j}hr-MGOXA!i}g0 zn8Q)R{5e_vDeR1X&JxJhBc;{=$A-5Zn&SuwB>O-sx=C+;MieU`Md03tx>_s!6A!=U zz9e%~0WInYT;`v4G{b}%2+9Vt*+U)nR>SZ89l!EEAdcr4>O#MS`)D9Fpr7J!=fN*c za8{NQPl^w*w8Nfl2t;bS_~#uKWR+p2ip{MJZK+i1OjhHxhy{|4l2XX*v!TOMLQy=5 zg^xIO-je2GkvFIe&b}ba!hZt!JYa15F zX#L%$Hyxc4Un7ci2d`QO0NeyFo|YOhuo`p3yUR(x8m_{WSHuVLaR6NsUWk1oo(E`5|;1*-nNh&CKbw1cI$_LUz4MpxF58ppDI zMSVz-L0mMD9ME>5Btqb`xtuZpLQ6%Nt%D8OEje;rd|GPBE)OfHNvObKk`|73>mqKe zsPKIMI!uI3hY0=b)WZn>bg;=Ts(wf+G$}Iol2}bV8X^4vSdqA7t!fl|m}av;)j}v~ z&bgb;*(IdDB_y>*Iq(kcTE*U00YyHL67lf5nqt;4la`A7A;%ymkI>REO?#8I}E1`5s#q%FNRd(l2Gkd4_~t2zIK>V@7cL<>51vq{MA+Z7xo!QwUws6j4!d zz0P0t;rV1{-o&+z&I`}0QgN z3!5jAq`l4rI!2n_MGX2F4=7-Sr}=_R1+a!8WW@+`u$FxUi1Ca>a&c&zXM81oVd8TiEag?|c- zi?{?2`<>2+ja24BsV>qX{R!!9{zbgf_IyHk3o(eUFtAZOWoe9%OUO0X+Zo0}zQZp@h({x?YXH?xX-k0T zo=a-Y4TB6274hLjRbF~JCb9$t4dw*hNsbuKA6ON~CTy_Ihq5cHH+p(w-A@DF;hjUU*=F@9+=pIHjOKCfvOdzGPG=AjE8w1 zdGSp3pc*dan~yZY;nkwV!?^lG-Na>hg+h*_R0JrigwVqXErVfz-^-Hb7w_xm#94dpF?kzlUo{Q{qML$-s$ z-plu{)lCs-08lk8QEp)nilNbGP_Ivl2myzMUl_@;LddfHR$3u;yq7U-hHA8k0fJiD_kk5$e}F#X$FGE=7sB-aaA z%Zn5~2X`T_VyVoy3|r)NR2si9Dh}TXn`bVBN2Vh-lsqI%f}1$oFMg_+f<*6Stm`_- zRj)PKcc)eF%!!;FdPpDG476(*t20Qb=8Qk!C(k_4Rp7haY3R*3b1^i0lnJ}-SMF(z z_?-=DF2~S#DBwe$J?rwyRIH}~WNkBa?xrvF7)HdJoJL_N%N$_xw{Asm&x^@xEr?AO z6LJSkV|$dqs9%b(zJ@J7o?DPZYQ7wxBG7%NM3o1dh3qlUdIKy3zbeS_2m5(nFu8e~ zadv`I(_?mTyTB4P%2etiOngRa@`~0I#T>l1=)4ef=F54HO7k8|Z+=k>OSK+Wz_@tF z-xh*qCP>n{6M%MN@BMXon=`b-{h4PrcY33@iHPvw8{qs(_kfXIMJiUD>Om?OFDVm>xIQt>46?P;v{ z>YFYCf8V-L4)*WhHnl@-;QR>vV;|P;dBW z*@tJ8cTo;?NfAnBZX2xxoY0POck=!8-DC7~%GQph)8 z$!3Y{=61OOd%JkSj>)0d1D<3Rmq9A(-fCd5A+@x*Mgc8w)z@1{1>0R#%KOmm{%Arr z&M`h{u-wyZOoUvg5LnB)c5j?2l5QY*U)3Mz#jEisGI+lhAx03j=!=T_X|JL2kXY}p< z`*p%L_RQGwUqK;6MeDsd#2ETYOF&^ZL=OnW>sBN)y=Td&H)@Qf-ep%Xv0AGOOfY!* z)f*`?#9qajhT@Jk-K1Vc>hbX`d&9kwpG3koAEj;ns1e`RPU=W46!1KQR>Y{ceiJ4^G*&SqtZ%6C%K{;B(z`mb}zESg3mTJpv>*~>v8`oh!<*t%z- zqGV0QuT6j4u^xCPQ{}l2u7@Y{3g)<^jHrB|d6_$ttQmzgZ0mtZy7bx8d+v&WiE%Jf zHn!!t3WG{C?icuf{|+Sj{g*Kh?!yQIPhaMQ-+(MOhUhs=(DKB?<>t_a z<1xymA@7}l2>3!LBW)tR)??qV;>11fJ{`-rabg|n)l1z!i~*)(t`-Od`#4Qik)*{2;Ad@1S=MS5m$~_? zOL?o4yH=Z?yWgQEoVJt{+hmWj7Yrs&ekO;|2!cG)JWRj`a)l0;O@YzLt5gCB3DWA1 z;p?{rpe=+FkN{OT9nK-mLj)T-V|f@C&xp;=ttm2~+u>T&5&Wt}iE0rg)sqMAVV+SREUh8r*p0aSDdUHc9)aT+9>71*DtQhA`M5Hh3OnkR3GwCV4$RQdn{QiJ zepeALS9zWT>@opjCc9jqwhz}D6P_4n5A`X+s0x#{Cmb&oE*C+{Byo>;)O)RA#=wcX zaS12Sx0Tcgt z!h8L|k*jm2I=!$i7ezd+hR8n*!n4#605DyfJ*~y~Y$w+pdTqMD4n0JjNf*Yx;Gc*L zvDMS)lw@4u{_s!u@Lo*zbBDc%y;Ic=9n|cRg>iKU=Sl2bIv)cw(l=4}FdMUh%x&)E zZSsys4`nV|l`b=l_3I!gXS1nM(lO9T!wI(X0nHrzlBqJh%|QWXor=Z-Wie8Z{gC*} z26saK$KjtpAy2ZHVZt!>J6IT_2cBs=fNJFaSw1FPO^D=hT~zDS-?%udBE4wLV2B7& zlR}^5G}JtugJe_~Dst0j?F#?gUEamzf-l%}t#d+&ImY0K3Yp1a6sHlqLhj2Jf$Uc^ zi1IXY`Dl%sB9YZRECGkga!l1!AznD-9Z({*7$cMIQ|p`xg*>C1HL~E86(uY=zJnT0 z;;N&914UeFsrV+(>FW+ov)vnW-g1@+l9YG<4mL`CY&rM~!Hzb8q!iA7-mksSD_X3crbuw+l6N~@2|8_A zYIH-jMzlvmsBanJc-X`04)#%4Y~1p7FRs5dUcPI%9R3vzrD^!D`N=NPE457I zwi@MSF38{k1faWE<5OUXC;Wn8tb=T3lYy=uYPOc}l1>|ftixhk33?P<)q}R~5*ZDW zpn9$MY%44)a75gw2XNV5fs_VN@hIZpDMkH_nVw~G=AZ+K5vea~&zj@BwJ9B%75>5x z#Q)eCZokP0ibb(jt3rr9*=^t(<+0n$1F4t8G}(2wJKoiyg%p{5@;H|iNIVKm#hYVP z2d5;(V_K;eC%9Xy@mMnP>@{PK6utkLnVu!LY*D@MR4)B?0RokC{5~B0o?nbcv=9y7 zMV0ie3B%DQWjt{oNRmV}>U@e7+BkD|`IV^7;Gv!#&VVsvhe8XuAk{L^;9?xE{#8%J z=}0lgisEHz7;!-*>K@`~62?h(>Rdwz^W@_OKz0$R?h4Lrog$k%6dM|AK|9DllGs42 zC(j|0b_pp#1Hah3z}H=~{YcdXs|&|s#s}X}X=cGDXzH|4>v?h8djItkg=QJCPXOKp zE-lH2vRW^teFbZ<6sZiuE!F&qMWJ1VBzCh z!FvbOUCFau|LL2aFA>Ql?>VpoH&kZn)_#5PL%i>6EFKnUEiLOP>o`}cT_jCu@L9am zz;Y+*n5x3r?9s;FH;<(`hvi<7lmybr95~cnK21Y|SKW%}RbAK{Tks%GD4kIgLeTY0 zjx?G$4O2rq{DVJq$}_FSWj4@Hd(5CoO9ZVWf+0$QLof@0udiqu(i(@I#p@V(Kv|2} zOhM6USNVs|Wv5s^$)f#NM9}a|IU|s1li3*C?686?-WE1UGOIAORWYiA7LatiiDefK zBlX$@^!ULe5_LoMUkd=OXy)5t^%@2d_y7`v=-SgmZ7+|CSkem|7_;odQ)L;&skEOQ zMusg}TE2fm&UiRa24?r!b8@>Xk_1@dQm*DJC->W*e?aC7d;%V47d? z1fxfbZPcx_EnkHdBQ%Oy{oVu>kJY(8{8$amd{l2GTmYaH}+kQwp>cqX2Dw zdc)S3AphDr7i8g+pUFe8Inf(XDyE5TyTyc<86|%@H8(BPosgT;gh!Uqkj+v@U|Y(V zx@INIoW6SeB&;k~R^1Ny$mE}|(18*1s*$pDnvUwMDVEl!UGrfiLgD7XkcvSio(wh+< zHhxYBTiq<2)i#YXgBy^fVLZ4iL>(wb+Hgen&C_ZJ2k+-Uoi~ zMzJB+xKU#wwP2i9~VUs}b>gh4J0*}ZI1Z?n)j7bdl*Yr@XX6hUy2O|1_h73DAJ!tQw1gCUKH;Y_}?hmB?TyRug zdXf3QftefTS>~dv^$hZ*adkuUusKWkd$oN+Z0am<)cNmSJ3NtT;TNjN*2UScZ9CLC zA&F6_b7BE*+N zg}&o9hHe?`Ks_meyDSk7ctV!}~HqRGOLfiuEtI8xNyNX(=ln`F6Y zkKxsmLTvnRc`2D7$dgeV#`%*qa&QX_jAR=yBwT@FQ#1Q)yIT_v*(LE7y7D~MOrfc4 zE5PuL6UUf#Yh4A%QKkNs*C+6zA_jPH6jm=AEg~Y&8O5ywVC~q-c7)USr`V&Ev+xi5 z;+gL}yn)ul3=#464bhB@roz=(Q&y#18mk}h*$jr$m`zX5+LyS9#UhI&)Jk;Dm2#E8 zVTD^)AqbjSt;Q*e->%#|=43FgH`Xt^EUFoIn%N-s>P{dw3PDU7a7@eEx}LZi;aj%V z5mr$n6<6?4?GMZd$@lRfTZ44+;x$P2YPzjXJ_E>^g~b@s4_rjw@+_)EwgJD(cNhrP zyXMAUrAUI_zuba63No6!Xuf=ut7G~2PL<6Iu$Rt<-syc9O(e@}*|j7OjG<`u)NUCs zY|RLH)?mJvksoOzK2WWF9?G`9^|P;%m=(#9fG!osir^ZzVL^SZ-MPlgyh^GWxyzmp zESLRFl${iToS)YR6g@wN_KWD%_DOgh@$ zYq}xvbdR>TTLkKoUOkrwzg6SAxKE#+?PKRbLEhRY=+#hRs1_=Ud73LKrKA}pO1L@( z3+1$+&&zdJMU@zFAciAyEe+4u5_+;ew478h;D;308*HmAmikg9@TR%EoQgJYbI#8r zMMR!=X?dgeNGt5Cgbvk*Xsk?dZdW6W-`cclKliatlFw%O5brvt!N-)#XjvgazfbZT zPuLH=Ev+Cot$~*_;n1w|!`K;`FY-EZGcDQl}Zz3f=k%noVWNmW@F6mXWOe5#L6uE7y zAbM{~vQSD$E7wH;P<&5`p=QZ=v(>G>*M_$P^%spu%_Ozjui$HCc%&@8AVQD#k&%uI zgvv!-kR3fTsrV3h!3GbUtcVCog<53xV}m5|tJ=8yio-63*J#;G7k!6WlqReCBO4Ic zumde3c>aPsK^z1FKXIIJ{*I z1{o}BS=3qKdh}|@5oy6r`~Ai$6+Ncb z>zKS7#49%po~Z}Bm7Exmxaxb(`fT-6Qj7#cP=7&D!!2dq7K9oU0<1=Hf}Ie@8O?<} zJBU}8faI4RjnzPkS{E%ZP@xV+F@XjAhm|~Mh_DoJz&ONSKAz7MUhWjIok#uvS$~Y? z^PjIE)`UNDO$b-xhYwe_X_NY+QjFLuQao4nJ0Qog;)Y1K%*=EUx!xzeVyhM%xi1^> zohLhaE*8eS+I?S#>+{>PK5jVrW8q9!3&M?mo!rS5T;hQ+UFvw>$q6&Smrg{VX0_!a z$rtIq);KMO1mrIZB{)HnxEBnVEETtC2k^@62|g<#7FBH9WwCDFR-=U=$6j^ zDJF|;+8F_0riE^zmaP9Njr+S=K<0ZCgJZJj;23*x*S|N5x9-ute30eE5^;mK%~^`6 znWd^QN^)f$Pvd=q z;VNPjDcgI4GC5k&HLajDYDOTN|PP9e! za4Ptt0SO-q%ZIERs=SZ0*WGUR;3~qrj|FfpRxTT^nUm$@&2I|=q!YKvVb4NJ3(dRQ z{!4(5{p6j)E9y~=E&SMp;Ln{Vaj=@JA>D#nD_F#QOwn8ERdW> z_tFL?!;RsxP`+pC6qF$gtsJ#L@ALNlwI`mmcWDXGYX%hVeAc$;cXkK38l%7b*CKp8 z`7#_&=gsDefSidYyJS__3=8p($X!ULsVgqCkMrl1Ov2xZ;d=Smu3WrydSfLNDsLkU zM3O(n{VpXx&VDrKHd zv4xarfNZqFSKcAo6YYIBNcZ9KK-4zRZ&_G|f^YPI77HZ;*?X!TH`bmwxxNAm1k^CX zMQBaUjfjQfO~REUOYPzUYTPpl4CI7%&AGy|@%)L_sm=F^#2F$kUpAaEA20K%sk?+j z-A4j0KP%_~S|TP(ll|)`XAkV#$PM!T3jTRQlqo`&0rORIo1q8LQ9jv|-OdJw3Mbt#m2_{Tv_jZKVX(wQjlt!Bb)AOiEWlcifi&A)NwF*e;DRjkY@;R9`bVyMk;t zUD}HdP6_uy){hR`OXq8Y$Y_CfH>Yxj#VL8AT%&#fz93jj$Gp) zT7+5@MbJ_mL3x`UIS=X6#81W#x@K*7#@NB-9GHX-(>d4TB}F^ijbze5<9D;pe30{n`*T5_ zqVR`iTW2x(s#~pX-L&KLDs$1tA(VX^DTs03yioW^YaGI4_pwR4CkG#t`H6G+Cg*e) zLK|2cwKvMjAB*)*ea63D1{;MX4Cp3QL0m#J9V2v0<;lTMf^C-JTMGaW;7OkD<7@{N zD}Z>F7gFCmul*$^wLN-hI!WdxF=t3NrPKPzE4E(Oqm2G(jGdxOMz0ku5bG+$k` z;M>Y2vww#;o)Jsg0&TO8y53PG2K0f06!Qkce1Lt=e0$n6so?FEz+FpEQwVGK0WDr8 zgIG=fx;Hz|*zKjzZ~HR#nnS_n{Y!;-rgvgF*9s#E^o=fdjwA@vZN~{LFN4NMx@DH1 z3I^lqRdE}P{&#sD+D4wXG~`qFoCK04)K!V?q{9tT_+zyVUmJW7_aLW3zL({E4_8km zNxwm|RudwjkW4*g4x8^fHm+#{3h4*UJ{iHC4&?+D^XT`S&DzUnQIuH~dW$e)}B}55?36 zSVQhFYqUo7>TBB5?vTH^c-JG#=;lLQd|7Me4xZ1Bas+%I0{*+kua0;=Lxl(gWXk(L zl(Mca|J|iDt-X=F*@o=5q9L0)oUy7s9&I;Hy`k3DQ{R!?tlLF)l+h@NG8B|S)&a!z zyF~hJ13(mU3fP^xlunD7Qt?0?f}K0%ufTn|+39O9HfV3a*?RsiUglVTi>ZBCaZ5)m zzm~fgCGIQu+QJ>}*|u!Zm^f1_Sr=M*@?YXvde?sH(p~!d;@|4tre*j3=V6wB-`00c zpS9d^dBZJ0Y4zB+3-Ei;-N&L@?#K4Y7ega~knjC|;6NgwPxHw2Bu@X!+4DxNLG9se z!lw8AditWT*YoY%xJ%9sX3P7__UWsJu>9iZ_3Z6vwt-&5=0d&fI<4w!r1ygr@#VY2 zTWm$4Z(U@IAM0~y{ZPJ>5B~LX{KSCo>t6Wn`gngk#(>_y|0O%W=iU5i{`D5|=k_V0 zfS}Z&`0v-MO9E|ARzXkJXYI441< zoleU?S{DE9_WRHd6%wwl>FRNPZHFNMe}~h@?d`rGgJEw(#$V?(Je$DxJ0#VjDsq3$ zp_au#O@Pn)-BnE&9$!IpgfMeT`nsjdc;wpmGt;;7P1-jqIKNIBo_fr$DQZF-Q$zTS zR?fa<5#w7EG5?SLXaaq0bt{8sVBOF)q?@Joa`xZHLN`zZJ3cW)-)1|jzBc|PpR|=% zj%Y2=bDgVK0#=(_xb-dH?g1`%g`CCO;X6njhwKt|@3e9m2DtpK5ZiqGMEg2t&2VOh zVvp2p-6@RQTV+hP5q0$y>q;MWn>drTH|c494V(aQ=4rH>j6$9i0Lt(Z?G=!U4mYkq zJ>IK`+jK$U+E1&?6ZK+-`gHNRoqK*UP8LC&FfU3C75 zADyUO^=Xn#g%yUwyDBCvyGUmwOx0lAkw*xW-ie~s%E0%V|HE~=7PKSn9+_m z4<~}lfH$9IyTlMyiciZXD_hMe^=0C6S`W2*;G%Imtw%P&48l?tU4WZLjpxqH*`TMV z;)!g@E1OhUH>XZv?PSo9oJ3Op@FZHsxB*@td)N|+*{0|tLYrO=%=PhO_Hr#5^w8k0a5`>hn)Z?R;*Dz>*Nf3@o1%@IE+^Mlb%roM*U6oKx$*kX1~u)4c+8n zOk6b12%UtEB0M?dF9oGepg2b-2bgu5t-ngUm_}xQ&gD#~<^XjFuaD_U<7R~EI%N2c z!YCtY({mZ_n3TxK$b4;x*u{}Dhc7Jv94DK*OQpM1-~mzl-D#a=j-HfvcsVy!S{Wu* zi2J~14U#{}9IEI&F(t3CZ_O4QluR;#f;Uf3{Xu4b1AaSCLm69+JmYYSf~ACS^E)yR zXRTLz8#?SP5QJ9tv2o7~%V!awEFzf4nr!yk6!opDgP=x!&)Ox;lSUnMx`_j!3i z@3Oyu&Ad(?>ee*x`Bbx!Y*#yX1^-fN1@6fS!~$ZR-~xxr^2|Pizo9=85g*jJ{V1O& z1{0N;O}1OjzhE1V+iXF)bKcJNvU{M?(Xn}M+r0NZHsLk#gqsr>3-n8)In!b*y#PH9 zC(Z|^FiTn&C!}JZD3X0Z>}XO4Qr+vN^o9S828ZMfQb}P*1<`_5SXzYsgQdo`8#vs{ z3J+Bca}iqWhnYVtudjLkUVCuGhusr6#kG4bZULOqY%~Zs>zU^C8ys50qO$zTk?pVb z(yTE2g~%=q4a><8;sf?}e~cZa)n}ora|zo&`B+R8g5Ssm^c}41GEHqJ8L{Wq48%zZ z)@u}|OilBLh^(P(d%!r2IAgt6al~5rd_KPIGZVOYzOuTNgfi9K6@AsP#WYGHtxXA! zcfzm6u)%tyGB4Q-v@$$!8B_>fxZj+<5xj%#4*GVj6+~GW4(NGdozuOVO-Ml<3Uc@i zeokNlj39&6P|#4#N*K`Ni+-{h-A1DeBrQmgDj1OdhRzYnYLQh5Bg!svm$)3t7jx^^ zuZUsDXXh3z>ELEum8H}Y%B*#!h#Xod+~&9o$p~CxH%Vwsl@~cC>NPgdjUrtKp%!$l zxOap21e&P@1He*)a`H@X&mKkI7^7lg+N*+B#XW!3Cl*=~7V;*958gK^U`6|e3*=k3@WnJ%2AwjFIV&HD zU0_DD0$yg+B3%G?V_SiGvHoQ(65jUafCOdBG@=yzL%IO3x~UCUpa(og}?$1>4ocxNLidPQ9z&f*xdEg`$Z)8w?9J3Mw2D%ksdrn)%2? zud;DY!FCxYzn+kYbHFeq=b<-?Re zuo)!XJh8C`xURo5;{k8Kglew!`aR1HLK2`C(jf^jh2=$!xJk7ch&yFYVdr9}io_KE zJK+GxDBC94=G8=Uly8KDLG>R4FiMvzdRo$S3Sior7LWcGH@3jvErwPg7T5yn4vYOQ z$v$wK9&8^jNrYoY0L|GOeirk^@pk`PtI=v&GO|2OtCn^PMRdP|rA32o5fNh0FSi2{ z@hq<&<#tOjTno9D=m>115fbgz(gG%&%q4po*GU+Y8i?nWrC^^e96YarP+s&Cs$WUnq7PSPuMKVFtBVZ_DP>K-s>`Otf zlT1QE4t|gPC@tS|R;|6b`yb)Y07u0~<-fBk;j zvO$R+ImqiQeRYhG$6ut3XTQQwgRU3CyEsw$lQZ&rnzvFqlfTAIM+g`s2H9fW%c=~x zVXO5#k=I)*G7~-~Xq6R>Zcv9OI0ocz7RFwl47yCrDQ>^1vjKx*k!IofsM&ZxSqYI`7pZRIzzx3EIMTel z(wMgKu6PZ4f1Wvl(J=17QpZ*4jdPw-PhGI^G)bMSFfLo2tC$uRo2QaDvN$yQ0fy~M zjB8&tjku0_?EA8ldI-N7+QV|I934Vh+-flNnmz-;R1f&Q^JX)GPub$M&^0g&LQCwn zk)P!(;k{YK2e(WCr}G5s+wMwttrvWv&lXU=$YnS2N9G{ySEX$k1??;CO&a`i=wZG? zcs^;@Dl~ej2GY84k+BRO#ruH=Q5Va!sYvX4ySqN7De6b<4~zXc-?|Aci`@)Cq=j`N zyzcSj80RdT3HhFEb`!p$$~gbspcO^-qb2ITSF;(7PfKz_bUt01l>*cB0ovJ?LIwZV z`Pso}1fYtgkHlJRu4&~+q6Lv(n>9p0i(D2C9$+<=uydoA&--TVpRv;S?F?NZ~kEAjB4sCF<6UZ1+|&!9A^&6AzKl!*1~X{iQ2AfZPV@EL}2@q@ahllrK@ueINob= z)*txbp?Uxls4PqFTeJXB?owmyXDW9CxJeBD0iQb)smL+6P$)Nky?Z-N4AYEFqavBR z4}GdWqWUpTg)~T^=U5?F)+A6@O_0ItN&f(n?un`t@I)?af|P1Og1bq!>uE=XnFS)( zKBP$izYJ`a3EkhaZ6-{-kmM1Y@Y)W=nrYZ*AVwQA9YcnRJf9;a%$3F3r&SeE_k3EU~-XsTvYb z-wW*XaQVn;%b0Mv2n7l)^!szm1ipftVmnr4e7&(lLSMh9G(Km4lgfoDN)B&sYhV_x zT5hVLWN)e_UV3$Mb(Ch*h_J1g5Lwq~9E=be@_bSB-Cgew$bbK4A7%DOOtchV1x_5c|#YTu~J-JLjZSoOaY=@ z{v}_V0GFxYYZ*;qFn{mLok#TH`lSP_Tf-GXD#^BkE#~ecl2xt?|Ihj3MK(MUPA!F*N=VC1Du6sBd(1iX(!&_Ts0f12(Si>>mI%etZ>tM+FOSe{u zj*O&`$;^GWYRRy9sv!UVL{B|^BjO{&S+9`cbTrW%M#8Q=Yi7o^#b%>tk{sB|Pk=JT z;bP_H%^khA5=-4$SCMJmi%~j;y&Qen^AV{sPYRN_=~4pm2wt*UkNJ*2m2nmSI;-DQ z6q-$P7oFIWV z%jd!nGc7Z9?HWMXpa|m!!>Y8AeIHkP0Ln)OW-F_3Xh|hWq)DCMGX)h;VXTG4KpKoUI&P|8E~i znEHjo20PNvb)WDU0dU*$c=KpPA->_H{QlpBDP|?*X2K`ru&Sn;JOfm;$!FL<$pilUp`*05DE3-LfPl{z``|tiWLF} z57_y(YKnh7;|!=!vF1V!SuRdgq&V^>{2h?4CITJmnETpB7O(N(^Y`h~j4$4GyR~@u z7sQp&hTnJEkB_Rbn?rQ)=2a*o$%PtJ4b~E&3@7y>GQp~;%~*FeTSeRQws4bm=?|1{ zPfNEc#x`9ir)Zk4DvE#1T^AcxQ!NRf5r*nwVy4XJaft$V7eHrf4C;7ki)46%qMVMe z*I(lFIP*?FsXdaB%YeSaY~iX!p4p4=t3T31FeGr1^Kgv%Op+GyVk~~xs(o+peAP)s z;?1vtqbE=HVEl7$-<{xZ%1Ey20MQw<2d`rE72Vu&S5Q;eU{UH|P_GLW91ZTdqbM)D zMAV1L!Hg`-g(fH6a(eo5z9R7UZ%O_{$53Q09Bd!_B~3sGZHeQs22Hd>CNr8uVTVk> z^niQiSR6b3K^+f=e+#%%&;dlMNJ4rlwFN|o3Sv<2Y~9l^FJu8v!+ogkKS_|FQH@SY zllxaAkqLDZ-<6t%Azeo@JxfaPaO{d9v<_v(dw^v6*6pI|L|$qeNqLjAAL~IDDJ(Oic{3v};#8K* zpln0L0)zv7_c6adCt-T+nCZ+??{xWDNbk3td7SH*Izmvg!}T$;gRwKXyKstD?^QdLC_5fzq$d!#Msm?jdSeGkugX zJ=Avc$oDHYrr#@o<4P;Bpq<}rqZe8@!=45p(|2_m9_=mH*O}D{9=8O$CdW8?&7TbOSOuebZa=8SMW~3dJI#p4=sDt!3cfiAdzS4tIOt zKWRvO0+1VM@48C-+e6K?_S4Y)caV49+xYN!1Dvg313R@HZ^9%b1}S++J8Mkrhm<4_ zGz24Q*a@agim!u{JmhP2^XyZENVeF$gJl@41++{Y7DC9~H7CyZv&N^E_@z%RHjQcc5aA4yfv1S7Z;IzQKsv_scp9Ix07 zk*-TL8mBQ1O<-w#KEnN6Vt>!h-Ubl+Ml~=M&+0fpr6mvV+>!qAa1q~0-WZl)5Uy|O z>~GaXj0Jth75A%r`2ZIb=Zv;s&B{>Reb9I%?oJsmSfr< z2(cj+>yKB@tvGHjs(&(0dO;bKUrDRqsh#SZl+DZ7fM=AJFc8Aethh{1Rq8YSy@hTp z^6hXtPX7A*LF;+n#31~%mXQC@iVRC2FU<`x4T9>n11J2l4$@%nwRJ6@;6-H+JNb>z z5Ac7(w_n$vzyCZXfzG%82k~&XvirZt9p@*Pa@dh{)Egj{{XCbfCEw8IQ<{~uhNu0d zl~PAwm!{1OD2p916m1MK5KBZ(a$DW`6W9$3md|6ufhhIrlVCsu1cCs0!0}HPZK3e( z;LZ0k-PpmOAUTJ`DoE+RMm$>4X8b&z5sz^Lx`N>7$|Dv0`GZi;0a!ekV?2uC*g(6jQ*llEG zuI`8ezViY3vNBIsE0z8$Cr6g;BIRq9AYr;6=iet42wKN*d zD=Zdycgb=Tdw5yHtPHxjz156%dsCfC>*1{W+W$6U_4`M0)(*4Q{g-RU=+jCz z3@1sMU_I^19!3SW)9`p5?$^-)_lj~=dkgU%&-Ox@>AnhUfYleDB5FmUbHoUTpH>+C z75*;+A^Y?}A2z=i%dP!jTNdg0i$>K1qFqs;?>g>b=Imsu$TAjk86lRJjT+|m)JwYq z3IRh)(F|E6{`61JoK{L7mGfP<@0gPAk~k87OAG;9o*%E=Gg}yH_g9~d(H?0jwBdh~ zO8+cdyW;V~U&Ms?+Gug&iT6JD6ON0w^{wsW(j(Du~T!N#O!@gz0^wOyhqqq@5 zEFx|@jf)5GN=M^m@#CC~10{+i#Mq`-M==^a`3tgmJPwMKHTfj{H4fwXBMi^}p#~;sJ}ta|6C`a3ndkr=P%p>NcJ%rF@uZvu zZ>$i^U&r1U0=-&NgSi3S{@_2V8aa1@GC3bAydmdp8S2nHL3CIdV&!Oe*DcD$6Tagz z^k-UqWkSvOv!_0;t}(BV0(fE0nUwkw;J9L*)vIwa1FohI@N<@iq!IB4P%**5Ro0H_ zl>e-Ds0K3Ts82jEVKoy5C+s`6hc6f1uRxeS_MaW&zR*$+QU9{lpm~YMTV7? z>kQ(;$7ye!V#;mtM$We@yqIrhHXKYy?fVt^mk40gzQ=33~$AhYuKwVIZt?W*Y+X~3Aiyq~NnL}g{( z%(^~z>JAkZH%Icf?|DsS$IY45hP1$RmZrilG>aM1tBzEUlB_DPg=+U1WW^F~fnp;R zdh*4Ri?q{VER&Y?()~fo63Ub}07jTW4-INcOWAfJgWtn$9NuE$L z!2J^~;FTfpsP`%=Y%vh7pa1cl3jpm0+NEm*YQTh_fGHq>0NV|~c`CUM1>Q{dP*odM zJE0dE8z|AgE%JoAmagD(A19kI%x{Ybtp2TMq2__w+~ z8t|R}afdAFRC3wP?*tlisrd1`Fzls|spCjx~6E2#4(rt5L zKW)o$L{$;GrbXwf#x-M$a}4|DObxM4e@$L$NJk*J;}Zj0RJF8e_~*~qX%T2EQae+r zb=bca(|1+ED(JgglExdOH@hEz@LN6(V1<}cXuR}m6J#a9x#&-3+;%IZN-Q@lHkTyG za%w^s=@6Yv5Nw@HnNGH?X!H_fdKnXUyLNK|>s7y*3il4D!(GT9#Nusu8_DqMna)-@ zW{$@r$EP5@tDDnN?m0~Qu;}4bTsWo=j#&lgV#zY(6or-ph4aCp#WcF2XdSXFYu}Dnjk0e+1Uqcw`j z#jBgh_hr>QOi1ibBQqM#j=IazgJu0r8GmJMqZWCS)+mnq6vxxQ61Cn~dHmNR$d%D8 zXq*XvT~Ho>N{^Rz3=h`;5RC@abfZ-M^s?sJP^F9}EMaC>S|2i(!y!Z!a&90F*1S*K zf;cJl(}c*1CLzKLT?OBtI8rC&plXgb<>O)7#Btco>PaQB97W!|Vt}g6QYoHd@fi)< z69(Za*Q5vV;8fK}*>)3olHx$~@E)AZM44hj4F$KZ|N z;z0mo?Xjstor2vqV_4!jM1kO*owFqi#345%*nd4Lbr1MpTYGp>=|Tn#j<3c%9wKld zg>Q|28~XC5=N|oB@n(lEkTP}(fdebRE+MSH?d@&=4q6aJqHo*U3bhrb-VH?D6(rFfF4L)fe&G6W-6OXDtyr(uXwh>LH`zEt>L`kL8O@x7ZD@ zx+QvGU3n0VjIr&?xfJS${owZx`?bT<4VPJ-2_Qk;|KoFrNZ4~UOSls%%kbw$0X%`* zBi7IR@>txl^uf}zJDi+Y(B63m4J z=fh)TtiI3$h!LPV2>|~vi}Mnc=xV~qN$Bsvxy8tvmR9l0PG-=7__0z`ehYI?O>oB@ zC#4h7wbSq>m4N3VzhC|MDM$WG3(PnGXQ_qGy=iokXZuXEEa{WaNB`4vmW~x8K?

zsAw8r%!x-nc$PQ8-z(h4S+Y_S`gZ<Ks4Zx3M*C>oBJNZuf_aU@a40U(N`e1kiS2H#6{I<1O#q@w@=M6 zy`-(Uv_-nyGl)iG?%+W7eK`V8m1t=FTiDJ$cv4`}Utf#RJTlD*FSZ@IVB%5grMSQj zfOL+;8JXguh6 z{A_tV$o};5luk5`Kqg5}R0ibd#}xx#Y%8?=4=R@$`hB1JfVL>0-gGE{0}9ZwBTQ;R zF0F~M6+6_1y~t*AK`(6!a@!Gw;{;}%W;9Xbrc7vZfh~>6u;E;Id+)q+jz+M=(@L6Y zd?BMMBins6+hcU=A+BArOQ3;1t_-j1?qDnT$8H$&g9<|4!Prsl9DiUmT&3=0^cVTG zdvspP+2c3@I#sm=Vj=awn=Cv`B)r&c)BA>!QFiCWcaOa$$2w?Da39!6L9{3NG)17} zYd1$rlk~2i=~=>n@z7l6c7vs4xb+&vFw5S-&p~4|wg{2PMR~B@@scW|(VQnGrMATB z?kW}|hQMZA?mff)XvlCvF!-h&p{nqoer*>#Q5r`;9+B_TcP)^XIM(!efSEY>XfS0F z;5v~Y>k`Or(QlrZat&Ee&yJxy^d4sH%-e7!;b{KLMb2~oQ-(~2>wiESXHgIl(b!Ls zroB`^0UV-qY$ojz-0yrv$d_W?q&hDN16NJ<8P84foMz7YxKr0}Kkh{10a~wS6_j+Z z@j&S^t?jjra1`#s=DNWJNOm>Fw|@bs9OJjB{T;VzR-3!v{v4>KaN9!&AG=BB`*JYM zP%6*vtqCm85kQ290F{GXTA*+jaFVz=` z8RB*|km;F+$Rw`NvCz-sNO<4c?aOArig?`Z&9q*F>=b>j9*v{C!o1rZ%-ouH6}TQW zx8%A^Rqn`2>lNP(!UGX0cHbRsbqD(+-5D5l4yL$bdI#)`S}fd!VUgMuNH=7kRwpT2 z6+Txg7cz|DPXCzbj;sU=7&T&-NSEs!|5!RKKBG0a>t%8&Q+3Rgk^8iHteRiTFw5M4 zC5EL#>#7TbC9O+9Bi030wl$Osl|3KLyk2E*6Eg5g6|&H<0eBU-^q|-sMlQqpc+uXrf;)p4!Eps@Z{LA*8xB*x}L8)xmo2y`u!|0u8@s{0J-f>nSJr zf(`O;b=!QZIKaC;tZsLCkqeK!ZQva8+^mx#Q~O}yQe!UJykwCc!db?Km!6a&W&w5K zNU~PDZVcG++zhxm{|4l5ce&~yI+@;*C!~SAliYL??g59^Az_IrSL#hW+rJmonce!|388F;&uO|(CC zlSK2MweC?7OfOKZAR}GXLX#;Dir-il&jx)PoEO&!P}JWQLB$G#LVwd16h!z=VW^1~ zs!gK{455|N*UqL-T}(el8%<_u>MoKL_c6V>iqm<}#3FSu&NC|Y3}HpkF95{X+cuumE-w>@YUUF`=R9bd-pF3o(dW47_uJ9b5cdoipmt;y z3^V?Al$D_B)9%)2D6$e4t|JdphdOiHTgXpl|D@j)RMwq&Xoc$(V8N6)EDm2zvN9vh zLvJO4i8mn6aTq|bQ?W4f+FgeJ0ciwgc?TvGw&kAYA+cR(c2S@`ZoZ`7KhZ ztuGG1-cD_&Ur~1aC!VdzCwI#>$3mJy$wl1&W_NW#;jEStQlmriNLYbyA&rqKDCKB; zLE=n7fZD=NzAC{hGp~fXEHY1Wph817<+ZMkHY$ErwZyHIHVP>$GN#PGntJ87g0NU3 zU9}X*lwPI_F+5^sGO0{dBmE5rnN(aSz+^2MrAAiRmK3EO1cjjtZBwVb4G;Y6;zWfK z`X;npICdG!K)Y12Kv$Et8@y&x`_Eh_$DRPxSzMbuGE(eaCN@Pov)9TX8@`dQOtOD< z7M$Uq1w+X*g?W5>*r^|G6SKruYqjciDFlT%!5{vu1=MF(yUtUTzH}(QRe-rWBH6CZ zChV_(o8@fP7;Cg~u46R;%G{Ub*7*{gw1YxQHREq6X|saGm6&IOmpB;Gcv+I-LP`i~ z7AC~B;@AOid+2|HS031ZykesQ+pNSpj+Y%X;-yjMskXMB+JXOMb}pfGDH=eXo;g+I zWdITxBOW*PLMRYs)}qL$S7lyGjIEW|CwL=S|HU6 zJ9#M1s2fH6B=pn_N2dkU3pLr%*I3_Zwq^$mW~%>y;MaP*!D1^5EfxSrW`hhuWKH)#QzT73d^8358~v6%B~bB08k@kCwo!xFYhN-JBv#++l@KZ(Fdw{5 z&N9Z)&wR*csv>kQ1RFs&t!*F!lYnE6yOeU@S4@_RG|3lxGZiR3nZPF-jhJ|9qI{{} z+9S-Kll4y(jK}mm!Cugghh%S9UsBQY{tj`TDCk0%rcOnMdJ)ZH1lT!nKwfX&e z+Y>l|Ec#PtlUq@JevSG>;UGEDIwm=$z%bK&X3C*a6^p4sG6CD%NQh|?y*N{)aBO}C zGNL5DFx@Kq`mXK)M>jZ=16`50Mx9b(&UTM;mJBu~C95=am1CN>1bMP6gCgqK&z6r& zzr(%Hf69qPo+-pG-@6ORJ7k-01D<(%0ElNBS&{G@_yVy&v`Jh&b#&BoYR@wVb^rTi zW;pH&7lCa+nM$w_p~ztL3E5MS_;5|RtKRn|`29afWW2kOS-yD$#>zDeEcE9<%dRk-OW6D=$9`TqGi!i7y&uMwWpUcwXq+(WL#uV~6?UrVmNP?~a za{IOEWnn9%!DvI)KE|f@pJC+Uo(J?BSGBNl#u>%n~K@w z0|=n>OM^{iwXkP@i>a&x$8DNuwIxcG4D*T)_WMX6P74csPpO$Pv~7!Jxq^O3ak95D zdn9)`>z5=fz7^_G|)rJoBKtc7yIs&;el9oc{8( z#jf{F$`N@ZF)GiaVAv1aP-LNV9Q%Vm1Wcd)Tb-w1k78+#p^R{rE|#0?@ZZsNXFWk1 zr|a(Fh&g;2qfN9b`pYcGLR&)v0ZBqR8phk5rWtIBIdc1fCysF?Fpgi8i^fcvMBsM8 zBpBBJ?h+?gk1`9+unX#)Cu_EQ@|#{N4Gr1=ziIEQ`)TBS*RF^~A`{GYY{E#l2kX{G zG-@*%;Vp=i$R^AbVCx#20>?i-2FD!t?u>r9Ui2qmCTC*;yi73gD=(QhcBLHt1eC;$ zfVlwWh&!oE2WqHs7hQS1`#zvuGbu9~;4*YAMs=NnyAFt#zrP~X3>YLwC#&H)hUs&C zo`4<6P09A6@T0DXUtmY$?9PI^YjB0fMjT-_MQZlq%J2155eYPtTIL9|~unA&Z!RCv+3RFPU#2&9a^xtbI?1`AOsD{!!y3GsDaKhLNDC2Sx$p4r$(s=~EH@ zrqm}2kLbnrEA^|RoD=SlFpw0}q-*viHj^?g0t3)Svj)x)1)_?su?wL6RY*-{m{Rz? zm4pz#S^tkQw(xhU*z&( zwP|WdvPBY-O7lOYi!W}BAKW1ff8X6lS)8y1rz$x9z0ggisB8o5*h_TtEwntKVdWP* zbevvtWBg;%37qZmPf}CzU{Kc>Bmb2q60bfuTg~soS zk{lDfLFXR>OxBb@>mvS@Z!W201Nrd`hc#pZ03>yGdvkIV(T8t&^V|C4Nwx_L*OUW1 zhnwr>txQUJHx1?(Bt&o?BqiqLOnKz>G~8$_=Dd9OK}+NQ=x^+Dur}RdkiJu*0bgw@ z^&5D)ZQvSg@B3aGN1j<9)m}Gq)_3%+UOk8HX;Q;u@k(6h)dEE9mk7tNNnnwtpO?qf z+%v|R)FwR=6vV>$ppxr-pWWK*{E9bd?Ud%3bnA__m+VC-+$tkwnlomHX}}U(W1Iy8 zvT&f61-AI8OA{YRyu6}v+O*bI1$;ELoFnf>G?(zDC`~#kbvkrdCF4LW5@t7c93gea z6->WQk*l&Tb7A9+`QudRsBJ#p=`%GE`Z|2omCStK`DT(&hJSR#8;LsWh6$8U@d7im z3i#XR;XHBAnMLyZ4mZO<5|kX3G&Y_O6e)3Q6j@a0A+woEK$KDmbRu`ax^Jn#42CJS zmU29eua$&dg0EDnN(vg{)^}p8T|T$Cv**Vr{qszCU59S)soSJoBGTZgq5-ViFB8XP z-IbDJH??h*&YjA0hQH`Nrxikt9ItIGNpj{4-nw*rb!yWhZHY6AJBue}+$c5ct^2M* z2fh|5_wLF)p$GEhztM;vSu*SNCD`<;BU=pBIkj$=?N+AAYZ~YIB0I?zpRzrA<(I?E9?HmzpwT%#-5+AN(K&I-Es7i zmS|4AByb`ofw^y;!$=b?jcBy~NjAQ202Z2v9ikc*r*)mJX%&f5OSZItjxYLmpNe)s znPoVWr4VlY^)7)ca(IcvwrXX z>&^`OzM1@6SVDD_iXlM)S18OGyIjB(j_QL&q1P9aY_44|=bYgDSga}cO*KT=!6??z!dPzO z5d&y^MU;a%N=u6^b45HfL0~X(aPJP#_=+a%zH5K8(DbzIG*veu{mbs0{#YW_H_}XD ztuqI6v9fgmz%38)4Wz_T z0yz(2yC*4x^YFv`z*NuGwCWW3qvajLgXNgs&cMaZ(kqN4{CjEdyEacY0pyRVHyje{ zQl#2OpS>kW9Wt(7Ck;ZY5i=xXJ!SMI0|KGqo|`~P=nl8$Db#sbeKtmj87(~}Q{r!h zI`H_tg@$kJ53RejbMQDH6Io}z^UMjEY}o7t~r~$UJ9J${>H5V_Piw zE7Qb~Nc0mNv6jY@tp__|rjTgb;#$ixH*rtTKYy|Kd&}Fa~m1NtfBiX z3iiz%SaxI(pEH5#I9^9@KM0#dn{GhU(wScQko_<)Qa>-p>#o~NpyQo!yEtA)Y}*MT zZ;ImO!0O*WPr}ejsOoF;!w^rv&}!uKaj*M|AH>cV1%^Wu^1g@B?EoY01qa&r;4vaPHM?~XHo{s7twu){OEio-n`L`2x3YUkv(pZsh?c+K? z8{u40Kn6tCMYjdjr@4K7FyWm1o)@|(Nu>L$RZ3`zACb-K;^lrqJa4Nz8veRR4+&_b z1S#U8()2_3AxK-E|KTh_0EFMr^z8s*vFFFN@D=$&c!hs~RGxEX#r-@LZk%P2=7^UWjnR$Uw3r`6n<&|RF zfMs(BfBkGoDDjG(_YkQzCf?8jTFX+JCt=g*7LhxSyj){@Ddj@97F=7u%ZiDiX zM~R73^o~*pOU0FYpQjMI+jO87Q2vfc74MAi`^NBZwX5ki{H#~$bV6Z+uwh4O-);fA~N3MFdgfBky-S=ph2 z-(d~W&QdTXK>!G{o}f*%vl^V`i4lua^z*{o zk7f^kaucdn(k;S~#Frp?H-Iu7(dL3Vd4Oo)!k9!ySOXP&14f_%^SQq`mGx{=MhvEsClXL%fGn4H5xEmM{dztRDWX(M6QmKNh7!c zIseTZXpP$CuM=k0m%ERg#jo9VnN;nRye(r!Nm(2Ci7Mx&bZbBmesDx$& zk@C9KjpcasZPdool-fpLYDu5(U3DtMQ(`XIJ^S_4}H@N@ZwuaJjfNppj`**)u zjZA+9dXrXOcP|lA4MS6{E$&tD9wKVNEfcUXINcQ$NKCFF5u+=euZ-A|wb7i*uZF0g zA356rP(0c($skn5v!U;TcQWF~?}EoJX`>dh=|&^G`D>~QlehXcb)iwxiqvL-A3Z-t z4KZVz9F_a&JL6GSo80){V|4T{3WU{e%v0#6KnAbC5{s_fp_S1#(=qco9yr4TUvR~b zETNJ?813`TV2ep*k;EQ881Q9@hP$@Ui=2g3)It?Q&uUXF8E+fe766?Af!H}hF*(QD zW=+cF69vNoDB=*G{tnfKNPf#FEIR=tzTwf(=@iEhiT#)8C?-Xdm}NBPbj&pJf14Vy znNBFJzZmqaQ(I=_tbc6J&i8ycgo!7hyffoMktIGKa6( zvp3vB7A(~z$J9Ww{bjFE590VUWD(#E2U#Y30@y?RDx>+iY}s+WO&ad zciD_PwQOI>C&)gsSd@{O=KP`DYapkezpSFy7%lBYcbVpTuu`_BghiNVot>n`*l+T` zMn*j|g3^GLZ*Yl@Bd*Ws($WARpf~=^UBTrz- zGa%k_$kK}d_K$OZFy(KYi&Tzl_8syJ%i;*X_c3dIm&m19GC|=az}y;ourb>ML@5J5 z1YW7^TTcGV*e&OZu zG+7H2X509z{PF2a`hbt`S<9TANI<6fse)lhx_t@z`*_Vj&1b!0RGR!(qSv+WhC>gd z3~%mdCU&rCA>}8nm3Zmd+B`wtj`B!Lip8l5edf2F%eOWdd7(8(g3IhZYalNwn)_&k_^y<`My0t+18n?|EWCjtj;gR~1 z^5*`+)fG6xC~oXqO5QBG+Jo$_`!6)oj24mgJ{p;w&GsPjG+717`cN9*;Y@$~iT-5! z2zST+SiSjc91s>i{utfX+V2v04&g$Q$Lvuo$gCF%yD%E-ww5yNw&Cbo z7jm8mHKammR6LsQev`lZ0FOTFpriV3lF_$ft^Q-RtXyp8J~AmpXZ)?q+ zgKZCrMg~!4E3H;Y){X3hzoyl{EhWGhQ_C9B>i}U6)+g`VH2~m|uF@+G%E}zhfC~uE zw*YT`-^WgeDuW8%?))p(g2^qQ;GJTO6reYtcQ~MTo?0sk0}FY=IYD`Yma)T&X48c5 z(4cb^SqsNrTMT_@$d)|w_c2qNL6sjcE`U@d%Io(D)K87BJ}?3%?IencBgh&0%jjkg z$vd!JZ@)MH)G}c%5aekJ&*XZ{3iI5)<1Oy;*7o@nC7+>Y6tDQ9r8-Id>Z;IR>!Bo2 z_HI8Sl4gR)|t577`k{QEajpW2A9$WcThWFo?BWH?AW_sU+sM#v6U zp;uUhogaFlPp~lhv{wa0UB`*z2^!2`Lu4g@xFzyt(lK}(bo|+S-1RPiP5)Dbc&^bcIiO=lh8!*)|csjr=-*U1=EGqS&lC>pDT-73%vx?*5Zr z{DKFu8KD{o>rva+o=#pBAslL9+iN`KYrMC^*9OOl59AzRpEy%DpCfl~<)_ba=^f=i zUaHmr)#ZS#_Q1Bvu)Rsxz7^&G)tdlX%D}cqV7uj5zSOJ$s!IS`ZGf%iV7uc_pXKLo zNfi&u$|uz5=M_|oD>-Se>WEs-HRYE&Z~WRz0qyo}?dA4PNA^z2N9XSP`}q*w$Er`0 zHO$BjM6C>3PWozTYH_Ep<+b2Voi}e_c7M`DU%~8@ju3fJqF1^5QzCfPPY}Dcw5Yl} z@n3OHz16lAQ!@VYvGK4+GjR8J!QiaB1=c``f=zq`!3V)1#WYGahVOp#!adx3AI*1{ z&v)Omt2jj`!3XJg;;E;q#kSrB^!Gx>+_E!>jS_p}vpV6k{+b)JKKllNABqT*?`ew6lz%^PlHiz0%GAl))1`bqL;0oOS-evY#(7r7^J#o(xB^7tj5>) zAtfkGpt_9%hXxv~L9nl$kAV}0CU;J~tw<8mNuze6pV5Frz_RU?@SMW|SNcpLUSM<* zpq^H1B610-wg9t!5&&@|1#`{_WT=a@4Ajsi(m!_5|B>40H>k7}!0@eT-rpz)kASD$ z?h>vMjrZ3+=uKSpwQ5-(=$m=3dGrSql?FqZug(!ee1w*bN7DY2=JZwlbt+E>XRV;1#N;tqBuwhxpprN^ z_zVyn9u(6cf}n;$hzb z<1Nuf>ra=uR+l!KOHYie&)ejoT?WviiF6pZpJW^iT{2M<2-I+aB&+2f1Gfc@I#wW_ z0|lt<;7Q-%NZTT_x1zF+M%7VSC*&KbjGtu;7IoI&iSi0H+Y}?PWJqC9-JUqx= zBW+z^^koLc=TRQhfSAjF1JpU0orRcOe!;+YXf9Xa-boWb#3;)^G>t}SWAD7mII&(G zxH}z~KyPW}YGm&~uW#sRW@G*T$xC#$ zZdX_q4jZj^9X_D4Rx!V!)8brLlk8#K-haC_nrH0eGbEzHi3lP82-gCbO1iXuTxoN? zBION^X_e@fZ>9mXs$J4Kqj#t^YgEy~d!Ly%_lE8AN2a9S0R;Ka28Kiy3IKn8r|l#M zx6U$mQ5tRMWpgh!a3>0F@1FDt>-9bhq*r64O$}wcWU_Xt((8O$oh(JD`aF?Q%TCt0 zSV7){h(u(MFh0ZW>~wduV}gN6O9^D!cAntE%1(YnJRB@c?3EQ0^aI7Ef@{+1*+$SP zj0cX~{^I$~4PmupdwB(zxR$k@0-`dDG$g)sMkf4Q22O`?%%+IgMckttH>eAing!21 z9JRhvc=YWHa$k9_~wulh$AE&%6_qv z0Xe3G^85VoKpA;?+19i3S!*fw2-n%0Xn@jv)54J zoL+uOT!;)>g*FQe08(H@+J!2unA^Kxr63D6tQM~+PsxY3Bai}=7LP=Q0$zp z{zVoLll#r!7?Ten3e^{}3%(;}gjJB1pSibtc>a95FA!rMBtOuDHux5)uy^BDfYoMo z9P`$!s~jX_gPXqt-wu8-NZ&|@;cq-dH)^Nq=lp@X3V6yKCcY136SGG?axhnN_bL<3 z^OTpd>D2x;7fEy7?p;0Pr;hHE++&R+>^kzv__{c#bJYb}9W*;G!}3ii}Xky$}5KYyQdoa=}cTi$~lJQpCS z>lrLO{1_Qa=1-EAEmVwUXt|8!Ad=%4P{l%mWL`#AFIq_HMkJsZug46{l4%|oI>mIQ z0+XecOVU`ANNt~Y=nI2(;mbN%tXS^}t-*Y^GMMy{b0R6v4&7v#F5P?Fh$8N9AehoS z*^IoJDM||r72GRUxIsxABc_MmMp*z90wf9r_lgYGc9)}3Ea{gJy3vQBItVx+(+8nt zx<`6S#5y`od!TNvrQZnjWAG*?0g7eRl6awPvU)`0d6lTpMHU;~56-e64Yxa!;A-p7 zpFCMMFlhswtj{i-E6OysucnhP>6YzV8m2aum7heKJFQn7_sJ)|=y&wD(pjo$KGekK zQ6Ix#dJLq~S1qlekjZw60zwX9az!=`e)y^_u(%gb|Ne~FGAkS0*`ajWPgBm3TaM!% zxt(vA0R9cJfCA_P^;ymkrYaNqRv9UqNkqMEFr;q+(2E4?3(x7U zi~xQ8{^|~g6f$~(@?uo#rv&Cmp#lz@4)=90HUexhbyPmH4oXGBFtQONS_MdRn)?Ui zF@|tKB5oVG1AX>Hur54{8uw(`L(1^r!_m;US1kFc-2FB=FEAiPD>Y6CHwHNKpG@b` zMFGrgCiw{v{pm6V|hwn8q~oHb~+nkLxfg8Qz3 zm%ky+?b4y{?k@&L8xnV%DXGY?d=5hPFohD2vD}ssb2n2$)+jFWLsgZu3KIZmi)mkG zumj3miDwzv@d5}CToq)As!1}t=?Yd5n+(3ZRyUPsUS6r($ZW7O%IMQ(c2bGt!Q6}_ z`o$X`5glp1JF9CrrpzW@E0`^Rq$YQP{;({>K+wn%F=&xVqo;<|-z@3Bj>Ek~bhCpP zCz!2LwxSs1q{HDZKOS08Sne>wAi3t-d2Ir4$Jv=445uXyNcIb4bC{GmT^(Irc>R4U zwQE`jo9$S^*|UQTUf%W4+u7-4rk`%{#`&k(M){}d36t)}NnQtPhOoE9#K|cS!QQ2e z6+nmRf^oq5I0u;F(vK=wc%2xT>ml{V4s?gj00BeGv9SGe5+(b0qL~{RxKd`7s#toA zC2^QwLF99{hGv)yAb?NC^14Z}s!j)Qaq;2HeURq($l+>AR_M%nyZ-486HUvYt8WZK zjMHsBz7Lb5D~*+cG~oBNs$KA*b_-xA0BMIJ2Wnz|r>MO}Ggru39oBoGMiu^NfROxv z$1IGyiA4^#d}UWgotBm;c|58yhVfW(8uiBjB1l&EV;NX06=HnPXE;J%^|?88B{mj`?oNI^s-1jDaZN)tlSVF`yG_f~JS!pOwHg?3@hb!nV3OYwsi%rEifv?^i{ z{Pmm#j|*1pqPF!=WMhyxgU+Bl1q116C_G(z<-6;I{DDz5kldjv8en%m`MWC=L4fB@RY)gS$#<7R%1m87z&fKUnO-`F=WMCD+L{34$B1 zd(Lc5p@$q>`cgT&X_{0K)MfAtp(4ntwidajv!vSXJ5$yNHQX02(q|$Jh%*7;6$HU_ z!MVMYiEemfLnlG5=D295T@O6K`nrR8JWWfn(>;JHw$;Y%mp$ zmnNJN{>;EH@+M7R{uHTEpWgC<`OOvzoRcS_Yz|_JlXcL6rkonm7=Df>s+ScDVs%`1BMp>);m=-H(4VPbEZ_vU6;_FP6bKo=4M8n>3-*}HYYr~QDBw`GIa(>%DD z6THC8(IwaR6qs|ChPAF!jmou=c}dk1?K$fKO#Xn@FR{{?u3zKAC5hJ6v7NT3wSX&c zZP1Q#{eH#d-Mx5+%J^@!;*`~$&2czuzal*UYUlo|P8N&R)5vx~g>J=CdlvF_pASoz zD&C-a9C;}^Vl)~u(}Z457bKc&>;MZC1!(g6llQLIV0PR7y^d?G{A*Kn)}k+$;(;rX z-6gG3s_s74wR_CE&){lAI7|9PcvLE}KG_tZMXG>%6Vo{AJ^?*+scv^?MS1NIC)PZ~(a{G@OuPBZ&({`@Ux6X!` zGyKb_0M>(&tFP(4;_+Z|WslqM{pZj9&-=$7xl_}Vmu7%RF^&3jt08U z&dL?%&9UcmxOx*wS8Q{X<7W&lHjP^+*$PvTf%tbBb;JjCrk6`(f^+K1tKjT=CzvFK z$%&|?I9OqkknS_xZvSu)4!G#4p{_=kgJi2&4Db_s9LHwZ4PArMF(J8sX@CCM$tCWV zgg?$8g zuh(aQ_)shNNyiJFD`Z5LE(-rOgM&{?Ss+Q1P!>zDNKukj zl}17buHKO-CBaBix4zP2EdP`=0!o#RS9`-ezcGm}twZu4h9t1GzSQ{Wn2HeKON#Lf zo=%(FTk{MmnM%NCO$z8=9H1C7f+Xt41lMf}$GJ^-S?HS)?QV*2BmIM%ZaSN8GJiL8 zhIUotlM&wOghV2}Qz-q)!3SACSNqESyC;lpdJW*G&*pXXhQoe(4(XvTE(msyY%6uf z62$HX5T?UB$hq#xE<6y}V7cJ5`{DxD5zTN9GQ-N$l0$NS>- zR*|Y$93e=1ay!4cxXXM$L)~p?8f0p)}l{ zD&g11L9xUSCkSQ;b8^Ps3S@1%dnT&#%J_F;Lbd`{-r}60hN)EO(BF(dw_F`P?Oi?C z@SN5W&;z=6`g*>m&a-_K%}_esi08&)f<-RRd)}-~Mru4;&O(Zh82nfdfKiL^FIDp& zErRt&1i{zmO1?U-nfXw4wsE6!o-Z*Aha4h|$Jz7K65zJ4RFC-_dYEhm){-8M)>#{B zNgMknrl-B71!Um&ZrB(XKiO>`q({gYur4u0cHS?}Lfe!4ngwNSyH`jLa=RpQL>^#4 z-f86`m~7Wb@{-~8j1ueDmISm0UY?>Y0~`PSG(X}0r=B>)Ah6sZ9N{!kqpZGdUR(lK~L`u!5qxc|jtZb1Zk*^5c9>C${ z!&k^gho5FGBrv=IoY%dzt$d}fZkFyg3|$fs4+K^q#R8v&JAWdX6arY+b|k6%uhE@u zrw=DP5fKj3fPT``RxXQOvObaXUmOEIJtq8qTL1^xD}zG&D?@=gg?Xp&J%ruR2VQny zz_W<`6k!V`0~Hd=BVuxVouIX8;69i?2en(CVx>qW&H*0foMOsP>w%ybMIqbBz2CwCn&cJd`^ zi!0=g8YLzP$4>f!$%IHZzRp2#7vWrM0~Uxk$>(Bs{sCY6T?w*t3l1S2ogOM11fUu@ zP;qu>f!Gb`b+5o}C8RNIovqpUzfeR76ODzarzZs^15ftD0%d~%2$VPYxU^`{ZLGJO zfm6^;-Fd)sb*BROp5h-39}()Q4A5s<_eQ(lcbHu+UQQ0;_HV9FA5&LbM{jGp^|?xY z$%Q@@qHd`);(=@A154O$YLhtLxUeF3uK=`hwY6CseS3KwKh7u_)V;01Ai@yR<5`u) zNbIKC+4Y0KyoRyFyv;1CqYm@{hSMmmvrSwXL}!_Eqn7p!w#x4uDdZbicfeevOY>d-g8WG*AdTZiDsw;Ou?TvRcQ2G}RGViP963ylEL!<4TO3+};V!XX zlRm=24mn1OBBzPlkLtBI&3U$*l;qw9`Gz%2!u2o@nX&9g4`^Vn<{Sr7PQ)icqb@Mk ziH+0=sm7Y!%HvI~RM091*DTSvvKWxtHl7$*mGbN8z5G#mn|F@+GeyI4M#4ss$##Ix z%}-$R1y;%}+jV1&5;kxQ^!!5PAiL!Owf*81ZV<7REF_dJrm_+l+c)p`IRb|ap!2JQ z3GnVCd4Xlt;Mu~1|1^iUR~wQe24y~5Rp{z;QDs)OQG!lhu0l(^vqIR*T2<>V1* zl98RtZ)sa`gwiM9@;dS!EnyH93KWk&UR`(?0&Wh%eWoD18opHWh~Yv*(I~WWSxoGM zI&<2zxmb-|zx8MQG>?3NZAb?&M>%3MNK@LqXF-N#sirXa!+c970dt9Vyvjm_c>M-! zfJ05?K}PX5Yd@^{N^^|z2UOi2qCO*zwen!$^N{5x=dkB!V9u5Wj~JixrpHdp!4g|D zZjPnt>tG=up&$ot1eMPZ&;xDj1=7}Sn)Lm=bADm&dA^Gj2-1a^m zb{iuipu%+3$COBu5Ylxp`>COd`I4jD&5TIEi|OClpNBq!vYh+i^;<@OJ`_Dw?D9E# zi6v@;tsAAE8Opgst@kBYBsIZ!AzA8@rVL#XV` z4oE1-p7^>}jJ*6fI>NWxJ`)~*WvU<-duP^9r;NJT*{xtAs4>ij?WOmUR3fDeH{T@- zh)-S8-eqri|C60u9H$mI0RjN{0s{bG{5N)DVrFUd%TT=HCu9Z~5QJae!sSVcHnu1W0dxcZ!_>i0Gn-Q-ZqUH=q#%$!b-cHemH>c|(dQ&|9o<&;YI zDiGGGbi;AiBwI@w0`fx7qnX3C{u|Eq6F5+jGy4acnP?=ihZP8v0SU}rB^k!gOnj)M zq*0g#OL#7U0UPp*jA!czS24~H5o5(uJ9-a@A=K)rqs7ZL#Lp}O6+Ny8p?X(1cfvaN z$dfG~fVF(eB!=wTy|Kaxq)jFc5zeCq(a^9Lc0!td@1q*qk3w+o-j`~UdhjXAX1Q(o zFzq3D6XgI9S14Wf(VAb6SqA4-*@0%A9-2tF2J2b%VwD(VjH4RabMq$q($P~azK$%C za=mMnNUqv4Dcgo`IE+&B-Yh`t<6`fWGM03JXSu}~N4Cx6A`m7HPb|iibZxz3*!0P| z=s$sH*x%~(Q~Wf2UOi&niWY?0f$|XVbQY$hH354=n<6!SghaP{H1z@YpQM64!@0wQ z0RY$~0|3DKZ=|v`F*30KU+zDJ_2sfIa_{nks<{aUS!?@H?V?H^9mOcqW^g&+yv`Xo zMnsz{qPrx~$CB}R#o9m(q84v>GXFB}k*w0u(C|?AP}fkcM8yf>y=Q;S$a@?-ajiB8 zdp^BNl5c;hcxNLD_e{c-28LJ%L@#B47U#&cCyB=#gAH}x3)OFQIC*>rXjYSc`1bw1 z^OAkT`{O5-Op?uPpQa#SAkqc47a6xa^(2cPKrNVOha2&hi5r_xq4;O}qM7rnBATS% zn9)f~tpOW*6}TFo*G^lkR<71@o-H>M^gBleUh|AMqNp+$IO9jMN#*@69lQTF(e)M5Vm zPDRoJ;H!a3#I=@RiV>mroY>*q2N*llsbMfVpgquYNnqj$DDQKIq!&{j5F9xxrKz#D2;#X+4&&Zal zG!<}eW%goe3cflN1EE|&9}7gZWrmdz+so~%yZ zd)2l(pv^?Qp1&{fHL|VWb}`11L;Pi)DiP<7=7UkB8iJ)0W&2+IAG+QtMwCE_79HDr zY@2&*+qP}nwr$(??6GazwsGgY_jPX0O*-kHPN&mJRrTsxxXX4^?;3WY>i__rVlMId zY%*?E1UtrHm>XKFsk9te-u4xJ&GYn)I`_~4xa9=w>vjo*G&?TZ;KliOt)1265li|+ zY_B!z*~h~oB#4b2fwnJ~=3GxUnPCR?pwfjzS7~|k!h}jej=@riH`osx%<*k_szqb` zdGAMhLl7`DC9K_rv)zNUw2c~tYZbk6Xc7=ZJ%+PE8I=eD|{414p+-7}tZJ5?B zXi0q(ATGipJ&@|NgPk@rnrfyoD2Pb08MdqoUxZn(u4p!}^ZzpnEtENIg(@Uz50)7& zEMN|^Tpst%?Im!v)R-*WuJP{w?@`5ekb<@l0AWy4yOs&!K4ovVwKd70DJJXNFma&< ztpkSpT`^r+9jlJsz>C?}ICh?MA?_SaI_E zJDKlxdr&0s23FtrxzLd3v*4VLrqLAYFn3ekEMFIwVg7Vf@g+dbBPc zKM@4Qgd3Gu4d1w^xnsLXT_+*{-S~?iNstiE+Y~*!!3I^BKk3P2D~-7`W(FX;Xl*zD zVsLe!*%ihCXy-W8Z1sf&GYtX=BSEdlZofSw!b7-%LtNe_TDriMO(m|!$O@fl@Q4Za zv9g%@0`~9XbEDBJK;w&mSQX;W?QUp1&-@MNb#q%U4DlvBcp&uD&tCo7#Y!gzj$SLfJ(xSSTGd1_Jx2=MZ!=6 zcc!LzJIp3XU9MeCPs{CJV_uyx-4d;E6V;saQvVAuXf`Os&_B5hkDc~>R-u9#?2_0# zWU46#^WjaTQLv{L#ib{)<0dXgMRFQ2WC+=i!R(8qj3?YpXijT>5Ws!R=~H7~7+`!T zxKUeZk^-!CyAh|5<*^96!{l+IrfzN0d}@Ed+<%bX0M6@o^Hn@@O9HXN=slDmj>4)4 zzxI3Xyod0~^8UHW+yNHsI)=N@b_2$DXNo1M0PAVDn@y*Ui0^k!a~{*e4v@>J6^vJF z)h9#>yh=LydmoHAn~h%0!I<$%B3-Bc8NUE(kOfPLX3wofvI01%05xaQ1-WJ_$N?;{ z^{~7(;R#+vA42YbIGo5$Iz*tyj|5NQR)+OOssNKxQsf#jWFR)QfHIo|-84gh%NZe^tH>E|LUTIO8MmCcBi@@WxhkW^?FwHhZp)~r%`*Xw1m6mY;@;55YbtPw&e zBKw=PvRdyz&ntj}auJZ^9CbX;ka+MJ})1O@_-44{YLX>FQ%K&EsJ{)Lvo{;k#K0FzUe#aAbilKtWMohbn-n-aT4EKxba)j~cWxSL z^oxcEu+@&yIw)v$;dUBeQ;c7qbav6792dLn?bJFx%7ye zduQ7$sH;x`NGF-0YNDWpzHg_ndIZ%;JQ2fOc~=zwq?lQ?H-JTVIrRS-Rb|@uJpby% z$jBV4Cc0@l&um>|tZ$929xE50tEPqFBYaunn5^aws3G;60_SRtz-4#QW}Ejb`_dl~ zpHHV2oi@+)Gc+xj(fpo%pquGkX-R?`=UOFOUhQh%1;yxz$Q%K8_BJy#!@>Bd2!4q} zWioGF_>QlY2KOzFGm4I#CNL5K)nK2;MA^g-V^t>`k8JL@$mvZ1;MlzTO!n|uu6;NSJV1z9im&jhS-uV%^g6npqL z=>0&A24;{sQfyT12J64K)($o*rUEPVUR^T14YAJZ(fm>x6ix-|Tr;v50V)7>Htjc! zR6LS8@wYrdd0eeTf}x>w#Jq_KIW}-r^1R(ILT)RuBn=&r!#Z=HtD@az7C)$n*cPi@ zVg+72xN+-2+VU;mz)LPcF)^bMI{QNSZXhne7sl9Tu`KkEzrRbOHfEK2?ZOvju$-F5 z;8&y9SF?y{Z5Rr^yr!>=nvq3k5Rf2}v0DMV&*0oE3{g^Z6nsy+cyZ;B^ zzfr|sy<3*gU+_l*_5VN>)&}OkJe2=HKo$M}*yMg)>-Jf!60bENgPi)gPc;xr&*|AH zES`Z$DQIaFptK-U#Qi3{SonU3NyHe7hXV)_-*g{r-QgHdG>GtV53x)jCijz>1-9Ml z_w-xXSn-DB#xqRA;#OgRn=+rECeZouWhB!J`b1}tRINH>ZB1abt-|)*&CS&=926d; z>NSJjI;Lw{g5$D-bv&{nU@u7rCrs>jPOp8`xsl8mMUt2-1OILxK?(;VrCdHj zLTb<;H|b0s;ltL28;C#>vR3J~fxoybTDcau4DkR6NS-r-;(D-qqb| z4MRfa@uf5V>k-pS&p7-y;4|omACV~|J~oP(RIKw4X?i_;u>>C$exDrEgmnt|`oiui z^`#j5bo?eNtaL#%49fybXUJFSVOc~+ka>gx&TYM~(FpEPq8(eKXKhmFd&r6-cI z$H%9st2Hm3LKPBHB*kFHzjoMIwWb_M#c~#4VdRkKeFHYa@Nw&FQVP^aC8XraiU$oZ zi|7*ags@Pn*f`T9N`#&2wU8xGI#i%?#a`t_6z)X2v}D!Iu69whSfOoiBv+b|@P zf0J^!4;sM=AH6Ig)cCbm^rZis8u)k?t!FWq&+xhJPVTu7le%ag)~pDF`1A5amkBU` z9nvYy#1R-VEzjD|_V-XN`JV|q?aBPQNW7cxDLGH5h$RoCBSXa(GQP~ms`jwsAWQeIwUwhVPiE%3< z2fNB6tL|kN>s+rkxU$OTA}PC^05uTIK%hGG{#2@?#N6n7eoj>Kl9J`%Wyu^^xES2f zb>R-ZBln-JG@kvNVDmuN>}2~KV$o%o4{e=etfY9x!pVuS2Kc_}JP^@G{`sb8fxAa1 zujez;4p#`)$c?!BVm(B!S?qQ;rsTpGt4!IEKqcEN~!tVf| z^-93b%(xl$=iY|Ln*!cC5jrEjFLcr>&&8F}A`?orXGv8f#QnT+7haVm7NtGnIJL3;!*BoWdXe};y?GM3cop3< z-4pu_{@*=gQh96Q>DM2=^LGz{@xT5=SR0wzIsdkdXeI0aw2bc>UH&#c5)u|`kki?A z@(Ai*7EOuprVTK$bY5|yR1gwqg3Fc6mF-S^0}ks_#{C42yX#Ch7cqFCzo;E^W<$Wb z8KoYqgrdrkaidd7ZTEy@3)YD=RC$;>0K2!R&(k+idrme|3A2)8eSY=L8!p=FU%5dlWEK)T8!YZQ zRB6p#UrMgzB^*vT&dq=sJhw#^QlnMk>va_5G?Q6smm%_lhF~A)s^7GeP`iodgaTw_ z1AWQd!X3Jx5is+V5wl1TB1eovI*oPiq@B8JxQZLJU19@Q=;mk z{`!tCEXRQ}#rrOBc8D!dHib!paOQgn=P0UPFmjOvoISmGtOrl|?yC=#Sj+q?Q zs3)Hy!G=@c%h%T!vgyP2XC-B&G}X8#nf_MZYk50;ldwz7VoE9(&cU`Of|uiISt{)C zJ5`t9I0ZZ7BxM76)hszPdMRb~-^$&>T6EkezpHF{U`hX(2j|kHe$bQb*=%}HH3Ek% zaI&@mjpeji_`4H+m)CdZ?TPj+9(IzxV|*77p8;JwYgU^)QQm<}8`V&>gU~iAXvxF_ zmt*k z=b%{2pva!BppCzEow4;xGCU-9{4JHg_n%X13>4*RY8`swdg(_E+vGiJ85r(s1x6g{ zpO0=F>EB`vur}P5aAYBK7E(;W<~OuH+}@r!APgik>9bZMI$8gq9uEprL&;7;-8}kW zj*4Z#IQ1s=YK)v`u=^`DCMi>jZ})TV|EU?S|?+T$x`UyRV-a>7-H zK{YcO&=uX+)3TPOfRP_*=nkY?)Q$Zzr zb`kP0vi!h>+6&{BpM8@svj9)H1>4*qQp?F@5Zq7r3ySP8SL=pbcy^Z$YcJ_9i;9Toq|G!nFwtw)i+utIm#UR7Zxf8T=FD-^KJ+287Zdg5sJiePH;G(ABw61e#bFsZw444n!7o4lA46ZV|XP# ziX3nH*O0@2>L`t6ZYXYL@}z_cfyEL~IbIcS8I@#x0KW{Sk-~3oQoTe?2J@6uVDD5f zbGOkpzqYvtg>4*vN-Jt#q_r|IS6DN%8@=8>0q07@RN&5m+@1fic8r zmO$^v;Ty-UdKAozB?Nf|ep>r+!B7!1K~5ZZ_uM1M7J3q$`gf0b3})7~WR!2WJXad) z3o}r9XM%+eIBq#*M5sd(e1+L)%wJ-i*((!BRc4_AlcpVk@&bY+CC=Q3CmyZ3$$u=( z)b7yvN>nONwz8MpZF=LzT5`?y9*V|D<4u9qFbac5T1Tk!gA~UTKvme;Sv{S8r4PHW zk4Fy&d*0FOj&v-}GtttnpIh7E>-!m92`m?#>*j|Tg_^^m7QMz?QgjQIvbiI?Gc{hu z?TWdcgC_PKjhs;nM(B#xW%cX!eN!O0HIt;bzCb?|pW5XPXx`>I5fa+-pWVPno7f#L zd&;c+n0owip!}#H&cv&e{LZc{mqEzaZ1q=G^|nG|H%j(UIvp#X9EjJD3|fY&io#TE zBJ5D5Dtcx^ynj{aocWrDB7mN@&X^d?1E{{63)}q&52IFwf}MczW&?syCcpgx1Djv| zw!QN&XgulaQ=C68%m%f#ykLZeycLQJxGnc0Y;{uA9V&2t@@8<&-gmk(d|rcIX%c&H zshzS-b8bSpwUppY&dZLGe{pfO8Gie={uwD-m6adxwKf$!0mq`5s#BZSve`pIpWfJq z@3GR*!f%l{hli?UKVMH?wOFMz(K{aJz7Gh_-zRB)b2z9UaHau$M8?h7phG5&aYt+bZfBB9-OxA#Gpgzj6)uh;ameqSnsuCxrKdbjkM;S3Y$I?B8vp@6*dP z547wEL9#ynieYk}xauqLZ!?UX6vWcWMBQex(S34}QxgqgSH*1Y^@>JNr~xeVk{RGE zDm%-bYa?IfkRv$YzitV1d8Z)#fS%B=x;&$@@3@#Bp~Y$Eb^};-m3`T2^ltrkqi{0M zyQTOtS-jKE?d80?1IIUWdzID|lHYn0S+?qJHzA)k=W&cX z;{`Isr)T4vyweJX)`|&FaMJ&>mU2Vsb>xlv3H;w5YttT2b7w@q1IP^k0PBAZAUmV~ zPh#*pgo3?2T{qeiPJZ5yGd8;qgtA$P-VjS&J>KPr$r*se(eX(*ucYu?3yQpl>m993 zKb~z`ct23(l=F~K8rj*8W!;eP$dRFffnHzl4n&DwU(Rab@mN{mUp`hceXXyrU+8$d zp~;25xo>ae!0J1biYy&Qd?_DqWCYEqGVtuoCa&e=5ek>nvM@)>$qzXf)jz&xiKc-BQT&@c0ZScml zbi21UH*I||r?g`3IwqB_l*lJLrv{Fep1!^dN90CCRq=SYwz;FOSz>X!|8;)7N(YM~ zTeIS^{6L|2uyk~|d;U7`AFrLxY;?N5%=m48Ow6pn?-=>EecepKn$FnpzR#K#{({5Y zi?y>8FE2!J9`mY)XR>?B zvE_x(JJO^pg&!L;uC}a&t*tjKAG*bD{)E6^+?SKG-ljbR(_JOgo6zNSwY9W2tS@SL znIHeIUvi=a#`2VSzQ2~9h7|{q54^<$8FFuGzAV4p++Ll26caw*PI5o|d%rg}V#Pju z0d@H$KV9Cgw~n)AIk$Y018gx^0%Y zz}lS{o*fQ>?4aE`LE~BN4eVWtYL3OU-bT`R5gM_oG~3)4*tR5|cys1j zHE)D3+90EhtEF2xIIObvkIkW`WJDL}4BBw8pY2JGU!o?aDpV#rS6&_(n^E+GYrfh+ z<3X-b9iigydsP(7d2oC4gtPOVTki$tr^@Ey*^qEXRI~i@zJqlov`yB}Kt5Ir14qpl z#`IyhBg8B^5fC+MX0_90OJ`c`b}Sdll|E;07tZW2bI%@ybB}i2@qFkpCVl9)#vSUC zjUzApnPey?##9b-fB??PM4UDzcyb z$r6(esO8`~*Y#YD?6sPgT(}fNGT^_g|Kto0xMe5WcRFRTCl3ijV-Vlzp265FI-IK~ z9tV2CmLVjCx=`?V08DvR+P^9x7yFfs3})Uil3Rw zvpz4f4Q4$!X>EbZqugL2INVZoXCA3QWT9*hI%PCv7rQr%DAEKx2NjX2guSi~f~qH#oTuD?VOcBrA(gLV0>zkc&tDIWv!gy z=(3?;W1&(AO$@wdP4aK&h4t7~ef&)_a!tFX_tJm*+DkD=!QW~E`M4^xBCqPCmVs)B zU&C_2QQUA;3D;x&4}A;&^_+FgTfUl7uNFl)Qk-sTOoY$6z9XMcoNR=c-V8OA@-!Ba zvru|9R3-srrISYp6B~kqhB5-Tb(K(o13hjhJ=;46Fdf!)GLtTW-LPXNe#OdJ#65;V z9yLi0kZ=W*aYK*(Kg`UV{#EQ+%)re?3l+@9XjL!{wxFlHYKn?K;HD7Icav&V8%?9= zRIHL`?FAtO8lE=9dzq9jZ5C*&Ql~y&cx*x zYD~ZyM-oX6=zAwEP}{N#o9ZQ<*rf*z*9_A}oq_9v6g8!N%JifSCG~>Eb(B=vYe77% zgS0(Y+G)+7dXU^8q|^ln`JOU)CE91@_2o_(A?*<;5{}QI+9`J|X=FtF4QGQiKARL;shc=T5gO@*zQ-9Vzpi=NI)Ps`W|)r~c5F6_b{Ay*k!dgmfT&^oj~g!`EFIz*YGEF5AL@ zAu-d`dRK83r=(q4lpEPyDZ?!O$u&naq2>`*UCt;Ej0f!3$wk#Bl}O8FBs5(M z>@Ui0kT(}TCpkp(aZqBzn9t@AyVuko`e|edD7tkL^=1(D`m6Y%r@6WD=mkR)b6YC> znn~Np^}9f}rI(zdaId^hnpr0zNfy_D_!EY|*q^7Q8H7)8q#;v8k`_q6W&_NmIM$Q! zsp&~rte~Q_Br&%#S=gAbupog#i1@=%OqamGp-UOe3Tfz+Qq6epNmJFZRW6Bplq|?~ z*?X7}m*-5*qn5U+ANv8D2c-2?p}qU6?CvDtws#2xqvN7A5)P!Kp-i4M2H>Tay$TGf z%;&`2Px{3TuT;e$6GL1*(0zO+slLPmO8^NywiAL+LtPJGls%F(2Qjq^nEk{k6DzYY zl>qK3nub0mb^NUeXarlcS)ScDKDP5`aa5L(9lHeXQ;Y(mkj2;v%0UjaVC5pYW3xOP z(iRbhZ^$wBNl`H=u?sV5?20dPVi_Rpe}Kt-6>%$wn|2`0K8UbP6~avg$M}jC0roSZ z1$3gVru-OkF7dO568g=!P==1xYPji4CgtPnBKe$jVJX+JF5C z{YLz@5J)hN0ya z>%3Lr6qsrflN6`QWKRLI3oO~ikH^@G8}k=R{a%Ht0++Qq)|xl1|#UWsN$$A zcYeo&oTl>nlkdq&CKk9@q%jGK=1ei})^Lz|fO3SQ`Y2kz84#001?|N}?xwUUeeMgu zgiON37_xQhrdLXr6E)EhNi%$vt5K@YYnX3WT!Q>$kHhXI(=ap$yP>9ZBWW;YhQ0fT zrvnpF3XH5drY8I%7`N36EV&ue$ zCqKJTI;FR&GdnwAuEGFO+V%**xtH3cD`Y!q|s@J-I70k zDUx$*$y%^+WJ=N4E5-oiuDCB~0HMby_l^9Fs?1V2PVPd8Y=%?0b>OW2ePEF5O-PG#rEgZA1eS+)M(lXvD0cVr_83Zglv(= zP(l$#GoYF^f#8ZK;nOto;OeDJAgQo0Mv+ji^BO!?JhIRFAd?V}70N^l8aS0NqIj+6 z-&d{&Dmf47D2F>*B>~hZB;1ZYI7@=Twl^gwXv*aNt7W?^-=`fCH;`{#k1gP1n*}GX zrPR5$p6YjN-(;KL2DXXcCP{o4%N6S)4Bub03;2XR1+;(Nt@=ld zt$EErCU1pI87o+ov#v ziO&AHdGe*~fJ$M|?tWz5;f1|mkyOJBRnhcR)$vfC=YQwgL3S{bI|ykZBio*-w6+eB zLjt7m50pg0$3O=Kt2E~ta3WzmJJ&Rk1#pM%)NglbRJs;33Q-1a`6D!AaIRQQns69Q zQE1Xl&2WHcPiqpO0L9%?6JS2R%Qdu^qE6W2Ngn`HLc&(@#a0%_{h7kad_2MC%i^^~OR&f2FaJR+2mOXMM`L_UNVgZH(~Hu- z2&f?ux5b-8G6q}+QMQtwJ!~U4qt$rrLpQ0?#x%i==%Kf^RrbgGx_8rO42)a|PBx7P zWu@?wiZ6^O0Y#}sHb&kxX2G{Z3R5pM0jYcO>3c&_;tV}Djm4b1BBki&L7fN4Tjc5) zujsEnZupC2=!Q5U(rYiysG_iZI1YT;L?zEjbx}t_7*}VVZV-{O)r}*ANs_rGSqbei zenk(4AwOdt)%};!>4<8-4u3M$9BaG@ndHWyWIb<=R-K8)6Zq+`hHWZvpw*ZO@b;Px zC90mJvkVSw7&aJy`0_EyjGbhHg5>9rpvl-)1M!h-D$oXO3Uz)e*J&!L!$<t< zX6hc0k2vOj!eS3exwqt+tHl_uqNL6!9jB62>YTSZMpp_~f(br%aeP^6Q*bFe{t&ZT zKi@~SOt{dv4#DCX2lq$~OmAKy_b#AKK-y_?LNTK#z(qcz31aNvCD+zzn|?kcOVF<# zd|k78=<-q)d_vdC3*b|Tpu92vDQ8-T3X3$dZL7qQY3OXT#1W_{;*`KzH;^_WqxxAZ zJegcAb73&gqlS~g^#(Tc)S;*M(l~oQ(QvdHF(<*y&rbR-Bu5Q?*#iAE-{+D{ZOTw| zWjX;5@n+C$-EC(Iuifu3Zu@>{bDedi$+kz*?-O+d7k_5froCj#LAvv39Z?s8sSVE` zN^^*b0Sx5H0p7>W$I^(h%_HyxDR9YX zlO_fLgTb;4LrlvI_W~G`;M5H4&ofUB$9Oz~qA(Gjlu<*Yb=G`7vT}saK~rMN>>u2P zE^D|V7r!<&-QFf^`QjK-CB3Q)qi zzZ&i@%zxojf;R}SjdQCCRfTsHnnWRQL<@D5t3KZY3_^okgzuXu9k+i$PH8Xe2&rCZ zbw$Fl&n)EMdx{#fdw7d}c7pYPU?2^Fu#ZS63aS-gkL>HF)oN-=3<*}(WpJ0rpE=_3Y53oaR7>QK$LAl#z@ZU=;9)R}etc3o$VP1K|Df z6}8t8#!|vp12;Ag70J%YPby=?-eH!^$#7obI63Q9DJ^bODI2rX6x9ULyID^eYZ zOe$_t0EHS?M40j4x*b23>2 zhsXKST$ScXcs=P=UWqtfj)7|&tz4poZX!vmi}FHVBBL$gGO)-NhXrW!g~?)sP4JW; z^jXn$Mfz$2Li}DH2y)68pt6e$L`R_9EG;n*6XeQVp5}{4u_nT) zM^=2ijN6q^jgQ~T;#q&COt9|NrvqJ|Y5ln-?zG)ic9+rdAD_YapfK}>B|eSryT(sJ z-36`rBH7&vJR9<@r3qZsbLT4`14aiAfSM|892CdFfeXMOMp+nz>J=SY&`3yAJMaX; zucW6>&H&0Y#sMXW2p$vBAZF$McCX;k-%tx)F!r_#z?w7KmCRD}p%foewYJHd7F_`u z05kRGt1Mm`fQLny`Q1vzgj#hoYK+CWBJksqY(wV4A)_iqvu&V9cP@%dWRL_AHJ)oQ zAfkqb&)waEOEEvv&LLd~YR<1jhv=&@>?bBzQLuXCajSJQ`Pd5?@Am*5`8)pep^;W8 zkzh=s3Q=Jz$!N45R^svo!mI-U07beh8CHZ0C)jabu#;uR+%n}#i;OVdKly+paiGc2 z>u*=I3R%S!Jo$$YgtaN+(h#p$(2l~yf7cawwS|H7s;?twuxQ#9iT_;n&t1@o<=&K$ z+?u_ZEIYDDAc_cRx9BM$ywo$w8$}m=ZDN_Y-io_-H3g1b^O${uHM4k&l~1sJYuSe4 z(_p$jzK^;wS-54r_znd!Ro?>qzKKvM#hp)3lodaM3?-FQu*hNi7DUX)-*35`nDWm- z-qdWwz=<;Dk~lNGx{rxX%VhRq(%|lw7COC z+K=)h=f(2G6Am7_tSlF`y9Kx*UfdrE`F}0uw5FlM%w^AwKJftL17fX=C@!)zYj)L_ zd9d%+W?&QRKo=ZdUPHX$L|vU(ZFReWB;Q{VsoyF?!4K$QnFUiAm0&@{O%V%ojapIO$8n0zYi|#PXVJD#CirwST(TLHY$F}bsI=?(iQ?tuLRcO3}8eu zPY2}@#v(nTQm74ug3}F#!1`e&ybPd;l2NH5teM7FbgrvDq!{k^wFHJCyvUXe)ozKR z+dn}$vEcW_Jo%PH$lAd-INd!V)%TRuhD34l#s*x?g@aBADayu7>8i%UC*tesu2 zN%rYAl;{h@JIZztzxjt28;xHEND2xh&u0jh7iXyFl+xBdxUaq$sGD#3qY(Qtx#Z(`uH|4 z6>GLh8&TB_z%U=$_a6eGjzv3QES~Bk$ZI{yd?JjC6agph^^@2_;Y`gE2B7+oYjKf-&)omhgINr5$uPYku2!3vfkqr5XYo%Vm4VOLk1F2mb05T;HUJm9P<2 zm?cL>S5|!W^0bk&AZxaf6}EGTE2)~OgUK#P?ZwTZy@|cp43P(=B;C!800#O-39E5O zTb`X4zaOLhVW!G*kXBdsFop_YHUal!wt;7qc9G!_mT^{QJkte6LQ6+>Wn9_7g#0c= zH3B|cso8i8r%H^K2W^K0Kw7egAK7|Rr3`A1eQgmbH^>$LI%wE4+sag$46C92B5YQr zKZI3pOpIS!Ol_LoRu<#JJ0ifiwtrMUSIB7nTfeh?#yR#s1HI_Ifb_AUNW;wtujtJT zvVY8Ijx?O_9nML=YmS!QC)@o|h^7Bh z=cWRyw?&5uvMr69v&gVo{dZ4jk8_K$7fJn6;(ub8vLg zAhc9C!Gfh0#`K-{I5G5=?C7=r30E7Z=>$el{I@dsliJq)FN3ZD^Kq&}6zvHi%4><9 zX2T=f%05jL1u0VY*)w~X*E!D5g_TO22?x)^@R3Iyv*-G zWR(iyNH)Tgjf689DDKNqs>YUQ8dM#s2F&>BDjpX#)e+aWVKMEMY{&K=8+SP9UsD6B zti0dfZ|Cp&kB4WYI}ayxPTM^puikBBg*=}3i;3kkXK$Ffl2Ez|uL=Fm5IYprB}N&E z?Tld92jt_nND)jaV1Hh}<~eTeT-quS1A7<4Tt|b-NV+FBzRQ5Pm2lyVz{E#xp3`<< z&&pTzIU6dDOd#Y>G{glukCm(T6d1;{#X!hc%V+T) z6h)JlE^t4Xd$Kk*1j+&!Eku18pbIXk!2k90 z9MN610DOD}39LEfJ{xofiz3rfntKLy#D^M7 ztQN5dY>)@gS`RUX__`i?;41kiJ$@#YxXu#Kd@iF+r=MAou1R~-Yj-kc&gn3lBb$q0 z6##&7=f2uRGo2$*n1RLs$V_X|E~2@J*S*IbyYb;0a}z{=@0vT6N{3DonLjq-lrk;- z=iq8AJ(S7}!SG52kv=U;6oy8ua(ksEwMH#ND!qR;#hiLKj|Q(>$hg@D2%3ffxK)5g zk8apPAnfNa^qa1q;4Wz52nJMWHp7pBlw%cM9 z#({XTSUB1;XcmZ4t3)7Lra=DUqxPWp$AL%y`I z9Xn{{_ST`*YWMc4M}-?X1Ybq!eFda`FhpP5jR5U0Z_rFHK|Qn^eW%p>CmvtNk~U?_ z6Uv!R__|k%5Lz!ut$$WKam-Ft=1@Pp^+s}K?jbHa;_Vd3#okOIi)-0?h)l8m7@6!D zOQ(qkWg^`9bviF+v=FnlDRqQ-DEx^iV}~M=3z|s{c0CT0}Fq1taZFre{Ex zY$qMASaxratzPqOgH5nd=3PrH3JaC;k9xpR>=}&kSQpo8efCxD70ht>)Ahe$Ms%_Z zpig&PeYWA5Czv>x+#8H2ELxM-oEg2%+fDSIm){F!b{_%r{L}QEEh#HCKl}3h(bXC9 z=^$V!G!ymp(R)~!&{?_cO6~_4KZ^?xeRIs*+@y}iihru+W*k+RlSA5MRd5ef(?SZshKb5p72ORmzXKMY$+Ex> zdIo6jCz(FSO3`1kb>d9ZCO>Xv|1l0p9Z{IPe#jSo11kv18;HNdOTl(azvm%5;7EyYj1YF>DzvIM#}#48A-ea z9VL@J{S&}DJK}lz7L$YJ@qGteotn#owdL#c^L?(%yK|-e+4JIb(HQ5++WOlw|z5thj+Zb+y1;cZ*-Tkyw`28d`qT(F#a=laAp((wuS~ypW3=Ao!T1E6&&*P z^lU{u&so>a{dxUs$A{n8(bd`U`Q_25yS>S)wSDOTOh0uL`}5Jt`b~HEc{TOAQPs&# zYtwUFw^#G}!*J;G5{l`)aq|P9{`-|gl z%I(kh19+ax(9(aGcdiC6kr%(YTf67JQQuu2uZEA;&-$g}VT*9drAW!mr4B#OP{Y$p zAHTer%#F}l6i%u4&eGS3e_tK{+<#0W72F@{-s;++n`#hZuWNaHI<_B|;7@;e-hKa@ zA;fiTAF2Qf03h{iKl(o{W&VGLs*0v9mN2rj2?rKxh zS?njAJMEK$2Y2Xw^3}FFm5jf{;m2?Eg|#v&;l!(zQ=B;sDfV#kl=sj^vGEF2ctougX0#bfzL zZ`0rXcLSoRsj=jqsL*9suOiejyM|P_wL-gpIu(v|_+%Q8G*$VC6aaONZDh~sYe$!e z+2xIKVcSxPYa7-vU#f`Ji%?RCC;Ux2i;2f5oFes$d*#f2&-`M*nVsVICTnnMy2Zk% zkw*`U_V?}%aYoFCB7n1#YSb!DWjBf22|dts@hTYyTY5%ev8&jm!r}L$=&P=XC>E*X z%*_ZMF`snc{7bU3vWzQb=$6kF+Quo;u-C~YM>|a8Cr=e4lDO4MvxQxPTMA;5)N}HJ zIGOL0s=W+=2#s3W^*E$770%6sp)3@RLXs|IRGf@GjE?pqgy0dJoi3&(OJxI+8VY01 zFg>B+!9~aA#tf40%$Ad8*&vGkiG0(S=TB_#O-^`B?4<8Rf7|TsOku1}b5b5ko4(~p z69PqvB09nuHN1_Kf$yS4tfbsyHCiSy&av^=gESrnDkg=TP@xZ3_svmU&ozwhRte{z z_4%}L^~OyZR-TQxW5TNcW|GeyKUva|iKu!q zY-dMgw{AT#D5r_Ki>jNKiFN4wKhL2xE{#7bs_PL;} zkr6UYXITuMhqMQeq}O3zK@Q3gL~5-GV|lCj`Ij$Gpa{-js>vbmzi#G4kL|2Qj@CeX zB?m^HGH7H+Bc{inFc$&)4&J&mTn$h|m_^nyQ$^}o?(l71y@J{`{&cynp}MPlp`2S~ zWuY%?o7!pAJ7Xc(&se(#;|wsjX1+})xT!M z3ilvMRkHrHp>kXGd^0T8ial(#wld7zwqIVI^km86;5;Ft-^umEeBlgwBgE#}GjtzZ z=Fhs^CNQLUp=x-&;v{lAw&1N%Om6THG%|8yeW3k!7v6^ZBuIt^06_XRfX4n`JWU5f zGh-)ryZ^Yc3~6{e7P2Az@a77fY@+BHX1N`^m~y*Yk^UdL&M8WeU`@hp+qP}nwr$(C zyQgj2wr$(CZOzU;-aU68>oMz`zcMo_;){$8nVoK{#EuuZ=oV zhEX&F@}x~?s#XD|s+G&%7htSEqj*7cCVt#xfJI^QW+a7RvDt`*PmqW`{%XFjB5&Y) z;IKmwIkaUh#D~UWIwEI)L(D?1$DpzJU3w1DeE*^j#YIig@DSj1PWgt#A-jN)=}1w7 zghs@1W=*alkb&4Bp~y(xJ>OAIuM}KZjBFRKTuT?rf+-@O|sTl;z`S(NQ z8gZYq9hl;}w-B!eIyI49;o8Wa%!OY~%k(ZTh}6Ua#uvd|YFAd*{Dq%STIAy+T_`ugd6{q@7gC06=~ z*(}OXcP_y>R-31vO}q8FyPJDtEVY0MJ*Kj)qTaRY6_g}adO|Jw!~SZlfTzHQ4w7_5 zv^ITB0t>toObYO{?i%^;R8i}gu8a&zNC;KU7a{$-A#VFCh%9BqZleq)CNL5u^qsB4 zMU+<{^Rz?FYjXPbA^r!2S|JK((Zv$L))uoMbKJ;HS$WI0sNzRg(=wnlGZnv9jliHT zQ_9ZCcaGMWMPHQF@hh(B^{$oGwjx!reKE5q*G=o%4&58S6Wa6-RnQLRys_3&>rQiw z_tXi>fwuR27c^^*U0d_@Ddjii)fW98mAd{>5J$oMa3yY+gPSio3`KN3I2z%N7k*FQ zXN=yDgTp$WL9L{ZSh%Ng5Y!SIsOc{o_s`t-(feQzj;`WV3x4P4RH)VKrkmuNPy}fk z>k&}u`afpY9}Xln^<$tG0G z*x9aRmu@S-^1hY^pV=#~$tJ!w89}JHBq6=7X**U;Qu9-lh17Gq?kC}Cs`7px#vlAL?J@l6 z>EKuSc3vLqz9#5n`ibhqcdVT^KwzIZs^uD?8K$!`yGmnBEj?!LC`qGU?9KJEVgdO+|Z~nnWlpBR8x{vQ0~7AXk#$w%*?1j62W^{bVyI5jWhy2+R9kcjiVnB;-o#T8oaE)A*S2|^B{!q%zNONfv*ea+3Tu1{8lUZ-PlN@+V^byiCNYCiMtCVWz z>riS{3by10er*`6>x?3z;s2aTD$uo3yL|8taDUPumiLbOJF%#rZXT88_V;ue_pWHQ z9X_dSMwj)sTxZ=d5pKJaV8fi9Q+ym&1U7OHhKuWIf;4bV1vkVnQZegX60gX5`6025SXGN zI0Mkv0h!~0;RDjb0=AP=MM9uGW5L!ts9|9MvZPAUV1tKZulFY$Bydi5ke_d^yaR>u zgYp5HfSPv@}8FK!z{<`mn?u5je{}H+aSYKvlGsq zn&(NGEkW(FbsBjHcPB_8En*J=O*&048?!pkHUS{5`S}x2COj6&f;m7D)>t33Q|@bM zwda$i>e@EPmuMw1?37}KtfMq9*CtnYFd6WCIYP_Fd}_AJtP_2r*r|&@A6wBqDT(X6 z%c}~-Oi6>&Xd1aP8rVX#A-Ke#@Z&%6uD#RM*F0=VLs(I?xADCA5Kl?CGQ|q4VLX+z z%#L(nTzZwy{}AWVD<}y~K{;L+FtSyZJc}}h&Rt@KOPOMd9=K_^ZIW|mPH`#lK$08i zm{U(?BJsPM=v7MX8OF0x^#neW5=V=TcvgXYTBWt_3`Rvm=pPtp+0|^`QWQWYQ}E{> z1}S1pTH7{soHShLUfU)t@nyCUGl}Rpwv{L3R=C50#$F>HkzyUZ?r-Oodd`_bj7?TL zXcrPx4o^uz9v^(NRdH>y?VD@`tZhBEqpqL2p6K?c@>i9EYV!$zX@jqp18fPcfn_g3 zI(8SFMc^tt*aXKAJscAf@f{3vo8xnTGeVo9fgQBvz(iF9vFKb;BuI{8caDaq}3Nz~fD4s|LMRhZBp&eXmg zQCH!n5kXvoSY#puHn&#c|U*HqWW4-ZtH{IG|a#PDBu zg-E2Pv?W4i``DcV&jpEA58!tk(X5~$#>R}T4yjG+*^EY!8oT8yaTAGKcEC_)anB`6 z?<_w;9xLu1!rper$`KrJBceui{^q3g96kF58TV) z5tov(tux~RFvJ??dD=#P{7i2sriW>UmF>fhK0dqkEDXWb}oBRdY?F9PSyGz)jolBQL1W2(LH;l5Yx{zQRTma6`h>y-TycEtLE>#$%gT}mnT%TwE?zOT)kzdV@Xf? zoT*JW1KPT4#+}Uo)QvTjbc}-5G{+~&Hoz|_^&(Dm<*V1zhreLeg3_gQ zqzM8u&;m#ywEz7!5gylOOx;6v&x@k84E|4+s`$Qi;^@SSD<0uAnujkhY4!k_4BB_4 z1lzA8mceKjm2OVR#oB&O+~+6+Q`hh3FaWz3rjIx)cL5&CQ*7@u3}VB1XA}ZqnzsLd zwe9;?INmRd<3$rTJyrl%dw_-6pU@kJ`7lJvAHBwG__j1@9w9a_9L1C~NRTc!0I*&u zll3g?ifa0JHe)e87kN3&!f!GGb42i<7Jr`k+IqJhTh-=zbJJgdtE);!FSdpZxn^VW z>uvYT>e@7u@o}|m>pVST;W0vZF@p=3(&UcPDMLI39rIn#;3(*dZ4FQd|2qLGr(+?4} zvf|1t@!Pu$o1#QLJhdL-nA{gW=#G4i;Y%Pu3hB8kqJJV@q~tW5d*1P=!^0$wYsV`|b~mH4v@rDT^d~A;OeWdoRzgN3ViQ1MC@0%EPF*lz z)j;+Yw8TZ$*5eteV}uZGWC^q{h#rB)dg#K-IzDRwIpI=)961BFx`AL5?!m{dk7SNq zkr^B}#aZ7D%@z#V%1b38lw`zAAz-Q`?I zeQ)cplm7z#r)F@KwtO7<4+)w|^Z$T2oLpR;?EigbcCc=o{?$eGUl$^ORIj0r{kdQpcAi)4=t&(=^`9}2$=;t%P-F=ezCepbHM&aitH80`+ z;UQB4cm?i?-T^KD`T606=NoZLdC;jMI7!o~43N~73KZt&d;8 zE}_v+u^9o*WR4bTNSUyzG2bnMdXBZ*ZfEB0bbKyBw?q8sNa!s#zY19xVzqTp}j|r=P?lHI**k&H1=`1;Qnra$FVJhRCFb3#XO+)O~m_VN&XT z$eW-9kjDR5GmzmirvXNim!fHr?*DJUYt9gm$xZVtmNp|v_aN6IYC`{3a!*sfB&qQ% zk*P2j@+Y0d;izuc3lS=YRs`5I8d!8`xWDx(Y)?9^QdtxFbfZ-{>Bp_;wFj9#5$|L?^%N; z8F*|LJS2rzE&PZT17?&MCNuzT@qjqaSJNocEtF?j3ZblZxDe^j^8gDP$$BtNbP5({ zt_AH%P$~^u2XHU+1JnkYZZ>e7HfH{!X!qPAw>0W~M%$N()T0rLDw~RQzrji$;z(7k z{zy+SfR&d22s-JQmw8&3SYkOCKMXwYsrW4}eR0e~Z#+K2GB>|5P}tlF>fA<$`cc+Q zj=8Yd(ZESR6@;l$+|p^1hTlm8yu@lXM8!q1VM-a^u8Q3zwnZ&U9N6f&tg`* z--cRP9a@Od53)CJk3+*uI=gw$W+v_*T<_&|@GgZydkBq>&bD79xr_1VWY*ie?P88r zDriS7?)b)X6CqNy)ibub)(p{3wrG09j2LJ5jl(G7TjCDRG?Z65k1Thl=!%4x2Ts~zlpA-vXBH{8Qhj)2yyHJmE~++1%P z{R>6ZDqAuDi6H;s zv6%B_#gbVSOh0-jc8}p6-)m-tJ7@;pjJ69nJuc|79&hWFz9CNRKOx{xrf&g}Q985E zcKqw1bddEvt+~t0=A+V~Lzd=V!VeD{B>5pYQcLYWGM9`V_6Kr^HLbKextdg-Q6^WJ zw?O9_dcyiOsg&XA>&^VlIub8cvNDOsqt*j6x>rvhyzXbSJZ@FywM3}HAVjqOs_Xsf z!^gSD$jXy14WYSYceNOfpi#Otln3EyiY^qGVE^rTElr@faE_O!hwKzVfI!{K_}7mml2MAwZoLTm#H1pB@y3@Nqf8kRqYC+-Y(~dD$SDtTb zCcX?^=u(9{8^804@LVfcyYgLIKDn^)<>blEezy+6k)&PlCTZR7j-2Lr zPRG!G?`4ez#vKTzP?OoV-14Ztlhu9pF~-r~2hQ!!Nv z)t`Crje&ue`|n#E{e8kWd)N2>*_sDIp8Brv)M1Xxb2vGDnB?)Z#S4Guz$r9%anWIQ z2M_J=jw{8v0|?n}Dl-W?s?2~}62@s@p_H-j>b`Wa zH67&w_4LlH59f%p%r&{fSE zZUHUWKZEsq`(wYpw3_FH+AuiKy$Pp4I5T0w$(otz+nk?r1&Q<#5W8hrGF1=9Y1@I* z>Yi0(J#JQ6TiUSt(Ol?m)9iyfvnix1_&Vsuv1G;WsUb{wNX!vBl>c^yG>H^`@w(t9 zVtC!c2j04s4at&?RqIrt2K7sW`e$~1#=51Hzxv#hulzz@$)!l+gXZOXYVPz-s_qxU z_V43Uy}p0ps5WUXwb88n%EYPZH;2#rPw?S~DW-4HQ*=r9LVP}c?JqnHBY$?6Rfqh; z$51KUyhssE=gSi7uO{`E)Y&hFZXf>YOIdlR zBh@Dl>d%jV)%oA~GWX0|e5&xuxn0XabIU0D@yGN`S~e;vB#`D*)o6F-Sl|geguCwP zXNHLk5h+8eLg&4wnb3j{ya6tZ^L0^jv&_&KIs@Y(??U+MbKS{AeYl;eJt&hm`{`;(+7`^I8#ZLiEUF_423f}mN_TxS%j z8k7_#HH9jb6}I;EEpGG@mFmRtStRVk3~lODA5iFSfdnT(VWF;a0rwdidgU!M9kK*K z(U|85k%8w;(V^$HRQmok>Qxs~S%pNyv_p?gMIvXfL?k6=6U$m(<5Z}{sz3N(R6|i< zdXC%`$qq2BO|6;~3iX#_`bpa9S+Yy1o+Y1+6x*YF)ts50%J?x4*ftbPdRn0RN>9Y> z^x4`X2J-rjoO?YP`0b*O*PRF>W(nn6tbkGCtOi{IS*-gkM8wifC1mPGmJMK8mufZv z>tO&ElWr3!N$ENXGE4?!+jF;W=|^a&J%{ zN)kFk<4+d0yNJp$jZ=_@S<3JmLOU)>4Z#%RCP(BXwIWR{9|=qBQ7nk^7A~+C{eo0l zX`D56p`QIRJ~-J9i1N^uq~5p)>czNFp{~41Zo>4 zk6Go5Y|FAyCXs@k&Gbf0WGqB(lV)O6sf;!u+d`CWjJedM(wd?^ehCaw6L3m3E(qjP zq-h3HSyXB|me8aqPe;cCql4R?uXbT4l!f(zd=x1W%=ZkTgu}Lbp1mYaJl@F5vX1L=x3f6?!e0 zFiiv+50#3?N4oLt&V6gh%#QV}a|@x?wuH{9R9`DpE0_z( z!VXJ==pZ#R+X}Uzp{=49Ge*x_*klz!<7modvC5yk6ZO|6`5_MBHr8`Wq$>#H6vfPn zdy==npx|0ovnh&C$!)iTCMmV0ii|{IJ2Z0|nZ*!^insvu^N{ytY@hiyW1ct)dNLA7 zG9ql17v-j;CQmVXa5M6X1UUB$lvN>_s)9A-PNzU;>Q@3ke*Q$`AtIF2(rC_)QfbJw zfTX~e(xy1W3x3WZmDKy?%&#eJ1Q4Y{gViB}xjaCUGCC|ATWV#9oVdE>LZ(QnDsul^ZI# z!;1fEnA(~4Naj4r+PT(D({mvrG!v>Okjbr7vAY442Wtyuik^Y=7a~vyTcxfpR!(OR z-?K8w^*1VLB2>B)wU6M<%vLBB|FEhsS{~JMrp%%Mx|5QLC{}&wb#Nl+81Bn%)JPDf zl!1IpP`9vFp6S=4dKnQy)rMMAtOyh?u>e-(BT+{Lu|*0&y5F)M(nThsl$YML%}umC+gdDr7tCzT8hb6A4;2z>5FS zm;6+FZ(qD{6UL$7qH~T^E6EnlPjWD5RK&zktzHIad>}FVK&A{Qic@Wk|*~&j6eBsgvdJ}<8G1+N83@S zNF_Q;ch9J_h_b!7QM7Ye&jwxbJKzWxlBm*A(rvSMbh$-c>CAs+^i7Kc zc{8P28#iZ&684fy4DP^*vO z9C|LkRqy*@bmZL(H0+Xzlf29bE~}w@lkQR^O#}G3j4!ZCEvbl()@hu3HB}FQOO1i3Y zb5<>OY|+)G`=;zo3L*kq3vZ1Z=kH9^Y4r0?GX~&|&Ywc56aAaw^dDD%qtss4oUM<^ zsg4_b@6{ftWEr_8eSnVIZj>r46HGNnCcR%)Yj_hbkBPb>`rh|Jmw?8WV(?4M85TIm z@ziW?bAga}>8ds`e@vFf3j?oIi`rus!{K~w>pHC;EK!VDTM(lB+FUF^BR~m(n{) z5*@5w53bf7ija<>@ogYfpoa4>Q%LBkRSztCe5+lciE#&{%f(yA^DT@13w9X_1%vGQUqWY(N z+B1wM>PdvQ7!npB7rRKU;E5*^p2+3!EkUCEIW-Fx`f-9OU7-d0mn{1>@rR|d zKc}+`^ED+j-5cSloYqi$JsX&UnD4%s)UI&Ri8t%DK=hgh=Yupjy|}9AGb^w2FRxja?7&`$Y80r7JyYZ(olWyBl9?=N&O+ zKGUd|DG;B{aO_{*#WALuzwi1xIs9LvWT{}}rW$5eA^8%}c`RP-isr?a*x!>i zd;ZLa+GnY*ujZ2I_9gtW;jLK7dT`LNUT#H#TjdE%x#7<~M&^{H6SkBLBwf+H;2+7v zzP1~0C@^2aZLY9iww-W~C}d>+;`W~LGWKA`@%4QF!WY5*uuL5&p2y$5$*;BmgK&5Y zoL5w-K8i~=uBaj>z+nY8d_RSu`Y>HD#v$Pv4kW;X-4=n4R1{c!(lPvg62OYwA6L-&uJ#f0d3-fn7~m)7-@*+W!Ge6S^X$xK=;~QWJjBa z)vEHJwX8b)klDWIHDE;5pp7W#cNUL4mx!=U1UDQ3s>lKT>C5-xVjmm=Dk=wtWxWjV zA9Gh6PU{G$F#b+zq2W%!hMvy+XiE;2%C4F_0>5FVQJ{fd&n}(1>^YTX;ao?;6>z2% z%OI=luZoc*Qm}$EgM+GgX>;QM#vuzFkBmc_*`$M?v!m7h*VU}|s_}S+m{fKAhOIDH z=kIx{r3b3a`0iLmAgH&6yKF!Yg}-&`uf8-%%K^wJBdk|S>v*ju(wyy~u0wsbDMxIb zqrVlp!2pBpI)AXZUVNbijNKPJp~Q(}n@#Z{b{6SOs<$J#f-9(2fOEC6jLlKs+n{p+ zUXy!Y`s2H{1^(gM0q^-r7^jWpCra-7>-X++14_Boe8bPb2yYPLZ{$PWK_Zb>a3iMF zHm~9FptB#v5wQzo9PDE7l&lAkjD5$yftl9;k5V=0?cETtinS?114x5HH6cn!vwpvq zBvM8ny&P+fHN8;Khc`PXSO~OF$kr=yAh6p9BYeD{o_Y4*?&$R* z*o<9&y$n3K&@z;9WM-D%<$Pw(H(1zqk|YcaQ_ab63RUsB`8Gq95a;Y3fkHG0Cj ze>i|QLO9MzjoWj~)!=5hz^L9+8tYHayn^|$!G9o&_`wiU?B#{`*XI_9&^y^s>TZhp zuQx%8Q?Ds@#~d)rA@r?)n%GuR|5#Gt0{1`O2kJ1BPp8hD>W_F_F0(f{wv;^%N?r2{ zek(XP4!MONEDf2LobcPQmb;|t*ir2Amu5J_YTpHcu^C2&&poX3AET^4#6f}rEV`T2eQ`@YTU+GF`mE1# zQ-(l7Xx4)GLNo#boV0cGLh!N0U3s;JJ5OkLH2y*qpKz0g$O&g#gG@M%!5%r`uX@L9Y{9MPL+$ zN)Sp#D4=p31)uE=8^7a6#SngBG{(ZO62wIiWvWVeK_3Z=SX2tW=kn%_N`Jnjy<=lM ziQ=p)i>&AN`UDme{6mzIkUX+Hp^h;Nkx3KXm=TAa!CC09+7&qxZx&_k{2+cdT1VC} zQ46isdEuVnbN30N*`MyZ0ciZXk|GY!*Fx>v5{kh2_3%aOvpa?Q`9Fy3j{Y-TvUbyOv}f z@!oY%7G13U7Cx(%)$4Qat}ED}yj$5eDDgrLR8v%v=9l22w77mKQ%N%C9B4Cc0XM$0 zb_f=nb5dJo?)}_+?fn(F6YptQ+UEX#ol^8+$5H1J#JeV`gx~SNV4=bl^pv1npxo!d zz0HraI4g*4F2m-P`p86PqWZ#}-elop=KI&;_2A#^J zV#r=Vq6j;22AUkrafMeZux-R0{iQH@9%dY@u-C>Zv7j*zob~8aT1LkwvdA{wc>2EL zQjpB0(x9U;^0UB}ODz~zeV~sZL;kwu-h^2hkJAg3sI2+oq^xjyR$bP}#UUUAJTZI3 zcR0r|X7YqLTAjxbkTC1d-KFV~4NZU?GTP|{H121y&O*+)-g5rzXlp-PG| z4UpjL$3K&38|5&7_>hQo3>20)CP3#KbvuvNMwMWG4CI??B0%qh@;Sujxzk-zo>f&x zh;|s+IphSzv7ZFo)46AU-)D~Kcl(e;i~mz;dz+=XTq#Wt%JW|U{Ypz<3 z{x*YB<1_qt_Bn}Z)2&25EwEM+*G|1Y??6O2_`|#RmI^X+|Kkni5$$b~CM9RN$23Mz zi!5S71!V@u@4=v3IaHL!?I%K~v^1?2c+4`E1R?={dk@+4)R8~{`^|@`?g1SlJlzAs zlnxkoNIID9-NwY<0134hPUPztGIx}-jK(>siNyJj%9YzU4oVu<0YEMk+7B3-X(Ygh zF=RDoy)N{cZp;W2!{8>h>M9xISrsytL4l}cKh7Dw2-ISW#l{E>W@fAqAJ54ZrV=}K?~hr=2cI7}9Q z4uX_N0Elr+AO<;35m+I{BvB88!~z^yhUiuA=L}6Nci<(vn^v=QtQ3$uR0VvOttg<> zK{ugu##uPEQG7&(5-*Wl&oFb^Q9adh0<9Eg%ib8boZ zixP)Z^Yza)5>7!9i9P)U(DCStEYTXFf>{k!CE+wXAQf~vw23rKn$tq>PZTDzFjj$y zp3Xqc^pNqGvl1Hl5DR>_5J)t(rg?&KLTlni7ekqVS^@1xP{5+qLz<48;~bz;hK%rT z>TGj~6`_^kzSa(37NqJz+W+c6oXmoF`bo}cb>Ry9Or%SpKD`p9;bB;*ixaamljZgx zgg((oQa~HoP~oR5;Tu1LQ39J?UNftdKpjxSM}qLD*?|3^kFm)od{Pk51=Ze&pihmO zHYt`h$|@zSgU~8fOukMH{`QFpTbHE@j;q*1neJXa6PLXBh1W>xQ$3HMFn<^>G#eDYZvSVNu7Uu4j$W+no1F_e&;7i zR3VvC(0T?DY*aV_SWxWoOZ4Pb67jh0Bm9iS)r%apTyvbS7wIIP*LW5%6=D97j8 z05?l}X2Lq*aCb6pE z(mQpdSccoVc)Q zDDy9<_9kHgA9NOp3XYCYAx@D3AtMj!vU)$-dyv`GG@dfV7Arv7P!d>tQk8;;kqmie zK!~1KBf^ihx7UbbWA@0)KPF`gqAXGgU9S{e5CCjP4b`*OKmWnXlCp_SO0#|}cp!p5 z>QLV-tk&6qEbFSDm1Jr1j#h3D)EVg<#B?kct;0|eyjP9}TKoaj>_eooWXuf6M=E1) z*PvjvXikA+gkQHQbzJ&N2N4XDTE0Y9&QOl0_E4pi5osg zo1fOdlmN*^O>bcWG&A`xojK^kKw{Jjp*##!mla?boYOMk~Zium*F{k zIyB3BBheV>>5n4SqM=14Sdu$Mn8Y|A7IB|7{$gsx5g4POeaMh|VNP`8hHM#1Y7YXc zWIQ;`nEoJ3b)AlE`JPJyKnNG2XFy&+0ZEE>1o>RKnZuGGX225`l)LB`THDlw@+TYL zwvuGFLc&Xu3%HayDJ#ozMfoBxU8%{uKUp>2tZA!*Zpug_%tk@8{1+4Z=->cqmCg2Y zo{7YpNZYq4Vh~rtsJ*(1f#bXFANV%lO%!5$L=?Wyo;tL_^Dq439tZ zT%aO(owSNCJoZL`)0WefEC;X^I>w(W`a{S|YC$PSsR|0jOd=B9*Sir=LF!FL9Dl1c zV^kxe#Ou`9iz$x)$IEHm_6_Zs8Rp+FwbBRgEmb*&F@i_|Ceuf{r%DJ}UwYdpjOSw% z$-z!@O786jd(J#^LZr385yWQTXc-sF>gQvs0s}85p4enuq_Qvw*yCZ8Q&%*+6u-UY z30RTi(A+7bF&>TX_fJhKHVJo7z9591T~#1}(r~A`LhEky>%idFha^$@Qc`udVrZ4J z=WNNU&MV0Z^kB||fftLBG$h(y+RE?Sc#Z?nv?eV?8s5_8yc0cB((&kJ1A~u%Y&QvP zNe*W=PMx|W=s8E!lYA|qFpsf!5B^ruG7r1P{0=-9#!Q8G&{Okca8w2_Nusl6J}bx9EfL^K7Q!!(r&xnZc#txr;rJOd!mc@#YZS^4Hv$ zKyVOHWW`U@(#EHF^z&2>6|5K(V-KZ@NM!n4>WH5O%@8gUd2gLSN2{ zwivzRp-HCTLushgsKuax=1#4W{qp8K34ku34h$S-wgH0+-A2}uReLhM;6&Hw#+EHL~flUPMwF^LNi|}5-O}M<%`GV1P{KPw(nv4+~es{ zu^?)p0xufi88urTCK!ZK+&uKb6U<8KLUW;THhL&G+R(Re2zJTk^#az7D=45J$lC>= ztDuL3xfA!YNw~ZSStS*M4@kp=$#-!ySm>YOM7gw!3!P>*>ENI2p!Sbtw*_9&Z4o!HtXk)XGa%<$-J2V8wEot|+{yNAgR)rA{6J`m> zt5C`wJ9fPyWKK}cC)eD#X%lgQh`X0k@%3r+3pjZ2=3e$sauYbOO`~6q=JC6qDeBSD zNNnUd=2YwiPR^d4=&W{O+Vg!<5+{EB*8$;b1VZ_@Vmy#B<2=s($&-|ZC9gIaM#!2K z=pHT+gB3>4$*H8+R@#nNu#Ucl=Vb91LC%(hiJe}FjcJndt@231@m=1AMb*Y8wKgc3 z#tiIf+p&_M7;q|@L{bh7e6CNine^t*#9@{>Nhz(2u9?`1^E9-AUSAbnCQ3SEpK&T} z26N@W2z=Se-sjAEoU0yl`ZTj}fhxS)ugMHJN8=nfuk7{$TiW$E@b~0H*{WyS?6M|` z{Ne<6`&$Fg>KlMOyUEV=^={tQ?=J=*RrfBGzk6u(m*>-oG4I%xmRFELfWA}gpMJIt zRb9aX3-|gAQ4qw8SRVW~UGimpZk>{!mCablU&g_;&J$$k0Dw{B-&FHv?Qj6^wF6cA z+qn6;a@Z>+rCoSfi{ttQj!hQ;Q!pecf7Jj^8y}?UI<1NaZZQA9eTn%D88ecc8IS;4uKpN?=LzM4IPcza9dfSpLFx1&&4*H^QU5Dtulv31#~z`7@vj9`9OrUK z_)h0}vu@vXzN?ob(ERMXa*Oi<=6UHAvsMA$j$o$(0tpS?ZA9RklCjJ)Po0QBbgv5?opk+?(>!*56Qmb6M z@ayf#L>vk%y?-BH-y(R;5nb!&a~U-5J?}@uN+=I%fBjlr+`-#-r?jynFV{_XLV?~F z&uHMFP8!5t+@Du{chz~j&KkwPtY+;O*?;6_>nudNZjeQuRKKFU_4(~z{D+gDWq$4^~G`M9+#>ZoXqMas{dBYUffwplOGSS(0`w+d5!K zP3m%@XFI0Xy~G7kIq;^K@ur#aWGVXll82AHMuWZFTtCq8Z`{?T91#BEEOn*T9TR)< zdWgZ=x_54STn6*~S=IcD&wJv-6&xhOWzL)HJ^xj+`yl@f)Q7Nhmp#r|2M}+1p{s|? z>qcLp%Fh3m;5p8o`S(UyP9^Bzc}3}Gz_c!hdgiY3gQ=Yd93`N6FQ?nDr0)hsZ|Itn z)AFqWSxvcW>TP!K(seS(I1D(9`5MSSKh7bFaDn&-`DIBa|3KBxGk0LD!9LllM6j1* z;8if%n>O>$8v5DIZWSBzI3H`z9@%IsQk})%gKL?0WnZAA>`{ilEGL`8bkeH8C{;o6 zPzTee{fpmOb4}aaY(9yRtBmS1K&}%@T*0u3ta>+y^as@60Po);tg+jM@%SE5 z+R<+C&@5Wv$g%2rMlVz)Cjh{RNURUPshI}*MCLMgMLHRHtSRcOsw}W3Ui$Nw^6I2s zpt?%bmydLi()5?NuvYDP2j`di+B$_?pMni%Pq9fW(h3aQCO2I=Uq3%hYGOADafaIG zkV`n&>(GyFWdYCvNxNJcz`w1LmqLv#jT(*23#mfFRM~+(ec@eG#Pbb?A;2^9gCPRc za(FWEu7)GUQhk=ei9T_swH08UWaS9V7{uKDPb=oHE~ooz%%Mkch%UIBS^OXTz*$@9 zP)!zK?^4!cOm0{^PDaU+j7!5$$0{8uT}C*L)3LqQY*rJ99VNi^(iljrhVb#=El3ME zKdFkQQPm=9lJ#*yDJ>EsgaGpa4F|nrvY$404dfP;0~rBm(2A8v^2I#W_`aU&8U>f% zfq|4?^OrG6%SNu%kZ;opvZPmN#!EMFcp*dedYb%8$<;|bnEsttP~}zJOir*$i6NrC zpUIC@32W4j`l`iqe=2OLV&zE)sF|M)0sgAHat2bqx!@;j4z0ngEgt8Bwp1j2QE0>* zM>Pj(ZNO+PtoSFr*5q*8k^_JIlnvt4UWARgqtW413SVEDzEUMMzhJ*9YZq^SC4UNR zjKTWK<(8Nf>31gZwlPcEUX0(A&e>8*wC$P8JtZD=Nf8B}IR2TLdzEIBdyyK=c$;AO zx+i+)6l<9H!!NZp&GFSrlYR}S8FfYNQ-nKeIAzypdy7*E-QVtk8O!AN*&mB_jtkAyM`|cG%e!XMMc6-G zd^DqDQNe$MYTw7SrcZe843w#!wG}#!?vu2Eg*5QDmsuMSQ59y|%}(5X9go@ZQOcmC zaV)t9)W;5-@89?n=THsUQI8?hlzsS>efyF9E9i{hMc4UI6pKEAD!Ns5rAI{pwKmEQ zbh}f#iKK8FEGKpY&*L8?HI}BGl6Q)xKdD4Gy z`Se;+y^-F~l2bRK3Uu}FYf)M%a_MHe0;M{6Lh>kS{gY)^@UY;*YhMVI1oc)Mhg)As z3UmPRFhSBG(2|Qd>(Ut+h_2o1ae*jcTV7<$rW7}#yC<8-tR4oMPhW2A}qL{g)-U|#rW(Ey3Z3=STSf&Ao~j9?Wp66i91sSkt{D9V(khH z$gRY}%kn5k&MkY^&^k4=vRgH0VBo@)-!7X9{PQ4^*`j5bTe)QvscO8M%Wu-=a1 z;Qn6OzFymR_}u^X=<&yVTqOve+yiCTOmCnEwz_QgT3!<=RZ!qH%)9; zv@Q*00e%VIyp!1KWjgL2-hbl!1!}&l*gkqv@%C|9y3T5RziM^zb`c}UpR+}5VFB`tHC0Mf2X3qadr&tR!Fsp`X ztH2{z6?$f~z%G~JG1rR|T5+j>t}A=-yvGdm+kEkkCBFR?p{l3X^@)irsqZ6&noS;d_?2`r7AJv}G((`J8ZP{2dC}<3W%Q z`vzQhEb}hZrMx$%GyX+vz-orgSkoa*O?RlHYP{%r3s&vs`s-V~%OUK9KB;<(mVKB0 zj^6LX*X`s??Z>!8I9Ly`fe-VjgHtp{KIiW=oV%QmGYX8WMJr-dST2WcpU$;y&r>)?HRws z1{OtA)h*+?YOCM|6m=8r6RXrB46n_wr%q+*&*v>Z=&2K>zz_3d4mbq^yu5a})X)!U z6cDi0Qg}qd^^=8#QUyrGBDLY+2E;2A&*HsE8+$V%(pnP7^^v3tOrwk@!mCU& zg{@zHY+x^tpHiWQE?)}ye_mTI%zhLaDZYO6M4^44jIl}mm7Rv-Z-~vf-D!K_Yh?`T zbjD_HH}e6E^EDi33_wNBw=EV0%gh%&%%u;r=Gp89#uA(rlE&i({hWob7~A)zW^vyO zV9W8N&`-1-1W~}nmfmtIHGR|PU(J*5I(~+@jPxL;FuYpQep~{(OI4A{!Pli_h8s4uo<^^0YxT!Dnn{&88SZSlRKN+tM$YLGJl+8qUPZlA_YNC@CdZq( za`(3C?;*g;xMc}sDg?DeWgj*~`PhrXjoh#M;TwkT76Pt_7XD=2zoiQs*rfwwWP0`{ zM$6U-pSx*{NDH);vLe^jEuKu~O*= zFn6fzy<%yZpYfJ64FOucl`PIH&ckf7*rJU^bz?+x^by6fvdPj0q}|!J!du4PK?L6% z1&B~q-&TpsHuH}`;@+PqmX@Jxt3hP%>9TzCK*qP@UtQm(vM(3dE!$VL{9&8Tcx_)D8p(10x(*(vq-f0Z5?`E2n_UF!MpAgaI|sz!LRU~W&KC-P+B4)X zGjWZqPcDP-ZG5$F4^a6vKTq!uJex#sAkObPlBmFO!*Mf@jefTE`YN0l3I5Vn7cGs5 z-&79R{q&Apw4i^++DR6OZ{Bo0SN!pty+-BAnk_QTI=mnj@$1jYQJd{e?WxiC3sz8D zB)&>=)zNWO?D!()t@)_2@(4OT^-Bftcb(ZlmDyE&>lk(jeNDhmCw^)o6c1(EK=uhT zrJ;&?uE}y=ofKVkEh?h*Bj?n$ifrccsH>`keI?v(iA~>y`Qh_Rcxzz1W#eV+JZfbB z7rT}FAz|rI^-qm?=^=4@bjz?G{O(NBsl{?&Z%#GjRyC@mZQMFktV477xHt|1i1#9m zYt8#b{c7xvcpNa6qFYkP;Y)*Tx{Lh!ajRS^I5K9Om6&I|WLtnw@-_pLa_)FfhDd^N zd!xV3*kw@d&Pt%%G{)cedm`teN%%;IKuY0=(KLQjz%O3#Z<#}u3Lq_cj`g2W&*>xT zK`Tv3679SrS*zCp$D=TH!uZNvQ8(BW`*Amkxaf+;t-@ebT!E8$>hb`_)lySfXva~q zA#Cg?^&*}D7CAO&nL4O5@m1i$@NW|25B??l++hWqY|qgZY4#%9BZc=wU)m>4RNTS` z$)&0b2f78#xZ9PrL%gW-9Op2*PAFaH10!JTXwf33Vp zYClft5-2}gJ%*v%IF=}~$LY<}&RnG##x)Q)`rXQv!p^LX5O}*q7FSlKhFH7F`I&7_LiMggpg- zeRA-oi$hFynPjOpH(&chz`}7`Xkh1-jZc)rhGn(0acmpBZFZi2h%vb)b_l2Fy4WH# zbCdh>csBQ-hD)NQwJzv+x^Z0`C}SGfxwEaLulRD!w5gSM)rRZ*pdZSzsv)g8KeQir zK7ShG8%HDsE(#R@$I*S)^+jjSf(AX|fCl}DHsy{;H2wPOQILoHrq~~WEs9OQeb8~L z71Nr^ijyopjuF%LrxP`_J?|fJLB-Bkh$*=18Aaz(KeZ|QDO_O;Wgc*(I)d1;9;toJ@eG|Qj&zq zivsVbh$<-%KY37=Vl6`#0jcP&3aVhpBkv09UN1-{_Vstg7Otush)I63Tq|qa%VT~+ zt_d5%j2;H)(^Bgh5|p+Bbuuy8r?Iso6?laqn}xWk~MlpS1H>Lgmv)`B?n4C@3N zCyUc;m8*<@k+;K)u&BR3!|7JhZ0%3Hc2S`no6w0nAAQC=7@-|}KBh8cmX3h9KbG3j zOB*ic5q>$!bQaR_6hwa|Rl|5mE7gZQ7{j58_UCKIQ}n$qU*}cYpRfN2xKTd*iBeEl zYQkN8funuU?&K>jt%a@P9` z1sq=&V!k2CWn^Qt*p5tn(QABC^$HcT(%z`XWvz(Jpk0n`g3JZ*p{i{W3us~GbI5z? z1Dw%!7Yk_A1cX(O2oC`|{R1=LYCbpx*UFZG-uh~O!!2-#*KdywRk!HsS!h&)ysif8(V)3Lk%w3iz7Gr7^=ZL{rkcT1wJ z)s3rDpU%)OXs&WFdYo^Xz}#5s)H_Nb#icd6a0)5D88oWYC2hQXazhkOl?0JsPT)^( zZbzGn^m5VkbwtbI10N?F%1y|cYq(}_CoHv##dl?%-9J5TjGU^E0EV z9(V^t$w2;b>NKfH8Rne&mntNBWIi@u^sKbv6sw_pL278&S^b;QnoxO0mI>5v1SOj@ zM|P3F=?>#?4dEN*-hJL{dkTB^+hcd${s>L^9-S0*jlWaF{-5uj>5`hbp*os^ey!*eC(e5H5r1ATT7cDk!#8Dm%bTjxs^OiKpE0Y_;j0!xfp!a34YBZP zH)01m<2b7Dn@Oxk_oVKlBQ%hl8U9Aso1_l;D?jbQkS#t>A~6c@?bUVz(ss3&KG^ai zmePeU3uD^H*b>CmNP9HDj;A%2?&t_Y93x7{cBf2T$u{ZE|HFaZGEWB~wh|5w4J zi?f}LfwP63?SH}}+BQxb9P!`TdVNuD)2{mz5B2OWCu?i!zLNV`+zz7C#xFGD(s8ka ztfKZ1^YyMA9~(Dkwts_2sM5+dD%#sMONgA=|B}s`JG1vD9JkSp?`EZ!gJ%y2#u_VV4xIxSrJ}Wu zPYEKNy>W~wcMj(-x=}gNR4)m8{B4^oyC!hYs34+YeeL?Sd9YDHkM0p*+_&y1KQQuM zl#l1v8yet()F>B+lXyOwvp_94vp{vj+q!->jie9hEf^VM-IaFNyF5(eRQ=T(gp47M-^EjAq!kL42oc&H~nZ{Hq<4Ck!0 z9tp}$=4oUSW+?^|Zq@Za_b%ZMirErKR30%bo(jcT&3gqbHzpI>s8ZT&l!V4-@UuN& zik5gA;e~L59V!j6w$-}$FT>04s1$B;FpGM&gf1Lk7+NhT9$VBdeK)bw$v?biLgk*_ zv_Mz2G&Z(Pl9DB%d{#%)^Js;(xA<54ZCY3QCh3}u;tW#4@SJZk(@S=qvO8bDY`H@K z6Dd`2V?LW;?QH$NP3(SlJ2E!5P+2#v+fs~O49^)NtaKLB=w3n(wp9DP@SPFeBHto1 zZr64`X1=W&ogP0nGWG(($p`3NO@tqPVIoLy+@dbsc2;)$^zeQ4;JtQM{63AoO8A%9 zXTI+q-xBD%e-3iKuZ+I-Z{51I>;SBAE}yqHid_0~r`iQFz$`da&z{d6k_wMJp5u*y z`@Ea;GUmjDC(rI|tpGC_*EU#X6v&{QL_P6;g1R@d_WD=58`aGcUfqLYj3{+?6hV#D0DvHuuWbM-)u+7 z0<~Qov)raW4iYv>bac3mp9&kcmi;|995&1)>0FWhEWR zB%Z61(uzHCie1Wyz0;8bW6rmdo(1bV7UYz=Fa((C82qRalx{K$CHa;O5cLMu&&xqX z&vlFU8Uv>^6iyB0hwaMBN{7a1XwI`%tP^J%6j5I=LWk&GGG&ldIIc5b=Ki<}cRBIB z#Txruwt1PvHG~c(qWKm`Yqg9Gvf5vXGJh-}piLTp2yj`v2WmzitZ44IFvkBvpP0fk zE_0u(7?y_Wx7%9j;yt>YF!@F=CzGhsG3}UyJETdkYRrP6j1C4ymW=bEa)506TLM?G zeUi6UD5@juJ>Yn&+^R<((vYlFZo=wL=FAX_-?!oX65@~@)Tsu6&87E%XD%?XoZYgI z)QgQGPpU&~MLyH~AQGaRxZ}3+ErFo2+UK8<4~wFy4r0~HQ92V+SX$?Oo$6Kb&U>YO zc6eNKcXOK=3&-8|@@+0S`rQOM`wAs$4ceD)l#{5a^vEG-n63|z35M0_WqdSa#qP;r z?GhlDVD1?d>7k3x8!!TRq+xR}yYWYqGs#|*<=s$>3rcaE9f&zZb|(f zA3A;iW8$lov&9f9xE3ifUu9GevrD}tQyG5g_8T- ztkyKgM+R8E76ds&uFH-x&^H2`CuU7<8+&h{b2q0nrf!HMu=SD@ z#1r5zSyXCwWq6P`_-AC%*%KKeBm~i7CN^5vH1;ud&lgqeDOjJn9+EMah2%5X$6J=d z1(r}bXX&ayez|qnCx|PFbD7Fcz8zkcKv~AbM;*T%qSpA_HQS>m*N)wKJ7!B7|5ZgF z$Ixojp`UtG=LaAtu(pWux!w)siZQtW;_7Fz;}V;uoy zezgNeUVYf5Px`@jPzBE(Ntj*kz**LbpBh?#QxwT#QCg04!YVmsQ_zl9NO05U^!R{~;XmP=jgH&(!l|rW;zX99ex^cwna@UrK_<`nt1Zyep-Z?pY{n`W;mh7O9+Y(CEa>Au7!brTz$2 z;%y! z>?IlG@Tp^If3!1J{ThdvS36zmzk3W}5UOT^TM$3aUT6TMtA(IYOvN+mxQ}cV zDr#45%}v9)L4SXDyX622EyoqH5A}UvbKI$xvcu!0KcK^WY7LISHH}Jg;Sd(Ki;Fq! zX+JrWPrOsSw%cA&+lz;yZcUi%aI4g38YHX1N@lRouNy(aUL{7nZWTvHBXX&qeCa|h z0mz-!xxZIh{H2n2O-sR@5b$~7N?6?@jT4vfdd(n((AXkqfQ;0Q`vZ%USKuVD(Hvhz zs}Ex{(O|6*)F!(90TB{CwS@T~6?kD%ZLKwkpz+~cu^TWdn1nC7Lb{Rqg;9(;R8Sep zza)f9+OVYie9|NJYT$&=UJW4 zHfx+sr-U0&Djm>mPq+ytwY9WD^;^#!R%aB&$?jT}8`{V#UrS%$1>UCOA)aG>qq|b|c*SofQt*eT4;qK=+JMTmI zAei?%o->rSHbR}g*8OxuH*!%3{EHx4V!`3&9NzZ==#S|pve`t#AOK;dt3ycZW9gEG zlz&p(O;S`!CU)GKKv^W3mzQ~uwJ9p;_iP^^1JGHrSE|k=+Z43pDUus(Gg1YteBKw0 z;otV3l4Rx%9+xh8N~-+%$c)e%k*aa7}XYe{Now8NrQ2O?}&Snd{ z)bOj*LA0)Z&K~O}ybh9mTc`K&{FmhD*2J9$mf!;u{^6x__sZf)om{6e+XUlBGcPOT zAj%J@fv>K;GdH`>uCtok6-0B_>lqBZ!O&!c>1bE22o@Ju)S<*piW)}PATGxSy^xSu z=wa3M3Yy}zRs>2X+}c#qB3j-_3qzwS&4ENxNiDHz4{5}fe$Ke_Aswnt=YG7>BlQY*%c417l{kYI#bERnv_I@r;&^2mTbm$y;(47 zwoJ>A`?v1}FT?Hg-=_l>E$M9NctG{R_jG*k1wEOLM|#umH>Ixveb7>$F2VHAOX^z! z&LV?Ah9U}6s!u{&!nf3#UOLJ;YTJZ5?+J*0Iizti30J0ILAJaruA&;Tvg>H@E}gjQ^X656hJQ80BGT#LtuF-bSr8UMDB*?~d<4tVFu-9O98{7y zOl?7Q@ZVsMrbY(x*1Wtw9*6d3^l-huq{UT1D9?#=?uB>LNVnuFl_b&Fi~*3z8)i{)t+RL0^ z)49Yaoe`!G#CHE(#C({L)=Zds`EZp;V>$jwny7Cm1nFQ-?S-c^|G;|zKOvu)Km6lz zc;e8ao@ShGVDxe3Z7zc8@hp$f>FGDE)1~!YuTunI< z+KQJ=PTWu&Ew|%K;aGmWlEC^{CX2cPftW;V51SHXa9A+|@8gMtvH|)7Jec6=)~rBJ z?;QQ$l^=Q4ok}`K6V^|IK|NX*P#RyUf#A{fFLYSa{~?)9-%Q1By&ja+gM-l=8^xT809hKouDp z6+|+t5@_v}RDrlsO#i`HIh+HTISsrHZdV4yk0b#gQqonK4A~T5^!7_G*WUdQ{#qt{ zSe7&*cq0Iiya5C=8P2N+X0~*9Mbh8IP%6nu%iL7#8^S<39t@f%1dTxD69gjH?DX5m zPA_PIrj$O*K2itCz+r$xb~v~z7`bv>b5rEe&mX>$oQ=X&3L!ZUHP*bozb*q=A5<@P zh(;R9Of%deu};y0;ZJ?7`Z^HS)Fwyns2^vZ)3?H=YOBG0(2Gd?R zxDEuhr$^!}cay%11}-4y3qkkXcArupzo$fHvObMk zhHInahd*-`Nw7g4D@wf^En(GUr9t3bs5kHd zg9$twoE+kI-9nRlbrdlB@S(Y{m#z(TillT)b-cnw7X}%vEk=DmC&4h06(;&x6j<({ zbc8)!G(7eAgvDywt*(dm;ISOX+c1*1L8@!L*zh(368Dk-6OgAYs?2!iJL>&IME!)Ss-^jQhIZKc99n$30!FRS z(m{_6)94C$&IU`eiY4N(u#Ew5)?J%cI^j%XtPp#qz|uht$S*3E9* zqHzLZ*hL|v>bh_hR}G`|sZB-bgGo^6h0Gi_P|QA1U@}mW#P=Vu2GXjqDwPz2tt?xV3O1hJ8DF79%ZG_6RJ%&OmHNFGnB_B@0_YKed|i zzMO1+R$;ZcC?5uS#Y}b6^S&*3%C$3Rd8^1mQ7R6HO5oQSH1GO8gf(6ykw44HG#Jnc zR&l+21*8yV8K@M3(fypW_aW}EM@HF409*w&!i@KD=G6L2qZH(rS%JcyPePmP)B^OMz2N2kwB01!n4_4NRKY z#wXY-LNbSmpF1QX*l#Mvm4FO`TbW+EK<;TqiW3l!NtUD3PWU0N^zX`ed~pqV7^Z|6 z%tKT5%v@@ngdr5Q&6rHHod8J^l0T1SzllSGAG`sMe6{juJ`T-5@twiR?Q_-XbgVGA z0jv-pKDlN0Q6f{SL^TeuFJdYWeR4P-MIXo*Rh+V1^R`Gp!Q_#j<$jD9Xwbw7gK!sa zL}dC&ML?R<9Pw8EzN5Fd7p;&k{0Y&(B;hLv_xW#LA^-cm9@>dTD+c`h=#lvCnpaEut4Xbn7Esi zyyUODDU(9(ErSLmuFqd`46^t(L~0#OThqi;eY*q{2E)a1+j-6a z7H`OqcAStSZ`lJ`LY7DEM#4tsSYfcq6GYWBv?bd;I6>patkd`KW{pxkh^HfPB`r>& zU=pB5b|S6S^P!|3t6{Bd9|xKW8?Ls8N?>h2-?*>9xKHT}M6MR~v+L5?EnH7P*1F#u zn3=>2_b=}cPK8U=uN!1^1I5;W1o>g>8pwx86J=f=%d@@U zMg*y03l84_`WLL`tV*m!2L0&}KSN5*Zk-uARVRgQUP~QN-|Zwq7A|Y`lR471RWw!C zHEb0oV!$aPE_$UV8;4|~e)#39Zb{!d7HnIKcI|ZZn9NL)b}eISC<|`DWTf|A+~B!K zA176isw<16tsQdD{ca8KD>nRa`F0c{ST8)dc-iVKaZqKdmFe!1um}v=3t}cvE=k`U zPe=hr{_H!%ejm7@hPNtYmnxF!gGmDz4IjsT<~)t8ePPsT09#@oIrt+g@5uceKIj45q)~P39_?dQ5l4qeLG3LU_((D*2CoLb6lJ`o`D8u zJC8U`Av=sTqB~=K$}Y^th8@sZ_4>V5Cz zwCtuLO%1A4aTTm%BaOC!(ZyJ0GYN<>Wr2!m=+sLuk50QJC(yjD{=Ee8I^ZSnUe7A7 z{OOAPHKKW)@@XgKJvl_hm|OXv8{&rhjX9wmSuoWa9t^z(X9|aob?U^<{Q|=8k1tGv z-rWF{7x%N*^)V*BTNiTLSC=mDW*m-<~GdL47>OD3byaT z4cq2tH($F;4~0FqcB>xJtKVxNYzhPSZSv1zj@2qP-VNV#JnW?0M(Ca26FTm!)nNH` zJ*q2^)Ui{I*9Jc5dg`2+FVMbIk1+wi@$CbBVJR=?&ehd&M_11o$%d0%;|l3_&wE!Y zt&xQ;fgQ{-`(YXP(?`@yGD6KqpKZ*u?w!@+PC7c-3e1Gfo7O;1XkZC_NDrUgg`T9^ zk5v164!KU6?X&Rb2dz`GZN+lihng1Nj`wDf2RrSL$C$pX>xv)ne+w$VJNby@x4mxi zm)eE?KLiy+1IPcsx|&pM>=qahex(!x7l}XZEYzT2$}o#jga``)8wf}?%56~_nj{X! z+(stYJ-8bdnjNwmjHrGVH{RV|*KH|_maO2T39=)Qe?Z|4y#q1f!NLrD?n?GK-j8K8 z0uW6EWEeQEwKIYEGIv2hhkU#w8Yqbm{+S0R+|!%0W`stmHar2r6u91$_vNE&vvd>o$(ZU9K>@As)%?8cu6QKzt#nJ%K z-6QQ-Oz}VSY4*EC7))7aUgwzw{q`HZU4!j{=FUj{;8|F7}p{G#*Dp8cl4|C&A6YN{5srs_!JB$oOZ_S7OUM zbbQqxra+~?v8#rgBB(ZZ*~Z&_)3){JKzvm+YUjRX%^ml8=CMm_Cagqm?eV%Z3B60J z(ZxQ0eQ~*h@zy&_jSpq^0Y7+ni!q%iDUlCuvIm%vm8vxyaZf?Ted!i^;HH!gDg>`u z1;$J8YYUlUN4o+H3AE8?idLv2PNk3fy~YjtYt1vrOJkZGmBRC-7)nnbOP5JP{XBho^#eZ(8Q}NWXloX z>w$i$fRjS9(lBoW+JG}y3(F2!L1#>%5R9nNovJm>PtzrJ>BR1OyLxEW3Ef1qgIL*J zwC1?gAkqvq{Y}1FP|2N|x_K&tbLs*|m19E%xKJx&HS?;8+E17`2`#QZRmD*@=j`i= zev+yTZLI(8lNH6ZF3~Rw~Cu={+_8fQrrk zW@i{0I9dFs7;=l(!*5-p{^vx=JQCR-o^$N-a^q>(fWug;2Y?92wLavPTwrzm@1R6~ zGefpq_M)QN?{HIFA)T6@o<@fJMuTT}T6pF)G4*BAk;BByW*tti z(eRV2E%&dTUvwz4EUH*4OnaJ3G!5}-V6l~vqvHCmI+OaYGLS}B8Djg27ov?BncnYf zvbFs(DnmmlN}0*|dAdD%Na>L*05><0vJp%=%ozR@xR+5XfAl)>EU7?9Ph(}Xnc+z; z-{M+gZ%qG{b|e-DA7QPIDdk-C6Xh!3`W<}6SaW#N+NXIKjU|=cCb1uo;VuD6TOG2F z`ndsQl#DnZ;0wJ*vUi z5O0qFfY0l$8~IthCL7^Py!Si9=Cqu>SDWnMNbo)&6SQ|Wc`s;T92obomnsDj8NiBJeFE`L*cXFeiXbRY%hyB zS=-+!@(7g=<$SV5U*@k5eXFfjG(ij^jAG8f%p@3gp{Zrk;}XB)*0IUGy(8tupU`a8 zQATC`D+`iWul28C=(eX};?rqo{=Qv)K_3wpC)N$M(!%6+3Sclz-s`hGs_c8`Yy>SA z|Kt*hEn{Wo^aNU05_`xswVjgvF(9<`PAw^r3#~ZkPDM{$urgv|B+raSAJ;P`N-5~+ zc0y~-&mB(!FED&@9d2)-yTGEe;!2p;C#m_{6(cBD{rPc%4o=@wTcHQlAs336ssRFa zYtcMY4&MzaVPH|akY@B+c}*kEoWP3nLALF`n@g>S8S1AC5B}m6H_kl}+@8;!6p;!=u63{;2?vZTk4&OXCxU-NB$XTs~+~)ih?#Vl2dT-*>5w{t41`$;Yi%oIM-vMG%6Fv~1T6#9wDj1S3)v$AF z0b5lh6}e++M%z`eOr?Vg=frrsV7Tl*KA>$wzS;pl2&PF9yoLI{P^ zr-(?Nq9#a*krGF$2zXm`p#F`mP7okgUow{9Fd^mCSX}ja9N6oG06@N-Aiz{(5rPk1 z4A+UT)}N+0F}DhQp>jPT7mMl$wVq5u!3wO9^cJR&02p;E_56Vs3CB|%P+%1}(9~3K$AnTrf;vTGhGaqruez_Yii8$km4xDe=*(!CpkdWW;ml8N`lhkX_}Y zfxx9puF|1bFiS9UWpkI)HP1Y991&NUZVdh*p^vxLO%$chhT)seTN?dC);degKHE-9 zI3?o_!(Zdb#o}L-Dnleh#mb(#Tm%&mc#o6rCcUJ}x8fI@6LEusJ@LmzEiD z-jLVLqF%}9IIXfN)BiWXTwK`^R2Ftn%GrE^?V9|EkBU9Lnp)S8#vA}GRIfCVDZ5W` zPDasSTf94vhZ0ey5$}~*T70&{^&9Y8)xI6pJ{mmfdyZH<)SngqVH)wp5gu}h3N3F4 z($I>@Z2&3bWLFc&Wuwt{vh?BFD*!j%cEs9F8te?-)rv{N-!jTs2>I08iU-k7V(~gR zR)A^;0Y6Xuo*!admMH5hJ|~BcYZ7Fv9aeX&;vQgcVCZud{TQ@2U?eQo>SU1_)Vf0u zTzOwuEKzUIJ0XTIYrKutLc?s`OC_W-BU4B2R1eLMA)UgDDP>F&xp5B6D!B_L5B__2 zOonzxkeByC#t6cJvT00!G$9EfX7>h%6f;0%r&bOa*%>QX6 zL6FZ!gHs6K&Uo|NO*5yD^wea(Kf8)<9{3(fd&nox?B|VL8;AE~t8(}}rpWZSxMBUlK-ixh+2CoYTKivX;tzDRr^9VdAW7yw zPqVbU7^d-;19iVzSwb!zr_Q6WJK@)n7ipZ)>lEe3+L^*kEyUgu-q7E6r+I9TyR@Lk z6bRWsLq6bU7sx}T6r)1Zgw8J6KcBD=5tEh--8X6Tc5eQkv6Mmw9gjDwco(wQiG*Lk zm1e3f$2)d{C6WET2$L*lK@Mn5qZ0IT#L=mm8pZQ7%+^ek>G;GWAjR76l@}tJ0N{W!tuMlJ_WMGJljh(&o0ZXP>*RQH|-rol}3ebpz(_wSpD!PcixO6j- zqZB|g!I=bq$M^V=pNt5}2Hr0Se?XgFJuw1>n@k%fHj@tYQmP3%kxO+fQ1=R*UmrWJOJ0#a&ch97Bc5@{)9Og!;s z1nRVRTU=6D9z@`I=!#`E{HNjD`|0_GZ~LKab{pYMvV|4H5fjkiBA)Q8xAG16B{H2M zlk+bI9wlFr{h;rKj^8-DjXmP8UgPVCIM2T0_)+G!fXM^+I$7w6ph*WGepdRAGM+;* z+GqnG^wJMsAtWMVt~=>7fQUgUoj0XhHp^vYp{pchW>4GTVY1R;zET}bXixM*pCKmE z13C#+Jh|OL$i`@XQ~^vnBRcbH_vDHei~F8t81K{lw~GZZ4y#Aul|s@tr*6o?!~i(U9|K#QJbsUGx<`;A!IMjubb8 zLE}=Z*sT3?&R&pWirVZ?z4$1eID`5O24V3+N)kBAPa~xeh0q`lj=_G!0@LPs zgdBu_Us7eMOO3b}3O`b~92rU$^$k9UYUGzskX}#eu+P6lGDz&poZ128=gT0UU~Jd^ zYPg&$T%T2bz}T9BCSO6uBOXWyu;Bo0-fWD*4uzC_M@ZtyM{3{i--Ri_xy=&K$ZxaR1<6T+C4rXPc^o z%``d=iW0=i*tQ@g=$~a{GGNt06Ng==LGw2z`NPAuBv#~@d2=GGAY>#R0&_?5l#ITX zOl9ZddDv6c=0gp?V%E+5-S7wxUje!2fO!VJ)02&F#w-V`5>vr;MjPZm7me)G2c8!d z@i-$%X4gIAsbfWM8GvtGX44B51KY)RaTI#mEsm;I41P!66T>^U$e}j%GAr~MXj^C< z*A5-8`W4AM7bQ}BP?Nite+K*8Z#dZwX)DU^AnYTGaTVIk3 z6x00EZ{n}7-lhZYI^Pnp#wsZv+K8;zB8v1VgrSZ4uVLls%-|Av4yKKJ@vYu_(HQiP zdEKr1a5;h(hChtqrC(q@6J%pjczE zJ}irT7f0J$&1zL3d_wvQK^vz5&h9WccHlhS-&e zH!)1b2!u{u_YyF6GExU98yd+InjJ5iy=J3lX#HO2-3mngTw38ZFDh2ZFH}@#=hFaL z4F8%0dbSJu-tv+rYGPJc*26v~<#m?3LD z7sWv?ao1M#WpBon?@9XxLF?@lpg;AO{KbEM$fUm51ci_Rh(}i5-8{IC%ciCm27Gnh zm@IGEFjeV_GFBEX+QzTYrlc{B;Ey#-ByEZh#Jmfl2&o!0PAp~BP)S(gJM2O|@+?y8 zlP_9_UdRZt_zOTSsZmC>mt+e|--Q9+QqV@T2~})7dYUQ z<#fJpqJs^LsVPMB?1@X+H#8+0OBA4mDs*4v5p#Ji>%5Jc^sPpX zVv#g;cwPxBBr`ohy)Oszm%S++6>Piny0@;{)Y-qkXUyd^^eQ@lp%aCRuHl17V zAF4%VdyUtYRhk8QBcx3mS1S52eWuQX3PMp-B2n+pKx)d!Z`<>67hLh5aCG^UVMaNb ziFZ-jDk|1$Hm9f}YFDaswKtkYa}d=?mu_8iH$%&Os=YJ=*=H?YfD|?(58UIck1muH7Wyox}&oQfqNfhDs3oA>S1@j7J?o?UL0Z2JO zR8WMVMT2BP`Ics>B^d zZ3dJ1(aA?cm4K|vZXjxR%AZ(-+=}-Ln7cM=6!S_(4k72)xl3$5KHY@^}l|*EAEhBSZA$y4jFeupAFW z#Y9|756^h17s7tH*Ur31G33^>eA?zj{ksFQF{i&=kgwBfAAhb`RqQHhWNBXbprYA| zLv__GAQuyes=B0^^v9Z1Z6{)ZOy52Vnj1;%iz=?1t$i=a7=U2K&=LrJ)ohR~3I#OE zm}=+$=sKtFP@pYK$F}X{#I|kQwoh!^c5-6dwr$(C*{L3*A8vKkKiDs8&&6+cHJ&&p zc#~}D8=W(}p^Osu8P>VJcdZ%@rUNBr(%+kGhFw{t4cMoYweYJyIc4PJgvbn`Av7Qy zJsQEh`cK(_Fo&^VclTvsVm3{~L*0t;=UoLJc`?{a$K>7|~)jx+< z;tR@{u09`H?i!e0fB8_fa!GVK6v*qdV)upOemitmUfx}`k0k?z-1UZlX&q*x8rX!% zyVy%IUM0oz8xrv2A(O~3P#!&x67Zkm20%!stfR99UK<%LtSO-$LmX2lPpYOjTFMAu zg`fJJ!UNZ^f5OjaXbM?b{9qm)yHYn}&NvQOw|xH(x2l40C5TO%Ux*Y)ptE$otj;D=-LO%Lc z4zB1U!d|~ryZF*^f0(p3p+q<1;!H4X7s)jb^cuAKK5n@4YV_|Xfe~96$H0D%rc2(qL>583cDUZeAZBdi zlS{M|TUDc^uD(wz>WdD|=+XZY(_UA?lgbXJVSC7#G~QB;)`r^Z$lHqz4W%y$j@PM3 zU6;VbyckP%LlIGlCQ8VO8xctVH2eAjh$wiy37Z7D7YJ7!4#DS=3lEQtMdJ<{Rp!x; zU-{CnuiimTvCYVmKj4Xo%|^tkIVE8Z9Hz?jfm=~ld07xdF;Ek5T#qNBaZf^`e)7Pe z4)S8A;!{KPx-~O;AH@7&1jomLK~8t~GBTAV8A+j~wzHTdpO43tgS9U%fa-I7bm)h4 zeXKss0m7(10A*VEHFdL%)){du=Jj3q7;S`4-Ja`bxuG^Y!;SFscRl1|oBa4``?9*b zKAcwie%7L3%e_s-$FkjuZ#(AO+}ID)%j~a;9lfL-pRsN6plc=2<>}po+ep>L*-gcr zk^`Fo=1I{d2fv;18uQi5=qaAgaLaFXZ}MfQUKpS;bT=4&4RNtx)FT^J_Om?-eq+q+ zCV!`i`^h-$o6wu;`AmuSIC!H&Px=bzj0m~ECL_ptm}Z07{lQV}m+p$XR2*I}8eGvU^%~gOE8c&wyReK54d-c~ zv8g_zIVo=43n$}??se8KaHR!t{)fh7W&Ku1C(4UvL_gFAlg3+|;dU&i9=SoYOn}N8 zsQEpgQgS(0);#wh0w(yL*fx&+^i{N7=Bo6sPtr=}0`bhIN$Kn?+m>R-Z+-QZGc%=W zjkZm^!n>U!iitV527! zC#UhkQRV?Ak7cYFTbBjywTo%>nB@ef$U!Hf{&mv@VBjg8*EiDKb}e^VnuP9K*6=-3 zelAw!B$Eh0R5jSi&X1v9+b@eWF|cR8*D6!{hOc*J&1|-TN@W->jR&_%E79o=&=K|u z)##YxF@NloFd(f&R7dl9uuZNoAf{KiEoQL0+PKTL+J%-SLM&W4xILr{ic-X$p&F%- zGOX3M`GF~lFEULDgU8ZnRNmMgq>nCw2=>`Q%L0LLgB2^Hy09H?tl3W^bk(r~ck?EqEB zNEQh-A5**{i6~w@$lC~R4D0(T9zqvte;4bUOvy)>pj~BVC`J4vuM&515E970OZFd8 zi%z8w?;$}4!WY@t_UR zfK*Hq2}K?%WJS(n;4aj~+x3lH7O||WnF0!eBD8}l9>L?`=N3PNw>NlsPP=b@hg_+4 zwB;eHSNe>=JLJOQ>*nS*@n-;u!C@Yj4bc-9f-9Dm)sx8}e(VsOl+FH2i-?%4@ktT9 zRV!H1mRDY|{q0Z_Np5;2ec2VTSwAW+AXRn4A|l2nI@Nt8Y=Z&TPnjx`%@FhF9<|%R z#+}kLjtoz<7I1GWv2~ed!_2^7`@a1FY!E%q7+JsW6b?AiD@Py$=f z8a~`mc6MKH)&-uo47~jJs8^M+_E#w2o){2d4dsmuW8DQWm&D8HU3$qCmiL7>LFH{k z*i=lT#~Oqii6{-^Y{edhdagpp-WSjYXEiP)n^`*GDT=p%$qcmkJMqWEVoV9ByvO|0 z^QqQWSw66>5MuGQXtG*S6IPNjyD0-V0~LK=H#lBr)un5?$|yZ>@AtjKrwYR}e4$TM03TMrfbW{U9f4^zA>Z1gpDNZaguac9 z$Ip5Zj6^<#Q(VgHKH71*oQI%+6$i)xf-y~nRP3csz5kpOj_zu#A^i^EE&dn#I74IO z-zj0VvW?vaE5diRo`4%D$y$9D*r}hpL?`>0?rbj2x%HPmClE~z4G{`OSdvk}rO$_W z*I;-y05QRy8#iy~$=;nuS5D|9xyuDkkSaqcO%kQ9qUOgB51ww9{l)&QWQw_mH?A}x zF*ex*AX}fa*yE7U}jTU5exkf z5fuSW?{7Nz_&{-S#FNKI-bGxFtzsOyd@ z7Sz;+*vJqGN*8wK+apvT1)9u<=#6c<6F(xl#MT!gz6(2q%}S?AS6)?)r^21XHM2sj zl=+kims91Zqx`y`Gr*P;P`K$PAWSEMWEZsY6*W%x>$&NUhbxvofPA0~{Dob$qPzI9 z!p|FtD#9GET-}7Mj*G*R=yPS67a9;1G1ssVZcQM!Nz+uTo7tIGo?67pFSdjwB43Y7 zh$jU{0wLHOs|ry2p9fGHI)vqkmbNE*%QgEl)zgl zrJ<3ONizmrQJk*24xdSK&goR*>nnKQA$31u{tdKa>K3c=Q_}h~9ot?GTTqfVwEgH< zthk#65w}z!-k07+?`C`3pM`Bbxw_lCJG*-W_Cp`vSNFCABV%~mF`M&Tx&1GD)c`cH z!O;tC`i7xf9VTFmqZ?w2_9bTDcb$=QdjgD!F$;vdNTK%c5ooL0 zQX4bm?7ah7=&r+$xnhPnwr@y-4~ZQ6G=7~hM3DWwF6;TrPf_Fpiwhl?EcJH>9L}PX zytzcw5n-vhTwf2+4z#m#t0kXPAcn;T#rf-!V`=?;4uvc@`3f-T28HDMFBsG-rQozR zb_zD>sw-jV55Rv$5i2(1vI7VJz<(LZWBgBREG7;P?!R&L+k5)IadfTM$GJhm!I}uV z>gPTcH7GG_sG~KA&DrY!+m(g$8%J@2Ype!Qw^@d4yF#E^cp->+X&p&O7cmgEwC5*YatW61?Qo?P8Nd~MBWO7e(#NUrUi z0kULn-$Y=-_W{tswg!Mw7Z&(D4)No>MAUgEP~KPLg17Z?bcd)t;LY*)51X6Mb0u8) z{AddKpnNCVBe7p93$B<>VQ{E%{hH+A$^ajg5+ZqJnYBnbpn z7Fq^1v5x{9tcy|d^nx+l11yApI7(F%5K*X?#vJ4)#6Oad!vvb8Bn|L)3`3Bt+A zQd{s7H}^v4UyoxTUw?Yc#x3kWkjwmn9?V`pm}dGcbcYB#>2rMG_=3-4V)2d1=@L>bW~jc`E}X^RDp@lc<@6Jw5Z1U!76 zmCBu9fa|7+bOsCTCt<(9KX!!>+M}Q^-n6Qn;=5`QQK^5O#NO_43hqMKQ!H<*MEK=cKSD?y^7yjXMpp4Gg6tWDxxrI|8QD68x*>{sM#gjgaEiOP zg+PFte?4kOEkDQ^O##wM)I9g7+0baZ(WY2)5Ra&ib;~5y_k^^?tQ2g=cLt#j5P0iy zo=gm0r@?|-{v+E#%H&LzGqc@gfr>vlAU)9`5jX`Bg>{FBsNR2eV9f@;$wyQV7)*!s zLST}k2}57Rg3j8g3Eu3X&xUu?$^K5bk>=Q~NH1_eRabd!yrADYN&Kj?RR&4xdhi~l zUzZIPiifeUPtm&0yz#!#>7Dd#LO^|-v3Wa#P8YyRDNRK;Jx=|-lVMy zBy(I1Cg3oNhJWdD3Ru-IZZTs>&_7+>6^^O-K_zZ4FtN5`;UP|bh#wbk8YHMnN zK-%*grP^%apl=c(U`}-e_9&inx0S+az=~@}d-KZB*~k_w%YDi1QS2n|GT3LgEA5o? zP7IU|j(NMhsDFE>3>_Q&$em#5XUlK(?M5A$i?E6$un7z+_Fx+2ZeV&b&XxpMDuRlRPm zii#TxwGX~Yi-BO!hc7~i95yGIH`6~S+DeYbHwTzq9!IU54fVUz(egcw-_bW%fadoy zMuLFmfRdaV=k5ip9`x_j4RVu!xV9ZyOwly2SpVA%fENQAu$ck#iR2UCi7Y;)^yZI| znsnIhKwU5|oVS3xuy$DP2C2a8Rle2YFy}=CeltS1P4i>B5;@AIpMA_S^W;@-gvrrF z*hFrCli;&0IfR(-YIvSz6rr#_3UdEBE~HmKg_^LlfJ96!kcZ;-3ozwX<8XRcLewWB zYJ<2G3_S|^LP7|A<1l(D*Qj)=8$zs;mx-Q5&6gicA#Q=o)ywdK(o^+(CFNKUsG=je z@ZbbGI>7Rue^+s^w0M-h$;^(c7?s53ix;u@6)#&IuE{0Eg9COuS>J9h*GZR=_jEnq zE(eXcBWK({A5WPs+8s5lnoXgaPZAy$Gw8T)Tm#LrY2gxQJ+z`H7RJzi*{M}4A~Haf zu)&D9uAVT@-534SFJT{Jhs>v15b|4j!P*SbFABVhxcReay}}f0z-Y2W4Qb~7(OqTf z@BhrFy|COEpwNRL8_bI$nKooS!US{Z{l=MA@*M)>7Sw`BpWs7}UgFa~*AV|$PvcZH zCk=Z62Y`waasl;jg31quKViLPyGLp9L;b1C2W)H^z$oh)6}>mF#T{kfDq2rVnnbpl z?3dqVwEWFPy52K$iDcA^a&!s$UT&v^fzo`fb6mjQ(J7;g!G`h$ccLfK|eRs}O*7QHzYxU`T$8PF) z&a*lGl(75GN|>DWEN7%T_E=Brz32j*)V&ZxZqGE|gZ*;UHp2tT2PObE(_IGm=7 zs(0>)73c0nZM!1Vf5q`Qrf98VV8<4u$Qx^8= zOPV~!hz3b}ExC8VG%7{!+&I}BXLjDXELTpqtW74jv`h0cBNdsIh+kvYupb>{mQO1_ ze@;R@DpYG&?qSwH&;db;|m883?UCxBob<7>+2l1g-@rRc*^}a_P1oB{>}Dq{-=-B+S$b3!NS(*_b(k%|9?rJ z06C~3M@kl`x__Hg^s>x`zKtUD)pOr8jAlM6D-vm(vQf>o&qqulNsqkbKJ!o%L5leH z$CjfQSImOo?HHPZU>b^#^l=chFV+dkL6St*huuO+QOgbm7PAybf*ME{wZz__4r?5FQHlZdFwt4C(;jdb>94ffq|0kxxKj>^Y+;+UKB|}3Hy1G0EEZ=hmw{P zd^)cIvSL7kfXBIsNHx=+t+U(1%Y9)gWTbEl>dzEl(Ps06B5fjB%_>Z}M4Cv2y03U) zg}RJCN{To?aSi;(>hnqI!9CTbBt!#%2<=tN#sLRzi~=L(hsrT4?_8-> zs#IZh2%@K(kSU` zSNtv}L$ZR%MCclI#ADje$J?y#(?+DXlH<~ko?T(%7;;}>MUZz&cu(D%2!zO!9;kWf zg%V(bZF5&th0JdD0RwSCy_P?tKSyWX3}$Et;u0r=eBO>O-*?8Z`}@K-r@iG>cJC@& z{v~Bj@8-%?rE>A)Hnm)^)dH-E2;L)CswK2)u`!G*hI7Z$uVd;?qN^Wf4rSl?9z8O{ z+a@Dt=gfDZlu1_9(B;p$JI&#nthWKa-R}0_&SqIMrYadNP25j2hie|wHYA6`!x;&% zAv?+?R#WKU)0|=(i@Gy6FvBpis82_F{8NL@e@3#Klx`l>=hGPF=OHIEQj@G?6zbuZ z&o<9kM%O0hC+kqLzqZ=AONr}ZmosN`uhd}D^-2u1Otd(u?)O~7-Y6$YQCNmnt}={e z%u-0Pz&C8f91x&KOxe-ie>vcn{6IQW+qr>>#KA$eHMeHF^AL8`ru`j09nM&iNaBr+11kAN0A8t37l z(4BtB)Dp-{@nVR#OeFhYW*T>`vQYj+5@;1>t`nd09urjm*i8lFgTzl!P*nHez!Q;% zHeC6itWYIyvwj#${E#nGBDReWjMI=+S8Ag&gs4e~9-@%*-b<=T&|;&g37BL>;DBf} z4*moli@NHe^JG{}m*^4b*wEzRuCUz$yS@Q;tuI9l`D=A>kgI!j3_!Q@pc8Z$`PR@N z1TC*=F7SsRytuw64Is+JDl|{*C?PHa$65HYPCKU>?N}tV!u`z*++dx(%2Q(668aOU zsixF}`l8K`+7V62hzvqyU7YAJ(b$ud#`w?^z=QD zzG}T)a2z?0pGI;G%;G^(*8;FGZ3s)kqC8TP52>M0*}h;&?kTcsI6AT2W2v{nJ>&Es zq?vy!AlG`@j|yumQjSd^4}M0af?j>0Oz_&#MpO%t`T`oExN{G)>j9P|Or0D%sHpS9 z?fs)=s@_lC3v#$-EqYR$Z~M>CU1%UoT%FY$sT-YvVq@Cvaa5s;te&Hq9YUEntDl4x zT^?Q1pf2i;XzJ+RcbuiSGVc6c`C9XPbJ6NF#IhDwgV@Ag#vRE>^21`8KqoZlXG5Cqk?!-wIPVU}{8J7-+D$5hV7TQ1+&LWu-T7wqy6r+bP zKe#VJASgLU_Zq-LgQQJ7-tq~3=LRJ;mL+KWOAE2mr8j;_hsoRTRUE-jT;v@+{jH8Poz4cE4R&u5X09aDs(3lrRAUN*q} zl4vHLuZqE;Km@!HL3)3?29UQ@=nFVdR{1b^{F&zYAO}BdpPZeyIcEXNXT8$|Zbiaz z;f9*w6v8q4?5anC6K9in;3X^=zgX+B!Y`4iuKVSxXIR@KRUVioji5YV!qxR|1FLy751Qr$&&P>cO*N$z7!ptE&ByDlh>% zX7d3QA~4z0Zq)VG+;F-eD}?V6PY`yT*|J%XMgdq$S{7@3rZGCc{81Zw*JL{W?|CRl zNQ{EUiWv0Nm~yDWNUX`22xS8|A{EJ!W{tk4HWo4PaQ1`|GC|yxwFm-=xFnRUF)z3w z02!QU5B9#%A!Ha$f^LP%2V>KZMiVgIIZWs2{~?*iNV z4Z?Df;yeCQZ{<}OhVe0D{T-j4x#T}#ovAc~ql{J{LYauMwRSB3{&`j@3#dPszY;pB z!S~d21Lr;ck~b7z80_fThN<|JpqdQanb(wkOm}tPiF_XQ2;RE47Q3o@s#87Ge;OOZ zp$}bG=L_L-(nk{hpzlwkWsm{BW!k@HfR2y8KdteSt3&!H^lg3`AO-YKfLWStC%1LY z`ZtrwF0fJR=pE^+ zdXEC@4~K1pq15i|ZRMEhZTu7jBav{|zyA!eZPM@sF>;F1+up;>WV(pZ-S9Wy$>rEA zYndNtbmPEq!$?W3sF}wV8USpl6nQE^qmH|?pn6kDQAIf^U2(K)BreZxB^qBQTK4K~ zVna$Tz`g`Buk{;4lM1>5!LB%5ywnXMWgK?KXz>B(ibfhK*e+5$>4fb*JG=6(LDecL zFxHj3E$%yIw<={0#}DemUexRoWgcqmsy)B@X7svaMCV}JbMGwINUonnUwL9gSN@3= zWR6_S=zKoCw71UaESyTKB+}?!Z#TPc-%oE`JDgBi6IVr@{E|?gAik2RM~!9wHVskM z{k)8Q;~!w3bFHd4CpHH^+yDF#8-Z^-wJ~5|vNqsk6#lKP^2prEu;@B6mU(izUB3U@ z<(6S?D)ZT|CaA&vFPiZ8zu=O|YFc*KA}BsDH5fRBO~{_vqWMT62=gTTB7db80)nnG z?mVnx+dGU*%`(&4OU22XvJ2TXpYl`q>4;Fpy#*SP8pW@T@Lu5`px=?%nXE=$I4ha& zt~r>@rr)Mn_Uze%cFUBd;Qj`its7_C&rJFrSiZXs$cZz>lzEqqvUA?MkHeTtM5>f$ zrHm7Wr*PJo+qyj4d)c~N<0jP*DUh8Cf6$UZxEPKSgV5ygK8@WaR+thD7o+{N3XLTbm|7th!Djp_=g7DW>>!L7tz&VGNsshRKbAs0a0UX-r@k2?5ay-l$5Ql|OG(r$`1l(o`GMObauqi~S z4?fV>5>Qx-A$5(5WmGIDEt1uSPUA%K!6rea6QHDOSbUjZipU67F61Tqx)vQ$cip=39a z?SaK=rejs>LwqKo3FN+q%PkTF_)PTgnbUli9v&4fVG5a(7@Aw@%d;mzf~!C_U`M#V z)-(0-$@ieZ&$lZ>HZ;;<0h%}3C5y2cln6$KznxXd1O8WWs7ZVjFpmGMuHt+@nUiZG zw}x4>4{)NK?jg`il-x7^73UV5o1u6~k8YU_19cqXlSeL0@}=YF40qoBq~w${Z&hp=1{<~@z62Z+9Fj^URC zre6o)W(X0O3#cy+cwX)RC?Z1MJ|W!E2_Wi(V0tt!x#Fjr?yg5*PH3_epxsqBYMoit zRIy0U9WTwi@zi4rfgGsp{4elu7KIaQb+@TNUQO9PFJ`INGa}v)0G6DcRZvR~eZJw& zwIM$2yX=SneJc1XHeE9nFS<}ooOR11<7!U~-Lk5LMdPRJ`DRG-0&^7`gQ0Jpj=lt= z-X*GP0F^lUV8;}e8Cm%iv-EaBbhdvgrV>(ZdWg>mRI|_s5D(BHA})Y7BaRON5X2$E zTpNBUJBD8QHheunNk$mBX2(GYZS>v7-Aak|-$b^j!$pyG|H;352chjQk`~v)BhzN1 zIml^!ZN#NLth7k-HJ0FE!%wQCpmcNF{yG5yxGy_`q-Rgp=j*F2+pF;^&-;%aE#J$n zA5GV(>7z)XeH;g4;XS;f$3mk(N#DW*B*uGvz=Coj0r_`(zwS|0@I~Gm#Wle7#$pi zmT$(ZiZ{zA-fQZ+H{QF*uw%D>_*wC0CWdF~?;@?au@8#eZ7U25teO{g-;I^NC25)o83Ph%_(&pj5M2#U| zVz=LR9rl;5x$*_7>5Jnia6j`k0G&;;^y77<30qg3@6DXGS)KlO(Z%3Zd# zFD*Nr0EHxi&tzBL?0n%63B3pyFs44hFtuD*T>LCVyH|fk&tmxBV5DG>05M(AIw(v&joXzXC@47zN4{G&5~F< z-mkY^Z(E@^-`1Pm7;i$BAoYTxIb^ZmqAA58EMan=^p^DLXU0j>^^c?GBDg*`oHiSg z*5zaDQ;B0J-EF|IUTp4O8FatvINhGEEopqVjmy9H5|6xJf>j*!WDRpp9J26HCvS#X zxcpVu=nqFq%?iB#ezK;O-`Kwr@cA?G?za5*_=+M){sGr5Jc5qlVCddMkQ8_^q;KJ4 z2w8@;cy$iZflaKD>yf<~^K#K&4go>u@JZh@drFGFN0>!tzA14^{rW3ST{$pL{b`U^ z8HTr>xqiG<`ncbO=bGzqWn&w2@^Cc5!B*?Idb{$Hl-c5;f35BiSK8e)NXmBqU3Fy$9N8>i#ZBNdx)Mlv=ZmBH2z-xqrKwq=}Q=;G6`;^(T~(*>4v zGccx=E zTilpv6e?H0KN9zbosF_<&81z^C z?~g|fG>4OyL_fX;ZpmVMOK|dlJ9~rfT+R z1>;+~;K^&+;|H&o(N@i5LzZJJ0$%1I*9-9X7fje)>x+*$uXtN;4lQ0)$z`+-weYtY z_+wwamFkcgp$8ASt(%q1j!JecvoXVjP1%L${+qH_`i||;sM0{@!Ds{78mZv=gtK;O zG?|@4ScI~##n+>c8a2<*mFTWL>Dg?IYwzKXciX(zW+bp2YejE6r#~-J2uovpiWo-? z)N=f7Dty$IQ*t9}e28Y}PZ>w;12C8kDGPpCP54){Ed}kR+3a1&50ehy&Gui8>F)-+ zBh6tz6*kb58Z{zMrd@)jX;(`=@H^CeZPhn0*Rjo&^o?!>Di_2#QZ(eEVuXd>H!Glw-{| z&e_`pjfNvH>^10I+kWHDm8hgG033R&SPUtB?U#{c0>}>E^ScFEAi+`P5GK} z+b$7N5)!CY3K2D-;_ z*HufT19_jvJ04~!_%PN~S9#Y&QG-aLA}jx%x^?Z_&PE7T=yc+`UECb0IXTT_&A!M# zlHX8*)8cq67qz?MuBfUZlM+6bf#+sFr{kWRy9#ASJOY>nC792g?4&I54x!lN2bEtm zLQ@zObLAj&rCfZLR=S`@t4_dn6U)kub@9#LU|04~zb$46Y%-niA9 z?ITyu^#T{?IFyrc?79T>a^OnQWdTUZ5x!X3*#};L4=)gPWqNT`pZmb(S-X$8vsii# zpiq)*%R{}>EPYWuGSx>tI;xY*@DoeZ4fpCsLk$9Pt`8{|JLot38yZE`$ z;UNr`mSZk55z*^>hnUEGxOH|7DI4D+Y(vE>tkQn*a&JKAQUsY_QHxXIhlUw7cy!ul>e zcp#QTbyL`SoF=8Xd8ck^DN0>_y}LQ#b`^A(o^DSx#TDGUMtUstt6PP{kL}dzRO0zV zw$cI356~$Bo$40gDpGl!7C&@@Qldr~lAh5dI#8-RB(Hs4VHbUWF2vUOK2P%PeK#%|faa;fy#-K=^1PCMzMHIes zF5vNwWRAm7{JL?P4$w*alt`@D?W>IpfMuWBkF4_I5MM|~tgb&zUqby4G0luQfJiVtHyvKfeY944_D zAR1neuhGWXE2L3c_lG3wIK`Z;1wK~Wx&Gij?G=O#Q`S3o4d-iOxrY#Cf=t$BIKs0a zkX-9ZegyU2dT1g6zqH(A*$-me%2I?!jan5(P)YE)r~!~nJ`gSZWIZ>Kbbd>vn}w{u z=En1yS>kl`k!Sr?-~{v42JVxQ{-#^N`1{CyEdXi!{#)La`5@PFAz*%RR?gJI)i(*T za0P-s3jj*IUGLndU7*yCcxxav;#*>|rRW8-d8t}r3w%QN<5yrBBB=fZBr+A7qyd9! zA4#4~LC9^!%%afxgiV5l6AQ2m^0~Qu^u>9ETOg};c@H-IQZ>c;ZZG695^=!40w*ti zOUegOow%trI@mI8kLjC&r9vHcc^Obg(Z0niU-B5#>4liBsvV^&1Y1P1RXnO|bu;E^ z=cnKF_JUfhZ1qAaG)~)4h6D!jHv2e1Gw{>6a$LrOMMMCIZR74_(8PQU`tWj^CidXo zU4O&lhO$PkzW4I$A~-M>75w}q94Fx5Lnhnmvj;wJ_TsF8WQxvoTs!0nuy zZ*E1Bbr#9f+^;vS^104*yV>Yvm=&)H3>f38c|Txy6A*5<_1G24i(En znx`{GHsm*~qH*k}Sum{KV&vgF4suGKf=<^5Yw^POL+kKY&bMhmoO3fWKHn%pmWHfs z5bjFS%_I@VM;2D(ndOYl?r+{Vekw-=-l1^m48UlWR*e+Mn5veanS00e4&C+;vCZhS z*%(I^;2R@RTT*J&sM(0m$-Q@w78OoL0#U|Yk+iy!fTwsiO^#l)YoBZyTS7dl!Vt-D`3N{XUk<+IuJedQ%KFcX(OT8K18x`__Ji-6jq|k#jp6=;l_Tj8`b&Z1_aveBrsP>G-Gj3TiT# zOSM(AM}LYcj+q1V-)lA_X$5SlsDP6mU(Y__FMbTIp6Kr*7MLNyq&U{BNUI4U*{Z!Z z5LO^yv;cs+KdyrqKj$;5XMjMdYf$ff101bcBcZ}RR|kNJEyFp6jg5c@5};VAiL{kT zGr~V9<>7||kg`@^K}(@hksSa}FOEu*>-ew|syo>dVMHjT+lVYYPh4w$%|19sniAZbtuDF0S(5atI^Ly<~<0uwp{>( zGs-7{O;YFHvY+`m?gD2-P`AR+Zccx_Vin%#J?3xk=0rjLa*cU3A6>!IyH$!1!XP?hBTO^MbdN$O{jp*~}&IwoT;Moo@P}yjLe`Ui?V8{RVbk zw9t`#42z0+m3HyHQ}d^fY$s+JG?dfEbX+FArE_p!j`smy=IXjxNoD0^7fhjhmr;Hy zB-M>j{26RNOe8dNdce^>WGdt*l@LWc{uv;nzX1S+_)|dKclMevYUm(}M-(S%jLPA& zxpy65Sywr@Q0=kaN&@nPC%F?Zspl+yOO#zm8jBfX?3k3~SO9oof_0F1h0cI+2M6Gv zs2Ldy5P)_PPJie8FFQd4NUZ{r1LWi@autDgmX;14q_)A2g};4VLK0gr*1k$v*2h(TxZcL*_& zBluH#5$A!%Y7uhA5u{lg35TjZlhZ1BTY?3U)Ljyd1-qatA+*&>4)T=fX8$0OTg}0@ z18TWt;eJdi=$wY~Ba_O7ld5C%HZt_OSd!I>tSHi_1A$^?izf^!_8t~Y2p0E3pQ=&J z8WJ_&c2a`_OM<*B9_j^VIS&@WHYL=DTNalSPLRd9xBJ-ji?nA*YbdDTe| z=`cWdU>c)t}C$94gop93r zt46`>n&LASeP*0u>wr)$$x=Dtt1R)AgV0%p_(PJRHVR=CSyYeH6>4SevfyRgnXG-F!^QvHWe zu}k;!`@p12P#;!~OEpUsCeGRY%mG6m0bxra3G$GVF{mLgE2=7!nE~4Yw~=v1McqV; z%~gQ@mUziCR5E)`ttz81tpz2WKkZ$$tbj&BVvecR7<}5pws+5CG|d<CbCnIv`J& zch}$fWBC@cR*F`4edjf85m*t`T_anR1Ejg_u9t(&bxY6XckVXL`4%l*)S&cKw8AmZ z4|rLl2)rxrSejVmNUO#V5B4-kMOd1PbiY#aGfLwOJT-XTT8dEa&~cS)K0SyI)r!`; zRPLZ#bZbu{79K8%YMryEb(R-c<|!DopEtI;GK#N9s3$Ud1>rp zBNJgg;22|FV(`&4I6niK=T~7VRxrq4>a#CMmS^w#*vrB=iMo6)GrbQ`)v)&#wx1tv ziWJ|zomF!6zv;+GhlXlltD4rM(cVMEQRq>KJ^vI2N1vvM(p%kzY>#h5laYe3k~4m- z?=^Q)0wc6ErNG&^N^9&yfb8CDTYTcGt5?mll5Wf3zrls#Olsh4r+|ob1+vcqkyRT7 zQ?5_RC!mk=3gn^8|D>sQf94p{gOc8b;Uz^0<^G3mZtojxnh?0yo*ds2o%GeZZ#KedBikexC z|2F*VY`HN;y~m?hp`mW`eLarj(_Os+pwBNyj0XuuvtmK~Z`#cgS`>n=<+Rv~&EX&) z^01L>cloB`x-6w14Qhp>i9xF>ga;b3hf1&mI(&5{sjn2ggLttpj=e4)`5UeZUjD7! z_8k1A;Ff7M0c|QnAxJ|<9Awx?=RpKZIl!>|T=W8|JB%<)FgV!hR-;IHCg#p42w6*( znte276=SU~>t|q05W2e}lN_eq0ajpf@fXg#+;%=$ESYD&OGs>A4(nSB#$SlPn7KYi zuOU{72Wm3_8+-++i3Ux_MpdQCN4SG#1EpDUAZR7)h3*(X3vjdz5K1|WmCr739JAIx zi%YdwD|=sT7tAxFWuP#II26a$<~P;AJ<;`vINrhlfSA*3{f(IiMmy>3ng}e*{ z7QIl*#Ve+g^a41LI|9^OR|v300+#8Cw9J_KUQ)nAL6}i$1EKauq5^uI3#cV(x1t>% zG+&fyjcmARO+FKO19maGS_kb3zsfjO)3@oMPtsK#s1L%1xc01*FsJ&>E6*D-KLiy~ zxuv}vzBBmQLO$@_5W-`~4t76HQxpp-Imb!oG0?k~fPho~Z|Cr65poc~D%YUS zO3RGZRS6|ikyoCBO1R$H>tomy(XDY`%qm6Hy#de&P@G_ZI=Zs%COSEP)D_EHS6E;@ zszl!{`;~zJOj}6s?^S+CBuLfe#bFpp4zukQ%_FNdP10uT3V=@fR56mNLbW?sK;2*cl*CV`-ElGYg&ph@w7~*(io=nrNEol z0x>mFIjLw+8zB?QxwlDEh8;D8$8OXosjV<&&#(p`^cmV?;USm~iB{2qwnb_As8qvq zF5Sq!R>lE_Rs9uXv+(jUlNXxNF8UC=Fm>HDFDI)CO0W|i4Wd%oqrwQuIo^3zG~&9M z?YjNi^@rR(U$$y=dMrUTt7HU~&D{Dl=Fw6!a&S^fstVBa^id$BSbjV#xA`~c=UN0< z=SDzY7aH5xm@=`};I-_6+^j09Zj|iZ7>x-RwQh=)RMFimg}UOqXbP_Fg+YLY5EyfD zWFzvUt<(|#g*xQ8C?7Dkjqk%xUw)6E0*tCzG6GVq5etjXtJtd0QvY#|mM72Tj&WP9}y3OHQM-OMnSk02x3MbtY3 zpaWD%<%x*LSP6uCM4sG31#iNI-h0F+Cy;q%M62?nsx!ealov!*hZb;13o#)vN9f(Z zgla5qkSl+HG$qc*C<|jnMNAObkcW%h*n2=e zab0x_Ub2iPmw%}^GSLL|sNAILFr1*I*ZM@?qKv@gLK(q~Jj%x-S!dvpIKQMLMsxt6^Q&IR}ZI-@RUC#62$KhQ;@iUoo!yE zDa!MnGjA^vPb>{y&>{gBoSw6k$0o-YK(Ht(rfWC=h_-=rBP3}^&qxUu>RJ8SYgQOn z#xv1-k|06DutW$!IyP5Fo*rm`zr|RT3Ck%NR!XMh3n;(>wHa;LT#do2YT_aisgnqF zQyeUj!K{Xd8?p68#d6zr2k0jei)96I^x6WzU-O9vVA2w@_0OvY&(Rs6LEG*j69An{ zfr>MmC=cDwRbaWzgmgvl`9aTIbqeIuss$oW*^wqJ;rIb~M3c#ytzzv@A1);KZ_ypO z$LyL}>dditNu#s>ZsiP&vgSr8a%D!(89zYy1-w}DV;^exWsad$M1_U=Iwj5M1*B?r z=1)6)*rXnDVsxh7`G$Y-U6#|DbpZB~3Of>3sX|BE6%4C>JO|(1QWNqRmLDhG7e0N6?$IVB_X-2qZ;tg3TLT)PS; z2+7>%70+^{7YP!HvbW|Srf@(0h#Px_u9|Wn%B3y;!2sD6OYD-kVFA>4iJsK$?U<3N zbvzSis$CSv_rJjtXYAAE<Tych&}{MgEp^Oc21yD@u&W^p$u zzEfxbPwNxft~@1c3j+C|0IGe;FUSGVB1k$P8YoX7A#~e-gS2~^g|XB^4Ud3q&;!dD zuu)rK=)NfqI|^xadA&fGJ?oZ6VQ2Y!q`bg{eMv7SHZAR}`ViFFGv^LTQXA)3E2Q9N zB+8P9Re`86=ngB=H%UJB_YN%90`w~W1xcrTN*$E7a>tu@yCk`BvDZfr?tC@w{i@El zi>C*`_V4hE*T9cgU+&K4ZC%HwseiuI(U@H6xI9$zD`IC)dAeB-onjgBQW0uBLDUWjg=z!+-6G_|St#GdO=jysNSeFXT7I6accJ{0~b# zGiWqLrtz+^&K3`WMt7#QD|dR>O;)VQ^V=m~y}l-O>h%Y(UY%BR`p@Cz0@WL*#Zx7C%iswsR^={>Vcfanl zLUF6dI#ASODUH2yugj&u`QJ2%QHaBqZ0VwmcCddfXszCFle=d%(*NaaYkcyn-1M~| z`K@=jmh@D<%Q&iCRcx!Kb)n35i(bFJ-?YURYgRk^#Z~(Cg)Mazm;UG(UO<(}cXAPj z#f$Y{q8sk-0Q`6~Mhs3)0DG4Ft=?+v2kp8+fHN~Oto4Un_`*qYGIm2zON=P>IW{LM~7rSJYEt4Z{)y&i*p8@xFdH4 zzR(&v?1%K%HSMDp+bFfN@XeY{#Tt4SXJ_RG3T-wa8=9U@dN$pL%#$=o1IbB_q{~0R_-8c!W+0$;HW%L*x z=&?qtu!f^Vx0+#k^MY++nYwnPs1#whNhSMwr%(ChUA0?-C=|360jUZ>C_~0|EV$P9 zNdn97zAIat@}{g9vfqd2{b(&!*X0)^e=W8zZ3E|J#nHzjbspWlT{3E*LPV5p_NaF- z&K4aNXbD~rW3E%_KZ9h`J2{JAEC}o8DuPs~-BMmgX%!p(fwT7W$yD_k8kuR0N5oZq zEA|UvE+4B3-=fpe$2sCJbmr)iQBr9F&j{dW%M+wyRmRL1#r>WWP z!(YEF8bkS}`zv+|bMGIF>;!r4czSkkt(-u+aL`-*ts>o%+QIyayng)~^lHzKyx_go z+V9{%@p_KG;d?(jtSg^PscU{%{^-5?@S5uEy#_4*QhWf2!*>hsm-o?mhtkh^D?$&u z2Xuqd<>h6$_5QyXvNIf8++sAUsVA-S<`<~?xTUj%pPk@T9Il{J9{Dc&@kf=S!((I5 zRnt@hb-bfGIlkoa;fDiV>TrK7CXgWbpT;3t@&gyzk9v*}k1#Ub_iHN8v09#>NcrDW z;jm@SB~>HpGr4BanenfRV9e-XnCM~$29LanD+Jk;aT{lf+uJ}chiLhKispY8L5LW^ zFh(MLH~@IkBnWS=QNc(50(n?Fc(8U3(yES5jtMP4nWyA#q^D>Q;75)EP7cHb&!1Nc zl|6E7RZ3bQpEweESkur+Gkq~raZ$^mzU<6TuHB}PE)mSreka0U+tnmpK~i`&>QA1R z2UlHwVrI%tHnHlywH~)y51*G)ixDahlYXWmMIKSrW#Nk3tcdjF3oI(;w9=7~L>6x$ z>r(XNzyW{7Rpx|u?6|Y*VWc7TaN^^(Au^?S`_Phqq#NGRP`zC5)7#Am{b*6Y{lgPe zrhonm27`j3{&%qh_1_ax_*kI382c}T8!h(#0|SipjU62RJL0+Fw*Fs-_}&vW;+Gwz z6qBjBQ>IAfNGivqtQBvj3SPT9=@g4d5HcI;B&me9-t^a2Jr2-bNi9#C-IbkAB@`IW z4qSo01n+lIbbXpNdf5=ndKl@D(_i+ulWc#N@15j&IZ~R3;*cB}IWb~nfJ5d2 zDM}hhMwg;b>ir2-Zc=iOBpxmvAD4&2`#%4~VN?mV*sk>?ceKj6aqB6RfT}KN<4tP&7v$4pWDn_y~N!;q!f4jZjb^@UUEC`aa22UQ6fu*X&*#YdegdoP4jQaFC~g741z~*ia4`AvfWPCt8DB%}(Nr-f3(JqX zE!l4;EQrWp!408)n|jle+QLjEV&VtiUIK7)8aiVW0Rz-TI&gp47$l7)!3YQDOw;_6 zrUO?$A1;3fV|$Y)nCvxw-8DcP;Wgz8H-eO6`Z%eaIzb293(R);EI~hD1q4q{jR6hg zJ@Hc%B42(tb3h|TDw8>I0u^-p&5(}(oB8W;?^Jthj%wQP`yRrJPu&}DFN2uXlN{_? ziM;o2s6AH>sWk6YT*EWLFlPn32FXf_aKY+O#sZLjokSKBq8K#=B_$O$zUWYNb7Bbf zXJ4#=-_Q#m7HXaJ9I}85d1PSV&;#$BV?tRV*|cv2c80jZOiCw#DV+#hgjQ*tpCDBX zJ3IxmDGs@eC9~W>xOsq>+6y@6-6>_W1?^=b2$+7c?4IEaBw;AUGN8*xvOdD}B6m^Q zJwyvb=Wr4%ZEX`i&fn8)tL$%#N7D6}fN}1%(>{8k zP?3Y~s&@#2FYGHkjI$0Q5Gl98`jSi$#S1?+aCwUlKBhTPP;AO3-r#B`&}@R`vI3@@ zFM-cDU~b!n(dgAlmo!+V@2~&7jQm9qxY6ddXeO8@WN;K!o-~{Vh!*0-P3>tO?3~$% z+?nuFg>#4eZs6sYBBDrofefOHuVP^6zzMjbXsa$m)2RUB5wgQU;Bl3lkR~nSD2KLy zNLwt5pP({CzamW0a8E6Ktc>@<=>fPHWuOpGog=7JZOUqF&}mCT**35#=Zab}0HMr) z=T82sy%;uB<)29|)zr$5)U^p)J6?~5h05R-P&4ya@qal)_khAeAVkM?aECO9hWVU+K=ocKu2FFPo%gG=Fy-}K1v@pNds33v1uj@Q` z*MTZ2K?Or*SKwpFo`|uZ6eiWNR`pRZrL%1VSi$r^SfFwF0A_o$DcWFvHNgpVk-;`W zqf;{2jjQBIIyD+t;ZJ-8y6h?yIHB%f4nYWVadBfLP0tRp$2cW2EDga0~SV2s<0*Rfq)8g~hOQSb`!7!GqkHLoHhoVUjULp(P&Mx#cyTki9ZB9=8V^ljbq#3kj0Zgnek%GQ!;$FK)myQ5opKb6%i|q5fhXgv@vuKhkmYU#38rz=aRoQ z1PHh$Ta<6`6X?L#X|RXvmlG$W1t9VOne&M3Us^dBrZET&jiU#a2sDiDv@#r$7#c(` zd^%Yu`X~PW{i^5qLYrincU@4C!|NXN<1TZKTx0q6gsw-@deqQ+@Wp78=lT0q4nTtw9F`WTA-O_Xis$L9HX;HkwR%_ zKtPbL8NT5YeL9(fz#(~nW`@Knf?ET5>Tiq*yag7e*7lt+SU3X|gz6ZD;W;H^pR&AV zSpxKrJNif1V@X4jZwKJUS$#-Q;;ybC)CY2v219x1vPb^hNj4j&@VCpf+J4C+!ZB_{%TjHV5NuJ;^~ElKhIVnM zxaRA)pWrmi7`Xxx>`zd45$Qo(GhV>uJ?+1<1OF{?1leY)qVIM8BL^*{#C$x5M)cfj zK$ZgCbjJ5n%5f)+8X576LHn^=r@?7(+zEpZ)gmsGp~y_h`a5pi!eDAj_HmW9$-sFi zGJz>cvjR-eF7Q+;Ux*^Ad+e#`n8a1Y^i(X^$K>W_oImT6_t~xMW;sdJ;*Um$liOW% z*w@gB?_sg^@V1pSRY|q&Nr~8_3w!RjrNy87y|O@wlqQrcnr5a@#eXLqo0M=^>;LT& zb5jy^V!v6X)f2@L-?<|AT5HKHWWk8XTp5aK`V|`xRB-9-E$qy*5q-{2pl}W7~d!H>i+C}lQY`#^hTS0 zbp;>D)VO8*{&il0E{tm)%2V9!Yg0HL5s??QY%V)g$-g;73Iy z_#6#OR?lS#kJV{^YpQ}HBwb2{T0}OSHukp!#5-D2iD-zU)=6(yQ3hCqWQe+ z&%+W(?E|}&X}Y=t3h89}5}spdo=(B%GP?XZMp5Uyy@-RAbeSH7n#IbQo|z+;OgfE6 z%M-Cn`23|tgDjK$;eTfNTNF!j#&XI z7k&|Jf3w7_Z>!Al52eeq&Vd3Wx9KlUr`E9Vz+FM=l(hYI$LENWFS6L}<@Jy5s@w68 zKl1e{&IQ`3@JtBLvYsIL$WcE7!z~x!nnPY=anJZX7g=p(+NnG;9cZ)oc`35u@v6+y zhSQ3hrwVGAsH$Ow!$8DDL*u;EK&@*kFwi!uil+qO)K{ZvtPQ8N&Krj7nNKk)rp!YN z=F0bur|cP2MN?K7Rk={3u&og5kt;j`UbF0`FfIjeXgbnumRj4Do;%9LIvr1;($EpC zsaiC=WTrQm!fE}6C0GYoSil$_Tb@Jv%0=nt!bySk(=QX*)mWZE(opAlo{PDZ&S^M8 z*w(Rk!YsjPXi_UQ2I~6LyV%E(_FP6gcMnLN)!=%lQG4YEfT}|D07nO$Xc+=Euiwbky@_7a5UsCrg}nW zn#yjo020Z^G+a<0)CHHodS|2hr@WlEax1DGKr>Ov23{-M1eNUv+Z#sV2C#l_#20E) zHrr~U4jf@m>OG{n?*gl#X;CMAUGe&wBl?nZ?X3#Brd(du*X+6Hsk4@C9Yef49L0Gm zgE{P>H9P~q>p$&nQeuUPj3WF;@yE239{r?ho?GRHHYW*4T_~Aax6ilE=Dt%CF9YJ! zfQ|I*tm5wMU8!*qvAh2GXfj!kP9;R|L?{&(Lf!ZST3mA`4w^CFKEvH%l2=1d7qK4d zmMyZ2#$8qj+UblWF3yWw5qi~4XP)lkr%QB}F(JPGM@B8i%R_ASAFAOqXW(FY1&Mp3 zYO1*mdcYheIn|RSC_JPkA-|*}oGZtFdX09EQX=p0w4TSmTGTJYyx%>Uw1*#B@tVT8 zIpST+ptAQS&ii#dSdFLgM2W&5S#(%D_UO2{zB*S?9}wNBi!{&)E2>#zMqHTWMK-t< zn1XoMAH52mmdqfvr%{l%7rhJ{ zo19TAqUTBYgS(Z}3swJyvwsfbeD6Jq5r12%4Ieurbvn|-`iA#MmEJiR+Qc6z0qu7^ zLJU>m*|Zc(flcn@zPjI<6>~Q%a2|^Q|Eq$u&?3Lvs}{o^Z}*y9s5l@OCw@+O!ao?W ztGwS!Z#@?$DNGm9ZE5?%-QZJ9(fexvzigV%%V5jckB86O#Rz%Nkf856`G?$t?IvGF z^Il)ST3h-yI(u3D!yW!FZ*Dk{6M2|hXTYaT$5QRfvOlXB>U6ubVxHq5N(*zGSTO#` zl#k&UuP$chXw{;uubvX*bxtc`KTTwI0~7N!a()%0+d_p~#Xn!&TAeFxG<@m!m9uP( zng3SYh626dt+C72e|S+f;;Aia$Yl?lastD)=t|I_(?6V325Bx$6Ys#Mdn(SbVXv&X z?FZdD>NnJfSK1fCc*3jTb%m>2Mf^iIdxx#AJ1V+w;2g5tzpr-Jcl6q5&HG)8 z5sjm0H`%wN@f8MkaE!ZE)M=-)1;- zXj0Isu4qcT;`mI`-%&2^FuEAd=WmZ|g?~FfCb^-=&`?t9j@ACiiuc^5kWPuN zxbRW!d2WPNh<=kTm@b@;WLBX7+r?2mrjNp`16C_FWHsO|qlajq4p|nytE@U=F{VL~ zD!IZ_Rz{KDYr&2+>8REIV7xOHYan-@^F4MX3e#J{pQoJRcqC~4ZLRR1CRjzg+)lPX zn6|_2G?~(A-ZX!AT;FSG6~MW?P)k7_CueP(>lALstJyCW;kR3U5_4_2xwC5{ zWQ^(>8)JtR^O8>gVTm;yCU;RAJz?(SNTM{53G~fRpLT9XRV`=@r*4JL&4=)_{wzFTqsIFJl zM8c$$ZknQEVnf@jNwe z&6wica`VraP%egEvb=+mPiWD*4uw}Thev~iVntm}9p5Q2<(;rhn56J|vEz3LFynLY za<1mUXNZx`ueoxW6Y;Vp4EO6~jyoiNC*gF^3g;^PbBDLxWy=nf=46hal?{fzBJ-yy z=&UlkCC?|xg3bwQ}7Y!e4uAalESmx@pwcV@DWr->Rx`lCnofHL96Fdvu>l}qpi zjz$z_X3vXwhGnR*$r}|32BHeHFk>btk|j|#u!s(c3c!sq!c@%JJ-VgqAj=8a)xkPmjNvQ~nmDM5!1KY&#km|mdi;*iZHi<*rkp!Q$)+etrNqD1 z$GN^+Ie2Z0eShzd9CKr129XCdC&PXo`1mlqJG?%qkmu_D7YN&@uWi{;nA_%1)ucyu zNACi!jtP_pe*%Z{HUKsHWO>=P*t`h5oGe%N{MB8=k|oe8t0(ax?aX%RO6GuH9r(N= zh%>vgMC{(+$_4$c{l;8P*6=2HUZt!Iapn1t*RZ|n*7&>v=+ZFMD%Vk7bVwf7trAiY zXGYn;#c)Y)JMfBr1LHQsZ#x`zUn%VV>DT+*^!<5zfcyC!ed3R^&-<@8eJ=0h{-#cl zf%7dQ-MNZeH1Cxw9IHAnC7@p;|-(%ru2*Z(&Blr2}t zLA}wc#}>&h3aBa}d+V70n*UvUH*4|PG^E|;9{~Y$kw<$d+ODJgN3PBnfZ)0i@^GT! zU7S|!Dt@+n+i~AJB0$x(E2zZ&%wA9yYj;D?ZdC#E)Q~PMC^KWqK1=Sy2=^ukQFH8q z-0O`hauOJt$rU9EAm>Rea~uMQk`1Un!WU-h#;%%ZDC;H@exu8xduF#_pv$1#=~oQE zd>5KlW!Dicstkw|UeV0MfuJ85o{ROt&>cpkC- z87Q>n3V+EZEJ!f_FFnsOp=nRyi#rPKTcxZ!Vp*25lB?39#HozU;Jl<=da_85YG5_N zz3`10#&Jd%CV^L}CxD6lJIg^$MBm?LT(2LS9hROo&)Mf^{CGyH<(aAR!_caeyj9mO z`&cVGg&MtPaA4P%XrVEXAxk@VY8sBk6?-*3OwZnmW=YCU$8+VK2c{7^)bUOjAS!e? z%)H`vx+rtVVVzW00KsLNsXyD;zUKdVyT@&2zl|20Zy=k0%N{96-|NA=xLpxaK#a* z1em(O4m|OK2ymIzaF_)&spSx!Vn3W@Ib3N4tJR=qHHm?BlD$I>u4EPbW+{5dVhdgv zv_ep3kT3?JhDgRFt%7*L%pxvem!LKo_4YtGTf^-UIC%IyK0L<+eWl|AMXE|8VOx$X zgBu9+G(giCoRaVC*7wKOiL-b6UmB}f$Lr32AO2tyB!2mH{r>x&ZN5t0H}(^D2xL3| z#vka@InF@;za6q?93&FXBC=H$`S+)lR7gH344R z26X{41e+}>W)DF`QSYc8iH8i6vLhGS+y}3U;bQ^zz?)Rj2&FQP(b^)#9F^!o8_?VAIUQA!}${v7;mvLpd;iB=VJn- zhds)aApP(r_(GRZftJ|NJXe4#b*|?Z944z#M#pKBR90J>3fX{%1-8Oha)sMz!IwL0 zmh++&>sg6EA^UzV}B+uQT-OmLV?k7bMo4^}1z6G{3huw+{V5<30oZO1nMpS_3x&iJM;y zW|)x-t?9no#S7R*w`;*m;;j@fYU2*ltd*%kswb@{iku+c25r_~+=X9}DNsKXKeNvr zPrxG8YVf+I*(Mn+4OQK126AD}_>zRLZtg?~Pk`R4g)oy{3c(ej{Y_F+$B7Ewx>>-x zE2P!y{!Ui)7tWI< z@)wUvoj_T?M6O7p{{{;V=R0Mp$Tn|sT5?N0Nvy0k$Qojr-KgCj0#V!>(4+M)7T8-~ zjH01+NJi6^Shs;$Zz8lDw*6$zbqLCCamTQkbdcdBM9td8OJoxYS++aqDX=~|5> z%cjPE_SW6tTh*6x%nKr0>PPo=os0bD5-X3#;$nm^2889^^uvtVXN$N!_G zPNfJ_qAf7$rky)?55A(&Mj0$oy>p}U`?&=g)%sK}ck)ju`;Q81=K87hwJX?r_gZlx zccz%6+ndL7r)K*8ZS@X2kUJdq?kWp3V4Y3wkP4!;My`Xfe^4kabgSvBA)2@27l=TQ zE>O$%L`LV14V;qI9+!EWztz4QDBpe0_shOmF+NLq&yQDgK+6(ShKe3Uz+&LXY4eKT zmK4B&rNj5wZY4wvPGOE=VTgskr+>U@U!s0qEXz_aKHbOIwUt#D+Hyj?e73ha+iK1a z7YFe*cQ9W3{rjf>>)l+A_}JnBzb!C+ z>X^lfch>kL$abTdONz77 zjJTFK4Z9PCj;m5=G|PI~idNp3T|IPxIH@ltAKe2zvl1*Dg*Xyh0`iVGocA}fnngon zLLGfyjjl-}s18}AC2K``pgC?dq?(|}UFozWYv?1FR1g!3ON6>iG!=nN<<^KOFbxAv z;Xx;+zf+zOo0zW?+mMLWdU7D(^QR8-ec!LzC12lfd%|Q>-~v;Dx=$IjSw}odWm*UB z%1AQqxiRpz!$EI${Z~wSTQ`)Bvclj+Su?NQHfTGp6ZbLaxuV|w+K=dp|H#YZxOnrD z{5-XyYi_DKP%9;FLw-3^ZJ1f^RCkr+Hs9D8D+7s7!G6Y^ABOWG@n<^UPn9 z+qn_rN3ZJON)`?mCDNJ2=yA5O)E-Wcq~1*vCpZ!}V375n^?L+me39vaJItNegYa+v z@_9peJMM9k+n4sBWI~aX*q7aPG<36rpp#?S zcM`BT999O0SbqijxLFe~XP4G%|#DB50jtc+^2pz!;F)HkC zfEwI{I9=zdf`(4}dz4$SK4J>G_}_12-K{S;oXlabmZ&2@86I*-R;srBF5`NCg^gKQ z&njktOJ-4>w4rdXEF~x`xnNkQQB2_W0_KtyJ28+o>fJjuDwmMSRME1FRLo~5;ZtZa$|lV!Ptf__-~Pgv>niZ1sS|8Iy!!BUz-nw zk%P0Xnn4QUNbe&6(ryQ0kX{b*x)4a%q#@{=$o+7=ZltKY@UL*BjH*&*;?buf1)wVn zh7VJ<5S;~6u#Z+%8JVF~15&+W4he{JL(wEzAa$PZRBkrmOc6H~VvsQt@m!?k$^e1_ z_M9BLr8^H%r5b~R=z8b>BTDXGaZfwQ5F+{q2Aj}ZRPByyx-&%H!d*NJgpAY`r3ROs z7*1A*5Tq9@X_x3CiMJl%QZT?Oiu_gtB7#NWuRtYYlpzuiBuG{sb1^#f;RziGn1&Ak zyopFlbEmzmQ@Wj1g!(5Z^TL}u^t%) zhokx~!A_*P{P*cNRA`94fW1Iip2K}m1sO%?h@sI8tNJECG+Z6BjzxshH$ zNDqO(_kdsjAzwf=xLzxukt!gt-o@aEQM%^5|DsFwqRT=5qj7PbJ5{ah;Nhn!R-`z; zEK-g+RrbEMlT)g(J%LLlznvIu1R4OW807+J&V0W2m>_#wF4My6i!stK-0i?s8M8xw zjnX!yBY)+c-#o1R`hq|dho|&0p(cHu5q;9#zA*?Y`tkpbcP`2UHlUGJcLC-@5v7Rf zrrT**{8R+Hw3c9PG%ZEd$6gcdbQsvvtQr)=|OWl8eYhBI_tGMz#ryk6i%%^XJkIHDU zaT#sbWT!k^ZeP_nQT>y-s#$LQte|18$}+ws68$8$n#=z+r0=*TpAUnXJq$}WlKH1f z4XZk0sLL2g6SX|}rl-6Ni4lMKef?4~|5ZtS#8uAuS5ZEfH}^GRO^=X*zp?t+4lQx1_08HOtysa-ngi|!z*)sE-J}f|`4T$`gBK^yt32d%(FztbCUIh{3 zkD2#_DlLzz_8OP9JWh<9Ki&`>kMefO3%!O9tbpo@IJu;;LF|`ySNfxt5Ns8ehg125 zduP}Qn<@wk!~gbpB8e+NvUwTV@#Xo#BB<*? zWw`fTVFP{qbA0fZk(h#z&#ab3F697lAoX79K^;n|2eW2RmJ#H@o5Tz~;}1-9G?Xd= zkH{i5a?ZmQ$;+Nm-6@!UlyAH*eX)agtf^MIvJBEa1IQ_r<<%JwN9LH&)jd?!U@Og> zn90Hk8Z&s(sGNz+P)bhY&?radD+PH@)o*GIOmI^_i7zk)G&!brC}K~kv?E*LE?TM_ z$6BP#Qa`EXQ&T9)!(JAvU9K4B0exH}CIlZx4j*j1@c_u)Lz0z*fXn>Yl(=JQ;G7T)t99fOg%wRr!D^h+teKlUF@l)R%4$p1MAzO??`%>d zAqOa?y`Yv-c5XaOhL%V@ft|TQZIPn;0p2eG*66uWBHpVHXw3F;;m64&`Y8|#m*`Hc zSH)6xHDl6hCL1F=&2}Ha2Hh&I^KN#5E_4Ve$E0;|N(&6W(cee(ni=<1X+lwI z0F}3ieyCag<*o3#=CHd#(2>N>tj9n!5rJclKtI*t4+43LgN7cCf}$FmEs|~uUc-*X zs43R2&dD>`;X8~b;&%8!vZz|>xF#)H)#=!+?aE1LqEPjQpl~5*eiG>fvWU8#S#@eA zhh?8I)B7ILU3AS@&T|{iIW0}ZID@P=q7^ID^e~N?8z~S7Jg(N5 zH05idU%c-1a)3Rpjnw6}OVUgWWimduX{NH87fJ;U4)lnd!m#F^a4si`_)g6{e)9@H z(SntT0aw^BHac65N^Mz6g$?b32~u*&$F3lE!<1kp51nISR?$9X=xG?_R&mHVGtU_P z!^*RIC}ggulLP^m#heec&!q}AjFIf)mPnC+*OeE#A7RRW-U%*wIews!;yv}KaFim{ zaOH==s8@htmMr4jrLVV40xNgK-6{yGmWzcCAM>Se2wRld*Z$p31R>-{nM7q3#segh z0(neMA15 zoVuExd0FKy@Udi2k8o1XDDM_|D6Nw9g=M|E#bp%hJ!33+N7X~w{Gtv5Lo4|-QYi&c z>WlXSNJVA|JIGT?}-`*^v6BtOAzn-$Qhw2@$d_9*rs$d^kM z5IJo>KoiLxR0iBVFu_`Du`Tvc(qaRJg7Hj7Ft(8h+vIm3ya$<{qHb2^9?g-uxg~Av z_i+F;e2&IRu{Wo-H(LZLOCFQ6s*E-|aQje8v7QUbEsV-Udo}8kXYSWi()Ax;*}4hu zWcH6tr6n~zOC8Ehai+FcH{vDTY4X^G(o7WN*E@#Qn~q%bla8eSz95^O!ef9|(*!8) zM^lOFkkUn0022fhSEu_QEG%@Cjs0g2g9}HzG*A^cG~t-t_5&}aE>wbIHC-N9aV^ll zIeHGY)1aoz(yol&aEUUC%{7^zzSVX`r#PGbXjhihX3FjSQGvA6|8dX~IWrE}>Lqw5 zWBx{87-Fl#Yz_#?Yl6z6T4T7J30-kpTq-~lWqODuS(w<8rJm432TqH;z$p%H0Lr43 z0x}`TRBO)@*hq1MYIPbJ8J(@r1MspIz49SE0)d+)__L`ehpQciP^tq0u1P4ydCgG@?_34;&muLQBe7QjN%WJfE2xgnFoz(GJtPDbREBJJq$9 zo|zFUr`S_3U*JQNBh73tm3Pcoz#s&`u{S#+dcfiWNj6d9c359INbI2X5IvbZm zH9yXWcPMSRBG*Usj+J%Vh*9^F9_3p%`HRCSU(-Q}zY@lqO%sVEvru08%&LQ!+Dt(= zvwG+^XrSaFSmx%dI3unzC3Zb^Rpyi8C3BdZnS4k4PkK&avVhoIT;95*hc_IcxhHpb5FRU%(kMHZJB#dml5d&`qIfy0ju;q# zZX!vnuH{nw%rx;f(j$QfqfgaDpiP*RnrkpS5`%eS?6NG>VsDjUzyJq+NMYe697v(a z==@EFS_ny6QWobeY6U-m5C%P~a_>PhrEJ$ZLljEQXoCkWJlPcqW&ie&!99yQj`}KL zZ~zzzlP&a1DEfeAW?lEtYAXU3=9Jb!88$2mv@{P6mmkvTM+Ep>BrQs?Kkx#bEyo6+ zJv=8r0e>uEVJbR=h+9N*#@g45lH}z3t?R^vik_H}<}pVL_vr-6%iO`m$-TakJLd6@ zE9edgoSrQrY{m_~v(&!hfRf3J78oc6zkd`X^Ha(s-H%tP6G(O0@4en5`rmDY9T}74 zZ0&7T<5rz#CM*7nGb#5wjFKsbOMkw|&DDEzl&q@s-uaRk+J59CqIb7%|6U;xwp|YV z75i@qz{NgKu{to6O)c!J(5bKYxq7c5+v)WA+uwD_x6&aBCu@IGF`nJ^Os0Id^U}zG zz~zeP6&uNv1)1w*4aHZgp6jKXEoUFCUU+Y+GGmA5(bZUG@3S7+q~GuJGmcZy+B}*p zE0U*&+|b0%2jg@wm`JD>!HI{%AxL6Ep{K5yo;K2>V@a@X|EJlEBGud>>0Uy~?eKm)8z&^XN04Halc3C_DWIIOcJ((lzAYTiw`5m%2opPm}%oT-$-% zQ=qf+ZN;}#bye-&6S!`l&w8czwJ!ax%)JzBJZ*QL$Sa1;N`#?M(3tvZ_}n_QEQ8Dk zcOq_c@5*fQk~bNXk7Q%nR4uCWV^%OpS1f->sL)BIxRt8uSnOsTSt?m}@D zG7E2&hWILKx1e1tx2j41MNa7Q@?WLF(A-S9PC+!oGsYoT7tk9EZd!RE4BPf74QY12 zNJlof39<9qYUc0&P4bHMCl&ewr$(CZQHhOn{{m4sAJoSm-4Y{hBSk&9Cz`m{^Q;KE%^M-6hkd3p8!O%)vIl(xu#yxNX9^cjOspf35Y2n}qvBHLjGs^MItUo8K*#bFV6GDwy~g?Z+W8r>G@j*A+@1qp-M+1FKTq!J}NVW%U|&S z_9!*d2k;F70RZ^@`(5vU@!gCK9qb&O>|N~t*_3isZtXW15We&J27yR~(Un4)C8a}G z;@mcxE32f+B%x6{#zFs(x|whTuekVs+K_+QXlQ?pkN0r*dB0w1)tC{sEc_(_qzXkI z!fAK#Nm@u@9+!E#?KRE3D;XCgJd}_Fe3p!&>|9xwcvF^uh=R;DF#(cG-20%>y|Z9S zrMsYSE8jftwsz;!4nPuVw6DRsT#V|JcGVpDhs0Fp4|YcX7K>q3ae_(1R-p6?^cj*Lw6i7kPGsEr z88wYr3^-dS5#-t#Q&PXAO%c9w3o_;XK;wup(&!=M1tC1OXF1|%PK0=x={qALkS%zY zwHobevo^K)v*dGY3FBlt{0v$c72O4k@0tVTFj>;Q26(Ab=^!tk9_)spj}pNk1z0XU z%y!WrbW~YfA_vi*7@MEp8W;pdE;2!Yz}j8ce@~Xyc{y{Q4EY`ga~UIa&nkW|R?b>( zaeOQf(pu(-Gm@~ho@nl8uItz0@6+mUXO|G@^IYdUO!B1rvNyaP445#_2|qvPnUd2Rv#G()1=qD4* zg<3PdVxd#3f1SEjp&gflL9ifH$;=IF5k!ufB;=fJw|HVs5SqhYJ+g-us^lv?hccPH zl-*Pid#mzawniM-P{2y+1VTxKMpt@w1!7hO|^;p1ab(~V$MLwZgu z+*K&|Cs53x}sD5oqzJ4yJI60?%{`fQOgM>(NVkc0bAF28{L~LQi*R zy$O2ynH~M@ZueG_0?NshU5kSk4L)QgWXM=CDp9>ns)}KK10d(NTslt_CPh|L+M7W7 zRC~kPE(?D~9Sg(8JU_Ut1# zv%;iRs6Fgz6EbtNtaz1&{QL)>nj^k6j=q@?yrkJ{GY6~ zz|Gd#@9OELY0a3bU58_Dj3x_N*h`32VSY9jqdVDeqhH^o1#!Lc$Id0Lt(8OOC$__& zs*XN=()K-;UI(ahyt894t;7BJP1er&F<-%;z%vDtgC!Y{ZW0hLL^F87hy}QC@hFnzL3MwSp>5R$<*Z!TEOxf^*yReB zC_jc|Tea|1CX662_;HFsv89l=Mv}~VZw{og9)h;>Wu4mA*86mWC`IkmThX-WkEH;U z2Be`k(5XwR!|r^Swd0j}j8|_Qq7+{mBpt_vzc8H$*%MC zYa_249;f2Y7dwp0-ukI2kQjwi7t_s(F*q&qg_>oLD?e&z{^lNNXF6N9UA4ONM!b@baN1i~;NE z8_iARcf-wgnyB2*k0#nkMFpF;aA>c6i(3xfv4TAMwAQ`m+Hu4S?$AwpS!nBlzns7~ z?5)vhy92s?1&LPloje|=bJsDqG(vlvKJ#XV8NHTU)JB`WM&T-XpiG*p(h!!p`59gXIpJj6ZSagMn#0(%85Q$D zz{}8s7c93K)F*?O!VkoRgYA4Bjm|H1w+d2b=7R9f{dsJ7{p2?q)yz&_+>2M^#tA*d z=l{OpAG(Vbt^Ss>u|>KiGOJ86#NGb$iRTBM?D)2JKcSEMi-kP#(fKdPMiWg*#x&$0 zR*#~+sif1Ce_1|!lbT}6MZ*HTiZZR%3Cl1?qu*gD3d%Wv;MW28MCorYL3-v3C@u45 zZ+oTQO>C9a`A&P|$u(?~tMhM*TBxgiIqk4_-~^K@B=-Lc>sw&&gxL7Ud|g5Y06_mA z>B96J|-gBs|$bE-GGKC?| zL~0Ill9&o2s$rOd405`nm_{6P#+)<{@hC4RulL>yq9{kQ`o-ax(CA&*hJBR z2!xBX*CX5kuV|N!eulbA=z%GT)COD?#GgS9VVW?cgNIE#=*WbSWl{jnX;<2X&IfP{35f3S( zpoqUtPq(`ZhKd<8R)TH}0argy?bBW_KPOi|?;Pc?h*6f{9vn$`vIAS2Qrua$fzBUy z6EQTd7G4M5!)h84csNW`QyH(buzL-Sk0jk>pA!GU6R7J#<{8Yjx7;nNTs*E(=Q)1* zQ)SbSKLj|^qzD&xyJ?)oYUT~$bfjhP)!{p3DoBcZ5iVTGkTZ?rAh-pJuOLyv;cK|= zT)7lO7rkm3pAJBEod{*M4dFqGC36%=h!{?sB2a!fq^g9ZvjAs)&(m%#XfqUOBIOCV zsKRZLj|>y^}5 z)z7Tev!RdLWv}%)H~iCMFyd?wWh@9>(oA6`q4Qf}4{RzM2*#zh9?~NpXmp*BBFZDa z)HfL-kS0BliRU-H9Xl>;3w$`^#9*9Hsrh7idol>X^~wc+i`KC3{8ue=u?c;@&7oNd zseBqjir9gbO812kAP(BdvzBIbveKTJnV%oLY>2ga^7A4jOI`cIsHF{I8ZoT&(=D!> zZmuaEe$gasf`BEahB+&fj=IkRJ4sDbRBr$gT@~mu6)Y3f^|!jfL0`582&i<-#7q51 zjLG;@GW%0oEi%U^cHH-Y1^F_W*i_0{+MIhPLa5m7)`B-NdEFcjzgY&qi))tsH~S7K zABGG&j_~jTW_$elk)G1lJ=ZEPJr|>Ce-2~L%t)Y+2<04T=5(I>WAuvSM)>!2RK%^km)fg%Ev;obKAX#GP1v-LgZSAG{NumP% zijfPdQi9BVU`4>0TK?V?R0S8kiyQ1dv+Et%(&&)jv;Y3||t z{%M4-E8SI#)Z_%s3oH!?p|c_}7?idJ2JfT$1P8x!_q@}HnW6Zn;$5F}usu8UeQ=`$ zTNrHG&wkZph`j!-_HZK*v)7dotIMjfl>>>L@ymOiUeP^PGZo?SZYFjguqz9&VV!ZL zI+LyXkBlcG>d1ov1tLAbF0uix+o+R+?3omy`%Y3wf3E~eNCTr5Q53%CnyPFy@C9x? z0$W^5`tU5U09I-bRBp=P8v=-PwkclJli|V@(aJT(BbcOu)s}0hmz)dVc-+s{K@3cr z$Ma5}+n7LlFAYtQub1iPGg`c*%Dmzf=H~Rs<@I8CJu*kQZEe<9|MX#jj*?E3*rJoL z$llZmc}X}Q(oG>$Evf9b1tC9gmzy?rs&qwNkeyaTFmc$478PSO_vdFBp(kz;n7K&p zb)PJ7(};2ZyfMUJo9QKrKL02;&LQ6)lCe{3f%|q0$Q^7g!+zk6NjErllelu^fluoZ zA`pGG2DxwGmcG>BHa~oFrDNZS_)^v`>#0BbP}}OB!QE+w6*l4cIxNAg=<;7HPe;S% zn$_7O#BSgB2_R^LQeR1Zxeor^(Ch=^oK-cIPp>G5J^$i2 zD}6G&S!hAhEycV^$SwR@bFHX;!B*76zw6JuEi3w;h3yvst}1T3|NKAf+GlcB4JdE` z0Fr;HHve->urPEm{ZB)z(eSq4WJCDf%@1;2k!UsLgsq32JTS4%D)`%3YQbcaQRCGt zGATl$R8qcXx9szREBRg`-U-W^a7MH9ed9row{6#^4Z$}~6$6=Eft^T-SaCV)C`8or zr0{8Z-ohS*J>4$gBsUj|CWL5a7$$sxm4}>9PAxdp!6S%njf9Jj`#s0c$Mu6JtFHB? zYY3PGG72XJDLCuj=j;0YGs4P%_zWpM4(a9ixKQ2a>FfCZa-wtzA2G}#wXKHjwv+{l zm;MrBk>fx(?~RmH(z*$_T)3y1#APucnvgb3(noAV>n5GH0QhO#a#C;^0zt!R@Fk@t zwI*oqD1+ld3~E{!JhE}trpcokwPQM!j+vIi4F1hnmQ)P+B3Z?nxkr&DGlvjMbI50# zb=;yw%Rgut<+ih-y~$0E)u$VB3&o@0!GD~27unS|)z6}!Xi@*vB_ndNtTREviwPss z|4HRWhrZv>Fl`}{@-Kt;N=SLoCkIhl zk)srTv?FIY1SMbIUOc8^N`NhPkUQKM_x*eKN&YFlQU zlB{6GYoC$KA1Y9hvV?$gM8*#0zh=V(z{0jDa@Z=t50sFx$I1}bxsnBDcKFi%cC9^# zPn=Kap zW+oJp!^9W{O-7qpa3}+2&H{Ax5H3ONZCz!b@OY0}iefj6)7+*za< z!P?n7%hVIWS_dz3;yd%MPXt_9E%o5s0Q@G2Bvp@9P%hZaZ`sT;JR8?z=)w)omo99` z;9rl^Dr}+z*IhQnBBFze^!!dK>_pl_m%Yv(3_-x>{g`iA0#`)#(XGu7f)-KZ&MnEF zV~kV}4TS4ON5kgatFN2Q?E&1H1kqb}L5f#;Ara`oM(u%0$%k>kP>9U>^Md1>B94OI z08-3lOo#gxoAhJT!1S!QwYdeX*8admqbfxSM_8031mS_kn!VLjG!?MBjS8;f=_1)L z-Fn{rasDX>seky7WFRD*BPn z+c=XjdvQYx*>D6*|Vq8Y`4Qg=jBm-nuQSfOFJbpSSoJ#Kz@)^wLJ9x9GUG z2;lr@^GNsinh@K{=e^ax4(cYF{}clj01ox35x@|)Ez!sJ zSHU&C0ZfIluhQe289j94`{ZJ)ZYF(K^W6rexhMLY)A5S#oyMcY= zJ;F^LDoY}txPRLT*m6H#DUtecvdzsYIC>v_}dfKN6_i=tar{bSOnM0i! z5{^yfA4dgzdk5|ZvPyO3n92{3Q-J(^#0pci-@1j%^V+alj$UaywPK5x^K;vB?qbG^ z?U~i*vA(^~Ngf5v`E#OCL9qW*cKY2%wqw!#Hgl@!qb4|z)sr@wz6~3iN(<-V+_U2u{1Apjn`URcL%prgKui(AWr$@ftDhTlEdp) zE-up~@|#eYr6?kFRad}4(8G)ml&B`Wg3IIS;pN4~^OX?)NxpMxK68(b${nEf+L4s< zGOB8L7bh*n82kyLU(lDToM3x1de(Y4!|~g-uxV5Ev?+-_!v@6Mp};KKRf_GlXH#J! ztE=D^SQ*GzB<<;7iKA>~@_Rc6()(-RE(Q6S3bL6Wj3=w-PVcQ>DLL>Jw_Lm@eTo0y zd$!`DoeQ0R^CZ&0?uqfg=1Chqtks$(gTVQDKO1pM(HsNe{3K8;SiP<7i1OROR)vym@so?5mtzZIqcR*Key3~9jL$wIxH%} z4<8jD*aSYW~aRyjS?o%MRiyAuU86b1|_9igsI{5f~R?ABE(;0h_Tl4B$zSQA|EN* zG%?jdHx|`T?EaCV;+aObSphaT%~Uq^|HHa?8W`OyWE#SIN28^nJA3h$l^M;!JrCWb z#^{0mf~NMOTS%s4g03A%P^Sszru;xQi=B>JoKlh(HSIlg4h{KuTbV1ThVV z@<|hu77fM8%Nl6T$sHlcEk)*9T*x3+u(3yXD0xvsVIJft4WDXy#`IKGB}TkhP8b7O zeYIhjc$7}aVg$k9;b4h{#M&3j--rrSYT%5KwFdxP5*GpyEE#0DD9k31t+s=!HMwm0 z`q8Kg%u>R*bHO`6Xm0!?$fVsAR*vbnS&-CoZayLE5yZ(&ilRdCa~k7lo)62Zu=5mu zvjgX>+6_QE@sdii=`%%BM)n}Zzr9Yi#AW2_bjpDlC=_zUwLvP8A$`S25~{5-RUO9> z2s97HBjvF{qMC7t#2n`3cXnB36*r(Dee5Qo7wES4T{>UE#53(E zRqVB$mh_zkS>|0 zsjOgr+(UV#8*3?|-AM?O_fJ4cd18S3$&Q*TG=~q3v|M6{G*(%df;&sBO(H=Kod#A@ zL_FTh%L=5IV8?bO*9%Y)))mK#1;wx}Vt*|_zyS~4l-Ku3Y3(tt15ASk1QI`@d%2X< z0J~$oB$&0j?5$5r#*`lUaGl8hJo+QV!2Ao)@hlWK-cm^JFbnr*#R#pnQ&wlgSR;QG zn=u^JItt@x?wErF%5Od9@(oUHHmJcVJJ+Q7^)UnfxKE! z0=v*EKiuSg$QuUIw99BOjKZjj;heP9L=v3@lGj{cL=b0PPeH*&uCr}8n^P13ive!T zZ`>r4oK~rTy_n3H8|PxINc3x{+);B%S93D0VYRoVHk#!l*}SQG)NQ>20iruqY~fd% z6QR1<@n^NQE~S)$i`8P6Son@`SY_XA$(vK>;osdp8;Vfd6NWGXFz}WG?}IX^y57E= z%}bieE46NeFQ_=!NG}%uoY>9A)oHec;$uz(@qb#{(~T$yy>vL3>ESHm&rkAe}BkGX6)yp+`?SQ!!+HN#xI*b|DhG7%eDT!DEX#7j ziR~|$w^8{s<7S7PGetexsKoM3P}2#RX%{J-w~$({uiaGEF3xhnp=^5YB_CA1F}h8P7I2@A4)n!r zrj?MuCW%G3&Dye$bUe>#a`f{$m8=#s*_Wrl8@4qStKP@M+jpZozyIBl^Z``-k^WtN z5TO2lMml>FQ^)^|bg8Pk&IfELzNhNtM*|@k+|x4LO)l#;Y|>>iv@J8L z>@esu^wF455>=~rf|#thuH9rn}cS80hbc38ht9CI@gYRyCa%Fr=6v zE%PwJ{NRZvuPU4$hr6u%0;nAtGZxYy>qvKmqNh?yHDQc;_3^#Smuw|Vm`g}qjf}DA z2{pTDnI~Qur!%(+@XVPdk0-c-L@70mFm1V?TCd<5Qc)f`y)O()rWnjGaX|5*!YHTR znmh!=IEP~%{N>3;LNk%pmOTWvj??<*r$c-;#Fb!UMw;qrGw07OUzZZHZyM3Cbf_ti z6j!~i?Vi}xwqzjSmdIEt78p*0tkOlfDK{s3S))$wI*SSQ#AxC+bZuLSa77!a9gu?Izw0Lj^5fI&74p`?ZgT$YX@e(uM;eT`_S54m;n-#(TwG6>7 zqsjQwh6`9(M!3{fd4rH77o-%PZz=N3SJn5UcFo4MTlBvA4jr_@%m|WQ{~E_KXu9EPTWK9?t6UGwWQ$?)BYs{E9A2!^ z21Z$DH>**HlQG+R{nft3)AHW_*?zi>h&ZEe`XX)v*}*E0NLj3nk;MCx+i29ckIXRP!0RMdHM`Q%T*JJhp!qcwEa&EI>k?= zQbkoevQ~@MBWJo3pSifrM&26?KNhI|Mw41qs+1k=rtXbR&vjhzwVQ73DL1FNnmQS1 z+6Bx3cSS#i3m|+n+aE=ZnwY0(>QgChJUGF>){dl7V<7P@nsDEJh6;q9bf9epJ*$fS zqD$t(1%EXW#zR5;F(7eAED=NCAjMneSb?Ay8BR69GG7CxPLt?+@&ezE-OW$Y$&<;` z$qi2dzm()m7Lg z+b4j)GHRuxyo97mn~k9b{FQoC4ro6nbo;AZaTE)km#S?eVeSc?0LgalJ?yG@(AoJ) z-5Squ{MllYESWyGONwQT}grIeAE0O{lC#MYjj!6 z7Q0Y?U1eBsM#3WpzgFL3=@E00I=?@IBCkce_IA=SFY0G4% zw-C}Jt>r`D`KeqGhP=eSeVxl0hNMZ`U78(hKKpX1r3vcVp}zxZkCTJRqP+5^TsQ29Y`_RWw0?F=Pmy8$xSUG!#-(VoRVd)=CMJ zu@X&Jf`FD%t76l9AVq3mQMAbvx^u7r16IXLq3MHC$0E`Mj)mdK0UT?YqB#6J5X=z? zVma$IDp)AAz)2OO7IyvtPqjg?oSG%wODJ0M@L1OW!bCchsiAePGhc4Z>1M8oA&Bi* zfqt&t+-TR3Yxn@7O|klLZ2)I54;m{Dk&a5LHA@F3q}O!f13#*kpn=y_gH%9n>~9A5h?pqKl{`E0RMD zulg~tAFcLjAC#pg$bBgr73Hk5%TYr7UG7`8`h)sWvGr4JA-?&V4BsF8*}HU8e6?SU z;~p1ZN~0)@Q_|8`<^%Z9_f!fFZ#LR8(o3E%FRnlJB~T2m^X0N z-~WB`khE@ybSMw*;4l-HT)oP>Q^}K_#cl#z(#k}iI`pbP$#rL5pVTYFnnpZo2jqu5B%R!S}w~A@dyY2 zkn~SZvH#bWz~0uu(#F)}pPKqA%f>G+mF82M}w@DW!`%~-jCqLib1w3vxW{XWJL}`TQTn!QxP)W()oX%l5Ah= z+S2Tm5ZM+Q1E)`6uqasaKv%}v1Snf(QZRxm2Ju1yM?GjA6wf9)RbtUbCCd^%l1-H^ zzSfy9u5)v+#D&!3c7jC9#10S)PN&b}v&VNy=8Hkal?zka1cm`Agors1c~J@8EJbxnkoE#WAmM{+iwM)=|w) zf+?w~B1+xtM?*xw$&b}=6oK3O^rOG;{7W8$XC19C4U!>o734dm6*cJGeg*JxaZ>@I z%o8q?-rZNH&s8p-F{VZoAqo_*8$k0ZwfKCeb-sK}_i#{^qe2ldFBMRT*NWYv> z&&w6Ul-siqal=cAmc0#FZB`=Rc(J`!okfC**LU5G(cM4}b>1`+O{>mah6^veUF%Ms zRure8g5WolnZ)=v192<(RB0{U6C0D|66o~`iA|Ny82J-Cw)>*hA~$e-ljwJGx}1fC z@x_>;xs?!a6GO@o^3y68q8&v_`Ky(X65+HcMfPonRELg1QzCS4c*sB#m z5b2+x0#VUMe$e1z+DPg%{rwb>l&|b-g(ZfDt|=;K7#i66;vYbX9w%VJ(>vLrBd&X$ z=Uh{MsmRsI<;nB)=bTlfoe#YPP^Kf$2-UjT@8yRuaA5HGP`m~ zTxox|ze}3$lB1n)#Oya3z(OY7d7}GPkhgja_v*E08dYo1X1J5DK#W0LkE&$*B@im9 z_J3jfy^-h`YK(hypa^szT)#D%!J`7l# zeEeA&g+Rs0g5%tOTnS8}BI`xuz`=W#j1P&rK{QjIwb*5)a7=D>q00y>pR1=gk-wAmIFUhnb;qJJQ;&=%|OAt1Hl=T$8 ztXRNpCXnmXR001@^UVub!>U6{Oqa_{S?p_!px!fN{m}^RUoD~@)sGA-EOF)N{q(X& z*jqFJ_&o@I5f^U*p#L`P&|6PeJN(`-b_~`s=e-o@!EJu!_v?!Kss}&ZfYcd<2R00q zUg5a6ii`rlz;kiK8N?6=tatf?Atu~m2yQ+c=9$v$6}E1yeCxWX0eEuNXU)>#9*JJ) z|9Xj`WU271ev4hz&<2wpwzKBhZ#aZX81P?V z%Q?!vQltkLFF{g8QQzzSrE%7QsPJoHksGvk&0c7g&QF+f!|0FC7F!>8MURg$K;2TF zP198_ws{OxjY^HOuS_eiyvPY#;(Sn&;h}6VM1#1wC;CYfdz=o`P>&N!v4S()UzW8>X0D*MET!=D|S6bAavTm02 zIq2=12cBU^b0MPyx%$&S?%s zlp_tZZBvFIss2DnX6!M>DKjL07tv=ggnHoDRdtd=?@7N7xAY8vFTXNFnjBNQzMzsw zBNBfVeCZdFdF=1!F-XcS&%NVbHbcEE~GPwGqK2hkiDuphKz`nq7EInmTdN5T|bb&O* zvsyA9>?jPAP`3aTUFJC{;bWoTYTII~%_e>QF9B(IM^8r5q)e(QuIhF(&ZDfV3$%ii zkmrVPzQ#_GR%orN!gth0ugos~n<$B{<Xe0fJIi<;7VD8Wj)0U$scbjkbt`L7a!bAR`8#iA754UWt zH!>-KfWT{)PgC@<9Zp%Kb)H*Z*H(O7@%qdt}R9&KX9})>>SLcA@Z;`AIG7!U<}rH28nVc(+rhT zYXmUe1=2b8grTgd+_6lOuv85JmpN|S;Sj51t7;N(!KvT`p}ysmNYf4!Xbc<*g?T37 z57{0HIVTD$mQuxwp|ga%;{- zkTzH3p^#86NZ6B|^ZMMLTN9SsKViiVSH&M=2F)29B~2|yc`Xhr#)i7a+=*FIR|>|s zySsyzV1>PtF_O2ewe@)|CX+FnoVSw-K15Q2~TaESf5KWztdwHlrq=|{`5)p!vJ9jp0;0qk{-T@ch zDBCdt*ugrs?*~$ENvN5fx@5pJB_#S#EKA{l-BT;t3XeQwYX} z0|WDAPO2Hrc2GOx6;A6FouzeMRsk1At3Xer9Q+E+?~Nd58qnqdcaYlAIFyBlTJu1Y zqRC8>NkZ;|a-0Ru0vG|RdlN2mITmLVcw6L{F5|{{qlS@KDKQ|yZFnH~(Uvpq;<@uX z-DNM8FZdzbtZ~~^ZI?u<&pCjFc3B8Yqo6!Xr#H1FYrD8-=xgRJHbkWP)-~F=oZO#| z)rgn2bwAa@p;n+EvX(@a*PA!^zVTGHoe{{{b)C{B!d17E8mE+7ap+Uqf|o z?wnvAym$|eXxJ6_Bew;+CBRK@{OnCN751O>Zr&)`_#0$do9ZJhu! z>oju`TBMZtm(k^iB*DStdwsZtmtm_AeL)eBa%ef5#x!S)QFypcc$!(i1cNP>trEI5 z5f=VTU~nBJMPP2OY;ukXvzEaecL!_Rpl54je6O;tHRJrnh7J2$xi!`9{>Rn~{ufHT zfLEsts+G<=G2f_46-#5k2om1KcV+8&sAAa73aMb*R!2m zVDsT0$IRNj)O47xZ8|uEx2A!<>mP!KBR%C#y0zgHA{iPUIAMR_DUq1=$jBOp9t+?i zL*O;$LUBHmHHpAh!1Z-77-)zK=(w=)Ve+!y7EG&XjWC)THnX|3G#F2d)6R;OoGWtz zcC$Y))$=y2+1Wo`C-ptMHpPb}R_iVERxBR(PJQ!BVyjj*6TBYI3V3?Lx*%T}P%2B3 zEG~)@iHbu?sM9zEMUU1J(<%2i4kDym?cgRF9}^=+ly?BWwlW{k>yF#nw&q zP#|-Qe07rEI>|4ZvKvKN&B!rMA3&78IgVuJ??4&M%3WNdreU+rn0)l+9yybD?K%5RI$^UDn0UubJXGc! z#7#rE<*Kt%qcdcFhQrYHd=!!!wid zRB4vv%E_@>mfrk+@xCp87!6d!ZnAg{vWsf1zzcDt9`*i8u*YlZe2boyi2eegRv?*jU+t2c&OqP-wQW@6UNT9hz9IpPuH!{V@Ftg%j3GbzV$%3{}8 zJ#Vx}#A&ybvh-+DhP_&^?Kq*%!j4IpF`+UL?fSpowT9|w(5VF0BTbO`Y)%jMdZ@PK z+S*h*$~|YS&2VP=Dsecx^l~pO&V_qI(QiXRE-j>OtTiWaFG1PNB#k3bW>dkH5OOrD z<~#-RVPB(p=W!%QMjX=Q=hC|k-m+fWYt@X9a&VcMOYb=$((&a~|ADZ{5Edmh!PRpW zMj(wom#9&zd}j6)o!QvCY{J*;_VK+x`kMZ$--xu>_-YA#+P~x9E@ZV%VNR&GlE##;X&_d2`3=hdhA z6vEU6O^A5?q3PRV4~Nvc!AV15SD+TPAV`{^%XZBQn)V%}#~s3d-DGxBs$4cbQ4wDA z)62s+w+l{uP%mvKB&F9o42vPG9yUOhBgZa?5fO`A&CZ~^-QJwnhzLCzoKCPq(+%+J z$&k6r^4NpRI^;1_8wKelX;Vra0O0�ah=PVnddgZUcgcv!lbUr~rl#63LE;&9mI; z_}7DVlOlXT7A7>^qBv@J+eCb1f|3JNW=I9CXbj;P#Qx%B;Vs*H*&LA=v!-}=L2F8&N+21g%oMtxcqM&s*EobyL9Qc`F&^X8 z7b_w{qL==_55pDlKpwNNo8jV=0Vp=v5FoplycJ~}C>6Mqim|{=5;VeaXf*|Gaeiz@ za5Zb>DKQ~Ej^cbkH;nbfob1@>frU^S;9F{%V{+ZW^q`Ec+NJfPMA#$3_@$~$HHsOX z2TzHc@*bcT(a_2bl4_NhDo}x1$vJE?sgqGPBf(E&2=uG3?`)MR%{QO-hD&rF;?Vq@ zc`FB++c*0V1fk3MNb0Y}8-{QWqQ2r9oY$gBpCrTNK^uocmY9+e@*iBMWbnVuXc3e4 zB_m7&56u>YBhB86AEM@QVa->XsYQTc(D_>jq#2%}1As!9O95kNYT`S+-9%ABazD$_ zvm`w_5J0kfEOs0HW1a3IVirX`6?}?WI48YU>iJygJZNBWG7~VKs#zb@FF?g5rB>Y#b~A~ z!ftmQ$`^(YVmKJX5uVXtFV;>8jDv9UP*j0Gep)c;TOR`T&734q>j8BJWtfGi0b3m5 zuuQdpRK{9>@=gLtQoNk;0QZn4n^Mg;sLPbko0IMkr!JOM`|YLJ4jE7t9mm&|hvYiYq_tNzjC-#a`4g$z-~}mY8e2c1Kp_d>7#Lwdf;v&o2d;}Q28LMf zA3ooA_q1h3DyAagkVp(Z1}&DPEodXF5wkCT(!Iq_gph5{K$7MN7?)W6;7pPPHSs}B2OniqtTf2!WE>b_>qEN?6r_|PY@rCC?jpWbQ{G^ z%t8vo3l;|MoKicy7*9zH11fMQ!XY^uZ9Wz7!j$zxQu&mmtRU;WEfpPt52UA;5JGH^ z_()EG3^2TnPZDrsiVI_|z*D=^k{5egH4Iw^JdHFtA&m%J3I{iL?{!{p6W73aw3h@7 zS{h_nFn6lLRuqwll!uh^W`K`%pzjs?&7|EJd7(2MWt@)uW*lQIykYY`$S|b7ATJ^# z$fi91f9N{L?#!Ys%*M8D+qO}$ZQHhO+qP}ncw@7Yicv|rZhz}L`u>JJ&KYa3{ml6^ z@=CnW`SK?1vk+CD$jJ@e3597{z9HZ_g1`NW$qRYGWBF_PJC=W6R_duvSG{I2eb}bT zg?TO%)}QmR8HUf2zljLZ1YY&PXQ-?u;au$`j^)Sk10_X*TraOBsk9*#V#`{Wmx(-X zk!pUR%!TPT#avfrgRhYUcE#}soCjYtOvxPui`;UA;h8Sy$g~=HU<7aFR^pJ05qk0( zf?YG?R*)M<6vu)H-g56+R!2~CMk9SlBR>Vzj*X2Brk)dZxf2Zm7j#KR1`Wfe`{eC7 z)4$-vT%3r-vHkYVX}dWP>dKp$R%|$t9o)5-o1n7;<_@noPAjV~Yj&>?cLNkS!&~9N0RUm{O8rX*va+ z(Uhj-FqT#5WTe(^MP_XSwe1N>vpb2>bR08y@jmxI+|pc8DYk! z%3SLqJ1f`ZqWmdyx43a?qFcVy-oq6u2c^@W+AW+;$ea>fFIBXSfEC?L0lRkl78H-V zpt^x}2M5C4T{Cfk4Pe^D4#~WQCqA6b1@z7QZwSQ%UV~D{2_~i?l8vS=fTDkz$->9y z{?A?u0{>pUxB}}$w1$Zxvdy7b#_~pNJlW{_VT;Ck7Tv3QUKErFY*{DY<)MEiwy4F>4l`8z> z%*;n>)Hk*X=@e!Ip`zz%hM!_;2(J%wdn>z8Cfkh8}Td#H_l}Xy&z`OzJ zH%*oT<8TXiq7sZ*PNN3)4~*A94IdX??^YbCJHFW6d8{(mfr;~l z#k=KqB$I!}g_;jv6z@!gj}Rk^>&%%R@6FeThf0ra+E{q{{7;6G;Fdc93a){uFFl#2 zzQ);)Tro+&MU+Hh)O&&X6y9JG?0Rf4+(CNF0Od;(9u(?1H#g>^KiU+?@IF0vNtJ-+&^njdYe*qfGO@xR zOIgi-GU6-`SEV&N6%Y0<94fqn#M4*aA6p-lVGqt)HO<8a#!cbuF!wOqZ9#A<9!KS*N%*h~94qz~7HPh+FE)1U+$cj#( zDa1V}jNw46j4eHSQwgXHk^}{f*8!nAYT-GK@*s0s+J#>pE0wOj8QSEqzR(M5CV0-e?`KdfpJLDq3KH8?^%`{a)za~2%2 zN;e3#j?%`=s6a$vMjxQ4-VbR$=WLja2N{B<_0LMq^zbMQoXz5je%U|I{R)CYnJ)T7JqUaYdbRhBQ=ne6u;fjoLqx^%QcDn`4fj$U725%aFP* zTwrNb0Or*Y1=5+q0g^mNdc$fOM8{!ZQ99zgI&U`kmPP~0cRdxG5SmQ#K)R78SpuZm zfF9J&T6$KW+kT>x)TDR;x&=b^pu-gB>9pkp+Z69aSs;I*c@fwCD>P?qaL9owReb!5 zsiqi)87sH#D0TW)Psh^BGa^3Q9Ph<#4@DD?e}1lJ zBq@LbBWC?*E;S({=-r&<;%2g%_uO39Ei-txri#=+RL!^Q@N^)I&oYe}v_$|x`FK8Y z3QBEIh7dKn!JwVtE4EN!ShwcjggtsO@i_j;w2i2KbFFEd`Ra*mxUq&fWprEp=QM@= z2^VrS9@s5Kp;|C#hu^dYDI1NixaCxh&WAm4QnySvy8_yB?9W+8b>4N$`|e7M1(C_PrYqqFj${t=u6vH(*kcm!+m7*x^IF(8*-;2hHmYJ z8y+iqJNB1UA{CW|KKaJY5aX|7h~4nfz41CDvO`}@$k47|R-12^$-&*|0Vqo4tCgR+ z98G=X-jl{{i|8s`-B@I;cVf*!ajjwSqNuM-Fg)W98X7Q_GOCxH{+UEC@HHB81>*;E z!lHJ0EBM8{DiZqmcITt>KFCsI=1TlIa(+6_rzK_f=`=e8#ehc06AG^aD1LCZd>>nr zbgsDJsH00g-QPqy;}t7hhWjCZop>*cZyJA{_~+oUNZ`YjAg=%p4wH@6IWmQ*a9l=jirD2Wuc6+!@4LcpQ_`FMc}p5=KEw&E z6#4XzjHDTsFR%P%ZMp-3&D7}j0B^ek0qIVBYS|Ao-pCG8=&T4O0e=#0UsWsNkXo+| zUw37~UiWg~?zqvk;9Zcb?r?c`@v<4v5x~V}rISJ^o~B(NtgHQst@|_`5aw_cIdTgnTFXdG6Y3rQ1dJ}%D#i-xkr-(_Qxh^1? zL5>Kj{B+qSM1^lm^rSnJ)ALcfWn(QZZ3)dJt>*=ZLPgEv^(>eGIj(I8JP0Xaly|lC ztpW?noD&|0MxzD+f_!XvPCHrh+0E-3O96Ihje7OYnVpI-24$}*h~>DKN8adsdqGy= z|K&4R+)_bXE)mO-_8@(BF|Y#YxK($q9zXe*0cGzywCy;|=E#}J!f|jNvAea}E6`Hg zhIiw8WbX82ck#$ug1A$E$o3=ksqn$xl`7C!f}>>~cLz~-D1SL)=+nivt*0FMENxUZ z)Q9tkl`R-5J+ZUEz2=Bk#Eu3RXh#yyG^duKRx?gTDZ?8(Pau&NoK&1mrs_zHzz_X&C?sLL@T-@=lOT?O*c1xe^T=5{xu66s~cRsL`<#EOD5JS z^V!EOnv)F%i23iOjGDi1zdOEf-vw&crZm2uhsi+mVYzxjRelgA9bz?#9?ah^z z)R3_p1A!PIj97OFoS6PY4<}~gm(pDdS%V9B2750(?z^6dzz%f?z-^=D8&)2zHts@a z^rF0cm)L#EUo&#V2pP?-vA=woz#-NrO}e{0ua4@pC2wK+QJ2;043SF8mw`n3BX0-e z;H#}ax|DvPZ1CuezucF|2{Jv%4ECrPu27%zl?*i>MfYSaGor4}P8Y|Z^eI*3PgpLr<=9Jg{x&2uOprDF4*P7_gcZoOWCUylPoo1R^eZ zEFl7cdD?D@qAtOLbJ{Vag5j#FzUx$!TEx0Lfnq!r4wX#uw zDFqJLL;1qzLn1{`9(3fG4?dZ;GBQsADg3?*o?HLGZ_Nde1lETy()a-=qZeNBnnXCu z*2FENAmpLrV||j2Ps^6X(@Z=5l9VG8W~arIl`FA*nDib?k49eK$50kpo@&4JOQ zIi8o18-m}IOtJ~f9UVqTZwe#VKqV8VUXCSia@L(k3k@VV5luj+fv^ir%^!cdrFxNL z7sw9ULIY9vpb1p1GEy(n5GJr<FHUmZi}<#)Dto<`g#NZ^Q}n^V z@w+9sO0;sRhY4RV3S>MzS9C;lLu@;T4;MzZ)AaDyJnPC^vk}5%n{Kx2^^@h@5gzQ; zAa-N#2HH+rZ|6yZFIR38`X7{M6}fMZTzUiRMcf8?8%fUbQKVs}^2d&%oYw(C6&UwI z`Pxl}iw3qFrbKjb)sZO8zfl8|jtU`^60bdasap|X*_J)gZ$zRm-1)OH_s4_^7|eyX ze3o70?)I(^T2Z8|A1Yt`Xe3kRsvFgc7(Ww-`WO($kAx?XSz3fB%u(JxKVFi{%dh-8 zSMp7|0TS?FmppY(hwi8S3m*p9UhLdl?0X>D9z9Bc&+Nh7#}-AfNDHlWYw!E>kc7d}BFUT8=MH z`_0RHZ`ie%5OHyN?|4vs0~Y;5!DQbW1Q*JJvjYO{?fz5Az3URpfrSRekrQWQcWsW$ zFCZ5oBFuifj`CWudjHt=u5&_bme`U30(~;NRckJ~WFT3p`c> zxO&jNeMtETaaY`FT*POfS(r9LVS7#MP_;lzCzc>=qt={kKX$=6366`SzSO$KS!=WK zgY@k7WaQW14~8eIsjxsN!jC?~KQvEVd|P?#A`ely_w3X`nC zXS-i4C^f+kcs|Y8Judk@g;}<2HZ#}!;ecvmxr>tQVdSQ?HWf<%`y($)(D)`)g)xFBIyKN6m41oeT9v`3Lnq`(C?!kUAoUu@GnU*@+SERwk=xASaAPsur^9b_3cEYj@QWOyO%jtYdBV2$s; zS#5?#z$?_~>|Vic{ziTo;>=Ey`B%9Sg8&AGWY!QIv;On3HcM)ZUTeTl`u>rUpt-mz z)n}=5a_0!1lFQ17sN zg2|A~rHN?*9UR)2>u;G-HF-T;qvXTi+B_P_^3;`tP1>-eLE7gI`FuX*bE;F|)q2kK zVm@IZNK;Vcc*~`VQ$O`0sRlrPnDh$%ox3zX)k@t)nFu_~&Sv68{MS~I$L7H?{9ERP z`b9dw1IX(8Qeagi{;+7Rr8Cr1F<)|@5Rd1V1FsO!a!E3ay6xm08fFqkpv$%|fb1R3%P(Ji&hmg;dvJ6ZB!SN|xcNqW z6qXw4R3RsFzX*`8*>fOBwrRn-4M0Q*!uAR-$!=vEqlxp3lmvV87>+pRN^U#FMB75( z1w>Bp35D(5XHvO9UHr5J#pJY=9QqsR!aW)&GX@|WX_`EoMHgj}5$c!~o<`E4yQO;P zWq_5fyuzM`A;+X+{^ATmZLLXS1g1#SM$y}CDRKI_@GLa|eCKH#Q-OkuSe{^Zp)gXi zyN009aNYh3D(Q9+#;K6>mHk?H2+Loiuj4`mZP(TUn#@f19MqFD?DooPOu?HPw58s$ zWT^$MPzjq~6utVSY8T zs8$6+kD!%K6eqGNK|53Tnm}aDxS_^lC=;V~eKaesVjgXdil*U2+H4E{E_5{B57;>3 zmFqlpT1p-?qC0RefpHue_1&z(J4JcL3^s#(;+btv3d9^{au; zclZ`l-Rh7hj#j7pk$C{P&dqq&73TOXnE1kl{ePRy;HhL{*w?a;R{yH3qx;NkY}@ z1bjI1;{7(y^s${#)}7nnLfnt)@)x=Z6|76$7P}(!9%^fgf4E%8!CzU^7fC~P`IAN$ zsTPeZn_N+t>zV=C$dBD7XENXP*0R43D2Y`uT1!OjoQPeRT6a#5>QrbRL|i zNQf_w5W_cNSEk(}eQ8*{be!NoeR6W37JzU75@J=3^@Z_^7`=&MWly}o*oVg%7O4Qh zOR`9@{Z*?~)m!n*gaA!jwC=!iX-z*So0Vcs!|l#@ONPO;G7A&krMRO!mQLA*>+0ce zi1cC#o1&iTysiLg8L^^tCjPGW-I8+!^ zb)#r$UC=8ZD$~4mLKgl&mZ1&4kjKb3w%gWbY#meDm_T(>Wsf$!FL_R}Iae)d=h1IS zaNDB2?KQG@5n5`b_0u;nxHf88@M=Jz;Ws(T5+R2l+!~Ft2TuNe9}ldTAo}qK@7N4b{AFPwn%lHLC!eJ(kWrG5u}|Mr))EA zzS}(4^rTHgGv}OGnJ!&RAX&C&!v@Wa+Oyy0OEd>D2OMH+@MC&MjCNgDLu*_!o2c)GRt*5cDz`f2QI*1IlIT=OAZEPBo;v{U|XCP!mDRy>z{&{@(w6X z0ic7BD@Brt$8j6dvc5$v3Vg$L<;k7tJ;$&1_HQA=6Lj40r`&0Vp|o5Pj5_&fwBJIyMzvs}XF}_W z@*|I=wDJY1A_=vV)c4|O_>GFp0VS4adPw~Es$_%m?ohsu9%*!4kpe&pVcX+d4$fkZ z)M_S0Fv*Vv-yPz)h||hD@^UMEUjG1wO7Mf|+`*PsHu2xaNq>Xlo5X+Ik1EptjzZwiRftpUVr-NS*#I_0NwCgn0!)Dw1MCt~mQ-#r!(~zBx)^!i~ zNh&wf$4K+eA|B%39eW6Rl3>UFff{PqKKaK}!*KVND2GK+r5&tpfB*TBSbrNoQlcQE z$&^v-L5kAWXhXoaOp3H7)$*^yyr@N8n)>rc;W6bnWx}#e_#?ackprrfn%Z1SDLVhl z>;0PWjzL6nL`y_g9P{FR4Avsx@#E$c5RW6R!QsDs2G#bdV}#TB1@yi5<&)Jl3xT%ua&j1NZto}%h}z9&XQNu zQ3^CYDdDG|5mg^$UJNFxR~ri$XviwBCV|F}JIZt<2)gF(aWdbvsL3`U3R@fk(X+Fdb)XA* zT|kO1hsqPTh5>iz;P$01OHA29f^W3+z8b6t<*VHPG;JtyDl4yoe2=p?crs$7s_^fx zS-DF`0K8avGFXv|vyAc_QWuVtRQ^pzUkW4U`1;+XRnfDzC)zpp*=)D>mES+$aGU$p z2~XgZ1Yl6wiyy%NEq!YQIJQ-BG-53zqBm6>-Yabu@%3RQRtK+4_>T$h%;}gm?k5Lj z!G+C&2H1RK3G@%JH(AN?*~;PRiFd46Minm*X6A=1czL>h@bB^3INA(`d~s>^l0d)7 zokvw3n&LV?-A1_H3UiM)J|@qm)wkFZw(LX!A_*of7;_lE6baOdA5IyW~2N)_lK_{zjXvp-46HHbw->%tXu{Tk>Ix@*J;-qYZO$gS)W|`5nBN zW4}6p6v7p3#M(e?=UW{?qgA*2vfwJDcOpKArpgniyKXnK*qq&41o!Ar`i_DBQM+xG zTQ1Viz!8`@1P%?kry6`E@bvtvp4uv1C1dIW(E#LUF5pU;Z?mO~lBMEj2CaR|@H!f3N&r4YpH!=e3`RF!ooX1{gE2-Ad z?m`4;s(g1xIDvTW&@?ugEIc)xUPmCu!`0ZJTTxHtMP|_8g}JQg1`WZ!jKeQ%G@k?v zIXPB1RWPFm6f3`74py4UMT~c2dhGI4>FfdXvErM;o|35$P-Yc#O>wc9%F|l;6T+XY z8)*~o!*rv2a*p(1H!F*ZyeY!xVcS+6=qXVwqS|*?WaQJ3i3!IMS542@GfQ zNkc)WzR)??TNh)|M&n>GPp?P_ACHN#+fW?poHmte-E2EcwMJOMkO+v)qRdvFYdK!I??1=BM7M1N12Xdc8p1R@(tc%oyYcwtI^{F9EGJFj&;f#j@u{> z%9weQl5Jy4iKMWulF3Qh+KGd}N7Qi7rxu|sg4;%Nihu%PPYj!tAU{Uvl6PJ-GNFYw z>$X1bom`iG_ zkW2MLE4w(}8UK)a?>PjM?cA2sOpp7+w$DNMe$B%hv2n`SCO24otyDppQ|c8h}tlt#U>a_T_1T1F7F#6#GGJge5P zNzz}!U$n`k$4(KMR1JdA=~$7qlNgUR6H8n?Z>ST!%qpr#S!~Hy9xG?T2x`Uig!deE zw1p~KMxr&GvSSJr#_EuF`8xmi01wniJmPBY zfi9F)ZgZCkzn7Id~GS46Y$Dc z9B))Kx7FFcj%tplAeNQ;#fc=7^wIW$n0!NJ@o99ClZ+e}rM%|m zPy*u8ODK+XTRMTQaM(%eTl*@N-6h5>b*Fu4W691IPO&T{KjNg0c|OpVmf~-p{i-0x zgKF6@h#&CQ(qL_y>~Q!gM(wM_gzX0qsZ2S}SfUKWjF{`=7Ydc-A7-C43UO8n2B9fXjXVi`LK!B{^M;d;Zs}s)^PVmpr9Ki*&Xr zU?rIiR4H*Zn}i+AO2=k=!0!%RD*8y)Ihosz9rYeucjpx}kT-80IB4IR-+)E%sC(kK z=A;{??qj6)5kU0oPRQ3c=4*-$YAOt`c;;1%j-!F4xr#%W8#`DxMmDf&R zg`lo@oil1k5@e>3kS$D_>5(Lt=QBN#?;bh5MxScbZ_W@=$c>APx7bq0(Xkqn& z5fz@?LtCeCgvR?(#yDL;`=l90k)S^=hZ3<3m7ru>9M60J3Q}V#>IU$aXg5bk$3K4t z1pNHp*ZqGEAC6}h1_f>?25BNl%OGz87%U!y zTV)i2@#98awqY)gUUb`y3oYSn$QVt$hF)ND$dKtd5T}8a4?R;|l*#1lWMpw+z<8?8 z`u8dc4u$b}V!4J=FTE(8jzN!u$gz*n{;(kM`*uVYqZ1-PRL#HgUoAWn<<7{jx=>r_ zHts?rZ?zKM;}0z8CJ?)B9b?f^el@yg&}la;BO>?2KzMxJHgOtS6WQm;2Z@NCstT%+ z4YMZ9JDE;CY1|1nJ4gq=27G=37995ay*#LWf|8TLVYk>EZ6Q}pMDlFXSRK|$WpTJ; z2%L(0QjOktNru?o8>=(NAu^0*wrdS7WZfXU8 zC#iIPk|L0BP%Ip(-@l*F^%#Ib>jFvq10{@rZ8IInh_8;M{mM@ZHW(hun^l&QO&yF% z7JWi6NmOujzM24Fj; zMi7dbqJZUV-!!bvm#?hB#Hy&tDr3hjS`^}Lokf=0HA=8)05eh#GH$Cw0_A%QM!^&$ zlp-)*;?N;=gBMO4#p8=b2SRm0p(AFY!pee%Xo<(*MZ(HV{k4;s!uE1^Zf4Gk8x38M zJYGJhYYtmDcfV;E9yH2rsk#Ca$bN^iV<24Mi_~XdREB7U?H^4b8A&7r_XnZwLhGpD z-EMYYqTY}#lHpH$r|$wEkoHSf4)M!4`0S-u#Bwef1T`6bYtMwS@%W-I+swxcuzVec z)=pD^QcMg{T;8lM;ae(HCBw$vo{8M39#q+8t-a!f`liopOd#=LSL|5~v$?ITK5-v` z1ptBMUsftj_nO~NVt0?>?CoF&DA!mnWn1^c?22?C8?K4px_ZtHUa@yny^@(eec+xD zYo>aUF)O*nn?p&0`N=T>JZ&9AT8P3kz66Xi(1>5I4CL|iqpv}|9q zSR7^$;)>d)kk4j2R#rn`shxRa`N!9S#fK)OHZQ+6mU%^`A*t2#)y)OjRcDL50PF}8 zBtx=kCoA6E(0p|ryBcIm*5n$uK^sP_ulAvhXm*f^jjXDwsQyt_I2z?vt$65 zK}qr9F8bp_?3O=-jGsStI%l^{+`KE+giO3vb$ zQLmv%Az9T`A{z($!e6X9{B?z5!cl5tU9b1K<2(YsN-{|d9imf2WihI1amWEy)Hz9; z7CS|>9BND-lQdAxtYZ3}qRDP9UzTaF^igur_T`0byP&WG1X`L@F!}55YH%xj&D-5T zsk&6tOEj8k=R3(BDq$Q#7Muq*g2f&jlly13(9(ghjQp{R2e%N(0)EK!eWX;RcvN#L zw>K7k?N4Jlp^EmeQknW=V?m2qsYHcz;Em(_tJZ@F7X|( zi>B)>Yolql(e1AM(l_=jip%I7QK36LwYuQ1BJeI~l+~0$t2cz{<7DJE4g1s60eyCsDJIRr~l|(1{4&O z8cA&c?3H_~y{Cc;ui59`j@R4#6Ps^KEQssj{ado5(#8lY%t3ZItFVvEI%WMW@M+l( zUOW`{OS9cB@V7lF`u7pY^42lPsLlII2p`Q$uBO+G>t#7m_ zF;NaQD^nO08<(Os&;f54wh1NPD)*YR3iwY`>X7)(Vjpm^6E~$Xg83BjC#k!rF6^=V ztn)e(qCkuq{jdqsUl`Lyt%A}1h!_#(H@JkIqZ~lkHM8@z1gzGYeH>`D8v=S1m4g?b ztPO8cc&QKZbD0r-Pa#LIQ%EL2b5bN6;uN>u7?`F0PJ-ciCjiK+{p)H@5)giU&FR26 z2G5WJa0kz~GwAnGdv$`n?j@0f#VCtH`U181!&YfI#`$KP0Z{ckkO`}+d7Hzd?7tR< zWzhxyP6a3dDWuMr2TQdZr#Wva7BiV=Ud;_Q#NY+hYGRSwwE)R;C6$ofvhY%(i~0Z&%B%K0b8pRI4Gdk(t6hA-C!8YTu$u!L{Sz zIO-~QR};=b8RXGYri4?*Gu%6>9XWiRkD-H@CxW%MHWCuXPl!FHRA}hubNvKe;o>FK z0;g_A)Tp|HOrn}c9eM`)D{~_TtqIXPd_;rXqXGSRd|Dt?oY{*og+W_L>wyHWIJ?Kx zAaS0v8$_$RWW)j4133q4&N@%x>IFP1fHAE10gjC{`z?lwS*@aG2Mv#Ai?VP|$3{7I z#iTVfkjM=LpC9DX(ch{uG`DZO5u-a`l^^x@ZzNF2=OSD4Rvs?$4yGwF@~aDm(-HU( zjAKM&3rP)nIE-L2Jri@hC%C?p9+VXEHrr@bRa{MOs84KNWFeR~cXmkP)`Jh1zUjnq z>gCr-08#Q3m^xJX!trTx_$hj?u0S=-)B5^&HHn*TE8&@P`coz_55=mxsCn+7m#fkhyy)`^S ztC*a3l$B+O_0s3t<(RwF2}9QAP58-_K}gG}HF;UmLB9q%ViGRAqu?l0*X+>CzZby( zkCJQP=zgvza#mAEG2m3w79fw(#f;(3%@V9yVpV6pra@#63ndUSqt(?ur?ISXB7Xc4 zCpYi#ureOR!wS?4x<;tj!Zo8|H|S|GG0r1->{FAvCu9;dM_S$^3Hhgi17u(*2gL&A zNd_s4ov8|Wj|JuJ=|qsApv1@)psNR^&L(nq?IB)+Yp{;F&O-S8sDGagJ@5!QS$c>r zWD{2#k_)w?(C`F)pDjdRtPEO%Ze`^i)$fwj1!*K_4*?$R&+$C=UkVxMZYN|+@*e?B z-3KYQUKK`V#tHD8?FVB6!!gWau2!bwhPGD6#kF$1SovpS93Pmb+ zbMNvomjF7(u?ekv*&MCYTP~S@dB;`WGM4!{5o9EN|3p3_Nc?1R6$q(i7+FKhGF1o; zlI;nJ^I9DCoBye|uR0xzG2Sak%l#1+&5-9SH!wt2Evxif%xQ&ch-VF8ap+0={y76j z)#jf9?u#42R4C%R6o`c=-}+P&v!MxtjPg1kC!K+*@a-x^*H-SASK*XPQ`(+I zEX^rF!t{xamVa9Gog1X^CLY0ofLZTie;oPC|aH+7$B&@&6JVGFLOnAzF! z-x>!#XofEqGEjZJIdc6nonj1K=`5wP_ISElm8{S|ocl_V%a1%mt{)wO-ME{nCPXv} z57KDSH{XlV`$%uXlC2Kjqia#K*}+WgEQg4DI@&&U%-n6Kud|D1(^y~EKV(T#l@7|u z9EetTuU`4b1teGiGxD;Dkruk3949f%CiW6_G63K2*tAo}omxkRL-)J|9z>kLViU%9 zRPFgi%CbM3a;?$;6xl~%g8_IoU~O%##2MzzW-2X>_T3w<@b)T>(43m1`4IC-x89ve z!_w!Ds#8$1|Ku(ZPrju^1{o@AizZ~{X7LkIUbY037A22P&H^&?^9LXuB@w1RAZojskk-vi&~zf2MeW3NHRam( zoY=4TTdUdBfRB~p=mBYIhJFX`*AB$0gVs?2SNI}$hEDemyq*BN3;i@KowHrcn^WbT~KY_YMj?s0)#f14FY@_e8!uoxq*YT2RTa61!VdX0^ z@Ub2$$il>2a+rGkX6wUIW9i`eH)LXfr3i_WWfwX@h@S2iA-B!6PxK~QMs+_@b59&^ zO2Ds{-p*Hk5Uf@iqFb)bPcpeA>i2n(IETfawjAVJdq-Z%y61@JciR+E-v_9K$Unvy zp;Pd%lm{~zioP^*kW4QR`G6>AdiI2uJHhRdb7&WQ?{HiErdEKKs&% z8fs7Aw)X|5){ubz+VSO1>AeLb0s)ze{Er=|Tm^gPYqkdTJYDnB>l4sc<5b8l#679G?jOF#$GcO-r_Rq`#gsm%&`XYqrjEzSUW)@OjJuPih}6V+DPObxWhqOUQRYwl}{- z`N&mK>0jhFbUxNyF>XpqPEh!_0H6M^Q+p{EmaSq#v4t&}0x(eeu1Mry_}|Y*Ydx>$ z*?~>;DO{;HC79 z>brpO4~(yRiX?dh*@lYnUH|3C+_@<5(UfPh+!xcX6Re;a9nVLl@8fsCdjP_e{dXW| z4W?5cu8Y+AK(khzSj@_Id`HKD>|C(;tWZvdUGk9Fy0edi=c5dh>_A_3=HQ{;n`A89 zl^kd#Y3fJJ^TAqxHfBHQ7 zxk=gXoi4wdY{FxcJNY-aohQ@mxWNw42Fj?u{N8?4R&T4fE90r->vK(MvyQ8N0@g(iplR!LP<-9{x6w*VzdV-gqNE3JLq&2H{?dM0 z>6L{5wk5s23%-B`cvymkd(t|9+f?jV*E}GIQZa()c>Ug%!w{l~9EPIl%LAmq%VN^rk$Jf`3%Wd&kdP2Phe}*d1tC&)c}KS9H<%+l~Nro zStHdWRUpcdGSRUB*GV;DY(DMsrTQvX$7V#3+83R~sTyN_kf$KbRyy!Ogrl6VS$o%O zkm;A}ss8OqayXHq38q;5#_9&fCY>g1y}D^Gc8e0mnXkU=sES7L_z@WG89L5vNbVi9*dTB(QkpY^?0xg(qw**>N$C_Np*!l&*OVXH#LbfRR+fz$fzq-f! zW1l0wMys~f;iB%Wut`S4_9inQ_TL2nO54s;L4CQ_!hPki(gaaA!iee%@<8-?rrZc_ zQ@be%mOb{2VlMT22vn1ZhD0h?DC-=+6X@372MkM`#wZc}hbh6{@=Hf^u!ldLqtAv~r>1)}N9k&1-J@%M&Gy0F#?xYw;c9a*R-iY>;na1;< zt+!qRvy+gtS`j7n+AK4sH-tHJvc_MwGlMkWW^2VAQJZT;R#@SCwnLR{9N5vNB=sDk@c^p`JQ|#GrkBycZ*GrGtc`tmC}Mf16|4t50w%ztoejtQqYvA(PQf z6n-4}1MPi!0=RelqC6-=+=p7mqgvdXo|b#He8lU?596dDk0=`QfLK@uBqrTf>Mp8A znbb0JXU0YI)1$(3Dtn{eQ*f{6yaGioK$jKakB5n8}*M7ae3x-FsfR5E@)3j_2wg_Kc#3b| z_SC!L_d4Ne5#j*5) znhpY)EJA{@5jIPxoiwYWazMExAptMRE^-$3mI+zVL362r!s&k$di&Gi9#&z@4J zM;plqG0KRGb_w%ut?vqAuSh2DWkeNZQ51 zizl>DaVN!lQ1i`VECnFvk|5+z9T_MxJ)3qC5vQ-V_alNe(AcnW-K1FqD0IO4YoWHS z@%RjsPD<9r7$0c4x#%3XXK#aP-5W-E@EK!`|AcZAut3^7y`CWrueTyIGh2q-rko^N zL8-{o9ZM_0YYDfnH$angHtOy?GcRC5r3^HV-pA0y=#~nLH~k9e861{tTnEFXp6do8 z#nc8em$Mwo7wW;M$8$h38612Y;kv=;&cBT3Mj{U?j>fo0ih>lCFo)uf;vYWKZ5rEK z0f9WDQ-}FPwgV0kT;i+P$irw!xY{;tSDh3*v80G!skP#C3npzL(NEpA%E&r$#usURYQ7GVh_Fc}0(+Ce zDaotvetq5d_xh;BCZ@tN2ak63qc`!;bd3JWcE=0v{*2L}YV+aj;k@gk$&mHqQor`m zJ4tM#^zL=pl1Q3%Wt!A3#p)}n-7?!?o(4_nq&${^THEY*jeJr}3q?A1W88g`npR@E`z{WM zyubN)F2omnlo~MkFT@iF#3BC1JT?4$Ef6jHgq;)(3~+c|OtwVmlp~HsJD0iZ-V7%A zQ^|}Ww-E^i?#;@Q7p#XGa34Q`kQmIEZ=l9}p}!K_klD&~oE;Uz+&&nRTcrmNwzfo- z@^=zs8lYZxPtXBtFFPjtt6X0}MVZnUfE6uJzpMN95f{5lK!eFFOV{c=3I|+4*!w#J z!Mnq5&HkCJNME4HIMoz3Z7R1$i&Pj>Jc>({sZ(@p@eLv!2^vO!6o9=c@m4em-R1Q*+54{ShZB<#MO4YG;Wf)rgJ2 zzNTY4S|$g{+L)v9#6)V4c9k=?Q-F*m(tETvuE}l0Fm}qLQ<;88&5mYoa#!7joW`wB z`M37bGl&M2I_EP-0aVdQoh8+3$5bd5NH!!QTT8KJ$mJr_{p6nz2(SyE9zi$(!QXsy zYagws^L?t#6toBEZlg81tTsF?TUriEr(gzYXu`aaz9b02;7qymiPxJmcvvE+P30>m zd)aZ>rQG|KNkC;>Il$76Z(q_HzW;}s5HM{4l}2*zXy6Flo9R-`$1v(Pr28n6ntx1eEreq8hHmQA3Q|yP1jS$XGxSk}I+%|Gh-9Xo&&>Vjm7l7^+ zyuS<2jY!~bU~G9a&+x<5z+4bayyJt+tmCCpTojDY(oEd~wT`tDD2A1Zp-jVu;bRT9 z-U(JGH{QE7-ZZKd)|~pLeHiwC2$XJU)y$fojP|z*J9<2OkFOW) z&0MqIT)G91J*o}=*?4#qO-y1?7qht_Q6%>#I%Ozdl*uTl%?Fn*7l&qIk!cvi!H!Mx7h?VScWocF-ffuUNIt7m=_SYw~*(663y40LzEMMGfj-~IAv{3495g{3E6k4(N$c^jZD2$s_o@ z4v9Kvb)KxE-f%jrc~-bdz;+Ee$lAkD%@OyX$w{H*7pL+XI^6>5)h2i-p~5(hNuSWh zTK&F98*Cy%s)CMVA;S=2W3cm+F4@z{8FpN2I)2?gN41vZsa&HL8hVpoBO8$MP%`-i zks!tBN!f`66Ae9tQAU^1$hBK>|M~}t= z4ntMb%--ysF}e?)klcy};lb&0MN|5J4-4Afm zgw`K4CZ4gCP~43u7(w0YSaWDV`3l?9&#|oc9NJt1c;%Cyv zQTj?F;w0{L=iCloZ5?h0=Y@3OB4V_YAwX{Z8eY6_`@} zjD|vrmL%2c^M=zRF%f#+w?p{52pR2936>p^r|^F}$nJ6J zCCQMoCA`tT?(V5Qf8IAzYPvf-i`K;4ri8$}@cLKb2HR=Mv0RxchX(rk@~M`q`2Q*H zEon_XB$oWWCGA-l!1?E`ekitF=m~R4SCLP7#F1{^E#Ivxr5sZrC08V`qorw>`Od`^Qw>I6v>1RJSzsD&Y*7R6`?eLs4lG<6{TkN`WpkGZt9{6aoEE8Cha zW9+y=OpaD&=i3%4{&9anp>0Y*xnZna+vJ}pWp|%in93}_oE3a^-RJtVEgtOCB+PzO zB&v&r+Nnt=-L`@My-Uk0NSh=>lO8FVwPH#vi8wZI4`P$IQa=?*CRX)rTfQ>tZ!U?2 z+?4OZa8DkXAnW&P#K6ds$%CArY$#GTxOP7sj8&X8J9?{D+z-s6&`LB3}kpOtZeCO zq#+Q7l4eZh2O29}ouE;iLs?-hsgyKqPf~n8K_ua(O(@ZxtFMuA%>BajJTw))1(7qE zPOCS-_|QE%NDHq%KyXj$(q>3%nSv_gr7k~H=0?d04I#!#{~hWGXn!@=D==cWrTCXU zD7HCGBR}z>9_1}cY$Ahpp4?vF>O_b%%s7s2i-<;O=p>16&iBa=m6%@;D>tdF@`?jZ zsy<#|W*I3^XF8&MHcg4pf;!LKY%(Dkq72iSh9_(F9tI?!{{5?b^o{KZTXMxXl1K)0 zqDxE#RAS1v6?7>dsj+vCz(9N^x1HGxqe)Y*kBpowR(@FTNY@M%UVkOIlqv!pTQX-DltinFyhAo9F*6RffQ5V0 zv33k6S?3i{dkG&He^Hq6)dew0&`~k{WE*VYTYcI1q{q$h*#g8f94cj3cO#B|UW?l2 zajQ||E?=CegOAe+QB4BWEtD|;kP_&1vr`<2B-Dh&wDp=y5URuIxmi#W_|Z5cY$KFOzR=5E$)DQ278@kIZ*?eW=iGKk9L+mvV#yJ4Yf+B zM18snCo_*6Sv3buVaDHDX3GdgD&J0^mESOM>$165D-Ripp7mn$==U0NdRei-Zp8x7 zif3W-G0EISc1@?vK#wuQez>V2;LJ3H3m2a&Y8(9(x#_PNRBBEkk^IU!IF80iC~#^d z_yRTu&5vpK^@>XMig$?D3GcvON``rnw=3&D$o-EM(Mpf|9WgH`7bwSz8HN$@-N26O!dc#92g##p8LEdPn#uWl=+WU5S?A^yJy5 zIVALIMs7Iuk?*ebqWE2o^E>G$jdk%fx=E=c|KE3%EVl+vlf#g|iC9YD}A&j&y`>We61{#s(>pvBU zE$a9vMyWDovIt@$Ez<6SNDEFzfU^`89Gc6+joX;qL~IOk2Q*DdAq#$WwA2uaX7~)v z%Rmj|WZSJYwb33x4Fmvq>&n8ARymc!*4O}cB6k;(gu+$eP_zn8Tn0?}mQ3>Nfe32b zyG($tpnDARaLM?;;m#zO*f|PRM}NqyOj#cO9F?rbuUa6KX~FAwdwJ3=HZCJMd^CbX z_N$$wDm*@3c^S&Zgbdzz6DRBD9((Gbf0KeK#&+E!_OBYu$_5plx^ zGV<5%ypWdZREI(HF4A)2Zq^ysX&lc6(7F()QdK`+f`#gYk?9hDO=+w(bg^4hLzo8& zv{c|vtGr~8rv9sQAkn8yVPLf+nZ1Mq3qqEW*|zGz z>a}1IoA5SDlwFFrdp_shMO)uRFbkovGAD5hfoe%Mg#k6(G*xyL6o~D)*p1Lo@?=&| zswYi35a0%E#adML-uyLf>kj4V2~r%!ElL3`dr4))`ujO14GUf=npF4vLaB%(MoIwVHh3(x~)4=6J>`9;;z8Whin z{g8a7pkwx3a=*5SZg{nY?R#w(2z0$lyyb#$L5%EA$eP9J%T z@F|%MB2l>-QK@#qruCstP1&QbKqT#gT$*=JAj|B z1n*#M%EJZ46K6Y0e=72;ue zK)i4uHnr%f;IcKcEjM6$+-jWK+)LeSHBz3+wIH?Jxik1yRqQ;1vM%vDk0yJ{7J||oxoZV z{`W8SrEfD!C`-*RUl&N4!m?ljW6_ZZ9|wm7Gl7Xo4pk7{g^6nc1&&ht#3H-{v)hnh zdE70E4izb!PH8g=N6OMFOBNBh_XQvw#u0EdY5WFefP)7}udn<)Q-}Jz!jCHRY(A*o zgAa8ghWvU8YgoZX!oA>fsm!BHHw}x2W0Rn|w+{ewQI4^Eh@R{Bg}D^mWG;zT&IiSd zS@GWkQCz^>YIk;9q1%QvPbHXw8`G_UF*lInfulQjCr8Gn%){K>(Bmsgxk9FN%k?9= zE4I)oxLqSxu|zhDLk>GqWDs6 zsDCkqx!(h2=Xv)-140z}W`(-eTCBN$I~-$&ece4Ll6jQAzpq~dVtBu8-C46>|NT7Q zypW&hWP0&tXKD}r&e3zv*|>3}_intoy4=~brmt(Zar+G%JvkWWloDzl>pV)1e$5y5 zxf*bARud9=EPZf%o4dMcKxT^hE$H;cSffI!Vuy!IBY(b=^vFgrcZ_V&;EDkdH6 z?u#Oua~lK5i&eMR5o32lA=czF-QcJ4sE~^=73T!ydVe}Cx$E`0-?Z<^Xx*FpcMGvL z;mQ7@65%)ZUx?hAvt*7Nyxv%{Ky0*I85sO*RyGcUlF6;va7mu&?L#%sp+@ZZVE$aW z7rQLg?bKaJ*fjp=C&`|mdAP%BV0Am$U|cf7i!7t{O+An1|itkmva zt>ksneU`G`srCTnX_PtJJtfMP9@)S)MebqsfU5pbSKX`<@FQ5&x<|KY3?h?lk&|Qa zSEJ)t%~&xla1S*SSfoJ2p6O4weEV8ItpR9uVO6#T#SgsIeShH3Wy!Y+`N{{tzkIK{EjidE~2z8R0Xc=jITbSXW1Se_k3~OhL%(GMeq=O2? zt^kk35d`hiBwRcoUf#gNQfQFXB-V+U5nf`u1}t|{vlUiTAOCy ziXE%!&?0v6rmMqIoy}$C(+0$UBuY7hKU^P23wQX5nEmwOWIMMVc!b<(!wnJCN?)($ zObcPiE5{d=x=*W5mD8s$NO$Yr(C@ZoBFz;}JF|*O)2*U58v$)uR3BH(mumv}u+Pi6 zp*?%mUQZQVRlvBQ?P^TEmQ~0!j?6&CONl}Y?#&C@6;Sw|9MCfY?vc!kz+#zjSy)7R zb1kQDOu;PyXj-a1vXr3)!RROIY5)AzpRpeWnrZO~|=chE{_I zyE0_4-8QdgZIk$LSvo(R65TLnd>xKMX1gcSz!30imW#S8B85nMwCWaJXu== z=q@WmSWp|Yl9&u=4#+BpD?bq_dHBnJVPLe=!G*c#3w-SJy9@;fuS_Q!pS)wL^lA%b z+_rTyOzk?Wj~E6wdh9cIbyzrLZ2huNTGq)dnG3N3t(p($v}TQEK#JQgnfa3mwdi;2 zJ(R9>3PVXMQT5oFG`zC-Ggy!;ZqE%hUgyG6$dEbWm4McP$IJ10!3QS4Pb3lQnCMKHh^ z5?!p?w|e`!Yx%G_MA_T;MRV~J_fCWR5IX|y-UTiNPxQ0iR5CFT_e~_d8)Cg??hJ)xd)qb6w zEp?djK5RI{p;scm=7N`UMAH|SaoxS0YO|(gMbd$F7i`!c0+E-8%-=cQL)afGToXDM z+DMHZ$aSe0#YLSVC)elFGpT#WaK(PV5_84k-TaAjE}8RyD_F{#)!pQsV(>=SJI?*q zJDD6=O5y66QNjiR=lccES-`d(8J-IsU$ZoQsd^*^XPb9G+m5>eF+UojJG--)f-{sx z)wqaM9%!foXG?>5SN5y6$?!BqgI&3TG(DSnJU@dVaVrG#;yrG)tM0jba`)d*a$9c! znO6;AOP3^k5a0eJ4Zj+(wX4hw6r^=SuR0aJO`2U`m(vGWMVoOGj)ZjKz@vvpt;7VHdb-K?b@oghZ) z&d%SA1&Y2n?D}VW%g~`|eQ_x=X1+Mt$$UTaPSz>}RSbxoa<-#E2_K-&GYst1%4Fn`QKE)6`*Y;R zg|jyqD%_|q9~20mx10+1T3rN3h<8@%u(VFSWV8Sv=>A2xV^4g(G|T!(75Ur}!^;ul zzLB+tcFy`3zbT2)ipr3P7Qm5$v#Y;4j6dM^s1^5+y|O3dOa=PY4k1Jg%JFI;WnKzj z4oG5=9gC_&2?#Fu1=BChxJC&nR7=VsExIAGmK^D~($_8bB9>%@F@3?yM0VKd^Zmr^ z+9N1H{iHzEmc@qMVc|n>wfaTamQA3owDmE>-9~JyqI40pspJB`OX<%h-?KRvwW@UE z5_sueI?TxBkrqbX-z#QMM-b;r3#xHa0v8NaR5*o~$CqMffjhnmt-7S<=b9YeMyO{h> z?0Pq6eM;EE6r*JPHuXRR^DBE-&UOsl^v#S2{Gj76Tk&M(@_NqQTW9&vgZv$UJgUh* z;t5@w$5T5~Mw+gAdd7I5!sXm|_j_QjG)8`!-Gf4{aEblSkI3|D56_E%bIm1Hf0<-a z`rvvP#vPXZNia*@={Dno6tz}jj29!&>r69cXHr4}?cLR#J;~%GWPzIeLK&oe>n~=s z!k=}Rf<3ofmQO!!iW}u^E}He}7q0bb5$;cNl(tz$=ow7kaiz%}YV#apr{OVtxM z!rpB{tN!5uwC9IqVRm+#UU(G^M2IH*^OANh>n8UALE3tPT?-I(d>JvYzgaQx z6~h1awS-2atfp`su=$9lMK=b^-r$PDpBq-}1@ao+N+W(0g|BEnsHJ~=* zZw-(1$N80pdBg>{N8{dbW0>ulJ!u1n{T+tfEYzs088-2rWi+Oc;D zP(%%t6Pz(-57puAh3=Lnp;bTM2-r=>L zF5>Fc#(p+Nl~n+LPaVo%3DmS4?&dK(c3wYxY1kJi_ohzzux6~V#RtXMXsNi%_xZwl2`JZgPD8j@e!WtUC6#Q;0YNw)tY+L8W@O-?4*({Z5JKqI5 z3k!>hrz{G2p@3%9pzaxRsv`G4M_i%(sjZ``XOO;9snR@hZmeXy6|fYz!XHiPBa6m? zg*wPvJeS05MEf|&urYQZpkUgnJqHHl;vQX%R-qxG(*Mk;g^|@W6uUxO&3=$EAR+pl zx#8U`@8O5DsUia-QlmVMm7e9qFyx(d&L%Zh{}GpCmQK_!Qga5Hn&1w5L>t&Mwe%Kh z`;p;;i`otN?(>yWk4K6<;Te%Hn6nmaF0wMuE(L*_QMTIwi|iGkTHnL0unb3UUnbH{ zi2_hL$g(W^UPi9FT`EkfCe1Rq7>4FH0SKeGHB+}QX0pf&PFF^*D? z_{t+L;3RsbR2m8mL}@Gg{AY(Q?FxvvEhQ_1!h+@>y-OAs&VRWP#OG74`kwpy%d~0{hQ(rY)VBMG6n0;SuJ}&e|DraPT@I-Z~ynn zXW<9|xL_(MG-3=mGr)phWUAE*I&_oOYGA4ekT(z`6n6bpoZ#*zm|$qVMwVB8>&id^ z7mn|iri(j*;Y@OV!YG3(_)BTM$>nz=IU-5u@PYd$n|7pQxRjf}0ofsT?H|;2 zhb{)`m#BzAZCelTQwWH`F^Sc%?nj0+J(|)<@B?9NZI5i*uHc@wFvW9!r)x;G zL_cbrHgPK@DH>dPXsbk8U?AVFY9}Njo@H3f&aKF^F@; z->OJ3mNiob_z~ptSoP6xSenwlT2}EXOb_Y8tBzIib-Vd)iG64LmE{zWv~Ze#KN!=0 zZKaO-7a_QV6mQG_n3tKQOShmSyc?1hbE?twxo>;;p^;%aoS@PLP=A^qY2AHv6X84;J9!LU)$u&Sdyr17`q!$dtSj{vJjMUs} z_aJjD2!*?a!TnB?7E`gwZogH?&wVrirE?zUrX%AWn6D{w{&c7X{0+B? z4~}&UZZ+27K5xqMszNkgC^w_!1&Iw}0O#A=#}Bk~9Nh-2^HiuVRW2KZ|5`=9u`$M` z2>Uz`QK&qlb17D0Y)NkHbdq}~y)YGR#?!+>-E;XPOaJ5gGvu#CFWi04o(;|x?12%t z+0;$asp!}?XN6-351`~RQ$UoCd6q6m7d;%g`)v1lWGlJ-p%KuD(@}?}%Qi>^sqS>` zVxG(*@Q^M${ej5w6)$CQEm|PtDZ~*Ea|*z&qi5{qMcu)V7Vn|G;H_>%7;8-Rs9shQJweLY~ypo-*V{m1ZJ)2(A zxqQ-yvA*7jCOB%;{A|&*PjF*T2ZexZ|a^xjk0 z+HKK3PV<`__QSh3NG6EVPAQ1RE5?xsHF7Asv~7V95bM%iyn$M>z&5Is|3HC{r4HicHJK9_1Yj{O>vFxG!d z-ovv@`yS+jFTr}u3osD0k*QywJ0g?cC2{pJq_C^{b4*>Gbj`PS{2s2Nswk55e^CyS zc+1nFZuUJ;Zx`oZAQMr++1a9rM*Bg>_`Bj#r`+XPvEyg^sP(}>`~1zPR=@X%VSX(_ zu8q50X5KZv{NGXC&tcl$L&#mHP&AGa<27FSDePYP^qFNtJuqetJ>vV4c8zK5Nz1l{ zAbYBAfi-+~fjVR_jU3&Be)Xxv+^v*l`vqb8Fgzk51}Z2-^O~PNP1Csdr+qA(|12O{ zEq>BMi7TC&zW=K;1PKN~BnAcm@QwJt$3BL7md<7tzqO$z6^&mm8;bW!4Z%E6yuQRV zS7Yy{m6|(bI_*_3*_E%lHlIYJWFQf2y#0Rimdy?!(*Am0|Wk%*pt}+39I% z$gWytc)V>DlE`(KqFD4={5IEIPnFEf*mrpgVGrz7pb-70^5eG?4die0B~(vav6X=8 zu0teEwUUJP*gY`2MrW=dA%Xm^AmLxkoMX^qrez;%-w^<%!0cMbg^EvimpC^}0p?zK zSf4S+20m_UkvLNa(~3JI7Ti{{K&XEl3G$IG zqoT6YsnZ@Tb$;fPH7-oc1@m}l&~yAjAtkk$*rU1J1k0f9DhhpGGl4ie$4p*kZ!CU1 zxnB)w3TVew!iQN7A5>ldB9@Kvr9ufa3{y@90aoJ#W;W6?WtLq3y)>mOBl?zpN|&^m zH0c8UhNPt{vc-(St!TRDrL{5xg=Nal4Mj%;D;6@1 zqu`#8HYlT884i)MY@C03>47E0=$r-RpmAX8(TFPn!$!c#C8-X1S#`yCSeLm(m>wR@H*Vl$H`d?rUsDy=8d>o|W;9I>mp&YIYKb4T%kt@a`RV0>_k>89GpIq=^+Bz0${v+ixZ6m@iQrSC z_{JPcUvyE&lv5LxSE*;NZK;KBf3v{6x5;Lx##xT$0P?_17mmTr0<$>2IWTjE@ON*- zJ{K6WAM-X9eM-Q;gltmJ?68f?erFZvmJw6`#_r^5i$O#DINR_8SWWIYqfHXJd103) z3EhOR;g_oJ4>_bZcf2{s=*hoXxo~Ih#SA_;kyTv)9_g{e(PcS6Wy7B%UolsgG0z4T z9I=899et;mSl~3UuU>~9KhkV7=6Q|%nfC1bEAX_^994A7{dAMoAK45yGK%3M2 z19Ejl2o0x(I{!vSU)&1P^YVmxGXp+dC29i0sxjN3vznCZoP~QQY&RYfG#4#M?&~dh zzH>tq%+ZT+pD9}5owmwPh#kPsRA=jadVlYGab-WYxKuxsQ$Mf^T4g zI~?cZB2xyzz=az=07yuekVQ5z+`j1!r(iU55unNN4alpN2`vncdQutbL>jCwNV^pIwI zFn-lh%O&&MaGVEQFK7O)1r|ytR8x|{v>e@H029y$PkI|2Vx>FN^X0E09^^MV{=QGm z@M3r}e3KOu;Miv^fjbJw4U77zT}IubOjn|BFy_t`d;m=!D%%;_+LvnVD_ zb#WShO>}g8|NiWhj-;Qxp|Dpx@p!))6WO&`m~j7ZFz7+zXVdv_su6Dn0D%6#!VNt~ zz2B$L70)kpi>>|_x@FoyORABNzj58X-fo@ku`or(>EwKKR)+$bOb})fuJ{&u71h=A zHNDFZ1j08Lopn60bX`UYgaLCL)#tmmzP^6D+t8j)Zop(27L3+Rxcgfh&1o4OuEwI=)sQ+l+-RbeoplfW%?kmrU( zX25m(;oAjJupTvSHY(~aRcOxj64$Nd+r|)+k^3wAIn@;29X}EzREiw#)%3TuoCX+u6FSr6)b6J$merMRNzG;+iC=7)M~}+kx+MA{ULHnG}VOL6O9$eOETPh^y!Q3^YWt+zxWD5$Z|9@G=3(v+p)^8 z=&Te!SSXJ0M@k|+bJ5uR(EH0`vJ)r@xMrkvK~e!ysi9|xpil5p?Pjx)6qS3b;b4pu z&@Z`xF6Q(Wu)hG~twv*5^%%(eXNG_Z0y|<<5b5li#;6!Nq|;cTU&~m5;s;mX8rgnL ztfU+xV6cyJ?*y{e>(Npjbi92?SHU-M>%g_E{u+QAw0Cp8U_>44&P-Gkgdt!-WmPYefu&Z*%!7m$HYCqn%%hnKrozO-z4H~l zjTd0|_3kLGZ6fOV^<(9_RTW57BmZ%OEgtawO+dFk!?wYG=tUzhl7N#G-tUcFsxJ`+ z31<-&qnT5WpGYTp?(~Ylmm`Dsqd&4~*<2r!71V9<|I}rZ6F!H_ZzD1jaoRJCyW@Z| zSVv5+((5H#z(4Q*Ad@0Pd8Vc4r5P2GGa%FOdcW&EZ)Eoi`mo<;J z0rCp3af^P!uZ#$J@F5QYcG*A~xU^Uk;g;dLTs)I9p^%LffQUcgyR&G zz_rDV29Ss*>K0nWI;R_ZxBwS}cK)bO3}c2E8HjH4C!Lh9%Pe(X49`2)$@E4QToslk zyyl-oxn4JqtHg0HYcl#(im&H|r%sR8V@vT^ZrhS3XBxW{s}vb1w0oK)e=A!Ro?m9J zb1-r|c^UtUOV$B@&rS~>*lUk9Gt(@5TnmOVSv%1-#a8}x?nrF9Fh`JZ%{MIDrV9?U72Qr+rULxu0b9zj}*r*L@8fJ zB~TnzigxiBOx!wP3LEF$8Y@4xp#WQ9`~pjV9`s+t0|32oh^kX0muQOhiDS>tfT;_{ zZ4Vz|*3UhI6f(LuRq%R&U5Oi|vscwwP8-5LKfDcKIJ#W1H-u$;FB+2uDFGdUVMXF} zzr@HpTMfHR2>RFDw6V%et-E=O!Rf$v!#(PH+&zhy5_yK?gPY=QZxy7G z=Yd5!iCo|&_~F*D;^{4A>LTK*lC{_tT9t*aZARNGNm83~WRpD*PG>B_h6AP ze7SM;e7}o7_yI{aO`&!8fp%8`h0^J(Lb}wvZIHa_d=*_i4gou08ZO?6O(uXQ@-~e) z9gJ-cI^?)Cu_DeXQ2C7@3^t6S5Bp>^IL~yFQIBl<1-M>~K__e>tYdFiuI;s(P6)V*geL(15K2n9!u(SWvT(0rbc%d%mqh@S9*Cn#pHY z1$7n)EQ_cTWW+T|N|Q5-6@(|THY>rRZC9-cc#T1Z!ZgLOjX3g@C^xwrxZLhDPQ^ay zMDO5kWptntH4SepOZpok(R^QyKgf8A!1)}m2sN>QpjAvkDyBTCuTPN_5q&L(dLfaP zt|Ze05?st$rLKUTr$3b=?iCR9OJzaK8>g|DB7(kvo<(ke!;T9_@`Wyp-qGXbM`mkPnoTEEQ0J>j;w=GE<^P ziH`}!64Yyo^`X&uvg8bDSe6cGS)-FUJJAb*mm#j7WPQC&<7HAgSGiF-EaT1NCU3!$Cq^ zP^l6uE?!#>mVq=CHs6(&)k#aPRb&W*(zq`B@}58=rfzHj8?;{0(wpFrjmsbNX??T6 z(l{Ml0`#RoL3V!$Q&F$M3rSe4Mea!WyW$Cje?8GTkX6s zeYS4y3Qk|DwW8^Ufit4!g6s?L5aYzjzD{WHOdT0`9e^41u;B1y^?oxxET+u>*cw{g zoPi0+I_M+CbK^e$Oa8TXdayU;l=uDpeEgnsesg03WAO?w4^;6^Vz*{aPwO6@&bg4P zZOi*mZljtqsiX0NHFeg)%jaN}ULiZm&Dxc<*V$C}S@^I?e8jCH5wJ32J0v!mvyQW- z?q7G_n(L)$TtVBmB<4SS$a(@X6{hkr=MMlQH2pdqqy!7|8Jru4&sDpS-G>W@SJ_bR zEf1TJDF)%@S&5I=LEiNhw+;*n)wPV}enS}4DvX(VdtS&MI$WY5TDu0UJUQDXV!(;v zq%pZfbA&ll|G;yJMYuALOKX%_x&dkON(l1RxLFr-Ew*|iQTp>&T^pO$;nd~+9d%

hDq7fp|T%Tn^E zHy0BcLO=A5?!g~9SLv>on-}NKtlvW#%>ex?RcI!Cyngp=NW(;JxFG|#YAcsu(J(AM z_J=bd3dyMyQ_bjYwpX@mgl%`|Dy}kQ9u_&L7D1k1qdj8W_XqRF(BpL zCp;HXYNhm*HB1 z53)+qL7C?#ndV+7z*p4MGn3k$9j3;4HjbhX8DlA{)^mm1xgZgBX5{yEJCPTHC%Lt# z%?d{BE^T_|G-Wnoz!f>?r!X~Lydq}3DU>*J=w1vHz4)xCeTn=TcS0fQP8B7hxurSJ-7M>>RV>(aAaERq5P$(m4kY>LDXJdL&rc|#y zulwiWS;L8bzEV;Z6}V?7C+?F_$2oR%XUuo)>dxQIe|`NnABCT6v^`(DJHu2f=bdr+ zNq!aDbG>Jr)2rTa408g}_zMI?Skl_X1DRmZ?~u4>sn`)N3JiG zJoh^O6;>c=b{mC}sn5=j1|W|T+!t&*9YGTyUO(!OZ%hX^hy2I^VXhVgH((A?+s`Gr zh%j0I(AzS!VEt!co~8jsj$!rK`e)a{2}MtlzSk-X)5uj49@e`5fxKc2#U~sep`eGHk$4TVfWaR zD5fk^qJ*c!Cy3Q}6pVS)A{1vKWEd^u$=8aL=nW@GhS!lt%7(dEg_y32e9vHZ7k&q2 zo+ujzU=SY4+iHWh;r{9_C!n&bf}!Bn%g z=LKa(;-^R5tftfCyNveR{`2Ph1tP}SVO#dYH`z_T=K5P0QDZY~yfMA$#g-xIq?5J! z?cGF#u2wAbC|am!$so?jqo>uRg5$ZL&6l&HL>FR5Z{C0UIe_QD383LEHIRg@?U!c@ zw`G}Okg#?TIIE&yD%0xBGH3SgF?`YNC+MF*X6ihk!I~Dq0G~j~)&30samGZ;=Wf-c zV6>LllXXeSDP&n63Scv|SU2MVFni3}z&f|gmx;o)zSWLu*QRvO3HJK6D!f~~wA?=c zcv8MhSf~a7jzp>1=P-2xf9bs9>ClWN znh*{`3QvDAWy+)e4%EIcjHpLfd7il+APwp|PxfQrvh7U>jc;7gI&c`Y*jaYZ@L*73 z3J69%rLAg~IoBT94B-^08om8q!|*43^ee+p5@MHqw`$&y3?buus!nkQzVCfYR@Ei62 z=sKq$L82xLw{6=tr)}G|ZQHhO+qP}np0+vL|Gw?UZp5v6sOO4|%)IA(=e+~WZ8(^! zM6o0>-Na5w@?BWE{|#DV_MFCGY-5$xUi}iR#T)*Q7kUs4ZuN;p%K=WsDto?X9ixO7qNzVJ%mI!V6}bgFyb;np=q?g@w-v_3FU;b2`ab)Gh!Wrf zGqAswP7Zm1iz21>z}$u|(@k72dZPX0Qd?m1PaP~mifBuu6rx1> zTU}h1HN=McKa=?-G!x|$c@sAa#WZM9O&GN2AA@jIUI|cot1vB$BF$FLFF%V1IgsYP8^Dde-9| zu^e>bh8qinX1knU(i9$vj4Z4;(sR*w5npPIhU)E3-waX{pp=nS+{jc; zy{?p9dYN%cIUh{fDDn$^D~8;A3UU)$Zq^$9gC*m4#@8Bm^w~b#)!ZFi+dWN0Ov9Im z^68SX$Z5VtYABAl5gk#TCZ zb6kZmbNp(*3i1DMOTV#!`|ozN1XIUuoBihfRLy+}{A_87w*D1CTjA-de#B*~@WKFo zyqISr)ogh@k$7Ta+Q!ZAGqKQ^D{0{gj*Xuvv5+F7SX`2qoju*;?&{A}DzzE(#M<_H zp{bSJzoP@YQ?6Yiy;bw19#qFifvV2@*hfh>Y~oaEx7?m+{g>nWqrO}okFVph>TP4& zktTS9!D)xpT+`IYzwdl}-(NpRA_9iRoGd^5Xl+YDs_HzCY|v~o>uK3a#|d2gFDrh2^$lzrY4xu$9F)S_17b*4=k zq!}$5B)GBE+?=n6@PRU-sDATq06p$WJ)S@Gd(R1M+x9B=oI}^yX4Ge#Uc7nqAzIHS z3Ky=ncE=i&r|8%x>mO$Ifx=SukAqLH&s6JdvnBdc^@`f&H;~d$rX9~!;AZ|I#4U)4 z#MBFMQ4M8mNMmCz;=$Y-Pv~F3kr34#o z`w9K;AuFub7^WVIUvEmXH>cNO-F+S?ejN*&CubWc`~!Oh6X-7_d6|#r&QVAEf2-Mb z%gw;c7EmLh>#IR#`?w>EAOEJ@NKG_lZm6gtI%v&%|L{RjX}H*!XlnB6;hYr}>v>)T zqd<8x*b#f9g>HI#sqJ3D!#k;JA|y6d0@=AB*ORJmt=9tW8$t(rF@ehjiBQpgsxGW| z+J>~$E6m$Vkx&W+L_G(vWZwg;=L>^lP! zE!#-yQfBK_h862XjDiYh0i;v9;EtDNav?tyi@s7~T9+NxsLpB~^1hdvfpwo7jD;SNAtkc@omOo& zaMS?5$Mw!{6t^??z-DZ4FjO3V=V$iREfN(02EandhX%>I@!44{2LAN_yqR2U>dx$*iR#o(c{ut zgGmXwbzr_SKv^GhjxnIgA|r|Qq7{zhmR&Ut&kHvaIv)hE)rr{aFyVgr$wplFYApEm za=G{q9(Gu9mNr0LH{ls6G30qN(!d!CzSb!Y!O%lfd+2OTR~J%*mXj8Gy;c5#%74AOnqj9S3E#=#n&hj zlKWPtEC4v@=7o-7Hz_K>fuX^i=vCtZ4p0&g!TyK%l0mAJz8-|?s9O?W{fZ$VRtl@e zA%nFBWnkz{-twi*Ugyn&FUk%FfwgQmWcg%3@>AoB6M|bO4_nw<8>PH=%r!tNHqQGM=iO&`z{P&@6ru zk>%y`-%l^)iNVT(Uo5RbM1OXOK zG84E9Ezp$U0}L*LBn;ApjD;xCfUE6F68W8|t#RQNLetXY!lCkEp6hUgfAWRQ2tFjC zo9J41*W9E+Yr&|%sb~;4An30A$1UP;yfQ~0jx`U;D{)!xd>(`aUb$XQXrb`jEtpWm zxlV1!#}0ugWb@BNLYs5ye?>z?v9G<9*6h1G9wY;RS~rO7&z|avpdo4T-TvJjB9I}Q z;u^?Gf;Yi7I2$j!i#d1yD9v54h?YsdU7cN#4UY)ci|2b_e{dHD>=PK$X$a}+MfC8s z$ABBQAH>DG>{Dt!olEdLzY}h|o#e7lOV?!WCG|rM)EhubsUiPnEtmM~qJNPES^-fI zV8dM;r81=-xNk^HmE05<{@K58^G14uy(Os<#OHx|BOfW|FURoGlSBkDd1bj4b}T<9 z{`j6|0g-@T{*Z+gM>#eMW!x>I30CA(P4{@ok=*}lehbpjQtt)?7Zu74drx-^3r$4v zEKpOhWLntF<&m0%-b-8FpZ)lA=KF<&#i!st?9(+w~5^uYbaHQXnQ&; z1NG!Dx6LO1Qn|A7)p{O|Ex_e}1?ho%?gOcY+D7gbi({TGk3rM{D>D`S2HzJZT{GV{ z;T(4gSGwbl?j!_4RPN3sBF+EOP4qs?9vKPlIULxEg79DmW*^Rs2hBl6s1m=>fD3cu zY+?rqfjECrmm1ie(8@Vc~_Z6heIsP=sh6cRhU9dnF|~2@{T$iCSHr- z@ii25o{j_7NrgfH0BecMf|pl>z$EgZ0mMbDcIOU{7*qR&FkKhq4lUG3c7=e&hPK8-;vUvbwlO0XP_*XylNjc20v6aL{u$Q2A8^*}=n;am-q+=#lKT z(c(9_$I8+R0ie|0`&m`kXL~Bd;LA=I%u{Xe5KX_9&x*4R_p1LL1_HtW@$m<-{V`r; zh>-tr$j`107wyds=B_Yon5O%u$!|W!TI!pO&+p^<;qS3Kqu<9n6+T}Q&)YhXBm7?X z=gZA^S&;~b0gW$c1$+nm!@ZE!HdAeoxn03#u%_!LGtA$UuLmdJXWF^ASipBeKA_3N z2PUm;ogA9l*jyj2?VRYh0il*RA3iq*DaZys@4kWhTv+_tRL*XnNXvXrx-py^&m9(HZms{WoKW9J?ztJi6w11Af$8|>;_75wV>Sk$ ztGPJbZrI*7VtMeQf7H8MQLkL5UFJ-ZtoZJv$a^LDd_KN>UdaEa(clTDhKS`i*sbXo z{*3#7-Nl=@S{Ru)JDU6_r*lQ^^FKMAKRbS)W=176Q;BIjog7n(EbuJqsUR{@Kust? zspeYcx&)MbmR)|IvB}m`iH-VN3i?1WO2-MWJ3DS}%1s;A1n!9^B*h8)ru^v14ogMM zdnJJim8pDu?<~t7F`FTyj6&#LITQp7B&5ierh-0fvZ>Q~(jxddc46n`Yo8Tf%UmCb zF{$Fxl$~=01OgJ~1aROvJ$|pJH`hTKxXAtQh-E%c9|uHQeI2}g-9GKgA3Rjy`1VZ; zAWK+AXknyB@%Utnzw>9#d=T!cqyg|wB7$tY%uoXN7hOSgIEH8|(>`yEb^_ADSV90E$YW7N!0rNQ_MO2c z^d$1(qNou&i2wsmAu>z9R~c3Rx&;fCEkC^Po>mx)}hhy-lFMmfzl2Q;v zf6!u}Z1t%U+W18U0_3gMsy8ov{c!Zg?{ZmuA6d3zum`;#Y%q7+_dAsyk(uV%@_4p@T;Br+Iyr zBE!6kC}}c7X8n3ewv^{bbmw4RfX1drC<*-X%L2x#CI{>mmMCsWE%5o~=Kex#wk$R_ zIG5|wiX&C#YlfuJl*}W1ek1-H;*Q)2lNv&#>lTGUU%nB-kJ>y>%QH1WUE%H$ZQT7f zx{^pMm8_Jg`-yWk(_(2b6!c8p{;6j_0X(S?%|Af%hz6{5|L10*3l z_rvrly>#!HCnFK}10gCqm2^Owvy}8(n|;MXsvpzMjvD&bUkL4KaY)44FVl7n=dXSQ zJCB-fQizXISlJ(8b(V*xsKvF#fHsk;BqeqqRz?>|olJ68KEN?dh*|B%!@;)TG>(b< zZAcvy11dRNOAX5JAjM;WR(u9AaOlA&l7)5ukmsaUt}X@24(89g=wN|#XY*;1x3&L^ zuoXctbtd{}$_p$mz~*OSe_dd?11;h9sp_Y%3J}^jHv!v79>@`1#zaNFKbqh@LyvZ< zHA$JJP&b>9*Z2N%AF|7#k@w>FJ!9#Ddog~!Qq*aX-*O#rUHvEYVn0!fy0Q2-zWDP@ zzOnI&Ka4wo403^KUrkeMXi~t$@=Edic)eno1aqrz!DNo~6BTe4MBXtuvaEdWP+eI( z&rQBYJyY`1l}H0TY()`k56HwF){OQ$Wk!!ZhjsZtEnZ#U*$&X7OA21-;k{mPtE#N3 zvGGANlz#{E6@7H*tKtlyvg)9OC89q*(&rskZ7U7#OWWpMPM1!(+;RKGQ^_5R1IOgH53!)AZ$ggFi)qt)iR z20d8D(!l5J<@lRH(`Z<%AvaA|kt zp`cu_BEX&@Md+MT=r*OnaL9{J1bVp z_(Qz#2R#Exm?{?0tj}i*eTqI27aW(t$G-e%=l;h1b(=EPWCely!ke2^+Mb>)j^s#J zSd6&kUT#E5i3>dvIVKTC;>I;nggv273C5E}j$C@tLv##zLBMQ^l#_$^XM=}>*CXiW zR9qV3!&+7>aNh#jL2&+JEh|n8Vtd!6o+j$;;O{d3ifs3oJn|EytIluzOOmkl* zMs|q=nS!`wlR~OgtCq0oJ0yCd$gv=V0{V!Xh?c=Do77{n4MfCIZe8&%lsgiQojJve z);T2+CJLn4o9Iy^QH0hxqAdRk*np(S0Q#L%QyIW$Td)mT{T?Mue^1~4$en@zR(|hB z5GIA^7hkKB9rD!#V=mCC!wAx%NgK_0R-Vs6?RSaml5b+Gx}coyC@PjL43TOu2=aE| z$QT#V$xAnJ2?hBI1!_6akI$f>VNMMdOL-bh-BnWGOs^Kje;sEoU7UPs?k<%ZS|-b& zt;9)cn2SYcA@{kU)0nyD^1~fff{9G1xNFe{`Dpr-TS`*6F4gFu)!U*b%FFM%7z}HV z60k&IZsoRYGs&E{T;=_jMk07e;5(2ApS%fI*{xf2=a;swTN2jPIg~mPhFYR)m(u*T z&x%%imHL#9b6+cxT7NI4nZ3tpw=h_vhS99NOv;Ez!pXnn)%@!SResLjZA^2v>7D>j z{s%Yvfrzv4P-OberoJ!I9)5YJiZD|_dhclb&;>bwxDz>BYTOd`jR&hPjEBxt^5XH) zaOe+}a@(sX5)$fjQMB}OOan1zf68`edj+(1!A74MAuR}!I;4K=(n=!KwY^A!V;%l~fu z^M2jm;PvrhW@Bui?6J?buGhKlchV#0=t=_z>}wvRA+Fq%|D&_mSV1V01g@wsb=tt3 zI2yn1`u%+OxvI;$?#R)z*LB5rz302zb^Z3;36hYt1TQ|AQlb6bb}*&a&p395G#QAL zjC#`J#I0^en5kHQ6JDG_&`o!|Tk3r<>h^x`_aPs~0`)9zl9tm>(sW|J*bp^g3E64# z!_2dTGhAhfxN|}iC}1A=eoR0Kff0Oj;lYaQjwONPW#HvUp3I?@(&EF0ji?K6jCceL z2Hn?T`nS?ROPRK&(Z6d{ZyHLqSQYheovnJ6i9-&m$}z7)Lzh+4`&

WMxXfU6=y-Xie8&Sh$5h#}Sy>abhVl z3{0;!7LU~~?%7WzDaRx(eFf?4!G;fWv8xc7!v!)kB2vCRy~KuA3a#$75?E{x z`0}wFq(9DF@or81K|~%?_ko#PVMuJ^!ZAXd{7BAx-exXxh(3nQ_evI8yvAU*RnrsD zDjKl<^EgVQwkO4n8G(tjs_I~w2A;T)E7$ql69eCo?|I68FFQTX7KV&P|y)OPb)xZ`r*1i>trtzigbYZgwWGXEZ-(Pq8b|}B+5=QTW`WL zUKqNF#1u=0DQt-L;o3nSm$45H)D~-sVN1X`h;m_o|@(Z!D1D8MYub&>*7Q_tv2 zA@gw}DixmuVU!CbQTZWF##11liHZKG{Z$j74r2W#tIE}nt3T|n&Z}=$54Gm{F7^Et z%*xE|9?pvTf5$#Iaj65Bg8%?9|4s(~7g^TW_P@!psy0rCzaznH-F*1;crQl6u_qg> zOi~d)y)@;7K2r{dG{$d$d@ijQNkZ_|lwI91I>RVQ5B;_CCA{#0Wy{skVMxgz>+*@48e7TV~LBDu0h*0XOrq!Bp%Ks~h^%Mo`Xor0ksRZoA)C z4=}tHL7KBySTRYne?Cg{?dTdCF?QJ+NUu~zKJ8u5j?77rZy8&&R|H3#>3*|n)+5EXCrQyJWNikYI>FdTIkSvfv zI~0C7r*oHKtv2F?*fdZdjgN$gw6=?Perh(zFA`1OF^jk}e-60Eh;j~{TXSR(h}S77 zrEGrIlmx<*`Hm+&pAI>XNHl54T(fG0GpN<()tXlNuKx!?^P}_q_Oq7gy1&=AA@c>L1LY~FOnlP(|rG$Z4vN6L9+gkxk{|4OIiA-Y} zxDjEJsqF2nYoML)9Nw!UxB`P3aIsXrQ3xmxf(q&N*=6$j7|NxBHyZ-LJE#RFP`Uju z;Khp4ghq3wRE#7cyFHI`j5oyz=U!wfYWq|%R3K??XDO-eR`IO>U$G;W*(;HEsUSZ0&|||QF@7Lqs8d*g9IRVu7dhojOyj@E|(Zv z(a+e=yazw%VUZY8bvb_twNl8v`yFc8=g{na?Uoda<6T?!D7RyZ>F9f-C2q7`CmS03 z$2#_l=(Pj=D5eM%#TAuHca~g>%=o{LV@|uV+R<>Gy zW;1qP>6qVCTxX@48ZL!M9UNUZ1qrW}xvFm9LWdc1!ex#F4irH{!vfSEjCw@M#JXkxFTZrK*cILYb`%Nrm z^9JVVaMfc9+K3io{3RU_j8r7glenx`A7Dt91W@M04Kcbw*}Ex(;9#TcH2ZTjDhF&+ z^;Ed){!nTGPps8MocC1SJ>fU(Y5~+9S5dcLH(R>Cb19;e>`y?TFmQbA@b+9YsV5MH z*TPZ4jpgp`ZX15VlifRO?8*!x;P6g;gBoMMmq1I-E_d)Hl{kOZ-c!A5GKUB5(5_F7 zq*PQN(URP9Y|0qOV8Q3akEU#aCEW9ky0xb1WBoQjo{R94x2AN5C>tbAbej4GKZVAV zRu-&)F93yAUZ4_!DS-`CHaN8^Z-QtS-{ko+w}(h_yxfZAPVg^qZWlNRqCu<#%sju? zp5zLe?5WhU0ehlCNVFr#_2^*JVQ1&7zd`?J8(vQWLaaaz0DzeJ|2W|6oNZ116KQ|# zW2Lnvo_hEC1qGTo4rZO?{My~bvX4M#kxSSQbn*0y|7}$MlR2KKh2D=`;`f;=l|rLI zY9wy=+B?^ZA1|r6Qn6K`bf)NHX<~8`0oXU;%!1~N!^-18K3IMM(m-$c@s#!Q8r+*B zZ*7DTEQ)-HoqH08WJOX)=C{`vG{NF5wsxP|P!~8_H8a!4s%z8xF@PZ*^QMV$#^gqf zw(33<9?bBvtM~2f>abzVkZT3ahom){H3j_R>)7q*?&%?}!9~=n4ga7Wu-SejfiNmc zk;{!4I^bSw8v5l;D0&zx*=K+QAs1b}{w($|qR$BxWQ=5kQETTFLHsA0E`>OPU=a*u zBg8HiLo)*14x>%rT-YgNWWem8YI+uVFBW;6@Qy8-+5z0+C5TaUn_-rWsQD1U@jEP? zh$LmLf;H5e{wThYTV@oPW-|g?$Qyo9pljeU%*q{6hDK`|l^{Y#?t~A?VXqf1ANF&A zgici10ckIWf_^CSPGHMZ;gJ*1Oy3XFYgu0|_vthB?D=yvc7gGgzOKez5PuqWh#S=i z(UZu_7a2t3?QEbT`~-PLkngFeB&PsCm-92Y4`GRWI35e@5PM0f& z@BoAD=0H2?mnc$8dTrZZ2r<#9CCywb5r1`d&v)Qha<7zH#Axzp7H6J-49Dgc(-(uk zu_D|#xz>(hkN&4UqPROYZGFuCfl&4eOLVIe`YAT=*uURo*tHmunDi0^B0 zZ)ORwTgoH`kSDNN8tfFU9{hCmjGm}mAt-sj?!3-KFdrK(yp1jTP23F+D_GcK7QA=v zTxiAk-Ln2vt@Y?kA#x^l;vOzj#9CAvk=ECN@$*=oN(AECA1<(#UZqo1nynJnZUmut zR95u6;sHU{!RgLj6T#MgWqb-n|C=qVqMd3qU3Fp|3ZSM|wX%C&niaFrXo#Qplg6ox z@<2kV)L9#3qKpF%(ypBq?M=C{3aE=l83sg9^v>aVwX2?QvESTbtQ`~l4z>H*v43GqFpu?FUIhLOmM7V9pZqQeF$mA9BUQyK7b`l@(gRZ zo+&ndiI|rb)dn!;SZ$!Q1iL)@JK{Ojz`|PEo<|9G+>Z1Q!X~OK_g{Md)ph;e?aZLh z<+ZHY3X%R6^quZJRTQi#xaKBcVh;cUpJ81XuxhFQVmBta+UgE&eWY@* zj&x(N>rBt$MHJA}rZ-%%VzmiN@9OH}VlO_i@yfFMimknJwwZT{sjbO@LNeqnyQM$R zTzn?h*|O=p{`&x|ljZI+c}or0>n@P5NOwrL*PgH3>O!RzG>B5Hs46hz%9^Ky2uXt) zpO2_dg_Z(E-={Fal8*qK3{rHPoU^xB#&t0(+u4(Ib5oi#>S=ZTKKpVL6=yNmP47^q zDn|8iv?_{Bd`6HZNTaN zL}+6MY8$mwm%d|*oSBxGNlj(zg#U)(6F6&n#OHa{Vw(hH_3Tf2#(r7jCflv+B?6g%dgRlRi|Fi@4z)K>g$~)y57DlQ z;bw6tM8N8@j5cNh6|IX<+NOFggu5gkoezzW3L_(E#nO7YBLEB5MnVrQUW4^|R*)VX zpp#!Q`ZWeB>ef%7ftUJ1R+7yqa0USRTkTg9x zzZhJE#1=9x=%!D7?h9-q8_)%Lw`faBVQUf1(M$rv|7ZWO6sP-05h#tsZg7z5JRl}8 zUGfMi?le|q72)L_Dg(Blt`U5^0xDhV8qSer2PkWE9lgz22oL=RFaVYWw<-|jw;up_ zCkOcf8$w1<9!(w~akev70`eUta5IW|_E%>ZD4&Ocll)~AKfq$rz=#Y!v~Li-?h#?A z1_0ZR{X{dN1O%6*2Z*)R4kI7izTtr>-?IyYECfXfQNcO71ExSw+{4Tv@im4j&tS8G z0LfgtA-AWb^z?vDKoiHbLlgB*;Q50^>wq!2gotVQYioegpct9^P9lleMaB$ehNZ@! z%;7Vh&gDMk%TEc62uczlM%2TBHA8?a(NR;$x1R*-@i5rciHL{1tMy}0`G%xMv~$@2 zjDw2MDxgA4xwRuipBqiJiEbZ`V>yWd2%NS;Q;iXND~ex}GjYasjDvlp>ZAw3iM&!* zos6!XcmS=3P+$(m?+Rc+2a1W}yAXvdqhZju!B9scDf7mkjtU2CD`QbFdJ>)haRSkr zd7Fk~xhJoay1-w6j|B*I0$zKEdx&4oea9;nDleR%Txdg;r$4AmdECvD6j&vOJdO@p z06{T71O}X}SZQg2hH?YD^kFDqg9N0(l{SbghyWpXUdPhXy8sbT5V@54Q`P3c_4G$7 zoqsJu>x03A{R1LdmHSorMb!o|$ZEJ`6Rj3QmPo$=k9W-z=^Q$lSY)a3<5EFQ$iU1a zah;C#nT^&QRfj}43(^RZyrUYGl2p^evPhMIRu?O$cAXkNMaD@^-yFp+P8{_=rl_}S z)+BY`x^yS$cI6(z;%_};qOW~PY+t;O&CKq@_Iq}~#Ix1r3aLt;_W;Bitj9=lo2Wn| zxT>@W#5^W@-+GQdxBkl5G4VLSkKEMJXu-8VLyJ)%)1=D^SIl6Yf>m^CHBuE9$6J@4 zR4K_0fA_49@e8iPh%H&47|5eTmn7-uj%n70GS~&m2iEcz4*k_FZt8kNm0+MK5p1w} z;l%5TO7!IPl<|Vh*MF`kV3bs3Rc=sME==SUL%>~_KYX%PKfkf2&xT-)Y-BbA9Xr_- z2Qj&vOj|Rf#gk5_r9ndOw<*Z%tx1J<5|%;ds7~NW)zgy|;Kdgve*1)0C{n$UcRTI? z$S9X1^UJ`FMR;521`N^CwBs?}0>QpL;R$j6@Lr)AtK@&x!sEMkJMea@!m6?_xyBZz zWbLLVRi>?7-8=RT_p+m3Q-QB(pto~v`;^DU{!EJ!Fti$%t(vrXd<8QV*3n7$)HL=# zM8MjqPrlIGQ>XBLmpF%PAewgd0zI%rHNm8XxeSOl_e&rFcSHCz`+D^st zEX}D@yCumf$e(G5mbSuy%@8U`z9}28=hx{Q6%pU9s#9wXyvlQ=bUC|UCsC?n*H*$v zHh!?y^NS$$IPsInj;`C<^wlcM@!1-PX|G-qf$0w%ha7FW+53a#Gj2$1IqHW$IwM`{ z*i4I8kg^BOpe6Xaom`%ucY|tG-EFZrV4Lr@x$SN1W(?nq;M-UxIKI)2k;^PmlA$Tj zm!1jP*G_ei3sa4Jy`eha>Y-cS=^rKZAeb~wg-Eil|4tq+C zaYYNfC2}k>-$++_0cf|S>HQQct2EvW>=L6XIN)}}gLA5VoqdvA@iB?$K-7z_mWc(& zcrb}{f&AR7G8VV$W=Q5o=k`rWEiQlX#skO=)3QFYr1quso zi1tDXP~K)#Hkp*s1n^rBGBL}`zWXYz8vfM%b~^rYP2r{f$^vfmE5FPT9T!#cn~YCnS8^d1#L3*n1XS*uj^z$=%L0KnoB4uj+3$q zq0PYuhE{DgdAk9P&0A4v94M{e=Hk$2UBM$HDS0s2%ABi8ZfOT%2lbZ-4%+>61vDWC5Q$iY%|}gtcl*Fla5izW&+yXRh#UqpF~* zk2vZGxi)yA9x3rCh(vI~Yy=5uj!ARDDGY)p4!xtvSmu>ktF zGBQ=|Ua3bD&sLgnUZhbNnx-9AU0qGFm4l7h-#j`PzTI%8QGf_O`p(iTii1JYaF3;l zbQlbz;X*OD3_uf*!Bz$_l!^Gbt2`?3H831nnu%-lDAB7X4Pc(w0~x7o4WD6%o5-Ta zI{`?vLghD8oovZ=WC%?yKNX1(AWW?cPS&% zM6YSI@YiYOr8)zT#|ojx5?jonIev+dRB+RBCCh5=MoC4IwM3RrqDtlBH`ra^aP9NK zrAl?r{AXQ5`YLs8Fx7G+IA&mElRlQ&DOs3x84=o0_vu#B#uBxvGr` zR52;OI{Tr8&Z4f@)o*f)c9y(F)Rr(Cb~!K^xH%RBGy%Y*O@l$)W9+UwSykRXUX~d2 zuzE?|A13btRg(Xd6$Z-$2zrpHR-Vkid>jZxc|1;=CUy}n{tcRmg$4Av9Ob!2Tq#TX zj$8-BYD9ceZUaTM@dx5OE$@)Y1mzV_MN%KYZF^jBSl}xo;vG5pNid4^hmew^xLY?0;o9HyeCTPvFrJxaGW)~=Si@W5Qeu{B-3DW@M>uFT_s>L zINwh1Kn!W=MG3*tvRKA22b;3pzEJAZ$HB!YoRoD$jv3%eW zpapT~9&9<&h1bYR5e(={_k4GN?G4=fKFHqM!z=xOQTpDkd>ddHJR5WvoF%4=67zJ* zU+}6mUAi(7r-XcpGEf{?Igr1|03Pdb?B2`ZU1TT9fVdRB2yTk^=`nU;A$yZ=&3LpcWPFzdJ;R~yo|zBgj*F@La= zf5p>bY9_00b5TG?!yy$<*6bi0(ML#q2k3kIgKbmmP>*)h>t}g;`QVvi9zyK}NNnPs zOK-yMm`Zqny7_>1IFQN&iUHj98_ax{9go<>MdKw}&^}bg8D?{6^tY8QSI`-5 z@m8OeDb7;aXmqwz^fmJ}HXpdEtSYqU9d;I*_ghO%QA6K7Q$yfVRQO#Bn}4PyyG$({ z#4=!ZsaI%pu>+dKC`NY%tj~-lk~`Fo)hjZnt|x_-GONahY=g6PoI|<~FA3{eGS%## zamz378FY5UYS`0pYTR``;U9 zn{sD>0a;kI^$#b+$}y;4NvK^sS8f!Y<{W&11}_`-ig{YVZ)TnX-4J^-ex|!FUufla ziFac8-05E-Zu%gBV&6?CKQUe%0D(c_j%P=GD5&?qJ5l{ke;9AsxA+cp+&|5N{1$lH z$X)(11^Gpgal=`>UJQJuL;J5XLv^chbgOA}&*j?-Sz3AI{~0v2Z@qX1z%b(qEj^%b zPQ9N$d|`ijf*mc2HGGSO33fEPA*_wa_upjmav%GAkFAEX7i-dK6w7|0069nrztPI2 z-c3|U6&x^O+V5A%4;p0`DDM~HnngpLPl9(WuNv%-X(ktJE1C~}&*=pNypH!4bJ!zQ zpaHBHYuZj-cUxjZEq}3H4xV!KbG0jQ)gHr4=p zK8>#Q6J=h|cplYFs(=_fG?_vH)9-)dO7I*PN=;d@IboS|7yyAZXR(@fZOff|0nht= zhJv=CwZU8hGy`E0=z>qOCvpbS5BqVQ0NRF5AqGPFX;U@LH(C#;td*`vkJe6FPUdHo zp+dFkQX8D(>6&0i(;Cw6a_)~DFSp1Y!=6#EsYklD9jnFGtKusLQSCTGv#?_Dj|trI zC0O1G2z~cqM@-c5=Qj(8j#gDX%Kj8z`T!&&E=PI_^^>4<*Qs2>O_R2D&Njoc%ZeWoxa4XQJW~rcuGf(ojY*Q6%h_*$f^5AA}d8E=|&>!dJ?47))x|`I66fVxn|IE z9;MK*bd<1=mT(6vs%jSVO}XISqGrE(@^j~*&I`1eld5^|g9;*wm<>74x!vKLG&$2t z8vOJp%RfA^2lC}7UPOFq6weQ5Vg+Alh{KSO0h#xK94KnuqH|G-e{fk?kzczW{3HJQ zZM$HVC2otmx+Gb9T9jr{i2L1^wffYpT)Obr+*;ihHaD_vROIg}_1Aj))1qhnU| zTytI~zBp~f`9dXTF(BTCOmwR?s7~V{%?a+EbNk%Cou4)8Y%WvePpRzZ%wtk>`Ln`p zrz|RC@k&brSz2aK?k5HoMCg6|AjiSY)8xwbKBICzlC<+Sy;y>o{qO$a<#N}-Rn&hb z{l>JbCr9DGZg~{XQ)QWlKv`@^%0H0ub81iOy0oHjSeOGxl#g5bdeIVPoXEj8LSb}& z-3M4!bpRzlsi1cDSca_N^jG2zNmW*!=9?622j$B==8C6eB|Vx9NW3_hN?%K7UYd40 z`x-9N*yhgcO%Vuc$dDV+vUGF_DQ5@|LS0)`094cj*6UH>Df5sarKz@I{E@h9~lg)pX;NQpnS1RHZL#Q97&7({V>lwn>U<`SsW-b`_6KF7XDJ+y)Q+R$ubIP zBn}C`D3K>>8Q_F8Df;`600sG%x660>(|Ov%Fd%3Y4Ek~vA7JH@Sg=NOkRwfl7@n|4?7o1FH=BQwNUTl7rjND$!9 zE&v_!&5!wdYMk&PCtxcySRbxCx98aT5vkBUJ4B{Pu^zpl4^DK)ONQK1VA0||Nc^hW ztYnb}LG!qgfvkQGt!Q!r6LTVfpdh^yog^x!37l9|VRJB%j}0SUfgK1`hg-B~H*pN- z9@GM$C?Lgs)(7wBymzxR%ov>ao9OD$Tkx4KR5 z*MOi~FqPi>-cXGp@QzPzqtznKct z0)1KA0@f7gVoMD`n?Eu@4&R4qIoCSlMozZcq#Mx)(zE9*X0|KT%0Oxdapd+xxz^5{_`H1>%PnF%dm01i^)eO@!(q<=|y8|_}HqR z!H61bBFu#k*FcmkROJ2db@s=iO_u>_qzEjPJ;|0R%25MYzd^Zd3l*;h37%@$5tTV3KNZG|cY zO;GR$5F}IDV<7P+#}xA@?uKM*u#L@?5WI2B{y-H$N60ImPh`uT7R^t}y5reZh?7_8 z=`wd6hEVsE)%Y(Y-Ryi{+G~_lE?m}7RGBOyY?bwT?IP4j_@>pAr~b_wq11X>=8Ik4 zUusnCnwdY%dfNc`Ndq1`bL|FixHd1V@avcfd)r>EKH@z3(UjV<4qI3|6i;nN#fI?M z3hT(+3OZ&dpHHO=ukYo@PCfSPqFj@`&-YT<#enbYOkla%ev zx2B!n#HkgD@k*CYrIk+h^P@>cy_;%Krp#3bS#{&Cs;2?rE0wQ*t%~B}d$IY4mi_+V zt>^k64>;|#ic!g9!)L)2fdHsyXQGQ=#Z0#B+QsYY_`AIsmIwKdpYGS%{e_nEwAHSi zv~e7Vj+m^LMma=dm~7zVT{HC=8`pqFoOP0WMwtF?lc=7DY3RSpVw1e-_~r%G8hNV+ z9;d@4o2SOO^ZIlx6Wdg3P0uqfpq`19@U8P&gf@exOADJ(>G>fY-%Gz|RI*gI>y!@f z${utBx|tEUScXZ-SkL&HCH@6l%lsm$O)|@OkEJp~_Yh4FCaW~pD4OhW&b}6}c46I_ z8?96BfjWkJrnWowT`rV!g;2*$`RC%xb?%{ z_UG{>$JsIZR=Qe;{fE@dTvy7+mnd9{fl~6yq0W!F2gcI-7G%If`a`P*+Z1P~#As zAE0=Zmao2OPUh(sH%MNuYNRf{!Hco>@^}o$o*#*Zh!7Q@D_gkqXn$a-=Yi=Sgr_> zsl>KewlLe6A^y0wJ2GUjCb|~Jn$e9RAcqC5CvcO6ZE^a+P)1`_r&A)brSDVpXJ_Bv zc`tNy2Uq`C>qZp!oJs9_sHCeOV;j$K1oqjR>a>oDLC(##Gf8sSK7RNluhT=a`V+nS zX_lvNc23Dm;a}5AS!8l^@OA&{9d3j?o#Kc6`?7a+=x1kZ z>d4?9TF)08WpjDK!b1sBi*#TPtYT5R^FfN&O^G!Kg3g&FmD5!Mud|q+>m9w`G~e=< zW-L8R)CjajLz|8*fNA;$3WW3vRluJ~G%4-N9+6AJ_Kgohes=YfE^~VJn=^3CV z=%B<|9+L#lr{u7KDOfqyw83E*$-zEA@-X={TFU!#TYxQkA78D6bNV=CdXSbiQuZ$9 z`4T_2rQbxPYjbd~seTQpfdi}#v%kc%89iuc&Y}&T8$@0|?q(fQB-H(;!sJkndlI2{ z*-y`;lvx}1vd)_ahic(AfKr|6oSn^mI`yk)e+SaTC}(4u55C91dZpWONEF3!Cya|Rb&bG?8Q!EWG5y!HZ~u)E8+{l{JRX)0UME!kHr=`~U&( z{vfo+j5XH@hube4KV2p_68K)MduTDz<);i0>w}(qy*N|m5%n78ZArL`9{~W}0-D7( zOsh*|z=L-DU<}_(;+0WMIi6d;(XA%weY_nsMZU;MUc!X4Z=#=bpau~a#u#m5g-Q`X zU})s(Iy8lMLF#1$e@h2Txz9E{N}>t!Xhm>y)gwO2<8Hs+(i zGy)uiqDPn8?E6Y4$BjLuFUCSvjOLxlV~nk(lmyx}ruj=x3IrA>c7X`}!wh2@UdT^! z=b{2NLBlx4k{%U);(sb8Cs2Wb68b&=mX%f{*m+|cX$wz->DWU;vtgTmI1*R>i>`Cp z7DWrPbS$ec4Hc!|~uM9LT*-iU~N5t$~J zyA^BeJfa6srw~1P#);zeMhmKgN?9(6V}UT$l~7*z{=CX9O zakla0O$hGkM(S|*5I#Hb_;5RVr|&Finz&zy;l71eDdFU6h1AWwiA-SSEoR_D?S$d7 zXY+Hw^G?>B(^JWjoH8$; zl5<^L&@0c0U;mDQp3pYemXKXoDUj3nW_$V+f+~IThMB+?k9|*-2}#@~XZRd-8B|p1 zbf7^(<^-3bqI0~5;?tFu#G!J}EW4wDRN&ug!vop918rSYqM6rhyujSPW7V){lqk7s z<=WEK z^;1_m@S#5Ez=$eh=pG+3mU+dCJATlmNusI(JW|G)B?aT^BWZPueh|Ge=^K#|EC-^{ zAf!%xbgx*Gkx{uAhpLixp#tC1UBgSN4GX-k@^P2Y`fyFoBjD|t=77O(XLYiCkcEs% z_jCUjoKPUEUsF^8c6;G_f#%~617PM|V;*k}y6Y^$F}u9t2-Y^85q$GJ2W|C0j&rd) z!JJ1D9oPF_1}ZEXK3y@uYi8+b*oCKmW`tS094^2+-Mb;sLE(RKnk zs{VCuGmg9dTreDUCN^Hr;auTE!c?LJjz|jMZ9d_e?LTuvz*9`%^L_vxi)8)C5{r_G zPca2kS-GmlDa_M!K?PS#@Wv>GJH}iau4OD{`W2?64!g8@jfMD!Q76#^R1Hzo>Ve*t zg!`o}o(ex*Dg6^?N-2u@X|9!>T$eC#29E(-mWsn(zr-+@*5g7ig(ystDNfo6}MkRfDOLReN&|KTN@Fu~jvkm?#SfpsRMetrz zCF<>{)aKl_K=N4p^jz$W`|3FDhqJK*DdB6weq$>IEfWnVCq>rr84^7YvTubeJv6T? z-)nQKR)(opKj8UfOuzsCAkk-xoP`qo*Taj9`~RH7INSW+lJN-dm-o_WYTsKO0%Xk% ziCX~Qt9IHxHbJwGbYR^4%}*_AcOy%ys)^(|BDu(c{oYghf-?zsvo|~KszfH|6W<-@ z=GNBNXh7e+R|lFOQXG#XX`#pQu$~Nkuf=a~0ec-qSpx*o(PHE_K@2RphITD~w6I|p z^cow_lz!=|(l9X{Q)}=&ms47mi4BjkJYCX3_mC$!ni15O>v4**v$ZJ-iJ|xLb&v#L^dzmr&(<`qLx@WqjfylJ z!7p8ptkK#&9;RypUU%PvWh=R$xIuN+K6G`3Yt8ymck$6I1raqQfs{Ye+J;)HMrVK|T8E^X{0?6D_=u7Fl z@$(<)h`5iLEfyHj#J%1q-i*@7WBt3fbv-ElxtG1Z?#ZRHTy!1ueHwlBU(M!Nb7gF%%Oii07Aq4*c85wB;a7#^QCsYC`me0=UhB@e-fE>lbpY58b_ z&NyLBCyuhj2gbA(p0!9(nI%EZeO9;R#d*Q$VAw%0huUDpj~&v0Ah9_ba5T!+!-Mm1 zytWKAUvbcMPj8zdG>ZQ6`l+DJbq3KRWyxe zw@1B}nZE`=>ik!6OcMAqrT8*WKuSG6rg5`|FJ76ODxOaQsA0(61YG& zV}5;){|Rm1(BYgAjFzI3PYk(`@|4(-;L3yzNz?5H|B28PO3#wFan59$`X7h8IpIRd zK!^}=$TIE@Y+L5n)KFJ4Gh?hFYAKV#&{edYvx-s(N z=*jGxeNKMy_HY0|$b5;>JKS7@n0IsQ3J&e^5-nzFWWR;d^@Y+{z6_Y*ma8JHp?(i#Xz z^C$s|m@6!`4t0Pe(7#f;}!%I%21_AgpvHah9Yr3%mA0Py7D@_2y z-ZCi3D)7aZRn+-IZ-L-X4GC@dij<`iJ{du9Yl&MS1YA6)Ho{bkH!@c_Jgto`@MjX^ zH@wUleyMXp#v6EqLws-#Jj^xVst*U|GmP<`LD*D%p{W$IhAn(>g48+Pc@z(rDoBi1~)+g^0!8<$Z;vJ)+fq zCW|S>TGGAEOs-dP@HglT5#lKv;u-4I@T|M9f*A#RyDW??B zTPwmI@EOwgw|n&H`d{5`(cU&z)K^*r>KmAl=@?hqk~j`vEG(Nt8Ogy6q&RA(#XVx# zjdvCya(Wa!^<^DQ-95LH5o?qtU~3u>wqQ%hFO!N)YO%M z?jh*4+mB6T*$%4ID0*|Nfk}-z&=yNqR-(zXbyXurEhnD;PJFWgmxxmZU37gbwYEcC zQMt>1hpy9Uta++6F z*}PvbpTB+Bj$RCm??3nD9??alA}eV=r8diZ{B zuD@?Cn$Q$ham1z?Em#+13&iS+PPj z7s5hMAR2jHfI9EqNiP!~r52oaT+Jg-?^*<)Vkxhk=^u-8gnI0xVomC(6l2p;%xTTx zMz!k30tU{x(sPos`E(R+0^l!@&fAIL-6_=OBD8i(7oiNI$(tN59)<=`IUBZ7I?<+? z<4frFudU0SB!wi1fIxH0twRhn2H~-LR9Cw4Q9@`@E@IqlKYT5XqL0~+66 z?CXHxJ%^Ur5NngAF_LVDbh?Z*Io&|bSsIAGcQsvlJ>2h| zv*5@uIg4x#Jr3=%mEfJ2%f_rk&EQ0^xN9OVb14lQ-c+~Jlm7tW} zXi7}1e7aT5-wN^*8H{N&gD6?p&mp1}dPRka!tqbg0Jr1EUBQFDe-B5)OFS*~L!$PG zYIEy=hi&P$v*Lo-Wc!B`3AnDYfcvuPpHgA8E~QDewA@yqK%hdDu$BZq#PQh>n`3xU z-Wy<+0Pc(>PX<8rgC18&i59gb04!Rtdw<6LJ~fr~{Urh}We~Rg>oz(SFbiY~PwIev z$A^IXCfb>ZcbtsaGQ_mA2L178phS2GXfg7Aa&Th8_z%#MCf){C*5NU!RqIt^d!(`H z5#yM=&G`r)-TxOGN zC)Iyv)hhALBFJW32MG~*=M9g&6<1MFM)i;W0h2}EA42C)0i0(81~aj_M2ygNxPHVl~Bt+)0Vti^}k*?U)7~8 zS@@Vz#MBg9YWU5+0V+%U=GFoG68sye6|fh8?P`&;$U;FY=*?vSm6$|k=}lx(Omr$F zfKV^~{?7}i6|f2fPU+)FR>m5N@@mNib0r~y!k7R{A)7joe=qvsqAt5^6f^cq+)Y}* z9JttinnlwrUgn`po%``ZLbSXygI`?7#jH82F_1OVv}gO=HH@Hqb@q+EN3QYlPuGd* zp&X2><(|ak5|U^DQvl2+G@)^-o@=909)>!=N9_$Qxn0r-pdpOt6Jii~vPoj~3^brq z0C4^9%%N5^6iyd~EE>VZCb5b={9^~VHY7Y7qP5?1k^cHQ9qq_1-$q* zAL9IRov@!@quv61-VxZ<=bH}b-r`^iy!`@{m6={icm>hB?cudF6=xXb%3zuh@iMRq zLKxGG)fkb<>4YMB1gia`?9~z&R{(x!7F{5C*!7spM;MX{z-Unjl7Pn|swa*GdJGz4 zqPZUd>^tKG3xFo^0rwOP4Fn3XR{IJgTp$I=Xj}kPap4D}+aSjuu**qGWr%-7B}5O? zy>J*Mh4VaAZqocR2@4y`AfN-9qr%*&t%tZOluVbvJ{KNglSgkY~%*!FghU_*xv4 zDuW{x;fEn&ZJFuJhhEwMpyqJn9)7=4bxfz*S|yGWb>{-Xr{6$n)||}MTO3~Tdj(+#g&o0#R%I)w=c3h%qmTuNqZLmNy!NALx zFqVR&zG=Wa$F5!QjNP3!_&pOdlCB+JNw_DbhiuYu%})hiNxRZAHsFfq-5UuezX>kL z!cxE-?AoaAOx<{URc|mimcmN}$lpxk-D38E45f4)xK&#Y&Z&cSwv9(hYo_j6z|GxV?v(SF66ic z5;S*K@sohs5G5bWA5~ljJ9TnBRu6TU$Rzn;$}*6`v#aAj5R3NN^nt9Q*y@&D$?1`Y3c6yEW`~P-=xMOO z&fRY)<}&{BXqT$5Luw@@C2zV?Lr`+$Q<3bWWT)_>+X1^pz@9~_zYGrA;wzxGBhaPv z9iX0r528aMq0&Uyk}q~T8*8D147Czs*=)mY0ZDz_^gO0HdXa5j5(H8h082ElaZ-}> z$*0MY3E)g?DrtN%jYM5aiUUoZ_{%vcj2&I?<5}jD%0dwJ@lSnKxEE^fD;cc2?{MCj z1vN=b`KkhWtSV!ZaGoF^R)s*XLWL-tRu0IZWdjX#2oS@e))w`c9JHoI$LaPlR$eDO zzi^38uLKy8>``|G1X3gvH#9w_ZH;9n4KHqCG}6aXty69-0m4$yS8pE2m+Y^XY>!%>1}>~Z>kjEE57btU`<7hq)&FTHqJ*{ABYH1{FTAmEuG(5Ha1YtM}pd`Q#&G!T*dd7=oBGJIoM8;}yz;^r+;n zcwI5T1rEhOcruO|U!e(MStn?`Oz}AgueqG?Bru-*dtvnFd=SW)_7QHd%h*a%`8>*~ zPV|s^S0&Dx5R8XT&q^l@Pr}pTKvS9q`QL(lU!T_>lCEY9jD238r~l{Witb!ToA<7i z^i>n8_dt(H@^hvx%^oD%OJCY(+Me$O;7ki$?Bc-YBim}rb4H0T*+&^Qy9RUJ@aXY2 zp_}PPZ?3~&6+vyid~CZ+05lau^pHR(`Kn`zuVYKOm}iR~U`Gt;7EvB*H`G?S+Ox}v zl$#K=+-IoN>0p97dteKtFZEQa)hx9pL;9kfeTAcZ{XyeFsJGA~PyN{3bF!B0eNpZ#g19P>#9|T4@RUQ2oq- zZv^IVY&kFic4evgat61?S=dkwc{yGj;ncgNI@;4I-f@TduA32S`}h9p9m|OP*C@O9}f&)PKg=d zr+@%qO68|si95nStB?Ty1@MYv9a4)r6snxmn#rp^2iFM81aXS>J29SB;7wS^HGIkC zZ=wr*I)SN38W7 zM^bes2Z!NB;GA6=P+7}0ooCT6`63>wIyh(Zj85l=J*^YZr*Q`IW%zMx$(lvoXXqqK zV{Xvm>fIL#hjRd9P4sOYoYSwLNez~_rio{GKB6{6<*wMh>*2aWj*wVeo}onb<9@vU zZl9MlTkUdMjgPS615KlPf{@tE$-*N6`_VRnY-1as{FpV_CQv<;0n8)uY zI~PxB=5>40*33pzZHSdEN!GGm#jRZnOxAchre*Sb-^@`AFL27~{nRwX2|a#Kzc%Dq zcKGp22rozMdwZ{)+a_cIJ^*{sOwV}jwy($xxcQ7CD)9l_-GD)xPjA)B`+`J6W32EaDQlvmMljy0gQl|G>yTxh9WY+6-eR;T9wZkV#CeE9%(%P zWung1{QIN2)n#YmZ@bUlu@~40@G!y*R@D$_#wi%Pjjw=hIf?{JH1D$cYn!0KRz|h1 zE&O*%pH_3|)xI9P-6naQ?e{Tt#eO%No0FTjgS(Z*z69Ua-?@9sra0m`k2{;$6$NwB zx@Z_0eOxvT$>d6=yn(QgMl8cbX~yP&jZM_pKUc4dDW#q9EF^+T!bgkN9AF;$iMW5)QmRy+59fOC>y;Z- zm4TZF^UsHcE7!&`WmDL%Flh_Ckb__?ZQ}QLmywr6`gK}x zz~<`5{88;0t3XDD4!3t>R=>@tnCR1tU9Rvy)PstkNV0r*B#jTFR@a0Chb8jS{1sYu zaBNKt)nJuj$GN^5a$(jojgmQnG~&1cF0Mn9rggUYCiB_(J^)dkq$730zfkDPSj-wh zYI8}bbF@ZhfJA552@BD{hAP*T%P9&R)P}Ygc!i$D+RDN}car6haBtJxd~O}AEYV9} z%bMQJbJ~7%%O&83BIB*iA5>U@EyF5sB(^%ZUJY?vILi~N%?Luvp4+>@6P&xuhKg+~ z(mLECnw%5nc&~46@If1f6e~c>dXDRa#eWF8q_~a)4#5}e0O`Ik(r%E$^LxBXG0=;9 zQ=6J0Y$vGq`2Bm;#)&qH>FZGhYGMU zR=f+k=3y%L54u9Z;i+}&gcYtuF!Ho46NP3AvFg{Kt7s(KR#r#zo^FsrLR{pOH^|X& z(RYl?CCiHNX6k4asETgceNHjtRku}T67%` z=v6VuB_n7Bdyp6yTyH~{N;L4pd_K*CkPS>jvk^_kln=o0P8PA%4Q?;Q9m`odw2W*U z=?^VFJQ8bZ{TD2lF;pKHEQgRWL+qu}$;5Ug&)SMgtSr+ns$ob)aZVk2we?DL##2UB zxtvVwY~LgBn#Rw98j_%Bu;-rKif6Xq&q?)0d8FmRg=2%_Cjnx25MHF=6>aYVrTGS$ z_9|4ip42>cCbo?*I;Gddt`2S$`m!u|Rsf0M0ai!sm*AkVcrQz-H4XX?=qB4mCNH(L)?fyJfRs zvR%f6>t1tB8`o`72=#Q~8t2_AhjV?jKG;?I8nI}7PJIX0_667c?5GJ2T%qbGZ-j0& zF}mNFpfvb?XJBX`TVD|$hi)?dYDdAQUM^FPL3~R%`Ptncx-h{(1{X2ewQFniXL;@UTl32M)^~P%VW|nukPob7EpM4?4GVLaY{0sCac(PS8|M*p)q=S66({8U~hDW zNg>F--fp4;-A({@-!UzJkMcs%4xCmvhuMed%-1GT^I+$U`(C-<#~-E3dd@s(**r`; z*yGE==OPrFJpW27I1vlxdhg9H5@r%DS{l>4eM&emY_FBTbggn3`L1DSy~FP9SKgs! zFI8S95y!;@%oJg_qk+jg!o-N!fmy2%Z_0N*Bx#t3tXKz2sLc&WD7}it_7zQMl95=r zsf$fnatyQAUt_?sjkKA_OrCy04<67Mvu1#C9iBdi#ym=qOI*-cAwyzrZbS)y)7~4% zwefDVcPKwmyV(YEpNQGCX*(Jy_9u3~xXJJS>2h`b+AeCfB0<|Oyfcn&BTW*r*xF<^ z&nR&Dg%k5=#=-om(B~AUFLTAOQ9=)2p>vnO|AIn|Jg)y;Ik-)Dz4vhj;U!vl z#77NJs<23Jk$bNqAW}CEj~q>(KJAgogdVmVScSD26fb%F?w$v2cfF;w0v;gQ31Y3t zI)L@51qAyass4(oYG7GE$E}ef)oFX0$HD%w@FA$w8+PEHHEOJvVn4E>K^+A-Sy-&A z?zX~yN_)FLsHMtw%>_mIzTGmwpfvuxC8hH|d5M(@qiY}HA+f4%xHBzH?IYEH2L^ui znw~($ogp8()0Agm zP+3Yq=bxU4?4$MwDTyOm$?#+)&oaF<15g&l`$9tpOA8!tP2GeXPENpEa>R%C71V;J za0>F#A;U#MBRlg=r$KHa7$KDS>6N>%=!nQkP#k)1gpjBS4pi*Rga&F>s3K_eT-|Ux zR`d^=^>l7<*H~Owla^6aO}X94IH%OmrM*z_VF}dGY{FqGKfPb+EYTGWyNGhwj70?-ooC_nAAqk$&>9z6RwL#lM# zIQgpPGOb(z5p&j zwQB}E)J!=B0&q*0c&e4qGvDv;5)k$4&H~2{V|yAMvaGSX)=rOC{P-|1`V~jFKMm^q z3d`xlPN&lgl6AmrZ&mj>Wz(KD3*Pw3PT2`-4P3pRLtS{bi9U$Al&fLy6%F&9W~Al> zSfvRYT~`a*X>dJWYQZ|l+Lrr?=Oy;V4<1xvnc~<$cP5UIqcQDqQd?zyEOdb1f!r)> zz)W3gR^|b;7L(Ub0GD-v=QW69vwo&=nw$P$XL}z}pZuPXpK5S=soI0PN+&B#AB;bc z=89W={sc`e7bM1>VxP7mUGr?mX`9SH%}mI|>NxgTvkoryH>%_GWjLAO;MRvY>2RM0 z*@;=$BTdwPM|HR9LC#L+W3&jZ_v9Kd_|)hHPtr*#UWvNU=Mmmh*4- zBj2*N3A#BuM2(4v2k#n)C{WLH{MpUlf34`&4+3<(=0DQlGkYjw_et?R9uPpp!D__6 z6a`++>7poP+CRhv@8^yiUOqAM1#r)wmd0E5B>{OE_pTg{LGe(ZoAofaODp z3+%yPc3?=SK7FrS$PN{#``%Yn($vlDzcO2i%^V0VRdr=4 z0-T|Q^J)7E5vbR%U+QQzSnyvv!TOh5P8PEF0DgLP{S1%OUAMe)SBXkIV2xY)*RRYx z$?Y|%n#NAeved|7FL{H9pWb7Dltd-jaPf3_B=U+yNwD?Q>B%gog>7}~IVn?Q7?jPpTr5>)yi9b@iS$hg<>XaQeKO?bCDM{>_-e*h&ju06J{ zTcSKg7fvdS?g@WX9u(32^=?+daY)}nTnjk!olC2PH^}ixRJW~ov_Rupdd-}$N^5D< z_^R|&uop;)u|`Dk6DYa)3Dt$pdlVzZ%u!&X!8G~-*33xP@`7p1D(}zZOq`{$NC5rM4 za^ztk>&e^OQRGU2i4~LB(el*0+t2#@A9%a*KWoV|7yy6@3IG6%{|)NzX=7;jU$1Tk z>&Jbo^}+oQHDX6KnPyS0wWUrAx5jmm&E9em-Tex_@KAt+q@ff74TQEz^|9^l79cU7 zqQh>jEoKf4UO#;3px0sl;o$)>n0${$jc7?UK9aj2l3=%(rzjQvE0@^HOVW2khpIF! zhz#c z%U-Xq0}3XS<)4UJ1mf4@Wrn5y$Km5gRn-}bFAaCb;qVA zPScFYPkav|w4?dKhFO702D&iX)D=jnGFJj7iZBronzLT1*Z3j{@>)P|jP4Y6g6cZ` z36ypCwJROU8dCkAu+6H{i12G<-Eh3FWr|4@)LWGkt1wb8b(2II&KYiFohX?K670s5 z2ym*)xSs})$}vw+i?03?OHZ#Ey3Hc-?h=)ku$fevK!=QI*P|&?R^-LPf@1Pxp}~wo z-GQjThzhRu!~11Fnz}EhNzT<;epUoG}pQ!Y^g-8lScVt0&8ho4Mn0x^+(ug#7QLiGLC%!x#V#n0*}@w zo#-JQ66a(*ZLz6L$h2ao&C4J}T9o-j69?;JqKUM{3K1J<@;lh0X{wzJ;uI^=bP zf-_#mZ!s*y4#tHzgREI3wNU+GLLvx6PjW+)Iq4A-DG92vQ%dzU2dN(wvS~Rrh&-(d z^YjLVLZgaB(|i z*jDBzYx}Bn{9$CqV&(v@onL=o+51rv-_abJ|L)-&=+FEe_?L9u+{M{o&NZbPZ^>#L zEE2aP>En(q%QeVJi8{{K8u8aFbplI?Of~L(0yn!2xCK5<{IfiUk7Y>*&sOwB&e{I3 z!Ya*aXKPEE4&)}E&sXUD<~cF@e@xn9XtGA7lhiE|RKJX~npoEjntF}I_{VcF({*z2 zV4p{4(w1nG%fhU{a;z#cBnALza31D*B{8s~k_bB=mZ4G-(*-03TS=Ok50 zw3LBp0d}TZ_b9$S#$3&8gpGV)>RR7gG|~=Z?%Ex#__-EY{z($+qe=GuhdikS%p=s4 zNe%3)ci%C-(S4i7!OX?Z`9fMSVAP9EI_!>e>a9a0@=!u>T3_gY)n=?=J%(wZxR{7l zI@p?3ZNl4;Hrr09`yjG;oZ#3;+EWR|7sS>&M))xLui15k_M}AFh3<4oETcqjff2QMFq+ zt8y4=tGfGJnoRH53_)H?ul@v1BGFfbW2N~cK!@#}`A~*Gj?8!x){=y&gs04A+3fb= z(JWe+eVt5rU4p4(ZUH(DM-l*;q+T`y!`B!B8P#078J(zomX^1XXQflDzhMSTgUm-V_nup`dm;ES0S&2#LuF7Wrdf z!|Swc-k_k^0vP3{t+FZjX~$H}4X`JJz$GKW`})fj`!xgiIi3uD1^d3Jk# z&t-k#1^%CgKL7PaTg;j7du<4xS##>5u9$$uL=l1`*Ml8g`Aj^Po50n47zhQ3fN42Q z#v?e&o#DO)><1#`nap1U^ydM4cOvC+|LJCZ&`oy(u6X~*`S-VTgGh(~1zk>W7?TyaIeH-9{`uVcpUTUB?UvukYLu4+_oBmeKTu4UV`RUA`;L{W z$fuM&SxoiKBdp9x(dw+SQ&%(D{^(xXsNB6U@LjwlY|Cg_VP>Nhp2VYGf$a8sh2>lx zEUh*Giv6v%^%wD;dmDnmSYa&g#;wxqSgO^34)?hi4Fxt$>>2rA+re48AZ!TpICdEz z*f&~CFfxMO`k_BL)=dPwUN=KMgXhhaDVm|G9il9UbKD2_L|bn8PIoqss5v zliLr3KM&e$-r=fq1WTv4OK;Llv~My%4Rs-TGG%4+=N|5+o7mi&Cb~L~OFTUjKuIQZ zap;M6MghY~ARL-Orw(P*4L$$!=fd?il9kv_cY{kW27C~87SEsSmY1Raa8|{X}==+Ky&t^0iOtPu#B_tIA|Y}+YvZO zewi10-V6iqN|f9-%Ck)w-xBfwKA_GWZDYo?Z?%~54H^W;NV@wNWF@YZzi>ZBmLxrT z=_@r(Gj;W|BS=MwNZST1 zI;O;Ecj7bO$4I>{W*(XlnJer094+M@n|RQfG!|$Cgw_Yuz@>@tgsEk#%8&_{1@ncM z(EU+lg4S8>oxq$R?Q3t4_KkI%_k0qGplTF|1%W_*DrHwwLiM1S<#&>(B#{! z`j_UWf=)|%ypI9C3y|~+_$8<^3BUwvGRbKrA=p5%JCMOlR^%NKjBQZr3M0yG#MtZ2 zZXu|D@8%#DwD9nuTjb2a!BOx$fyd0s!ObCb>J26nkl8bQt*c%_;Jw-+p9ePs`7q4* zkk6XWKfD99)2L=3Z*jC}B%J9j$tfd@;pISa+T>aKeB-#DWtWwIHgQjn@VpGD*hl#6 ziTL0%C$UC z*!#z}xh>M?^APPxKjTU?R4zIo)fj3pr=UrBwpeC;eAwG%)$g=VgcW4$ZH$TEp45s9 zHo!FN|AMz^f6sGm*6nQk!|^`cyzOlG^XDx0O$d+U7E19atQaj-*Q zm;3}SDs2}-9B3$$%0=JuqZ1Z{`<@kuGm7vckR1Gv{R(8g%iVr*V>IV}rz@XTUAkV= z5NKX_(N;2Fo4m5W``#B{f0@`%b#V*#Pua2$%^jx>s-ZFTZ@SlgoO!x?Hbv*;Qwy?) z;&lirGOf*__}%wHH2ohRai>FHHYSWsW<;Xuk61*Jl3@@4qW1+bT&I&;|_}*L1nM_wVoT5&iZ*4sG#MME0eRdS;%=a!S0vUsHPtxel!O88u8w z)1(Eh(u*FRZU62}kU8ndt6qdtw`ZcHjnYUwT^%2ck5)#i){d8g^)9kK=cJ*kN&p8jQGe$X&$;l%EV=NK*pExW zQkN@N?BZf$$t<(aig7CWsr;NONb#}XE*f%}OZR&)P^9^!A1OPmWsZpaP79nG$Wrr} zuWMRsF9Ky#D}TEwRzFOuSoP(}K7{qgMR&8S(f+$}OCrs#W7U;qn?(AQpj~fKyqzCE zmN9eUOsMhvj}~f^P3B1VAK}6v47qYaGf}DpFgA}3-yCg12o2tv8M$r^0jhn)_bB0t zd6+Od=IBC2E@hLSMStzypoyhdIf{A1L^lrFN*jSqdj(&_Zd~V7@p$=HwYu-J_H(6i zrLtT%F>5E{GsB}|t@rv{otv}HY+`^xX>5*f>V~+3{`f=erJ8MW&gv`h(FT8H0jnjm ztlvba5i>x05o%P;mj&xT!d8sj)eJ2{S>#~B>h9wJ8Eqnq<|OdvZO(bVVe=T=<}rFV zML4F}Wnut7yz}&t&#w&o z8UW>n^lH}L-*Z~rUS4j`MsL^m+Lqdpny3+^JL+}cuNpt*PPE-z)?arGuV5$CY5POo z!_9frDlpw|pvID2Z96J{pLOo9C%xH^Ol{N3aR5s6C={cs05DxP?Mrr2g7TK2#}eg2 zlhm8n&rS*IdbI6UHFsFc^kb{-tS(fIap>x50UNf)ZECA+_=fgrXtVUlYTIBul)bh$ zQG!X!n;=E{+71*{RJ0tC(*;{T;XzJK2ApS}*H(Elp6B!_oG&HaBTxX>rA-};VJeX6 zx;z1v{3QT2c}p4jmT8}tWhlb+D~LDTy1RRxe!2$y+Lip9qBPB(JAc4{wHkr~B2^_W zEn2=f-N=zk=xTJTyT-r7k>;fC1V zOz{HzP}N5tX^mRxm&t(MxN>wRXz%Su+jQmTog%lrvwhta=khuyX^9-Fe^Sn-KaVKD zSRiWdYcU~YU3B(Ge3$QhA!sjDsK>kWNO&A-51%R^H1ksLL5>4LzC5Vn>M9mo5QGnw zM_>uIL%XU%;b==jVZ4`sIEg;YS)_0F0L*;6;$u6ah)ii4=A6E4f4Gom48n&pNVPCZ8|LN`;3;Z0 zOAZnT)stxoDCeo(6ceY2+giAcZVsVvaQVl}UCMK4_}@|){f2I!Rt6hE)}NcDeijt$ zQ0u58v2!Alj*7JRTr6E~H0UYU^96S40=nF_p|&Axmm;~iOt-54uAuNfA}hpc!RN{- zkfnG57;*h_TBmM$$-v7qA}6Q<`3FGz=P#8nv#2A0YePRfnuBSv4MHlS{?*v3>RIp# zCWO8z9^^3o$HoM{((;~|r+eu%tdrUd@FY1eS^(SrH;tL%&z)Msoy#ha!>bntv>chO z-Z^{}XzMwSjB(_l!>$vI@P@F>@re@%zAv`Xb;vYCi#u4e()F3TBQ6UNA{1t);=bLa z)4k?*C)|$$hRaho=RMh zSOCynhmpx}u1PwIp$+caq8mE=b_m{wN>?? zo(Xth1765G`$l<}5E4|%^hTioql)*n(SM&0TOU+RF@)X~Xz9hBUW+nba2z-YV~iXx zT&K2zSsvhUOHj+vY}^`T104wRt1i`za=~Dl=oZrOu;g`I)Z z%$*}85T3tiYHMg1^PHa3-LqXBQ7b7z2Acr}qZ#vm=z53n(86VFH@0ot)(Tc^+qS)8 z+qP}nwr$&X@@L=P+0VYc*_zF&QT2^_2Z4L}=Dd>^-es*8JjKp6mq)n(QAMISD3=`a z5heJfQC?O(-pgt4h|nEh044)YjaeFvC$EqlNFmij13d>?MEYmgx|-2B(?OBgUxN_K zPf(~P(0twIf3l0nCv#!1%ktt40yv#%)q57{#ijQmXVNe{i9RSrD}gD8=*I#LLEHa0PnNbeBhOe^0N_8(B%W2y)61m!Rt z=D>$%jfDS|;`ySd24bzBf@j^c0sLN@Cb5+9VzK-ogT(t(DC)UX;{MF>B!9Q*((rkr zo&zk-h#4tJ|7P&dq-TAD<)f21iX5Ea@ESd8jeQZ;{$kX%Te^9D<;W!G^z+Fve$LgE z95YHlD^!c?h?;baA4BOm=^rgmorv^YKO9q#PRh}GIF`{npcPLxs@M45zzFjc59ge_ z$w%v?SX6x+DwuurGdgclnN-c_aBKK1t@KWtOv1Ar9h$SDNsH4su$H#YgL8Nrq%-Nm ztTfa4yfup2{*Iwnq|de0Xe@E68%s8MKkfOyN^SnBFU{&Vfyy81f2Xt<|Ar+R|3|Df zsmT8)F46m?W_Qy)9!c{V$ZDXo{TDWER+mqk4vR}x21uJ_MZ{7Tr?lT_@f%EAc*zkB zbZP9)5Oe6d$vJW&sbXza$Vf*?9IdeGMq0?iONV#K%a@c zr$l4f9=K{)+*WCWqFFteIN3tRnI`3pl_e$PO%j29^4YEY7iRV;)(pVm`2@da+lwQH z2iBAm0#@<<^0O_+=fVAZ|LP4Iy$3)LbH%7I%sY9tVg(wT z)cJj9g@~P%RVtkDtlwhX#nrzXX*oOKNw*7itLab&~R z^_G+@Zwr9dG6{Ih81N1^@~;<02`qAJore9NMyBHVg}c}H_x7z%jucDyyi9r6b;?Ia zifNg|$bh3qMN!7a1IuPGN5!TCFgBA;%&oi5exd;>t=W9v{$ai_a=#k%6zhw(A?GV} z3QB1P(lMJs?>wk@*)un)`U5?_qPbfHHmo7hk{HBA8y(MR?7=_2KS4K$ZZm)n{DCaK zMUt>>kDz}`DMGC#{N;i)_3U9PRLo3Pl{;`f00$J&f<%^P6B8}Xs!6ERJ5`{R*`S=J zq!lT95owhZ>HM{uh8!sy^ZGY+gNTKlr$oxZCX$(Fr?o|!VHu!*6j@wH2X5)As_4?G%UF-edNn06OjqMy zz+9tl%%+n2)rV(2L5K1rL``pj)E2>L5cuDUm@nmBf>G4I=p?Fr-PHgi+}DIv{z9t3 z27#~TI`wr$H1@FRZX_p7BL&$yRpmw?v7y@3mI!IO)BY3Vhdfu&%%?etB-;+O{=nRv zw@&;hl&bv8I1>1z!@r%bs>FEiIKEa?aLDwc-41avk6KeuSh<6_`LmN8LElRc>&SrsSekcmvJ(4&Dh||KY?6 zL;5#YB(B|!^;N3@64!fX(Tb=%GMQ;X5Az2z=9fbp?v)A9kBckk&eJNno_6QkO1iIV$$ zZTp|Yq?Pq=ed?f6SOR7ARtCZR2% zAihcRadi|)a*=K_7jmvo0dER7V>e0+A2I+qC)=(rOrtWtcj*Ut#(N^0*gi% zscD=@Kaorz4!hX^Ud>nd8W~Y6BCu|z!B;(fT!KWMO7o5CEn&#`+Z@Snh3IAt?4H&^ zoH)xPO&GejbJ{@HSSd){3K6|bHfi?5Aa3SDjxkLk2Nvvs^T?BzJjBT46!tgio*e_m zQ>KWdXrl5onhb)CCf%hvA~-8WoXk_?lSFTt4M5jmO7Z}>(=S@e?zKyWDkH%V0kSa|`#a#m#=!WSKrq|&oT5(E9e{hkE)&#{v?f|?ay?dcV{AM5%3=mrH z%&V5;TQVP=B)Y>tz$z!{Dde^oPS0*0v-xk@7)syrr%M5!O&aimpDkf}vZ& z&o7MbUnG67aAXO2qo`TmC*Uz!F;OHRoT;0#N3( zTu|oW&^ruk+zI>MpyA;n1}|51+EiQy+a3R}W-*+jBq*}9R>MA1j)rzM?hcbGqf1

E6ffMk>-m>fd~>;f9=pwt8Q;42+c<$fWBQIl>UL{PsK zx*#{g73?3cN<=7${BYnTm?k>6?Km|lwvA!|PyaJY!}N%hYkCM{z`jEcWBod3^VQ{I zeDV9feer;X2XKfoAY%o?s?;VF*biZEyoA7Bnopo1ipR7ABdJ>YF81H z?LfqD1}b?bY8lNt`(2g8YA_7$C#3P=2|OnBw@GND`_7-JNhfJwTCA+%Nfj-J7f282 z2@cabs3j3Y@ySy^Dz&D%^Oe$kZ^hOQhMd3K>z*bW#pe!3fHg?aOz!U0$`z1`$jvo` z4z#J=Mu$|!{^f;TqCq%H?xioBLolt$#qvFc#9f+iblC$V64XvJWudy-m37P1c5f?n zkcyrz(!|r`9`3K$sDw-Ckb9aW&APD$cTCNuC#AC*i)VSp(VJ-(JiHzv2MyZbw-z^= zOMsA;erv~f?AH&aveoTaPr6RlE|$bIHNGudC+NF7h2k0`HcF}~mDy0*>=|bTGFL}o zUe`R{FP%w$PNd#ht|}I?h78 z?zqF&bLa8IJ{YpEI^#aP%Ej0L+!tJDJ_MHAU}ES9YY$5N^dZw%X@+X*lFHiccgicm zd$`{oy5mk+t!NCUN27VN+njn5*A?$e4>NGXe*^wsx%BVSlQ#P|TY3xYf5UW)t?bNy zi_w;v$A84q&rXhja*ISgEhlU(pyENRX1c*pd$U@2JLO}cHi#terheE0dc|A{YVTl>|_vDkS|@*4|~edH9ATl=WJ*U z_wFQOGI6>*&&kv2vB2BY=`(<(>mYbLQLG52V^2e3V5_EPM+YM)z4yf~(LnM2 zXkVTG*B5a4xirK%!Zna~0bVwj(D$&-8a9x86OT{EI9Ay5FWR;(Mv#Gr+&;Tda4e>i zAZITxv1C_3k~d=Kpfuhm#I}Ku1Bq!}&OJ+S--MCk19s;NIbluC{U|v1l!!RjoPsd= zTNzI#ZHV3N(6}w*&3^HMGa4HtnmCG7QeXeL6h#vPUyjC3%1&*MBZANFTW1D(lO@Rl z^n6`#>TvkTfziq&nOr5aMIscQ$PS#qArLBOOuU!Or7FyS4%U->h|)E7X5I6B9iCx9 z1L0jH0Og%<`Z_*7k-t*3#XZ0qIL>#K^V%xd6lSpQ?zK(<0G6BK7y&GKG`Zr(65Zwm z`+@JmC*^?|KX>Ip^j-GI^PE=8K?_mIa<(TkQbwjGu>}5RBnEF)BcSTAJ$cEPwmF|m zct^hE@WGUJ80sXP?%}Cc0C+&+!9q|ly^TZy+udTQqB|U>PN!o&U%p{+ ze+W@8ygoQtqtQIMI*DGhTFKpQH5pco;E2Tu2CMfVS7s_UJ5sjHTS7XEH_t4;;f@%} z(d@j~3`+McN>a~GVB|LMfRQH%e)5C-lJ(d&)xh?dDi9)rEK`glj=^BuyTp_2sNsy# zOF}w<2zu=iN=i5F7I2(Y6A-vq)|4ItDGTWpOMWpVEzM`(>h)6LarV4p)#YOLJ! zb)06rSmV;Bs(5W+TO)pfixgL6xhp@LN<%P~hstOz)1Xv6HT~K-7?SnN(xU!8uZ-eq zKyvwTFIY}{WvQMFyJw`bxnbV=v+0&OldM{G#q2;mQx>?gDba4zeOx^eEpFhx0JFR5 zZz;gOUPqMTU#p}Funw+n>Ct9;H84fgL<6KYz(qUmg3F;=m`0{0<9F4MfCrcXNC6Hk zKoysC%o)sGna0UxIHWz^!peHZXR5g7-11*bOBmZLR4QgTT!#L5tRvx zuQIc>Kr*oSUozrg+$a@Occiun+nxFOj9^RSggfhUa@{se@N5`o{8TvAI}P-SY|1Mg ziG?Xi>&4(xL1h73-B^ucdMD_dIG$cslXyNcV)VmUD(FOAQ&i>{Z478NlXxhjShEI@ zsGwO*=OvCo{+c$o-qtFz3kTHx0-6l~RYFsjI}OwnT7K>`LzYln(sDHJ^;*HWa>^jf zL)QxsV4;(^6C}-?!kQTfl^?uYCL|T6m|BXNIrYFjm_~c-D?dTGM>);nwW#|(lG(Ko znh9DurwhIm%8K%VaAKc!yj8<_Lby6>xUF@*#_MJ?m{^NtezVFGcc?Z$7cI+QufmvF zQYI@muPht8^$Yz9b*q;|fJ+Yu1)~@F!bR6d>qlZ|1EN48YlA0i49yg;jQhQSbf)=6 zZ+=>z8#xL=>)4iO_Yi3R@99m4fGrWy5j>n&s6y+if)MdcijK*L0f#^C_2-P^@3EV7 zHaEWA&fixXGVf*e+vQRN(_TFhakN(9=}UzH6R{Uw^U+`)?Y8(?jF!sA{ExE=ohMXd zGZ6XK+4vQazQrL*!o$&U##+{HLqryMEt}@rpspNs-Ei4Sb}Pw2PKYCFLK4I+<*S7r4fVXC%XT(v_Yi*ZDjl14VnLE z0Z^r8?V2uz@_k*C8h-^wguqH~y;fW;JxOM5)v%OpLQO*yISvT~t>(=9+;WqJc?&Al zOuzRJr0s?GrOkC%9}k3|f{>&C%`WnbM1_ZYK)!On?#O-e-g!gbpqTnR0*pT}9#JHY zRaiP`xNNg>jh`cCbB|(f*4!n`B5Ax%p<*#w&S!eRdSQ{GZ&^)*ak%uZ&hhSBQ1j?F zlGTuJ7|VT&_u#o=V&S;MC85a04v5mQiR3S35kE?#;}C%>4w$?@?=E3B<2V!703M!3 z*aiFo_SYK&6q$Y}EP=|%HHIle)r3j{=#f^!F4{+$BKQmvX4lYzIRYn>yJ=Kih{M6tPBws}7CB5Hc$Ok}WLq6Fzi!#+k` z2KuUl-JsNdxskYG7s~vejqH1P3r<&4oM2WQHn9zqr)_jUMJ=OV3#MjAq)>-%&RMYWd!7s%~M2Kfn3@$$(5vq zaq#>l#7WYaX>v!fFSVP5RIpJoH7-has1y3p)6IJWnku!q)wHoGGVDPPQjTUMk4BMc zToi|z$j4QFV04e^;V#aufxJZyzw0pNFjjj{uI)ug;8lc91S+Q`=u{u2~JlECvq48 z4HLRMTHVuqAgvyC+ZL{stM+a5b_z8JCAc8&HGC%kRsDsLii3;--g>P$$oT&OiD#Ote z$@ZFRZUATu?gRi5o?U1$nz7m(kDhg4svc9uf-x_qVvgNfjuO(9Lq-!{|vWnU^ z0HOYYI)f||BsR??-nvCewaDOGnl+L4SKfJ)RRu8Eqs`We@epWf|Ahw7#QQ+w@m~7d zN)7?aP_l3hypRC|Yw#&Jy5}h?{aB-yGFO=|U*M z{2kdhYc_Sjcf@qN#?|TeXNc917d?C$yv%`}oYUQGp^%uYQWx3$0({v*^L+p~`HDg_ z)HaEPxdaLm$*|ECU zHGg~4ap~PBY-EBL=v{1mq}Ge$(md-8J_BBI=m8S2a&St>SOXbN;C9~KCxuR z8(MGfKkdRf{68hm_o5YoyY1pK;t^U49`KdaS)dqbUXbDWM>Lp96FDbFh}5V zNg$fpX86;x{DpYDf~zgSvn;jJ(aly>gUDSnu1<=pY+^(m-djHp_{3eyefL7>F5YhM z72VVl9>(hOFHJi-*`5Ad!8)bV?B8r2u10rB0a++&rc}y@3JTURVAnKT3Yq;n+;yKM z3u$SLTjEuhpu)W|IEN0rw)^5kDxd`W1)s~d=q_y{ZZAJAnY*iO3O)PYu61SBZRJy8 zFPV*9>tvRkC&34;PAkT46sTdN_0~9Kx$14AH58a0k}~LNQC-b?3iQPqtt%b!?mR4E z_JzJ2Kzn-x&)+1dQ&dQbQ;qSWU3d$5UD<^;P{#PfNR45=VGr>%Gff6K%zHDwWikVX z@P&M=@LKQ-4A-?K)ijS?!c6dr(EkRP5pG0*KisO4+ogzNim^$>F)TkmR#i)`tFR45 znidJLt`~GaD&PLl;@Jw=@~hfglwGr@-Gb+8%YAq7k-t6IG9E5brxXoqhERQh|KAbp zw3A0L8yNs#LgN2tW9{&3yf*!9SWDdIjvFHNJ5SVPN;m*WkQb|$6?V=-$wyI^xMRj^ zW7kM1P$VHFDg?f}wWX*o{a@#h(ewC~T~S%1E-uil1^oLSH92Y!cv|1Wxwd?D)jkd73Z+AT4g zN3tQ-tSDlE7;Em%d>!H(h5=D&U<%(M1AI#QSXx?Q{41nR9z~ z*e=uihEc>lbm*E3jgIZ(`5+IFw4VS+{YVeefVIuEzk)v8^R zCrS%JOLHR((-%Ss2A5*h!2#x@gQbws`L+6DF@%`fptbqs)WZYwJ;s037Y(xBHY842 z^Kh7%3~Ms6KmAM`VjMPLpe<6EK9E=Gb90B-_$4OI z2#Mi1@P_xqjh3Po^muxuxtz89BN4X#bRlMSknumLsZlUt{CaT+T3yf;sOnK{)CFAK z5FKAj5Wk06(T5@I+l$GLqN3e~xfm+4f6gNW>IB_wZrt}waF7up(36yHVo9*S6ngiS*YxeZToz-Z`?>6w%pkk#52rhO(JGb96b zg+ng$rTi&6+M;E{e?}8A&mJCfRT6Z%P{r_98QK(w2Dl#9R^x7U(7t7#3Vnc=pNty> zOnq^PDohJ;Pnz$y48AZ>LlkAukbshY$qiD^Gq|H-rO|PoyruVr>;h_5?r&DqZc*=T zk+T(x07rj@)1q%Q;Cv+)PwE-c0izJy20)AA*)6A%qVQRmEELtsO<6~X4=0%z&y?zMO=jxcgKFeR=)aLjTkrGibG`4?qx8Dl=bAY?O=kD3gaZb~% z$x(tf67(HWVW5DE$!oiN3`&{cO^^s$cnIrz11m=lK zo_NK)DvBEIh6DK?Dn7w|_1rn_^D(_O-G92pkvF&Fm$AT3vnS*IjI@IdqV&j6JEDHF zL+4T#*Z}68^C#0iEbt7rw+KFg<>L~AR!HTq@W3QARBhMv&*R!rqON2sWfnEa$?Y3? z4KwdW=NTUkH{J1czIOD=p`(filvFUqoHSZBGUU^frMyttN-! zQ3nAiZ1UpZBsN)`K<}#ZR;`F6mx?3b|Z4}deUY>3^5r7lo|8HDLV zfJKcSkdwG<6kjOheMv_!)}h)JSKwR!cAXS?Vw6%{h1KEyFP$v&%W-Hc7hBBG>B`5u^8yCGOU`)=vxLr2gF{`9~t0TaU zuVfEP$ykm8>AzP{uh>p7ti4_(E6zSH8qWg-=Ct3-*QzHHT_N?=L$hCG8t1VBT{z|9 ztDLMZMQaGD3>C3%dBKt=9_F>Jgn|kC96a|xFktD{6{#`LoEVK{{?OZ+x}C7u8s(hwmZ6vLBS}}I3)Bq3 zv+b%k{?rcfTHkcL|9tR)XNBPppKW#OMq|sIkVLbnP zT%wa%ZFq}GV(TRzO8*J1k9#xNG=)OV?1Qk6CS@H;ZaF8_pbUXqf(Qr zX;%?VnBPmcQ(oSMxm0@x%hf?{@@xaFx>*76kpmE|EBQV*C>_>j;}w)NlgY+3C7P#s zu{u*KKBpBTZ0qGirw%vbFivoD_0c37t8ytT7FCAy=1Par&4joRjO%1QfGN3^nF9Q_JVfjvQ_5-!q1O>06Mc9J)*&&EQaAy*gwm zAN!d3DM3X1`M9)sy-l*wE;;6JU-sk|zG{*>R~LlU;8zq>+oWT!RPs`X!!-)52}OWK z*&eeI`*eLMbWs@`;8nc{k=kJ7hGB?r1TeW;UN+6|Vr9V8KGk9MGO`T*jMXNnix|L7 zS;#^8LnbKmB$Namp`#hA)|FGGqI$RD6&C&QQ~}_oq}sz3=9lmQ$PXUFH2l!VcHdgt zrz$e-vi&VXI7r_uZ%`<~zG4=Fn$z{q;2w`0Fyek0eJXPHVu#R%XelUs&+Apm9`+RS!WLoi?Pb&+oUr(FFKi)bu@C7RHW8z~9us1CM|n}xI;kcso1QJqL&7KltwpSDyN7y3KJ-N zyrh?GyB4iT&{{|QkSe`3so~voD?YoxnBArPk%75pR7s*{((C-9b2R@9e2(TRnb7sb zP-?_PUlQ8QXbmo|?%N3%Uhy{?NwZ|~yjaNsG9RFPC|3Pqm?#S-$F*UyRZHZsw^ze> z#XL?7&o!;+?p&3Zjn;OF!3m}3%ypG=nRo*yJYyZ>5sJcDn~#Y`Rg2no0~N8>VqaZ( zfTP7`?g{OK#biJ}5AKQF#*qfkd#6J3+&kfax-L$`HD^pafnFgVcMRyJO@M4w_v#!` zq~J{-&oTYbZzDrJ-aYTPSPL1$sYTYJG&z(^2g386QWC`jTAcMwN;KC?9})1T_%*x? zPDE9dNCk0|%lE+5a*v)*XIC9Dw)7QeOZ!S{IAAeuram-CeMHdNtZ?4QGqTj2pM3We z$Y%%d1rcMfc^oO@#o|sQ4kvC}i9kvmPP)38J%51Dwsn?TK@0jGMMmmcQ%q%6Q-#!7 z4#8<&VTlQ!IIHo7qVG~&h~QFolcqyhl280^;lLUx?vAKPI4BTt5_l%TJzQf%f|Ca4Mm_xpqu@#L-7?(pIgfAZ?6|4VG^HJprNYLQN%0wpn9{@e@aN|d{gA5D0*^!}7_b7=(P08h zW5O^T>h)~r{&^K=HphKsbQh?Kgmq!P^UtW58=Ba;_2S2QwV36~z^__-p zfMv9lVYH%veP9X^i9qzJ-3X^DNDTVaNNFP&?1$2?Q{nXRpULM9Bb;rGoUlV*wu2=~ zntTB(9DM&KGcbeMd3U;8L3!R5GJ<3)R$r}vrJ#a*Y&sK%ZUIY^c>iTR6 z;$&|{=j&^@CsFOz9gD#z4iK@M?1FX$`AtJ#+xL3AF)PsKjuFzBKi8u4SpqOENVeY5 z+Ug(^GY1{e?rwmqmscz|{5K?qr5f+-A($Y>CbT0hlI zrt12wWTD_sq%P zKPEEwC?k@53;ETSfu7bJ`rUOUhGBaT`no|V;r`K3ZYs$pTvTq&+#)(fo#>|x*pXIC z(a_ELu-;7#vPKN{-qgIu?EzzB&x?ECzLcRIZ+SoUKk1Q807oR zp*Ui(q!jr(QSHBlH<=Eq5)>O{Z|gA>liXOk{jRh31-%z%H5xB}guKnj7LV0Qn44BX z^iJrE0ZhWY!84xz#C|=qU$oGF6zz){MS>H;XWonq0hAI2ms8@p7|B}FGlO-h(lryh$`Nz(E2YI?# zwcvpOOJEt62ANIeU&4s{uCq(^qzBa?DoRcsfj%AZ(E6{gH-qLcRK8umMBZXJ{eu0k zXfuA!jH?{4pUkI|1lJbey=;L`greiNtpR-L%+I~VUgEdj__Kp;+}6*EA*nJd@3%pw zwvTv3*#`QDP_T`FCa*DKW`Ikek)Gb39$dz-DzP^l4&q$|mse`>2D)`+5$1mk zqH^2F`Ny)L8g6%`|2mUvn$m=;yot7HjA7#pW zYil=aqMfT`pBftvYRN66190#Iz#@!c3uc`z2B@{15=8p6!^@ecbFIsjw$iYPAE+Wm zGmn7n?QRW(*K&0vC^O17)5M)DE0)sFG}Kh$w3Y!pmc! zZrd+&s?Aihd9?5|_bHCD3_4iT@;QTl7a|qs0PX`u*kqR%ryHkLz~T-$-y>77kBzO31SWij6?N4dTWrzNfpR~#~>i`;Z$`juYQ zT}aRZK^F@essgNm%TR6Ek=Wz8+67b|3d*p_Y%l8UFCxm16~C(ox+4WW(7?dZ}meajWYryB&@I`6^9K3EdOie6$rAIXkQ$y)1p5jUp7N}w7^ zfJ!I)-J`(A;zjH@&^{mL08;ap4e+Z0CaUtD2AS0ov$Yy;K_TNClO3qOC8Jk-9r`nS zwbK$Y)g)5!RUbFZ#vYlaS*)Syb*vIpV-f1`EWe^4Lp~e31MSo}>9z}Q^MS17H)G$>vI45S`on&_}q@ahd9^IjtV|9&cyKM#zpR}~8%g)}jgDuO{cF^!u}klL z{_7D6>QyA;6Al0Xgc1M%`+v|Bj2&&Aj2+Akf7g0b8kQc&BdEVuLyUC`l|~+_!AR*C zR{PM6GBA0mfxLBrT2aBa%|G9elS zDSBP#oPoJ{uG7E1RAywPkU6C^47fAV0px;CFHft(*Rrq!tA9bQYD;>PGO5a&REna$tToWharKZuGYwjOgEXmxd# zK$DtA(Lj&+ifZ7skSi)tP~2|rVbSrX?>WyfYfrTm=_2%8HXDrxZRO`kBmGQgrxa*) z#i$Wx0MbwcY${Q|kknt>fDQ>vLcOVy*rR7KMSU4dwrKEU*#Yt%F*f}t1o0{sZp>kD z`Jm#PVqQUk;BxIj?(!Z}6zcdhbLsS~-0Gj^b*F)RSko=pzb|GNMWxssjC$&NI#K43 zmrbUYkJd}hRo_}RCMF{DJQGCK!IIG=8Sd8w@zS*lOs0-S4$&`t>EXEhwaQe&b_{h2 zm}EkvvD|*QX zVf=wSF=(3rpe~9D9}C8gk3_Xc9U&AJ9nx0y*u>aanIBV!C70~9HWJFvLeG|g;_8CL z%UWiM?SK=e_a&Oy$>xaJNP>Bk9hHgs&rkIP&$X>8S_|Vq>K+tqWp%i@QgbHJ)k0>^ zGlZa{H2c`3+1VU>$7G?Tw+YRP%^~mR?m-D3Lz|mtd6=J?q%pJlczqt}CLGJn2iPNq zk~U=#U0nS(f&|7Pu+=gg#pXyGY=XUmWQ6#35e=(%3BXezF%W@H9(q|fsaz06lC4#& zDhtxesZ`@^xT@K}pA9(5f%WHT3h+oo*+bw!qtlagC zfJ^BuDt!(89?)3#S=dFd*f^M^`e??LfxLNqowzeNU(fe+K&FRGldQ!Q4U_fb5Uf)# zZLhRWZMDalLg9R=pCCe=3tsYPsS-gV=Qzp}b`L@-V{+4?X z_9~`~B%AygfDTIj3^K^6Zf6Xu=?Tt!Z@v1PrSMWGlP6POD8B%0x(PFgja&G)GI$iK z5NLyD+ajG4c$oSi5GMn@5bPSnv+01z=oxQJ*~$=8i-2vWnmFj#2{VUmM|U?q-2y>8 zp1@pfP>!G%yBET$UmXBp0Haaa?Wn+yU1`DjSYx>z^S$tVv@SaBGWK=5+rl{ynSs$j zlfjL>n+dK60Jc7zp{lyrM9pP?HVAyfS|^{uR5z4J7TH`Ho_k<5Pgq#;RUUT2u^<(> z>-b+RlfE?GX^2?}iRrX`*!5wn77$=k9j`Es1iiBGpG2et zEP*YB1T~sNoJ6=zv(YmF;=fp-5QF@|LIV_}%(>?C-9abJZY6&*&f<5eZHRZ5N6$#gE1K`jgj59p0`ZS>PpH@F!ePklIYOnaG`cHi;0J zzqm1eLeSldh&qY{L*?Nx+v!mJ&JoInPW*?2#_kc;Gr+L+?tu2(4`sj+C zsk`u{TTPS2otC6Qi5K=rqFXj^qZ^#LH5eDDnaf!yLes}#n%CZy7+-LJ1%xnR{CFc- z`A+{;3(d!-j#1-8+A~odHS3t=`{}x;881^+mz|5N9%8l-fOMXjHicbcc*`DJob;g zANB6-@7+F^fJfM1A+X@{Dpk6=)3yifPA<;&wVIwua{$mJBnKNOFsXv$fM_Zw*v1s7 z!1SZ7>8G)E2JExLjvSmf+;xiic1wAWrK?gG7y-g#m zRbrF_=W#;;*(V1M!4n{&L%*e68-OWPnO9tIb{d?^^-QI3yJN*#lenD0*oX2+8(5Xw z2!uyRsF;Z?eub+QoBL{7e*r4lW{F!>Lk={VVkQF!!xH(a)C6sir|gXw#%gC~Taf`H zI^BQj<$2h3JKlSm=5;W8kYv;n$|G6Ga+~E;CdRsnWS8Lh{P4}&-~?J%4#DKxIy|ce zjdcRC4#fWR$Dgci=YI!Hi#j2{!x&`~FOoPcQsF-XiGsJkv%5#UkQNIw-SWSRdmBSO z0+R9_gil>B<|^)0C^bf1o$@}$-yei`N*~pCvUxjbBfLPy`;!AIaGIJ8>$yn7><)of zV(0<)npAejjE;8H$9G8OKtmL;-TBw)v<@nJxLo9da{5C{i97^V3`i#|OODIvQx2{i zQh1_`jgS);*tqd5W6nMZY=K&(UT*~)zz;=o@p zmF*~e9S69qJgHm+{@5JE@s+SJkgU~34@=68#n3(7A4VX^EbLgIuQ+HyUlE3U6s`)z zARtrJ92$hERC>)EEo+2kgJvL8eDBEVDc-!zP?bb5Nn1))w_j)4zfoXqQqRD+N|jpCzm%i4dYzbcu$_NFzxRlK?`4*7kh+9P_tM-Zu4Z{g{cDdu3ULbs_X)V$H+a0w%EDUU4THD<<^tx!Ak@@* zsC(wl_WK;i!<7y;Leid4M5+~)Q_(Y=TN0&D&~3}??nrE|qPr@NLyc3n>cp<&1i9k} zfpLtQ2h)zfh}?sQnbn8MgC;3oXbHZ8{b0Gthdylx-z0MmS2HQUAgyzHI30|70$-7P zC_W<7(y`7f@s+)Gs_+%8a!_h-gdy{I$8Ldh8o@y+o#kR*znxZJ!zvXF3!9Uju+uxV zWFArDwc%Vaucrc@QcZLq-9icwo`Ke>QRpU%y|!|p;kobd6B`a?Cc$MjHlEV7Ke)Lt z%d?*k=t=JmN2%Q${I({jZK?4z)hRZTfy*iR`hKisrzj;)9u1dl8`Fe-2h;Qf%kah_ zZmty&!Zw>*8heNP4X-EDfM?q%1#z2=aXL7agC~3T#~KeY`sW_kg_D)aC@s>4ijcBE z1Fp^>+kCV;WGw>uF+w5zLuPg{GqXdHF<&`KP9$%#8h@mzPs;zp4|)DP zpZwZmPwBkG&He}V?1sr2-1~}2o4&{1srD-*gRcFJ@m|EI#+IBU>^H3$o?V;@HJV?_ zdT^W38?HfD%>1*6yv=ekY4A#!OB3yFNS>b!OFE`$MX`qoG3Yv6D zisO$d5@lRr+m-OQf81wBCIe+7SN*PTC+R!7>t_&leoyTcy$c(4z! zyLbIn4bb`25=QS7c>d2>{10X1{hI&J8kv^M7pt75^+*q>#n$>WUr2Ti-<>V+c@CQa zfR8)byX*4jABOSa@SeM>w8u|}z8~n^Z|6T?cIYNvt`j}r z1)G{LH+_{o8c#hyP2D>8Jz!1E=%2Q^Cp#*iJ;3K*V-pUm*LVOUjzXZXGkS=@r=e!Eew0=3ew`1kl$`)ox3a{Dsk(}BSA zpYGY*LJWI0B3}ZWPrKY=$l=?9?f`A_U4!q?hW9NrVe~GH7~FnI-179^=s>#YQ@XtP zo7SStWP|Ye&hu@;Kj)ro+g!D<53UbYx%jU4j~H0%cQxN;scC9e(eff#rX*XWT!%GH zKsQZf*sWD*2W9tkD13Zb*EKuwfz@`>EGOSASJQ?!SzQxD*i2(2K`5UG?kERIqPhx#4sr11 zFnxsaH00ole#tf!Wn1@_Vg*>YjCFc7(J#&!_ zRNpNd6pBcMhi)Vb9q%hL5D1rM==;PQL*9=ct(2uTk!)0=^3mC@xX~u&j+`$-vr>Wq zOD)7CQvEsm{`{O(Y`6&Gog*B-G`4ii9_;fl+|;Ex#lym*7yj<&1SJenz8k`b9uA>G z@y|bm8Zkc6D+G4!wYXR*6#pr}sF&)H5FNOUsOZ%T%~)ZlE^JtdAXTn``VbA4zqTw8 zA9jx#(nG^=q>wV~ohonqL_TOUIxROG_~Cj~Rcg2q9hu!V<~ zHku~@EGVSonI3>lz!mO^Kch2T_%VNOIC9Z|Bhrz!gWVaJ6y>Fm_8+{4(r_m!C{m>H z1#s`;Ogs4Z=@gmr7EnthvMn05k*Yx!dwkqCsg|E3QnT@Q%GseYq}2yCz1pwbpIN{h2X6>{NYRl0N0_;Fm%0*ydt#&d2+ha@PC+c zzIbw`A|VO7GH15aY|*27=J6ZgAJ!maZ-36PcK1A=$?tUmeR%(*&&bF>IyGDogYdD} zzFq!9EJ+&c5q{<^=Vgd|F?&?f>i$3Io&5~podv4^@tLIZ~Wmx{Gq{N z2vdK)L0AcaepU>`uVmORD9Nw_cEWc|d-L+#%}37t8Kh8yck>(0HmS1)6X5io0@At> zzO-~J>+o0eqI;5*|BUBLo5qwjh>l3*+bj%@GK_+FW$@B2Jp;x5um})?td1Le_`IvG z<>FbD`WOF46`4=Y+sS2Nl{T*S%o$5e1;~@M11i&!JASuLqhaRSIX$bi!mi6+_n9QK8r3@v;ALk@0Mp;4pGHPr1krjQ(u0b zY>C>X*Vaw?BbnuuX>lj@*;e9fn5VUz}+on0B*~#x%B+X?1&H zm>l4f0o3VW$YWYOTC4B$xG9go1#mjgYXCw>Je|9ZN~{a=}2lH+{Q!YyOjuM#Z{a@BUuGcQ@ zEQ>Q3;{|eS!r&F%rL9`Dc4|^tW)~M1g@C_Hn-)ZdH9HX_xD$%8h55%#0#$!1zO?&0 zfo*m$CW{kdM30c4twelzCchHa^TlcelbHT8)>~}wBw!E9(*2MfmbL2IG z#mGJ zlrmL_vt>*ZS>;9~whQ*J7#=AS-dJ?QZiM;c8lOgMWWX#d{K)D=t3UFhxGjFkaFupF zckqnLx~tgY-y+rQvyCLt;{GQ|y;ucdjX%~nW<)vffXy5eM+8$}G|6{^gK6}>C_ZnC z>y|onNzV$6CsiRKAq4kJ&4!BVPz&nX^UUXR+U>=Fn z2w1_&RLBzoawjbaAj6fS1D?u(CK`ZO6;m%*0?*nkm#LVhQQ{kmMK*U7+ALB7K2aHV z&T+Oo##vrA)*ZV$WZy1SiLF9z%ly~9$qBc{px_D7Av5ho<+u!?1alyR2MVaZJpDwn zqKT;Ua)3tinTq_6f8;=oWxed#EszFT28-oG5)uRt>3_Q%hS1j0iJ<#noifb)?QZd{ zc_iP{gsv9fgJY^%g0j_{)xeN9nT1Fg#dMx62m*`!bKcCS1Sl%`@8C#3BP)~|4@80i zjZp~Mbb@M!)aV`fh$=KA>+)lh6*ldbO|aky3WGMM8?>;DQ)P;lE=mOd;w1jU{d0 zDrx1lHreCaS>)WnD7l)rLEY~4WX73N9G0U-7AQ$$2@*>{R!G0r(T%N!aMukI0utQ4 z!Ng6R4d{kQ0TVxPDXowi3VgVst{fAOlC;_yy1LFg=uc@dOWn^+m`{V~qaI{OK0$nY z;o=h4%U6|gLaz!KnEe23j@aSlbec?tM@V|8x)9N%Bwq~eU_lNXHB8FO!mV&K`F4N~ zS{tSrGuFla)%cAb2z?r49{v_dDy0<84LHGMhR|HfEUJX<2)R$o6nKQ3g;HOjT#yfW zL8;;S1$nG*y2|M27Us+s1^!BE=jAV83=Aq$lwE))@t4zg2WR z3x90kV9Mk8SP$6;m+2a$X>1HiuCfCK=)v(A{F6Y1d1e|6u;oB=ydSv+7;GkNSa#S& zMudAMAG{}BCv^P#6izr8uD$_M!}>4Qve!0M72A}`R#(*li>_smApe2o(lbbh0e{8a z)%O7#Fc=D4x^Jgg?b=?4hCo|MkNqtCd1_i%fN@W>5;f%Ll~iuO&PUd+WxlYV7xdv ziZOPND}l;-sGyt26%ayuvOt3cFI^H0rm*DS5G*Ov4spU#jN)K10~>&QwE?ztkicE+ zv3L^;rF6WG;DIKT9Qsh)71%3Q*aXne85+^;`5b^B@#+h#N@e$GOrV^P4#ybqYoN`_ z{V6mwtxs+l&BO=$u>(x?XdL!d-=n%r}xr_%QxM#5Qj0%6n*`^QV5TWIu z*sF>Wa8WFkr9rferPShM4aG(FtKQjzfLoi4&oDflB6&TszDYrl4U7pqU=CzS0)^@) z{W~i6OK!=7r%>d$iQq)8;oj8Nj2w!qQ2}xU?3{L+G)>GP0(CfF7V=xJO5PivDDFj8qa>2fX!;;IIH;ihiRk<_R(7*PCu~V<7hw?4BI< z&q0NwEAJ#KWhi3-L!zUz9-bs`Tn^{T{fox*SRMhL#))Nd_Y7mLcjAilqJcGKGr%kC-v;wMh`R2eq4L56&H~$HnrE|}LJyL;f}l)bl;PbG6X60edT88w*`9?X9cDE)OP*=^Ny|WmfLM z2HX>xt=muyY6cQ3!b1cIfdnUhZMpB3-HEWwNB}#u3pYZPMiUk|NmgMwI*&3iE&)N> zhea=xJ*dQPn)&Lk#~2l%AQ?XP;vbBN0e9>rhLSD7 z@}awxK!k#3f?sO{??k0vqV36)9R@ksg=e*|fvXLUQXWH@-uny%T1Ac?3U0_lG~cP5 zvA9BvGNW`qOencSe4jc(P=_;OZ5SX2N|)U!Q?5s+Kp~QDX9g>^Ry<;bq*BdWthG2} z5YeciGC0W92M#kxB@vuuLa}`lIS~O$r*+#O zdJ(9W41eb4TVmMbv4JQ(bkor@j&Bo7R=A)W2YC>lUSG{{oeF_H#?+BGqhd}9Z0wMRtzTEwsFBHX#DWjF&3b!zYYm9V zFVl|eE1_9D35i~f*NX-GlW|TJ$hc390x}~heO$kK0=XaJ4Z6Bq4yeFTWWa&JrKoT~ z#)L6^SS>+UY8)n5sVRpC-W`T)VlyNJT`SY~q((HQ91^HEXs z4>Wrgs$IQmyztU^9P|`+U!+UkdBWm+zoGjCbr z9j_GvroccAXjF;isb*pG%PKyCSS0dA6*3_*e2J#2wvo+hU1_0IN3@Pg+f;8?YQYst zKckS(scg!xsCRs+M%eQghUalry?Ju}`lBvfDuV?gz&bLUZIat@#0k=P5&YXwg!>xD zvRIsv_Ye0V)DP3-U^grD$W&d|h81b(x*pQ{_)-!3tiM6v3MF#U6L2US1YR-EoIi%i z3?DE1m_`~nDMC+Ua~-RwX%5HEYYm!yz)P__x)M=$?prxYM4~f!~ zg=W1nS(?sVbtEt;%4A}8cXa>L^QYS?PT%rK+}oM1xd9Ft0EiNpV!XQs9R!wB3^bX1 zyJl~Hqe?Fx+gZ#@oB&cOo?CgB9STV>r>snp$|;1=r$&Q%l-u>Yv3B$t?!6I5o;&cP$K#od?T%kAs;XIj$D z*oxd$qw9*MXiClF?CM_@{GZ3Q73x!L%MWggr=G`A@+U?Xng=V(R;%eDrXTZ|=Ca;W zw{tq^r3^K%_01{f zvU$c-cxRR8$;H8IzwG3_z|rx!O(otmFF^RV=UXu{uCLadSiV@i)DvaGUN`GY)$JX| zIKOK2S$fpU2O(bPwpc`6_`U<0Jt_@=7hmPJ-tgn*eJ}Ic@8sd+<>cYuyXsGG{B4j1 z1m82D+})20whEki8<_wc4!F6vxU*D^_^0Pd?@F@|c_XQv2{n;0ZN=N4(f<_|^%dj{ zVMR~y5>$kD^HVK!HnQ-2mk?QawzJf9v)eJKXLO#4C@;CFocwii<8#{cDF4eeF6r)^U#>XJJt*{)wJ69#64mP zhdH&Gd;{;EQs;exm;KoflMKeq?6EX7u_5er zxj`>}*2T}Cp!hYr%=`X*V}4eK?zLyg>Kt>sG?IXukxIQ=!*pzQs<3AllBN6I=~;qR z>qPgdi><*rJ0EA(=T4uCT@rqseHrF*s+f6)l=^9^@xAib<;w&eza`(#A#`D{AP@>npXlA?dy0dCpk`Ab^_UqGa$tJbf4C?#P zv%mPhUn{9vITfBs(WJRsTzjm>@V5niCqHgY^vLe4d3~@;*ngt);_c&)PGafm2psV2 zsoK?ST0O2Lw)-3NRcyB|B^r|o+}T}O?tuv6T-#Qu{O2i{t2yV}hYdvY0!8q^dOJNyQf?FM!=KcN$`Se3Y*=z z-4{EI-94?Befx>;2~54ROw3~CRjNLqVP=l9xFAM*s=6TWS&jE$*QTk(}3fAV=9Ny%M2Y&hUz|789$CERM z2LHP^9QG6Q;MpU4UzG=3(jnKqCf7(@N$wC~@|+T=?ZKSk0DZ+=6Z94BV??ynXkBkt zj;Zm}EK#8fi_lZw&bDdF8P+zMiV^Es&RFlb@20bSMTn$~{^d{!RRt#^jDVG_ z2e8n@F5cS(_D31Xs(vcX;!|O6pxEEvcDRwLTjLhGm1LhvCkgCA>>!Qqg3gA(x~lPO z^NoGnl?IJLOD1d}I#u2-(`8???!H4%74(}AS0-5Nau--D24Z@+TFW?4$M1T2?=Yp36#MprJr~h1e_~5PHTK6S$Xb(3N-3 z9(Ej_+NQn|4G|SjAoRe1;>3oY+f)BJqJJYt~;O@&UTQ+s|FvhToJ`U1hriS zx9dO-V?@ki*&DF@l|zkS@5M$xo%mx$q12=xQ07ICT|IHN$)z=x6!-v>L6g@xBE$;- z{_K7&Fs-hMyAbXpF~M=mK+3l7`bP@38M03)By^yIa_?r1W$UiT+R1`bXe=WnqIeP_Y@`gD1Gxyoq6Ywmv(TtmE9Pj_tA z3;Bk6;}CQXRDv-RLzk#foj@lmP%4Ke1LX$5iByGV1De_gLPQkO#M=Hr*_cWa0)xRx>~iZz8`6bLI$F-hw9j+C(eQg91G5kQQQAgczK7*V}{Pk z@U-y3j&4t80op*jco&oK>ju6X=p6zH7yw&7bG_UZm{KUnL*gG+ga-CGN%b#@E&d0m z1!8fu7hX@K$Z}|eoI*4`dDTFKtgW?jGb3)UOy8c1y9JrXn3ny@1c%&V#MF-szRqwb z)wCy$j=4m)6JfsJ0e(ciU@0sHh-&-CmY&0HmIuA}Z%B7?K)1teYxwq5xRGHqG?Xg; zD0==KhPdrG-i{o8%y+nWR9VrK^~nrqX5R1eM&vt0YUh(`u6F`NxKQ;|pkZBTp}9w{ zwWRes{9M*R+nFh+coRLpcZ;WJr^6f}4hC&S85c2cRjOb$!-1i|e*MaEab2Qvq0?C( z6BQQULj~{#INFV9sE@=xt^2R+1hE4ki_+x_JroPbNJ=BVF+S8N?* zFY#BXaxW3Aj7C1ha_+=`E;elmarjo*ta#4&7qg*=5apqI5j3KBkJ8bdU1n$ z$O`3OgdQsb&H*`OBcwoeQWtBjgQCl!cwCmEGyTt_s2F?kf~y5Y?3REJ?fFYA(^Q3F z;wb13?Qzq(PD-gK*4mkDZ_tM0gmjHFAk9lC6YwJ6N1@ZMS_JT zPDv`9gOG3_8}C~viMk-XlbJZUCm2d5)tUu;dD93X;ZBDptc~l{FAv11ng(h{@B5Q2 zlvZY|<(6NPba^aJxiwj~f@}d;O^preSAaJdxzymXj3w;qD3taNhP+GXYpCqa|A-Zr zh_^DV^}p=}Lp}8uI5L2E+svLFVfTQX;*e=lw{o8EH-(m6)@2};0Sl!NWR?`csn^k{Tw5EcoNBn=S4&z8ys2)8CUaYzgetaQa?!t0%+Z-cv2JVor?#(vYC&@83bYt3 zfT}&6(3RU-?}940_Q{pkM6k@rujQnCTbH>rbag?3AqB`$>I}zloy?ILUVa3lWt}d` zNsDipr3T8hMVvKnEMIJtL=SB-1axNLXXTDloi#qY`D?`rzpN)TYA7f}$d<@Iv!iLR z>LruQ_$lUnmq%cec(HH>)RkLqi|+7?=B5Wg82pt=2YmD*f3XF2E51J8ZHD`f)C@mE ztf(}FD*ziiH62D22 zYn{|{Po}y&Zc9%XzC6aecJc!_>Pz248zD7`G5XB#{(ME;fQ1j-*>*;WZo~UaQ5EIZ zCcmR@TjxR%Eb(hgm)3-%#ok-(>dMPXS>01iLFkv*bdVC%wxV8eR_zi> zLnBkoW8v8ifbnk9NCc;r?5OgfdOmqk(?(t5OLU)mzUd9r&#(Mfn}KE-3F4z+_HB#g z6eO!!u!TGHWV={_|GMb7p8;iJ3*KHiDW#T$V1<^ir{;{#kLMSGTYNyT~zexh{4LMISX0ZGM}n;lvW;Vas_KSO_6sPgGMl0#nCeB*Q;<1g?cz}dA-P-e*(}Wlz40%6CBZoKxf`mdDH(OQ2 zrkauaN#M@cPtMXR&AQmW)^m;5_d{}aV>pq^IKE?aW+}{F=*d+fLvoq6y#u%giH}y{ zr;GRF`H25NN2V!br&ANZ2=g}5{~m?;Wtdz3`K|ZOj(3nW|uin!+4ymo1n z_1XO&honVx89`l#) z%$;>iG!Me4W0(OWj4eX-d+54p8q(Kg(jz8W9yecaHxQ1-AYB5Iy%VUE5>0V;MAXkaZzv_1V8+`I zsfc*SYb9xkZTbf>&?Vs$o&kB5nX+Mu+ zqI<$3Rf>9_WyEJ#*-eeaPp^QNP;Bm+;6##~4Ed-oDk4Wmv#{Fy7-Xu2MBXxG075KL~GWR55E$U8?W7l-Hf z*9uKDQ12fNvKFm~u}^Q ze*w{7N2I5T>^QwX23TrD01~3l#F_daz`-!{_P{gtket+OV!8%--v^~tK2i&LfC66% zH7Yc`hcI481j#XzGhFT;{tx8WcsCKjpuwu+D1k)5HK1D}bd5=HO7UVINO$_#dC!Pl zuzyRA5Gm6w-!O+bL;?7*noXjjS9~HM>qu8oN2N91s8=r^!9W0I@J^8+mHx#z+t8k6 zh!wFWnOZb#kAX?bSxR`%`7g~Vx6wNKGzq{l65+d;%~>vx=yaM#PT2zr4AbCcX4v(T-BC*o7z&jCwik?C|sFQ*|0*Vx8 zM5&-tZNQ{?+D2@3|F0N@ReM8Jc!-55%2oxzvf?R=di&pBcy~&`D8@=#grOR^qNEn-L-&9SUk_c{Tlf{*qxUeVl*a5WQF zVT?>&c_<4poa%M!8bkC^`KvKg%-SI&oMA|Xf?&o9XJ-qmIZ@2`K)CSCMKQ#v`K)QZ zlKsYO3)k*W4)^ph$4*X6E#nH;M#SeD@>0d+C&kB3lEz}0lXYS3X*cGtQqPf%Rrn3@ zH-143Ao#?01{Q7rVG%ZSvgqzJ`0)lYpg?sT;b>0s46*yAVsakF;O38)v z6mOH0U>yDxk9Upsbxe}bxVon9s^trS?3EeA@oOViMO!A5Ps8{kec8%oy&q z?-y%RNmg-7uK4pZ3D@K--(wSMgr8be^gWj=;{2CxTVcSu zaskL*^9$>Jl(JM?_$iP+hc-nBVmS)3+!>^>ZmZ_8cJP1r}X;HhL30FQH(9w-PSHuc*4@Y1h`Xl`p0~=ea0Rqe z1R2L17iRhA-@~`xXl~awH)qqa)tCqCNgytYh+(iL_*ieRq4e|b7qs@bs%iz%*K0Oo z*W@eLl+q6a%L+SSN1wxP5d^s(I)Y2H%3_r=c*Nj(?2NT|@u~$h`AG8;*nE`eAy>t; zl7%1k!#ukLyUFaj+OP*T;)O)7yZ;pyU%FL2D{1Aq!I6+rZF;9^_N}I(!&rU5XJ83dM+0eWg?7Kf0a$5fJ z`7!u6aM9EhmNzvJXJe;Aq3u#*RJQQA-;ehfHU}RSTxVc9hIApq~?XVHM2wlRaOk>Swb3L)`QY`Z4b3QL+IMm0}#^Pd=D`x+kK z$ABwtI^aajM%~2D>GrmrIcXezpRav_7XMv0;%&~`mjC)82V+;O9e;}F5*m#!0VQcw zHT9fW-duOjf#%re#|m0Q@QzunS|#VZO`ysHeWKN6$zXPaK;=ogUrY@H>CW)hKeQ3K z9}4mkUJ63Q=g^sr(i1-aW=CY{iB?-JoNday5YyMcDEB&|+q2UhD)KHORghYF6x$Es z#n;0(P`vF1yf8nKa@LQv0Z56}XT{^u3XN`A2?rI{O_Jok1l=paGk3!;O$XVAyF&nU z*w!n3y=Cu{9pm^*d$^i4c|x|J$@mAS&?0XYCS7wWecNo{+Z7h>5nXwut(R71_fJ`u z<|G&8Ni|r@W}kp}@6K`Dx*e{xPQ0*HZjmosK0aHtf7gnfIEPT!;by*NgtynT&b5rz z@H*%+gCS?oORYT=wkFwcSgCc`O&O80Y}Zk}*5d3t$%*k*L(AAtqx*X>d`V0`h>J`c+Kye$5Dd3{z#3^2Jk&wX`>14z>jz|80k7nm+SO z4M^-a;MX@-AT#4&J{3*HMz*=t=K&5PSr_OCi2kiSZg%IHbl0}%Y9!43@quFj4b*-;{^3fW1j*L<5DZ6AAG z*h_uU*7m*WKF+m)_IERb=t64m3R~GA_Re^%=6ruB58AtZ|1Tz@k$vn>BoF|A%`e9n z{lBbFn^~H<{8qmyD!SHN3@AP?b?xo&QcB{pP}npKQv4VBS7J>K;9Ibbv}1*hrZgqe z6g&2LB;;~fUD21MjW-LNXS>9QowdN8s#l_d(xi-K)LPQn4Kf>pd$}j;lQp73WL69o z+u3r+iZ#})<0~k1t=b$JGJG-eWaRjSdO?$vpX=uVg~hi5N>Ju~zPsLic?Y4PO~I*& zatGdS8_V{6dOABZwbcevsD?U5V=Lb|)o3Gv;Qf(O%TPBLfwk4j@ERGnrc&_xQap=J z(7&#e)nKnpPs2&P?Sw#224z-i^ml8`;m}L>+JHtPNZ0JQYzMx)A#{lO7QA3p;e;a& zNOC11FShjzOgY=M8?*PdPEufC%2Zb&LPW|D+Ce+1Yhh^(P!o9wa}gy=S6P5@;D@Rd zSytohOm119VD}W$8CbXdDH8;g0 zWnoNQc%8kSo3OGj=`%7duVb)LD}^U3_ym}%~F;F3Qz6t|_G z<1XKHWU9)A{>j!UR~^6!x8lo|sbu zqCIkao>$^6fv{)ZckvaaQ(j>x@$W45_cE^gqM!5)mgmDy;Q#r0!(qBy&i*yLSpUA@ zaQ@rzwXt_LcCvK%&l!$UwRZkpX85|%Z*a|&iZmUWd1Q2jBn(hUH}fZSF`$B+)9OrZ zt9BpZ3x|d7>v1O?bF(S?HI!&x_0IBo;qTbB3Ex(rufPN^K!$NrORl!*ZYwtQ{-llUjvn7S=59Jnbd6SBB70%wCT8Y zxxub~3?wpW8|iR(fzslt=hXZ-Jb@kTUjO!R`oQ#*(S0MBXp9KG-{7HGB@-F_IgF1k z_J{~-Qgy)B;Or>h@-L;sxWZ6r`1m>*bDotpe+e^&;;aaflhm@_sCEGTI7hB*U&Y!< z9ii4aJXALM<{m81Zc!eJlB~M1GD3<2HKha*U4VMUa zp)q(-oJZ0ZO3Q>)=C&1tisf%?@=3n1F&Zlk7gF0*luZJ2mo zycn0GkgkDJC*OCDY-+eSmyn!}xYSPhXs$|J)>ue#ayh87e#Yzd;g~yu`2n?s3V^yX zNXtSSoN95^d$h4-y1;*OI~NsRxBa}{=ZoKMyLSr*7eQhC-R_dThTY(tY0bmG$XHmY ztTjhX)$6`z2)n_0^vW;67zKr9CB0QFl)2GKo&S|4f(n%pyt zM!UI7wRvA*HXU1D)sb4`zHo#}$K~0Dz(UL~nA*hHAn=pX>c>6lWhNCvEwNQO&4Y6c zcK@MD7mgY?Ty?zi0{r}wt$zW>gG{lo4PmVjLLJt-ld`0qwlk?-CR;@3U{@tb&9LY; z>UT~_Apv>S1F2Ch^_pUy+Ohd?iMuZYR$ZiP7(w=K*1deYLhBc-ICML|4s3%T?zkYW z)d5~WJHwlcQwC};&BPROxS!7g^bks&Q_H$I;ns|yg$;i5ZzL>o$Xce(va5NAslR?0 zunG@rU77SzJ)TEVJh-3h@eL~7>*`QJzjva%rJvKT-=@FD6@Sp1-mkEmH45dtD~43%^uTR&mIQ^*bcWeXgOurZ?ZeAKuBPtfG?ln`__ePw4-7 ztF``Yp=JKwXqdlO+J9TmGP8GbvG`9Doffi3M z;+j=zWM=GT`FfdIO@tNTN4Y3QhS+S8>he#iQ)$se)iP`FbLBid^mXPu z%<5;qP4@eix4J`7CMt_dUp(5I{UcjjY86Em@0Dr!%Vl4h=H|S^wi;FEgc`6a`LQnpWc(4l;0 z-qAUy;&7=GA(G3rBqVwPWN*r39-T}0al}MbJR1V9^vvgj*Z}wts03UQIx7)4T1+;Oa#&7ViWj_ zwT!-jF2)y0pa7VRO4*L3rZ%K;gMkC-R#9Fp+1ijG{yV9!pC5B>%xF8;b2C-}K5uH0+ybOTu|T7|8tKi@1vDj19Sc|nH`fETO_UQ(a@H{~1p@e~J` z`nKhLnotH9F)Ib$n)cdlQ=)dYdSVKO1izGsP8hVnX<{IZpVG@XM3_f1BSo^LO1w;; z2$l54K+l>h2sX1&`?K^C@VEt*9=<}l2JQ;Lfka}eCXRxLq6DIl1Oj6rMTi#EJNW?g zr9ne*%wRrLMSmP)pay+WeLjek+%Pbytdc1b(s`Cr$G$9+sMv=`X%vq%>6cpIU##@Y zgiimH_mV)KWuhVsS|0YTzhTVdE{}1R(0Uep|WAX zOsTqQxNpL@M|>Q66v z&*c*QebeK?lYKK6sk(CU-OVY1wWjYa6!h-az^>=ZuDq9WVE8M6L)`H0EHiJo`Y}@y zG$5`g!mG}mCh~*h@qsH!8&(G3^QL$M!as1JSx!N314_Z~FxV1B+4Vf)0MnpXoN)+b zRSv4`kQGrflI&6siM9d4SYfii^~3e=cvv&Y2oepqV4IZ%P&t7$?!wf-Oe|;vz8AS z-xKPeOk2Y9IIzc-JeQte{u$+2Yn09}yEr-XVssWszlX>0>BX($tcwRdY7(+u^fz)` z&DoVcx_LAEw2?vVVC<;=_%Q2opN}Q^!n)Io;6Y^l3vn$S^np5vCj6vyS;$d@Acs9D zCgj2iJP-wrmkk(ng+s9`6tYl`n=i&8Ji~F_cim@Dx$E7wzH!J8`Fm@4WQnhaNWfz5 zJ#H6Wdh9Zs-F`rZz#9C5gNWz&fL**+oC9D4MG%pH$g^ILSKU!QR>!Ijp+U3Hb{rCp z2E|{sNh-~t%`%}bPNUb{4VKtqJLV>h*3LH52&i$d#sMUg)E#V~{5|&u*%gv`=2{fY zXvMG<^JE{cSg^5k^Uf_8?K=$WAnZf}?ap08$Os#8A!X!A%=i_Cw3ITOyNqa+^xf|E ziCYMP)Zo7CAi(a=Ua?LvkH-u9?fdHMqzga7DBNU8H33-?`~4(SoelS~7Ucfaonvj) z>pHKRJHrl7XYN(Ht|c%mp`{u`ScN8UM5rSyUGQ1g|EO-@*;E;7(DaQ>Vb8NxVQYY--#+dGcJRWtuGM&IMf<8&e|VI6iZxBky2v-vVg{-WQDpfg%;@qGDc)3_?5dlv>WKVR5Zi zQhgyyr&%u_6!Ctga`)D*qn=q*03|clI_7L~61GA9Bt?PzZ)NQG?yl>sz8jVeMt1u1 zpxXzDcdI4(MLT7j?s?;56cQISF(&xVldF7=d{Spusf9*efb1=4_V%WGBogrn(h%8p z!L?O82DFslI*<*L!lzcr{&NXagD0nwVg!(h^!viWm5==-yU+^l^1{XL>{m>WOwG<1 zuIJuPZkOLxm;Jw2yDF_!Uvm6Qb*HfMt#6UNq_D%k$LFj=J2^InV*6%0?m zHlaezZo;HUN4>FPZwMg?`a_T#o5A_8k3>O{7kC$&Ckj? zM;hZVk_GDL(vPOQ4g*WbB%UO668*t+a<~7K?ByQtX6P+?>?UP*Ai0@B zgfaJP0$0~}CeeOJuwr6q0)-$%%ReWZe=Kl6H{IAF;TN-nHJW zWwyM|z;xRB@G#Ew^)-JGOMf(|0!9!k!qv5WsR-hQ?JD6pKe;(L>hJoF?P+z*89aV^ z{ai0T_iuf&f`9EZ4uvT2o8CMQ4xz&BT+TP7(c_fMP)FF_qNnv^Tq8;-*H7jI-=c(m zJPl*7hEvjAZGD~>I}`AFP-br7OlVY`NStTa{QtiGSE$Zbp#Wd~m(dqX0RZr;KLl{_ zFt;_eb8)7(G5ureBcq=IpfUZPD%=;l{`Eg z?=JtIj7E`F+!Xj5x zq{!-?c@ol2GnzGS?6MHP>d@P_jJ+QUVVWhA*)ft9w22pxTStgd&An|MYt%tWJ3BjO zGT;8RN}Zccpq4r;HwcN((`Y|m!lx6^m}Q7T9=%IA5y{c>9>VkoKofK^m3C*fz?;*2 z>&_P>J^%W=kQpSN62FyJFvXB1vl#^KQP9`Pmsujoyi4FzuE{;}EUgVdr}gOJZw8)f z#Xm87a20L4zkAvFN?i zse;39L7fR~xF5)JI>)}Rd0UsKJCeK)msIT5a%LnMrQY?EMqq~?|OP-Q!c3kEf9R)lb;%mVim*OYzj2K27$^1Su6 zbH1d1Baej<83wRu)f+YoC^qu-YYB3!?e&S4R--ArBb)rn6rprmxnl+KQ(;QriFLaZ zhnF3)WfdQ>tg7gW<$L5=g48Ms8|C6>Rqrqnk5o;yyoQGB=LPg$v;w>N0cww-U@~s% zSfN<0cK!4h2Z&RbU7G<^YLyMaZwDsQ)rDpWkBgz2XN}J-+kqn63$q}dT%jG9qi&1! zz3zzRe)Q-my0HHYD251uD|AW1AP-&GY$!&e)2Ak}!&=Sv zR}|qSaPInr^aMoBUZU|X(~Q%X7sP=}ID5VcC|420r>ZEiI61UK4ZUt)yM+nUYz+jQ z31qF*C3NbC)fRaZD0{fsv821XGtdd{ATu@qW;7Kyzju*po+ZLKeYRenqGOB%Sp-J1?yRui}S?#f}9V1$+X}+!M;h<0&_F`G_~VgJV4C>Ses_E-2n> zWJy%bu24r8+LM$nRAt=?kbQac0yFNCFwJuH%9>|s+%kZi(o>@#J`PN{^*JOajd)ua zSVFzQ22>RUi6~eVpbMLtX~bK}+T%=b)THj##KL>2?HkY<=4R_kPmQpvPBT&R!75Dm zKo20dMyL#1EO)+x8?Ls@8JtWR+zPJ4bwO_e+7MX%#EYy3v6{CPju8Z`PJ^#tAWn-` zNGvu~Rthk*5<9FClM2wSF3cVv4v7QrBA5eEg{+OkDu!pTtFR0oj(I#(Xl~H5HZ<`1 zTD&zb312&X=M7)I%V7=0CLCdIxT$3Cgye+jxFa&CZ0U=(tBE-FAbCSq>>GdMgK6+kHMY$4{e7?TPr`ByCN7muN;x(k} z=<<(*7YIMEk^vK*A2ft^>wf)3Q~r1(?9Ptp%?+UY z54D{;dTU+JRyiz?{Ay=w6=H;26*wwa8cdbrL1@g$tLv8Ezy{v7qb}3;rUL<)ESn2Q z&(-ehOP@&aq>mpzjoM4rj-{f}Il`P9PO{n&~86nzram zYoax~eYS1ewr$(CZQHhWwr$(CZQI^`{+CW}r*AToN>;sADy#C%IR>mqWql9pZ4M{k z{zIbE*gFiE+qU*rx>fpZX7&8`S&z?qYwwRxx1F_4otx@Yb3@0qw*|Eooc-oeS|#6S0&}d1~9xUgp|b!cP2r?0d(CM2O4>FWkVCih(T?EX;rDj***vP%*3PA0B*HCfFIyY_@n7hIRD<(H}1WVIcqc@iTc4YV$V!V=`rBOWmS`oStrlAkm~>7@+y zRCRD#_BW*uozH*L0~aDpSe;kL2OFdhKQ$!i_(hkyL>NbM{#Anus~!nR0%P`2hzq-XDnACcXU2F;4@VvcLlC@=HacCn&vMtbm1RtcVNT&JPYF} z0b9c{ZGHXx<*gNC@&59_kgYex>k(kkD12;*<=h~VKlQ}ynzGc}{q%+o8_ZK(T7EWD zot%vJe4&iF!z6}t;*vOE?bQT|GBAk6xItT6@f@RLlD78}EF+Yqp!=hZ$>AT=#^9~Z z31LaO#-oYaEAbZN@sc7YD_h;#kRbSIBH6})I%T5B9n?a2Q<1tRxFBAJhmcidgKC_~ z@5qPHW%A?=$;C^`4bFjv z94Tuu9YK-${_rNa5xego*Gw;f5sCi5BdvKpUAiEZ|JqcTW*uzd;T6#o6mhtgRU9ja zL3o+I(N8rfA9iIx?!|GZ)qYp-A=ASzG#o>Lxd@-2Rk?)m8*)9Hv!Vn-{g3n6#QYHJd6m>qNWGT%SNVxCm1%`#5JA^KN z)fgDDA44Lmg(oTplXnc@5yY^q6z%bkQ3YRu8NE|@!IY-$4yAS+)w6s1A=2qFCfz6L zST0Bc#zFWBW~hhwJcoW?rF`5+#4bp8j=h_YW-)J~_^_8d=1|^&^WM_@K5dV&81SQ& zWYtq52wCMN&QU%Z7ZLTiLq|IsUW>$%iWhTu$SrG1U%|@~*(Id1Cog6iTa+a(+ovRg zRl`6q&y~HH3;jweP+y|gkzeccV9SM!*fHSlE(N~z;&kodv&Zn1gSh8eKkk-)&-=EB z^&3<=w>tfSL&%|)LO0jHwMyThevG6kRZy&35w~e_3utk8!=XCR!-o590*t>}HJM12 zXqR4dRHZeu&M-Yj{4&cpmU_7qJNtennjtk67ls_R7vx`xeRuQ=v4$L8y_yy82OHmm_FMmHZS-AqD79lN&B0cBqCkPb z`{^&#hv%I~um{^cv)Aug=vw|cx#NGuGwfd8W1dg9%sbpy1Cd2(Fpw)KXoE_f(So9W z_*P#XJy}Zi=LJC#k_OeeP#Ds0BmZu9F}X~>gu_UX4RW76@8E%;?kBZ;#OY1W~H%GRRZ7G{}6RT*vtE!ZJEvxBuXYTF( zocpGs%J^iI>Md0CbW!zQRVGIL^uFv;(MC0F^U9HJnLDbfCW*9^Pjqxyk*;d^h(C;1 zeIY6z7tz8kljc}Y*mzRN^zJ5qmQr^5nfK(SlAC7|0f>1^`YP!jV^=?J4N#e|b-tnK}M+2od~ zIB;COe|J>6>5+-DgMDuOr2I}!3H!hj>Hc{^nibf_ZqazJzUYa5rj@%X_fzpa*Y@x= z&jJ%}Ch@L!HX3!|{!B4sDvGK2f&2S(TSz&1^3NAq>jV4krK9t;I{I-v3KWDRew^i& zG`>35R>bZxf`)>x7qtu{4Le6`HRN4quukJKqjyrFbQt+_{4!%-p&xC zIp`b{j>z}MFXx|%fTc7B9H8X)kz_LhqFLtSAZVnhgU<#2-3M#b)MAYWM>aI*QXXN~ z6K6=&qBZDx9dbPkFfa$l;(TR={C9e)UMEU1&r)x_{sJ#Q0WU4t|rwSbWGG`m4VRHlbLcgM#!wdqNff`sVkoa^@p*6;c4aWioJ=*4ZTs&_MXl<@>};8;j7AV1dPgPQwRQ;E#VWd+MA!MDiBMK=7yA$*lm`MTpM!)%?; z5`q)pc~)x)8j@2yNa%8APi4Uo-A0C^EK3D7|HvCZKx3L2ggCz;KZUc16mip`$ytPB zJYr_RxflFP`oHV#1i@IQ{~uBHtlyL9E1zJvxUlIOzTS@0nbpXG-Uv^ayq?+Ea0+op zY9bm02dcN`F~$0>1xXC9fHzyf#E3XqD=^FZoTA7qXd*AKiuy8FiT z8wbLIZj>@YTsW?SmDZdHt6Mcp_3xaZzGi|HR^z*8M1=R_AqPzhxZ!MB$>s)IOZio* z4^k786D~*Znz#;Mp|G8;DPhrdbg`)@UEcU6L%0dEof&*anF-x|qR?DCKW!*dm@=%< z_RLP_2y%_Zi?6{}6Vs|A4tg}kCM>s0YG~IiQ0gwjCPr6QnAyLBBZD--uiqy)MGyHz zq&|X4#5*W2kG{^jLlk6&l1`(erm5H3sB$)xDWIRMuKyNm7I-ez2`XC>{#9`M0R-|y z3?pLbwW>)tMJTziypDt`V3vX9`Kg`U}N((LqM0yydkQ-Faf6B|w5tiQqJ6v-Lf_({AZrI0Yx!nbjwN0i&csic zsNq03+3x79Oo;X4iXlLw5FrAm?LydwWAzcx0j}245CLQe?-$xXYZId*${<2ob|MfS zA8*Zx0Fimyjt8#X#3D1kED{>AH$wtB31Fc0xyTr~u1znzAagz?MC~ja6shhA!GA3x zRW-N^R4~#NG8#oO*u1}ZZqS14os%FyDw+Mt$_M^rJ!_}4TPL2GN&xvUj!8>ddah*_ z{;KTt4I}*hW;74+KwN2`^p6=GaJS2Y&S?6Lo(kOh01{msd<65u)12}LiDX70*xx~n zBQLan$o5B$52xOfyKgN@Ya9;gYa(opy5l*gLtMypA|$AKAr#f`R2N_ja6gMSt4&@p zh-l$|fS+lIENAnbpVY> zV+N0`z2998DQHqJFb)})en>@5J}GIx{nN!P$cSoYEC@_Vu3#nJZ)pw>00^UEkFD9% zX0P=R3uW-Om*7wUG7w3Y^)wfn5vB)ela8sG5$w7*Ks>Dc^tdo$K3G{EI)PI7!GS7b zbEZ9QwSqR~Idc{0Zm{NT?y$Q$(<6-+4Xc89;4N6@!lzVpN9XeUV zSfUsmJp?}4Re_NI&mwbK(O@q=`PeTgy3w7Cd2? zd-q@q!0%o~Z@4kx1=}SDne4Ff5JKCN5nJbgl z+$HPvs(p|e?v`#VB1W9Bb$om`Q<*5)XfK{-1U}A2IDhY$hMH1koG)#;v^_GwlKsp( zcOdbl_RwE&c5?N~{WEuYUkCJB@rd_jSWZrI}OQI)UI*QlU@A<_K<)4 zvrb0sw5pSjiI=`+{h2xL5y&I}*AWm9R5Eo#^P&(*+k8T0&~G?;?%;(R#Ct42Yvfhp zrXg`O-g8J-+irU1EOIMPB2~*hE(ZzhPt4T2Otwq_9fi9A3RL;{a~FD;rh}q`i+NToT5+)ODM|l{=sqpbqhq3-LODm2lXxherKHO zL)8YL&sOs;@a1TU-zCNWzF^cXUDg>vL6fh%?{`+qouFF9RdKdh>#PB3Y&=BsU69vE565yu$SWKDBmxbF@ z{M&rLWdJPYwAc^rjLFAi?2&Yiyt8)uXg$bvUi}_#i(7h}wnF{Ly)CYrf#;fhWh+I= zxf@L3)>*VB`!69_#^|h=7{nEyC)P+%g;p`y$qcKYr83I8!aB-bJTP3U$Z}lOlt@|| zF0G=3+AuTS>is6;T-n&HntvJDxy~wG`Du_aaWhFRes2*?EPb#xGii=m2vy?lNk3L? z%yCjN*>ue%)pBJ=7m`V7GRT~a4xzwVP@q#ZbV|NK;!Uu!_$05kOtJj((_BQ5#o{6U zY>Ii&5V$tlIIYm)6w6l_UXNv*WXvPQ#JuC#mIgAGZL}jK%-PFu?kPp++#a3OOS1cE z{<%IpZPnSnESe~*J^?>`c&>`kimcp!G!IHsEw$s-ZNanC%RAoD#f~GTQqk5Z5UYOTdIotRfn{nci>}olwfjUQuqHfGF@son`N`#zUC|fC7K{sKO zNQX~+=1QaSQ?QPw%pm$sSWdo)^FT_m@E||CVETwZjgJ|*=GZIayjdpt!{Jf!ZK%m= z48+My`&zAFjf=l76y&ml9ev<|e=x;jkMc0Cue(y7Xm3D{b793*AW6hKeM{w}HFC1f z8p9fo+-)e2Qo5#GK1D$vZp7fu^28=%Spdo#>%7ldZHs2%uxE*<-d!?&f)EJ6%FnJ8 zKu%|w>j;xVp*w&XE}0Hu(4m1lS81VYoznIfSu#;o@Fp7^Z2}8K*JSeq@Gd~S?t>mX zEzwe3+2Y|0t$~1-XKT9{1%&$9AXDFLQ6RO1Mtc>tgpb5(DUw7*NWaZ?x@fHrKWls} zv}26F!v=y~S?HY=>k%ip zVjrMy%SMjL@Qn{A2MA#p5Rx`??%h2C5B)0iZ+a0)Kw>Uv$VZ+33R$(KcNyhV3)56b zc)$b&^O<($1f+W+NY6RIb^z)qHaAH`q+ACq!~*ej07ZWpHs$s_oUUx1SyPZMKX|B* zVS4#Rh6nX&g?5B-uD1L@f?S7q#psHp9UNwsRdE)!6T~T&^104Saq|^YgbI1)E=k=a z@}K+u*id-2{Ky`wPEwel9Pr?Sb#&%;PS1+OiW)Ely9()KlshBf!y1cmJ?3WBk;rBq z%^uoQb3m6cHSJ-Gg>4U^iEpGB;4!}G)`9E5o(s$6_H^;nweoeBL_GQCE$#*deE(Ru z1Q3&)80^oKEMN-KKjo_L?Pt4sB(F1}?x6lQ#Qw9MMiL~3-j&_gNYx8 zvY#+Ojwx{gi4*^bceZ7jsBEvbgNQ`CQ>@TT9M7Bu)wiG(2aIJSaiZiZ5fDsyi&JVG z)K}S_4@&T0d_tkhKCq&e7f{0(m!P4-Ak-;{q`p=&t|o^E#6v?Xg3bn8Ved!xFa%R; ztowwz>{54OaK^0@%(lC!r6Tc5)>Ih(#@jGEn=eZmQYtMNxWGjSq?5hpfIpaz5m$yJ z{X;TQs8a+W9`Sc>*Y?(6PKFOp%?Vb%tJqG5L)m=9*;S=9|oJ-tPovy zZ56A670QD`msg2^%w$Ps(QRx*@z8zX_2EQ~T`X`_Kd~_xOow6vw;YC->w2a zV;G6=H&($A1~_(jj{_%pcpVqXzi#{iEyrYF%xkHHyfl`5P(GfSW<#U_1LhP}Le<{5QT$~`@7L-3 zwJllS@KIR9at2T*3DH0yM*N<4QN2XuLFBpwP_Ym!{BzjC0S*g*T$=64g)N}IM@(-> zmzp}v&2taTE^oa{AT<jbjW937=uVS%>6wh~T$wY>_xUJ9xzN3n$v8q{RO3nl6S z$xFD%f<;C@3l)5vsU9;E$%9ezK%&5aq;Hn77i8?olfGZU6{^;iuc7Bz(uoeQ#{(Pp z4b&o4P*YJDW!*9pH(cx|XCY7XI;2IN5Bw0|DdF<{rwrM(p7!tYPEY3{jC{yUl4k8- zMYaT0`U~wacS7DdDnm!=g;f^-AOHjDX7O=ACyiIhqR=-osYcF3O-SrO{Xss91%{nN zKkd5+W?EFyTwRQ7aE$>by*R&Q>=^YG`OMxkPj-@HTEiA!X}SeXg9oF-N+SVOq#weyMq zfp%q);S_S&fv-gIoq1jz-^Cp58G4Go%6;ZYu3<%G%D}LkMd5#2AE2_MC;eU9NRgTZPddJ$vG^D=+$h#s99}jDL5ZqL(@E! zPWaul_lCY8I3d&wfEGb$G9;7w5oW4_M6!3{M$TQ-@G(h`Kix3>AZ9frr)bM)9Kv*C zpSapw5Atsr<7p4SxQFu+W!4A+%ed`sUM+F^w@Gb=FvTq#9B1-@RC9CXhnm^%0u*oL zi+;T|Lxw=5eZwJfC&9Uuc7qyJL4*wFBvTC1AtNt1C4sY!JLZBbpB4A@VlOj%018iXJ||&Gf2ARHq^$u+#FDQ;4nPvUCidc()iJ#$Wb;51dT|vQF|Ttl zE7_D&YSU5;31sHy8CbElBj!E|yyauT* zo8pn?Y_l>Vw3uuhsw)b~D}`sSeOjY7BVJLVgS#3~5NofsCy=>W8c|K26`k178R|V? z5LiBPW*B|(WV~~ESMgIC6$%LVKT=EXK0R)ND0gK{)q2T#&y-fpp5mFDzGIfnWfu;N zHD^ZUJkQ_KlzvpP6URKf`YtqiDRm-wg$?1prjRz^$gMHqPP_7K_wxv?Qr?7Z{gt7!ak!#ziX7>A#ExgrGyp z6y5Yx#>+qo4jXX`y5H-A}w zjLXi+GG8w&v_qkAgluH5Ua1cRj21iYCIXnW&J;>!5kE}||DZzO%ej_Dn49B;St0dt zv23S`U!rXXCY(OB<*rt9F=>*keH9JDRNelQPD&xWDE3vhHF=OXTj6QiyDa)hd$bD% zo$*RSFx=)XXsFx}Zrk}5r~e_=uWbu_u@Vz>o<1)tIi5xv^0Pz=CBwDC|0p(c1DiI- z!ev&HmXiCcy3aZY-iQS(h(~k zG`*PZaS(IcKGeffIk9++Nl=(UT7?7_oc*oCrs$Z0t&`g{TQD>llc3NO5-)_!&W`O; zR~v)C*G-_GBf=S~df>i+iTNsolP^7U&v zd)&b>&rD`KL8asPCSSJb0=qH!^v3H=(2Dtr(>w893>u}k^e`~_3lOPzfeBUogaXQw z!Yd%%J;!WFu!rVUjzOTpa?{CdM9Z*0Ey2CvwLm*zJq=WGnIlEKI9n{iyTF33D5&aG z@Lll{iNPoLchrs#McD*RH+89IL0&nm<#g8NtGy1_>9=lC(*XxHX7aLbML{&Q{ZQ~> zN)c8Q*5oCm#7yz4>r773omGd-vC?KLaV@Gjg;c1LYVbzN!uv2e(D-qzc*TZR)tg_Y zB##$)=%USgphBRNMp$p8yd^x}Vgw^!E<$UdOuMJ%0+pEP)U64eBfMnn{6nQ&Q1nAO zsMmDQ7)fm(^ubGl3Hq+OG9!8u>XZPNOaB1Ia)u2ME|%iPOj+ z9t{rK;Ct-+fvt9yr)Aj@IQet356F@o4w5I_E5rbrjWN$3SxPZOBvjD+1eYMgge z`NkY|dwddL1;HZG<}E0Xb9k2%xsdzk-t$_ZPtQcdmpsfKed_5;=fq`ipY6~Y&4p_x z!L(CjU#`IKZ;MQ;4c?w^@6W?uj!uHB?W2QBpZ4ClEhf|S5}8-IjCv=Vp_kDQnqO!q z|H<3WKNv>64_L|yY&hrw`d+D$@jOu9!xW*yH7cB?D=M80JY39MMqC}81uA@@EK9RW z#|^OW$Cad!#dz!sm4W%!`kx~=`}j}Csw9IfX*FB?#}_p`iFv(QPJ@G7$CA|TPoa4_ zyJxKAA3AO4UiF7GEI5T!>*2tPQ3gHDL$RF72!`yDgT-5*^rB_8{lHQEWSbY7i*jRJ z@=m21dP{O4UoG0(D?|_zZ^ z{`TJacu^cpr(Z(mEiIQg)TBgC+eD+1LP7F~nYWzy-EC*66O|0NYh}Gt=JV*1XCQBt&AY}5tIaJuCv$+~S^IN=qysI|$~x|&C^dWJ#Z?pS zUi0bU^a}hyh4(AKcG7~?VCeu zADsZnh~F9qfu12szmh$4U6ivom0}S3lA4tSxn(m`5M|B@FpC!&E*+b@FhH*%jZuI` z2{17rKjvm`L3eQE*iJXXVKPLoWEVP9@m$M0&K*maNoUmKdG5`ZQ)7X#!(JpEZV<|= zLt2zh6yL+9xT41`%&OjV*dQJrc)a8yO+2qX&q;=FBCXReczsGj4Ydm(7*_+SQ&<`G zM^I~t2Lx_RzC1beePy8-y2mDp7Ir~*qy#eekC3E+AXlKk&30Ldi`vXo6c|i$Bo5OE zHnL9XAW!r;Cs9vD15DbNm|B|mrH@E{$Mvl@*-jg7m#)FZHe}GcqP*Mq00#;>cZV+s zj8EPBd$;Ht(M+55sS9TfWq0p^gX-6Knh4~Y;N&jsh)rV5uj+!c>?doY4YN`Z}~9f z9V@_L0mtuUbU+ui{DHR-QP9WZ3u6Z5(9&*tjg)uOKD1?L7oVOv3HUr;BqwCByqgvR zDrP||^AsRp&3P*{sYsw@5@8}{puC$N<+_fo#zD?jO7?aU6L2wem(>%O^bYP@ZR`HK+%z@n`1i0{XJ1ZE0rMI2&OYy?T}fR7a>TmTM%IRBBSWUtst_x z^k0HP!3x?*ToX5Bu9Ht+0A~p7KqsKpWJBVRBXrDZt=%bGuHHG^$Pq$QSw<(rPHcg~2y>HaQwroq~Y)R5x-x zC!&(cuOZ$(JdEteRxSSaAA?y@i{AnttK0r46LhpKhguPOe;aGt#Y(pmbs(Qm0CI83 z9F4MWLbhi}zTt}~+32iwZg_Qj-W{H3|K5kDjsBtxKVn*DJ^yxp=>5-zo*N5e8!GPn=e@R;7s)PpxkE#>ACgYn2vd^v>~90%f-r zg(TI6=at^wZt9;|9KOn;3E3+evL>Hz&O_lPp7L|s{4P4-4^L{*8PNypXS#v^%ad99$Z)@H^;+YAcCT=yv>39j7q$;lx1Sl%qsz=qZcyFI>+N3+mWHMGXG6jPg=F{fO2ZU6B6TIn!yFz3M*WB!`G|8Dtq zcPH8W9`OF0@w$GWV)>o({qFJp%=&#b-1h8r#{I7E{QNRt)fV>RrhK2e(vrA!ZM~pu z4!l&v+n9ZZby8~?B0r{~EaOm=^8JWvOhCZyw@AK^*pIDw1Q+ ziYBy>X5D}G={Zli9HAXpe~^)U+U`xA`^Yt?c}n2XiuR8dYU@nGfYxlJI@1}uN}gnc z1(u}Fiyupw**&5ZlIXpGP(?&+=+J~$tTbyd^g~aVr5KJXpHz}9-N>sI1oC-z-Bs`R z@KG&-=uBqb;3OX&|9m*2(#pm9aXIYv;%;-RZCm&?pQ7LiYhSl>eJ}gIVK2Ns?prcPygJp5O*VU^k#rgP|K^zY*Sc%M%gIPsHAJM zB6?Hd>G^W=w6k+(`oa42?D?FlY=|ExCR=4xGWelEyPR< zB;c2zT5v))EDBvHr#UtozA`5p*>H>}l!Pr-3%05QqO=n2E8*hc;RCbde?##qqVMxE zGt&dgqdgg2g_oCtkhHW#ZtPAC0yUt!aup|uv*+B3E*+330UzZ=-uKLD8x5l5hKG$H*P9?Cv7 z43P`0gi@0}Ci$>uz4I$3J;HpAd!njVGv?gerZqAC1! z88CHH3)(#lI9yk6xBuL4#P$seEF62hi)?_t4X^(M@pd0FOWtMQ)KisV?t3Y@tgtkGW91Ua^juW^6Ckam7sEppUV6VU~VwyDtE6chtcGSS!3D2_S7(Zx~epr6_$HJ zGCx1gs5aDp%nHb}Se1c2r=Zs0;BW1Y&)JEHMGG5^uaIm0{FYf5;Org@%Pv zlY>fDI5wY!!CHVU`HF*O1K1fM1qD1Ee(n?J;Vlq@OK(NaKY{gf7A%4#UmhMpFt!g;T5KsZ3qWX zAH9;XN$)_0+ubR+X$jv16lUJ238&MENq6qqG!z_2ZI@&g%HU?b6NB5}JHn^p1|9Jw z`Woj@3@kmsdyq8y@eb|?oOmF7K|2YIfX?GnR(c1wUKQjsnZ`1TR+LYWR2UWHViU_M z@kw+7@hN8}%U6#JQid%kqZXA>OV6nhtGP{?V@OHo zXB3Y)rBW`bRcab7jybbb|64wCDOX>W+k1?Xo+AC%n-l7!&;M`rm8kG$slHpY#*C27 z-;|uYvdOe|T6Ap7EMD6IIZ>>V5)HGXd=C|5COWe!ENCbqqFKZ?UQ1M7OO*dyN+f(i z+ID+C5roxbnv@&MfKr{ftCD!k-SL(IEP6{2xpAT6nc|M@T*2OLvP0ioXbt)e7?|Q% zA=`M_s*f_TL39KPg!vj&QDEg~-cN z#i4Yxmt!=YQRXVd&g;F}710zF!A(l&(*HM($AtxO8O*1HkN(b0qJzn$U#c z&UT3YcZ~OxS#qBcP%cq{S#f}?Gbjx{to_V5fG2xi1u=SkydZJ@Rm%$O<)3pq92x(? zRzyvLp0u-M#mA8@;1008aPQDTq@QLn67lhbv*Mx~Njs%)jc+{;<`adY|1A*}2$y-Y z&x|bLUgjc~62*Z=3{1C{2)K47Az+n~%tkil9d(hJo^T|mbA<|}dmwBel2JgiXo!eJ z@hIRyt4NrLqOptewHN<4DbiU_KovEE7iqYb#&%8S{Kf-iKy-Brmd8P>!!MBelo3}N1xP@a}X)7tI|L1 zgyO`Eou1JTgq?F=beK$8$kv1g+paHiP`O4K&Q{}()xg9j`e*IR`5=>GZA zx_I414RO~v;QD$iBrP(i(-nWIiFYX|>fezLjN&Mr5Tg?SW+x(WPC?k@y{8*^G!>|n zbAII#;QoFb{%=^2Q)J#F)8Ya#4>4#2a5S^wj)n$GHLj>S~H^;Fw<) zM_j!@B494Zs|BaFPW&J86NU|5sHk&xCi$~gl<<``9>V(`GJ9h17`G)rN{Y3GRbJTa zs6PRsehYAPMAPq2#jMD%a{#kBYYy&cvw+r?LNJ*(a#nn-V_>ACb-eM&D~~q{S^Mg~ zK8gXKOVAe!=y(ssquOFwWL+2S_nua0iNgIa>&{|=SII~ijH53j_Q}EUpxm%f5!BWNeW?gh z$%~V7;Mjl+d*}k;ulI~WU0JiBGb1OcO7`{;VVOY@H)`k=31am^gwN~G>*&#jZKFjX z`^=@;UV|~JLL#LCW$|!@O}8N9FqEkj3zPhtOZEWF<7MaZFA30q6NtQwuwY~r($z+X zd}jB~;~}Oc-A$_KzDynOXxE7AK~Y!sV;5-2hPMMLFJ}hjFz(!aV9Aex)n#z4G0aQ2 zR|cc(JMys|g58bL9lg;vUPMyaj}%@#AnX{PCym$Go9kxvlSr(Mw3_-gA=nzmu0N1; zgVgijY)wv~2L_dnb|na$1YU7QY@?Y)Y3qB#C>jOP?BDvef#Otz3%kTiPaqWO3(;EOoVlqHM zu>ocX!B`z(751{S#H+aD^pg?`RUzCHq_=sA!0>*Rs;Ve3Q5UMc`khq=p^pDK%E`IMx4*vLF(M>6xu}*Kb7_j_Rd|i@Wl@Hh3Kkq{}9EzH5O&tPjpkUFn8s zk($!~l=Kf#W+-jmqcvq~(ilarP&_*_#+Ry2U)=8sZ`U$kOK~WCNd4s~0I#!_L29*) zEw1+ae;}m!Jt3Z=pa1~Q{}3ntUxu!WtJwa}ttX55*0k#J!Lx5Q4L5yj@=JAJ1!y^kh+DBxC;_-Cm}zxL)0!U*~UKtmud#W)$5@ zn4ea}8Pq>i{LTqXwDW3=L~t8QX!F#PQR!`Gx%cYW<`OSC##+^OM*vqI{HuwJ)+dGhpI_zb(kVH;C;vv955b9h__0j_NK7;;Q9{{e2I1e$ ztp&yiuhUp8nl=AH2KhTVIvqDL%&%PA<z>VR3^?3@-V>x1p_eqA}aIIm>p#I7mGq58u$ijozmyc-*&D`S% zK(WxB>OZ)E2HO{sr;r@P$ss5a7H?cFz$wUrqnUi5gTv^pqhN-P*yo!I0D#EU2!)y^76iyCP54GCkF=IPAVf#kG1DQr2J0hMwN8SZQoBgbR;)*0b`IWjEVg zG?tbB;`*vQHqyaZVL$pP^XXe~F+GM^s;&xNwfI;sAkw`` zd0uy_T@SrL?OQ-Y!K7dc;uYZ}hxENw#e@42;OIvXZOUC>1~)fcfr4u;b)2*)nwkfZ z(#^FeK^1|lXh&o%2D|&NDE5QmyXeFPAF8c@!t{*B$)oms;OFIk=lW%|;VII?FxIt1<-Iuv8 zyf$CaldP=t&1I@o&&7!v1k(1d?NMsY8{!M{tJ_~vUjmM7vaj2PahY^H+G*_Y-K!7# zp3v8=39H^6!CX7=dqKB=YEy?3TY?CfVg)5yTXQ=V9(GP(QYWZRZr-e_*WVyrZcSrj zsU(12H(;`e=btJT+Kptgvy*{!+**F+8m`>J6-0(9`~r34;M9G>{|Dfn%)Uj<(0ndG zT0hjQ^=wfwEcVTTFh@Xs{We^grfco7!ed2~pMTLRBZng4N%~l4h)AiprqDbvd7XE7 zBMQX{6?8o#v)H)aWSxVtXGA%c15^&C?pnrVOJn^~ncz^cwG93j_d`btP?WjFYC z_tyHh!dCXxTN?wqq%5InffeLp1c6gW(95ib zunVNL21)7yBpXPX3I0_bW4PcTs;U;Rb+#Z53qvh3@NDOYEfRa!Z6XcY`crC<7y*GH z%H7vQK$wQeOO?!I=FD#q|IZpFnEn~v0hZ0ERBKE}eTL|-i3*mp&`Eqr! zFm$s1J&IEG8}IesS~#O$m_XYE>1MHPimN>ukVGnvF)q~P+6;r-22ek8Pdo`pBm46# z*)$EmVVPy_(_pgY?S7LhXRcVQMlJ9;xMU*;gqO*uiG-H9m@&)vQDBlk;g3psDb16i z1ooN}uqXoi%8~KcLkvk~v2ZWSvzgOA$c_zkGqA)Y#jyD_x(K2Y}R+s1|-`zXdQ2xywax@P=GqozBlF8I1F&fI__pJz5Hy%TS zdeiB^Ys>;yjFYzrPcn9^+nG@9>VSz9#))c$ZsIDRsSOU?P2kFKI0bFy8h|n@*qtG< zq3*bQi{Ts(=6XB2c|ZcyDD0+dD!1Psoq`%b{H9bHGrIaRYJ7{P!7yD-1bZb_y8p@% z)&Jbo1}SY%g}3BlzmaX5xPYrk@Cpr|30#3@d_398d1CmFhfB(OAR*gIiH%!IVfDp9 zgy=7QD1?;~(~}PoEO}EFO_@s5Em}zZpvvHQ5te3q3uE5PwBvvn{+R(T)=dctCc@YQV-$;wBZrQ)tc^aif&#P%$kfpr=B}!sM zgV!j?drnlYpnM6$cf>^E=@q8=4y+1TR8&yTH+P&h)Y2RL6kC+qMxQCNVy0uy8X_gj=MLvn&?g0qXcjpf zdgM(E?Y7aA4oxz1Y7FK1KIqMOrL$XMe3FH`y%5cUxOqu6GpEclhhFDToZMLXRUL0x zn{IXTv>zSn_Ng~K)LHj1=~3Sm_!jQ^!=>D0$1%<0p}2VUQEk=&J%;K2{d!El(AI9m z)T1;HD(IzbPZ0AtrxpDNV+9}Y1)XICERx3hikS7IKtH#vT3POuN6j{w?NwBj!KH)7 z_HB#pAA4o9z|bo`S66D<18;%hIOPGwF9KGD>-&Xv_4P>n?hznHDaT5xSqy*CPf(mP zMAqOYw1mL|GhNgAST7=B`N0%k^ygKi{-ltCD`exV5~Pq2_xiCJX-cF_-1>^JqF9nq zdX018$@PV$zPS}$d&4DY=<1=%_EU@oVu(BKW#G-rzNmZF*TixSY&;@RcQLUiwVzr{ zNGum>GD)g6>}lvyG&p>cJwhlU^a>s z6|es7z6{@YQzFn5wFtfxg@cRicB+sB)J?kTUb?DS07>^LCK;{r?bwd}5w~{Nmd3E9 zC`Fr~*c)dt_^gZC1yst;e$mSlh!889oq$4DcW=gzh79J9AJMll{4!Ner}>KhQmQue zX-^$8M@`EP{sf{u>;WaU~)>To_evtveZ?>KzHJnBFJS~b4F$95{ShxkaR+y(m_lQ(|MGHr~ zZAeRY!+z9_ON+~q{?F!1ZeQ08SFRiY+c_<=2BxgkX2C+pQIkNv_v=|7qKD6ZmjNki z2CER1@dg~SR$~|w2`%NgX^(17t*p_VE>!W}t*t)`PVV5>( zLb{6eDtgxkWYKW_@|x~J^wy9*K#Pq^+^*9-d}))gMq+HX6Zh_sAY~#{?2bM`FM|6GysSwp$!H+x6)vKi!RP*M zI`aEnf87u%DP*#QuiB2&s^1OY>1qZ)H`b_lNOnx*GZ#jBJiSqSyb~^)SaNe%df56R zj9pZj0j^Xn?~nr3!uXw2Gj@j*X`%7B`A1~!H%gq^e559T#gH!)118=f869$SP&p3WhjC`1*D_){GPP|NS=svXIYnGL;^HE zpCduh1~J$Xd={2mjCea zMTA-O3Yn8cSj6}}PjNHFbNmEn2jzY8bp82rlPR8o*!<&0L;zT&6LJ)XJy>FqkWTi; zF{Qg*QibusX>rj(Yrnn^(qU89yf9|fjsAmU< z6$Ow5zVG|7}bV=fH%9#HuS%(`WUJgbjM<(YdYFPaBmy8 z;qW|(Ea5S2+*ngIlnI&zDl!ox%<`Y?Hv~!~*+jtk#TO7PvmqkJQA&eXROJy8A_*Yv zm#51(K99bAk?>z8t9EV)t!C!&P!c}=8o~|%O(tMT!z`QZWr>!vMICUIh0z#A2rLH# zu0a$h2p4i>6oZm6fcX;VX~|)1EK~PO^KQy<1~VDMquhZ2-CrQ~b(6lP{9L^je`W&u zCi!CYVeGkjv{wRIj~?TBkH~+Qf^gga9P|*~aaf-c344?$OSaH>o6|VFbnHvjO`ADS zqjFu_O?_(C?S4~Vy?4n5!FMmScgRF;qXLYengs7nUo{@MeRBrvZp(yxl*HY1!BvE= zP;NAOx(mmBS@yysv*>~vrBFZ!{q2y-aQR}?E4RBrm)Q7F#>=TlMt-`?^!nD}^{M#% z<<9dc`do|faw~?ApAfjD{C!Kowstn+wZh3~^Y}kcc@=gpX?KEF;Ie(poo>y$JRx)l z-$7(Rj}fkZo&Pqb%d8QWhqN ztt#~iSobnxUxc6xF~8epv3JHlW=xgFyO?qU+bedzz`gu&Yk%NN`{E{8v9QU$>Q9#U z5snSIbQKU2QqF?|pWH*QLm~7hyWArPszg>>2K8-%`B?_^Z33viCPN&|+a#tCJ*Y#% zE{?f?udgPKTMrDzcC>;d?B_9>I{X;y~ z%_xK=;*1pCT?@b4#(Cw4v`iSLNZ_rAXfASyXcLk-JVh5Vo|D%DZAF87l9OeS zg-Bf}c5R6#8cqU;d*aC`$0%QHPmhxgM_lMDXM0LBOgao*tJ6uJkP7ThRP)(JFYhWlBQ$BYenB^;+_geB`RGwdXuD6Z6X z=AE&6+EO7P_LQk9u*`eP0#tyvnI%eY#su(fVpxfFld8zE0-fh|rp~|o=Q8l-Ecw(; zYBjH(Ub!2c4ORum(iuT~&ZCH|A)_LRr?}3EZg%No5&nGtQ7Bix+l_I573bq(uSZ(2 zIpNBQl4+Vl{@$~DK#RD{kX>o3(}1vtWWnQi=yRcp6W{^+i~x58I8Gp53S|RU<=-AW*&FJ{ndeY0M=q^b21Iv zOrM#eMswp=WdHNF{f?WBUEtrw%KfInyP>w5H$&+Gb@uf-=7RdLU^n;Gc6T|uA7y-y z#+CRyMoZCOp(yvVE#rlGgSQI@fzIT`Yink6-UIQ@8k45TD*C3floiTad9#9FAyR@Z z`tq%xxUlb{(5s`WE=eu4?$*O?Yu^CTfp5Co$SOw3<*9nPk-b=Oj5i5$$K^3@ygYQ;rU0cN&SlK1#mCc^cF%`a# z=z_>>?$&?nlU*fzJ}EcGukYhtS*o}}K)5JNQW0h;$`2>NIf(b((u%q(m!~So^F>Nt zo5S3x8nurc>AjXEqx#zKo7t=jlUxS}A6Y+m$5f9sJwaPZ6*C3XA6buhWF7S)?H4}; z5Gj$=g5vM~u;M>9pC1H3BTxEt;@gVFKA5ry<#_?lb_^Rc@SBHhhcPlo6}zjrl3&?o zw-TJt6$n-;+jSd=Jy#ZOqzhm_tuVHGrXkUO2n8Dz~pkK6E>yRHm1-vrjyR47q^1|Pp3VsUf=!*C&s#O-(g z7c`^3@9S;;1?n2V@a(_Y`Iy_78vbWEHCNT&IhzCFD__2_V-biAf}Gi%-Sz4`=R7A3far-0Rha?boP^R8Zh6!nlm;H{Fx(5S>g#gbRos{DUrpzsx zGFUN4v5M#SwKvO=7x7C_fgW9PYelcZ8f zM@Ulv)z!=@SuXmEEd+uwfoAg-EZN9^`6@16dP)IC$QAVp0doc!ak6jwi^wu11RyG1 zPOT~nB6Zdr@|q(LDn?>I5=DlK9nBFmYeCyPa_|?K`8?9JiT50fc8Ml(s3@4Uqc17X>6AHzW7>_g#K*S>s ztmF>lt-^w1SGVW(N`gW&TO2EFvdA^q1ipl-Bp}s4n_IoU23QDYOz(= zal`?KJ1^ccDcg5$S74#{k_dJQ%PyW1o(`?t?Fu!JbWF}8s_SRPb@;l%r&p18L|@LL zKZlp{BSGyFgNA3XcG~hbYH!nZS!d*3yy^GK>1ptP5PvZ)HD-_>jo0u{nl~>{&;IRG zuj2pc&hc`4yVT^&O*~ zRkl>==RI+3sX7#I1PeQXm3o;Cw1)$K|4w#;>C=t&8a0fGKk}j$)`3zaB%6F>IImY; zBwAWQ{*JSVQIphPfgEA@my)3&2g#4=R^i`;Bj+o`>NM zMM$GfU#765V~qXV^dHVyRT08n?!j6*Uw5A>k3$1E>?zYtXZsPlKj;u2K-4p4>oMuW;&lG&pI*7^e*XP}jx|@{D-rUjr;t6ETmo*i63f#s`+iH*wieq>;o7 zgvYc-$1**NAs3xJ#yc2YextuRrL7p*i6i{~SlbkZyw zjdm_J&#gC)iA?PKoo+qGsqhf(52Av&0H7<K<6gjr3?BKsqY8okwd9=Yw{^m_UAYc{#`$k^B#z zXp+&^yQ1#Y?3aukM?#jwzt2T}n`uB+wOzGGmk24j{)G*}1*YIFrd#4$Y>c`GR-6#H z1vI6<;uBFRSqZ5d23vJkt@2uZn$4i`504DCjF+ znveER&zzn*+I4t+l}>IzUy?lkb{TC=W}3iQ>-8`Uduht!HFgI)*U_nG7w4gIWkKWG z>&>s{P`H*ar^J|^%W|;)DTJX8(KF)G7Yc?K*Q8Y$KnJ@J6?;NjWMPU#Ys5}IR2-4v zqi3ld3Tm34DB)b55Hx24Rkm=A3YuFLKyg+U#6`eG^ zH)wOCHk_wb&3n|#blQCuwfg-b|IAnQZa!amPjH8{H7YNxP?4lOelSdnx)MVj&P6fF zlq@+;XTAo}Luj}Ci3#c5fZQu^y;p`KxcLOX?`Rh5O1}p<^SEs6Ul#IpLVM5chm|)V zu9bzy{JZWHT+xbj(xZ6mUAtz$^~Q zmR0RdjTz#WqEKfugbBjT+H3pydRJt_&-;+4R~xSUMKKt;pkmoGk0P)Bf_(P6#2*#J za1!Xgc7g%8-5rYSn!@gyZ_sDdvF(C5u3T8T-qr5?THhCbW2g2=M*qZ=R|u6sD&dt} zLYDO!bOLb#f}yn1#*+O-SpR&_gWO?s6!>lbuK*tYkaX3w03TG{V>%2M@EM%R9NyTq z0AGmks)Pq`ka~kQ-Tf+PUIDSLIvm*jfWUL$fHHoiRpFNJE5+u~JD>)YdD>5aoAgAw z`Y1gZ{H5A1_}*3S$>$n;?X-SOZcvmVh+J~YEp9{gt~NsWK#Of=pT;(4jO>NdG=FT@ z$p3}Njv%>RwSEm2t?>Z>i2qxnb~3cJHFVOqvv)E4Py6sc9V<<3^^FNszZ3N-#-lN& zHkq+;%lJQk5hNri6$m8NX@f`aSJmcgvlMOe6L_Bg{8GE5Ti#+xwz@HzrnEJamunBx zmmg}SN^1nS=u={J4Vik{u5F ztMM^7`gxd9ZKiJnG%e!B^F-w1v};>yZ>^18%^zWp<{~kef8yp52P(A5a~K!CU0(Li zkCzo}a5jwa%xRr`KICHcbMkZaw)3PgfXOPZ9E^*By0$RFdb8K)8@Kx5LH`&Mv!%XX zG(1u~76+_EUTMm)7`O%)4H*yH+N+Qw^W)sYFz@j*(Hpw$cfGdBeNUN5=#Zng84$Xj zA0~D7Uo$y+x{BL=B7w=M7K&Tr84G1q8z!7d~4okQGoW)2~5^ zyIG4*-M22oz#iZ+d6(x^yDUi5-PMR6J93K0kxVs^xLF$CpWd+$ufY<;xEEZsmUlIn z8Nb&PY%I`H80RHm=7%DG$}A!V$b-WN9DOAovM$0Z47RLd z;LIwj*?6eX$uiF%(ZKoD&4L?7a_J+B5#P}^JyH`P9)jWJxP=ivP6&Lhc#{Jn&U0kg zuLYQ`p%Mj$k2PM9n5fsav+Jomw{}NQe}TdWpruae5mg0Z<}?rv%qC#UEml2Ch#nWc z?vs|!QmAF_C3Kl6_YwA6Rk*gZUY)z7(cR(S>-^Sob$F+mevzcEx$Agx;Ucr#+y%`a z4&8ct8rdC?LWCXXW35C{04;!P8pn*cH(`euUyi8D?wPM|!|JYDO8cr6Qd$UPj6F)l zS)(;dF{zw5m!+$hf8k|t?VFJk?oPE2i(Sn3qxx>!3D2A4YW$$UP;t4B+n2rHrsG_; zWMNAE2C1*{AdQCSk>qrXf8DBLpx2cL(FqB6y6+d9u1aSsZU$eWj=UQ_6L>=siOJPZ zuvU9JzUM~zU48t*)05$^usvh_>1Pg~u7T)pW_E~|e4e}d{5m_i{}#fR`v)s}y07q! zRA+hK9X(^!tCPEzODiv3%#M3Wl=aOiupytMFIDgsKZ6@!nS+nedJu&lxxVHn(l1aX zca!Y$5xWRFq3|G^&maVaL#5z!GnAlC=qt*#)Sg|}nH^rbR}Wp)`Kg?3s&}?Vr5%%c zBKSh6?8T^*Vv`uZry|A&k}wY4!wk(ecelUi=BjVmHW$);i5`zJlx$O&8{d^wS&mif zP0lv^%oY(r!}sucJmK^SnwQ^Y3lY{^4@Q(r;&?RkB`{_$Ffk~bh+z-89nb`(@YO=v zM2f=Dd+Okui~*6^V5(GqJT|oL@SDI1lm0vS1ST$b2X@O4HF!QquBp6edZ8uNi?(4^ zv@I)g#R zXp0CF_Vx;RW#yf|dYI$EXM#vQb5u&ri}|;2oh)Gvx*}CvbP`h`jNR9c?SC_(&QU6+ zF6oIU^Z;E(sWP(-pCu12FY9|GNmru7$ze8nsrFRYhSRnZ*GVRhIZK`1ow5SiWJ#!u zXEICB?bi*95v+CSt5cj~sDlm~h*U3`^Xc)yhfnT_LqT3|n$g{5`3+3K@^FW!pFTVq zT6pvdK}nWUb5B639m5oTQx27_qM zco(G8Xt_M4^>j*J=_@(*A$8IpmGBgYN7a?3ciqjsG~LGS zI(y5O2X#q_VFo8ZiU)VZYe}AAdz=Rdr2oiyAToRB_@DZ&s<4V!R=EmaQ&nXhc$E!6 zjV*MKabq)!UB76!{u)7_g-!XwnrUUA^SUPhHg!Nu>iAK4baeTw`pJ!+Mh1>J8%0t& zwJekLrFG_aOxUkbY&_?h-}z>?k3HR(+0jNXr)Gs9R4OByR3H{wqyzJ9aM6;iv7FBY zEfxv|VLvl9(hSlHk`BcW(Z*|@dg>4QzZX%3BumIL(p(CF_Vi>8%ocN6H6wl7rWbEA zb`qCL)z@UxHV8h|JfPihYv6ib@Qb%vRqmG&LbV67`hobsZ?4v~qgs-lv8Ax0*HX^t z8$s9o)H82K6pDowqaA%ewwJpwEAhMKZMArE8$-r!o@*s}BqLauN) zZFbws6OQNx)4S9uPScOuZteL~JFnGM|(cOWi`{^O5EnY5I96VB5@-fV%9D@@D5 zJo3WHXz$YRM=J>)MwYHzutFoI*-Ucsg-jiMzV)NJ1+y$IUl9i{z^dtgw_N=t0;E!R z=>BDz7{G_qfCRPNR3?2V#9B~Pb9`D1n)e$*S?!PdIs=w=-!_g*(()kt3pB!KJaAC!Xfui{?el6={NpEL>xY)f(quCUT`Euh^r(vOt9{8M($wyI-&wylon zbgx(Yly6#%_Yfk1s@MpbM2UMWU^<`1K5#IiV-o@mR%)0H)0j~lJfJ7H{FxI41b;iD zARWCU_i%Wygdi>`RM;U*O45i;qGA>UBY>w4GgR>c!M?a(cGJO=wUno~91kRbrjGqa zP{!TBewUTV&w_grn+}MLF<_Wd$6-)Va+nf_`fz;M3u4+|Nl!z}jyyY*nwfs`(y>*j zTxfFbyZZ;IY9sjc{va>jdgVQrgrCX7F(h9GNiDs`I62jHld7V{hbkh0Zj*FQBqbiU z?YVrqsWl;&*8;OIqR)~f^;DbZKh>#S#Ky+L4OC~yJ0pmV;@AnEc~n_2J|y|C*TtdZ zdh0SJMtS(&t3z7JaiA_H$qA08t`nsG)oDDNf*Bhyl4@&j>%FqIxAhkuaW*m&*j`47 z0L*fa1S7ED?)!LqyyU4f;E)!yz(}~6^{+S_6$cL|KMBth1Pfz|V`HByg`0$~gi)M3 zlS!Gnc@wv~!<5-u_MRt|De0anh;}KHN(;SCQetMjBH|2VXIC)$+c z8Q)MUDJuz+oEHsuF5e(62c@tg(c&uPbV0sZ{G+%%ESVXP8_91r6+mYZ)y<8Lp7Ib4 z-6S10FwmV5c&`$SIH4rhN=7X|G?H$y+sJ-C_Z>lBu0+!J_KLG|cSiH~`e}t~1w)Y} zu0A*?;PV~W6SuDym^^`3sR|SHTeGsmn}xY-F4#}MBh-=V!H-ONYjzi_L*&Jq6P5JH zH{`P1&$CZUp0gOM<)sCwAX$#V5GQ9qss*!{;WWu=!D1#nQ#pg39BXzOZTp1j;V9Nc~eo>~@Py8L~9r;}PiDa$}0Y?A!C3Yip!0gSW> zuE?(Ivq~!B2@0xXHt7j#{bbbZZJJ4cxhkeI=%*}Ejk1#ppLKC0_Jf{CMKFV^0E}ir za4jdx#f)<@vk~jo{V!BCtm+6ALn+d8SY~F=z*r#cwE-fkNk3qr13j=P24O7jY^@?G zU$KsGN(|%k6P@;uD`L`zN*{r-Ohb%;QN82>RiPIY8Kr8b{5B=C?PWD=DHFC zjaS~fuK%qTN|J*nX4cgfhZ8;3j9i(CqPAoyLll!?O}jdO>OLYGe>PZz%nF$9X5Ok~ zTR$3;cDJ%jYq~=Xqps6858+{wSv11=h1W4$eEKc?Vlk73?HHGB{<5HE$AsuxRcSff zUX)ZUzn_dzOW;S-W6W8sW-AOZ-hq7VHCHkgF1#8Ua~d7BSlrVK>4HtFh-z-$tYj!J zv<7hd2pZDlp7HcM-eq&dYco9z{w_F%>PoZ%<4bVe_o7$-e&l5}Hta-)rWU>A9nF35 zP6qAfMf2R5iGpBfr6I0Mwdw69EyqCGuS)=f~u z*`aS(*!2g271n2)%Jv`N7mb#3kc+41OSyXU=3p{ z8>6d!w=4I7+}XVF=s*88kbetkOm0QYEu(fiN)Y{0~pJj z8%3vPO=Pbz?(v+&b7V)|9qsm8+S@O`?l;qhPlVXnFE^ivhsp=~|DFE?EA?V6sQ>`Z zGX5WPEdR-Wmd4KiVN!iRe$8SRc5mLOHN<;TJ`;?&^~%C>nnE`MGArf*c@}mF9 zjKb^%BLB@E&ZlnzUq;4lk-ORYI6i&cKJ(XTj!mU4c@k^=C89L&v?^1Ui}QEBUf-|& zW(`eTYqn)ReQ|0&bM z)nZrNFAGn@=Muc05Vn}$iM+>1x<~1Bv5-AoS{n=}QTHN(Z$i7I=}apO`;x0`7Ct>h zdVs54X_OH`8;Jfc-`wqjtIZ!)vTEQzn#63|m&N1Kq+{ZTg0GFHDP;fYyk*WD8BBD) zmMJHKC{w;DbhZ4sz5g}2&QyTE6rS`$E?o;vesxqMz(UhNo?sB3HEb_pVKi!$hCt0E+ z5@+#{Tk64q*#LAwi~!aIEac_e(f5*r&Mq{?SmY=xKz)GUS3f)~0~z>V#k=_Z>hP1l zCVd=qy>z`Fb+6wa``i+*c;EaXKmH%sCcX4zw>RCAqH2;Cx?|}l@O76%T7Z+C*1luD zC+k`euLEn))lw8mvgL}*87^}=?o8GKr``J-^7-M#L*qw>Q2MIb3bjq{wI~0vQkv@2 z>@P`8%hz%wQqHO|1YzF1ODAgBAA@J6PC)|eey9pAu6~H@;>VQV4BjjCadY)_37`Zg z?vKn`JU9^~VG`#4a}TJt{j)IfLbuYVtGmtpPVacg%jME=BdTygsvm+Wa|2go`3DCs8(3bWC%`a}6h91=tnp+h<_x7t6ij*ph7BR7Ne-oq zCHi3w97w~+0TcxQWxI@~hI^7Xex=-n5b4Zl4E~ONf>}7UYgxK6u%;Fksv@G1jffz! zIKaY!8UcPnp(~By1|0*+@J?bugDbNyN$fQKo^?v-^h#~4VZ&7NP+4EOF$P#jhla`A zFyvi~q(N%`TSpLA8y%6R*)C-k$TIR#4frVB*CH@=;%-rw0j>oXnlL4>`tlVF{0m>pTK%&+GGxsAflzF+4uk=xW%Hvg@1T zi7OI_n$wCTNC&eK%&y?i>yP_FwZ-Ni2!)ckdXDu0Jxv017+{(R%of!v*;q);T5ZUu z^)=7>;2oCn2Be7+?QJn=mMSiBcNDXQ-1jN4F- zT#;BMv0EX_Cy%b*`@rhmS*g%b9d}LDQ6KltE;&0;lg$z|1YuvK&p1P$^3uPR+WlmM zJykLR0m(Bz_sqX7XV%F*a7x1*)&l0#%}*_antN}7%2OTSb1*0EXhz!91ih{S5`P1m zsGu^amUejg4eI}i>g6r=wmRvxmZ8iM$)g;eTDHN%7Cg>s@w`Lk10ppc&y;|UMSvwa zK?JH=^qI58_M{ch)LRWl3T>!eMTtxYgWohHDJ(h z_K3kgRMk(Mvi83!+x`D1dV!Ta(f<=~whXE645>H6D%*o9e$+CpXy^&&0Vh{$QIyJx zC}8r{yPBw%q$G8^i7>$yBkhauD-@C7#o83%($r1!iL}&`$wqC)MR8GFbmwEMaj|e* zT(2FVLhBf7p-E(Nc7Z7~BI{9U-C_SHe5keV+5RWCEXK7h#@!1@?SG^2h~6-62XN4C z7NakfL+iQloa2t^{65TzxI1f%7pcSY40LwOQM z3&FR7!XOZl25exKuCgz}VG)mG#T3s3(K0&-2|;0qeyFiQuJ33EeUL^t=UfEqd3#j? z1O{FR?*Kg{QS&~ru1L+5PDFqfyz$_9yv{tsqsr$cV1vr6)&ugKFF6d)- zz>b-Fj!6RT)1o81qVf5MV3|sb&v>Syk)T~m8<@*m=}L;W{7q|(H6P5Cbnv8&57+s6lf ztfz-)zb_tBTrH|Y`^CwXf|itff)Ua^k4kNxFw{MbFmX>WpM&sz@r8%W$JL{}4v4c= zUHr3RN-*2yo&31j@4oNvAF;2m*PYt_m+u^?AGD>PFR9Pr?dED@v401cX*=7xIqrI& zx%+1uNht+d(v>K&ZN*DZthjDJdRHQXy)fm{citkZMW!-Xvba@!61dp^S_;8(LzkV@ zqNEoxsBFnE2Nxf@p{t!~7xU{9TLdM9frDAbq-1E%D@|LEUy! zThJ!U6+QJA7_1m1e~~e$x|YG<8zwfJQX`;4*QEFEb{vkkN@_WphpDnU7dU$gvbF5tE^NBfUD+KUc=lAU;!mBa#>X95a=Z?@+hdE2_n?~i zRArU`Z>nQk$C%0KYhaj${KsE%EKe-jQ2giMH$k!fLNaXK!EZ+RAf|huqbl~OG5tXz zSnLWEv3pE)e(qU)6hAc*PI%Z!>wy3%agqQln+vo7gHv+8TkF zH`zvE+W8JdULX%Nk$Z|bi!+VhY*8&Z zgo)>g2$YUzue%o8Z~!mHAz)|aV`~!l+>a)NxfkMbz|q6btK@9;AT837oEG;$@;ek# zJcfm3n}`W}6X^>Kf|&ld(mtvw;cHW~Q%EXEi1I&=muJZ<^!k0%MfvzYUt7%|5+&DJ z{^fEcNJa;Y`;D7>og<>)Nf&?FO&vMqOJpcPs;Lz!*1*8qc2aM~mLsy2HjH7Aw8M`1 z&8R_*$TD4Jk|+i&pJ8ML4d%T|VtInW~#%HwVZi1Qez--pHiB_{Q(SlYE!xq!jc*O#>Hh)PhEfw0G-HcbTW#pwR6lNe zLMvyh-bA+ynMl-nMpM|(dJ?r=Unbg>UcWgFU8I>rF`Z#rL~A>!MY)zn90USQW!&2r z8}3^g^bdC+(%DUOh?T=RC<3Js77%B{2qFy?t{_S8?ptM`l{}VPfTsb6Ud`~rkWP7R zh^Y7TTnqw8qS+V*B905G4s@(BPY*RwIvr!ZD~&ywLa&6IMJ{rM0#9`GEZ{+S^FMJa zm*1F$+O~wD8l^rj(LbLC6MhV(=*U~OzU_@2e3B)$GH^PHoqnDs3cMk(LKZ;D_IG7VkQo@+xH zh8tMN{b@T#ttIB1zFEh7^9(bAR>|*jlI&wcQqrwfGYtNF;mILjwAEYt$#wltCSM?! ztHmT<>+;PDvRn*MMGD~0rVS7fN?U=}5p*zUI^jD z%aflF?oO?3u<1FR#MuK0RP%`taqf}|fgeCKPJ*6PyEQg@#9CzE<6VYedi_H(N%KnQ z^rz{~=MYj&skbzhdqlhU$e^Enp8{pr31=(%&zt9i3YVhiD&zy%R)1S74^gu-b*eOi z_v4l1mw0*Zs*ezG?t1Rx14eUR?#U&(c~U6~B1ks6 z1?P!Z7AQ=j>7tCs^T4r!99eVyIF%DOKaCkF1OX#6GG+{T`NuQIA)ib{#RIaxY-DiB zD7pBmmACs4uPP{e(%E@=lYA^cIIakLcX6rV0Qd4${jtK@+fwH+Cs#0BWMa|)S~j81 zxnM-P{oFJeL~4hcV*^^^;%a%v9(fkDpkvo*T+=13r4K>5&jcaNr7`TQv@CMoHzN!+ zWqL65XZwU;1fwdc7kmjFPb zWLeE@Nz}#IYd_k)#Mo2BmJwA^nbIVu!`M_Q#at(e#KI1+f7atkZ(m&GQH@8hq+GFa z4c?nl9_AEcjd{hm9Vf`Rk$}j8aS;zPaSHD{z|y0!?8dg$(b(!>9NGayctDac<9a{v zYG@O7^6SN*FlzdGx&NBUCt@kl`M&k)E!(A*FnQ;% ztMcRfSarWXY@=&|G|FZUA@7us$zHzb`?KKJjdL#$nU$zDZ zV{B5$dFqKJ9gHoFFEmsP0Ga}F@d}LI8aTXma8i+h6FU)3W#FS<#W80(P7yM9xUXo= zf(vi*VB1$Je)zt|xbp9PeX4C|#a?_l$4e&ratd4Cxu3wRWX0-#FI1KN&{{$`em?AA zbLO|UJ4f&S;$%)_G;1$Vc~ZSPk$P5CLYz5tNTyBroEL(yoiI3Z3m|pa*xjKeLFqUy zPX0Fm^6n?#k5KEhvsd*0@pXDaPo<4~|K?$bfIjo5pv_8X@cu9V?hEH2WK(e)e!1Xn0A8$u|;-Mmc z<8>DkU8pg)Vjju(7<~#DSNuw3j=3vpYP``TFzbpUwZW;*gAn3`p$|o4C1IW(Vuwf) zia`=;4;#0y9}y!boj!Un1U+!!yjp6jd{tCEy6Esw6ttn8u6M@50VD(({G%r4? z25UM>CdPU?J`w;o6QS-Z;U~tP9lCY*&FEK@y^Tt^!dY4@B!^h0fj;-nU83go5oAWm zK5d>X*aWlAC_o}KxfWeSEpcWi7ga7kh|c;9)nY5s zdC^m~FI()o{BBuTf|Xf+$fIKV!Y3x5k%=?j9g}l+*k>M`9JMEHw0n>-yo&b&eJcwB)Ge5ExZjx({?GB$5PV2~q(Bc_e!(^mx@wRSz})}zcFUcdoy z&wozi{dqesI8x#85*4dNHTpm$n8H^>2+Ql@&SL&hjnK#xNN?}SPgp`6jc}BSq11w? z0m@L7Pie=~D%YiO-gf>_ScLc^6hawZy@q)!l$mpU6YTIQ5iq6C;`}Rz)Pt6pJQv)~ z2{Mzm#E*QwOj*ds)<3V;Z>h8dsOng4fI2sjd&J~nt%I#Pr(P~ou1Ta4nXqCi6I(IR z_ntHesn-v*YYfkvSQ-#31dHO)ypEB??5VG^Q?qk*t-n@u(z0H$t`P9(@#4dgA%(4X zjwCN1($dZ)*GFo`wtm=V2ftuS!dllVdJ#om%9OV0b$iNH%bnpxH;6~kV4&(UWTYS+ zG+aO&^Hc`5Y{$>g^oy0GwC%{cK##}`WYTnlcecwc6>bTY4fj`TO?OPfT-{ zG4@VO9YgksvBrmoYu%pfs1(2$&0!%JGz+6*$etb*LSVF!%-8gXi?wpOqd4RE z4TAvnxH#pNJ{hB7K;~B`ZM>%jG=D}WO_wI^rN}UDRS92Z9!>5qbjMTX@X2t$>5@Xm zK9xa@blLdu@h6Xa)XVEf3ZT9x;9Ut=j%{i(Csm}G#fHy(PIv1kEf`#QhrQxlbuy4$ z=G-!}3+6GS``_lzn{dF1;@27oRZ8yLo#rdtNMF^Elu0o>=LI zQ1dNVWB{_^F+bBJgQTHEG)!rIcfCIi{1MDle(+HIO%(EV6lbCUP;9~2E~r&ry9IfZ zEWthZig(YUf;vh7@8=~5#-XZp$bx|dtkQmvP*5n!(8rf`Z*hYjs&(HU{VR9DjLCFO z*hJ^9{sz0y)mVQR>h;4JrULWqS>c=^HjeV?Dg-x=W5wC*db<47mge`7BihbSAU<7$ zCwqY@n8e{1#LX^@q$uaW4nvQ`79#JCf@s|X^kI1FyH377Skm&|-4; z`keCsRr=ew2NJUz;p9c`J==gWSe7o~INX7d#eZ%Fh-o=p6 zzm{4ah+)DJ*=(tzHWbEj1FR7$ANc+iApILHBx{Xyh#i5fcVl5CFYthzQ8u z!_gy=?4Z)y*B!s%=`$uf-!PP}12(bYyYBX)33e$g}A%ndYV20=`QLo*#LI>9}p}J=qF?0Ju zOag$Vv&~;@`||HXj(1G+I)2i>5xI- znc^*4UR_+)&aZX<-Fnw||3SU>BQ~86crHfYxwKBd2N1Cl9^@3-{8NC(g{qvI%hwi! zF1JuQ-_P+lpZGmLHlo?(0F*rK6sX116kE%JTe+kM`SnZaSPjQw*e%h)hxre1^#awl zpip?A!^z1n)NV_oTu3M1(7l`TLQMJGo+hEyk`} z40qRd`b%@<`&m2cohz<0K?&BL0ai|V2VNqpQACQ=TuPtQf2|*RZFimGd%>#$z|vk+ z>8l#qyFEVa%gEd&Hu~@$`LSYUFrTz6jSVa9}$L?bX*+f5MpAAZeTt%Ah4Dt>!`BTA8ZH@gSOi!0~9pEjdz zm1bFb9ZUQ3^L3h=^YwJL+$a?F^>9WLLM5(Xi1_F+_K_y!$`|o4QpEI2HN5&Jr5Wi< zUQ);Op0tAhqFb(~Vs#uiqt(7Mc+#*;P7;{Cr_$@jziL(+t%$7$Rs$*FcC*#*pSh51 zi7EIh0d(Zn*7psdejsrNU}g?)tr}OxGempEaIsRpa12cC3kuCKOz&v_(AQg#jQfUt zs3>g|`dzpfW0>i4qQ7DpgwuCZQw%Pn7IVXCOu`FXU7XWX73$g8)~(4VyQU}-FgNGefHO`9`(s0E z;4HR_Aj}r*>Vz5*ICtBO%dk((Eru&kjGlmumS-#fqY)h#bB^J_*irTw?FV3%?_8M= zqfBX+9WyOfKeO)u+sBO%+&pGdKoBtdFDWpEP-<8G?Z556=t8CgGjAH%fC;b7eP=at zW+d&Jw5_+NzU&;^+9DUrX3)!KjELI_|0NPKTBgp2c;POt85Q@$ZNQQN=QdYJ-+5A1 z`9Y!^)g`6{+%p^WEcWHF7uy+tWsxA^Y;K@^(8$M1d-39`rJ;&BcbqU%(yXC-R6HVl zJ7G7|bJdZksW5;i8bv+7>(35rBzcA0v$-czr0eETn9Zo$i%#dua~={X*m5MxIB8|L zn+w&Zy~u~JIAG#LaV#s;pZZ8FRP%DF4jz)DM2-xLsMbg|59$oF6s{xs*FnDBm2WzF zo!`QiORRfe`K%2>x$6Pem1F60RAA}!{U9qEF!Bc_ijROf*&Qe|;FHH-LX6H}@z9Bp z3B^qOd?jVeto^|KPJYL8x($w*r_q1Co}<4Vvz%&~!ve7MNe^Az2!rMVd1!rJU9P>ZqGps{66c!}h9yMfRaf%{RkDaX!`R zsOw7!m3Jk&9Z}BtK2@ld1N;Xnjx-Ifmp1mTP0b2*Xq!t+mhsaTe9Tw;Di21uf;2#d zOXg20D>RqUm4yHKax^U(wpQBJ+Q5Q5FaNUR_4rv#TC5%l|B5g8VY%6wpzNII_=B4b z@hkI~DhDJ`rz-S!;GI;ghQ-qR(3)OQAF$?Sm{ez(5ey3LnSiWho7gOUIQ2ulq$;jz zEcd87YgAYyg)quqno29a}k3)(R?i2osp$;->G^ddj>g zuFgcYBy_-UJw-Fm=c}1f^#aIMyRdgyv_<2T0I0O4<|qW+3vx;?{<3Z?FOIMSL$mV3 zzxf`MBZG0-+>$#Cdl0<$Y;EAn1Bj!)uqo$1-kHrEeihz_wx)w!a4K<<_ZDZ+ zBxErUPnq2WA&GY6nVbeH>vVMbH^qSV=>^kp?yW`B*yYKjw?$T6(xxLtDNptD!o0 zO_&dhCCwe%h$X@N)N&fhUKU zekEovFcpQ^*?-5Wiwq{}q9o1z8YW`#}1G4}We{P&ew!mflc0R%l%<$HyF-qdD{A(y1svs|GQD221c$tr%73ICI_w>>vxM+ zNi}|vP^<4&=Psr(bUa$8ZN`c0GxyjH$)LveOin-j5JdZ1&Xr3Th8_SknO@6lQ5B%X z*jobf0}}w~N{#$u3$z`ZQxSj>U!cl&jJy5q*t zr_yMWG~)i*jrf=kd$vw~>$0R``uGNYtp!aQV_$Q6?7g4?Fi#N$d zaxqtkdF|b34m;_KQ|UHc5G%ZV43jP3S=q--vYT0g865`Wkb2r6u-BMN`o;K>;N7aP z@1z(aJ2{sQXP^|d#}K=#LxTlpPw&B4KWo(0MVoEK`{j!E52EiJU)gbacgo#iM{d`y z=(pJ)&s{F4^1&D*C=!Vx@%>PF|DAfs(^YeQ=Ex6m9-Rb5P|2=>L3grM87#-1RT27* zym6(B;7tRpK`+DpQPC(=GkG3f8~xn6p9LGFy_Fak0n<_ERx>+s$p|)HjJ8ws+NE86 z^8{LjXIT4nrD9+{v7z^1cK;Op#XbGx6!A$x^8n}F4SxHMQ7d>E8#-B>d%nK3gHKMM zea=yAdwq!RK9Dmz<35nQ=+Q{=C&o7IhWjb<shNzGI$x2B*_* zo6JLKXo);FgS1S><25N%L8BUr905*+Mu6MuWoH;5H3CX?B{bq(L{ppM) zXT4}5iaH*$nLxDmoDBukeluQHhW~o4p3&0K-_q3M##M43@CWs zaumgga1jLK0zQIC%L)MRI9mL}QQ)y5;Ph}UwgU(xEq8gaoRz_0Dhu}RZNOO;nz{ZA zM)p17frOmDV`q%_&-HH zzT-aWlSq-}gcpDAN4O}NH=BLMn?LOwC6^k2L$&IUuy*cwfwRN3z!4lPXF{bhUsigr zDx47t^1OX}MlBK)e{P(c8eM|y%mUHhf<5BHg#0YaN?FR$4U;I5l!<%DpcAI&yC&Qkq;cfuc{FxNICkX%WP7WxeF$b2Rg|2~bEFt8l*Q z(b84y=QPWltvjTy1<~O=ieuKeln%V2AMU+iwfhH`9!TWxp{vPMC)oI6eb=$EIWZ2JvadRp3Npx5>h%}E=|GH~LhLkP-NMXIIaa@A_Tec94wFP(Du7XGp0;+B$3^=GaS z&E=js{ryx>0r#AT<5NwbS~ls6IA#$!6ipPizA6?ew~BHC#4nR`Z0!*$k>Tha?PMTO zAn#qG_V6^+Hh<6)O@-l*q%7a{RCrjy>e&T#>ZPDeKq?uz7rWs^e?bLx2|E;m`sh?^ zl%_QfgIISF7Br-JuID=}n2}2y^by}#0sY*ClD;Ltq+a^RP>`~ZH+E*n8X)R2<-uqz zjz<86x_P0Fc=2bofM$@{={!+}&8*7_Pk zfw`==)i*acg-Cu2dk?)EP}Ydn+UJEoe6_oN&fmuKStgknjX?e&ek;#qAB;7+Z{#yL zK$gosd(@ik?B!K= zTKeNA!)~fe@dd*aJ8@juEY#aJJZQ*!{8D$@C9KXB5<#=pEjPv@mmS%m9?I2et5_k)k4c)uq=P z<=L>i-6G4Y3>t*+ubbD6^XpbQ!T?qrpIz&YA?X*|aBJFUHH+BJ36YA~ugb6GMIHJ? zVDe;trTno?Eg%ALO{m7A8?dd>9)W+~D8OSjw?iKq!vQD?3JKVb{p92OPP$J~=v z{4sW_5r5j&HsyZH#x}(M`*mX@hW_O7WuthHP3KNb;;=qBFgROra;qD#oGr&YC{Se-7|2(~-26j0p& z%&6V`09f?q4Sn;dr*MultweYsBWe@oGF=jDfHv5m4rZ^Hz+yY6gL0Q{R~8meWTr|J zM#Kr1^2jBjASSO4N1cfH21DJ@jzEFgxz81@Zno%TPFCwcqPLtnQBP`YqP~VY*gB1NCr1f| z_x8qTE45qvWuCVwb9t#y$hZnX!=Y?vE!17Z+Ll`d4G4a;Y zpXo}4eBI-`Vz(}}Tv@VJ+(G8GK;miDtRg+|4Sy@HsjBfDK%yk7hh+M88T@UX1f^FJ zrgDj&3uS~MAL?r@2~Dvy;P%HOc_|vI*~q-ATbYP9_@Yp@;eaD$uKVU|UeiMen=}R8O z0Dl2 z7342-}QW1{l11LuT>O&Ax?vw1_bd8$2SvgN9P(R3`;cS zBUNtc1bjyXg^yz+xNB-#fZcxt~1%DQZQ9iqAB+f*> zK<-&cBsAi=sVnS|1tFP&eb<@7N%(0!fxqYNUHSQ6pqKqca(1MDM5h?=|4&H4)xzW7 z8<4DQW0y()pPZIKyhKP+8s$O>jTLtD^iESvb%yLGg1KxYJg@0WGbIP|a* z4F&qs)0>>ro2U{C8Uob@I6@q!5W%hi3L$|D(nrwuYPeq<4H_2@BA&je62Dmnj2)hU zx#++tSo*n1dN9g5?^x2yBoL<^@YQ3qCt0~&XV22^Rs9ZCfg3h!fKgjxpb&<(ub;Q~ z8-t(-`ak#fQ}+1U(ILa{)2Wy6t%dLrp|R~!f+3n2@~aue%$#%J;z@H!f#5iM$p>~3 zO9GOOLAZMy-{FTyzA{7>Ed6R-x(iEe{W<|ksY?64m!fMRL^%oU2zDCG$qF{Zue^~w zIqg|Z#B9nK$R#NU;5h`#)?blOv&t9z?THQd|^3XqyNpbqLIf%7uG60}q1a!5W`inOtxJ>$J|lHGR%-FgSLvC8fkL(8H+v zKwXqGGXQGx`YhluAZXGY)=yNpAII!_2s9sZ?0ZP}LusLAM}hFrz&8 z7U97wC2Aju2xZ6!d2Q8_s?)_4KM*Y(YXn|-yl+U1oWTC2!-|Ur`TZWo;LFxon_9Pe zOljKaZu$ZGf2XNo5(`ayFaUrQ`2Sv>V&?Qeqg1q-r_+Wg!tXye+QN*uK0*xFwRAbM zS_m)|VnRWv4!jr!=1g8Vzc7*yvm~GAD;|V3u^h|G_0IN=UG4(Svqv{+f8$@>sxeNf zN6=nfJfM-K&Y6!li z?BUE)Vr}Zinx)P9h4u6E+v~=~#@(V$GYT!FakLY0Oh{E43KHQ3uiFVn?c_n>{irIX zg=E@o-d(mQJVu?NtMo)_h(hJ+<^FEXwv8bPQWE_EZ=_n?cgk>^Q0g({$8wV3OO2-N zM6|R<@q=qp8;MP3Z8}YTZHUBC_Vm))zJ@m{joUuncr(hygsW?h6q`Wms8t$!cM%h< zgyugIal?7zFsgZAOr=}Y2AosZ7crU+jWA3JwgH$JyxCw1%-NE+#SYmaX_%V1OP0+V zC`IMURTQ3lSUPx*9b@+Zgo6u0$jAxGig*P6`sq)VA4~Gk8Y5Nx^mC+(z>He!q=-7| zrt75FuAGnVV6uxtjnWk%u7Wr3`2K*6+Mg5X*gR85jL2 zbrx7tr^d3VIpw#qCbmo7DAZ^1L1jd{cg{j=@kxv2yv#J<&&=o!5-G5hAK5bJL}U(G z<2n}FRdZrRn(=1b9FB!5J-^AN%+)A7`!$Vl~!jE0v{ZC+Oljyy$2ipWEkI^)$wLsCcE z1jdY*lZ@=oH5e{&FX0T*kkz8Hm?-oqrUb2_6Z)PjW3F+RjvR}NGP(FIUW-OBexBCl z@JowgbZ_M=x2gSTpEbVYNAdn3PzScJE~`#YQoBOb{N9nVKg9}`WEb%34#xJU2(eUt z7OZ-2ouSrpt?EHw83@>C$M$tMe?lcNQbura`y#9E{QxJ!jOuaVaHo3?qK-FB?4PVy zmSybz6L29+1#e zpInGHav-{l8#=&H8iEa(fn}zQj>8i5^k+4J0jG}olTA8)aObz2B=_*QtgXl-re@i+ z@9cWjpaqFM^(ieK@rb-c+C8 zFFY}ug-LclH|1Z#V&=IG^NHjPk6O%jqiyY*4FtAfpf&@eMYToI8EX*p=((S7Z66a5m$}-uIrGZt- z2$R3I5kqdZ)Y6pdWP7y3?sICXrIB>_TUM?-4%87y1e|_f87SPj^#kIY#u`YzfKL`^ zgbhnL#5K3Md4ftISw~2ql&ojE7wRj4n4)r{f~)O!?7=4$AKA1ycOHBq{Bz!qdT%N( z#F5)%=`y2KmC0-k$D+3ZCO#F&M4GQ|P<@(aJ-4D2kcDCIQAn&j+AI=J!;Q}(;soIr|lnENpcz|S1WGfG6|=If_~;rLoyy2Iy3 z?RbH;Oz-YJI;1PT=5yQ{cknFOWA({BMCN2Z_dDnR9Rd*c(FpYebZrvAD;l1M=is%A z>XqL}?@D(AJ>^h!+!}MB8uew=@pBp2N^|^K#EV(duI7byuGU;vdX=T{{0SM4pf4VjXlYtv07O&ry&F0!mv$WoScBMGn*B}gJy z?1Yd%I}WigpG-uvZlxTHP8n99tvd?Mw_i+mq7G}dF~r3(Knh`5?gE>c;isnlD~fCe zv5gMNq4wY@<#|XbIBoHc%e&RU?E<>NC?eYgs3Fkr$D)}|$FhusH?6?#(NY71kMGo7 zttPpXp)FtZiM>`C&z2JSQ|oVHC&<@PAchwa@-~QHoa0OK{RmuM= zF{FqnGR;%NWsJ$M5Xy^HLvg7r*xx2DPBaGb2Jxjqod+ql7gztLQ8cUxpq**X{ zBhlMgy9yQ??0NVX@2IcU-pOQ&BLKVpy74#|clLh3HUCo~2TXK~F(XJcAt9v@U8sonsN-s)J%5f@^M%pA! ze&_6v*)T_)09hJ5W+2qysktQ>UEDyB6wRM%z@hgClBHgt!yFN@EVtegCJlV)J&CDi zZRI$i5`x-u?c46k)9sq%O=N*d`UPvhdwiiPjbAxlQK(ZTkr9viJ^mio`iV|$|*Y0r7URfKen$A;9&_78(WsV-U(?jZcA0IR_py-7X19T1(vKi1R3sOHXCr7n8j77S;MT`x^+-l_w0Z}#@{fx(Kj;a{7r|^ zH-n2(>U7UT$=QZk>L+9%Gk3@z_a}4qrvQC+X4j}5ij!C{J6D)T7e=$guhmS6E3)_6 z?#A%c-?iA7?;r6>$#QF3h_Bz5d$cN`;kTK9waz6lt&bmZ5$DvJs zqKa3Y6yku)Y;;{nKy;6Ricr!_&SXgGk@!s}-$Tbu7)|@u(VtcrNB_O8lEOow;i(=^ z)9u-OBR`<_h&*C%e9(2toN8VJvfb(U+D5RyNk8!KzokDad|}|7Uup>+`|6v#`b!35 zv<~iX7>GD&U;)Itx&d4BgL6$B6p8h>i+(%auV~OQaC(Pozi0aPUx>MIaIxo?+rJ2W zx4v4}yIv0bGPx5{6(w^|iH zhg=t*Z=n8bo@!Q3cXSo6xGL=I8)RF0-;aZCzyl{|JT71m8(wf07X9bFo#56 zQ#uD*VRZW(R4sRa#unN{v(vIOOfTu)oW(ZzOO)E0aL55c1N>1zn=tOdR8;9Phvaqp-@ zq0i?Aj;#sdGzUSrG;Tl7Oqg#+tir*ughqhhK|Tx1v+;Z`YVOa5Ez{N2NGX1);WmFs z(=m(AbvN4s+A}ou4hS}~5B;!Wf6dZok;yr66{wgxvW}t8HV4YjYEVKjl8+nE#oYgw zkfjalt@>wqw7+iKjZF8|x%Sgiznt%i^(&(n|B0<5P{$3(-#J&!}&L})!(jpe*XcoRENv1?q-DbQ++#ot^7 z&1;L_L}~hEi4nSU@CZ%FyO1TO?QZ$*wlgpzZJAQTw#T?@%xi6xW$70FBW3&jV{zIH zq%4hFa#`S>{PnzJ;xQzXFVCzb3Z{SsHCILAARCA$dWAwiLW@80RsC{`l?p8D-5770 zpsS;5SOw5JT zIDQ$lx2pimI`(C@@(V#5)I@DV3Rr=sNH4oif6p9@Tew(w4`b`#D@qpGLh6%G@HG_u zV3ImLbl$Qpp%<3udGUrcin^}54L?|3ZX>WYXL?X4enS-c!v(q0B@DJmsV77KnMhs@ zq7J`m=xFqv8htAc|7RPPTvLvFm*V##=*-uU7?zd&d@^0Kwk4m`OQgv< z1(6Qi)K zGs=V&IR!aD6Ht&Y7V}m~<`V%t*y4Gkiu6)Q2*In^HJ$*F4MYCD9D)f5H6|kz#B37< z^clD@z_v`8wXEQ?T-cg%E*1f^WsR$8Q*)>>3_@@kpm4e+{%EYfMbpuo%b_L^;zsf# zqYJbzYY3_NP$(^FL$C(v1;W^=`v_R+_LnsZ>C_q&#Ep^!U~Abr1I&wssm1Yd^5n~g zg478Z(~NWO_-!1#sNDgu$Z-O4 zlxWc(5!wJjhzM%(#@g&}DZ=S~r|!<>I7WiFQHM3c>38T=CY+x3YlDU1GmD2B2(&A#nL{^tQTHBwOcU^#OXKYmR zXn}fS(f5RYLA=F+WqR1WFtU{Bmc+F-&#%sTWX0#E;}@U5W^}{ylK=80zHSynPk2TN zc|!4Ymi^h{Ti1n19e5flxD?X{s))2(mp``^2E6oa9lr{MX=PX}oITNnm+Rye;JK6v zG)~oH$;+b_V)!3C!UGwDGWxva3`xWU^5SEp^AkR7yZqP zdVE9u*9Lz=JWTg;y0;^&YP}{7)b3KDVt!=2i$rpJ>-A8$bQjSpAnql9W*;n4!0}H5 zW0`J``$yY*VjDf$&ib3*soGc#G`#H1YEEaYT{VxTWEy8LlJ8XJ2Nd1Z$NA~Xf;D$h zt2jYIvS8}yNqB2!ECPFxxys+m=0uKjKjAJWbocz&!I-!!bL(W_FYIQ_y4*QyJFDvi z13M)t7CtJI9#FScKp)X}@RmiG!FaQbjb&E4m?EsCo;lO0pF5&vgo>PVFs2gxa0vB1 z+z+ytL4Q2vY`IN>r+5^$ zn1bXq+_69t+%?i(3Gn8%X}*bn)9%>N;^?~8UW3Fa-I6JE8=k#Udo#1E99H+tU@zMx zwBE91MS5d^RM~*+@Tlx{CE?*C6qQ>`Z{3?dVh_p7+<{YZW?Ev@JKXWboH}n!H?@A( zE*4hR0$tCGT1VhC`nmQrp;vX=*q~n+J*=f?|wnj#z@VBlCy}sFiz^ z@(Ul#ucTG)IEMDDwZP3sP+KwWsDvw&CSo&3ch;v4<)`bH%k=ImX#CeX4zIET#d$VnEaN!*B zH>)i~UaLg-h!J>*5w6B*0!qFe-Me zYq830St&BS#BDtr>w|YF7I(bFNB)MX^9fh2O7`{F2F8x!zl7}KeZ39tho*|1HPFW^ zx2+LB`f*hnhS<_RTol!#lRlCwXmiW_p=HwF$8_`9$F+R>zQX+Z{o?RN<2yqavCR}1 zjsl8siVdiQ{0#3B783)9!M9@VQ6~H6fi(ZPd9v;gP~KybPS!Qhdy-(8m(F% zl2o_fU?ATwzO(Mb;^ifa{nPFQ_80IZ_swruC6vpTb0iqu7=}%#HAG^c&&Bq-Yw)va zatHNy2|=xoq{(Dap{PsD`Z|fMMNQ}cYj6k2Z21kbEEOw{?*R?rqRy$czRNWFGunmL zXsU<7-K6MGFY{cOY&1VJ34Tq#4!y*_gwG{zW2`&QWiY;6hlB^NeH2X&s{aS~BR>nK z7_w;U+5Mn)^(l?k(f5@MX%D*@VmuK!!(G%rR!kyZ?LDMD!C3A?elMWUJ)M1CBejG9 z6&tU&Wi&wWN~v5J?AO5AC+e4TnJY$d1E{40XaRHdoJk5$2*nl69L@A4N+=HS9X(WM zYP&J^dkG28iA2)yAtQa>>>D?rr7R-mocf5ng8WVc(O8H|B5`K^Bur^am^uf4Kp@u0 znNw_fpxJ>wmi_nN9kwl$W{cA65Z9bNfBwYyo|=5c9l2))YC98&+lJWRbQ_X_Z_KC~px2&6a=#Z-Q&%bK66xF0WXgz^_Xg zIi2~w^NU_h!KcED;EpKSL~n-L(=mp%l#`^@NVNV@w(bbP1}y#1!R^^xiOXTB?uNv18LDOb-(ENS3|uhR-13Hn5|0BMmY#ZxZSQYQLJ z5kss&$kI22fggr2MMYhN)wGRmue4>2=#tqZOEQ^zoS#koT}O{M54L zF7sP<Yr-l!2U1FKOIcdE~a5^dBhc4M9Ok z2cv=pG{NDv|J-ix9xIt+o)8NMLeP)404%o=1cn zzW}BC=bz?4X^LYnyoRv)OYxctM_fY%{|mNOja*GK_V9Zw@7 zE!CSX@=BVF<0B<>80V;ZxdGhlc?j~?OR7vUP(CY*TsGbCQr2S`Yx7h&6>`JGxf(Y~ zWf}2oX$;YO&^`CPxDmr-ev|Y`MzCeS1=<9ARTueLz%H}%{%|oh6KAg0%;kerDByna4QOLR_QDmTyuo^#$swd!$p}u|I{{)SCf+>k~>ae>~E-p(Iad%DV&g zl&OjO1*qh4Z#Tmp@=H-#gY!jkQieB{K@e(*t1~PWk}ROga43k;;i&HXGgW}*O&XJCc_ zj>iS0PZ0)DmN3wFqtqiLB0Z?vdk74KNG7Cg#Lz!w;B7KK@r%94y3}0U$JB{$W(IIf z`04|Xac{~o-3YI2#D?m;V3d{D81?I=izvI?JI>pQxo%lGo7+0vp1f=Hll0LFNx{f; zJvZfZM?sk5%o72SHYtTrz`?5+)CEKoT~3gYYH(Js3_F|Th#U0MT(@beLdBb%&LvN% zMrPixj!yE<{*f7Bi{no!+B+Pb`&Qi9z%~>gEvY5-d=o>M-Cb`aSE2$+2eTAeAI%h*P#ou~`wXTO{BWx5!-ohpux9 z5-i%bY}&50ZQHi9(zb2ewr$(CDs9`g^|D_?fAo#M5$FGW?!D)lYmDxrE7O?de zYpW!J2_k@(;Ig4Q#!iQE$yGTF%n9T~6l^1HMk3&5yH+s|aLNKhTewa%QEf=XSv;y+ z1HXe&tr+R>8g;&!GEyUrH#ozq>o)!W1)uvzx4iztV!g;L|yH&YP<<#J~Odv9Zcq+~C! zVxVTfP*AH;OrWGTuRTJ6f06E1w5--EJ+0<5u{;(I>~=s@LIgRRaUMA`R&SSg;Zxvg zy{TDI|5VbWFND6MMLRJeDS@b+3p4ncYm83b2*FO#GW%%66q|Qb0yR(~-1hH|8;uPG|E*+uK!y7Rg&TH7M%^Vv&%EvZp0k_!iMK*B zdiGc|RQ|e`_LUz?dV|idY;t4{omhd|-v&oco@LzzB?#bEQm7S+R^mV!89$1XCOjv` zd5^b2mPN?IXdYY@zQNO#W`P-rIb$|WdVCk`6UG|VosnjjFX2eMd)=FOK` zoqH@XKu5I5jgYxLN6)zggDTE1He*Nt&B*Dr+88lncc}!!fcYdGKrOtc=JP}sNLYoS ziduS``P9HOdS|N?g|zi^(S0u9GnGVUoMJu4W+JizI-05ege%*hD(PFJ-A7c$fzLLP zy**^(0-BR__&seza!N#?x_6<-Mag#8CZyY~+j#_c*D=i@>-;u;>mY}m@kX%ggBaiO zgW}lsnDzCcCh#v8VIQg3qvg@7@E3zucwN3n$Fh$>kC+wh zUs{UWF=XSLxw#^K6lNq%swf3Ic=}<@uiW;VYZkL7Q zntH$xk@*IU$vm^E!&8Tk%FH66ULliOg4LyZ<(wWnRrxGLJ8k`u0|j z|2YtwWN_-8P{n}i@G_TPOJ=y6@p?w0)@a}S<7>e%wR?`*&} zVg*q9O`N8=7|-G+zj=gP3c7vJz{XsmOP%(lU@QF$d}CIp`he-m`CY;D`*N*y$FrspH z2V_oK6M+_0w;}qEs@~`~&CB0a-8Ug_U=Qss)&*X0Zlij5{+~$Oo3Ggn^V~G(yxyVw ztLndfxvbZ)F`t7YOb4)$);pQzKjS>O>&Ma`H_bQ?%}EBc3BU#)79k=F6X zpJ#w%Vxzl48NDTa@Fc!2r5_@`kUtfcyk~^Xvhz+uI(Hwt)W(u8@!y zVe%uB@zV1((eN8U-v`f`bN5K|N87GRuFXATZyb!12BgQkVT}SR`G1oLZP#ap9WrZg zo-QbbwFK)a{7qjyh5EKAG!p!Jf4v`g-KiY@vM2RGIEIBW5^&um#h{;X#YUYM15C__ zxPSm1g7!Nn#cLv&4Hni9#*HaGdipUUt%$ehitnHrCi%aK|S3n{h&4C&W7HzkR|_NRIf zk_GGA3?g6*xTRP8HT;n4DX7oMXMLne) zB8QX{lscX_yPPgi(R3+tKwJ{uKZF-L^BA24aVQ-ukHZal=Z+^S!>UFuxi6^n5?ZRs z20+B8oQ~LH6F{6(@ho-%VaF~ES1KsvvPPuobYN#DW|hvdMY) z;yi~D!R&%MQ|~7tKK!fC?iI(?jPU{Dpi}bC@euuss^wG+3)eh19`y?;e|y2`3wci9 ztk0hbFa+4Po=}x1-XQi)AR0WdkL?F68UW)+Gt5DydP6L$oP&>6tEu$jY7TTG0I1p7 z^Zb)ay->T97H}f{R(?ZJp}xEXgFZsrS`!g0;deL8#!N*; z={H4l^80YAQWNL%fwag|VLDp&b0lt_?oabQnt@6~9yUr{{xuxsgW`xjEdqwNV4Vz~Mdx1S@oua9R<%Zs|nyaig$aBBPSVQaRru zd}Dqv%p9Gcqq6kYu)(&BR=$J`nNpfMZ@n$ z4$!b}GGh&rk|43O6)?}G@P1Gag{wSNQXElnP@t1<$lr%U1BdyTMyI^k;CuUPM*Jc` zLswg_lySXg0WrE6^-oKSiZIF2M1;|j;@vm%}G1={RC`V&87(NtpnhFTf_ zAbuT+%6}^A-l2q)FKDK;nqT;#=B}dd*CNS0tO4`PSjt-|=M(l3>9yc$-HpVA_Tx{l@1oyEVQ`zl%NUu#$_LiCtK*H6Kj0gFl%Mu1IF8PQP)v@X zI*72cK`q4-1~q=VVK9buoa(EL>PvM9?Jg-oYGoJ!OMy>2GF2G9ob?FzkFzrt0-*@E z+T{mr+}Q|8{v!#y`3L0E>R2Oxr4NS5mnTmZ;E6HTu>YAkmn&cgug&M!wGARP{iObN zW%9*N_#@**|G_v29G!(fHF2y#2e@!$k@ncEC@u+1c+n={OCKMQ~N8Bu(!G`KI3IN)B;-lq-1Zn zmUH55Y{M0FG-gVCQg&#eM+tv3&-Xb-4a7RRY`7csGu5r2JAEI01cSl_-I-)?g}v0{ z{(Wn7RYsZE%tiR$-LoMiAcGht2g0K-u{>5BN# z*ywESM35wz3$UAlN7-&IS#!TVZl0A$7CDaXavBE7Od}RUUBTd%IG1B|!SPjR;P1VZ zLcRkTUN5%T?AVqQZv{#GR7X=unQ9f2q9>C-{uvVtK~~FIE(1n4JVAgmhM)T6CEwaA0WI<3Y{ed9I1OY zWfN6hm;uk?rkGV|(h6EHT2smM&=XSM?e#+K50vKr-_m&&PAOXgo*8Wni@4O z$5E`diCn0xa#^IBrO)92lIpApnEl?-i3!-X-j&9PyRQgqW@o!h(e6F1&h=@~0p@A; z9KUdW8SN3!IhIit-_)2*2^`SKDV;(KPpOHSMR`{)5wE4GA2ni^!Q!F;W8@K8dqn$K zb>^f;^HdZi(mNZ17dvbXZ7@ogmq2D|VzITRIkmR+I8SA1H2E%Nj?}$ua;Yo&f%(Ph4=e++S~DHo#>9-$idlyoi0U0fgh9;hsvs)wIx3#BP@0 z$zsV4{~#Ozc2DR%lVh36_^TN|2%E%uuALTU$EhjJ4NZkjDAooIgi}u?dh$VmGvyQS zFPZV;)>5gF*Wi&&gF<~;Vw{h)MV)YoaA@+yeGqthbZDYC>-+QZGiG<7os!_H!8%B_ zwRg)kUL3@p7$~g?({POxLkmI=25o{6$;Xlk8>=;G@>O*`LX zmEsq@*|Mm&lBkgrxu}tqhxet5ujkq_v2t{(Em+p4v%y1tF%B=L3nJ(Z*47r<&rfGC zomi;s#m^9p`Rmmy1_Y5;>-!w)$ffH>77XPsjUqPOQ+JZK7+`;xxgXWqtZ^_*HipmZ|*F|1cu0kI>682)}%He|SKB->7XG@;(tF0X7UQs%g*VaIyng0qFE^*+bc-sUVML0qRx z+y?i=w9@hheZ)p}du|kBL~MLmM7Gm7S6Rolt$z?tOxr0c8zJd3+hp%?r(&}GS2ev$ z3Bf9v)GXd)Dnf)vz&3XJU@<6Inc1w~J1W?`CA8e8EhdY=D0xW}K9GR3o$ENkYNplo z!Vho_x>jq=4ebRp*cPZd_9ucZq>IdXSQfQcHuAA$^ItxjLekJyw-qc};6K>An*B*Z zUcDgWUQCYMdxei=2-S5y%#y2xa=GiGrnsFi5Gg>732!P)26Wxz<*C{RntbgGo}d@) z!6?3Jdc8Dm2n^oEUU)1?#8>W-wAXWUBZ^_;QMXVkRLfdA-o1>p@Jz{nRjLH|n@@Qn zXugMV!#*?86s;Uw*U*JJOjP%K&5T|OuUtt$_Dlml(gg}BSfdzhN$)Xm2ewD_4&^(Q zUFAK0M1~ZdMW1O*hx@*4L958&*v^UdHabKB%T|zzmmN&e=5L&Ia3K9Z?Irp+Y5&NK_3$2TH{a#!w^pD;k6z_gvBF`S+vrlR zT8FEyvRPc{_HZh*CmiT)Lc@1~ir6E`RrGir9=_~Mou$?}fW_!F1^mj(NciBeRUM7oU}F^7Kszb<;hcLnNO20Eaxktu%V_*R5Y~Nv8VPW0BFuwD z;>8IP^M@*=x`CMac*pVD$UGY=0RfoLuy&kn~ zK$&b-noel+VvN!uh=Q6W_3AFF^<8#EtlLQ@6~Ob7M%H}v<+n$7NC z*;^wRrAkXY2}5EZgZJz4qY+6FL~#2n*Ze8tzBcvp=Ni85TEC{MJ=N#QKPVlcfL;Ut zPX9ft7=m!{L9u-Wp#ocQ!n(8cQriOgq2R9nr?Y=hce=(Ew)_RDcFpvvxj6`t-13TK zI@M(QV0$n2T^tLT#msvi>r)s?R$#2RSH*KfN6C9TW4s;0 zQF+)(Ej-T72Qzal=4tm6>>c|c$o_$uOvrH5Q7WuT-tcP?@671K&uhpD zrkj$l8@5&g+hw?Q9NyFFP=85qZgi8EF#&ZUp+{OO{Kbeo*mJy^<>)0U9!JJfbmglB zn7}=g9v}!QJY-~#M8U_=j_zJ>h~W-#Kc_fHdP(>0=q;{fA_L{HT#} z^{=n{0k_d85}Vu4`>WayLm@@kDld!5hU`TbdjRgPVqY1kv#c`dN3GBT)@s0*Z)^q1 zl(Jz+{ohzG^6bxaSX?KyxXLCl747u$ZJ3&FCZ56HOGDf+av>qEg7h
#eG1v80w z@|>NMXw3&oYca=R!*YjE>g>D!3`RFQsui52r|005}}ALY53(f{_&xxapU=Lx+h zwY!`hiHGoO(=8j-uAV(!>?>W4i9IE{wJB0U#N_re@h}3SE)|zMK1}F-AjA2WMcv&! zu1lV*;JUxXb9w-ikj=F<_uYQF-fTK!jS2}uTcdo^Ncy?ApUZFD(*Q^9naG9#%`if( za6ylJB>}_}l0y>0Y*ho=rJ#yj5Op8Tp$8H7eV*pLm($ZzI`G&ST<`NKn{Ee!?pjz3 z9g1Ox9mT&MvArM96RJC1IS3iVa7Gn};V*|zOq6`ScaN7x7fdn|9I0#`%pd&2U1JqmSDyBm#LDC^-IBv+UU*@V1g^;_`Rn{D zCM%51V?D`iN4zk0;l>}^m0GAktMxZutzE+-AKOy(AXT*!f`!@V3f)yHOqbsml(7FK zlk8KjJ^85j|5j*$@b}zA65n~>K;Ia=c2n;YXIe&ter(@9RZ|{0Fui!D@k?s)y4~W7 z2mQ^_G+l)8^ePBnWk|Tietk``@Sv=W;P((fm9#}kg{Gr3(lPQ^f@;S>Vn5F-Uv$IdR@)^pP}3?Bk7X}T6l827>PN&-S&++Wc|c1 znLX&0Od0D^dIdcq_D&y-Dwab-6So%3J9xWRyFnnn@nT+JcxiCd-q5m#cIw6&9NhM` z7dc+BPzG~LEnX;-6te#5(+DTlh9C7+7DGf>v2X8Ku_=2Bsp@pFh%TJDG0Th&iVlep z2P86~glV;8#J>2$>i`+`O_myGC5YP)@_|ouGIq0b zzFyv^9(?ijh7L^yvxnj0ar1G$Vn5(^>mHDo4w`##;9jc@C6?xab1a8?E{rp8Y9IV` zvwTL=9j+Y>Z}%Tj5Yr|+Oep5XIQ1R6CxqYYsL>Xjq#Ay@#xA3YzyIDxGT_h*rDy$U zUouK`|7Bjmn$F&?THg_ys9m){j3$|7bcsvJb77}unXIWf?wCJQ>VdJ~0P(Qh&|X+b zqwLPR?=zOA`bGMFpY8a3zif5WEtit{{=^Of3tKj`DS9&J=t$GOkoji9-6Zu;A>_Ko z!u#CQl~!&MTmdAOs#i$H@3(PU=z)0XaR(H;!D5kDX*REDv`tA?w(_%#8_3?l(kvxa5wgVMpNt#k1x<+wxY{CI;9s>{C+LDN z!@*1mQx@NTU}yW9189X41g6;m%;b;2R5uO}>#Wuh#>Zv)uW~m- zFx@~f4ivxmr-W@Mq)>`o2LU*vOKCQ%cvfWWKsp+|SY04?o;%9SLL5GX65uPuA&YOE z7DT3i+FYsq41XpmncTF7K!IdD9(0w(4^+QJ(z=*ZtgH??FAk9h$s83rORk6%@f=m%awMPTfma?E(Q{CI%4mX!IxHe z+aY}ofBTsZ;_OfjDnq=RVbBE_u3eI%GV(B+Zl4YN2KT+1%nu+YH-z5))xGdAm8}DU zXReu?v4qCo0GPGFDrJA(BqSpAWW#2i0yC7tZA$|Gs1A_n-zXy7{BM_6-QjS2dJwxL zjqnSPM&LDhI?H4%MjLJFn&#O|es<5lCWrit@k9E6?}>S(*a3LxPL)i3PQxMNmNods zT~^OwYS0x!4H^=yN;`iEq+!_;?V#jiCW=#k@}2pSi7|SuMQvzSsgqP-6%ytu!c#D;@WG}X%2nNiB=to* zS1Q8X+tVf((MV4Bl31yWzLv*QkeG|7M2$MyQaU7&7Rm*8B09@(GGNl?)W=KyCa?t3 zOif;4D*0+~I!=v2LO8S#-y!V%op+z?g9auGD+poY;?u};+Y1EJosrd^@uf=mhwq177CvX#eLf!iXZ z_|hbS;kawJ0JO{(9Qs0F;`IJP=hsOy)Sc{1^CG)*wtke{KIS-?Q6o$HQqo!K!o$uA z7}(HXmj6=Jx?8sk*@m!9c~-dO67&;QjN!}0#Wg}~Sv`TTot&TpHV`L*TNWmy1Rd(s zxV-GGj514e3Q?;S)*y;P1z#iMX9C%oA4)ck0#Tl&=E(f5H{%ca&~qD;VS-rBPAAJl zd|kxivY-dAAswS>-PG_tD>}k!mumty<(A1<+>F?3d40onl)IJWmRdXYcKAZh)Q4f@ z$o?+gU8YIiZ8@LxZ^0%8&D7KsbLCDqHU_Q}LY} zKptTDRm}p&A&0extl+5XBD<-2`Dv{ zckA_OTC#5|Cz;oGGNe=lPlb4E1_v+l?@|NnJBBT^W9Sf8ea-T#zQF*LoN+FOyO9Pi zCYbuDVdIhz(OlE;s}m|?TS=yEoKb&sJpdP6Hn~YO?fQ};gron6HlQpvYEw-vK2r@0 z?5TCgC}^>}i9Jd>VPLWq`{oIGLU&H_&QS1S9k?P~D7|@F3;#m+Q$*R}whvmhq%jt+1gdZ`MI(H>a%Uu-$C>7bdf7ylA6LEpxt>|Y;;+*6MRti#OmwHnyouRG1h@mT)bCAj=k6pBH)|2v&MFSb6I^XlEJmHZ(l zi`8bAn({3D)~CvcuRa=|0Nf6EYe%Z8B8ymKtHl2%YccIvt&qh#Ceod^RAFiQSfuI{ zGV|Ok{;mVzEfbBN)0mKgzEE3o)V<7!2X}mjq43ZWgGxTVI zK+pBzhUBrI&urSd5q+Bd*lvU7vDPxWCLj)6of66*xZ-33|Aj$L* zn)jt@+_aQs5<>DUho1ct-|GcP;k#Gv8YG+G=1_MCrMNe3li56+nBAGL;VkfN#SvjL z>f!-|!L7t%CvPu{qi=uG2XAV=1#tZ6PdU2z9($UlT_Apfgt{=m4C2;zLBXCDs9C1t zLhQioxdcu7f=*#}%cP_W0aAMOSLU#RvHmdX3G#q_omH%=*{0UWCaRUbqN z&d9nCxUokRAWi0scTZZnB$}@d(=%+HgK=FHBJp6@?F-{DNGoM3Q{pkWK?s)ORS>Vg zlGyfmx_TbJU?1q3>)5p*Am7Y=hOC8zb=Dy})$9nc2F-*C)SgIliqZtrf3%|hFe;`s zV~=S(G+K~n?TXM3g0P`lApV5mpJ-_Pee|@?P>DXSXJvs^-fz_JzI9SC zymETU{?dhJ)_@5>5FF^my^z;R2M#+`R_9#lQY58;rmlMq0jA%#aC7^82k71`oD@LGTM6IT*?|o$O_Qh!cmeUJ?GC26fCEmH!_p zo4xc*wKmB0m^f#y{6K`Rc&5m~ocx zv!1wvONYaIl{xAh$F>5j4tN5o-3pPt3WIuZVvz^&#J)@bQwrpgl#bI;{8Qj8L)#7D z*P4LN2FsE(8Wamm-TZw7Lwb%Cz^M(3v!zAMo*hh<3Ov@>xURT`9)t`E%cH?2a70|; z*p#L!L|oNnb4UN$ZowCN^(KjK4{7+vEUuD~$U>An6udjovDVSGqrTg4g-eGyVo@15&QEnNmDx)#@#?Z25Jevh&sNe)QkVw;Yfvh z%Um-MWCFJp0%HZhbmz$q#N4(I==FYJYhg}if#a6C;223YynmwpUAnkN6Y+BfZsM<@ z&}brz^)A))_*1mSqBlm^b4;#+CqqFsSUovyKc{cQ2NeShw?-D?@DSX@q7`=?hrXcv zklyC$Lb*`)VG{lQcMHNWcqH>?j{gfo^x0**h+%p+J_bQPb3i3^8TYgYsvK@sbu*Kb zC9oye-4cn7t%ECE@0}iKg9mJ@-SFGG4e@u`EKgYL!fNX4#==ycFc>njRNl_hVI+vL z2Zs-^Pq=UUZQX_Zk3Y?&nCxG`$av~@_mcCxaZ+*+57yioy}<;-D8otPbKOi$-^4%$ z2NaRdaT6@^FjOTw-A;pb%58Gt=S!6P8bqmU)OzINWpjAnfOn|(UnKQ zFS{`6PKBc^l}KG}mVKgDL}}06o^gFJsH|R@ynyw{k7MM_9WgyR_4P}~#_`c~JQ#AI z7{qUD^>{c8{)=w?%fMqDcZ)psPj%N)$lwt^BU|LmT1=XWV3`0%?U3xy4FTB)Kp;x# zEc^_WAn7+DiweVs@l2#DB#AkG=9R~n1#EYgk(~Ezh>v-nf_36hl+`9rqsgd4Yzewi zHp;9fJvWU)=s1<)2WeqOSUDDgKu_h9BYCuLnl*n6B{5$MdHakUwT_Hkm9MYS&RZGr zGGaO2Q$68t9?}QR-D+zY7;9~F)?C>_UAtS}4rGNEWfe2O`VOh8KGZu2KPV6Xl(}uz ztoTqY++KYJ{%2(T*wk&e{Y$wNf%`AG5@R{(s^H5~ixyZDB#?(Goh(lKe8ePul3GfHaP;{hzT@9r=X82ri!50*1G(*g z*pCrXi1(FHBN`2q=wOa$Je$6io9U}w42F^Uu`U|c;LQUh)#Il}T;0F&-i`h-AH3T~ zS)V%2U$SX|-+jhC4C*iC28oLbA%z&(-8xEbjGXaK=m+VUAndnt+u<8O+S|%bTGrQNkeMrX~jUL4Ob)Fl9AS#Ayky8HJQH)DOUG7dWpE$Uk(Y)=Vsh{38zaBE1#fD-irh4w}Y_V`^P?Ux~LO z?{{#B;*6~rxE~yPlipX%=y~E+dWB=kx#aMaA16oqjtNCVmi>~7haGy_S9h$BP0vNe zLV)j<(KdO6+i0!U6pvc5FbUr$N3_c4=Z*FZpT8DjBE=_D{P(*r%c-N(=xP$Wj>dRP zrK2{T)uLLx^Yqeml`fD(3}buEmZNIqD?mLG{lopR6MtUq3E&|mY9Ywnpg|B~w=gSE zm8xOyv1`LVtJ?B{J=M?H9AUH=F;*!!dlRq-DEloQ^@KPPC9V|=zmCp7g3+tt8naRh zCd^Y$BCVwd4OH!GKbe;6%kx7`j9elEMADPcnEFyQP~;H)XF4|hL~$U|UdWpT!LkkLBV+=C%wC)a z{!HkA)f_`zoAjKEp(bNF({&l6nRKO{AHMybPxCo3-i{U)t0sl*d=beI0pmhZhm8&S zp5!mns~K?`Gn7G7Er#bGTRH}lt56SXy``1jXql@chYDk=P`#31=SHN($QsnLk&tB< zZeRw>%vxo7u_w&*X1Jvbx+lCJ)->h&U_D7xgDsE)*8FRn#FSYBg^_W}khk9t&M-`Q zftxNM3bfI=P7yaJ-s#%G)SU0wVGm9_xzN;4P9K|ha=dEG3lm(*@8YVQ5@yJ-F&K(1 zP*pU)MJoAX>B8V$a^!_Hlg+SJr}5%KpwgG?Y+e)9ibXEOB=}W~3i6#4T*NvK>H_W~3=t04iBT0L=oH=XCjAb2J!otC2bxvGzV!>7om=DN7W8Fh zntf9P<|nb&<{~=X)?hQQ_j^Oxiq{mjR~iM6(iiam1$NqGfmU7o!aCQK|4ZZE$@+gM zNm`zc8zYE6y1Dz(&0rMqc~`{QD$PS>NZDm`m4sXWgygZ&xf8KX|_&;SeLp#(kms;fd}n#xgELh;8!KojBItf-=z+wD%tqz zV&x`t?Obqf`@L%9E~M?IV|HA9dpgd_8Zai9+h|?^Kg3Ryw)yzD%ynJXhsNSQI?N%R zOMwp78A@vo-uh;yr^g&i(ghlJpJDMH#4x3^oAJDD;NR%or@PW(R2S1l zyI|hLbJcyar-IC5{2yVrG0VQ@vrgOlXD!! zT3##W4Tj_D5rdt$t}$)dn{nE4y5#(r8G3BDa}G4+%y=t7c{b#j?nYNT=E38Zt>zq* zz=6RT^mTaCtN0iRiWHoKcvAfmKnpSL@L65uE2JV$O|XDpi0d&tP3Z|*L?9&QK_rZO z>H|63@Sz{`zjdBL3ye)M+(c{LpW?S42oGwwv@@ljrLm;f3#}^sgefeUkj$ni{idhf zA+6bgnuzuJtP*n-zwa0GCm^ZNni)}e+hM{@-tkn~V8xy#bT991Y&2Bkgkjq+!n{zu z^6MxAae1Y!BYoycpQlOI4J{d>ZRejDXg?iW6R1d(4NVV_=_{gUrs6GX0ZN?LUUAf# z0m(EXW??yxhhja&gplm7LGECDpzfl-^H#-X&MafCuxrnsChOqCtBhf@k(AtGbnXWo z#~AT2JNhFM|ov?Yh#nw8D&4oB7H$c z>(-RLE*UrO^d$s);Ysy59MyUs84^hWUi5+ix|yxb*H7ctHbw*6&*U(Lw54MjrWAy3 zdvE9&9LL*v?0Xp+Z>&vi2G`l0?Zb^XaR`;0UKyGdXla(i70=;TL+>~^WjWNpS4un z?1=zFZz^o8WiOQ#MqgS|?0G63m*9z8#{|ereYoQ=QsQ4D5dwCt=+Dj4spK+ote+@J z;Vk{O+1$k7^4%NBSW_N3W~_lSIQEjx83yaiM2S9bMYE+Tnytj?F!knF3v{O_5}vPS zABzGymh_L3@%GU*gPl-LR}zE%$(?K{jb69oiTD*@aF6mEk{;Mc&mlryI|YUh_r)}4 zN~fGp*pFCP=*aC*O&02G#{5O8?!4S^lAhB@`!qKDKHfaj474DnA0>SEWB<$j>R}GR+R~R(=Ui43ak`) zFlyx-9&la_O9*5wsKS6h4iix^4a~Pz@6)?6MIw< zy|tVD+GS>jvsZvlk2{4z(!<|A%bAb6FcCql->gh&bBcsq`Q`Hb{;km8FB`e64X+xU z;5vJGK^S>|fTYgs*0s(ARzlY)64EA4y{cc(Ac=tQ7m!@f(98Ji)IY|o=CwgR#+8`2%crD+ACljwAR&f<)e>5q$Vq*kh$J3{MuSihn(ZxTN=zV&Zd%rM{f6@%`kR)Pu0yrAiH=KO zU*P+qYAZe%K5}->Xv(@7z{Pg`$HUiUfmi!7M_Vg5SATD7&HBTM=C0qtg&$)FG=+K# zm#kvTuD=_cS4;I=ivAd^0siyuxFB`g81Xe@!n8tOy}&v!UYPeM-Ozo6Jco(owb^GV z45R#U+mD#gD363oEEjW#5K8n^d;}j^sbH7DQyo*nshJw=%Mv+oyJj-{XUO^*P%w;0 zVbi?gUb8QB0K|8amYipLx}VhKJ&lhJsT+1-cU=6jp1JiaP26xIfJ~UM9w#XTw32B% zJyKuZnpX5k{dP2fxUI`bR)0w5j=sdpO@VA&vTwsjK(p>s=Fi)sbQO2*%+i}HGKts# zg;aZOuJj9eW=y=)*}NPG4q+HuSI<@x7b()fWkrItQc{u0;E<3SxFh`V#Pz!Qyzv6S zR+&JN%r|14SG+3iUEa=*U)bROuHaWVZ7ZrlDW;97iqK03 z$jXk(bSH^tfLmoB#t4=Rhb?(k){SF-HGZvppV90U3;7G>DQy)YTX1yEQAr1z@W$)4 za(JspagPs*Th+7m%dy+7)W@K2jg2gq`Al#&AIsr)14O&KMGzMK{ok^ENl}S210lPF zz0S;3-H&yQvRgB*hs;h92;n4 z`G#cht&BVVF!&DBa|8xGjD_;Y=ncDso2urRcD^|1bv+YsuLpf)J4>aR5H>RjH;=t~ z+Lb=e!WaewC~NtxA|6Hrj3)WN@F@`{&`IxklF*{nYn&Do0%Xlj2>?tj_g&#(S(L*r zif^jLMf^|Bu<3)-OHHsv>3uVa}~SWDNk^ouVkZ(jDT&73`1&>G+@!cPODAFvCCA4AJexSgzs~Gon6i*6P7K z8mU5!`aBTZ095p$r)vD@bC27a<>60{|16~C0i_#m+~t8Mv058`z-4-NKsyGsdv5xw z1afMJo`LNZ5nRg}KI1LiHVj3j`;jL`KdupB0RagdoV?4c;O?UAs+G6ErGa7p9zEx{ zNfyw}s4$kP*=v)#IkkgMfmeY~i|NyKLfL0Mn;HPLfnN|tG-YU);|tf|_t*O^fwZex znkut=%?XDEa0=ljbjU4q=^MX`4r|;)#u+<1goy^uIM>E=aB3~0m{);Qta=K7bD6>FUE33AU&{m|KIPUdA}ZwT+M`h2gFnxf z$_b9#kRSAr_$@3i_lGKSPTy3SOE4v5jx6jws#A@r5{oD|LL=dTX-S`}-&yf@4eVdT zk!X-Ch5n&Sw1X^d+;GL(vpOAs0Bru^D@Z0u(l7+#qrqlEt^X3@io2MhN8&k$(OgAN zhA{;$7xFUKz@$3OFKagBAWHtV*r$4rMW3)!=Z21~Ebk}8z^gz7DxD=^$=b1t_$!Z} zY?JLbMGoCf=9+M}P%@7pcnX#pN`Fzd0=c*hiXDzqj z(k$0Z(n;gSb-;N$Id|ef0{6@{-6!W!V|Y}IYaJd$Eb^c?`VjZDTn=bTWFgK2PkVKVvK)@`ljtKD5*e+@*Nk>_eRw;yy6>)F2}Ep{UTE zOMOCAgGtg%2}Q9dxZvTMv8?-ejr?lFwdSVC?(42e_rs3UrZXxZRP_25LqXk z5Osn^jSwpMpfG4t6Kvds$~aIl_#^X(zliw-_lTguBer{E!P91(h|Jp9_)*F}%8)%^ zR}g5P79iwEPY_@N^kL5sz#!Z7A)^m){hZ+U{kOhJ09S5*?6kR?f{gziWj6Umo&Hre*L&ySp0<9GTj*e!b;+p}g+R&ZE$ZrwQ zp=-G%b9Wsxo;6wz?+}Nq_7-%%l^k2{bXYsj4r-(ncxXP?w65)}5af`F&wXj&@VcVh zB;s1a_pofch_pKch!?-+0I4ckU8bPs*if_IZFHHyHU^f-AZu$0@F3WWOPBD*p7+Y( zI0`+m zn+_?OkF^IIqce4vU*qZdTHEQy4Icy8#LC-O$yV=2rmoDpiD7HfFO>FF_bfpudaz-@ z44wm^k>#;o&Y#QGT&MIn9jT)_Rb0?dzWKlSU?HzXI|qNI53Vm?Gljk6&3$f_EUtBG zcw_etF3uNk^LKfAK0fAcFmGrveWl}Tb89wW!_yz`F$;Y^G!t+8wotG{m6@0yMhuoC zSC@Wlu3v9`-0aF(RR66~jOyg~@SDmmX<8m)jO6Fh#M_*btlQapcBFL6oyLx+Pz#Fnf+^^J4E*jg8H~` zBkG$o#fs^QIUR4?90vDvXDfq^D5sg(zK{qd&bw~sg>hu}ZyTLDfwu>vuI-&RS=I2y ziMl_d?D$mKlauVSy=P>a86&+v5-OH5?H;VYO)!+%%WCF=|A<%vs1$U|B;T;YzHaKc z(qyz2rJJu?2MqnQWo__J66Z2$&a{@g+@_hjc_=rvI0wh>aXen1s)asYz9!v#{|FNB zf9VT#zn@;VHw$&-rtN*EBinC%Pp2=J0I_Jkt{YCy<|76zY1PZ%{o)2Iq3Irk_-iNH zj7x?=0!(Ao>36M7_rfngNnLgQmm17V#EvfHt&AQ5?7~a3AnA_%jgq}WFpYi;QDZoAx_FtF2O}sc&tHKAZp!ah zfAoT~(aIf%B&U$;$J(6<`?=`OVfHQ<*w zGCaF7a^Y@ksWs;@ZtxbhE~o}pP^}K1YZ&VDU;q3rm#jCngBsTw z$#IaBmoPeKHLY3$YKRd~Q^1)2rS;$ysPIYZ!&0P4U3bH1lwhur+4aXTC6_`{&`Zrz zt8;q3!|6WW|7Wxwe<{A2aIbAx#FKLUg^BNjMjpvLp!4j_%D5=|W8KIcuG5)ai^wWXB2J+4{nYA3A*n_D+;jn+qVb zc%HH?!B%KklJjTe!pR<&d(dtx*q{m}0*JBcjSFrm-hK>fHF0Wejy+&dr}8hi2`0G= zWx8h|K5m)mhJ(PGMlWaC5#)*}8!9A>CVlZ_BsJq5d1r{&N?at6ZyTsR+pp4h5q)<} z8h^08E4_AN?Y;_5Zl*dEXR?8m9$NlU#u3FUBxFrM+pvFz#z!D1UR3>P@ygF{yTrR) zsR7zuZf|PVkh->}y1GHN7)sJ%j`24%q7Tn(!_}TkWyJ>04#vDa7-&sx!#}{{(>N;O zXRsI!XBd@goo@lYFrI}dQW>5kL=`#ajA-(rkjGT}$fwCub!G_=ZXz)rNz_wqX3ro8 z71(27k69$&lpoD@@j!;=^kA$}t;1@%aWj?HrB!EtcIRotTjtUfg4`}!1+2EcJRE`a z)7o%*qm#;w^Q9OB_)%opIcVp86w*#)&E#evZeS?NtReKB*JaBSxY*Q zmoo%g;4L}@&NW!yk(+Awt)MQn&#?M)H8AAR4w)X;BXo83&PytQ?o!LJ%lN3=rZTp( zw;hGUpyhdj;)-h)g40FVZ2GtZ$gvNSPgc8gfI#5d+QaLL^>@ zo%STyd&o2#3-J?46 zT0SgPZAufHRq`f@?LM6yoqnWg_ri4}#D^uv))K?=&a@c=aMNatwue6%nYH2T|1|N) z{j6lsWSAynGZHg>y78Gc4h3wu05kRM5ULay?nu)H)GZHhB2Wk(iK)V~pVZ(-o_Kzk zbj{#L&YzFNd3sfewO7g*-CtTHNb6d4c|F|DDQhg2PsT8qDk+WpOkt$ah!k{GjXjEOMW7{wqjc091~z}>%q4ciV%b%u#f#x) zN8=!$#iqBENyREdhJoDeEqiJ}25D~3Ld?E)Lj~k~qV%K^m6`jViSx0`zqVi^b!vZ& zE@#d+-^8NyrI71Q?>uNY9Yi5$YWCsRgV2SMV?d)B6XfRUA{rbgj%QG)_D9i0(bRN) zXv*R;t?-1fMXB~$!ik9>VST1oo1u05QF39Qr{m#Yw199J;eB~E`}`e;PZe% z^@#Zs2re2W4-Fm$qHv>>j~@}lWq}h_W>XNf0YeTP5`3f9P;JWX)jacQ|N_A6Iv_j1IX!}5z}T9 zIH8hN)5rZ2d8V`eJd$4ZCvJF6e0)ucoTrNTVbLV(+IavbaAq=AxqHz2$KVw9;(i^1 zB(izdD^Q}!tT$Ggihl@0dLZ)#m(leg&AX|MuSxz7+&%O+Wr~Sn@X=Vhaqu}=Isq@614bJHI%2j!jI*Jt!n&`=|&tq;|UPY za!RFn|`2EtQPRGTytPkZvi8uD)!fJxDcO%}N|R4k$bEtfo7pV~dYG zH?U<4`J&KSWs*jvur<{t!!$egUb$t3@bHl`qmk=7kBjS0yctL-1XRjzHlwK-N`?G( z`!2X4Hq4n?9>|7p1z%^7I)a_Z2<|EaGJ{ZMfl!JEdw54uo--=#E@5)E zaSsvx6W1|)R8~@|3vbsux`RqvrUGm!elkoJ*_xti4ZdmH+vXSg8I`8F1 z;%2gJ>-`VS_X+KCnpWHB6Ddqi{C_@ikX;iCpA0lUBb(vfKv~@Vi~+i1b@WB~^jILN+ATXeFj>STH(bCT z`<3EbTsWm(ExDAW)V_zUiXc(DL7~y#1nJ~6k{~{&rS&FCInkQOT~1`4A-#!%1`F*q z8A8(T0!kFK;HYTB6JF`JVVT<#E#S8Qio3IomC$4E>bi?k3=D&T5O$xP6UcrrA`G!* z=T&Q;+Ekn0%nUISqxu6)@i4fJlzTZrK-uJGYAobo-jtG<9{sCVAO44$5?IX6EQH;qFdEE+v)Y;P6pNWz($G1jh~ zzyZ3exwCdK8VIT3Z*xpq!xmHQT`OaK=F3tnxSu3}s@GKo5UD=n-rt%YDa@hew|`4W0|TqJ@t+*qkL& z;P1cSk_iNHS1H-W+&xWP=r%>stH7^5QIa`002+1HZs~Np`7}Vp4l7f6meaNVmKw^f zzMy|mnq6?FcnZT|$LIP3PY!=YubO;9&F@lCnLlfp+58 z9X&RAVS3>@?Yz7>bnixrO#(sDaC9{mKjxol+loId(p{Og84O6VDa5H+N675*zlRE3 zxc4QIfA%aR!{zVURoqTGVDg{cXtObTl;Z8MTi@ji0io;{R!Q)HwcRJDA0j9K4UJR&0&tG z$1d!Nub&UGgq7e&5qYa|_3S@4pYBD3=Uy{3Z59YL3KV$}A?4{vd+x`sP$k62eXC_o zcs~(IOO?yqtOkqW`?rSOwMiEGc6wP;IJop)l#+qPV?rv@q2dsOtWP~L) zDYybkwT<7m4fmVDKJ>g+n?$M^O;!|mYIWr&N>ZXW$6#lfqCj6L(z}uT~rF8N0jPr0HL17u&BI z%6E=@y5P3SeceQHwf!K((cAm|3`(gN_}!+A&|o(U>dn6fgp&yXv) zm9mh8E)?;o4o-CmKW_-zFtU$FQwZwLe>SEjx@-Tj97U#N%GMh(@~0mo7-tqI%WU=* z=kb%7$!|tziL}ctt-CPsBuv0~#kLQQa?IPa$eUK zLl`P2X$42B4m8DeUF3{dt&X$?7W$fc(wH`?IhUa@b~ON$mLz=UWfs)ZOK%IO{SYSu zdz%QFe`K$Ox=(W{-Nh+h{jl-k+ee1Yiv?#lj(Pu`^JV9j78Ledy=k{lNZ)ptj~gnr z-4pw-m6V=YA4W&j!RN%L^JI;fHxvgugd4pN55UlFU*KLoIeY)T5EYB1_uk- zVRyri8~285BUi>alz#rYI6fak8}sWSWA|i&2RyMT5yMVja*9h6j@AR{*?Bhc7UE_rM+xz^8py|cZ z%|@Qi+o2_-aa2BHA<}n5(orco{n-6(;dXD1o{lm*CIlBEU1B1AVmqJS?vUjW>g5oj z9p8`l`^Vk&^Zv)v2i0}p1Xm1F3^6g0uh$L83zDy2N`FFsHu5&n*NarBz{kDA{&qbc8B0d`?*-i@z^(b<@}3?fg^C{=&h!I`n8(`d5;9Kj|4uCWEB3+ zkz`2bA2;b~-H6Eh9H5@;jR-vuN9Fi@fGtG)N}#T|q1X?ak+w*d*Di zRcA-V;z zYuD$C)o0=K+u>~&B@orylXY&v@eS;EZd;K3w*)8LktamccsUQTZ^pCfRPnTYyw6Jp z5q9>+|FdG=O*&SZQ-FZPdHv_$mZiD9E8rY!9+XSNvCQJ z8G}~(Arvh{?q|Y4SlJ@PPw`1f1%2n#OQ6rYCicdxGb^%Kg*~W#{%>6nEG>QD<6=Ac zi+Z>1bKcon!mp0-B7r#V#n`=X#1Aj2H2QFl2Bpba1S4WbJeR(-Y`rEb65bTxwtIt{J+u)yD#Ja~`w_h~pc$&7`#GBV9)N8!VJEO?R9{*(a1fS2d z+cCAWCqIY3-Jh>?m5yw^OlqrjCA(fz^`7W;QJWNh;3i%vZI*~uyh{?;n$WK0S1Xuw zz8(z}6$?YQ1|ZDb$H0!tq&A}vIt4v?etd=c{6V@4W{E|b|D7R3R=IR>sON0oxa6!z zw#{q0o%m9qySymm`+hk-K0Ye*#XurL9`RJ5=;iap#*SWE7UJEBWi9LAdAG@8?y7vG zU0^pbda?Ir<{?NVa`_c}_wi)K(VnK8ot|=LT2WS|^iKi^+wVK<7;sHKrk0`~rDBa_ z^AaVpl-p#zhP{f(hvqzjip8T`IY@OqE??|Iy9D>`yDE}um}9+lQ#{g$w}*BBLyR9M zq(m^%UEwLc6k26ZhY&@STf3<;zSv5=c}O`x9&LZ(s|>r9E_56va%t_^7x`XGxgGmR)-T?kgFAm|3;oWH(9;o#usTRXt8i6pJ7?IAWe=3wwZsi)9)~DH?5Qr z?BWiEAOWH7VI@?nSM_L_9_uAL8}?j1Bt)6V8!c7Lfy=+|8A?85a*Ls^g|@=>U`I4y zcf)MXS$)xY+kREzl*bQ`{w{T=DCjAqv8Go{hYLkL6odj}HIbN-kVV6Iq2*+RqQhY{ z>^qGN&41G*T4iqvK^tL$WyLLXI%x5z59 zoLJ?71M%)G43CpWa86^j-O!fhXd<{RN|hh;z8o@pK^pK{tk=gq z>M{3%xW@%lf7U(UxwU!)1RuK|pEGv(%7%E#H=I23Utqw`{%DK_M7lCBZ!%G)hnfT4 zWjn2HT)}ZmI34kgo)K5{miGn`d@KG~mHpDQ0UCf@FhjV+hPhufYk%JHdMnSLjHAeX zDXT($O7&xRE=XKRK_%{MQ`p6Uj7tl5G{R`+!nYM6=J~m*sFsa)B9L0F75%k2H}S|d zctbdsvg6DH_4;LX)p+o-aoI`YdYx2e*~Grmw~`K8h?kT%J5Pz1YIc)KbB(#snaM$Y zHa}_9w1lf~qS>=_z)a3p8oz)+H3D7`SCw-j`*Os?Ea%RGUu`5!@GHuvlo}WA-9J?k z;0+ICTR-FwBGs1d35_b^6mDgt$2oe&sIE2)GtygyS;6T{Yo!i=^OA>JT&*JsP;=R0 zFZ**FlVnT%EL;zg&yt|;-ofLG2}pxsPQi?-;PZiuImNVr34@23|3knUV%8GJKL-v6 zo|q{zt=~??3|3^h<%;dG-BeCht#!eTgF1_!232HUr)9+k-@kzQ6DOvsif9P%TduP8 zU^d=$4dyo`ae9WWITW+F`7fQ(CU}epf^ct0DuA_77kyGSdKaIbD+-`cI)yAD4{T5b z>pUO14L;8`&A)3Y%#2r?*U&7`NXSrDE)OKbpV11vjKX z1^{l8C+mGBp$bEOUk4iG@x{xy+LpyML@5j3SAoLG5!fat9I0x_yTa}?Xfg)lUk*8K zg1#zD{j04+9iGSwj1y#026E|BYq+(t5=6D)n2s<*L9vCqUz*(%8I!{^Nn{g*Jy^45 zw2>%8^vbNL6x>3aJ+L~b(x`dpHY^{)xl4!MQ1>vWNLHHzxl(n?jGA?WU;7*$`}Mj1|6GYn?tD@O(ge)18E@(}|lNI4W! zmJ?Djcq9)uC^E*B85NTCUwC7o)P!g(jPeAaJ>Yy`HoN+MjcN)1o`B5vWpjkLzn*2_ z6pD⪚BPrt5O>>|Lq09`YY{?lAvd3 zIK(y%3FdxwMW3SNTUK$@7TsL&3Dt8q#PHS}5PJz2;i(*xnG7%)tfj^UT#)S~@5XS| zR7B`4()t(Rs#Og~IiiU;N$E6WR#0Sb%srvm2pCf~VgWc14XsTyFQwyjPtnKkP;IIV z=(+;^YyXbe@aGW=%qlHOs_JzCQ2{Qlbu08@ZVS{o9I;UD@CR=%1`xuq!9Wt8;_cCVe2$jKNrxUd+$Htk# z;_F0~)~GuMb#KI~PTfjNQw7Ysnt_V4Uh(?pN?A zB8DCvp7117UI`)|GKU)c%4IyL8y?7#r)MByH45zN#WZ+X^J@=s1#1%HXZOCBk4Po~ zk058ReiuMc6x>*aBiF>=r@gI+ui<9wwPLCL&5pztaldK7`ZO4}F~+1IgW^J;l4PT| zk^=wQ)0G)Hlr_-)!}5rg9bq}|F%OCvN_0Ldd|F z8!&OcEH>&uh%Mr}X;fo<%Q}KkuZb_NBl>y75-3(3vr?Ed`~n(0@8B*uP#>d+uzVMK zbCrQ4)+#xUp_Wn`+E;&D#sTI=Ouu!CP5?i19ejIrgimyw0zYkk@nN}UMeH4xc2OYug*{T7ejHJvFkddbDSc?m5hOJ^SF8)@F zqz(N%MuXd%8l5LtR>-PAObJZk{NeK(mAUFjK4llmiTLdq-sXK33$H^argP(w;vKPQuk_(ayY3=c&Pt*XdQb>Z3>_JOq z`6a)VlEya?K^DC&o=@**iU$EyZ zH6#7x+Fc9dlw<-D*Rk`PA2UOKvvu*d@SUscqE2lve{Pc)0$Ot@U`geVl}i8l2tZS> z?U@cT39WiEGF}D>e!q)oxgNV9&E$K4h_{DIo)0+%tb9^1rlCt>U7p1c9< z$hxneLDe0{F}xke!LAD|c-GHC2RJ8<>p_=yr1Eufz%;!6j+?l#(6kc6i zs&|-2C2l@$w7}9_$lVm-=btVQIlmRFkpp#X;R5NYc-t7KvXKTU#)rT=SZtE$NQE8f zRsC)}!LQMhc+N{JHdVOf=`o;*#{>SJlqKd!B}0VKk2K8*ut>(AwN@utll)hX0!jF1 zd_WeZDlxj$GTl|qenvFuDy7IF!9XD%cpmYlEAYSHD+tA9hGiTrJD%6@rS|?}@AgK! zh$B6r9(HQzY(q>Y_sqs`y z#(kI1uD^dB6kP;HU7{7m2{KQTV)d%sMB!#*BCD+Yu;9r$;OVugunrA|$2QLKfqjmM z2GKBc9+IW#K|%&Nr$kZmipi#;rAt> zL!n&zH-42sP2&Q@DxrHwlyE@OTu5betyxZ80LVu(pw3roKGz(!wvI}Q_OOJ6>2f5^ zq^Gr{C(+Nc>}4kx$BT@q0?4VziDuonjJ8+ zn)yoLajNJdUK)e^T&F`Ve_wxgcyiX>c&8{B|By!xK~b6PV!ls-jVa1Z00Yxr)H(Yq zksv^IpzaA+(Ila|mruIEKs0}Anv}l0O~@614r&Yg?K^-O zTxle^pDuL>!nWZ>M~P{KIqNh-nSo5ikDFvdb#5lXC`n{ZgjlLG)8*d+ zC{zcLu>4Vjv+A`se#)y`U>LJ%L(ozIReRa-4CW=n9|iNWkhHSob@c9<^58=)PKy_fSM=93S5GyF5}WRC zw{|SO;D9l34CHaGOKgL6wepv;R7tZY{7`q1ORg z$@g$0hnF9S9K};g5y{@pYH+LRvV@Yx{9$NALL0~8>C+m}5xWQHpuw^3r6*zed=`*O5 z74)|SzG{<(jHelMjZ5o4cfp(eUYg#>tm`GaboCf-38K6D)wI&#D@mcvsbDXwTR`sIYgI6%*Du)_ZHhUxJR&!T|_(35D^u8^nyS5 zMY(fHbrT5#ARWo+b3Vb3Eu~Mvk?N{Iz-9xnW&OIj z#mqO~R!5ew<7|Fv4D(x(ON}4>SlI`za?2bO1)FGnx8xgNRVZJF4^THxOR;;8_v++@ zGVvV5BxZIfJ|zQRm#uAv>*Z&ZCF?x!S9=Bmr))?zg2!Ye><=?6Cp~*ehgMM}zc)U7 z-spq0w!r(g9Z2EHDXe$g%)u5^9E9tq{b~e?NlCZd#8KYZq66Vy{}H~EVI{FUsQeG# zL^gY}wsPG&yX<$%KdA}wLbIF|*RgNkgJy3Apww>fId@0M4sj!keP8b(mJrWtJnH@b z3NExw_n@E^I`{gv?j*qfTK2%;uivdk&@5#(1wh|0GU*>Jf)yrUYRFtbn)R6;$}CGa z8T!C))(v|^J^!)5I*cYDbM;{eWO&8S_Nq@ko>fY%-U-&~x#(%_Yu7Bs1eNFBS~r4IzeB@hQGP1GV?VKl`~7#+HTM>)lA z*d!15&gg)YFVpblIL0mzyI<6}wGVgml!3oOf_}1zwrOL&YhOHVUdyg_mP}M}GIk=b ztu&zNW$Ce}G0zdk0X@!CjMe;iouEPxsjmj9$-CK@%eyt(pdnW7%APa*@1-e9B(Z-^ zIEKOO@;^ERa2FgF^VtAqm%bSKE^qg|aIy%{%}gbXi`lloqH>;%YhURQns)H|OnPn& zv?c1u%`_F5kiA{1aG%or*FCIpZwn7ax>)sbw z^rG(_(XuhlX#l7B=_URW;jMR=)*b5x@`dFFHHLn8dXV~S3gVc>F$4UeOie6VtZ~bP z$?eGq)o!`U@wW*ws>=Wti;%cPjQs(Z@su+16Tz6jUU5j;shnCLiFq3zPzpn_04sLwj>^;%Ndvvz;B&USk@|w)bOz7H-imMan7L#&q-P#)VaQD-H)utop zdDXkQl0*hv7WPyNzQAzJysQzQ;LpyT3_QzfjIBHjJwi+{n=P@AM`*~)LgWLe1zfLk zdOY;++@32IY?FTH_Z`Zzhe-+g=$5fbT{u9b%6AuSS>K{IHS`69rzP! z!D4&bG`pieKw?wT=AfRjjRz)zP{Zvwpoq$R9zZLzO!MhBmLCNmR`?b?{cb)p{>)?f z)#~p(T6M)P;WJ~DdtqAFR3iCzW?P}i%*WBRKY7t$Yg3?jy^R2Jk+dS3e| zIW)kr5-G2bGkF8%iEdDSEUvXRa79AVXW79-jm0Z8qyfFR;89uGXC$MzxgtP#!5Ug6 zdx=PBeS-!8`~Zi_qm^Asxrn+Kk`XpElQb6C3;J=lb1 zDPGTVu{FSSQ5<23S3c7DmY7hcPx691O4Rl2M42z8JyzKx<+ZOE82s!UZ;ex6j!a-R z3l8PzKkQy6`GLW?ze2>gxBmC`br&BS!fLO4GNN-^3fF}<>{*i6(PL|%Y0b51(Y5C9 zoN-E(e%v_DJ=Xy;8|O$Cy7n>_{>IQzg#y1dTmbd~0;okAz2lbx3qeS~_gCwm zaoO)I%8%T;Qha!`^;4$BVqe0u%BOGk^Gdy+Vma1eQ}6E|F`c-IQJmL6KCs!u}% zlKX>-q2w0m_$m3#=wX+Vo0Hh7*^_PoRLS%8TpBf~8`suAZ47@xUAX(uH-QR(eOlN% z9Pt1z$97hK%>p$F&XUMS-=&~#RllU)Fe^Vn z4%7&9Fc=dHY)XnbTpuP1->k^LOaiZWZS-qSDOJF~)c#1_>i^x>PviFXtG-kOYnt9z zz~id3O}OE_YLm?|aZQ5L3LVeI1sz0${jEmY@xeoO8?O+pA-ZUiw(U)6u;Eq4TnATM z2T+;VsR?E6C8Dls!{<;@nZro-V#xeMsiD|RLd-vP)Wg$dYLTIMbpq+CumLc1SN%%+ za4rAcrpy?l@a{G9534jnV2`jb1?0zGfr;7cXH8gd7`7rHzeRY&jy@+(fOxYGmxaP* zzBj8jh%|t%k5SLBQyHY3@LJQBCD&|5enZ%qy3BoCc9Lu+s#gg9TS5&Fj6-EvGnoy+ zW^vLN9ZJ4vy2Y-AGssoa-CDeI?d31 zlrvP86W8tm5A?DPm?_7f$RZvHjD??F3Gqnr0C})?egMR;{_MXgr+Kj!MlF4ArHi$& zUqu#EP%a$3#4w|^xbsdLsqqs zR2bstLV#8*OTnb~>xw~__}A`ZA4)EGkXMLvdO0sB(;M&lk{h#82Vq+>recJq-i@9)vg@yW*mkS5q8KeGoWfxdONFM))(nQ-iOY+!!G*FS30xB?=t;CTar1QP8>DmdrWw>Cpad z0`5O_p8W{RtP~3|+4C2z0%eU?neP)hXnh8P;qY#dI|J`{gq;hU*b~Jqt0sti1E=qo zv2{j}6DBcbjg-RAa~1@=Gq5fn-D_C#%xKsEnql&TB-NP6Hw-uIWYV`RfA`xokL)Rl zc8GTdrc3IQffES%TpCY?!y+x`x8COY9vAnA5D65yDqqjQSLmBT?0ZT*2l=2dW&2I*wHDXV*GLm(x^@c zV$CtF)c-7DN=}?fi<4Vwn^A<4rvZuWLRqdjUQytF4a;Twz4S~ezlH?}qPfOypA$+R zD-)hD5fKbrmAi(nI#(Vt(Xe-02gNF!xz&ccwPf=+qvpMLYDi@m4qvH-L< zW9Eo4-U%{;t0jKwAUfM3QreUu$yOO?hkb_*x%0U4`m#mDxnswmUrw^re_?dCRXUJM zRZMSfNFao2-I-(DD8yRmI|(_-bsWOpMshV*&BtLESsPWVAzypPliK{%DfcdK#w9)A zAvqaGIAZFzVkNFzUNdJWArJaQp$dN}Msx7kN7x7D-tq{cNh4B|0f4#LA1_&AxipL$ zWnGi(-x?eowC9wNU|)uAQrva9;(;<9f#B<#!?%8jiyUoIzeZ}kM?wvN*R~KryUdc= z1EN>|5U8CgR6&mQd+bTtpy~fkKg7`}B~90VXVKq)K)6cIaw#QL45A-m_t)bn3h$56 zmVicN8oKUqnE^vuFvl6Lo)ktm;h8~rh^LTNyeqIG_t;@HlD>47fM?iZsF~P zcAN_u0~=0^&D!BvwC}HYS&a`Zt6>pJNF5HVX_*hKqZ>stKjX-8q?bjw=e0AgO8Z>rrn*-yjTm+zQh zM5Y8Bt26_K({5M2-a$|`gl4FbnChf_;(rFC?hYN}vM#3k#}JonsZbCTq+HxhBl*Ljt@JLN+Q3zN$c2w$4nf;fgqm zn;Umx<&
g9R$fUNb}&Pe?sr{LNEgBi63`m#+>;|9eRu3~bq16Nj8R7G~!B#zPy& zcBk|vj?k0;b&;S9!Mq(^HFOK#=?M>O)KFYcCPWBvXJp$laiRw8^HShwXQ%Jgg?Zvi z^#%*M*%1F5N<5M1i56db#jn`uaTkX`U-_xi1+AD~(n{Q?2ZvI^a%J&pa6DNiF(R+3 ze)_{>=B_;Hl%`z(LcYC4>&CT7?fv!2hRk^bm`J+e$39Hkzt4m2 z`c?&LeLXg&p6X%_^?!Bs!2X4ChI&`Ojr)s?Wb`;BFL8#G2+qx!fDTWIzt{r*QMekQG50d~qQn_cHQ206wt7FMn*TC9wk3xGEmYQy@)B1NV;)T^ia@w zD;Y5=?Jk2=4C?;z&CvLp!N2&24En5?2L^7tScb8ps_vh0=%=hui*WyM3!dMZ;c1O%U z@bPi``*pFp`)j{h-&f%EC`O+!y}KBH@S}gWL4{e$vxgT}^MpCT=kn8VXpv5pfUA@Gprh?Js42h+K_0h@YxOmWtYlMtUk_>9UI;nU#7<-OFBEm(Op$ zIloQigqW2`B^<8Ef5Py7dcNPjjms(8i`Az9enOpWPQH77-tAp{oE>d|Nr(2~CW(|I zDmB%0?1+M9sL^Bc`(m(aG5M0B*VTx;(ipQaqVIucgzcD;0#{fEDOF=!6}=@_tU{(K zBUHmwhfccEqQ{7$D6Dvc!ZUO-l$cXVnjPDd1ANY1>O$y^5{`Bt2b&?B7G6x;fJmcHR7bb z4O}pH^7{O0YvNGAK5W55M<<}9!eFP8?p1SJp}+xw!KX{>L;d9xS!3Ln(_U#TukJdhLL6|6!q7DYcLSp`0|o zhBS{LLuPw)fk!bpG9QVyU!=!_FeKxVbttYn_^WC}l#pZcDrb�aYHY#2jP+*;77~ zT^n@h!iTV(O+P!AO}M{tgTKn5t2LMZVT1JQ%Wdt`93$wy|Btx|={xsB0D36RnJF(}c`Ydq!`%{E({S-}pRRuJt zdh$9M*>mN*EA%Sn(;RsVPt<|QSmy$EYhWEob3wD*Q;@gjKB_+vjV4PJeP9N|5&NSF zu>efnB62@9a4PeQL>ItxspG+FIzDkFY{zGZ(`Ol!&b@0)Z$Iqw4|y0Dq(~fh+^}rP z9bn;_uoL`~8lmq%G5tF_IGc;&c(!IfN`9wbbFyj&{7~a_0jmE&>_EWM{X<6>r`qa5 zsee|gZe`Jj)@A~+@!XvlZ<&zDEnC8H&(durB+@qGMHx`PQkqlb1~c+$6@AScSTd*%4;TuUZMv(Fjnrr*gB^e(Sl_Sk8RtwXZG0I zW81cE+qP}nwr$(y**Ev;B8f6TeSei6wUo1VGZyq2C~Plaw-a?gx2p(^Jglx51ibs+Oh-uzn5f|-m_DPAKVj@6`$ z^W33~HrcAR4p6LCwK5j$FGe(Tm#Z{ZOnMr^9Mmja0zl($ymf5SDTN%C3 z94J_rIqCSARzc1CBU{!1Kz75w@&M}cXh%N!0uvakbwd~p=n9&6fCx7$`)NpI&|?hb z?$s@JwyZ9_Ej=B5%?$R=lwvBw1t`5nbs*{BAO4VBuL0(Ovy7n?1EbWsDL^So10w7N z9A+f|^^JsSn8LM$C4B)hrj&F;b2^O+{e-VT=Rc%&tz&9N!5c2tX=_nRE19J0lrVuB zpf&{)I}&{Xj(fy?i*voCIHmf!m~$+|z{OGoKo49};}c5>>j}otAk_l?>5`R{eNGd9 zGYCdtuZ1w=-S*1~V+}a~*FW(8I})$`1nhYu0RR|s007|rk2z81HjYmER#wIij{k!3 zGMA;}##qDdBXv`ZBEjY!iK(k>wJjS8{CWvcSTaFuAOzY102Q=^7}k5d)QoNKHZu|g zhSZDw>rycif|RVRt8A~Smy^?zl)vf|hi-x&nQ|_797nl_fS~nms2J~8Zg;JzSJ$8^ zWg~-%3iB-SW_$A~iHe2^x%^e%@O?3CWS9iH3BMDft*=5! zER;4vac+ExoB{o8QV>EhFOJq_h&rm7*DRrcAYD?(I?`dbi{fQER_Q3|(uP!HrrShw_VPRAO5JUwgcGF)FDH30yC*}m!1#Uko(;*DKppe zIT|O{P6B}#I|2L?jGETLT6%#PNsBo|+iv{mCxgpuQciaesuY&v5ZY;FE*PA&^+j-h zx`PN1k|!L0$s}dMv;Mw(f?t?_;y4HhZ#qRURiVPHRc6-1Lb?Lehc+{j%uBZ6v`#Tv zx`&Hii1}5V5i8;E4XM*_kqO2e#}jDZ+j^sEj4gS1gGPH7TCPX`DZa6W3*@K2p5(!h zPEdp#4lF-JAFyfukgtRxzOR}bmUdPyczCK3vJ9csRhAO4;4Em~wWo_95*;9jkj|c> z7mc4dNQOIaKP$+Z1S@b~D2~xNEu?o869Uwy%a7bX#PA!g*`KS?As6ifBJSfx zls_RQw(VS3{^T;X+{j7}oK-kOSxGBW#U~_ZrLY}!0|Qd4jzR>X>0sd1;xmI8alvwF z^#~J#M+bb%$SgmUaRLI37FIa(@yb`9mwZ!NI;$xv-w_St|8bgulgdU!f$PqL4;YKJv_=33pEbL4F`lnw9oR|+V$k`^0``D3C29QM0Ey{S9e ztegHk;-P7{qPc95v{Iwe;C$7@$7g;)Z428f4l;mm1ajgB9cajsO&dJXYzG#!jPHC_h`Kh$pR z=438bb&kxna=;OV4Aw*O07*)xYYmPM7YZ0R9Lm_>IHR^`<`9*D$yiSb$jMJMJdKJhc9siT@OAYAhhneDcdi$qDN-};Eez^os%dkX^t^A z4vu9P3>uaL+|_FMf){XmC2yQHGz3O|&P2PHYeANW%Wc5DDNHM#x>>|{-LKwQuk|Xf% z7=EAm+g29olY7%`spD9Uzf?*)Yz-i_v`q-F0on)j1Xu=`Cep-YBw>G^!6w*=00yJ?+ zX;PtxC{UHzk_NtAo(1302koJ(72=H?e%(_))0%C$n7WC(*9ESSrT@3Qzu1?R_gr+! zQgXfy1lAH3kP9b3EG3QC7Z)j{$Z34RR1hWx3E_QqTo?$h^r0W7HpwiA?G4Qw?;a!j zW{N9#1WsV1L$N&2k3ZxfNQD8&fk$Gd!vJs&LOYZ{WWf6bViRzBl@wCj`VN64OC0w@ z{j^*JASf#;ze?3j9<gG|JPwR^IdWEOO+ah`(K7UvSe>zQ-9T#Ei)Ad@WzAPgvibKOjW zD8)W00f3;+fl&y_JU)!R!RiJ~lbQ%RIflj`TYt*MI%T>8WTJ3WG82!sUdRxgeRQcf zhbIUJ2mn=1?aGB52!bf{QAcZ2f(^0gOd{73Pc1+GU=>OtmJ-mt#>O^(c{H4?Oiq?h zk?FJ)t}|}90CN)+^`0`;Ev9g81N0@>T#Gl8Rk;QTOT+2)7KoaeZ;H7P}Mh zitrg6ax-h)%#&;1ec#&;YBK(40;)gZ)^%*3T`RBI zC6C=|mfFwPrX2Ho;GenA1m$N}3S=3oGd^7o;w>{>ZTPfuSb%gF<1?W(*S`7XX+Tqa z^qpbwuIDEpNmNFL$BsX56!mlPV@wpGU}wMw&X0*0O%+fdk)@AiEiBZNW~0y`02mK=7UmKgL(mY{MmrE_X{uT z5FGtY#1LJGU^}=ZYBa%$H_k{SnbA?i5lNKDYZ~1ENIV zKwc>EO}}snAY6aeeze=}=|+N1@c_`*8RZZY$3}0U{^B7^{p|Y&36sZ8nx8eY_<>?t zy|E;fuXW=F(&s?0n@TrW;MH?oNtcye~nk@Q?<1)eITI zai+hHvNJAe>RAJ8-=7OGvY zE&K!dcq4)bz8nk7nj}z2m;!g z12qrvF<$D_r1@d9V&8SX7}R;xNGP_DCV^d92o+6~&9obsbJbslx_8K9Zj6*j=eG@D zuE7q>pTd7cMA$?KBti-oK&%|%41=2M1~)Cwyb|gR4O~&Z;u%bQmfpVcL3I{biJ@Wd zq{RynzycO7fqxe(-RZNa1j4+6LrV8y@D}JSE;C+$PwAWK;Pzf+S}D)fTns35~a{ ze2hETpIj6p0w~|wQbP9YY2P132~LcUE{umcy~zik*>;?g$4J1l9hz@Y@T1IXCqX|w z^> zixVhEucySwm~0@){1b-_LiE|Ezf;b;&h{b|Uciu#32?RKC3A3vom#1(;QO0C^>{IU z&)MPiDj_1@tW6Zt#gEf~%A2a2l(Yz2o<&94ti>ZA``WVKZfb@|J|%-ZOgQ9|$l;bH z@BKkCxbLkm9#`c5C=5Z#RaPa)(@Ma$a@TqJ)w{>G@A6~%U9OC$p2CyuPu{`RAZhpl zu)$HOO2;_-_VIBwY75g4?HkkE!}&Sd+B+bv+`G6s5nG(Y+dL?gno;U=;@RKOfn{M_ zCv(Lsmfxsn(xJ5xhx!PIF!|wKm8%W920M34s2=o`sCq0^IH;jRghd&K_0X373eZ^I zxF@j=iA=t;d@a6ub(|8h{^#fhS!dq*A7%9nnK}b=zjX;W=xZLSKrw8Aq)3=;@5w4LsPj2 zf%HHWC2>joPt)`AM}LK~tjgls{71k0F?t)V8Z8(r&0Pak;&0fjnpFk(l#yk;w)Dpi zCtO;YG#ODWZ*$MGmGx*8p%AN&$HqPsk2bs?zmb^9JQSfR4N7v8A%8PR-v%s(~ z^S`^Twyocc)~XQ}58qU=a|-{QxR|aA+zxZksy%T_|By$nle>nG*#*)2QPT~1ka4(O zMG8Pr!{md91%tbt&gX~~MT9f_x$i+hc&OTqpNYsyb~#|gl7yic4dm1^IO?a8_qO0D zxnD7qNxV8z@2m2*s#Da4ILAK4Z_0F99Uj_dLslwS9C(=Ta z%Inx7XzRD-D0m1Yg&y zcX)5OB)6t}Q6vL^|M4ohiaPOI0bep}M6X`l^Ph=z9_238{eguqvf15wo7U&Vbh>nR zZQo$qTHlX%TlJUTlnit$QG*{_(cIk_*RSVssnUbriwP5bXfQ@|3=iJ$y zDA`zelV6~l^PDmN9xfN+iPilCJ$;A;)ghl-)Gh*wUoPBxRjz|ReBczLMEVNyF5y2U zXrzIg2B5m@_!!BqLvlxsgGI%@R(3E|-od}GrRfpXC|SP^o_p;E9^bE$ovmNZejZQu zLa#1@oo?64%7lxhw;!6v3o94cII;|P<4DvmPq4pzl5nIt{VqgZvY4|~Ml`$pI6T!* zn&5wH?Xr31q8mi_0NpCI5FTBxy5oXqu2oz~+Cmc_&aN~6O{-3c;;lSfj7}NSz|C=S zbXWN3{cqi>QHf*|n-&0IZwmkb?|N0XRTgSUV9akHdwMMEj<(m= z*VF&4^>vxjNAbYfuM>w8Fo~_l?OGd@*Diwk(jAWb0l%-G(2vJ36gRO8^F@Y*yc+C3 za<7qR~ zms4!bYtGHzvFLo+IoYoi7`xb=0#`R?wX^W&x3@9yF)}l;5M;>g_T2azSN!bw?v3sk zq#8%O9xTxV>gBUopYK!MPQ3{`R)Vr4KL!UjK%A-aZE2&K8$>SwlhV^f)8wjnT^VhP zru*cX8_pQg2S`^WUsqr_FTtOv5x7CjU)G9T*}&Ne2v{F{H?S%@OEOuQEUr1ig4E=Z z>k#$tbr2)t0|Fj47Gm;2sJk8h*zSj7u~wV+-kU2f&Hg?({HLaT2{LDySiU`WrjEN3 zTMM73SLsc|ydS~L;l6o%F!($kEHBv{E5dD}GUSe-O4kf39>77Rq|Nxv99qJLs@S2( zt*OaGx@=efP-SEsvZ`hp`RLOZeJy~r(6`9CJar%{6vxFTR(C!(Qv) zv;Uo*%~`tCGq3OsH`FMo3D_Dky+gR*!t{ZxlnZRVu!>$<>hkp^HNE{|JF^R^_=QT8 zTAo)bRv#QU9T9z=C!E1e=K~t1Me<6T0uvO4vc__?zm5Xt#}uSZN1qlndkEAR)u;T? z7JxAQJ%5ppE0Z@pz=lNh&kb}VkhL}54C!?b4O-E4R_7r@ni<@S3bsFvmkY8PnvNM< zHo*Z13-W?6CwIt=Pqvs%lviwk4RoxDzt{RJ>oo;6GV~$El2&{Je~g<=v=rN(q=HmE zi5*pP=_V<1_U3I%=dapf2XUanky_vY$eK}CUZV2ZN=S1&kr%@0=Dx$b)8qPe^}KJF zX^HRG{OaSoh=01iM%RnU@3KStqk-CaViNPMtl%U?qvAJ&Dqz#eXa&wu({vi$2ss}P zmk$l;Gv3@J0H$no0!`-0hhFSLb>CBT;BaV#P1@76BssKEdCS9*1Yh%>@3&Lk=jhI) zzHjRQfu~%7>GMY*x#gfwM`I$K&W&}ItvBD&Uz#*c^5^P{k!0l3JKeY``CQ(6C&)&RlMTN!Z>%6)1CE)}Z$J{#ZpEb|ZC@p+&l;0Q9II!B3 zIc*D}aHZSX*uv!g~VjL49IP$#^gn*rtcYKIZ}!Na^^b z=cHa>q8@Z|ST&{RK^`vrm>*yjIiX5|3(%cEP68{1Di}(f#TlKxWGa#|nAxDJ>DDW2 z7E~$YaGCL;fwHwB1BElGdfmWv8<%8m`3dG-4Jng=Qa3WL9lPGlTCFy=(}Lj*5yz8~ zg>zV*91){4TeNxxEn>xDC24%M?v&4E?$k4F$Q?3`fV7f0>ek4 zP(iFMIyEt^&$v>~gFkwKmr^*?7^G>?AAY-OM{@9#fiJU zWhKp(RAn!`A&4dFqewvFdwU(fBU|}hMK%+q85Fo)poP8(C3q|174T;`QY8~hktm4OVZvbW00&T@Zgxm?WifuHFWqW+RCjlD-}hUHvcU*I=6?R_~@P)ehFSOm7Uu;9qUN zfC~l$3b}URBU<3%fDZ)PZr&CjiNI5=yqMtG3+U~`w|MXC;Xe(+JnUayX@Wi9fwe8S znAaNj#qxuD-lHxfs2x`SuDAGx;|xTMV%{CXw1qu+jzU{$f}w+twqQ=0eNZa-;wv28 zf2MOC%7V9e?ITaX>}DCe9kxC)zQ53Ug8L5E zQ=6VNLie3s&{B%^lD)sWhSm;;>OO?Rn+~`Rnd-u3efh{y`D%FZqUQwPnp)JIq60bn z-$MVQt}=gHjSTbuPERO?dj2wDn^?OtMDR@Y=K{3bhXKXY{dr&}KdTO1H@*1#i8|MEJrK5Qaufpm65g);Q)`)&eRFnnJuF+KaWrKjW3)raB-yB_n;`!J>h>2MJgS5M0|z08ml(q<2cp=IaY zQ`6m36kqWD+S>J|k(EGcyct)aTQfJ$V) zLTB58jj%*0>3!pg0W9@;N4z1lLilJ9LOtm`$26I8#yo3XkO^o)xum)ZG>i^rit%vp zH>(@VsxD>({utDf)gU6(oZq*D8aVGJmHKKVQTSgr5!T0-mV-GR)c%}{e>cA}XR9Uq zl`U}6UF8RKQyc;kkLcSGE?+OEl&#e-1e_)d^3^aK6xWM}<9o?WseplNTn2IJ8;KIl zV#B_AbzSG#%A=}kw6*2Q9W6HWrmKuDr%G5@w(ylU6=&Ig%CDf zwNyL+RGIAM`f+Zz85Cj($sVre*1s*|4l(-F=|M>U(vuIhd-aWUYy&H43zgG*T&M|( z6gpl_eE~}it3;vKY2`GSqrC#dzbq|19KxJ&zZAP9a`05yBXO*-iWyDx8aavvtoPz+ z1Q-X;cmE&_2>d~+`Ru+(;0bw&bUt7*0c&CVaSZe&KmYGEP3YDgKkdxC`t!2xMsbnC9la|X}cl*4}>7~7J;Sgrxj)O)5uQ}c0FUI)S8PaW{(i{ znI+P_EsflaE78L&u6z&K?N2hQhwx(%&&r9w3i4CwU!7J5O?l3x*k*whI8BmveWaRx#d-K zqkhR+qPv9aVXw~N)2NnBTbk!l$IxJsg{VB+5>@BR83N;9BknSG1a+6ZY0A&uLW9oetXS3jLS}G5r17~*ZuEvVx z?2ZYYUqhU5t&`bV+F2YfJsX=`$*7#qbMB(he04#!FKqWXMlGUtS{4+Z2vC67ZOar1 z^aRB&5H3Y+Wp30#y+_nUX$z{?tfDn=4FDe`L&~*Pe!#C4m1^d4W*P^&3yYUp(__qD zFpHpBi5}9WF$-E6lE5(-+fS-a9LpHGUsr4nFGrt8n4*UIo+t*)#E5tw&ZH@0PR<@X z)b;lm(!H9i@DV@cVK7o$8G#8|BL=;T{B*vWNxX_@7su-&cp;Q+~Fnjf3PYVGKWOdC2axF#3wzj&%U3 zIN;u*oRX-t#;fT}xwBEq>{j1p&ES(<$PCq+kPciDK5O*~(uBD!h5VJ3imCgE%Q$`p zkh{Sf$@pu%FMmo$cRQ|&XNcR*fFCa(*}(IU@7h%sKsp!_vF-wZ`F=UnXar@4R77WB zutcq}k@6=q&)RysZk008;db%fh1{}3T>Z%IzoPnVEpc*TJMyux%~*h)LRtIgyATlX zzEBDLoi$*SH9U>X6kG=}8ScQv@v&-#3YDzD5BXPz7@uLViHL`x5-w%gSgP~Sp6X$h zga%&=B~n(?vp|P`8b7KkZRLsjtN>umGOR&Z59yxw<+h_EcGw`-uEaxBK5;7}BP}!5 zLn&{;DbI7lX5ZP~AFFn5m#Vth1Ptbr7+_( znwVg=y5d(OemV$|f!?#8f-pSpuj6u5nn1i;L0VLk*N-=exX|oW*Lbs; zB+sjK46A}Ld6IvQ{d$3IA=8j)y`F0KHmN%?W0wAn-Xbhx$+IlF=| z1Pdvzaj!q|`1iM1$cQ&L0Kxnk#GM>^c9WUVWzBM_?tuFsCCj1{54R>U0p!C48Mrkn;dmf@x5fVMors4ZVD&&^(cO_>o_0s=EqNG;Z>uqa zYo@*y13n>x(bL&!H046IHU_~F_vFS=RjE;Q(tkYyf!$<>DU zVB5s#gnU{LS^_48=AYO$pcFN+R9xY95EGA62NL0BD#HY0amxJfgpU4BVTU)zUFyPC zzIbrI68q@k1&l+PHD-4MRS1=1tRNjy3$=>6w~?ruks8%vNaFPJ{Ndr^T;t=TI;RA4 z1onGr`TY?$gWkrwI#tM-IK6&U;=ySLye>aHl*9T^O)o7 zT_>DOGhLVl*C3|Tkn2t~ssQ+qfi~Xw2W_(MwMj0E0}+mWd?^jDGLj4mu~jx&4HV;l5>bgvg*gmv?N4Fvhg(LUimiwXJ0|qkeLVTv|G_38 zlN~Mo-~$q^Nc8)V8gH)4L=b%MRK(!+BG?z~h-EOkGSo~C!Gqz&aMi*=6D69s0TuFE z-hK~w&pKB{0VkeqKgy!AkYtuwM!y9UYZ0LEHOXi>q@BPt4%Du|VN!W!7#_SyY$dB; z0mjcF4H9K4d<{YB$}pk&v;zp&94mbk@Cgc5nQ)CZdHY1SDC(1qHMVn=TL(go{<528I5oJpS{^NmX3y%p z`2R6%WjahB#^qF2O|G#=D^vEDj? zLxox_$vEx%>l*-nb9O0*Q`3daKK`d%)U=#C=dC=dmZ0}n7yc?cq$uyHXS~jCfpWy? zF8oM%El&O?rcjvqhl&a|Iawo(h2yh~mez8P7g2ZxTo37Q+C|KG2g#Fa2Hxfqp<<

Ik})W5+oK3}2~ZT81646Wo%F zVQNWAljyM}a0O6NL+~vn&O^^{`$eQ&xu1CCeL2pHTUrg@=c2M}I!=mp99JuI?_YY7%S+K}xdsWNpy`+DN@|q%p%#{C+QNM-{-+7~#UEWmI7Jcnn z_@;tVNxZV@OST=ok`9rZ;lU zBQ_Uc?m;uh-HBsso#T69N}YA(RmG$83ef9)L&~G}-+i4?H2`EC@UG!_&vQ^GU)|j~$T9kS_WaH5 zf6DB~4n{2jH~jveEEREu8-IvdF#Te7e1l$wynusSnIXCQNa|Ox6{;9n%VdxKz!_2Y zsH-;8ss6Ae(rH@*aV(VVdYUPXqRTMBy!=N{U4C$PYZda*V1#&4KF|y!^hPA+;Q~IF z_Zhul;CKJn0r=qSB|!H9ZRO(~N?QFCY%!07To2)_`?Y~&Ko<5&voQ^vD=Wjwz<}-u z9k!oVj_s~4%CZfo8fj1t9 z*z%~Gr&TM(EZ&7raI4NyqLZ^*Kh{f8uzV~N(%@9~nX(jRM8W4I5X#*ojkO6j;rgmh zX4b?YwAK=nU{ZZ@exG6MvUPe^0kY}=k?>{KpDs%Xu8+);J`5Uuu`3VjqCyB(a#J5{ zYj2L(;h+8X7okW!JdHZZMi6WTx#&eHi0^RMmX?~@T2D1t+tTG@;p&jC*e zx;osJwN^u>#)M3hXkHeOUl?BX6x2Q%6Kj|%%gfaK_Mhx>^vdbBLxM{4>commIrtu) zoqKxG<3@bFg6NwEru6+3`y8P8(}9WNJ&4pHbg!>(v)l{pLLcME z4ww3pbm~_wij}tWm9U(gK3FC6x-W~}-9I|JTi2#* zI-@Kd;d0>YP^ra?9jH4-N4>}Y_FlZ``0D~tM(%M&>sj-(&}SKr38R4t)Zes$duHt^ zx%7Qpx$_}I2CUEnkkYX$J6 zk>k*G>HAoSYNh}6(8bSk8!dZxj~uWHJ@p@Zb9EX@dWk8sr@c6HZzGHyqPgNtvP5ZU zJ9Px=cNcVzhJ6t27)rT!kogGSDHOHKOETd&bAL5M`qvBX!NtAuRc`eU10Y$l2Io3* zrIR?iEHnumwmKb|qPK)*+c)hvHmOKN{)M_>ad=G9)|xU+o*TPw9`+86Q_D(Q3ynrO zB_s$XA6e@JVjWeaCet6{8FiFq0HVPu8u@~S-l@LuRt`^z*?2!0?{}m2uI^%QqRJ1X zmO+zr0ne34f!x8Tph!?2!hZi>(bbpUeoj2Dx3MJatDhD-U^F5%Am$M>KMXQ4hpX<= z@6Y1r*;o9RpXk;}-x%G!$JKS59P)f@A?9wZj`;kGUX~Vq?|=)z921MZl_na;@y-CT zO)!aTg?vs$~Ct-HOX{p@oj8HZ;|2PlBzp6mO(SDwkd3Ix91G`lW8;!rK~j;h&DrHX5x1utX8l_73i~ z&qnT_4r-+mX~t3?iaEQ?9?E2{&sIAwJzv)Q0KxjWQhy6KYO}YX6cN++^w2tPQC?SP z*Q36cVZvbqWlA23D5fdQ5RjLll;WY(j8<9rS2!=eb2cEN%XtC&bxsEDHAM;sH|C~g zgs>f2x0PbCH_J{$GA|-eG|YMH`H+N)bs^$aAW=UhQ?1!E#lh6u7L&;RBJA{9&cD4Wap%PI2+`6``vN>+i~g^n!4Ztv$}r)-qP<+_4c`=&5@xh;O(?r2LI)32wfk~@y^ zT_A$w5T?Dfy6N4+jCHfR5}9EHg%-K@Mf=_>yZ0y7dw7dRMT?pU=6h|u_gGfBzUny7 z`?g>G2z-Ijcmhkf0d|lPhC|+k%=Has_cLUvVm>miZyo~+atd&kkG&>-*GOdvL_}p_ z)ZS7Y=uF#FjV-GnxSaQkeEDm4W_v4>;>QEWk+hom&hm?V6rWXuw#t=3o}-nuD99~P z*wT1xw(N2(+x!I+pe_IYWs=vHmqeW~ZU0j2bRwT_TlCZLx!r8dd(jql0}bX_?)22_ zzEdxsm}zjUqyb5fa%tpOiKx!oIx=k3gmw5#2_!C&H5j)oz8|>2xEA~|)1OJy-d8Q8 z#Qdyb0Y&;f>|5ZXkIvTTW!JwFE12$@tQ5v>&k$m|&85|QXK);mU!!I=Im8)sd_L9f zMD(j70KO#y zut@W=d|C5nPUFY!)+|%R2m#ejP{HiQXFyAQLUP4u@M1shRmw7eb^v-2qfKm3z}0YV zwA2UcH^O$J+XS}zUP#P+`)`qLA^`$m{i40J`2q}mc$PQimXs#t_PKZs?qn!^`WG3o zz1Zrr%(07Wz6MP-S?7UN8=>gDlD1HU%M8>uk$`Nmjz~kvzsCxd$l~R^=8Tpk>17rY zOH=92Hn`3u!(n6z%kUIVpr=NYi8U0w%}tk{U`)lOlp{)%~LF zjLUwxDs2c9t2j^EUh$$UXuMweecs8T!<6DZg|fHK$jHQmYg1WOo%( z$l7A|3{`UN6VDSk(F8Q)W0gHU0G2v@PodcOO*sj^&XK z8JweNhy(>DkV^D(w{8XDThCs+f#!?>SCO##KzN=fJ18557%-O+g0|Juvsr6Qk~&c1|iHZ&B~H3BXYTu}P{!gh=ri?AfS+i?a^U(8s&A z{OtRVL}-(?Ags*Pj(5uk3}>cH!m9FMAXee(EBmBaiv))`&4;gffz-0p-=;w&6^%goHRFjarACeKQRDtB!?$EhU+VcOsa`ExkTC)Pb-$h- zDJi;mxkmK9eZ8Z%#bxo3v>H*!ob?{Pf!l)4q&b)SyYyL;ORl{hd0D(1h!h!A1YPob zrgJ)lNqF-De2bb!@ej-mUdd(L z)s8^64xC71STka0!=#Pe0R3wtv`=~Z29*M!?eMV(Ik0DfMrRPMIYAoA%{s41m7*;5 zjTP9^geR|74Jxf!O1!LXeoeZYWI?;rUUBna_kJ}q$*xpM-Hf&CDjvGPA-|!-NxqYq zW2@M@nrUG{g{8C*2D?MG9h_dV)pkWjwq%;gwJaf6vZBfQ<^Rbs4)}@G4+?IEintDD z?EBguz^#-{O#*A1wjTL00&h3o{2;8@ryGr@hWflY>B^Fs;nvaoy~L8i<%Ppom*3mj zZU|@D)hP`ba)I0VkvIDgvJ#2_1jkB+5nCaD8vcYQ9-I_m^DUV@Q4d=v%v}t)B)Bz^ zN3syzw967~m!M33MdVCE3)!2h^csP7s!d--dMVt@7+xcz%mDE zeHDttO?i_1ycHZ^#SZlNU@cN>kxnJ7Wa(nR5Q@s_F~haf zN1e^to_zESSk9oH82c1#VPL8ILMr`q#p*r{FrmMm&0NN@28J#_)%jdajUg?r<$1Bo zbJeyC=72hwAP}Hv`V!6R!s8Cwr`QI^B&VQ!YF;*>ws4}#>|H*Fl6}aoG<8c=wSGtN z>$6SupcFJjrh65@p8|#3L`sKLhpppQ{U#U-)+zy;?dU8CfQ}D+DH(4lmoZo9pXh&ps5-lJ=ffQ>b?0ix{&P zS9doq+$#TiEjCO<(2pbg{nxCvhiqI4dTMm3)E1!|>=14F^4GfbMrl|HKA;2Q+Kw^& z7424_g@clDZSG>%*zV5T`$SRE2^XCsm50a{vIebR)*F?Y68ic3P1Rgrv&Jb!%$ZT- zR9pg78e=M|hWH*sE2Hb3d){Nr^CixEHH^|PHkOgBg!R)Lk<2~ibk46s3~Xav()PD52bEVN>xoED6PT> z5l1Sbj2P2FD7J^6Cq*|Y*c_&rh6fa4NimQaXrOzP)Q9qbQuQe(Ojh?XmdK=~3v72Xy6eiFk8ZN)7KF@Gh9gccbTd6JkBhDab zn?1x>!g%_9B7a8LLnklhk;@M=nhb1l8J|Q;v3#hH@&yLF`WKX*H+FN3;`Qr}~_o%?sR*;q@j)Vsv+DAWj?SKfENh zKKM16^dH;^`(XeI7=aWhTn$|@=<`6rfWNH65BA>g>~L|?Re3?Ok}BA81E$B+uzK(f zMy9Yl{c0*I4O``G1Q<{X7}07F-^zYwLxLp!71YQnKGo?+Y88UPAr)Qc4%-_lkv#OO z)z-~RhAOIQR&+S_CH_0)n0efh=?Tfp!K7X^)Kx|Y83Rm+>r{GlvN6KstnRuHVQ(@7 zMP^OIg7dZ@^JvhhtOqGe1jwyfSoVJ{ApZYN^~8pp5Qw~h;d~61`h)?h1lqN@f2V)s zQe~wnMxtGt+oT;I@9)QEWbFu7+&iP4+?1yYfF3z&-V?TJ_U2LwX0~1ueXdW)M_tc; zka%Orsjw@3QLDl^zWoRPP6BHcn;HsY;I*qfh}=1B44mk%ifZ>oL1~rEv4`m7m5DZs zTBSbOqnEL491=&Ix#3K-u_k>h!yW2I@y^?^gRF&uf9!;EE?Pvq)!zb{= ztlK3(^*2DKf!Np7QLxPG7Hyofg%;!f19~nt&^oscR}J20sUhgyp4d)?SUw9<=7tw>>)E19FWv#Cd{l-wxfG z0N(C#^X&~VD#*f<2Y@xya*(_|B?y5fwk7_2oc)6QFIK&@^*PE53IyZ?2?T`we__=Q zwnp|A{}I(GYX9HMQftu9y-B9ka0IRy=sDT7B0raJ)wqHSAqRld=GoMyRwl3D(69Tv zE~47BqBmj;(SLi`;zu0s#D#-EV6NG)gQ04U+Mw-h)%2wbaY?W%h&14 zAE(X9r&yrIM67tCxcQuQxqdwPUUEDm)R191O%Z9%LDq^cRT>I6?y}a6$pSkrt05$HqJ45a5`AF|Zl0l?(rf1eUujIuHRC+a zWxy~FF1=-AM7+QZM(7j4E0B4US^+~^yvohtE08l4F?T!<6DL7VwKBULcN*hj=6X2@u@J z#3B_hc9$akg6WmW%-=62jQ0ZwmT@-#&UTxq*Nk3O7-ah7(oZMAYTui2g-}y<5(+Ai z((91hKt7p$Rb6EAY8BdFwRQ>=#4Fe?q4DabxNEfzwI*?@Wpzjj*LR8!Cr{KzXRj|1 zrPv>It-7oP7&SxX?9pcS?cmK|xKLG=Hns#fI?p%WW$WuS4t-#wl_orQN1 zUr9PQY`HHhi?oD_sHej$^>H!^Zpj_nrGn;-GovMU=|TK22qugG!{EHz9!a@ccLD7T z%@w{2fphrdU2TIyBTRyS-S!8=b)7=;(iQqh_I$cq{SaWlLB zJ8{NHnG>E6KtRv`NjT8{kBPJXZ`exyc&i?AApHvFiD1^cg7gc0v~kdv@QEO1f`-jq zHaSV5>qvvO94DDdd2RW=)ROTc>I7z0P@>ipZ|AxFsameouF*j98Mnnt(1hV6QymEP zE|V0*?2RP2p(0>2#v6wgKBd~XopM5(_lOq3i&N#Js5P>Cr^|5S$sNPsnUP-F+Wz+I zY~TDsfVaRoq|qUe4l05`3CSN6wBOO!ch>_&RmmF)YMN7i9h!2ViQUB9ZqXnmLRnO5 zOQA_u#dg9#hZA9w!Ad*Jj2>63+7*h6hAa%Uy!vY;gp)s!-pGaw?m2!Vojpm|C>6so zcWYXDgnI@eN#va%AsXtLffP#)IsWBpwxZyJpBrx<+gS`7Jvo(JoOMmEm4DZgrzpcOgTR)w*{PyN}8>cWYMK_hR z`_a;6p;;usT>MKvYaCT^^IDSJl|0Mh9=sxqU&P?-IV=Ov*z@U?sLIYS4Ar|JsP6bK zwOej<71_Y1Qwtu%mXFZ~<-z_f7tKPjaSy6|~E7`tYLiv4hqipWv>2GcbkeH~Agf*3q> zkndM8x;ch0X)(MHFvx>~W}4;vW0_RkJXi zPXsQJe~IViLPfb>{U8z^H&{SpL8m~M3PfYyuVCFz?mk*yCsGe>9(TFbucbSSa9QL= zIZm4po0UGdm!}OPxt3K_UPvK!{p~s z{EY{gWR9kAeEDE`LDr4Tp|Bg^%w$TeQ1Ns`{lyv<3nl_6)R4k*A>;-W-9<9*?v)6Q zsRrLP*BlKahY%yJSQciAN+IFVV6=rNh+fAcQ+;+5!xXDTel<-{P8jo_%1zd0?@@RnSP;j&nQgf=*39{x& z>U3!(kcUcQQz5QdDjW;2Mp;=`D%aS)*V51fxzh)$9o8XM9H~@7!SNk^HwQRZdz`VO z5L^~{6me~!1|)@VG8>;Z`ex<|C}`#)uiuG*id34i2CsJ(nq_-woPz_s_FWepUkl~L zwBuM67flx0(aT@`Ne?&|{Jt_Q1}IK(OP>-oJ{Llev4K*DL>;lQwC&*Y$kCAu;De8hB*kbQ3j3E2Diu7}ayv7E$zM#B*6CGX?el zeXm8ptMPl57Ar1n(m9}MadVLBKRyF85oz`Qqj`<(vt`33h;74W;2GKd;g?Afn_0&+<;q;%;TF ztNF51ZkDd!UySJ++y~WT!T)`#s?b?ySlEDoU|;`lSCqApn~{r&vy~&@zc=gJ_wv8B zc*EH*A2?|lx|NReftzcyf-N^@nS6`M>CpHEk4`eWOpqxBT18?qBm900vp`6mR#O% z`B$gtE+TarGlIw=`4L}1cGzKDbiAj+ltkF+TT(l#k+ROw8a6~@)m%>YFz?9X(aKtn z|L@|k0_0h1A%h%{tYiw-NXXzaM%?(gHbX=$B+`z_)}(iRgwZcmCH!X6JV zM}V`?ShVI<8$t4VfY**wStC(2a0##bh!GRL`bK;n{(@-{IkV*< z{F_^K@yD9QAHsM55#HaPoq{C$n>A8rVLw%v!AeKU6sN~X?LTnvLj=|32c#$5Gl}wH zFLsQhVEW%lg!jAW(DS>e-Ug7xxF{e-W@(Qz{p;BE^fSkAK{-4-o6Y2zZ#~v~vRS3Y zeO`bnF8l%Dd+Q5h2J)$x{9(h2)57kfrTj=2z~d@-{ZDAKfXR63bj6JSugGvHSL6_K z{B$RWmr?P9F&7`N`#P`7_v?gSZO1kCF`$?$WGC2x?kB7?%r*} zYEyKKkCSu-zy08D*Jaa4oX+fRtKWt1$Iphfw$^+dEP0*&U1_Q2wZUVG;uEgQ2=Y%4 z!NtA(h6~ob=m$Q1HHe`B$x&EfbAJa~74&V*^GOb`dl&%#v}kF5%63hQH$+X;x&2i9 z2Mzquam+A1PLo$&l2ag>TBt08+ciPU~!u%caW= zNukC35%?d47ix0=a!VnWA3oNL&N}|__K%UbtdKk3YJ{d(Fr`VK)E*)-fK`sgjX!KG zLhRD+4JhxP2c2{AXqsu7)wPT+44LeKt5GXxO`|^kD{WHaC#%%VWKXt4?ZNCOh3Xft z6e3{VU4k?nD!@lx_^AF=4yi1@IS;?r*T=)svCIGY=I%Ob5K|J81l>tg?uB}|kQ?oa z34}+#SgyL=1OOYJX|x#2k?AR6VVA1P*Rg5Mt_o4*x4U6o=YC^Gy+PS`!v~*F?~Sjr zajDbZRA)Hjm@uwmb)lbHk1c%5AH0J!q8zTLQ4yG}KNSgM6VGJVIo(;dO^;h1Z7MgKWgJ1qtI^VAGm@QkZx>;GlKIp zWW;p8y<0FDM0Y?M_lA@LoFnpm9q!*0W*y+S%_^x7fY=;mv%1&xFD5j{<>-9O*LeBk|~a=y&%D zA?*_L>1ZOrf}rqY4lHlnzq^FO@C!ZG^XOeE7xTF1lwK>%jIE#td8NaiyOco;t}1qj|(>ayZ3p5f;pXEX7qAom@&J009AN z?`j4{MU#^MK1{*11(p9!st7JbrGw1@R_ zbV`W3e3)nHyvkp-vr{5Hh{AKt^A1E!>qLBcJH869zLwt`kt47PK9o2J#+|p46v1 z<>_g@5dHP}{FB(zteK4Sk%Edi8%>oPH#+An5Y)gDqKx|Yk+@G8$I$Sk$2obb9W|n% zWC~FZd*qZG_uLx98u0*OYskb0?M`?sJC4uvp6`iUC|1vZ#sjbqqTmDd0Fg$Q(P=Wn z=b^E|B>)=RcX$W0U{YHTs}k!zT&H^y)8?|@#+>hQjL1#HjTeqdY=PzM7V#wtPW1#u z<-wHK2NeKdew0Lp3zC4vC|8>8kSGIunOt%8ONj1yH#uP}`v+{x{CBoa86 zCye>mUU@wuSs7X64>s=_^;i4pJ?Uw9?_EKxz0dHe002)dE^1GSGl;p^MC9a<2~hin zD&23P$=W~n{#e{>Se@-8^9mLQ`vhMy5VYu2@Zc1$h`?Z>D@FUo9*u$dFZA*l{fQCn zdbiuN07P1_|40zlrx@WeY3W?=MH#qnrXXkgjC&!dA3o9iBm|ihfzH0(SV{VnkfC!f z9|`Vs4Kcs-gA~gEl@W_;M+Z;%AT>`{9x@OTEsmf4vyQ1+F@N&#IRz9gz`Lu*$OIAk z@!{w#Trf6}U{Goro+(iIcZ~3>uXnS>v{Om->#p0Q(95Sa$6Kn0s&7Y_^5?$dRL;Vd zEO8GP-^YisKU_a}B; zD<_xfZ+-#+&#$J0Ga(ps*!&K z%7awV9Eu#1a{nW?dhR@v@+4#+FaV(wGE)+iVa1Ddkt!FtQ_Yzp^Eg|6V%@fwyGpI@ z75k2|kIU%f9g3Y~rt*x=ChVvpj5mnFZInT^bgwzYga6Z54}}JjqRwbcgpDVRpaO;+ zCYg=BonoE;Q#CbqWQgu=p`%5}Cs6x*Z&wGx=k=-EZuin|sZ`x24^4++AoKp^ZJime zveED+W(+8kNs>al@;a8W8&$#CPhV}JuhWe|c5h48vs^EWM|l>!!FZBjTjXn6Yk9uvFGLy2 zB?!@ah*^=5PngQbRJ3V385NB4$8ouCz*GnJj~k&9_w?bvfUjhR1 zyI_>X3`Sq6D)C|uyK3{Wl_W=>6CL;?Dzlwm5xES$&2mxHsW)bkvA>nJ3`okto|}10 zch^a&&5~5%iys?4;D*{@h!UH&cQ}!eH8m|y3e^`{{V^tw!C#e~03N@MN-4^E>DA2X zk?~IGvCOQbI!;tnVj3~f_8Q3IrUE4cS;RrxfqI6%O-!MJx0bi{kvoh9M~f3bS_-OC zRfDsZ;R21H2l>Md)gtPk=IG=hPF`pn!s+}D9IcR(&++O#0t%ax4!R)UggU0M;+ zC!vP0M3K=a=Tp7IO>8hGHSUfEwGeDd!Zu5^1{)#?FPSPZ*Q(vIgV*M< zO;VWqgmWaB(PZ@d+cD3t3s$Wg zmY;Pld+Hl37Lfb>p>YiV`Zc}!6Y+z*g_)9e{>j+UZkfy@^_zS-jq0uuYXB0S*ToNn zt2?lgkcAJmsk)8EGg3=c1=K|k2Kf|CfjF!|Y|Wq{{o%>9Up=Xg5Qy7J`BnjAUC z+ncMG5cxY+_6PCH^2w)`?@e*G=8`{#)dBszv$exNEc!{ud^sZ1I^EB>K-9Ki^J@Z* zvlc4hZi3+$x$l3O!oOlYX@tMl=?NQTh|aUsaM&G-^0vj5+gi(HABU?cMK`=`$MmH9 z*Oh;!uc*@7D=U5*>c1-N=&IPae>py7aij}&9)iYG|Fs*k>6)1LE23kh zY*EI@A2+0Y*4CGO$v2l3^NtKdI9usgG}SGAu0S2U95X2DHSVy_%Ovh{fAN?$GGSSF zgB|H})}z;UwaWWk6^YjEjP|)e7JzDpT9PKyn}D)GcCC2w&g+`x<(GE!iJ1OZ5{SFu zLl%u*BiA;V(1fD`AaUk*VifKOMdTq?x__sCOG4ZBhW?kEtT42X8{C2HWU(vn8y5(%W}O>b32f9i)-B)LzZJ_#^z!1`E5JBjtI@R#v$+arGWf+`d<$U3fX9Oho1+{VFMeuvdGp z>eYM5JuI3ka8a#{4zpFBWr=J;OX-8(>k(4^SEC|86|*6`AuEeha$ zURp^`Q0z@IBP5Hc<4}8$XzZrZ#R2qUH^rpr<=(A>3CjbKU|N>Gw9l8n)-dO z`?dd__5T*U@3MD|4)*U65a8kRs?nM@|BP zz-2$i#OIsLdKS*kj0;NLVg-JD4Z+(HhPxME!n~chTf=I5DG49{h#FTGR3*oRt5}?`|(YTt`IjG=f)_c~);N?sj58$!OPO zpaJWLbhlaSeKC}og6&Gl!VpCQZ=~~V|L5FlyhikHM7@{nj6B7D7!D#gCT88u#=b5m zO`dKmsAsM8DJ7~}L)WV$dGWxBnOaNhq8#yzFLXt|IDXS&mq-Z3#B6BRt+=`Blk5LO zkbAAz&UyL5{+|b?3HQ}e(z=MPN{S9qDsiWq>n1HIyTp=%O|D7lN->pMtohO! zSvpV>_*CveRsD&oDmA3R%9{PHn!lHx-lkc5oLR|6a||N3ha%?2N|j><1i__oBq1N;Pf?2#O`D?G;o7vyrH_70 zxof6|Ub9T4J4S|M;3{b8B<(JiBMO*ylQn$Mn;gYk1dABjNdR2nlXiz}v?`Zii+_II z6f{3(NP{jiTnO<(o?f2e_ruFFF$;vmCSX^!(BPnA&GA)#J{Q>IMqsXUTQJfDYskH=Mc{uG|FE~Vv4qD&c6sP0nCQ?;K9k* zSX<3A)C8>shz*Mi?j}D7vg(ZEZJK&lnTtg_;&=6@5nn zK=2W2_&0Cv92}xU1+FGk#N7$DhusFu51l_Laqsu}-~wIYaJbX4!wX0acEye(6^kO4TNyb* z+zklX6np#oJ6d zF6#NGkSH%*J>|E_t zdT52_TFv&lvSRSo2`0rI$xc=O;D&6V#a1}00t)h?YJ0QpRF}NDGYB0$Q@J}zL`~k! z&IVv&Q&Sl552QH@G1O^uC?0m*c)KM_1o>Rqq)H_xR$`RHp%UwbZ z2Lllo!K1IJI~ezzyFh;z)W6NL|0{P2?K%sJ2kVwCya0)EGUZD+)_c=OF5BP;cQXL(IpUwix)nwAMIV-F z2Ri^iI6iQ`;>|^MB_f7hkPVJ(Yh@n7+*Z zK?Lpfi$B9!Qi*Pw9-?N2q4$k+qH5tW;=xk$O6~jN0YX?7>rv69n0Z0(CX|t+iPe=+ z+{*QfCnp?qE=cT7=IU0g2ki@YA3gHfv*h>A2>1!;dJ09mgOxL!!!(F~hb2Tek-xG~ z(?0L}%okn#l@k0e6VdRc8_N5t%1el(Y4b_sJnLgYY0sTs*E#rU$CY65 zf1yLY1Sa4Kg?mef^c!M$%)770*t(3*IbAq30bxk*>opt^mdTtpiO&l#uM1{cd6(%< zg?nr^EYR%>H|4uf(jA_T=58Ybai8Ccsy|1wbYSB9gmxL&*ZOth8sD7_WnFg%DKx4*Zh=n{iX(LwqET$}%MGZSK4L}6;5sBp? zXJB$=a6%xX>_YncG}TY6k=|pqp<@_PEu?XsSmrUB;yB zYpU{6vTa5NTjjkKuc#NGFzIaXdw^-g7@z4n^@o(8oU#sxkbxW8E|PyCRug?`8l1&3 zdrf*>7waMdUT@AoPV4#m&8}#>-4vX;SfJ+kxy`AByh`S zYfJ3M*y^t5G(_VTD<-(GCtXHLzqT5?Tu;M?vdqE6-kHlRu$eqebEs|qPBB?q^t4dk z%+t&7;~$mW}>GMx!2S*wtoN;sfj$F+j4G zD1x5P$~o(1HahV^z)|DqaXIUIt~n^d0_wGM%uBrSJw_h4weKafNadTa2|BUsX3nal z1}k8V{UC+9N@&zZI6xzbr&AzNTMAoMPc6eZkA}o~41B7rz)rj?N7Hrc?m&zFE?X)+q_n7{V0<`7{@`RkxwVBSe1ng7xq?7WL#C? zk~;R*;}7yBDf_g_EY$3|Q1<6L!8~5u&wsWK+Xa2&F%>aFXFJ0d!pC)ymDsU6GQ1^} zG}ZaAvoWjlKpz9}kg3&{b{sg*1{sfB9+fl%*SgnI1=Lpk6W>bsCN)DUopn*-W2h0> zl4zPPHv2Qcuq#~-pIEVjn}0Yqr%sB4fpQ9#G(qT z-gx$%SX9>d_AO_IZk2B_aj&(z`Fh?8&M4Uett`429t!e&&k(dN!cMi?-CwLwA?6KQ zW!gXd)UwqSNmRK-` zy8o1?Z-Ro<6roeg!|9vrF|Ofn`m`k*RSKjRwKE=(wkNOjAydwX+U`+DdW0Ffj;el% zJ~mO|H~^;qP`xW_JP;&zL6TRU8f>YI+x#aZfUqKH{Q-Zj8#$0&D2yJ-WoI~Du-Rlm5o^ZL{nwB(l6DtbLW(ba-~_0)vVo1GcpQcW`w)k#-)e@n|=_ijZ{Isi8AVNpL);%7azdZTHliNfC7pa?XuP z%?L%p=OXpJn=1hUl(stnNK~8#4JJ%nr*7kKb#g=k3TS@!!=yIJv^U&lOl>#Q^pJCa zin64m9;hL%M2tCUU#X4DRxMPaS8d4DL=AIGK!ky~^7w`pdwcZ-7 zqrYwn(q@aEMS2zKl@z@AmVhMHBF9qX4{4aAm5i=*Uz?n_2yegqT>GTkMJ9enUR!_t zC2tWB)7(&|I`K6Aw&>w(9a2($-H%6+{=A(tZ*}|nhzV}u?e5e1xZKUIem&pUeVzA5 ziT-@uy}iNvyL{c8Wo&@d5v^8@{vJ6H46Y{qKAzAE?)?S2zB}vp_4D|8d;$IuX!Gwg z=neR7-8b;hLJW>42>nM5B4#TV7%xLU=*~KHLr5p3XXCoyMFNdLj3x;I*!a1s~_t_n?wT(Q%B;Wp&*vDa|}OI;@n{x$g`hw>{(GAjTObN zT#MP|rp=k$y&y}62FKDAjzmM;KZqKmbA<(Bu`6B)PY|6##cWezBOZI0 zz2*p#{v+4;t+z>COf-4o=c6WZBu*ofSP=@j(CiUH^vuocAm%f%7A4Ai?V4@LxmbT< zp{VL+I(;v3OM$f-Bf3hpstjLAdZ&&f*A(Bu2rA-S`VjmYRDvP+yq60l8AZ?=DA_$Z z#iEHimmGI^gf5%}Q*8_mb3jqpoDyBQGB60kfoosvnlLMLg=Cf*EG;eQdD{+pfQ+M< z=5~2fYpk$R^n8+pKjKbZ#T(q9`lCvaRCg>(d@une!62sE?XmCEe6=M9*mITR$~+F! zwsJFT_p}>u*=0{I)D*-4bN@TU0!$Dq)bj(eL6%*RYd|)M+}idP|7ekV|74;BXgox3 z5<|D4*+3<3F+Ii@0okRg)uHny+nl&>zl^)q4liBzY=Q+a{vA~Jo{%~5*X>&Xe*|gw zo5mM~KPRKns%F@T(U;BtYJ;}GM*46~!nx2OaFujlo#ehq_Ax;+Y2UGy#NC#R(ch7$ zaE+yEpA=%usV}(n$asZ0YjPpP50AD2`r^L*4Pu}i?k6)t0|<1fdPx!lSIkbU^3<80 zC*eMhr*TOV<>|WT25`}f6YOGzpe}z?j6yd*y~!1)#ppa} zA!pFj?df~mv~3SC`i4a;(ACko%&jfpgx{hce_qKaAjx`Gk1b}aiValDI6K9Y&~nO- zxbY#FR)cv{FM2JBtN3c=Epwh08ms{Zt-%)kTLyKV)Q)-(D?gaQ)n*AUmw&eD+s{|W zjxa!tn-UNKRYQ1pL=KmA0hd2;kDj|vQgrtZBbO;X3`v}lWB3>C&7;C9*>Kz7$#yzr z^P7UwDbsVLt^L5*%8iEdhPIEMbbA{Ef1|_}`>mDRQ{{ZVy@+a(w#hpBui&^Ty! zdKEbCp_NFxe)$d^6}XtiEoXC7%VGQ{vUgeMZfg9x_&nGx-XYaq7MZ&kIhe63Gb5*h zpM#52XOmzN(U{SIw+wFC_eZ$2s7ae79Uhw7AqQG~lkglhuVt0MZ=)#Y#B4OJlFz0D z#Tmv-9V5Od(lb&yeleRJL;|*`?_9K~3A3)?OE}@FszyRVT&fH2Y@?iYnBprso}0Xs z+*(gItpp*XDnb&W>< zAovDOzGriVUpHDzgeR?iDK7hg0GjFP97cj_@W=YtfX*l71R-&nJN z@*z6OXIw)KYE3Ip8({Mo+o4sXQfypWsYL62&J^;hNd1DHN%WlaT}XwqZ@h*oa)6`U z#z2#}1N6s^J!K^0c{lHf+ov#0ettPA8fZZ|3|H>&TIL=lTGt+Ncb?kW@i}CTMsEM& z2{*IHjo*})!5mPCe%={aq@ylBGR?3YTa%-nlfTe)K&d%^Myx~G>qPOtX z&drBk-J%LcI6E}w)yl^TVOiWu&wHx-%}t$ATj$BZTA$nJ;>(HqsJA7T_hRn%qqjn4RsZH>_b8H$Nl%T^o7yeo#0 z{r;pjYojxm9lJXJV9Lg{4t4o|rVTt5b<#)?7k)*73ecBf4qo`v`{r1Y!jg;$ zKivY7)zc(a#_oa{eGYV;dWxuec#=~)of2*-E~N=k!|FkJQO2dqqy)Hj@Iq3W4nX4nT~9)-3HHMLZ}C$5d;#0? zhelZ{Q}c))wu z5yINt--j590j5Zi5j~RM|L8SRi?LTTq6s4t@&*kBJ7c$-iaZ3%^S#_sU;wJAnX!!? zXN{|TIy$er6%}oxj+tI$LSSfH^|^?;t2&Q5f-HQZyYO77eaUB#BpVm9#mwE5ei>{| zH1fR_lL%x75t@L*dLZh?IR3m6Z_RRrs2cg78E-i1LRQ6u;!4;L2fk z`Q+Z{$Y)jYZJ?E>0zvQ#aMpaJehsWzHg#Y|U@gV)**yM2r_??IVvrz_Rs;E_8A7;L zpUXZl0PmQ4oiLo0gTU6j%QVjrv+k{o<}S~1C)NFJb!2PU-n&3bWw1JGL*?~hIOa;p z{yIOpZrItc-a>676KFp~XZddLA4 z=eceS70Mk+5Rs_KA^q7D4^h4Z>JHXmgpA33M)^bt^(5-sYX*n}%gOWnFDfvUZ;iCS z5y8CS?_)_0j&9{w39XLcxl(whW5fvWL(K?of#!bB9mMjyUWcIx3Cd|MZ(Bin10mji z1NBxHLU2XZMx8_5K?x}etTG}h?AU_|iSd9YMlo7op!l>@!f}N`kb)%S?f&5HjhtkL z7FP5LN$>2|_&QO#s;2*2pTv!ld1>PaVBPm!9FA#gp9Eqc+Vrdd^Uc^ae_1B&KRUY1 z!uBm8)EH@~#%TY(Ia)u}C#PgqTQmL(^csD(WL6(25?Lt(V0T)Ck%#RH;w}e1Fyy&U ziGAp(;Fx+rV1vG0qyk;;7+q)UyR{UMKHmdj-PJBR9e<?>1>q5A#AMr#P{1w7~Qi=!?`rK+23tz`{jOCa|K2hiccpkH96oAw%yH)LUvoq1nWlJm=xQbD|LymK<3pQ$ z5<2SxBh{+CFl1QCPHovN=_f|}X)>$YeQs6Mdx;Avcu8-XcNoVosBl;+7E01A8`^q{ zxf8#}2Aw6r&k_D5&zbJ$8Fz1nUk|BWl)59Lj}8Amzx}Gw^PI?s6PaTiuqqC)b66Un zi7zh%L6`Wc)c1%Qb64DbO&t}i=fC7a9>I`sk=$&Myfnwj;ab;Q(rv1T5(UC!rJs5* zp(25^fNt&&Z6#h0_Nk=o$gxB8gL2t&n4h6&{M5U=6@2!1?QdBV8tx6AMC7MUR*m7m zu*>&%d?}s*Gz4(Yj4h)9gIUq=mQ742w&QeVkJgW-Id8Yb$% zMcy5ABA-zfZpOjgsdGMOaPIBd_;K~RIz)U8v2}soZ|ii_!0&B(YN~PXw85i(kGNWoP$*}24H%{2I&^7aDW$TT_VJ_JmTP=u}?@EH&ot=HXR<67)ppVU*NrERQVGMl zRn8lF&p?Za$oLSjNHU#eQ_rqE7L6VPG(YeK3HL`5G;E&dAAL+zF?ZRE&U6A_q@DA%V~Z3On{>4;?Itpd7`mw|JXzinsxC;DVQS7pmC!ix z)U^2q&CDGq-|~n5)nd%yw>YJkK%fv~tZjSFRgIaXBS!LOQ+)^LT<0z~R`bf_hJ@gD zk{rJ%)Wt)jZ7s>PY})D3uUeL47Xg{5Vj08ZcmX#1kof`$0@+7-d0d?^@+!{4Pl`eK z*?y+-|CM%(ocGIxoCkyz_hZp_P#RaJblF{)oJXBUtfJ79mY(Ark|phki)R1Gj8en} z`DviDcGjH1NbrL_jOxW(~I?NE|PWw8ChT9wWtquBG0E%to%a$ zZ`eq0Dj#qI0t7??1q6iqf8}EB%v>Ce0G9tnjcU{`?X#HBer8qM&Q0XXS-aE~$WV{Rc`nb%6sJNSasv*iq@09KG_IAir!=wT#BlLaB>(MN z^EbZgd@A-kmUhj;;t3(~H3LzC;rabMy<8REP7TgEijgR6?5Y1g>-s<3l)G=-QUeJf za}{sR1GLgO%CleCAkJ0sIxMbz2W%YEebl)V-*m)qQJNBxFt1iT_$TcO#l(Y4AcSLA6y`mp@DS+> z{Xl>KSlJsirZcNdj>2@8fB%oJZwd}A+Ln!N+qRt@+qP}nwrwXnwr$(CogMS#)T??` z=hl5y>tlYb-|jiam_545*i+Z0TcB#tWq49yXwG_NA)z7NR%nPIWlc3n0wpt4!?Nn*g*iK9h?GDY5i-$$fJ{W<~`Z zGC#JUyase*6xHVf#@XmG-U&rvD1 zrGlO9&}T*`-U2nk!;-5As9;Lg4@Kd&(zRYmB9PUUc9Aql{a_od&N!8hpBPsf>=+>S zG^_2gvtd?Pv2Mqm1fRR2)8Hx*D-x(JBl@ zsrTHt)@8}P=EvcVa&e7gL_?40H^(`t@`5_*DhE4oKaNeUa@bk$3^fA^6Z^9HjICcT zIA$v_stT;vFC~A(DD^7u`)@XNjw%_|ZjV~NxwSpG33Ku;$Xb-PXA;ZD2)GY`P}7sy7h-bz4uU<*i9 z4n~_}O_od%zhqcu;pZhG>6?aY23!h{G~DTRlQL?>mIYg!>&+%hm^4tt07}=zEl@(} zUF(4r>S)2CFu;+B%2rwZ7N=Jk){XM8PNjzhX?7mU z9$w6Oh_u~@Q95D?E<%Hw0AmQw%YYXJ%Q${Rm=%@NpnS`9CYNqu3a+z{veYRqhIPRO z9!siJ*_ZxNKSrdXFQI;tA8@rz#P<(m8DA3|rG?OFqhoNzsYIiVT-w&hba;51bkIu!>Qbs#DakX=kV zC4Mqe(Jm7=Xql%GkqMn=&P@841mlBpJ{M7Dd_p{eo3lpp7Gyy_WAJ^U4U}kmzC=0_ z4b{MetE6IlV;7Nl`ozqVPQg!6g~YIR_#uX&&3OGwg;h$5F;XFDZ>HXtdN!lQQ+)5* zeB3lGU%UVs382`1LNnG|YUB9Exw{SbSZ?4f?XU*v2KL|Rc`VzPp_@N%wOg$PwTVbA z&C(hmJ7LllGS>Z2fCEgHo6~W-$_4KKdp6sg|r!IF0kewfBzkYdU*iZ!LQ?R zZnjJSn{v{+{EZ?ZT>LKl)`~hjIbSq=Xys{kQeYo-EJV?Ok0Q~k*4e0grO%XBz^{lL zXw=OK;2Q0^K$5aSoD3ey;89kB8Sh<6KGT%<3bT}CK+Vr+3{?C|>5{+lQpr=@xf%$D zQfzkkA{_Va%E_2Eq^;q<7(R75+r9zQ*j$PWkz#jXW5%DU*NDa_@|Rg7%AO3wh9@FW z#uc7iA5oxGL02}71aDApgDM>on;>2E+N7WorcA9-mRY(h|Bz&zW%4nK zzv0EWBvY5xR$PG}H%#xKDGwqZ&~Yzbp_VQ0#UM&F)ci_lXuakimmaK8Uq6E*()fZ{ z%zT{1B?ilE0N+Dg;oKdOyjT8LOlmbDEy+waIV+^$;iSWP0#u=IEtI0{l4qe>QxQ~R z3d*@mT9Z~qVI?YkcYn!#0}|~hfM*3LA{)Udtiqf&2}7rY5~5Hbo3Ni$B{{>Eio{qY zk~oz+^+);LdVF{)j9yw~d@^ zkOr6GW5$4b7ND*!4z<&;E~^T6IkT_Rl9Kpu;am%8s4rR@`KJAs+fuDh)pE}dvuKj> zN9*u^5ZcAx(LS$`QVMMWnrivz{yRL%W7h>(y3+d`E}5!Izk zsd<~KI`b>hUj!~4rn+Ba(;0bs>ifAFmIRJrkxouO@55uJd#%L=@a-zD6G<)b0|cMp z@Y_tblG^9r?V8bfCZAws_OG+eGM4&2OxbpyZf28GU0G};%z#Cr4T0#D-gWW7)my*uidhU%u)CJXCAI&(lbVG3>?RyVXKJ(v0U9v0LDIs@Rakt$c z+W!~}`?YqRXgBWfbn&>AmEI?`|_`opmp zwT=*^@Q@Qs+2!U^KclBI4AnBn_cu#O>BTXkoG{TRY_Op--o8ImPvi=uSxb@YyW)5e z<$-8Cab)wtOiXdNr!yB}<_8@Q`fzd&9wLhhl?qvGnFa`XdW1#FFi5s$ON^FSLyZMC z<_VXInQxee1+cKzqLbZ_WmN#1&06?ty#zeIEb;!@mc6%7gi0D=Ypfc5`@r>)6jFsPF@`{HH<~SD0M1D}{=JBoSXc8I6cZ4m~8^ zFg6~zQ4)w;yiOX2e!3tq`lROU?EdQ1g`iF(1q~@VzsR@y_56OF zp_6Igz>h3|`*?Nb(f)IBd_R7^9sz``C}Bu;)sKLKBg~`Ysx5raB-JkLfKX0ryc4#iwDve z?p_=jI!6M|p?;3r6A*1DZ!|G78pN65&SxO~wg@tFpDG&)@fj?Lir6>8D3Bx9$9}i7 z;M5m$f7wsxT?WyMnaF z89Bn~w!L8y_#+lUM$rk8F0dkH(+?33nG;z1&~ZSZsqisU7PT@Uv0%PwXtc|la(w>X zdpF@iHaH5Xg}9AMMu_qL@8j_dhiX*tu&{iAWWfZXvOqoZa5Ng#@y=gc6_{`_wy0Cu zU*8G@>|Ch~NHd-qVd{ES)~X^K*PDhdX!|5 zN|`Hg6ETP@mIirKAne2Yr(BD;Ef9XyYNI@X0gXlis{MLyN3r;hdN*FSQuwU|i-7@* zTB>soEc50ju4M~YP`fYGe=I4yeW=73uz8-RuzC7k!7eU(`yNb9VnSUGer+ouO^~RO z4ags6`^~uVu#4V3A}if=-!LB=G6WzOqhu&vl@&bXK%|r$bOqt7E$!1^LwA}nO}F2eH!AkkDBTg;sq`jTY)*U;zw0zSaM-Q zV|K8GN|fW59~Zag9Ww@P$}AG?NA+}9i|Q&zacfyntp-)x_D?m#E?TXgfdN@F#x3^1 z0En?RZ450Uf<{_Krf%nBf~<(jyV!Q?TCv)JJT(@Yi|k>0TQzCGPvk5%6lmY1rf*WU zwq?pprHc2XQvs=7&zTK8usJ;D(3!A6rdyNt=k@B`Q3EaT+uK$K*o7Bn^~AW;8GT1< zU9N6He#%zcQ%|!dtyikx+s=@fwT6iz#*)eXkej(ly0X@~)NEl%PG{F#y-izmZ`ReI zuuolGe9uX5T?~(up<~`!2w&t}SEy97i(E~jkA?Q!)Me=;DQh!#*7B2{e{9OitF71Q za;mkQ+4LH>Lb=!+=CbT`Mq+fUapNwxF5e7%cDRXkJAJ)!Wv*_|xL<>-l`UcJZWApw zN0e1cRl`@K{NBgZiVNK&vgFK@tlc-CALv)R{eJYO_Ah0z%r2$dbeeI;?~Ob)e7Rcg zZZt13z*`&5S?rJ9PR_OREtJ=PlziCt+yEOVzhrlFE`P=T=Sr>Jl57VZr>IxGv$xZG z`2Wq}tDyc?UB4w(2owOoFO37(dzjgn{D#TXTbsC>I6Bc=I=TOczg2hbHrWt-U-TUG z@klt~Qb5(iFe8F!Tw$k*Sg@hByPdiDhudXGoEf8J-yOxVD{z@yaruJ4G~;Ilm@0MVumig(2} zgW-7*NubaOy<$UzmRY~O+~PwjWT z9B*B#R9`ub;?>68!*@H+bx~C4NC};ai%r>ekVn1^ohQS&QGc1=+TQep%~)mep_-e$ z*%^K*6O_S`MRh3`V=1LMgoBAzoE*(vawx7T0cRSkD@BVcjXYpPX}Wez;8XI&tdb+^ zc(1!&MP2zdLxvuw2qI1^4p+ujL{@fy4?s~db+@4v zTuT|Ih;^fpyqFP=5bgFKv8Q*Z8IV54dEk+jYqJf+>Amz_fIMU^k`cwKg!EZeEb; z5Iwjh#Yi9>BST#oS&#{3sWn6c<;q^=6O@L_O(WI&DCb|Mt- zMarN_4$5-%Zz<434zEpZlP0d^lV^7+DV=eGoY8*gzqO{%4@O#eQn!jp5wJkgNQe;l9n=+R-Gqy65?d@5oimGNHcoChakB9y|eC4}NA4ddMOKI&2=O`!Tm*Un+n8eM--kJUY-^ zT#(m&qJ80wYIHQaySe}! zZFBfqdlW>>R@m57XPX)Z39IF*zR$*+C(R1$Q>ag8?6NyCcz^!^)S6i=D=$}-Qe0Mr zrY$5Bx(xbKszmm61`3dA8ZC!}|9RS0|Px9s{{HtJ|Y$T2Ct-Up$kz z{NE~xQo;sR%Sl<9ipuUl&IWu@0TyCCD{08Dbd_`t(Xp>iDyCHV}I+X|5r}=GT7_zU^f7yrkZCUwhrY?sWFN z+;K0Nn&5Vzi0+MY5vCZ%P@D~vVF4a2U6e%isBqa;zlX9XD`57SGy zOQRF2+!*+D1^vwR4fD(95;3`s(y1^bu8C|q_mLn&X)5-}gia%vd5Qv`lGbYq4T0FR z^NvT6fld6cafvhQ2#2QZR!2Ta2mSlRgGSsrw%lO7@kF&0%YwLrTR^`-kQd~F&;s0Y z{wTIGxNMtZ{pG|X#9uK6iID&!tjW}w^VU>HXK}cRH(89D;$V3|$AJ*2C={Tn^3NN~ za^z5;TpOL8k#tbOF5QgBfkYKob%Az-v;2+zjIc?(M1jz~W1rgQY>N>^%?bWV zToH&-qKH*emBo-E#|MJ7k4Y*EhNB1i2VRF~^2ax)f~c0r@DzfFb)xSt*hId@s-%U8 z3L7JWxay@2+UtQJ6ZQft4LkY}5EeRij5z1<1oFn!f{=`=c_b%oj7qYyZ1VCOY(8-< z)&%rgGav^|J{Pl&5W4G_C<2ovnPfPgT)$D~_uzH4$3wEa-W4gQw4;h?v{;&8%w4PC zEAWV_W28rZ@-DW|WUyeq9VW#TXzw$kiJu?P@k6fsmEEE_FC8)ovQ=~p3DT{y3wbHx zIo2EJI{b09Q4#G_U;e_PGjdR;lq(yXpoi+hgSF4wVSQ^q$TH_-}GdIuwv40YZx|q5KQyglW3?xD4iq}^30`Ge?9!$X8cDT31p_#v zg4Gfa$Ld#z$de#KQ%LIgMn52_Ve#qS4k-OxNX3dA{IoZm6~<&v_IQi)<|h{WxC*kq zXo@~4ndI?M`}`0g4I?xuj^)?%BHqa;0LPIOs{;aUv2hlG~X_a%av+z zVnDq}1(5AzDcm3xMh!lE9LN|^l60Qo6DDtW<|0h^Lpyj!P-ABHrkfILj}8?zm!o5^ zWrq210{o_(T5_!c(v*i6H`PMtzw=kiJ8O+ zl9bykTwpKC8MrXyZb`N?0cjh1kn%rQSfxz(#$E~<_F$9m z0x8Nh*2aYsk({ESkg4KOoCL+5@(O9h6ocEQV_ho3<{jslgoI(hb%*MfoKdkGRZG#( z1*KVK!}3Bp$pNYYDSK`qb4k-{IL=hjDF~9fC~p80Xw+@`)HWuk|KjE{jMmY-o%D}^ z;LA7zP8W){7Hay+#Ma1kuOp}1GuxI(hH;|<5(iy5{ZyAr0$t^y1YLg$IxokGk^vCbw%;t*+L=G~xWIcr-z`aeDxN8?&jNi;ztmD`c@-29+9C__qHT*azD&bNr zk8e0-!j`B7uqUV@2baX!##J8H4TS<%P-(DHwj1CSvJH60iJ|%y2paYVcL+M!0Er5uR~@L)YE0DQCy+uAaW3HrFTcL95X~52bx1b*qfqnw1O@ryE1Oa&9HQkG zp%<&_u@ky%+V`m~gJerEs>XQs0!%J|uiYdSU7=Z=#mgctsi2)8?v->DGJty)cJj(} zqyXoWS_34!Hw2gtKm{}rnW_h|<#3^pvY!hwGzHnBV4X-i_#j9yrwRn{gkXvzSkF^3 z>?BuKp;m(>V@O934lRlpH;BFDdCO=Svy1%;cN`23GI#?tP7^*T5t5ZRB3w*DQ4>66 z;qtmEhpvQeng?&}U+U;rHqVO0K=KFyKx~aN$(rQ`OD-1_(Klsjv~er3I-u+9zP!@i zPdpDn@(KrzDM8gxj2#mz7+JldYaSGBx7hlZ>pP69>|Xw^p&N4QoU}$qP*GctsSVZrd&xKI4Jx*Et9s<%N&6SU{-U?!`oR&3n|bf zEJbM#YU5OlAYX+NrWBL*#49w%K4SxyW&bWa;=yl#GF}O9aea2U!=9$!b%0aK2)({k zY#1{?_AEsPd;$@5+>d;238z2Sm|h%EYPE4qtFk`p_6!{wss5WH!IXTnWh`yNCwHH; zFnfA9Zk~ii$s|&_7VtKcwo9pkUzTs9WJ5i8^5CXQ_M@y}wPYn78zIXYM9JV?14u`@ zUX&(M&8Ug1IrLh%{sH*!Zfp|zM74<*0ALdv@PD7axZ3>oV%I)4ew%C$dmp|~BeGQE zEk`8V+dsKjVjef9s~&A!+uK?S4wR6I8S=%FioSpMd%C>5dOyNnBy3Fq;t2p1lMjox zyGxrPkoxuv?iiRcrmSsje2)8ikB7bc2UyNQud;_FZU@f;dwvYRKO%{CCSYR6_NEHg z-Y$3bI0MjU)ONB-c@t{*B2v-o`S>J0W|UB!Z+5#s%5J_oUw+P_>9Ze$kM|T$XrHr# z8Bg!wZ@(W-K8kVykYAC}$f|C6I6cV@;Pw9a{6^TWwb&q@Sf5{mPwS?vAmB7s;qIl!j0#OLPu{p?a*jkl^~+LvAYIjykxKL073nc zG^*}etsKUyFZnT>7_fML=Fn)6cwfcA>Vznq z13RhaTfM*+UZ2$KHZJJrdpO|litc&8sFc_vjYxWu;>PF7`4h04*mw>q3nqR-GEbGu zfWfG=GN~v}0OBmfu0-E@Hc16{Hz<)wm-PMpy}qtNAK5L@=gF}5ti;9%sR3nleE1DM z)=1Cf5pYXNLm0eo_V-t=^9DGGBs?T57p3_MovKS?q6~HD_|sK6oHt+{>fm}MQ)V5B zY8e4F3uG7TVYuS;G`QcH7Dlhf=f=_V$LrtC{eFSCNDy9Gs=*ZQN)sguae5u5D3EZeo`ECvBYo@ zjzHOUXpU-6c++vyEv^74kSc(|z?z#IKxk*n3T3&{84*u5;VU*sjr%wzQc2DR={&_C0yt9jZbq+c)MvaSP!e zg!x?Ab><@lt~{plu475e_yWWm94a|w_n>^-0f4pwUBZ>~GM;G|(}=(b@}KZ-Tn9JT zGjqI`?LuzOAN_;<=HOzRH~EW#9EFOp=(B1~iK&|lYyjJM*ZM0Dy@Dw)6fE>xY;c7e zew&<$IA>iP)#+d%1##Fix1cu^CD+Auv+ANJ#Oq8~x%0A8f}WvC2ro>N&UYgWHr*1D z#KaPu&}z_d&~R1j5g5ogu7&;zYL4b%eiIF-eHi3#i3F>W!5#ta{CCEnS=4fL1CTIj zb3d4gbA1xh@$&5q_XWQ?VRVw~>ZfDv^{B`Slb%NQHti9X7OrK2J zq%LMS*c5v9X4!J-ED}ImvVS&ww>A!-NYrpzKGjf4&bYj}`^+n9Ve|h-z`CO-Z*noi z{}{|VQ+4}bYgkH5z2)y&l6K2T0XvoexXcc6mDEH%a8b@C%e%dJPW@T6a;$E%L-ELM{f+j89AWBJB= z^Y+?+$p;97(t>dbZGpx&nLLF%AOzfB1gnAAt*d|Pf>*@8F4>m8$YfHN$QG3?w+PCV zm!%zB9RP{3>GB3fdAsSl1g@~&N?m_Ywa5+!NOSr`F*0c(+O_3Somx?c$5vQt=%0E& z@$+6DS`5hfKjQ=100$i4@kD@3*(n|^R1IHEzyLEc`V!v*zuuaL-zb)+4#{@oqq7wEmbpj6(Ug1q;+ z+&o^ypp1;S`yuGE0EDh7%#?&ZA^b}5tqjb7rvyxb_&wyd5oEq8p>P7+1(3i0m@C-9 z{P_{=bTLFNz*O%RAAM3IJw5o1gOzZR>PV5v-u>5LvZWEwvJ1WMOUwR{OmvP|$V5!g zE?Gd{_CY9=bkopB9fn8B7G_+zu^A|VWj9bP%6&trdqkzr7oZZw{yp3Q3j6!O6E+-B z8q69K+Cf5CE<$w=7R`60+F!N1PkO0{n%Z5#is0CHHsydUoD}z}9YPh{HyrIjISfe4 zDRK-!I&z57zlWjXv$D!no%f`v=c3iSX7`l}y5$h-75E9kXV@bas8mL9Uw?0z2$jr} zN!SW82Z0mc3FkaoA0L2QD~N?9F;IFP4BE*18=asnWsWt-br7cH+Cx5M>NQi2UWMU3 zX%gKn18~6ScOu^JB>0g4>_in7weGpWLw_>tH}q3sX|lu#NKtGe&`u&?zBxJB=~09f z_I1;(;1jByNavW$tbUOpL9?=~f{m27pfKJcdbu{+QmB2POK(={4Eszdsn!flim^W2 zh`kj)^%ZtN?TXJk(*bx6`A4P$hgP94a6PoO<<`6VutR5c`xYLr+m5r-J#aqBjQM(H zKJe&8p%R#}I>8)?#>ii5!tbi1HAXtDDFnCXjD28t^*~A!piy%mpL@z6!(KI2yYPe{o#v^jGL!?{nT`usT zNyNe|2npF=B&Eq`-+_u%hHJO`1?liAHiqs%XW1>|k>QOu!{gd^)B80JZq zlmSVD$eJX^+jUVdw1~30^5{WaqJi#G5&nVkV<1za^x`)6n(+8#P);XG4{@*?_i>s@{@U&$pU7koZcHme-gII?ao5JBL0p(ZR7hvoy)Ijv zF_bX5W=hpo3cT(^M)?TrsIpUm1{1W(X8>U&T;!qyeRjwS(6y6&8#2ZQX1WifIY-FX z*iu1pQWZtu*V=|^LQo?cM@gDQn$UGsNDCcB<8g0%P(j;q(~EG$Xs&?!l_*bocF4y^ zmY*8{3XqSJXi?|lX~9Qpr4}7@!}wMfz8)MkF71Ck(i(r(U+70XtaNeJi15(-%id`RN z;XbI5UkwPf;{<(&jDw*7*tY`K{ZK%5Q3MG2(C_v6r0{R|=%%q?1 z;nadK81Plc*SZyA+Qa-qrRn4Jx63J5{aTk1<(7Sz*CUe`PX_pt@WO306Q+`w+9xBs zc2x8iTD4go-d%|IeEJ=8fYUdq%KfQ3jSj93*jA4SAA@l zXg$C=2v|jnU&0|6sGP%u8BUw%b2xj1X|y>lidchU0DSKdey-Me3GU8*C_{Z2JgMg} zgV}t=v-s>xb=spBPTo4sM1E87E3`xI@l1^M8?hfTmZ9%~%3ustsdx4hV=clY`IRIo z!j#lM7x)X3RgezG!=pTDAH>(N5TKcI2$N_TcrYcb(5lQwWD;7dx0)WX*Iz)GdF5zf z53gn;ZOB&}P)ey}-dHCKiVF7x^H@5TpRZeWeLAv?IU}+mz)UDsfF>{}V2O|dZh2O| z)=C-8eNxL7zA}jUa?%a`IW>EKIz|3(95fe^&3vm~hgM@Fz1QGMd42MWV8+gtT6w=A za~iF+o+RYJ@(K6H#=aKSc@lzNv-=s104galN}0U(On_mKBx!}>aWxQ_;|d|^Jg8Uz zaH31j{7wI_9kr<=AUqD5q0xdqv(v zYEn5hc0P$FD3npHKOJD~x`da3MR7nmAC0&UMVY{sBHo>=T9xPiUP8T#cJQ_dBMH3K zy6z+0s&l7QzD5n&^dzVLHY8$N5`6UZx3@Yv8B%t>zKXReFx7ZBw{zB23s{IT6Bmo+ z^Yn<=)fHN0K2d&OtZw^!=`U|)UK%2Bnh~5=QBdWuNrHdZWGv=*G~W7%!@64`wPcF2 zlwxSg;c%h16R}hZ691%sk6>T;s+9XSubgu#)NCUq>!kd2tg}<&Hm7Nt9GC1E7%Qld+X*S>v`^ zDFM|8O=i+up*#EMeFY8Z} z#LFt`j3syvu8K;RoEs9)oiK3 zI+~4?1)Pu4*7=)Oc>fOhZr)d3kZ`k;8wuNR3>+AsA zsY?tCSF;`X_O!WnUzQ}d&12|4-iV--{_|)psD_VTY-;@6hCXys1O?|6i;Ob8a?CsuwkG@T>K85H%siJ2K;kFui%11Ji~8 z-TO|=VF^0j`^_8?j6;FUQH8XC+5-RXv({k!Hx_&_N-~EObL&vzcChp!TG34?t(>qyy`$M3d(uk+p9I~%H@v3x zLYcchT^w|bK(j;0-3KBeP&~m4|3yjd>JA4spynE~?4_ab~QsQ^m`aZR9w|5m@&VJdw zcwW}>S30F!S&Qbn8IzMPj*#QR&d_3Q7s;07&>p#52im*{gcYHDyKc{mee~yuu^=o* ziY*)$Ge$OjNx{;Y08TND;cc}qUk|jbU+;LVS?>nF3sB>2an8j7WTB8sY){aLwe{_@ zDoN?a^fW(z$=DBpelDXXH9P1yq$qvN3>u2D>>*(B=Nsa?He5beAyW-1#iLz!M~ zG_F9XE;yPPk;f7v)B?J|7+yVFOz!z9oGVQf_ip?Jnn1-{;34#t`=;4MHnXQ>rvMVG8I$YNM3kP1DWhZsqtsa=gm z5~3%=S}y3EelD7_n)now>{LN&WRr1o{l&H;3T|7sy1u?TLqdbC-=6G4Wzs83d1DHL zmKqF_M}p3e;nUm~&zpETfe_kTv$l8BzLJvp+E*b)#gEU%2=ZdFvH_K4>7-@NZpkad z+;Z42YM)SNMDll&G*mFB*>UoGqBfx zdOa_k>L3(3A%cj+y6hG1s?RVWIKLroHz8WGRzBEJp$>yO=s+lVBqJFUK@ZpAoCf-& zpP8Ze`4d`+^u9O~|4pbeaefRQ>3*p}l!G<&DWqQy%8&4KdXJvWN1^@I+GtO&t0{SF zGiZ1ye^$S|WAiK2Iqpa5`$Vv|LTGiR<+u*_7ae{fcDvXEz&c=R^E_Q#C$2R`cnMFL z*AP}##^}<=teK+7L{ZeWoER!W-KLKJmWBSZ=9oU)rznn&yu&^VjtdgVoSMkgUcrBf zxMft$M6jgL7HG3C2-S4z0a;xRe<^S(h$Q)HdN68jmd>BrCMoD_u3wrLiK@_HH%PGa z);?aSC|6B533#Bha*we9C>TbTMEC25_maLv`eN7we5*$qNRCU2gew>(*XiKJc@1Y6 zYA_kOnU5F2i!?eH6xt>4gwg(}`sG{9rcZoAi7=7wv<=+x>+S3ole4{-_U2B_L{h7~ zX7og21tF-wni1MBYX{G$$lsl;XR${RKOr&P8`|n%K!2KEnCO>=WGkhpcyZSTvrO|x zaNmq@>1`2Al^)_SzM}f{lsro`RmVhi5Ad)}x*r4k7)Kv5hZq}xUFNqo_BkK8{!s{? z9Ym4yhINlaPS)42q$*d2^a{I4BrD33}b2t$Rp#r68R4097 z1`9`-6xmls9uvVaFHxbn2^!GzyC{MJ20=H5+BO2p^3Kj1)|1$DT@w`sj{6yo2T@O6xj~^ z2HNR|Kduzm)1gj^hbM11`YZajk>fohgkj11U1IOF2wzqw%EPms zVH-(^Dgj=+EqXmMWKoStGrDwEIN zd9-X^YM`oyGk2_o5j`}fon!LVI2j{Ezii>rXC!!?l7EF6XYBRbyjk7-9yG5Jfz6C3 zp)a%WI~2ZH=@5=0t)1warGTx4i0T&0CcTC^f?I;tfy{6aX$x8tKo7f=p?t17!d@=3oe^lh&kcQ5+RtK1T*X z3c{osS3*B76s!jOPKG5`+IB2^CKAvvLGe4q9=(hk;mV}R2kBo|a?Lm+oF)#nS0(sR zkCnJjOrl9W20CiugD&zfnF$nII?+ALY<6$u`&#OT`NglzR0Kh z6w?D6r+$_O!0LYpUYSKN!ccq>h0PFBMily%r4KA;*$R%dP|}2MP!~Sdp&1b`uEWPX zWG{;V&q3!r#Ly)s(6r+TKIsRn|NKay_#Dyi+N0o2{3R}6wgTWZKK&ko&dpCGPopms z3=>!HN(>IETA}{ISuvL4TkOx0CLwRk$9>b9jeC31h2S?{1dK5x;;ubg+E@@_neIhT z5}Z!BP3Q{gcOtRr5+FDZ3}F@+^DSUK-?qe57i(ep(TB4x(a^@)s%E>AqPR9wc39U& zs+*|m?D&{=kBTbqn`5vD)Vqcn1FG5F8wwD9{(4!`QzUIEtDKL@=K~2)+M$V3zjP z>rVdDp5_wn(`Im&_;s^PaDNnpE#LAtH2w7r6u5Q&6o8PI!#^DOdk*U5zrrzVO)vxcCA0>DPUmhEv33P>y4H663CZLpG zs_g^f`?;JDr-IgVpLYeblEyOJ53C9tUb#%@PugG{!yV8YzWrz0>@7boA#+&x3kEQ2 z=d|6uo;B`BtAXcs?@w9hW7~_F=T=JGkE&S;7`v@Hq)<7Ew`^p8{XxO z{KoM(E-+gGNUNuI-lMH+p{1?@?#hU4x|6=)!RB&MM?~iv^jo|(Ju+I5$nQ;y2f|SB z?GYFO#5H42&B-X|JzO~4{E?si867{D5P3i=as@1qH2==%fd)D!r4Fm|pVU~3|INJs zwlSlM=5nT@!7WG0hxnE*4KG&9n&O7@Nhwubidv$i>@N^nl|UK$6=2tSE7WKUH^47s zAUY@Pq!9ez53oQx zM+Au(7N`Uc>_(M^j`^F#bMcx;Y+zU+Adsy&SA{p3(9APH0oSDfhk9 zsEEs7>8MCe=>KLXcSDetT%{zZz}WX#qlPL(P@uYC*mcHAW&z;_A^MTFf4p|FQy$0Z#AACtool0|xQ%c;4~&oJ1K6Yt z9FRYZ>1K|(Eh_{EsJkm7aG`*fK%ES4GYzD;pkr!F^!2${mWgf30V)s8k%WT( zF>2p<`#*G@Q;aB3m!-?LZTpsO+qP}nwr$(4x@Ft8ZDVdvGD&ympJczCm-BS6_Fms= z!VF%kRd)-x29SYzL)H*6kn5Lv`3s^H%3LmjE>Y_y2}wYBl5$c zU2e+q=+;--7X~C-;^$6MnU_MB?56}v{e_kkif)ETm^Oq0?5?dQ2tE!LK@=D-8f>P< z7B!e}K}n|eg;SVfhsihb=Q+<20!FKRkk9uA#IPD477QkF>~hqWgtzjrMailv2^@$} zj|k?5S{oGjN!1+-AYr4f*e$a93T~wA?St7CLIuqJeb*%2T?R?1b9BA96L*~}LDtET z_YTkSB&Ri1n3M+!vJKO4n~2Px3s2n}HmMZn>1sAy(2~T#2jdeS&{!99CBzT$poI@M zMggmUv#jCQ!`0ys5K*6zoj<~X>y3g~yIh(R`12bJHOKT&da<4}pTEoxz(&6+-tvLe zUmupMr7s2cqFI}@VfC5|XSnZU>_I=Lg#oEY6z%dD=!(D`AS0N_%caIxE7Zs#Yf-rAj0=BpQou7O5NJs&iFjv?zUQjRE zaG9vXsSo%iDJlT!lnHysoE$!>_Q5hgNUBqiEMtthF^nKth4~yR$0!9vRgig2FBs^N z1xi*WRHb|`+%_OAbIvd?kXGRBCIEm_kmdscB_lODfv=~8|CbR->Rk~a++iJ!kVB1v zpx3f*=v@~LKiGr!UF9=zUd0i8et&NQxJEN^G=WE8^?)h$5HT;(GhD$^7;uG_KK)LD zB5cCFUWYnuu<8u9wMjK@$)!61n&`cEl|xf19`Z&#*5i8~~i z0{cidk#7L1KxgnI;f$-bIxT8*&>wSSda5IwuDuY-cKY&gLwOoF3e$T!2MB^vPeLW` zR%al3+h;^}Il@2ju&5aa(V+dg(g_@m{;UwP+!8tKP&oy{Oyo<(^_zXxg-(ck3aO)` zQyF6OO9lqbnR)UC2F|4y_czl`Y%{%*T+ea^L>Xigz0Cy|1fWSwcc1J<&?O!elmM!r z*EwQGZy?vQhb374uqyr6AXbqxY>ic-OQ55Q{6b>#n`i8wE7x3HUGQ*ta4%k)-CP?s zHr~(*?g5CB_Oh$@flRkVh;9k;vI46J$gOt)CG?PMt>bQ!+ylmQh9JQ<@hqdm-88!& zt!+iS>Ql8+$$cj)kM*Ke!t9L#)+6-?Ii97J^SUjs56@Ko1}Vz9!dIa0?l~wvdU72B z#XZS4Hy&;bPa{c`SshA`1o_Ht7QPswi8$WjNJl|E5|$Q=nl;>$$BDFNYS=!P&7gq# zU84%|92?gdud)+J%TT1XzJB#Q<~5XUTgZl5f5EEp@tdbh3#eq*(0q5_}EFR0Prmexh@evEaCpL_RS8 zDiRlUrgFfqznS1B^hyJNlknv%{^$tTZ;+mg^bxEoG?UkE6_Sv~qWKOGP|iuO7b=Ef zSY{r7Xs$Cf@h8#=v>cW@o7KBYm#1z@Dms)XUd(bMQp{DD(_&%iIsj3@*c5(PSXfqb zxC&qNC=XO9SJR~E1JtPYDThH)$mtBM5Q@Wyc{DJ_8Qyi9Tw_XMQXzn_5MbkO#=}$F zA=a3ih4_H0Bd2aL7VIm`X@((u#wHuWFVi));*?PJ7wr$XkvEHL2~ zQ>&w$M-Fh}4eOadW>b4=hmQ-o)`G{oaeck@{=U|o-40M~({UU-((P1h*$Mq2w$&_S zG41swPOCv(?x*vIx2u$hntB9RNq7*pqf=nYbd;hUF10j6dzSR~Ae;}2 zRzXHqfiyeyO(fICGGzb^QGkfDiyVp7ppZe6g4P0*TE!$+9g|>e5{7&dVSLOGk<(m5 zM}o^L!b}Yz?YG|z)}Oe-1FnO~5fSwia&7H!Ah13%gwg2^FQq=K+RbhtIL1gv^X_VU zkqPgmx8OL#GMs6I#eH?`G8Y-$?u(t7*4o$({`uqvbJ;pS?&-0s-55o9JE;^Zygx|W zA647!@#<0!nbZIdoDEhxhi-sqTdRPb`k~r*TD;JbIN6a`Y1Z#85ItI#pOC6r=iJ@FAZFGGshistxR$!gIi*QZWZ!KTQh};aJX|FtGXXUpG6nuOI69ld6xls8>VH zzgwM;KbJYK;Pjt;Gm+RAL1CYCR~4ROZ_Fec>Q*6U=2*W9zH=XtyuWJ+z~ry!PX11^ z54f0Gg?|8w%WxtQHZHS=5a-&8Tl7jpaXTNbN^%)5B-DBm!b`D}7uwC(A7?v2_QF!V zGGcw@(usqD*qA&p#-+=V29^`7Z*GbPM9YoXkR`2Qmo!+#DsYCb?$vu zsNHmaFj`VVZf0H^ZbjWR?7K&Dlb^*TQ(WNf9mCWvpbAxGWL{*5WTCz50ecLOH3-KP z)!>1-->;d6V;Ff74^NB6^st_MGkhWu!ZFLi#y?6KPeQnnoY+y^2gg4i9{H>cj+Rv^GK3iNs3Z7uOS>10wkm^BS35fOW6c6j47NC zU%X_7K=Nm5m88BG52 z@P@bD+ImqHw%B+P`RkH?Z*y8A@5GufK?u0&ankB#Y?94(uN-9BrYN=*!`QI__|QTz z;C=fum=oLXh3ztH8&~U7Id2P6ukTxrchdUEz6lls#+^9K+~n1Bjo@XoiNi?H6}KYf zA+&%D?z)1(xm@~};D{P2;#(3>95F6L)`Gxb-E}8JDF>$FdEc*?+wlX)0@L|wxj*yD znF>VEN?4%7INe*k_$ogoMZUmxv7XScjk-o0--I8&{C#fch`({Bg_KIjBGf9jXNn>A|Tn81fxPYjqg-pn3f&L1MJl_C%z%70q#Yz^-88dRXka zJ0H6xObTRd+$HL4C?|$uY?`Yt_V)gRc6tRiSYrWYO(#fs)_zOCiswo^)nh25gr91c zv))*i29?>y#PR9!8sb8g9=5b|n911Rd=-1le`o4fgbbL9Vt>_bw{?HLv+5y{)F7-6 zX%#(b-grr7$`c*+kg;dpX;`sKjiO46#=Z~5zwRbv-GSxC6lfRdcVv=4?Fjs8JuVvu z+4Xk9EfFhrpIMj8?9GZYiBp1dP)3$LFS$MrHT=~JSy~%NSJsD%No$sSxt(I3Ar>~) zUY;rv-Cgt%o{||1H#8!Y=bF`*Z>>4W0MAF44mA-!AMO+gC;Cr9j)HH`!`yNe_2 z*PWWpP9;YxKeS7Eu59a3hE?l~;Uizhh*VYhShW?T+UuV9$cC6K*s3Q;vUAF0eBm$T zXHcwcA^A?QQH11~3z~zgt7`De#J|$PsjL=M>axUVUc&~#cxF7IbI9I}RLgeC&*v}V z5|B8DySJ5cY;`MJ%2n@d2MD*v(8wE)gGHw}egqHimu7dL2h*<@1EcegZE=77zTgDq zatA-1e%^+=PIv8!Jz{Eyeqy?d3%&q{C$r4Czi%k7*={Ipsai88wvh&*|f;I%0~xZ{w*o3>u^3#PXRfmC@mF9NB`!p(jwZ z8h_kbJ)wB$YSzV9i9<5@di~S7xpC1xUOsX%G+&zzjB-t}NazC&>%n0z4t7g4C>XS= zQlxUhZ+SV89z&2tBA>YJUN#3!&#ueCf5>Ehxhb{x3yJULRPY0E2|d~o9n6s@{cdc5 z6y;XjAm>X?yn4<`$St_vZy?Cmz!%D9K@=~icK15~6x+SjH{lW6DH3oRroFZcT!{!P z7_xpW);u+kVqquy##E3~uP;%FE>JIM$fY8-r${>vpGHz8+)KD`P8I7H+>BGxp+RT( z28|z~YBv!SOy5nKMppH9aJUrrevnLOMD#rtU>KYp5~jukh|0S7bg%&QWr#Z`Z;Vjy zKL|9HW~TJBI3czj&+&wz^b(6KTbk$IUGGk+G3AaUz?H_W8*-Kk;tpOQ<9Go49`};q zkL{m1^yZ&BKrd;DoPLECN+jTg0DXX{`iLxh62^XC&ry^6hkZpMomRM2;qC2q_nzF> z9VDd4aDW5}!~5gpd2#f1&&A`OC>AZuoHKl9M}+9>X&Sz6S0j5de*z1f_Zrmi`7Q5j zd2$u!WyXwsXs<$&!}GoF^y@54&Dih<;m%xWvWv~n>jw+TRk{U(3stbHp(bYRJzj_W z=jhmtUhl8h&-UZ(pSP>`#k4v+d?Y23gZWhC^>O-oan@O^Kr_j4PTv4ALb!2Ffrg#U zvs$}^O45GG5(ZuHWII?^{Y`Lw^xburt{~5a_G#Y(Cv!2Vzc6@x_WkG_Ny1r{jQIIB zEIts(tZY5QE6(|PO$qekQ^$c zG*o4qbn0r2B<4jvZU**$XSW6NiZr|!RsHzM_T1iZd+Q920dSi%nvJdMIo<6UymxV3 zHn!nBy$1EGHkM>l55>j+FI)414f%)OBuiV32ff zo&zSReJ`&psSRrpP0w+jBEtp>N46K_Sr5XApCV%|q8{&N6)ORBRO&fperQC9?N#a( z7!$BM&MF(0CbXud%vrAWElXvenZ`zYWMrNQ&*LZVaei?M)>Ktx{r95(1f$ z^2Jrd{Zoq5NRmwj8!5~yqLFZt#5=rytY-`e_J&@hZrEv$^(qfL*vBWC~-T2HYccrV(<+;Sg`@$);2RU4uEOGv2MZ&wdUpB6-O*>i^ z`B8_2#Z62$UrQHfSGcp&-oQRfZCkPfJp0bwbo&NeVu! zU35C{>ve4(+ODl*I^W5p0m*TC{0kRgGlR|xX{{Itte`L|GhTktB;QU}-_)k5ut9sN zOdHObW*%j|ffkCJ+Cg8=B%%m#)6z-|uU4eSFhpt_z;vrX;V`d`B)BfP^gPtSC@a$< z5rKZUdPRaKtb3Ex0+KH_txn@RgZ%Lxo_%&}Rt}nk-KJm8mY1@~%c)u%PVO4z%r9T!jhnw7ue z>0V0Ijd3l>>pBdU_)AV?La6e~Pew!}6Tk@|uTfit__v0ia&eeRJqAj`)g(~mZ4rtS z2o=6%$XP;IZufHq_eKep8q2AlaqJ>Fz}tt;%REld?CqtD9I3;5vb8*+ut*V*XoYuJPo9lUo}T zfzznzKhojf2Xk>aSQfib0C>Xv`Jt%uruqDCgFbu5+1s6eeTck$I~YIYH>11#9qVSJ zl|I?g8O@t(DJ)mDW1TkXEgP0N(5|Z%YCPfTQ?4QiChuo}b1u^nnqM-7TN62?$e9C<1qidIRffA&Qd#bw+uOVWZztf2;9b=Fd(F4@YI zsH{=Y366g9K}>ZZswu8y3;I>@I!&_&?qm!j?)k6^t&3qr9Oi7mRc=VxGcT_g16wP4*|d;0$PHKw zYbZ^+m|F!lTYFrcoKe%B(;qL}KU!vYdst6Aa?5c9JaBV*77cgy5rok-;n0NuyIM*p3N{j8P-XDo~-4CK7(IBI8;8!-Fve#AFPO&hl#?OX0_7ze?_Re%tBY}u{swc9m2`^^pbEe${e zN)wH$-2m@XYM%e2WKE3CpPoYJ=;hUeqmnMCxV@gpXr3rxXj4H<)QH}jUJxYgilmAy6_hL9Ih2IjNQAuA)RTKA$lLiy~;YqsYWbbe0h-F|#&C zNpEa+@MYyVK!)!Sw`?cZ_b-g~kf~9r^A9*cb5qWnSS#_z*v5o!?8q!2;n}EhsibFf zgM)QoN)ja7KRnNtWn?{%S%p4-8a3h?fNMam3hi!{iF7{uItqncux&m+;22FqB2JFj zfFBj_u_H6ri+%?J-0D%3eLVzyZ*)VoCy`@qS$5fmRQ@Q4r?k0VomnLmR4Gsy0VP4D zD}yfxs10zb`6avr4gQ?A(@8b%cFLDr7TxYik|ZcAW}%99LJ9jr3H8dPC#xv?c1xZk zM1*loK3|fN>cw1FmG+t2JNPw_c?j!sYWre?)z;bCoqOf^dKkA3)S)x$j-bjFBK-=6 zTv#umQMI$E^X~*ZyU2WS{>ZOsLe94DbyXg0JBnqWcEqjOpkZUP1_R~z)j_D0y2_Q0qtJg!fqKl;bXVIkoWl-kEaaohEaw|r|J;uq z7SCrMI-82{mW=+n(#iOq+K}))!w#-1oLJ{Ku0lDB8}cJ(>ix$s*C!j}(O+e;I*q+e z7PJnk$eJUcYL-86X3<#HN_GgAQch1#}$c#YD=6n}Fsb*i^44qMwJS zYhJlkR5coh!TN)uiV%>(v>mChDk1F$$XRy%)X$M;XkKx!ye2{v6~0f5F*cvoEidJ) zBew3$7~AtZ<$No8={1+gU4|Z{qbg1`+8r;+!#vB>5k03N!e)oEtKYa%Bkgd1Q74y7 zt3Ahsum4KHr#(|xUxspSqkk`*#s9@zRU=d>3^{e0ZKm(JVH=L5t#$|2m?(=N zV9uglnJBUyBcEO}9Z|6ovnu*WS@6$Z0_#xLJ8}mMIrI_!q0@Z87kq-)8;pO;kl8KJ zsUPWPM34#^-G#WA$q*k zwT|u=iYA0~GMt3m{Bc6YrPwe%d`v+~B+d4xP+*eq21xuotP706Lw$n>4SoSe54u-* zi0%2=_B1jrW1~dAb4i(w8kKIDj~fSB)9XT7`mAQ2#aAcGfvB!cftpI6<)VYnC#8(s z$rGB-TInm-sm`q}>AVNu!(op9^84)mH4N5P-iR@)murJtI=?sjXkp``!gWq zI;vjDfQ`HRgiD8vR2b?DEOaU?%eM`qcKW5A+9t-A9HSWp4d%xL#=rxak*mhV0jN z@BZxz?Bt#8Qt$q<+VB4&r55a^wv&JZ0L0MzXEM8$i-qk!qxZ6=jnf7@!p}^d{kV?J z;{i!?nw5ppN2WztPwC>*jVT=R2nHZl@br5LBzoE3@8(3C=mrKBhYZFkg5qV~*;>^F zN!J$58u2G4y=b$s2kn@AtE2XfrBO7ix77QJ`c!ynMXdGMBboT8+?V)pgBTy9XlA+v z*6a6y{=5CPbbaJ#+-sI#glay0PR8~Wg;cK; z$zhVvOG$^E#%6IFrtf8Y?%8s^wea@V+jMcpYr+Lp5}JuvA_jnrEA{EU|?gu;GyZAmN;Y&u6w0rR4*e9vG=pqB1}QkkOAI%fIu6dxmYVWB>Ai#QWJ zqN%dEeX^IhtF>10wh~)y;5_?2l=WvA_<%)|7DTD1)Zt)$%~64Y=>-F3_jLxf29mwv zrEZc|4x15v>Y7MmYR{e`DjkVePZpJVt8u;{MBJAJb55>z#>49Xi4atm)A!{E|K({#gb(} zO&F~{=(tex=708GtuZURVvy)9>Z^<-o`!=xAXDcOB6GPRj8=d2?X1_KNTMhXLL<9D zj6&#kj~PR`ZVE>}qqI+dT9Tb3isb?s$jOLnQM?Vz)_&J9*g0@LSn};KRmc}qb*;sP zYDTRc0jq^($-S!xBfXLm4U<-GDUXGRYDI$E&M}RmL8~TD5armiWc`9LT71~gU^xCz z`!gvi*_+#{ae3_-36DezSx&7&@id2>2jQS^2TauhqU%5~Q~_#XkYEe$-AmqpL@FXY zjziH{CaY`sJa-Hn0UoBP37@34NX0&vewf3^{@@$~$4-`ulzcA_rz)+U&dd_pL3S&t z4lTg0pbnmY-YeP(>A^3Vbd{n@j}sbg*hU6;l#ZMh6AvuK#{JUhXwqQ@jb1BMXshKp z>TA`Fz-xF|1I~^vzx2Wwiec`dq${7!@UpfpFNi1W0g4m+)W@{5-0Laibkx$luFqIX zB#t;Xvh-uH(27NS4-9rMatevyk+y4G&)@oU>%((|IphC4B`8gAD*A=`?Skyi+6 zp;uN6r9=x`0>s`vP*dhLmvIMv6+eiA-jOkKCO6)3#1j2(_><8#jDSvT^F%9p%(4?O zNWv8K>V6SX4+A6#O$R)pLhl6#In{1r`<4jS(L zSV*&Y1}k8%L*B`^PX=gz5zuU^_v)-==bGUqzkXV_e0aMXY0-oAEYqU?G)sZ`cfB|w zzI}UlVB9emKK0A)+@^Jn+0&NvKqR<_nZ%S4)%6}G8boH&FHEDKA4f8)lCE4ssO3-D z*tpCmuD$Fv9WIO!d~F%O2lvspiaun%rHWk-Re@+Nq!$e6YZ<}RyLkhZ&JP5ssZIYdr;C5yrbEX_%JFx2*CgIYqlhU`xt$|hP&%!b?>qoI5Up4Of`3!x_ZX3%+ zaeYi0#;yqY|4nV_?dr!{S!8(uo~?J5QQ5#8q?Z3#S6TI~XGmT3O1#na?&bUH#Dx6y2J*q@JI_6-c8{iGRuRr!-|=W= ze2^iq@;#{+9JCxdBBJS~61$VUsJr>&Uk{D?;^zZe-56%qiDquuL#%cfo9~9p>iLI0 z27I1$$WMKf0Ox*?#`Ji z{h6++OHbL#;v%6D@5R--i|2wnYmJAQ4#bJ4b2 z5&uJwr~E@&{-=ZhD;FncyMH(9f6F|n$G9qP-2zzybC+qOk!d9oNL+!8B$UO~ETmmY zE>BcOdcAJDy`kXpmD03c19@|2#?QSuCX~y5#0n})k7ms9LbvcBuZbEdc??zUt@W~O zc9vFWWlqH<%8(BS`oos!3@HM;0woQ>fPnFt{g!|(4YNF2P+u~=xz(c@>bRM!mSLqO5 z#oA1!6&YPo&9KtM=16io_Dj~q!s;NUZ#Tl27l{2yHuyK|N?aCU+nAgviD1&&A8lrW z`=#V=;z-8vv7qd-W;YEa_E})dX zzzzhgb;0$+%;a4qVSPYqn558^Kvilqt{iMkv6}-GQvkHZo!l{7pW4z6wur4|3MFG{ z3}zarWd_D1TCuDITT2Qgd?`+oI~Ih~V&K7rM9E}V;e?Rwmv_F?qPwY$5nh8rN9%rf z{SG0$z2z8}QIND?I%ukMX|rttr&7^=-Mvi96)GNMW}n1^z@J+a?z+~Yn)IpJQAISS zpk+VR6C&lipS%n)v^7yHRWcK83zb-X;(|a9@M$8f3Wh_2KcK_u-xQs>P&rj$_yzV{hbP z(nxBQJb7<{d(2^{L^*|KdFul6$9SwC;F-d3rd|Pkq%I1*cmfuPL~N40zv(x(dr#;n zGC)402;+L^@WO&KmJMh#4WH1x4!c3DT>WULVerq2W|G)aXj#IdO7>^f;zna)=C}n} z!2fVyPBFM9UjjwgOLGNfdyLu)!ao3OkyK-|SDdp!j)Ay7&EtlRDVP1yRhca8LkAhl;Ys=E=bezjrz<)?*md8RzcT1MQr z%EP(5u#;YJmmI62=%2Ph8-_;rj$yMQj_k^AUG+YFY-Zcu9(BjpuE8V~8$U*^r|caa zcewSIPY*h`HuN1A+l3ZizUqP86};p)L0Uok1!X^=Zhj-JSu}yWuEBhYt@{>?=ZEH` zEaQulXO(HzKBa(c4JT>`1@cexS1*YF9&K>)LL&D{US=D2t)!K)m)FZ81uGpuy8<#BfKPJ5;Z2yW;+T z1+JJ^pW#IbEKZp6Fk_f4|K&+R`i?s?4*r==hTz*n!Q+(_9`&>Gg`H=jC>4K4p(Lm< zepwDkbbIAaq*=6QEa8!$&RTe1U#H)GYTbqW8y^-f z2s?y%EpD^ryJdrwY17uBc={t^kW3jvhOq!)1}RaId>~qcSZv&Cl&bv1?jt9 zo?NW^;|YG^fjGfv!BX7PTCqKeJ4d5U&5O44XQ!r9m>8@&U+RRX@hz|r!Gs+dfG8er z;_Z|mngKuC4TlG>6NPVXF^ulk?&SNCIWYGmd4c3IC-)RkFXgVJzh28h74g&PKw6v;E;rDLP#+zBnx2ourg9qBVXiHo57h&W(H+-Mw|^ z$((bPhF>9c(@s_B7+3zTs&6L+k|9YnDWzB;Nn#^htlu zLkUH02x#Yp{o=>N?aK(F@=-UZ$q?;Omx~j5nX+4w>22FdEWFlqQ83B$LUJUn z48Z48Hqi7;^#QXYOw}k=+D+x=lV;m?Hvi1310D(&3~c22CdhJ!q1KXs-YPV&w6e;g z06o5{%CTFL;Jnsk-q>VYo-(yfmBo3~PxEj0-ninS1Xu?-&}cg<~+AMpV}PhY~`fk(Nkle7l5 z^>8k}+iU2qC;xVX3(0{HT~!cO74G|Z^#EJbMeuyY)_S=i9PHokEzF!dgp|p#Dnv{v z@YoeDvs>UB&D(RY9@VSfDjjwC=k0*(Rm<6^Eg>qs@`hNrdC#r#2i2Bc*Bo+)s$`G7 z-E{N=Bx^$f_KY$jc45dgO1WjfmcdI&!Q?xKZju;~>J1Ro$Q{# z2FSw2sKkmSr(>7ZfI*HmN_-lgQOt?L`?N3i(f+c=s?HM($wDigF5}qR%+=^!Lvv9* zZoYmcFDrGQLC@V8)M$%jn@NmJI`o2y%o`YupEGPY$h`rZ9zSJXo$wAbk+Gss)J1_e7M#@Vq{BXEFT6mU5fk*2Bx-A*E7qWN3g-)<; zYgq4|W!-R=oip<18(EecYXk+-(Q7uLYMpV1Fz}sg2hXws!`c8g>q9#q1z0wx7Q*mt zPA$2aYh6A~$G}<#pM4f&4mU6ue>}DW0Yf@rytpMDS^h-)iGmC~bre0pY%Z{eHu{+^ z+E?2-55qI5zaW6{N&g4Kt=En80UI+l$Tk7YDhNAp;Dc|xliTXWcqbQO07cA%uu%tO zv=&=cR!hV1``h~`EYw2%<8{{^hjR46Gi%00P(y=-tl&g+Z0FksyekZ;W%>14mF5T?Xe9GC=+fn&9s1^6dA7 zGTlsDtz`og0Y${P-y3E#bhfm9PUD6=lkS%YC%&&mN~deXn5aJ=DgR49MeKOmg@tif zkjik2{M(T-dMoKxXD8&TQgivyUXaN`l96hLFe+Y_l>P4=d4^JiUI#U*d|E3P&D}M`M1(x5%{F>1wBJPAa*i&ao_&0`v3!?=F7KnQ7-3riLf0m5modUs@R3`rJEb0EG@xR(MlXOR6I@o(~(u#r%A zThkxv>CshYV;zw(s?gMFNUfAYkDh_R+!7aNAd>trAue8Of4G>o`};K!OS}G-D-;yJ z5`*qmJ8@rgiJi5{l|>OGMSSpHkTVIL;;8G_=WcADIqAZ2iN8229;#R1G89x>SNm$zhYM$^AfyN*GO23beCF%Zf&? zyG4ttPLF!eT1if9*AZQ9HG9=$izX=!C?e(pLdQBEgogZ@W_NRAgK~>;k2~V9c~I$f z{J}LM5?TYgTW`B>t;5^f-TeNoW{GH)I$x)u&+i@mO44W8-+DMc&j+Q;@x!v;b~B#u zr#vyc?WS1Yho)Kl?$?0l_n=q&)(5Jzk9ACy8kNVh7`d7vwp;w3cP9m(yQidGO*y`I zr>f?7dKnaiKr>VhyF2d&i&GIJPyVSCGyhf~GY_tm)#5?GGQ#Xk4K^$bK5r3Q*1^Y8 zOVpAFhsa2i45OyTIz|Jb(#dRuUm4tveA($GXJaWLbB&jQk5#gvmstHXIY|HB0`)29 z(oV=zVrh&UOSTbeufail2gHJ#1Rp4uH6w6&Xr|cDM)Dii5Gm;C#@q%uVuHs`c}2b< ziGgefMEFTyhyDWB@LT?E@`S~#;aQWDCH@K;9N=ruK^6xX+~iav=xoNG^Hi zd~*SZr#rdE4~Qx+zog~ehL5|^Mo(ysY5@)2*7uSH5^0?RCStWTlbXOx0ov# zZ2ev=Suv(f8Eir2DxNxDu-JmGhAwsLtD0?TH~MJ>SB1=cc|WM(v(FNRF?86dFs zZ0sQXYe;ucxluJ1o-m1fpe{kY$f>p181{B~!e@lS$x-FV3g|x3<^ikA(h~7-@_!h1 zk}8LhZ%S{F{5D65wy5Mc0s6+Zg-Q2d5I?1rIX*9S9wgo2o!z zE)1wT4mgt!1&tT?^AIl^Jnt1tBwo?xK?|4X2q&&bhQtvoLpI!sC`#9dL6MI&KP_kD zcj|#aC&!UE%K9;P*_MOBq!_)HpH$|Dn|MVo%7=3~kIuoYXFzJBBb3$zj2e53br)EO zfaSAsxbDR5Bp-AFH!5OhrqvKHV#S_fWXA)yDZ9;`*R54_AZ4H6yS~#k7xZ6@PhB?3 z++2Wi*30$jKUX+KYh_1m!*J{656fNo=b)HPbX7i6@MEYs)0QW>HxxlNOl|h{D$IZj zY&E?RHz2SdBuZD4?in?IFmfZ*#zH%F|Gg&U3<%!S)fHI+?Y0S`uO zG#8n-WD(dUKW72FVmg{6_(%d(a(0nXJQx#`ex(5!9fqO;sJ17GsVMFu=w$&;blOLZ>a zHxzXNMS|8}vUz2Ip#ae8AJ+xo6&IO;YRn=C7&ivJE;+MdEmjo3(j`w(K}wCL@ti9J z6=_=+CxoUKSxX-wfcEqD1Q2;o_SiS6PL>gW9 z|6-VlCiU--M5@@7?~Vmcb#uT8qm!y+Ns%a751uha z9_IEu#ZwN8tM`u9vHMkw{Zde6#C85nsP?6=;wz`&`^4bdb^pqcM)UassCxaSTJIr$ zA6uILcFX@*t-Cwj)~m9P-SMH2@C{1I44c63`8MEn;tyg+;Dz;p#+ld7pe6MxnM5yz ztXX)>Sjl0l%EB5?n`QU`UnrR;${X>I<&R05v8#_rZH zp51DwgDATh{l!23m5Oslti%@3<(>yN>8bl#rp?vXds_as`=2861cW1=z<&wIWdAu{ z-rB<2!`{x;_}^11N7F`aa|G$pon&|cFv_LIamSK`h~7n&Ne^sNnr@jU-YUcnSC96N+@V;mt|2m6+ROrP zUJK3Lg3*w321?oVa_$)o?}g>ae1TJn+pKjAi2QlLKX_VT)9H@naLI{L?*g#2$c$^> zlNR{ElJ@YDrqUZQW%`~x#w1?fMBIzRAilCI5%rrj>VQFUO&~)>OOK?wok}ZVXzxmN zBsV*$TH30lSnIo$sk(v#woXO9{3%eEGiP1;ziHi|gCDyQ}w$@5}KURI@VDlM-POgjOjbtCK2r zMPH3+5y}Llj)@Buvwfz6n|-}q7krxDwU!sRPkZxd(|O5v*|lt#y(fDE1LKT3Jn*5A z)4BGbY7*)cofb`Z*EiJmzA&*~HA7z(!g}2za7_4|REPG}Z4_MxV201mGBkCR-h$=k z3Utb@I+m#M#e$#Xu|NX~gPLrDL?J|Vt9XYl5H`GBN>$nr)${?sbwrH?IwL$+PHfOK z_Mkx;hBCrbd7)KcKiGT^!Gt1bOU$mg055OuB*@7?J|RZ(nhMN=*1u^~%;F@8viNYD zVSol!tWq8|B4WdOQ1%@^ScY&ddYuMnZ5`B9Zx>Lu2_F-ihRN7w`VY`JpC8sJpAW~B zQUykj=Sl*Dkb6Zr|8;@T;J!k==$A5XVYs2rqkf*r+3CIwE$Y-%uqQP(dzX5yB)WHa zS8to|HzU2t?z`i}(vb_IDh z$ssy>JPoJXr-lR`!ZgK6SU|L2eCtlAH)tgA?%Pph?vIQuL<$yBav32c!2iy!Syk? z6X^i+E;|l_<`--wQjDEmpa*q^@v8&zOpy4|=^30p9570K1|F1!oT|UFHPG1o>?sdm z%=l{;yU=^Ygut)vEctaxryK)iNfE@AuHmkLNOu?NgN%aN$c+s`Q?=&;a}zT1?JDBY z?VmDD00n8Ap)`Yp-=h$@F71OT?=s@V2fbzg4*+vOjK9Uq7G;cW1{ca&*lGyAam1Kq z6>i4{&KTh;92;{m5-cKFQj#I64~Etg+|~Fph--p+q64~h1|PIO6o3%Mt>b?f_P2={ zf*W>fX-9K5eKh_3D*1O>50+L|!OYfbM*KF`%-ZH&JNGefTMKhf`EHxKwJDecW?$TI zPMgfNI}w*#bb8{?m50b4Nyi{*@fjc%B`>=p!`B7Q$lA;ujyg3upO89xV;}AzC{N_P zjLe!07$6gb!5E&Cvf0!(`rf{`_71+cAIt5IGzp8T64B;{=N@Gto6Q67*T;;;~R|)dKG>@lPzSNLytta7ll!`Wsf=laUMvp%<&BgTh~>-%n{Fr zquN>+{0k&>5W@9<7(+|+e$`lmpBKNfjtN(Ma;cbsGN_U0QrajIa$ee#(r`}QyA8=M zdoQNH9y}o&n^}u=hX#&GjnQVsq&97gs*OS4qaZ;!!$jPR$GW^@0zY8BDkr^pBi_Dk zPP++F5`rd8hcrCTwp*uid4vg~Wo#~Ej4S7QA|9HfF5(>wfeaYOw<%#J!j9DqWoDGI ze?*|OoX`*BCI!eW{X`_*G9XF_HF(msE}bA9O~MdL2pndKD2lDR{)Dg*4&&HEkUme< zT5rgSh1()BQ6VgsK0-x+6G14|(Uo`OeIZ4PWJdfn`f2o9{QT_WG#Y<5wiAs5BRSWzg0}YnMsSyHa$wIkEgmjQj zBIR%=q7E?qqPprAn`ml4mxLc$;N>g>dd5wdA0OE6OWVS)@&xo$>w_Q46k_zmFBa>} z`o)BzRG7M{mDsB`TCsG@iEguX5~OO$eF5q}WwCCf)Q(qX+^En83@oWhfglWzq2mrh zaVt5kMr}utA@dS0d=iQl7{t6HWvh|Eq&0~}&K|Vss#jNN7_aHnMw>D~3U($q+afga z#MAPL!CX<$koBM<<7Gy&5zo@pfJI_8q=pE2@U1aqK^s|A{HYol2jy2;t`uy5qZQ80 zCzn?^FH}3kBT{g1gkLopTIylLLfWBVb{v+X$nJwE{nmy(~>^3DUi zHfZaQZS6INDmbAnt6@;*I+7s+FRY1ew|rSsMuZrSGbv3$_ktG2Mw@xuUUSpht{1u= z=Vvju`vL?j=v=Ow{&xA#V0dXyryG*hjP}(nCEJXYVgSLa{2b_>^sXIzx=;={ZLq0;Z2mTRo|c3^}x0dwgBYYw|rODuv?V3 z+gY1lait4Yg7ts&Xw9l>B-)$I)ppWwi+hd$r)K)|q)n&#Xf$1ZFkIfbN7YkUts$4ban5&oy9iyu4}8_uW#k-h-Q5-X~h|p#<#b zh;Ob8c2a$>r#8UtVif}CeeYUntkFGgH>9p`DzB;6v1xHqtad2ZCAgp-(6fHAeT}i3 z#hNSbRO;sPJc`F>ji9`Ps`S~Pz#lzxhkv$QVtoU&=~bWg zqI~xuJwtyxeR$&Rik)h+VN3a3jgxG?q2FvQG!F!+2PcH(xnwDIo;h-RI=v;=yp3f9 zs`bb$`TB_3znM=BU;%YFu}Zl)io->!M>+6FC~IlenX&o)4`&{_?nl1$)Ao+?)<*gj zY<>6mY>ca2Bh(jQ6ufc*?pPuGMF2j1^)O(_1PQgk zPr&^fS)-wI@0)TDVXgs8Sth@wFx%uX*2{W9xGh-0>KDUNi%{=NgCW3k&Du$1L8`SvnkA}+u3^&wLF zu^2W=fx*FkZ5EVdO~r5z`|V-(=W=|~)hTz(n)>;}i_?tjW@;hj>55L3yo3BoQZwCL zyz^bkuKG_V+~23_N?*(@6;)CQEs+oJ<6A*4QbA+QE&tmXg$#E0)m@#I=@97fydr*r z#9YtG;x>T0O6ALd-5HpZpQm!M=r;8~FbaGa51+KHixxC4*mCwS8>(`N7eoKJhxYvy@ej|cGHhyg()9=wZB zUb+pZFOPR~H`LOnzHe{k4zwOd=5AnEld||Xm^P7mf&!1&iY*)!sN3iGl5;ldp2N(- z%%=h#=IK&rvqA$FwU(a#{|fG3%yX{zE7yj6N2l-DI4UOPg=4h=(5t!-8?7IP*NM(F z<(RwSW5a1~EyB>6%XmDOf5V6?M8ZIkx|hBxdb~LU@J6MrF|}!PJFZS@lLxjn8~|df z4l=BJ$Tr9n|Ftw|Io2InLj9P{zda2aPN938kDdPkP)h>@6aWAK2mk;8LQqK*c+W$g z0087?0st2P003}#XKiI}baO9kWq4(Bb1!UZb8s$jdDOk_b{xl%DEMDbQ6y|P0HPYC z{4q8pvMtNI_-yW4EbX6UDNd0TeyQ}sJS{FwOn0b#zN zl-auNvM(U@T~-&(W?yW|e#!|hmhJAYt2WoeMfRx)W*pGjJ`sF7nGXFN`KHKh>*I9qp48<)* zcWvKSi@KB&d=9;}-(j!y^p*9#u1!f@CK$I>0|l#^?3-VkS!8O~z<7j?lzyBLY~R%g zlH%Rm+)Q~0TONyJahB?S7mPM1bO?{U& z?EqcW^i@Alo8M*aI%{Yan|zD?09^vZqQ7qU^(tGG*`{q*S=qGv%{6_jX>Yh5Scq5B z5#8i+e)H&OKmWz)le6=SkDtGo{rquX z>WcGYNmgZDZE-!*X@2IufoY%xaj-~&6Z*yCORc{tm+hvhet=~&Gb<5~WEkGrrW{^T z#rs$LX4kdLvhT~)abMQ!vkb?S%x4-K*w4+XyglQ#I>QdwmpDCOGQlcvG94io5W;`e zI+G2cr=@-RCtSl}RA%rMAM>w8R~0aKeF8$n6~SM9Bqm(C#n^Z(#`$B!}-Va%9xOq?i$v(^}cz1$7>!m|1#_LwM=Q5mQz@+ zOPDG}2mjZ#KmRF7R`ZVAUu3_YJ}rO!>-%TFI?JB@>ipR+&$9g4lVAViEPL|o_rE^N zjuy|JJv|c=NA1z)RkNNx&6}c`KCPy*F~EPts@M&%G}&DM%Cjx%Y||CnZP8_$>P^`Q z8R#3e-gdv~c$I~fyQt48x?8Q0+u_mmtLcZ+XZWe&HPlCFtjcniExNLR4QM)@nr4Ds z;>N5&J#6wd?pJGA*!hD?20)ro60=5)vkmC=}vfn}% zJca*%qb42{o{~I&o_**Kn^m=`23|#L{&t*d z{!fR=wqK2tefIPK-QOLcdklnEf4n-Gjg$V(A<{*@{v`j?;ey}(8x`Cpfh89coMr6} zM;r{>@)qV+14|*!N!*j$ZfcY7I^Z&h+}dqA`0v>J=v(^gJE~FjdHEUC$-~p`LI{&o zp-*hsZVX(~mo%!gJ?zLEB6OS~%V@|w$&N3lR}?%AdV>GKRxbD_ z5%%PRL^-~4qh;4NnsGQ-nDVp!<@!&nKh|GdWtUI#t64Vo(b>VTZVTYw`o4e#Xf<5F z*SF{Qxci&16!&I0(IzNf<_t{$_`0l|nSODxluo3Z$roHH%_9Z>3 z-)rn+`)6;N+B=?Jn+98Ug;V}6n6KPhQdk5@db4aiJWGIogxP!h>OB?@{Kia5m>ruk z-pkiveaxz$c$_CY{=J0!@iS5)n5m{P$Ftj$@dELx=l9WDQ=jzGg$?v&35ygQ^MJvG zlH)FVUA%#PoDXos>P)j|PmqVey|wT%%i^|f>$<%GIt$=?S~x>pBWsMD)EON{esY#APIfLke0q*FAnd#vp@ z%{MQza#g_~>Me8f#}xD?Er^>6?j226!46c^|C+0zptKxZH17NEj) zTVTR`2SlL-;8amSLGl1D=SFvWkg51;xsVgXwiuSzKEpAYPVggynUY}WI5-M$L*h{u z8i++Y(?8*tiMu1AQcolI7{=^94BeCIQ?;3g9B_)<=gk1M?QA;ckhx6#04x9T$(hP{ z<`>sAd7hgPu*BKKobB6om}Ed@D4Ch-N&ZkYces>yuck4pm#Yf0!$Eu3?t0iYm#?Rm z&F1nD=SZTKs=oWWZB}2fsD>&2?HW#E1_SdBINt`(qcA{@&T#lq^RYOG<`J|vi529G zv20N!2!*kuZrFDX4eN%PdLwUHW7V)ut`gJiZ)HY98qdiz(v6hhE`w^@!R-|tPNNGJ zQ6Wl+sis^TC^Vc7R)!QNf~v%1Hf`D4KGB_VhG+lfH6j|Siax2zHVBD)MYEt2XUdxTG%Sn1S z)Z1rKyU9n>)7j4^Rti0R30dp%@+ zc)*Ve{s^J2@KDl#wiDg;S)T3teNl^W5@uofriidfp% ze?XDYWLB-AtHDRHW|lyNp($@=}Y?BS@sx<^*G8mIl7nx8AY}? zh0|RPWn#@xx+S{DNAvmde2u8fBfv{zsP3Kf#JBnEb<>d$^RQ2b50)H z@_Y*QRHcdLRRimMxE?_{~`%Te>3((`^*~iZ( z^W&4S3q29;cpqT)qo0sHD5sGg!+g$hB9~?nLTO?Rd*l)sI+*UJ{^4e!bUnaWoWuX8 zXQGRMtr;5tndw^S!uA^EUzFJjnHP}(4%4Xfp@LAqaG-j!@A3K}!$M1HS8j`{!QM*O zMccDw zK%vcPw{YK%T2D4NPtiNZct~m69dVPwIh=7sXCbyUyDRa)!vb8Jwr|57qeN8+@p)M7 zLh<{=^`;nosaHS=Sm?T?Ot4Pi#u(QXbwE*NvF|Q^ImRBvj77*Q9pp=ncF?!+tenj6 zk4AH(rt#$RyZfW7C#gJ%w2<_Hl)8!mJCmqi!~u;}%e1ggIZ~bDlS@eZ#N#FomyX$^ zxOi(4i2#qOBnJn};4#p=WybvP*XH=a$`X^1$2Of!%+UqLCF*2@;pXqmG{`=TClfFQ z8L(Z&4jIw!_b`S_*!#cv%{MReez>d4XbjKiw@QBg1RYH`T`vEhVG6?W<-+r zD_(=wz@OQ<_s{#PF5m03Z$EtZ;CH{GiElRAu#WEV$?_ulcH*mh{ncMR^i|68SDocI z4;S>auxO}|LzWdS9)NpAR$^6jtLlexHPu?*aM{gzI@dU6O^)8U1Y;ZbaCE*-Yu~2< ztuOy@VP$Y^E|6%O>4R1KmiDziV!AKCo6jH9+~fRuZcyl*HAV@{XAT*pa8fe0GFB-% zfwAjQ$s*&9{^I-`%}#POni(RZ|QlTX7uEo?m8-oE@{WIGY#>>1OJy1+Om!;*Z%O=&tWip`RY%9{qoaq zKmY2>m&U-({CoNJXPR&$l z_rHDh%^v~!$Im|f;*)PaNfI(&`u5+y{tS>N3K?F${OqMoiglVwDPI|JJTzBSP(Oz( zt9UUER0R0rxAJ|-A^Tkz&IOi)j#KvucRB0ak)ER%_AseChv;;~Gp8q0) z9sGxUd(|Y6Q6aqX_@B2(`e`5M5&l1=-l_TXKf3yQD*liDwl}-q|KrndN%!<$Spr!! z-ggvyiphU+2_CRF`{^2mMZ3ED?|jn5{rPygo@ReGn|Ai$^jFj0PM`fQlCA&$>l8dc zR^-0!N?dOb9q~^_l8+vUy_!jmk8vTugK~C0y&TlPd^9~x>Tv=}|K*CD zSklw3SkOaBvg28-wwbR^|2-%!ZyKY*_lG3y$H+|2=ko^G%E$Bh)ye7O)Cr1dq5wEN zBOZ5;FHj8Ii`3l{!pD-^i=UME=qDte{G_y=>^s2z{h(w#urrIseE#HWHeWrNPiH8- zbw`Y|z`P(aT_o!{{GZM{lIIx4#e536jxS+VtotiN>11|~P{f^{%t9o5!(X-&g(cI9 z#OQ(%wL=}=`;9YHnoJ)DRZ{Aaq?RSopOQ8x%$srvJrz$DT-w~-AiGZlxfj*Ibx6&& z`6*hso=)sVp(%-gVk-NlLg6+l1=O)sQzNa@*!T>n!SQdE4{)HTW`!=0Scw~D5m*~B z`OE3q)#RW?#~h6MYdFPS8zT#^!jnm$V@NeuQ*?J3^GZcr%>`MiVXHkkKBu3P*+i={ zW0LK1fcZ2^u@#E>KQT%C)il3aCI;wrc{@(=)=)%|nh@onru*meHqkR=&O1sOrd6ki z|KN!K;1N&h0m+;#q9kh>{&oVdt#GCsbGt}E_hvQW|@qU;Uw_x%9Oo;@V% zqV4(ksQ<6|!mbF6OtGnVGv2&l-;!PHfhpRR>#{>>XsXwqPY&z|l;Y^%>~b@9j5~PJ zRJFgVLj-x4(-9slw#Yc}&1O}N3`rc02a6u*jagzRGF=kaE=sgrLM`1YD{B01Jfe(R z`Mth1uJs=7p`A{m?N6(~$JaLAhDLOq(AKAJ4d%u_odkN%=|9))qKH@qvu=B7Qv=m1 zC6kP{ucQkPE+*wkQk|DL;=FV6nF6~0t}kDqd?dU<_4x5)7+t#(Eyc)w=3gr`V#&Vl z+Pd9juqB{C*@(2cCiZeoc8UhQ_1!6p_U)a7>=>I`uIR1nYDpr|nEZ(%Iyr`-?wf2P z<^^bY=o?oZ*{e4q34q~-s#pLJ>s?MaQxuOEduU>3!VD+cuAKH|(Jil`1;3+7Mc=-yKjC zFcO9T*LNej0gFg#1OOY|_p~`8<-mv2ry@lTrKQe9dxEPE5(O0gqi*dWw0KjNRNr@sw{^M>^Q6Zpvy<|uu?FZzniVz|^{Fh{Fu zE|*2&A#w8Y$GG84Uc7j&$w32_#qye40(w>BkoC)9%Ha`mM>v6Dn^_>Gw_;mQysSD5 z?~a|(Pb0s36CgRDglWuETakp0I*!uy<8 z;}g?<{~x|Z{rK{`Y(C7Jc{g9r-;njg+jk$$A5Z@NdS4lggq*>&iW6S0>6(0>8IO~ zUaH2u`VX0uc9Wci5hzlK4)v*|lZWLLPx%b?kLU~*rMvyNt4p?8Afw0!&Q_tl=zGYn z=JWgc95w$}56iC}G#QQ4&*#TeGV6>CI^)`YcMM#|$?;|P&Hk@fZ_hq_2fO|8>|%O9 zZ%$5U5b@4I?7~C{{;^%2#2=3c(mg3d5F%Klg9L;2HOiH+`mK%S|z6}CWPvKxt+2_(<`e+ z-j+pUP2@?=mNL7gjhtr71e%5n3fYdrH3}Zwdgs!?^7s2qeV17NZpy0TyFS~tAK1nU za=mQN=%P7EE4p@Dp_7af_CUK{9_kdTTEf;kG~o0*NV72~Nn7oQ#)5UJbKol`&55t! zOq;(tvs!!bjLxy|VZi&}#c($!3MY|bxh!|E5wfHd_nf2<@pyiD4@>AIkjoq(xGi=E zsJ*?Pzk86{m_2bs?f8Cva)6rFBYY}ZR_ess5$2G68>b)+`t>k!Zbb|kwLxHIL-nDQoXvMRNDETQ5ohAf@ol@>*JVzYZTj?Z^z&T} z6Fs*CIqD8v26BJ^?9_Da)=$Ey_6?j8IMS8h@2jr#GXu+2&>3QtlQfD_4(-kca#Cxs zcLzVKS(tJwSJj9AQ~BBQ?@E3oT4_@G3xZCkX!|+2ShYEq|2Z9zpA*4X)W?e+b*CH} zuuz%|Y|Odb%gEb3Cj$_gwb?k!DT;<=AbYZPQ^8Y@9>a3-4Wkk<;8Y-;w66pS9CZ=_ zP>P%Ltu&kHDT_$@(){KxTxjbFP7vgVg*}>6;M9hoaAm#9ae3~0U+SUp+x@^o5dohs zU2p}*;5Q{T6zuW~)LQ$sF;Hn#JY}I=0#8N~-?e)H?3-$c0OwV+F8y+e^a#(W=c}az zrF&i2eNN)6b_ggDiVM$U!G-dmV4gT6!kcm`?C5t*ihV!8Si3Iz>zr09ottqztdfF? zvirkBfOQKDXTHUI(p`>=rH9*W{H7EHT}RJh34(6%>fq6dv&s5_ZBTS~x$Xr?Ld!KC zKl7qlUbm@_x4VHylf;b)dR(_21{&-UKhPo~13DrXCpy(6s!da%>M51e;8=a4GD!yp zfOA*hm=>o{Q^O#lN)aH(o02Iyg!M*-nI$TkFHp*kQr?IFp=n8z`#?d~LzN(WQ{Lpv zh~NM*B2|?@myOJ_QoX_m8nO`;g)^-)k{{;S8{n(LCWw*oqQObe)e`PQ+^X&1yi`Xt z;$hD2!1A3)BLzsaDv=*=R0M@Ukl7dlkSSP12Db4xBDXY<{Q_8;2ry1zzX;0F?-z~{ z$#-opb&9g46P~*Tr!MYn;JjLegBzh<;&YVcm^P*OU9)tt;Tm;_6mLBtcg>kq7gjY*yYtT^Ci2b02$`B*}%4rU0){v=`a{FMLMq>{W1G7X$2>UAgoY zMFK}a-d5q~)cl?k-{Fb*0cKi+Vpbu-v9?}$5x4RLRB*GgDKS!hA4K2SI4qv-6R7@( zs3EYoNsvjvRktsP6b!!<&bOl*1;obo7XJKt*Wln4{IBtYqt=dl%-zZA9*- zEM9wcFaY%t2Qps*UTaa-0WsJ>hr%RqOFzbj5>5^>e<6NeSe2ORX^)cE;kv(r?q-U6Hz$DhQ?_$m-}Lij2iny{`HZ*tH8G4^7Koy1&byeydTGy>%YfK9D?B08>5xy<0sMgypq_Z& zLW}}b3;t;9Dk*objrge31ea@(aWQP?>G&9lsMhC- zo;!=7tdc<0E%P~i67$Fr!9lwXs_+g>&2Yr>}{V)$PZ_7E`->tiVKuIQj~VD?u(lwsWKc#X*M>f~#c99HK1 z1(zN+XkJq&;FCbexjIt;py6@D&Iesc+D0P8iz+&fi0}yK zPNxYGNS$I!zFOX^4Bpwr9Tzy{*UvjT`0-N#n68u{fwh5)qef7-Y^msZ&Biy@+C|J@ zqJyA@1Lv@B#JnOD_W)$y^|-&K#*sT#EDMpkvJLz;O~*DGu%YgF6B8V8dtNwt&5wS11RRr7a#n zB4{zfQx!c`@oGh)m|lS-9#S$R7v?DaT(b=r{{EZ;d)`keK;J;qTTN)+kXdV2kPLd& zq6&`%x`rW1if>qM;}aT_)Ho51F_BYdXPSUAKa;(A-?$lT!d?QAyQ~5cfw9RWdpfz% zM|u|MuV2_agPYY&|d7+)q*$3+z=`*ihZ+wF_gt;R(^U z-a}`i6+DS2ie*bpRjgL|q8x5Y?jVRhqu;vdwt4v`(2k%16hlc7A|+{7+hWI|4I8Kx z&7I(OhOOt2uN@&!ZZexSt>`ve?hg#{@(T?6a1H$i`%Qmct^HyF?hcMIz(~oF)raK9 z$5H^F#YE^~7}SH^bXd5~WioDIXBPB4$ltX$Zk#fYn*`^g_s*eqp#&wBe0g0!Nm}4% z=^pzkkW{C$Ks9m_59F|2n0iDB{TlSr)P28}UtY~1@V)%fQ5$%9Uo%y#ayA^x+clom z_BC@Di^A-d3sOemawcjb0d&o#_Xj*9sFh_{N``8(-|t} zZ%tIJa!vHsXHFM2?&df}mPMzRTZLGSAq+4@Zd>&|c{wh+;m*(3z={MT<8b~CW?Gs4 zE^I8Qjvvn-<$J63JG|HpSfHOc@<_GSBA7*wL$K178BXEedQrl>;JB;<0U6HFwy-8N&vit?sLgX`vL=PcM?e(0hDU;e zEm!1R5kLlWdus!J+u*??2PP7d@;E1eI-Jl7KXWJ}1y?j+S6Zo6!_IgV$CqZjaU!Fw zM15kU(lcpBvy<#Pfe%R-=^@8(3-V)l1bKrE40}43B&J|}j7i)H@)mcvtC8cNT^RS9 z77=voXlBy|*=#P3+dv=E>GK!1M{`MGF}AiGY3IAhtosbI8{ z+RYk;Gj zgazFX3!* z-F72iw&-)-v7Ow|rE$Vqgarb909eL2@A^;uxE-F;^cnnXj-r=J()jbc4jdt#0TKu9PJQBKj=e@Qqn=&b=WHL zkuV89jo>0fPLp$HO(9GeGmVIK#&A!uBtwM z?h-*vM?B06KNgc(!I8+?NFo4L02HM zZ5uAoCc9 zLr8Zc&1O)&=DN(mRA_(0^9o}=w)9w0Ynyp#-(uRwFaTrZMBT& zC0Vz=@KnbPGvXYM^(8AnIJR#%9?;>6;5o7{?U{$!Y~BuyIF@0uS20kLYD+l!CQfsd za}3$1NF*`}=zN_3rf7HvK9$FV^Jdr9H7chzMYkw6{-|RVF1EzE*tT(S2SO`hzP2Or zUC0M82$mI?S;VeGhlmNc%3`D*cr@(G?hP;Y_MjcpE1}1+YGTl8eNAYZu|h#GIM%Es z9O|R&ZR;;ynx6OlOIt8()O;d&}TQBoi(L-h|7=5-s3^ zZ3=g-1%>*1YkZ+nL_3;~TJbK4E((I22bU<7ibqzt35$WvDBCQRU1r~bRY0Gj?diT5{d4h_|+8ssr5d* z|6wqjWxlOy7~g1Avhp(F%vTrNJgzP#1S0_7WFDpVP;NWQ9I&ny9pACBj1$`HCC5P` zEwT_^5zs|oGB3IgB=h(_7(#@x_A>{^4kkU{_oW@yJ{^{7Ye;x6$zZnS7DwQ|_tN{# zn#jfMl-#d$dr}%5E1ZIa4J)krwtnMJ0>P7XpDe&f+SFcx(wV(8fCC?tCYm%gDPX;9 z%;{PIZN4n)I$p81fv+E!*ZrAfio1xGihzEc>UgT<0|Rg`#Z_BTaK4B_neG_Jqi z4=A6(Gn|KsX>ECw=4hEj5?6OgbtFZK1~+md>lh!cupT|5US1g+O$xx|Zdx9R4u+!n z{D-!2YY)LAa`{SbNm00GzK9~%uqfHc7Pfx(Z5XYMFPb5!Uj|i#cD)u6BOz|e^0oKf zyV92kTcA#5O19tTFVq!A0m%Z#YYqVr>qT$Hq- zAt}S0QmllwxO3#_f~0tKBfxdhM@MTy*r%i97#{!~N$i%Sv7N=zA;?@W@}Od8?!lVC z(8NR9)rg>w-M-Ugw|KcK0J@Rc8~F_ccXXXnQzH(u?7D87{g%sQZnMI#c^1l;wkslt?a*_MNxQZI1>bLXW+TC;4gIFs?*b_)LIVD1 zbq7l+TIQeXn7Tft=m}EaT`(|f(eD9HQ`wa{cK`}+-V3%5tRx&<+hrT*O~tEqiETvl zot6SkfB^PByiKA`%x3jac~a*uYnGA55`X$I5t)W*B+sh{NYJ<{+qL`{ogfCev;3;; z(76l{ajS|2F(umlncDZcmO^-BzW1dS5GTFcvgP|G<@W;Kqe5tuh{mc60jWK0Ah;9b z0ZbtSQXl{Ph0Wlo&Ccoe0>|HkwBK+-zG|iD(Sh2`b89?c5Mp?cFe;FfIB`%$V(u(z zG?tG@baD$hcb)8}TIsyJHj1&5SFo^gQEv2)YS6ZD1aouDI{1Vk!!tTm9`lL&g&e&p z1woY3ngOuHF)P4PQCTKvHEm8_WF>;mLCJz#srKyaW7adMHLTpvhjjzRBjy_@! zlk?L)&u!M)5DV|0n0%zXW*<0A%~}LjtHWdtSHdp?of7W)q1+}jIrIgrT}?ZZ zmyx+7c@H{cg}7ke??D=lLrkHKZFYS7^WoHaXsY^fmeF#EzqE&_NwP6A>F#XG;im0g zdwB$QG2RbeiahR84}b8(ererhvjCK)0X!_;l5ld6G{*l%C}Q&vdEYHz8}6eV7ly40 zf$5SF7UwORVZ`|iCN5c;oZ6^;5b3njBdR>F`0s!bU~@u^*VVC zbD_L7XBCd>p`ieh%p<3;a9l-Sg2oY-Ak$E6uf%BB)u^&F)+G?Po`1T=I0zFq-1z?C z-OInDMDBm60rkk}{F0v3s zHQw!&V~~S@A?6gmaTEf#{xjlw#LN}2uSTX_1{FZD!n#9jWD4_B-S6wRi4mf1) zKT3>lq>3^1lQn;fCt0~?(~0e$ zl8VXcQ4`ts#`>HRw6*OKj@2e$oDo?GFR80NVq&3yp?xO}>!cj(t}KJ=X(}%w$q7|- zgN;ze;nZOQ&`3?s4u4}Z)&|}Qu@wyFx!=4Fg%;g6MKy$E5axq8MWmu0!u23#-?at>{r}L;*P-@gh||l`@5}AWr%JtD41Km5x7&nZaos`a$7EK zPvsIkWKshu%Z6_1Zs;8Yy`D)aUM8j4^Mb8q;`^ctsX~EC(}+78iTuq9Wh45tBS5VumfnobH9A-|Mgu z%t`dkEFz|7G69^fMgW0wCCDErSwh$oEA{j+q58!`?ux35a`<$N`h_Ba`h{R55x<7z zN@Q9D6xpu{KT?rU1*5<&nW$h)*h8Fpkw+0Nr-D&YN5NgXq3Df~B9S+>Wk}7U*zuck zgXdPelsnZJE5`UiR$hP)6Cm$3c?~SwQiRGupGIkc@f8_td`wb_MMFt!AeluI-}S&3 z?7m;c?LMUaF6UDeoIViMF;d3jb;z4gzd;%=-fdOlDn zAIRo?D?mN)RWq3yzCc~BY|NZ$l6|4jD1I{E)}3QcR`|q0$S5Ezu6Cl*FCwfv`~|zna*{2FvSO5(2agG%;Z)3VSXN2BJovs@E4)*Q z)8TzHn2UM`()i`>^Mb2|0q#fRa!(22A}@Rb_}C*Rz?n+W=hnID>#DAmF{yF3jK9LN zJ4*GECa&w|x?0d(RrRa{SA{*cYDr+Y9@;xn6kr7bUEgrhN{)_rc8-wDG$+7s96U)Q z$Q$Lc&CT)nwlTx?3gAOD^$}0b@x>K_uuYST%Tb5I`cp%2pNN3&Y~p;tVaSpzd4(6B zSr*)S?nl^goPT}MB|%^SuS&XGq6{QFbNrB)6WK-7&X+cOV4|-{;K9U6_I3PxS?tKY zXkD#+4JN13=Y^0$jC+7uDZ3X`eA$}AN!?n%!{~XTF5i@OZy*#4g%K7pGs(8aEki88 zSV?dcc(W3k&tEe6o`SXMWC9#AOaxV4 z4op0{Jsf5kJ1@o9^fAV!GlE+N=sQL}By`f?uv^P2gfRCcj>ZJ5OMnXUn7L^qr%_&U zxa{s-xb)iW*oP#@v>wMqu&8^3Jfl~o7^}~747qJ$8rqV_hd8612%~Is67-DejR#2D z-=tGGBb*U@bdnrNH)bp|l6>48W+cNwGnbJ(&W0sSk~T3JOXKWJ##2N#BBM#imFS~c ztkFi8##&sA5edvMLUe2*72rdo2p=7qKg3vL>X5)Vi-s`*?%G|k71F0g7JLKq2m~EMKdviQB}4F=r0X;SXcj z-6y$*Aw_h^@>43h?p205WI^>R(vs)cI(7#sXvYYtJe!S*(h~&ZtVYmfed#_9b5%TK zdVahE0>L!^g2j6^zXN+{gWkco#}e3s4@0Y6=|#lX7r)UQuf3_!uWn;|# z5^n$s7JG`E)tmPj1>XK+yipHawj{i{MTtI)Jm3&oUkYY9z*ySrVvBEk#jq!`(T&)Z zZ;vI&_yVSY6F$}iGd%J?XAtDaKIpf;TGS+07@b1Wg>-R6ylyg_ESg+25{?gL0rNk5zaP zIi%z+6FV~A6<7hr2X+F}Nn8{<4v<>EH6q31ds0OoqE>SSix=`6l)|vu+jOQ>fL^~u zjW3Sap$u14K%+ZV8qXw3tncL1#}yNIep_*qw>3U?*b2d;<-g(L-|z^KY;zF+X92af zr1@Tcc}p$3jch+0CGj4K*PQ?l;J$z{T*)(eI0i+mI5>I@;&%r81yWG>sPn#{067zy zC6P2SmLoR#3*VTt!CUm6(MkBOEDX}|({`NL6Rw2l*+|}DC}5p%tvZ5|zK>3=E;-3} zXo&k{w&c@X>JX4pYZ1mjwjmo}MP1$YBKbh62@PqFmI30tl+&J< zxc6g^f@rwxIu!f23DOJjGzoo>R^&5Wr4WvWAffEi0#lV#o$8{GqU-|qzGJj%4#cPF z)Il(`S@9$(+-DX4ov!&;{M?gkugH%uj}6xl6je$r4-onn_mOhGiGb(gJlsRfVmd2a zflM0^MB9`z1XcGgW^TYIh=~G`ebag*`noK4E^2JvDHaoA=~Aw%ARuEm>l^6m?7XAm z`a!o6BsD|%o&tGYinB0MM8nEF*kMJWUB6(Kd? z$wCZ$Df{1I&@3W^OL{es_?4yH0^O6hyvD2!0Wjq<*rky=%xunt3kHIdfHigZ%3=2n ziehb!hFmdNZwm&RYVwq^zd+j!dJ~S;CPCsUuJ=-@s?gq?a2Y|!K?F-AK6ZI{_YR9e z&|aPhiOVi%S=uZy39yF+j+o0pzDc);S5cPoB9OQ_)J#PsQE)TFX+nA>6gPL!)Qdwq z^=yzx(Gu=?1mkY=Xn|W8&!^_NgR*DTjx7q`Ov+1LEMVefX> zK@nzW;}6%C2qUo)4%+$-c!y#cY!=*AaT0?PjfkB);l)+c%*G;L;w)XTZfzL4Z;)C% zoDs*zQe~q_VD}cBjmpJibG0v^Q_F+V=2?Rm?h#W5j1K4l${ua)3(`sK>;Q?GpHW7T zVeWdqF_p?KuP;C^c7~C$IAb1}&I!x>ep%{q7+|bD=6LnptIir0j&FmJy(hSmtnhAI z@!P}2`d$T3BQgLxw{mEKL;$+vu{X2K?aE;bZ}I6$x;VNr1t3tjZyI~9VwmlCipT?N z5iyfv$96mVSiwO3l4y`WM2nwf7QrO&?pu)!5MLlkK7Apm;IR*f87rH) zfm8Rz2$Oo1eiVgUOdA~3(kW+(O02LKRO-@(AYz?6BkcVV)>31$g>F$tj1Zh?c09=& zbCKdfYFxj)HdX2cB)6V%4p+e&3o4(4O-MQ0 zVs7hMGRRM!D&?5ivddh88v_Fe5PgD;B%9h@_An0<2%?8!)|FJy*X1x?pCDaSCCA^X zXgpE^v9W^rpy*jF1wOYsmPsE&Sox2t??4w~kuynJID;ne}*+Iw+81kfq(E-o|j zVJfqLtFUEjdxzcLEQD>Ok%wV4i(tVcrd3Xy#36T4Ovg_{Ks;PD&R%i!z-(|jl800Q z_Qlv_UhBN6KTO3YJAKb|`nx|d!+CI1^FXF^c@W!A@^F<@ZN^ZCN|E}k-#53oR=i%6 z*Hw~=+V&_eDzzArYHp){8zIswP-I9Xa1`+dgHp<`JV6;q%g;6@4-*z0aM3);9JR#2H38)j;` zKuKp%DLO_!=0hT`Lwk(UWcaM(R7VctPBj*WshR4<3>q9rZi*|A$ZZn& zo%=TjYU@=1Fm8kr1oXRY5%AQ5#;R6}3hQ9aG%Q3ez1nzq+0`Ieq-0{!)zZ%XVKVJJ zB}`=~=>8h~t{SkHRPa$i503+n4BmQ)vJr<8lQk2LT|_ZPWN;X94x%Xby$TxlP#k+R z3+dSRG$^7xMPWRAyTgY?w8)657ss<6z*f%oCj^(O(rA+$H5l6z+v4l@HZblf_@m6R zg3^A>EZaJ!ko^$SB9oH^9G;x~rMDUS{lc!$Iv$m%zXQ%-KnHHCYHhu(jpOVW_D!I^ zut1x{dXs{P?W^YoVt(0>HNJ7)quY93)%y*JiNf0q{dH-L(z(#r9bW3EMgPX%OZ)5n zdd;%Jo<&*PlJHh?ysmc2&m0s0S34OgY{j7HIeVepw`W2e zZ}Y(|j6CvX8YGzMOyUjG=dFoj$^BOx4a~ha!jRz65(BpFgRymr*4iibeJ4L0wKqsO zl(ORz3@c$(QgZ~9sCJbqp!hX9gV+@+9XAv@!!6ckhr!sQT+1au@tGPq(-XM2^Es&7 z?*!m>I|jsi9mCa9$pE$W8Gu<4CEohawpg%QD;ea}X#()eD7#ri$?Rz+=6!vTOU@jX zsZHToK*|hhF@B?|dyZ^ylRrx!N=b9nG}6XbX_TGcDs~#V=tj;E4@@s zZ2_p$hx6IPiL1Mq+P-0bE7opZXLG93e3XJW!W`gKL6(&~GVkJvP6BX{>O~m$qOjq> z5u{I2GOU7&K2mXfziwIdsa{Uq>CYSE_fitU)TlK*oCzcJ%o|V=g_&T7s>C;5Pl?xz zkruL8uVF0oMFM)Du99NETgpua5B_mX?D=?^yiYf6-t6n%2N~n@X1z>M(fHL!zSj?B ze4WlorvAM`iG@-`OL>wJh+Bd4_~313YzU_O?+I9bwa!b3IGWzg>Mcbgy<^$#sU z(a`axSZefWyf?FS-_8t1CLu?JTe6Qn`>|YHp~q(4Dvh}Tpg=o*no2?NQOkW;9)q+JO$#!G zv@!$;I_`hJ*_3_o42iiBsgg0ou-`KUb9}dg275Zeo-4cxax>jEOY<8@Uo{y(;ziGv z?_%h9EV5-&gBSE{a6p9+_E5v|f-OfSrVi7_L&9Rei8pIOR=uCSRN+u5ZNTKDZc*81 zXv$t^BJ2`2+UP>6pTgNnd2o5HCT%$Dr6fb8>jGGBnLHY)TyvvO{n#V7pIYCH^)Tpg zFXb6LPNYkm!Y7uKku*8lgO6N~j`PSRh4e@59p*z6#B1$)G`aYKJLEDwl|FRQF_zj~ z45#b#W8P3xSrR@DQpuz>0@6X+P>19p+2ewBiRj_do({YuR-~^V=#on&vxs0U28^)h zo?hC#8P!#rsxgvxhB+BpL{2+v3P8xB!`AdBxGWT=lJ(R$E_Ud_NaGm-D_oaa(a?Jm zt{1up{WT5lWP2y1e;T2zMGf3Y1?AbSqH0oumf&`eU=Z4eNQ~z3r+b7g&g+s?*K@dSV_DrfJzPP&_f@qQ; z`_>JqF;p}b2(XvQMw}n@W3=4xDBZ20Ui$d;@`Y5aA#zGt*f$0pHL^;SM z@K7Hh09cca9AEV42y-wX87)zM13wKDnh~dAqlYJzh*wgbyO$Ad!lybC&Ls;^l7TTC zL{EeGD7~SLR$qflA-E@-z(5-Z@*eghU?6aVz^zhU_}n-0F5I4+`62TPi&DGwDd+{1 zQ-t{xOS`E&n8&!l znx#v~;JAVK?x#}dx|=Wt7b@co{D4Q#2a-HWky9gzAx_C^0NqBs7b<{%5DwSM1DAKPo;~ z6w*l~PEU{)q>+?6Zts(qZbhv3EGHgG@k=Uo`{!4djYuRZ2+}{G--aE7W~um4^508 z3r+4vqVXf;`;n8qnGpKq1NwP|tS*${75mM;F8iv_o@d?U^7!)Y6(@a#hlpgFllj#< zlW0}4ukns?$0wKbtGDkSpUo$K|7hMzB9@|Ig7xHhb}_%d{Oz=6}0F zziZmN0aVv!Io{nP_-*)wcOfM>HB_6XTqVkF z2qlRSm7#rIHrQ})-NeFs8^IZ>Ci*h zbXGYl#ZGMcD>b1R7kwD1^(RLk-5))=%+9ZlFTb15=gk#NrRfPSLZ7lxZCr?tAZ0V} z=C{B7^?d)~(+_`3e}8wSGEHBSxp18KXnwC!{RI!ws7`~bEXZ?`akvgSy{t#s->%Wu z6B$#*_1K@x`zJuyC5j1rNj8rCMfNXcxy!011CDi-eM9#gGrnS*;b6{qjVPfPZmz54 zbwFlWG})>IX0ptnvobs?Wo<`c6xmL*Z=nmzYz55CkOBLbHDxi>cL5bRz_r=ecN;R< z=g~sh`;RiTq|E-YUl9&ISJ!=p-ZiITg^pJE;ii=W2d#obwmNy6T>svgh3t6t{L)`I zPbPk!NwZBq=SKjcqmFjKqenga@aK;4d;T(b%s^M5Yjb!kd`?nIeQ!SS?`3 z{;Ck+f%NSD;WUTJuSyyQ>6CYj~tcP{AtcL8UN>=iiwyd|_su$M4tY4{;E8KlR z<@Z7SV2z|l{1G!&NthTJ-`s=60jfCgWrA=zOAU{M#as4@!44Iw(y`kOhW9DyXyJ|) zsbcJ*r7z3s>ST7bcq^R9J4f>mRdax1`vAo@LGi2Qp)w}txT1KQp!nzV_5j5y1HP=N zZ;HDH`0Xyli~yTS)y;YHpUz3X(*a(*;>3=}!-^pA8m z`z$+VnrE3Ysy|6=E+2nQz?-$B@s{4*@3t9CiMj=&~0 zKi*%?S4F=5B>&UZ+h=FLdUyX<5AfS}CywB!P?d3N&}6X90?Es#`9EE~{Uz`pMgGIp zll$pUAORFEciq)3w9WB+K7$=vH_|tF$9BBgKNX2o+|tLWH@GLi)c2Rex9dOgj48I| z)MOeb`!9UIATwvaj6`qQaXH;gv!e^SesB_vp`!~ua)iqCk4}U*{>g86o3_tV#;m{* z>SQzEX-Dp(uc> zl%l6gN^;cqWW>MTH_Kt~56dLcfE5qPi^II*i5(w7*Kr%Cw=fxQK}U^tPt;(4SB>v? zt8qOuo_VcO{YOjp;8#*3GiZB9&$cx6V?>MG10KIJjsnjzzW?wXH?O$(L2g@_k-;a# z7yAlMF3nFRPAkkGIPcgO>}0^iBZ>UMbo$VUOpbR;gIn>x=BOEW^-rA#6Mv~l_WSry zl_bX`z+S?D8aq%tG-eb^6_Wk-ZBJ-c^(!|6Z`S5?dLw{mJ{pZ{0zp{BA zY#Co@k=XjiLY z^0w@biC3oJ$pxi*{P;2b{t1OCkn;1FUuUdh&d|;-I~HA4_U!5Om(!;wQ=5wd$n<8< zvx$6!tGMq_2sybPhTX;a`OVGElzdV5yJ_2P&N)p{S4B_4)ys)sCtPxr<1;UfxYKFf z)9g5%HZ}jguGW%tTWw2{8&Og;IzGS64K+xMMn)AoT2KhiEA@I+tX3rDoIIQUTEv&@ zy4aA|x14_d&tHD^&1au}^76Cn9^-%c>lfdC{>A5Ce#R+}GMr6WyUu7&%&R6tl@2*M z%GRn$bt_U%vxlPC+l|fdu+(`@fAP-+{voyIqv!aSv>3mqX23t^_~#e+haUKOLVuoU zb;idC`k4O`1^Pn@q!%2d@I5Iy@Xyca&r|yQjQ>8>1b_BSpda$jFZt(J{PS!2VH9H+ z!SLVE-+%JR(S7cp{f_>^QHnG!KdV-^{7agjpP|qsH^^ckMqkCk({L3__T-Mgq>&cS zCt~|W!R>;6NSjZi1I={8r-PinUu@JMV7}dT#hrb-sqCns`qA+eEU?Q^LgF=i- z@%KX$;3PxKt|d^Y`kP;&Rf#4g(2^B~i{h^A^Y8a1-#| z;K-ulD#CBO6}Nwjiv7Bx&=qaMGy!uBc4-`2ElbuR;tL@gdfr`P^2=M=r{-l=^cUlm z$l9)&7lX*%2Lrfh<|W9~qiS-+U?zBTXM|kS7(JjdzL0TYhDGFrVY$28vZ3^{TmJGx z4xpLwb>$*IP)Pj#8&4D%_iPyPYNs6@6ropo=`GPAW>M4U5rDp3AT3}6w+@h-)PA92V?zu&Ab>#NR7H`b7)u|6cE0- z4yM(rm05&AX(J4}OIfQHEkjL`p)2{z`|XZ@J=+-PGXsj)-Lc(uiT=1yip&uzSJ|!k zr$gr$ZvyQ+IiPe!z$pFRmP`KS`ytTo~lbh4P0~xcrA|eg|o=w0!_W*-z7Iw-ER1&eeF*ad{L>( z&H7CRpK(ZSs(<<9^Di0r+OFm9g zDoaxi>h?ra4(@6LMcmZ}j7BX>K$*Es89ZD_fZZDtG7dp7DdZ@Mr2_b0gC(}A@y*@mah4f zArbw?FP(PJ*d4nmrFFs1#c&=+gN-G05}xw$;hUnPgTN^UosB}6&yTZiw&30P9d4@N1# z-tUxyI|@{rhG$}rhhhI#qBu1&{^!gCd3{|;@r<`OYbf6DiW|=Kec%;uD7wLMOuic$ zd!wEXP5q7d4ZD*&g>vm(5ldZvSKj_HVmKe~W*y%E3vLymb*0HkQ3~MAp>_=`#+`|p zUrvagee%%d(m?LXudB^_I=+}<`No&0GAT!Lu8((#2RH{qghp@yR56qUM}v_5W3_ z;XFj#T-OPRYkHoJI;LmU90@bU`SJby1O?nm{h?anN)eOMWP4rr&_oTQln~z^ainK> zoX|78^Ha6EDADBdySML@pnbHgFx)m!Rvx}k)X~A~Rl^R_G22vHykJ>8=MjFQ3LI%A zM8DyL{-41?uYV7*|0hlM879kWM4n*>q7^_^V42HeOX~z*uAl{s3_WROX7N98RZzd7 zRr8@lAv-A#rrxKFs=yX4Rs@uY%8m5Cv;hQ1=m zzFiMDFrv@70BNkbigHu>W z=k;7+Y27ZGY*8X5YI>6PVvo_JKflZ#AD_&x=977I_R-&Afcl#h*#%1`96C%@?Kx6)x#2*g z>MRrfJ}(=nREf4~8J?fB>pOgh4c2bS^=oAJX*E;;6FZZ@g-x8mQTdrNO zE|~8$G2cbcQoSfgw{klthGJY0?d3_%kOVo*a>>l$oMZdOhB^P9{9Z1Cw>B_={S;q` zrVt)}a&;15UOC)_5Wr3Iw$&E%P#_%c8BZZG#zLC^B zIZm4J_a+QGp9p%Ec%N?@R;eqYa~q4th?P|#gpSpQGI`z!DTs@opklDIE!vf72-q=9 z$o8XFq{$PvnQCP9Pnp_gB|fF_PBtH`a;-G+N~ETcoj*3{^*!oWO7?+?;L!MH%mgi5ZVdI81N^Sa!c#xN_XT}B}Qg^i`m zz!?^I)Qt3IwbUVAZ&44%e?5n>@iObLu~xqZp~k-J$8lL@H8BFQ*D3pWE}WCmKP8C) z+~V5}v|)%Su1rm6T98qntG;R%X$@SCW6e5Zzl4qm~`;_OytrC}+7J6&?+$qM|z*e!y#0uo(e?Sb_>G2KpfK z?@JS&!RWVS?_A7E(XeE>)m_DV3p-fbwtMk!r8od3T|+k=kWStSWG+C-#860v{?O^N zIitj#%hgWh93zZoT11nf>#sLAHTUCxzO+}Gdkxw|| z-!LTS+C=+6RJg<<8PJfoUubcsRVii|6}PC_Ct>Wui79-UIppJy$gy6b<~}^Q2q$T6 zb(}?wVy6sGgXbZ)S$ReSienPxJ8D+p_$G^zQ(K?%l4z>%EKr4Qi~@}n9ah&x3q?GS z2&@*4U}fn@{9}t|wJ5l#sFfGC=`Mcie&3MKH06gQEakv%iXip<*G`Xy^%9P<1U$}Kcyq@ z_*VGO(F)n7*X7-|*jZyapBg#t12(VX-cVBBx4rml=1rrJeVALcAO<*GM~U*`;vSL1&OMOP6sPjy{AdbE273(bE5k$tFqa}G%*-e7=otMYnWhQO|io*cMBiv&-JNdm(W+bDV!)qu21 zc+bOoL~Bgl*J)`>aUC@#SIK+W7o>Cpr3$Wm9%Dni9$F6hqTWSWT>I0Cr3y##tz2u4=5WyEh$^7m0Vx`2B2)0Hu0THux zxrKy%$~qFPo)+8WT{V!T!lz@P9u*7Zhrkfj$xlv_q~W?iUMQ#-G=8xbYiDy-$VK!k z2L;%z4E6b(ls7pohfpga{OA!RZE+NHR`88Ronv{4W zhO^;=@Al<8ikX-1R9*z@3fZ)@2BVP9By zY=|i7E>Uq(J_f)JC5j^i@KsDB{3fU#^#r-8R%{w$)fQ2;=w(&in&f_U1!iM1~UMw=9-M zm`gAcE-@^mQf-S(ltDEv#fR&v&yD>*`-#_d+PFb4dprn#sOeVMXOurQhg z4_dahbvf>1cH^XWws(B6-+80C_ma1!ce3$L!?JXc4PGq?K3>xnE>FeCcuxt;?P}o( z*7;fm*}coD$k%^J3e^AacBW$(7?fj+Q!5YmE)mf)(t_fAA_xM;YF%yios$cg0{}hm z=0Ea-DC~b8$Nppzl&muj73SunF4~uG)W`!AuB6-6wFCN!mQYZErr#L^U9RDiO<9}c zG`?3=CU-XUQQOuvy)8G~nZAR4h;*oUlrilhsNEfoT) z#_X0&rR^FS5C~)th~|hsA?`jox0^$wSuo6T9$9u600oTf-L>QCJxbBh-#~mZorTIf zHZcXG-C!$41Ep6BjG3SXHY<0~jw!KkN~7bjcYfsQ5|6xl6+rX*^ZuskFfO0W|e z{X^S$c~W!A%Y3(iB>ccaH#p{-eQt8@{=G85h_}h{;u5koIX(5l3L{UKKv>p!E%tp* z>7a9a3)%21(J<(uCshb9E$U5p*3*%ZXHJ_P16iwwkGH&WOo^uq#KvD#n_t`GvcSS9 zFVkkrj7XSe++9?Xlkq=r`|@dV-}>4_TzJhv%Rzj3uYr>7xjk#dbcS|EhhKG-OfVWi z+Cy13Qi?^xwgmb!%e^F)RI8JWHeXG^!yX{B#-=21-!h{F?$)u~Xqf~&PQdp1TQE%J?wxE8-d((ihRAko zyhR5&W~Qa+O4$>n)NKTTL=`blY$S5aN)mkWd;aa0UA3?~zdV=~C}$;&bJ#7n3saqp zBG*InP2<&kv%jX@l`uhN_t1Tuqo>|31|YoH4<<#wL;;#8i8M8QYvRuEU%jZWZD!uL zQ59==YL0tiT&-qR2v4)%UI}*LRO1+wIn1uQ;zl>po)jE)%jYVJvfY{{f>3L?KoJ~= zuzTLY7QHQi+O0+Tz<>xN#V(&9_--pV8L;*CawuX36g@5Q1n8>?OP15K)t{KSi}cW2 zwgnc>Mg#VTQwUd=JDYIXj%b;!8LwA)5)1@b=_N!(x>$;IvGfdd%J2CSsW}-_AZG_0 ztVz@YpR&`kHbsRvce5cT?x><7(J@XGjfV}1PVBaB4lqWZ;MWl!R|*vlijuzh>WlLL zCS-XLVe$5cM82YsP+BE#9>E7|c`XPjgp;i3)xttZJ`pN%ZfjjDDaX1*g*A$g;>%5sPygh!kvF8 zwv&*|@NMzhKAfuvz)8BjLE>ghAlZ`;#o($VbqiaU5F}Vhd~$@}R<#qj5o}X>Vcz88 ztaOo``_?u^0Oz~3K017Dm4Iwmv$Khy3f3-A<{^T^H56g}b+Ie8Y=faYHjJVO={{Ht zlCH{ayrhj-8qHup$6@_B%qL^ z5oAW3JVc!Wz%g*pf&Nlulole4jBQ=ZUTvW{fYTqZoz`B@*Uaf(4GXSBaIh*TR^? z4?Y!2u?*Di@f2^ZP?{^a(nu5=ZOhOnhL2KZ+f}!4BsR_ALr0o?*YZL8O~t#k`Z&Gh z@=%xIRYYYR!N493d6?B5>l>9MXtsn^6(drIH-ot)nOLtA11M|>0^%_4M`#2^8QOs! zWA+FWY=lVMiV_m0(K`yFt`}8~J{fa$do9YPwSL48<<@3B;b5qHRHTfm)AHX2LPgmW zJ5S$<@jlQeZi;R}`U`#|jgl?xJLTxJgkF#$b|Vk8!y8`=g?oZy(p@ZH;-!@`045~K z3hM?20qxJGW2Y>3t?vy-!z7kWE#O^*i{AR6tA z6-$XNXyb$iAq(NUwWS!fG!ii)o!M;7G@HW-@P<@U5T?*ok0O1nk@96zHg7_5tbS4? zm_uLT{DuSVtx40v=m;Htn+ma`m*Rj%uL?~hXkZ;Spl*F@swP#9}& zhL}=O4F0c{uSrANN5q&6DN)2FO(iEWPrqn^#sJ47go*+Fu)Gn?Sz3JX+n-ei=;oCs zL(jmL%}PuT5Y(5(FBFDo!K<3rvOKWH zw+Px;LV1kC^08j;EDE=yqn8mzCbBz?NNWsP1AcHcX^)el;m6@sA`w)JhBGo6vng3* zCU@J|MUgUDNn{}2`j!?=v~e+RS1f+RN1|PDCQZ^xq|4otl!;{g#_8?>Vl0B(jnGKv zCTC!(lR^|;ZEDaWDSp|_B`iDA<{AC$Xm_(d%HD^|^%*_ViyzL#PZxF%<>S+GX&%9i zEZn@r@7sMNX<6;+21hFAHJ<$TUElgsMZs17b^G0r;ufpMW5prb@Lt& zC*j;$L(xb9LQp@mO{^GFGGWAf;Ns{91s`Yt>P^hkmfc`026BD0;SacHM*qF6&qLwKwf@{_KHam0z z0M@G$o{^h57&m^~qlr4%zxi~$Bs-hUdCtVB=vI>gOma-_#)5T@xDhQegckB!wI+#} zT|?+L@Z8w8ud$g6!J~UXW?sU$F%S`0ptNRJjFx$;dvT<=>{{bbR9rBQSgSKmTlf+r zwA1u~(WN#<3ma}D&e+B{fsAzA0SP|(+ZD%V^+W;|s!fyvLY7l>Jw|Y2!X(jfHD_Qn zmkI35oG5CjLEa765^=D}VuL13yS4AeV7LmaQwn2byF zYX;%YSmAk;jL9eVrq`v!=IYTx0Opx-Q)AOkO^X?0wInjgxHv^!q1FTH`>vV{LQ6AI zAwsJ!Zf6X+l^-{~R^0}+poeo-Xj8XPbYIFgtf^_a}Kx*C>iU(V1IP(bMee;7jdYfFTs)o zDVeySt0Nq~;4{NCPUMKDTu#7NUYM6VI*ZnpiZ?Q?6` zVq7~>PC}vQXs3tl9Rg-RAZN93;k_hIby`)uUC9)f?||k2G+|y1f!#5ABj9abt|jc7%E%EM(p|L3CJjF-4oPEX5za=9I8kqK znWcbih}Kndq}z*A7(Nk&-lFM~Xw3$P%j?n@XlgpQp$M84KAiV@BoEhgF^63xdsN6{awxbF zsCOBKVP0*}$1ViLwRkRJa#5-bVesi8;I-`s<|Q zJRjiL5)c_BkE>C~KShCn&M&%mB&1c-1{6aPP2{J1UhqX^>V>&18e+& z>qu`Px4Ay(c!U7Qu9S0ggDiCtLf% zWFS9B*5Vo9Zz@81x-134(;ekg50Qf5e zf|e#=TuG2LG;IT8ux`>CNF=p!2(hw`sW`_Peg@z`mR@0yqoxas)=(oRMs|o!*%}If zLrCs+#9fLUD~)W7W%z-qI=XJuslubCf#WRG=^8tXI38W_GYERA2d{5Z<-*T(dzRzC zd7@sWYe63W%XY1^5zNVrJaB9vmXeNJSk&5hX(>FHTAXp`&+(mv*W|2hx!dO?Z}IXu zoT+P^%dXl1v<(}Ck><$MZoE7KR-YNq8-LLvT2nsSiF{SxZVyY8ye?$c_5A^DHF^C% zq^*Vo=BUd9B~+?7s4!jV6zKo$?(4VPHnKeb@25blv`NVVX*-#0TCuD+8K*nD=_Dsk z&tz9r6p#c-L?plfphU%VUS}R+{@VxGXWF`7^$ki++;jG<&#_o_>jQ>vb1gEfY!T8)Do z8ARLMx<|;M;$ZA%dEU=j9DqBmuXonJKik( z8j);^t%#qoG<4S%RUH|8{a8*RvP7jj*m8~>d^ITo>jv093g1G^ww35X!Zg<3s|i#R zbc&{!X$eMe*Qs%JLHJZl@?zJ9X}Qb?xYL0p!6?1Ebb|bsY@avv#8T?GCSMr4o6wt`32*_6&@9un3tS^k?BJKGWp@KNxP|adL-xIZh>QWfXEVY zR|l34QsZH%0gUBRym?^^n0t<4H6Vy^Y9R75r~7b@+Giq*Tnplkq5HK0?%5TMTwZ^%a^lMun1 z-540Fq1p0DL)|;!btbEih@xSyM6xid0Z@V9@4(5K`*+b1H)#N?ijgBzIets-u!Wq> zzA|^~Fj4^%3vNLqF?OMWDI&cWM?!fUGXUom>JTDZX`1Sm^seI4YHgQg^OdtD?E+v! z!LJ}Zx!TlDB@{bTcnWpXL#$DlYOgi%eEW}@cFyR!u>a_2=P)O_cnz{!TJs;z559Z_ z6^mfo)OU~j=cWtU_3CNW&Vtk1G30|05F9x&~>C1$*>f89@DiB4aeuGwNl5b@QV@-$6?5= z1l{JK9_?Y)97ja7WGMT3L?zRGnOnpdnKQKRRv6Y0&(hhl&)&M11aw`8);Y7pjUfFJ{kfG1Rz>W=*naKv$(+A`r0jwezx)a zOwtWwe6(J}C?0ObNT>euKQt@OivVr+sNvzh7A2ruhK>>yz|#4o%sRKaWf)aVDB&Sm z7uOnRF)*X%D)^l&f)AXDb08L7S`%YKBGtGM4*10CfUK@QuqJ=HPG`SYrSKiA50z(mMB1fipbR9;ko$ffGdhce87JM+D z7#p#)v77dmW}Hq8QLsA6Q5yyNn(Pf@q^3)~$ z7)$pNknFCB1h|_K?uCp5PV%K|4>`=TmXakbd#Z^(8VAzA*bSPtSD!3gb(1eHAL2rJ zp<@E!6ejSo6TOAfjTP7eCL36+>9Z-(;tDBW;iYNhxVR}vbIZ0x;K0~<`PML*=&m4{ z%rY9>#n1|5j}E$ZjhPiw39LAS;cY6uZa1X|I^L|c%L?+-!A8@Zaq^O^ZG%$+g0FoJ zXKmN`9Z(!|b+8hurF^*zZuBq(K+r}H071oz)t2Dk0jS9du)GZOkl& zH!EzGx~d(~n#s$-8pb4k4zuf0iypqbw(uw@2wOd!sQW6qdGZmiA(LgXCP|Gb%t4A# z9Eq7$p&Ze&A3Rm7Oxxyk{pcxDLaOt%$q;e^;H>KTI6QJ=I)CqAcNn_UwjRUp4B4m` zG-TA*V>DT6@sFS=v5s)d%t>UKuTgQ@=#7}P68tymJn{0BhowL!tChvJ$C#(?RFq88 zg;pk$s0DATuPWE9hGg-y4TOx2P<5+Jt2x@AEvU=v-PcN%sm2qYH1y_KO3;kh{FW5BD!7f}K`6az-Z%G{2F=~mpb4fytdd}u zO_CP7MP%xjXOS-*6Wm0sY`d63Zmap6%_zl20Ct6J>JR9q>>3&EG%2#CeDE};9&5-k z1zGSO4(l+0_i&+Rjae+YP%|JcTx)|jpi=DAP)uRkYF`{2XDlDbJo7aC7RGKAvQ+tW zF3yc8xwRtTF)+fF*{PZ()5Mr<6_Xrsebg<9i{3m&+#f}9t*!Eogf1xON$E7*?-{=I z)^fXeD=c}Q$nx@Rif$50`|`vW^aT~E$*FqquHoWbOzx^AUKZ|93EdPaNr@UXZh^(a zm+%GU7CC+5~}y^=XYu%;Pmght#*ZY_Vhzp@ySRLN5Iw;Zd>%>gNF zAVMpxZ&Nmlnt2dpFCk6A6c&q(n!Hei$j zt)ra!$TUSG-UM&nrUH}mP()sWA^1mHo30v9T%cPISWI0Oj>88^sIL*N25tLg1A=;A zn6f96m3Xb5P}NaVq$){Oht9`{vn@ z2HqZ4>)Li(W-1Xop*9cO;aBu>q1N;ztn2yI^)X7`Sha%%)uF&(LJ+1=QF{l%b^(oM zqibXUo}`6zKqnpvybr@v2j!_n?%o5p3$f*%&-7DgivQTu`MPJn19{=Tg!Si=ZL@g( zXCG!@u9uOra+mWk+u-HvLpaN0ATO^AHRHtG>O&TTJS0SzyG%W-*&W2gn?>b&Yeh45 z9c`H`Q?*;LMXKrKA6woZqhbswtwDN#Lp*;ZF-E4$p3VkBkS2aN4; z0!;MbiLOvF>ysZHKfhA6 zlz35koAx-qkMGf#|3OpZT{pv>clYg-ms{}I`%mbVeT&O|8}}L2@vfVOV3>RN&_AgW zv4!jYlV|X4lW(TgTQ|z7%l^P=8KH05U9)h3I7!Z(diYMG-^2Cu(5zG!)}n=Je#wL6 zW}!~Oc-VCk`gL#AfV1j<*XgQWuPk5B!|6U@Pi%iCz^CP97wW z5<;iA^DcPWgA*q&KkF4n!_2+6%6puAg9h{~zuhYJCpXo5EyaTrt8AGN)+y-6^Yqch z*=YJ`G&t+21~Je?Ed}ajvBcm@vJe?|t`r@r|M9XNW%{G((P*Gb_0QFGT{ag|)8W|; zsE~Gac;+XnXSm-}q_JOsvw*M@=zchO`O*1qbfKojx3Cm$P)r||m8*QkdeLzCwH%!E z_6((08z}T8glNJo9cFlfg8ra&2LM1ozrUj&M?aqR_6@FfJdf8-0D}lC>GiyAHmnBj zIqKVlE-JF{1<%KyU7SBkFV4;%Ke>3+AL7p;G>yCcvm#P!HEGqn>#nr=Mk-DJ8`TQG zcD3J3B4fG8w|gqx{-7)UOP0=Fq+f8=I%YrMZ}~PdMA@|1^9VH+YwHoezDWOK+3)5qxwwFbAohRd5@p*X0o+jB6=HspZ zoc{5ryW`zxusb=t*VFKZIV;;>3isPxSu78O)d8mC=Uk|JR?{A!tv3c=?~(cLQ~(Lb zF8ez8D?(oUo^+OT$s4wVt~wTZ6Z)=}Zh#QNq!{N?H}G9Iazx5}dz=AZj5yaBxvnka<<1 zQYjYGTxQcLp7(oC2Tyu5Isf(L=U=_}?#1_Ce)IJ!mD0B)0a=^Ld6f?J9MX=vR9Kw| zS&pe96Gz`)?MYR6yJDdLdd)8C)@-y~y;p=uQ&8yp0~(EfkwJa@%q8bwQ{MU&q9Ec<8~ibRKT)>&nHom1?C z(9alk$r(Pe&erBEuLGXd#^C@Z0rU!=I?|C(@jX7hc6;Wte|-7X_h0(sb&8OF`~%BY z55Cu4iVuK4i1J5#|8vaqjQ++9qCB(UwAr{K8Adj4Y$`|aHMxrPIYoZ1TpbZmzowqc zM9-9sL#VZ@pcF7XnrG!1g*WB&j7E~K*(R=!X1K;eviOr;YoElfOJ(|{s%Q?dk4U?^ zX))vd69;LT!(jr;powb=m1Hg-kfmz<_PjcfmjlNvj^ixvih^?frpS89>YC!b&JpEh z1$hurzbVaX5w3&tD{)CeP7`PBTjz$Cfw?-^-8$E?^W1p95;gAyxd)i4HdEjle{%W& z$iTm6;^ak4@GzKLL9~@T-mpD)?q1Je^jeZt;1PxJTyQ4gy40b@+yfNa z=(1g%EUvT|i1NKg*}y5X=2wGnJZDRqypTDja80G!2<&>2iJ?vVZP4T)nOsYsTSn7x zaKs8*)b<*_V>KZ(8P|o;#&{Zr>kj;Fc%8vD$y0Jd7wxr}(xO$1{^oj2By10gZLP`o z6w4&IKT-#Osucf%@+SOBPr~mb3227{+9IA^r!kY4HJDQ_Oq~SZii&Mn#jZ;&6>M7q zEoM-t-Oz-wO;2n{gM^)=dXVY{=d)R#8#YNmJZM2xG-Ph4X<13R)f2aA!?mA3CRUbt zUfL03mzO9w2T-#H1DJZ|ZlD0Gz>mM>^`M>5U)o(38A{GuN?P*6+vPRyJ+48-cB9@F z3&ORj*~JlZt^@CJh_T;gWPWT)w~IC*Nk~NO{VTe*dX7!Jgg@93&vGO@ zd+&LVEBgD3Z@>NW>%YR@`?@(1Asw=a4aoj%e(H@=Vn8nD`u)Ze`_x_)c9vweBeSS6^jC!9wJUB@&dR7@`WM{qCVI6NC zQBJ0DR>l0Et7B?V@nw1utK&$XMiYfAfJ1Ia?oMfC_km^aF@W<-)Y13#lj9HGe4;ZC zUN)0#origUJAd)rt3O|uj3B|P*Ri8b|ExKT9$+r4!(d#{%B9s2=V)+rHaZ%OdPcVS zchE3_UF`4}4w9pzWH1;cN2ACAX*4+Mg+kH9`KM|aK@Im1%X&F3U;oSTo1QP>@-m&? zCsY5E-O1^0wA-cW+4J2q_4n}3A}8tYWH)-f8@++NUlrh9CNc^jO*zss!5oDvC=Kid za=m!TN4>kHmG==C%QO!q1OA(}^(NoVvSpL+o;}}<&PNyUhVC6eThDH}-92nh$MEyEmyG)Re4CrY!>~9c4lX$}F9l7XlJCq;hfI92unh4BT-3beB`_1$ zS=mavAl-T~yI7(Qez4;9zsGA&^Og8tTkNy79vN+~_4vkmJ}`l*&P$XWsW<)tW{6JR zbA?Maxz|kHS3-8KV#fWdyT2sU^^J9|0Kl%bTpZ1YA-$C+(4u?dW?|OA;IRT$Lp1?^t-=vIF@o zz2-a9fSVOxV`48AWV~O?xheM?&uFLb`GjEXPI|f;{Nntv6{+e;blmWZfcpMhM&j4!I50WJlJKEkcgZy&;#F^r8cCj5?_esYd{`L#~9Uu3b-X9-wejRr_3<@t`2lQ@!$384< zgPysA0!+p%^Zp@qe&_PJXAJlP?(=V7y}087?geH_mUs_0&ftn!&4hN5Go8i0N!BFQ zOzydor-P@1Cr_P<ESr>zRx*7g9 z-?Z&2D~FTjUq1S92vq^Ra{!L_SPs&CGk1`Hj1*q$s;q9xMB_|h@-XJEN!kns3jm49 zb+Jl58hn^ElSQsZPDQyML5HX}qZ7Z4S+#DE7BD}=o`QgRwVcqB^3%m%TTMuwsiJk| z3gHx~xdMeC66HY@43A7X0&fPs%uzy)PB06xknsygEO~TZrgkQB)IdUWw0=UeT`|}^ zLGhW(Zi=xP^IQBeD+E z6loJMbR$@eah}YpYMSI_wV5yQVQOB5vOkd%K0JDKJUju3-lsqQwo5M_?UBVr1zHFA zoicW>NmBL2-DI~-g6e~ck52#@t~R7qflQEB7H0Xpf^2Pt6{!J+mxLgksC_ZVJ;8tm zk_au@kQXWJPTBkKArtfJ>_Eb$1cP&=NgZM*%zBs1NqxwhHC*`M*s1kfy{cO5uGA9S zSi_&BUMU3Ws7@qH$w{T=#SLJ1aC9^tGFj!BKL#K@I>>hi#qMCdJE(RC)7`=L@O(5K zT$t8&0xpkPQhX=!vJX#MVv}3%~ z;a()B##!S~%9`%hh@`$=@YR@XX0u{@s^;~$cW5pYLd;ZAtF=YT=@kc&A)*yfi`Yl% zE4<3RLcKn&II!s4q7Rm0q8>O`eHpyx*|XE$sDJ1W+y#3^CO{NA>EPRYypv^A23VHK zGi`N-I*wT1j`mC$w8l9XM(f zjou|dojO_uNoeyDUl<|`AHwjm80)Y=<}kwga$Y^_iVWspWO{3t1{Ff*F+ebH<}|ls zO!M|vVwYE7u^I<4YXow2&H8TBN?Y8YD%P%AI}hMVV?epi6FRYL&kVlJg0cM>=g%0Q zl}#$AJ2GGaF~K0v#GFvp(5%u(H1STPey^QMTEhO=2(?M3{?!ti>W~Dct}$Vt=uA4m zzf&tT)giC^bgDDynoNLy>1Rsquu4z$L-n(!+0=w|q7HcFO)-;4m$ySYaq8E=3G|Z} zVqM0-^hobh%@zLkL&x>(EWTnt$?Wo1?Cr8+LbZl&mK7Zi1^+ODj-+& zkYqBm-$g?}6@&2M#_%Oa*Fjb0^cNJj-tCzV8JhJf<~`g>(7FoSIjcnR7ic^%nutbK zw{Bq`eXlrEy$Yu42#=}{uBN`_0Bd@ycK2d}YVo*zRWmlbOmNm;_(z?7oo3{}TvN(* zPAgr$RZKig4-9Aur>*(N`O?(JOd^^2grKCW41JX|+LLS4^)bb1hUobP>F^Z7mQ>N0N@ujEHcu{4 zP~S4V#)JB`q=s7&Bl)uMnz$Q1J&RPMmtPZiqLm<9F{%c18F?k!lNk~C!{C@OXdg(iT#?_l|i$@U*^y|Sc~3!vxF5*)Sn-V1 zS@0lG7VJ}2wUOqg2@?F?=E`WPkq3f|CtVY8`dWzB+=@&HnGkettGgvr%b8lM$_6Zw z@J)e)e$^okSd(3?LDS7Lw{})8)WgVe9s(=s z4;LCMNzJh2+1kX=36qXtGxda}8BY(iMjxp%@)D6|0{geAbKOjdOEbOKTu(WnSgj|Q zGIhsAD@W|>icCru>igm?ttC|bWC`vmp9k!O+Q(8oN=(4-2TM&`GLt`RE2{Si5-g_b ziIq}(tV16>)xnQG#NfwV6$G3)1cc&=BuUASP00_QN>K<+b*OM$j}z%*lggz(`cO(k zXsSbAX(WAYQlS~VqT)|@5yyf->pm{&>r1%c`i(0ww=l=|ynmw=ouTh^RJ`XsoHW_G z!YQI)gk=l=57>exvd|dwzU0+`YH=BJd!&%JX%C`XJC5 zpt_OVw%d!~%@k&sP`A*vMjz*gWr}wYOziNSx3)hh>dMR`g zR0}1J9kfln*6@DmS+p)RaPWLL+I6RncDrF4)6W=TM@pw4%6$^K5S$oI`7@3QC}&Ldyu`(HngoI$^y_E4q( zTLH+8878IhLSVEfgjG-mJVBSs4rql*iQp(c&}TV3DI(xk{qnsJqni0O1l;_1@Z|B} zF};jqkkl0ef>|%R9rte3NnntY>?}Q#AN)m2G=qVCbjAz#A^tr<$pQX}-o>9`@gGSf zHQB5-%ghyQG5~lizidfoOj4=wmZekGa>;ggct=551)(=wf!VZbkxW1kHn$?#vc(zQ6u)(2bQOqc`$y}+6Dn8A}8@{GyZ%O>7>IZv@ zT7@%~fFA1tq*C?V!7m`u2&*+fg)K2#O|5lvd>a=zSR$_{#dt}k9a^?CWUAHX6&8+| znyrugcG|P$B(~0~JKv~#$s-{4i$?#TuPo{f+ujNZ;7P{)fBw^XN*sE9`B%t#`EM`3 zd-dfvUz4zqq}}yih)uz?ftazZNl}O{YIgokY&QMxr$|}+- zS0^@cqf`(u9xl5YdQ)R!M(Trb?5>Z^!yr6ejJ#O3v2>klF@nSBJU&LbhbYKa! zjKqu1B9|mtBf%QV)wHxt(?)~zVxI+yJ*amC>GSGE+8GEu)L{2d6o){Y;k$yit9-|> z$b)ivL@suxVTX6l68|1n)w6_vrXKHPNRf*}7NE^1ixR9>sm|9RbeKv-+f#f5STs$@ znaA?ht7HmlO^R_T)JjTH%2hJXC!ntbkpfh{S{O_^lM&WVZ9XJtmZO?;#8bzb|lZP|rII@2crPK>L&XQ0NDGh<#lh zUDI1&u7L->tu}BM03Oy7SohV+IL*;^83^@()y5nyy7}koHf9t19f7lTHnFv`*P=0W z?130gr5g`29*m98*mZ-3!23@*=sszA$CLk_m3e#r(+OtpIGk`1=PZEv?W2G>3;Ya8 z`X~^w--#JI&H$7KtoY#P>0VQwI;U&P)2=+GGWfARNBr=^4<}%{1@@LIX_l1L5B~m-ufG5Ct1rKPnd~5{_^HpndG+#7cO58D?U|USnPJG8tDZRwstc_~9Chxo z9K<6S%=NmNRO{Sg35gwzaD|t|J-UtiGugewzmM>&gO-nbDIIxE;|Cv8UJgJ*%*p4I zOiWrfsb*cX=FGijwY+=?`Y=;WHbS>1i`qNI;Mrv6@}sM@*k**Vu;YS_2Y!`%J@qs! z!7e{1O>afdlFmJznN|$@#ic}*2IY+l3egRPlKR zURa!R*F(j6+?1RF!)_TocuZB?hqkHJeyAQMaFtmBpf+khAMn%<0DL*4DnBh-{s&^NbH^I6c_4 zZ^;ZcQf;n!&hT|qa%H1_Y^iUVlPj`KBj1cOx+*?%ZRe?9d36lSE7PwjNn>zZHjn0zvB-4QcPyCUkkwc7y-d8_*7d4VaM=*n7L*- z(KKd(zT+ECP8mc?@sbR|2Ch0?7|U>?4l1=3)dp!Pu8t5J*M$<%X9b<{D5iKEUQ#Gt z<~Lyx@a+o4@~O!PU`s1jt;1q?Z?3Z#`ibGwQOgFp_V2paDam>w*(IcF; z$_r664^I)6{TK<#*n{f3wtz`AWCR@JdSk+Yt7W1L$Ay&+u~ZL?Me}4OR*L!dgt&BS zx#FV+4Dr>Co3_Mham_5p5&x#A5oX-)et{HAJ*2m*>REyqNj$l8czOa4r3 zyWZES+RCps#L$38^~N1jpsGbtG3?_18z_1>@Q45(2U;;-d7Vp%bjfCNt5Y+t@f2TX zO`Ec72*2=kE=;7o7b8~pE?XYw!i9`OJVm2Vlv=ZM0>yZ@+zJYUk6vd{FmJNF z)kg5eR|kfE+2rM<@Lmu&Dt%eva1`^2b=Nna;9T^+(m7}zm;~cHL3KxX03QE2rse*{EwbV7$=c41;ubsF$KN+Ry#pOR#bzDIF#(#j%jwYG+tI(J`%Wg0XqFT4h0TRuaYV z;&xq`HnY|yg{$E96lf{eX(&$~JF&32ifWoGx33sklOJLXi;lNfxGv`=i`szAcBe_+ z9j&H=a|OlOn!KWfjxn_kPdlDzPYGv6h;|P~Jk1}WncS%7LouHEvUY&oC#P(_n$ViR z%4?qQ5a0z{VyJ0PsqXTmqEa*IV^GdvHU%+2yKpBTrIGIyOtwnQSl-9cU@Ii7n8l4!jo%MFBAg$i+8tFBG)V!rC8f2Mz~t*_BD%53+niOIlb4* zM^YA~bm>7^A_~P?hq$EGw`$ZenZ=~dkUn?_ELN_n5YU88hp>g?f#`aZ@~OlvQs1uI zN{kBv;YLDJmi+o3`;JVSFL7$n}F2)1L(=1E9;H48n400AbU;s(WVC znQqD3mDOUax7%u?r{HF#4%J@4n?*tC26-bgOAMsm*&B)VP~6dQRP}P{#MRks<2I60 za#|xLCZYt62df6tft!TOc`K`t7Bj&U7c*6j@pLn>3qoi5arflG3EEAL4OoMq*eKBv zo&Zrnu$ne7$7Zmlf5;szwY7v^Uw(8Pqk-pmm*70wJj6 zcP0Vj4$IZW#O@LoS?g>reZ~H>HgYT=r&y+-T8QS1(v$wC&{GPo39luHEUREnGo{Oy z6$KsbQ_%Ue4)={(1IHC^?+qM%Et>;*>q*1#xJ}-ugu!l>o0^J*Yd=k(86A}LCby>QNx+ognnvjBrQQEpFmFX8 z1}Bhi4TrobPk#yx%~{te4@pQ*ts_3u6_Ig+K@q%6)MsZRvO_d#acCR^Ddsq6n}N{a zL3vMNheW zP9_o|L%nj;4viVL&d=mkj=>>Q4@gcFUV$xh9$9TJCH(CG+{xM3l3Uphwt7F<>J4EV z?g(hl9MaezhYA#HkT6-9BKP5;-A^?>c#mpMkFDGmuWyZzhKZqUqqS~@bqqEs#Re2k z*lp@?QN2MVY1LNGDgqTy%L^o1BAb{3n^XxUs?UcTG zL&Qs}*nkuPU38&~*r*}Y_&uQIQIf26W z5ljir;~=<}{%%@Tp7({RY-v+)L~0MBRw(n-0W!An$<|q5(eW_g2vtM97X(NHaC_6B zvEjB#)&TcYzdte7^Cv1q9>p#JFb!btUXSZa`e&z!_3uE{Ng(iKe4D6+KdELXk{AAs z8Yj2WJv5ni$>_kz#5gWH$NAyt;=<&Bb{XiVo$-CQb&U81sk-DMt*irhCngOBgQhx8 za$yZBJUUdg0q}%%AEkpKnU=U@KaiK| zStj?a*FS%p{^jIi_u8uGjnwU8>t{VDQ~$N8>&>A6=`U+F3AdM&htwuxX6P>&R(QnT$rHU$R|%PA3uJ4$z`Yc#4U@Z?EEj&ryp<-nPGSX0Ou@6KRi#cY^VF^n39(8Kypl-* zo>LI;lAxqOt)qE~tRC)+87ds0-J%rCcD2RistM{qM;&@a5?!oM!`JEm^ysiZ*p14= zM`xoay*(P~3w#$;-d%u(G?w^BeFbyjP>5Lrq_F;W<%j z!Wd4#Qos?7sph^q{O~cHxV!eF-N}8tTEvWuI7&N)$B(wz`>t%2nZaN2L>GLyQGKUw z&H3)9Nq2ph<5fvum%w`w!Z-iENA2?$;XP)bp@aKAsoM}j1>pSnu-(TQhhu=I(+&4% z)ZbW))9Vl4?EYcLKSIT#JA7}CMq9hg(>M1iGPpa_eSCz2*5U&+(Gj}y#?YHT{a;W^ z0|XQR000O8002Tz+}#xDKP3PFR)PQk8~^|SaCv8KWo~qHFKlIaWpZ;bZDC_!b!=gI zGA?j=jeYBr<2I7#cmEZ--i>J6)9#+`@lI?wVq!zDs*<}@l0{Xwz2i9tMM4r>B*EsR zs^p8izx^@+5}66IX12Z*@(1vRN4}5=y?XWPs|oy-ms75arhc2LXI0g2^EipyEWUiO zwg2y>zprfdEnYJ@nrgD{FMI21en?Dtsr2dG{4kmK#mR?1|M{=qefaZ-|Ck&uljf`@ z+wSR;M#cG?Nv+gmdcVKf?4q~sHknRTQL7&&vuNS!Z&&l>U9_}MO;r=?tbY5|EBKnP zj+M!S@0sYlG*vU%sgruPZ~JEQWU90d8aL)a6@BBm$q{`s`KT_~_|4=l&efYq)Kt1S zy_xLQbMw_#zx@_x*>As1{x&a1PYLdrA3-`Z3cdkEtT>kQTvs*;)tgmgiyo$Ed`SR5p zN9>o;zAwi`+7}O%iXVC+iJRo?2~#w>XjM;mY!Yf*s3%`EzKQX-pa)i}X)Axj`m9yvzrSu&>C2b=VEp}I z<`q;U-ONA!bidjA`8V4vexdU1!k7Jf&+hm8&7G&> z6BTQn@E4n@rqO$x`5w_lTwOLLt|RtdGh;FpdvZSaCG15{Y!*u@uh+{vzf$ILa#mMu zk?^}(aL2> zWun7gzOLdF+ommj8OPYHysnHoz!Pw()uZ61#Jl^to8@%D%gfz+Nq&&zUnKdxB>zK_ zfA!>!O3N!>-mm?nyou|xSc*5ml~uLBl!TZpB}R91IaJhjH_)zXKvWL^SR9}>6#T-j zW*b-Wsfx?atq+D z%@6c`SoN$15-EW9vQ9n2q!ts6V^?vmiN|&|;Zs4l=;6EIOR5y)18!=EmzHjUs%!?z3*r!2(``ZbyXR z5t6$&YMH5=v3LY?9cN<3uG_59C(B@l5_;_>^I*MpI3KDGlHGQv8dl`Y`p$qxOw3=- zL`T8=hZPD8IpfdS_=&P9MwA-YVUiNBKFW-@)199(+nv`dwsGY*7uV@r)NOJ8dZL1HOqzhSugNeNp*$dWZNZS=K{kV$ccU}tx`v)RY!WDy z+gi1$NeoGHLIM%@(@(zCnnpus@UPnKsHA-nrkY??J5`&k<$WaX0p(!UA|!blB1{bK z%+qilMAP-wm*>0H)^qP@J-zjF2&^ryyhi)o=I)xPUvB+Gi0)^s7%k`f6)C;ZJv|9j zsIuwwIbu1H%8n~wxG)_04|sXbXCn2bRYk&R|K(QO9Qiba+4oy|bHrZpydf%aUiJ(A zlft(dxsz@6Vy!ibJ=Gp+`*j55thhPupJ0{J3$!3!gA0VE<|wXL?WZ&9@x7NW`{mER zc;ic}J!`w6(g!M1V?QzXeijIxAb8$BuDxx&S4C~AV1@2g-K>hTC4h4>Zea6qzV~zP ze&>syrvBA}Rp|yrQo|t{W00S8)wFRYzUh;$b-=GbgECY~o$J5<+m}>j*wZX%?bxgP zJ3IH89Xo?YD*Lg&)RAO0swVEs3w?^gWa`P~&#*>m6qh8;uj0bn+ShY1cW7Q+dxi3v z-wQP4x&~{k&}2B?kjLVdR{ZUKsuHlP0z+Z_fg6-}cL84q=1ELY*8Q|wjNLH$$Q=N6eXEkl*i7|6TF2SXDrKKOo!6R@^Wu~CGWnwQUvXn2__(sUjJ>&f_( z2XLWE1qx`9T!IQ1qEOWV^u`$8C~80m$_sxtn)rEq^1~8mam5JF8{+(+nFNT@?E&*V zD$#pGXaWn7)<{oNClu+H$^|Rxvlc2LnA5ss0x&N8%E3iK6Eh|Y(OJW$;w>{tfL>e+ zU>oa7B7zNBJ9=AWZfH zOx$x0`aA)k)-vKrk<0F&rIvIsmGVS_E+ac0goMil5K|xvo`E%zv4&;C6sJV#jC~3E zT~sCIYck@E0<1Wx8FSh=WH9#tu6Q|g6e0#zprY9aUzR)wVxdlzQmldRvm|4aDB-?2 zB&ua{_S~AQkx;={~fdZ;;@9463-zS!uTspg4!U|)+w!N$>K8T8k6 zTph$ES(_KrG?f6_LlW77Rophgb-TED0I6_fE5TJnPo$YSB<9xwXDc{~9z<(+18-~i zC*-*)WVaxqYW8xuFaS{ofq}Aoc5{ToQlWN;c&QTZEV>)TnQY*sn`naS0nr8UMI}GB zK()!RkK&(R{NA2q>q0rd8W{HMWsACzA`$mE#OD+ZPHsj&#h0N0gxglUJX}NoUWxH! zUstm&7WSwHyE^a-a}k@ENB-hKFvUG7Tdpl8Al(`zm`k0oBHgsD&=8&5*(ELKu4?#g z4iDX<9Sj!K=xzAhr679cIHc;(o&>S1D#NOSX_P8eMkqfSa*Ww>ODzYI)OBUrQbM+3trsR*5uJTl|Pv{(dG(S@vpA7J0=@|hu z=LT{-A00gfbzF3`WTERdz&_S)P&x(4^y7*_8 z&_S7+gcV!PUQ*Cle7U9dr?4CDOx2uWf%w)6gvW+H)RKOBDh-S4Q7A+XvXh%(^mOHT z9o2;!51`EjZXIY2rWajNF?p-1BP%9UW)ocj6P5(s>ycCq<@PhEE@0H^nXCZyT(Tm3 zEvRQOfraA^D)RS?kSrVl()Ap5#rAL=(mCxI7#Nu9xnb%rQk4dSnES7eG3W6p5SQ>J z^2QG3{sBbfLPliG)CZ(@z;wd+9~(juNRZ-^ zv&*Ww5yXSK5Z&PN5kqv$;?pfyN5sJZX&Ran9myHfG1IKIVLfZeiFPzqqONC*d|>4~ zZY0G9^a!GomxDT{>*J2U+u=*?c*_d5u#Y2qWXBfOUnwt62P*AU4#JS=sEEt@Op^7K zq^vj}Q3(NNR=tbHkW{0l*5Ki4OIJg^>H*nACvpmPVDzA##w#v4u=BCY_N3g;%Lw`iKu?by;q;g+mUaO-_5NwE}h zt&~J;#USYcp_w@gK+&+ND@;Ko?2a35#t;|^K;ZA4n+v*=*b2hWi~wzw*R=^uIQH5W zsXwbMTPO_Jk(Eb7W#%u|1^2^fjOWPMdT_gdLG{h}OzCPzFImGwh4-ez1NEmvRj(Gw z=1k@6Y1W_t#X8P{U7`+d1OS2{X$HWm*bXr9b$~sxi2#~kRjYtpR>uHw7JDpjWovfo z)MlW|{$Y1JrV0KBK zt?MzVrt2HWxRfCosinRPWEZx;k#0|uL$K1gAOxuHI9@V>)vPC&LIH$63cV7MLd9ao z1TmNs1s}b&lA{@Z|L86}z-{ofJ5=3RgH5T}q}jC1;5)}Rn*oQBiTUu_p^~JQy-2&o z471z9rM$M^2oArp$2Z3zCf^Qq{EHwnPN54)ESGMP%1@WKL^W-6Jw&--&KaOVDDXuT zeQD!LT`?~NV6cGFxaz#*FZFTM9`c|I0Q%B>u>fl&=(4mgu_|!W?G}vbn%7o?Vc5%Q z!Zkcye^S+Die@H3OgM4yl1;OdI*`(=$=RfXSfAhH#2nbpT2FH;;CH4hFkh_n#5#&C zGY4#rHqb}^Az1bU+dg^Fv^(@*>OWdfWibmiPW)m9C6b#B67Kw+9PC#m$M&jX_dBxo z+-}Vjm~fRebu!Ib2Ydt_5{)oNEdnbp!L3t<7J_)-Gadu1uMB%VpdCx4OA4{u%6DI| z9B2WVh_qtzD{~_w8a9kNvEuuZ2yXQ{O89m2_+hw>VZmV-DZ<_BuDE)Jc=Um&l8V&> zi0~QFEJPlNGle4N!syClUJSjZx!~+T8P+cyImaE50FzsKxvEY|5Mo;}7pk12f+#Qt z5~QM;^osWs_9t5tg$8dTx^Qidx z%`xgu`nw%Oe$|`;^8|z*bY}bHQg7(URbV0IW6^m0hcI9bU~4a zlN2Cv`8FZa##E{Gy-LLyssdjHAcF@zars&>%M!~%fkA*d;tE(PWt^yR542$Hvaq_< zob5*yrdjnHWz^wCi{=vN?1)5odk&yY-LUPYqRrjrOwEv;dE(@O-!o4b;|5ZSUki}( z+L5ckI0BfjWFRZ{L__7@-(1rh8Y7K+mJbnu#jg6tO`j5?L)O0v7iZ% zQxSY&%4W;TD5$asY-Tw?M)FEi(KTSiRV-lvM)84{+4Sqky@Dyhq+LlOw1n!42*$um zum!=K&Zaybq_m-lB{6Djxnf2F$46YVV7jY zOi`vJF`4x}eMf+^l^IRn8%UkO{&C5-5G=v?bqRWb=Gzbe2pwi+Bc4NDm}SjupL@?Xd7LK9#t{E9~L`?bWsN^6|$ zR34)TtCBE}$|!+f9r(;|>oaJQPca4XBgCT4LL;W92M+^TrCG9FM|L?G+|yfuMMa)C zM}6GIZb@&sF5~J(W6PoHfy!hm@E9>H546lm4Yc3Nd2k;XopJXe(q&9KKisjcaDp!Q zzOZ%l^rI=k6oZtLv&u}t)%L2_ilhq&H>(5dj6p&Jslsi+N}VG0WmSYAsm+!^TbzCV z3HdSD)P;1kR-Hqvxr9jIK&xKxoK-5549EeGBZ^)%m`-Ak$WbGM8LCze!4#3YS%8{Q za7I!oQ=MY&9|um7F>%gRVwthl#@uU_68_1Gd;U4+z%u%x2R`x2FlG#^ zoynexr@Me;XMsyV@u(#rgODXaM&gwK<-tl-%{GiH$}NE64c{LC+HKq}3(i*n4z^Hh zHn?@%wA>|JC*)1QG7ptR`j#8N2nZVq0clj#u(r644|Jt2suF1wHJE0}dCS&uOZ);varvBpD!}|s7L6_i^sB+;lZuM_;)SZ6<=>GKU|R}ena4>j zkXz9tbR0(Xu(=rQ^Z2 zJTnN(-Z>%6tUA}t*|~SQFaY+#`rG6R{ZI55wmPvK7#Ou_D>es8w!BJErZ5SI&j*GH zK~anNETs4zaQ%h7L&KB^OY}}-23p6d_{I*COQMPEvxu|Bkbl<0!0=hsg)Q^;z=|0U zFacWnc04e<1fj0=&{11kH!W{bgHp~|szt=&Zx6x{uW5^=##|Ui@W`C{Ni*M6eSs-} zl31!b^&nRU5Y$%&;;_THJgocBgg~PRaXAnA%8$Z->tPWARJ%ACBL85;%jpsUb7qGs zh@%(c2UcboD=ezKi?QU^`dHW{4>xb(ibYt!^m^D?S?v*p`@x|2V@L6r^TsCpNQw?o zV0us&<`;Me8opv17|9z@(I3=L!0WlU5gIcZ zt^+W0(>G>`L0o7$i|N}Kv31uN?%V;2k9T7z)ep;>h%lQCkl-=*Wy$YUvf@~nkA=hd zWQ_M!j3{azX9^F@soWOaU)|Gg`^=*)tRZH1p_?sdSGC1~ywPxjwxSgWT3P3@CK=Zj z`pB3bklh}?-+ZqJb%8|!IL#LzzO+hG+Y#%&Ekb3WEd-x#t4wNbZOw~Su!Ybz$N%JI zku-q7fD1l?{gASrd*obQMshyNgf}3mN+iFbP90}U&}Y(=J?bf@jjtz_o>Y*g_hbWa zc09!L>8z9UwI7!!ozhBr(hs`CD`{6 z^zwcNDi$oDK<-Sj^S}a#Q0>qtkh}bkVUs<~&%~t&vb~vU@=%zk3_qTm*I*jx&XxvN zL9JaLmf-V%brB^XrK7ew#w@;LdIV0fWG+Sv+?g$Rj95$}-99~?LgNGsCUA*OEx_Tk znsZJ-5E2FpW{z1T&WNKlq%1Q}_Fx1X2uC+`E?<`rTWo!F>{!}X)B>;ZoEPMU2x6Y% z*(iLrv6FPV&!MNb&6#acOxtIj=}xv#z7niL!M0{`rC9VePx6_K#J}~)P~w6eX|^_a zt;?Lg#r;I8xDdacDd0;=jMfYjsS2A>-5p^?H9;cP4CDqVC%TAYItyur5tfA}vT5{u zkZv@)_KDTAl?b$xrz6aQe5{^c!W6d?w+?Fv_~_GD51C0GoC?L2Pp&}CjQ%|`3amp= zf$7j9Fm(U&;jiB>gIMkENFv%%oEf%AXXt3B(E|}fi6fXo))V#!qVCpG0L@`Wi&2}$ zMWNEJErB2?;vx*{na7WcB7Wy8%g{!lJo}oe3YKv&$ALQr=iTu0k3dmivz*uXBb#{b zAc9p!McI7@#V){rR}hQAA1kl2(j5`Qwj{QwCva<<9SR-(8h!%*`@=A zu>p_2Vxzz$CF4VKqztl&{aG9#54j+(AeJ+MDuLJ(AWv$Lu$lBJ4)G zNNf-!#_@tRN~wA-qwH2DnU(O^;c}P{GQ;LwX4q;Z7H_O>Y>tHzY~E*qEyHJllsL8W z=r07VT|C1T%_LYx}_s)Y9FH3N%fd z8m*$FGFir+)Y~LMa`znlk%iYPy38t+4KF2Wg*HhC-^Lf_l zt)PrBZtyMi@i-+;u+VkMgIk;h2=g0OOC`a;EJTo*x(gR0K~FbHAbkJ{_d^%vi9YgD z#BAOzG23f+*2_?igUap4Zi&fV5xT)Df}pPsZR!l^PjQWjrYfV<%>HCK{t$vJ*JE5` znPoUOiH+|VpsOyRnJV&FQ^FNXgep=GO9rkiwHT}#*J{rV|7Ew)k14O>byV@FD2stFujs#;p?Gy zn?D}NKaJ>K@1MMKek@t6B|5cLq1zg*rW&(G=;pl0 z5=KIP?2Iq(o|z&f3dtZ;@4tc16`nq>`3SD$+5 z$xa3A9Sj+;&{IJoQ5Qe*SlobjlkGF(U?9P~chQIQ=Roq)dFmuVYS#kk_$$$%V&G_G z7<|*SqTGRGwlZp17IHA%KY%Y0Jz;{mY8lFc&THP>Qx_#jbjnWhc2vQmi%`^{XKm$E znVpzetsYf}+*h4hq=V$A!8OT`kuXP35;%1=Nc(E@}{PGlO{;w*)?(0z@6yi zsn*Tr+G13OJQ_y^tvrVYo9wFIy_uwgoAD0bt)cgu)+Y8*BSi`WnxT7!^{Z63QbigIV*R+9V{E!68B#K zWXK#>DO(*@UJdah2tL$%zHQRT%mNzrnp~4ta$NYu>)P-fV*v_2o(gR|ga1uTTQ2uF6X&qt46n<;Ez`CEU3v&1l`lmz)I-o zEMtzht;hDg%Q#I7S9{4oGsv>e?I~a#cUIHPHyLhyx2(k4V2}Ouzm0KE09Mfh6I9rn z|Ga=)CzV}Hb5X!_QGm1RDjeu}`*LHuIB&DSG_Ajm;&FE{siX>G76UzJKMl2%_l?ooDy;`)Iuv=M=qG-r$RfqJnSiXc+^h*1}y2Fh+xR3hn zZH(5LtQp*;6Ylt3t%jBLfK4Y_!*=~waWFH=uB|=%r1K{nx5_TBx9`;k<};{J88!Fx zOittmda1JsR)wk1%qOH~uHvYwJE0JCQ#`#9s_G>HzbK3_(wLiVpqmBtF81%mVe=uW z(<}T511dN}PsWom;*AI5hG+279M5ZB@~*8$P`Gk`v<2!}{01BT7Ky_=ASJA)YRv9e zYY6UwfU`)xX`q_eVc?m_!z;}J!QFl?0Z~J-L~YEz{8gL$rI||efy;90HHAQK0CWGO zWK6YTxMQ8WD_-$~({nYRvKd$S=ECN{Ao$Zw^LW4fI4yxsRA0?7*~M;yTWN3M!+zJX zlWo=Ae<}qn>%b3FAgo!-vp@>y|hMdtYql&I?%%X=twH?hGhte$OXR&6)4yx!aBoBM;0} zwFiqcJf+*DF#ndUs#xDk^_OR4oJV--L$U31_To0tBQTbN*El#+tP4IPG2fm*hh)L{ z_&;)csc@P~U7=8Y(XoJA7Tu=*zWClroCxtBMR=FRV69@`c|}QSyGX3$MnIARma@W5 z^vBb8cJO?JFpuO!9`Ivvrp5lvTI9T@*@lJjo*dd^06Y80`T@<*=HvqMVx8xNX!F}> z%JG)>;yAqkzJ`W&)jzB^68ZBT2%ly0=?}c*N3YiSK@0^so)gXof)0X9bIHq-ULzQ> zVr@AS3P>`?mA*i$=M+cqnerD|s&pLzl8ZGAOhB1OdE^?KfU?BF9^4bEch==xsGQqA zzE%PrbNRZ2>C;ZsG%G_gbEGE~{38bwGi4N)SvljDv9rzux$(QHmhfCr9E1+{P+XbT z%0nKwEP?&g4dXB_?}E=R9ZhENA46<+e6VN>=!O;C%S^!rH4dT#nWR~7{g$AD)3KIO zg5qnxQ&=joVDe4AWXJ`!FYO+7CL6Ld26gYL9opllD{CFxHd#bJdcIjov|GqKurInc z^Ych5lyOUiy70jp9e7bqzKudMF-&4{6M$-as#GR&Uz{kJy8->V0$ z$2Hi1zu`LFIC9A);uE;&?O1Pub^bE~XEr{Ji61n&x{RSvYpT!F?TDd_eNx@S0KdAh zy55Yg0Y3Ubflh$O;fXO1bqM8z$qLM8!!*{mah+YGtPKW8$ew&HafC(`U0txcR&w_! zu_?hBjWaqgqyev#&6=T)z?^V{qF_b@!)rgcR27?l+=;xOCu(_g8wo#1*cTuc{j}#3 zqip|1!vV|fh;{F%6dybw@oU^u%0tPmqYDsxL=5AtA32i5qGdasD6UO&NgLdd-fUGK zIYdv73N(85-|0@THFsdqo=WfEMYW-AOH{Dry$ejoyWGu7L8LjIGaOWfB5<4@r^q?}}2dz~ew zeKqkg^-kc`D{=VDtS>}J4r=HHfZ?nD!7>~Q4?Dr-sk$!-(oq;2+Ym02xWsEeHi-}* zZj0ytZZceXiI@=&Kg3t6*pM$1Hlsc(;TGzNB!na)vAIX^=PoteTzPYx6-y(MOt#OY zny9o-PM)e}M(cQW7~*&jzL>wvURiltC3U(m--aGo6WkOs3?D!Q%oR4O+%^EVs#7I+ z=7xx_vkns!Pl$ajee1Dr!bh;AzFG3{{`gw&HIE|<{pySHw!0pBCcaTq4>(sqJkX3Q zzaSlSf8Oz$uUBz0cP)CnGhR@(@PANCdXw%T@wU4|Mb|k#7*E}Xe(H$6yF?<6pWqXE zJVAbKt~$OiTN}1NUr!m|EAu<;cx9i*Q@L9;zltuG4sh)~Tap@UpkX3==9*B|ZV&06 zwr@;F+dNNd_W9?n4%bd+*aLT;vZg4+*K{A1uc^T|s2`QC?pr3i| z&ls1Vhz)#CH%?1tFWNzzW5>yFQr3$B{!s1@%de7^Yu1~c{dI-s z)k^(_7VQFCuOZ%2X}ZOPxUZM>;FJK*&y`i`*NQNAu5XH)@3RQ6uj?&V-}m>P@0AYQZ}9@x zzg0Ruwi;%YA692SK2JY@|1DmvML(E_Kmh=x-~s@k{m z_4*eVFiUW&$;(@57CV?P%ez2^s(Ss`82ZQj&3Sb^Rz6-BJ?@VjEmzL)AnY;Gvr6`j zh=GU;5A@1R+)J(0^gk__v5SU=92uRQu5ZpZUy`x-cyXFv5N5}DT&e=0`te4cO&CVx z&6F9!f}?{uO#M?H1!a039po{SJvT2ZCwQ*hh-TFQ8Nr_`Csg^?^||Q72`PWvxijFe z%$P2{W833Ka0PCmJrkUlu&LDZegR z>A$Ry{sWxvpU5H)xO1B*xAgH4u~0M#(sYt6Gz@2P8$~IzZdcQ08Hjc0v}HGFpEO%; zV@{_|*dtHgBLrxN(UjN&$*s71e$gx>#+hT=#aS|@#~t=k7D`;6P$yF3fm|l5*bO|b zLGIrg4G;(vs}F6lHo9V=Gv9smFM=}0p+~hWm3zm z;c5vsTB)Kf0RGeK{T5xiO0uV?b-9DBl&3MNrL+}JIB|x{@Jo{6&AFVlBE)l6BN^l1 z;pHy!e?8>(u;h5rqL0Ltxw}HfVE4kNi6>>v7brd@4TGX`9%(*UbRk~yv5lBuxl4Gh4+(zLIQ>DEe1mJ z7T@;D(}j`_f<^DjidDR#lXkp@b?5#7rTTHl0Y+lrflVODz6V3p?t)rjl1(qjTC%FM zw25U9X*JoG`#}Ni^wb*qBRM!u)i+jUKv#lr`d;7xsjDEK25`*|>bv7{X6WGrGh&W7 zO^+<^+%3^%-fAO`*PC()vj*u539W>n#TZQ-*I>81Ep_w%135fH+1E;$;f#l}s z=f}-^wM6{BXu5I*b$=J9XCy&e3iOV*UzRG(YOdChkO&~%R4TsaT-qyizubgbK%tn4 zB(69pkf-J8Q&-!2euv1A8xlKYf{$xO;7w) zkxL*X{9+prmw-X``ZvzA!<^~IRn%4yuLIirF%qa+wv-bwMy?j{U$P^pMi_Hg;}zvA zKG6jgM=y|^4M<(r^Iy*TAAr`-xo(afg#uqU@nJ(s+KDtx{d-As0UD@*AWl+O<^tRR zb<^#1b$=k3%v`(vP}v}6aQe0qMps~sP-r&S6}x7VYj3R>ah{4hR~Lce5rY*t1x|$M zw$8Y{iEDlY56;Snf&vU2_G)jZm*bX#gP!Mw>qN?r3>@sjh`62RGIIs&@U3nfJ7XVa za<^H1BnWRgq;y<=aH+~I*XM<~l$zWPNFft|JoTCG{FH-(5)F8`;IKc<{Tjr?_^)x|DGJun@vVOn2SbRsc0o?eb{-DqOSNt^RXX~8GMG24lh^g(xbyL*ab=Si zvjj^v5u%Q8FWfW84IZ?Lo=zm@DaM^RD-kYHf&GBBbE50rfRsATnFP@SXmo3WMI|+G zyHXpS+u&C++}7gBF4;=sgD=pS28=8fBaX&iAaM@ETj(goo>MZ)*_czBA(tfU%9UwM zcB(aLA&gVv1^4_@^}m*>U=V3XXMeO#k^9?8f2OBTHiJB=RH^@b2%%);C2d?;x;NV1 zMC~{hGg`Rp4;w@kvbAFXuDJZFszyyjuedU;XNM=%N~RvXpZkezR5rgE+PctbuW}Mn zu1%C%htdG+9!tOPr$>iE_D9;>9w9eu`y~82S>~hY+EF*MtaHbSscFo(I!>jYg zLu>Q88s~O$;Tc`#^J3#qY9!N47m=clk&Xvo}O4Aq<8!aWfFqJg3Ygg zo0t=wR+;AdY|XHh8KunGsJ_!Uti@F-Ph{xEVbg3xIFy=PnISXKBZ(fE5Np}+qKU;w zZ&TvoA123v>i_k9wn~r074|q14a9-5#*T#zbgujp1S8YbHf$Cl)EF%fPhZf!45t`V z4+MNr*}NvlU-?TRGIgtpsS>wHlS)H)WoKCH23IQ(?&vUbk4&6jJgXOCiNa3uO4#y&cqY_i|vnZzUH(todH~JmGVjC@E z;1AZ8eu>D%Tv83`ZWRT4&&c%dR_3KW1gkH?J1%W^US?*yxxpsfHkN_c?jPnGB@ZmR z7fToTv|Cy_-%IJuF8cQ||F~XKnbkfZv-ii*iY&m#^fWm{ms^&2PJ1?aZsMC=jJX~I zbwPBvt^=34mn)mQ`YyE~B7(OpwZTfvSb5R!q@$lNe&Cm5f`QGdd%W`or`OqN#x$01 zUcV2cWO?O(RwAnPcKWObx_>WEc!GnWoAFl{s0-%6`eVFzESuNxf(rr#!?Zb5yGI#U z+x!n(gmyCA|NTgLz3;yBa$q-9a6pePKOweJ8{NEVDt~NPmMWfqD5Xo)N`KhFI-VV= z;ng2NQXE)wFxVJ9+bOZN+TSj$$=LV)g+BJ>F`VG-G4Sm%vi)J3C>MTMS5JB@Qx;KS z?0zFY``GGw(Oe+#2{9*5MP0%F{8(O5y=C2mKHtG3yf##MU*|lsirQE+I61l(nXB_3 z)46=3w~e?(ZolfNmsUk=q>9eJFw{Oh;SA=+g1b|MQabvcb1}!VXp-&E)qF;26}@%z z@&c6nYZ+#mQXGp+zGMlzuAt74=&c*>d&X2e&G7^J-{`6iyTe=y4gi3T_xt?M=xS&6 z|Jk)j-P&=31@Sk#_EA`qk1>2PEM%0?$m|oHGdo6Q)y21^F)D)+@2(ooYXWb52v=q_8jF>ID&=~#vKD7FzJ zNW28fD#xvctLH&s${i`n4Eo+DE#xhwOQj|zh})rRg*NqB?CrHKSrEr#AaNmZ&LmyJ zSu`}Va;hx2SoeTl!~`S~vqz=YlF*ZMH?xZ6bK!O7O9vgsW~cLI4_d9h{_0o zWz;4vUkvAnkn)C1aoAvPu>Z@7MKNEk8SChYrBXd=B!|ebn6|;M6<7>fGm7?0%n*v0 zx26h0DMrPfYeU8;#+r;-nQBtO1d=>ygBSxTMm7ui z$+I9KMj0-Bo7V46X68Pcjuq~SP&q&?a%&|lVZRuN?gJfHU)btz?ODwnxNOXmi)(-O zA`{-nm17W*9V0SOF_SHC0Os1j&(<7Lgw0h&#X=BlCw2@J5+Za5W1|482U-3q%WQ~B zlgu}6DhH|x9mRFw&Fl4XgvWL#Re=!@Az`aUro*>?|I&dF+4nr$DZkj%j zRNUDsj*qgMUq<(_Nrd7`I21-3gL(+D3LVmdk|4-CePdT__gDtsE1u!khW5uuv(ptW zxSDTm0|_j>6V7x(O#rmCHzL35smLCzLA02Cb1jGL%9o7mTAG<(pSx%rCOqvfU#ayv zm9bTdumcELAdS`SU*R&01+@v}qS`1(I-eLbr$6{L@bQl7-m6z2!nrrsY&MV>8F}tY z_y!aS(B7I^9tw@wr(fHUjs1XgUlDD9_%EX1dsOgie6pV*3zPR8mCeb8eFU=x$j8yT&>WlzQVnBMd?E>#B?1t#D)#)2AT(1)$Kj+m?5#A3}l}<}r zF3LMwz4K8Z$&SWU2>Xkx$en+yO(Gsjvdn?TJ(~Cr7`QeYAJL5#We+&sw!6R_%{OBa zK%461C_}^x(#jDVu$$~e5wM#F{V#|`rn<2NlR9nq$(`J-AFUslg2Ssd=qX!L_rwvD zxYcJAiLu5_V7V{QAvl9D$}sXd7|lAZ1k)n+-wJd~Yf1E4Z=x83R;?}iXxfj$aCWiN z0Dm;09>m5jiT+SxyQ3RY!_i)|{BL$VQsB%(MYyiLP~!Gvmq_X4eV%fYN5T@a_RY`4Xov2~|=@&8}Uv^D+uX z8Kxr}SXe}#p-!ZyG_>E_`(1Od*4_()Vmi?Y`j7otOIschub5J#M-NYq;BTL)__rxZQYmBbSTI+bOAE zlb=(#YW-sGl7dl_4-y*&RWv*r_Ugq$Gd2nnO&>+;QDjMx9S;+US#daS)Oky-m*EjR zHnr|Mb|mc5SmcbT7MdOBz3=;H{hm6&{P}MQb2)bIm%p%?43#Kc{hK&v%wMwoJ%-nU z(pmQHIv=0RtOA(~&o-jG@^!x)_Er8&k@>CAl7^*CrVHkEG;W}~R7*K`PVdXS_%w(RvUFw=Y z*Ia~&c)@0bR+Ml(6-bzVw9ms!OflAR6)eep!$>7<{nU8=_ywP=>coE!VaxnGwI2A; z+Ou?d-FX=Bl^xEAIT*krJ;)Yy#c=^My=X;yS+?D-)XlEa&8pJv*1+r5fCsx^Bkb_O z!>g=02z-rA&3|2a-6&w)c>y=2)%AWEX~p8s~1aD(ms_U8+k=8*-`eXnSvYs*S7Szz5k(jWZP z=4KxubI>nkaC%L~AMt%HCoRruBKeSL_Sxme=aR+V$eal?N@U*5q_MH_anP^l?rLb% zWgwTO(bkH$G>YoSf}vP7C&^d#bPM;0Wtw}(j=Sak`EV<{FC0^FqAOJm%|*^o{leSU z>w*(4MSp8G8aFDcsR^$f%Myz=a=Yq}&iXKc=pcl(r4l{R221mHPA4}TM>Rd2V2N#q zzc#B5&Q0x$XU9g>*8!?NeXZaGhCr6&Y%S@u+2vc_-iB;;=Et^}#gQ=^CMMTH=v=Hu z!Sq|`(;U=$HzQoZ=TDfTL-aKn9d5| z`A6PC+v@^P3&}vOftKet{^%l(hiWs{lDEMQU$eDs&<(|K-|BPKY@mm4lWm>d&}xf_ z+=uRq!)WwRgS;zRfv4f<`0I8b!< z`%NbFtT1#m^6ViU^up@#DYPo#vM4)@Kw50fwifr}A9sNg_-<~e$>q|Sohp0bN*6r3 zc#2(5hek2h_O4QU1e{cD%geBle8XmG2JEmGRaI+hYAi3krygikFkWNmR{d_;m-Y1E z>0UIhLV*TC*s$|pU@u?SokEQQR$?}*HF z+ihBzH*~Et)zo=$=|IJ{;qyZY?6fm`bT9SP3u1&v@{zg|g+VP&@kO+v^t)Y2|5_(q z`Xvtvm@`+`_J#R6m8=rOC8wUtFsRb9WZcWtB^(p~Ok@#41+1%ps3oFEu_Y=UGE)5* z1QcwMY0$l(Uv;xSehvrQZrGF%^~8kANc5h9&se4& z(NTT}HUlkW@|mW

rAHOr`#ls4@Vg@8OE{`}e+NIqILU2V(r{#_k%z5>TMU9!N!A zgKY|7f4BrtX%qC76;u0==jMls$bqLkF}GX zP%kf|ntFATnq;wISsF{mT>;8nm4n3NNv2Hq8lu)84K7Wlp4-ba*`5AtKScE z3CSlbcZN&H!c0!d@mZrn+VTZ)oo_8@@9SR#Rv9osD-8q+GHd)IOUfmgGP6RCcxvJK z)CZm=zPj=p4%dq$v=@`Byx!zB3*i+g#Od1|=eFi~CeHjzy((YRCvVweJuGANa5|$R zhEKl{lQMNCPLF@6j1G;<;S>hUSM1kGLZK9&Njb)G9v5%Zr0~UOd3&--SBe3$tEptz|)wHJ_v=2Xfu2O);CNiqSFLu_U_+gnKHY#l4F8c}m>=rW&ma}Tu@ zk#v-1Y~G2)ke-I8?z5$wz{OFXUx!H4CJ!e#J7CFay%L)(;H#x>-TD*leZ716L1W1H z20gYh{$|d}2bYM)6?r(W>UOAxk1QHs^E9>aQoRthRL0Qd=8n5h_M}LQ=*$33&>hEa6+Q^_Dk|9{-g;{g zRQGT)C$SmbyGa%*w3Ul)aBgDvg~e~gJX3(H!1tKi!_-@aHxd5aTr~hI!Cw;!0v=1xYMa;c5-qM9Ln0(yz=qHrn~qVEsoeTE z*j=s8mwSFo3eW-*@DUe9lemdpME+2;juJMB`FFIaYonC+or#t1mOTlDs1rOhlpo%> z`8y{`UU@=~!a8wy68%7`R8N9BWj&OX@1Ek9=m;8xf(GjmjtA=w@{jvw&9-{&XeF}A zW%`$%HuX<_45O4ymA&GrU_SysELr&d32Y^)HiS&z&l^Ju9*+7uxr34yD2E)ww=8{2 zH~J6zX~S?aD5@VG*ZL>N=&Q@dQVr z4%xm2!!3d6gRdlom{kh-XPWf5f{r+gk>yJzAE}WM*MOD!GFz#__6o5?rClRli;b*I z-W85v`D#p}MrsSFQ0uS-Kn)rrEazJJ#jZRsWl9#u>1Ds`JU=C zC~Q$tC2_oD92*&0f04dpr?Pf!FOcL=YlH8()**?sZs`Zp?O=LWFPwmR^OVfoOpUFl zHw}*i4|M?6X|bke=5VsI#eY5s)U8@^QA+PMXlqxkQ1D{amUpRZb_AMFz`ZZ8(u^yo zTnDdo!WMa+76_R4eNj*~tQ|+&uIO<{JA;4URUI{IWM?OpJl!u=_Voz3w<~4)gp@4Y zH$v-D-qvVk(;Q?E(+y0ApCv<`$%IM^!?GgLd}{te_g#x4IUB|h73+ANO{TWb?9AA$ z(p9W7W%Mg_H1*k7ud`LJOO)Ta3T8OG)0v-@b8Cp-6E(imj@XN8Xg=!tyT>1!B6DK1t_bP99X<`^snVpwqRpr6~O z37WR=JVIazfjXKES0CUZ1OcRNRwQROubHoG`1<`>QY>G@4GLf7oVkocwOO{;$(f|_ z!-+b;=IVF&2JJ{2la6K(m&t32Ynm_>S?jD+l!_4!=AA>hmg>mfEq4Ev_UVp_b2K9f z4d$?fUJg8=yWH%y?p?nV;CUHks)3fa$XS+dU}Pq!(;5k3mh`}>%r`ESoWa!6^e0ef8fQi?Q^bsvzEbb6ze`t{AiEIPBq_xJ;Q-3Aa~r_3Gx83fJah4* z*Pk*6D^hcrjPDeqT6aHRM&o>LJBW{urcO)q{`pa%1>odEXz8!RW3JYl(#bMSW_X$xVWs0 z-Bm=d7t92rB(RQ%6EbDScBb8PY*6C5aJe%=Vk9MucliTDn{+xhPgdUDd?9dTXj@tU zVF0~kkVQXT5iD$yspxjPq#wINjO*52~pH70ae3Z^n zTA}C|SSarFe54sjn`L$(FFTz#S1DuY^7jsq80CW{cIfPLU!5WQ!J{ z3Iz)LB46$!ZsBFYjn|>%tC3CX4{Yw8a94c_%&7ukL&OW(61O=wPFAjxBhIB4(6@ZXW*umIztV$s zE*;t$S2<19QWhv|s6M(|P$nC;R}LJuOM$I zB>2y;xInPfUUE4|cT|+jmi>4FH94l9L(sRngr}3@s%&k1&{R?GV$n}apyQ%V$ZuHV zf~+&=1>UYZW0@$F4Bji;Q)9&8Oj&P`ViPV@?iMwb-^!?;0Tn%9TX`8mbpnKOFJ%y)F4##mGMZcMi6(VW>qm-HT3K`nxw z;w_#bk=@*lHpvq6^r)KRyLre<8u{WexJLzlc?$O0NwVh=7f#(toA!kvC&)?|SEXcz z-9}^np`}Iv07TPK?;@6&bT;T=Z}x*N^5noknt92E1?R?$m`(L1nzWzA)h3gQmH#PW zWc2{BOHZrsf~)tJj!XTAyF7CF@#a>T$yjygcKT@UFN2!YiI~o{Mq35@5X?yceB8?k zU!!l_=n==X2c5=nc%l5G2Q z=Ty}{l`mk*##C($teO>K{HT`>?YBT%hMNAn(!DM0X zgDT5FeSqb~--HHfcC0RPVFLjPQ6A;~j=-6Ry2k&-Cd%@Hg)gGn63<0T>k479&|->M zL8<1DI#mzH@!Dnb!YHo~nc}=j^2+AGCp(iFxvoDkHpRI*9TbYfjjGAuphnq&w*>Fj zHoBW6oW^-v5^|YyTS&)B&pNO(Dabvu1xF?Qf|jLju-=E;)%SHB6RcV5lzqnn2)=A$ zt`HyPBO8s=-I7Yq3a7k+-JjWsw7`ITZ9{=#jQ~wDed@B45DGmw4f%Fc^YSj=f$Snilnk%f}k`??ZS-YtWU$DmnU!ozfrWpFmEd0DUYek zYH=Btt}ByfLZu1?7;I%Kbu-~M0~HrE7_&a)Qxtdf}i_fZny2Pe9S009v*e zjo{%by>+euCMJSjAb@+G9l`(JdnV|x8+vf$xN2Wy0 zdfKOKRE(qd!Ln%~q;^rx4(2LkkgCDOwF#%=Cely5Fmj8gGCY0jIue%n@Dr5rTLEFl z8aY*46svnE8B88E>jHI(a9xBlGeJovTsU~sE2#E!BLz9xFy!PXyz_HIH9my`HB3t` z0)up4Y6VZ$qq5LdCd;FEH<@t^(rG^*bkEP()u>+RN~Ivj1POeK!h zoG~WnRPY2j^5FgB`T&O>T31?XybuF-s=fVr4BHFoIUa23U6K7+cdp0_U zKVNY-$^I=NkDFyUFJf)@XQc2I;AP^Scx)O0lwn;z1aJOd^l4|D(PUr<)Z$vM8jXEV zmWOrL09?RPpE3rA!1Te2d~0-pGd`5TWi+xI{RUO+Y+JXoTsw;J&HZQ}N^slh;BU7` zN1_v{C$$3c#aCRJe~(E5Nc~$7R}lM8)JE*(PQ6v<_SbE-vgiU}OD}e4=fgv9$k-X? zlzn z;q>s#)hWST;X8h3`2&&&9an~U{VY_q6Ii~{c89_hx>>>|iw%#{d;7}4!C(YOVO|UzgF>P3E)LWID*~%;?mL0tBwoEb@>$_L zQb`PS2|Xlw>&ycF6M%tc?h;4VL?w%zsD=i{1?!BJ_Re>+R3RjB zl8rj>h}qeP5Pl~Z!`V)Fr}WGmtVSjM?A*AYZppsk?^KYrR_gq}sY3ge!q)7W(t%06s38w45Kx09f`T$>4OYV43lTgM47ucH0TYR`!vrzLt-MIz48_ zsguBjf*8J}xK`Ci?QW-CmTap5Y-Bbowo$HrJl!NTB_*hE;1r4#{W?pFtSXEdv)5i5 zeKA>jPMuHkk!SI|jjFOa+Qp^Ip8wM75N5{~_~J=W?ZG8@naeRVCyuJ() zBehm5i`J{|+8%LYYpc@YJ%;7mCkCz>b533QOpfC@zc#{TkAw2xW36c_6G7X^D5?tjsFjRX0&8kOH+H6A#P7y%>o0*(_5m9v5}H5;zrk@}6)tm&y7)~2&@jC_O{^?vAFr1Jb0 z_RU?NClchWnTwOp*XrD!sYnLUCV@)I&Ki_UCY z+fAtc5=6;-bD3NGn2?@;7k~v|5qjPSWxv=7w@}MEE)=~)@iL&t&xk)U737rfbbm01 zA8tp^$^I>*DZlFjR%OHMoTGSM@P@Ceu%s|r4nw8CL3i~}3DE%vJsc@Mv^$BeU7N=# zTJZ!gNtSJP1`52qUbH{rGlCb8sHrVQ>lc?4?8EHbZxvd2cL8D5T?M+{Xnc^iR*l*t zCN>w(yX%vLdNlSIs_L}L*w2(zv5Xa#VU%yNui8?tPfsp3sd1u}%2^dQPrGBfJJdSf z&Wbe8N*LCW(8y(#|4YMMfK_Ua^+NvrIBGtiCqJB_0@-DeP$Y5k_#^mLo5NINF@kTj zR|!e$==8` zeL%gaB+E<{%jlw6%gpoj4C zkPsQSc4kk{6W)H1IWQUZwm`@7{lVcInlln(GS!t|Jnzij$2b;Mzy=zBMHdGHz4w|W z10&c4Zk0ohP>>idI=CZ|yoE7}1vBKDs%=}nud-BQJA+UfX-{-2o58jSnuV*>c9%FRQ|+gzV( z!c1>pSYPi8+&#N{-dLXg=vtO3v$50lG$baa5H-m^%uBW8MBOMbvoLyHghl-*Fn2jF z|EptbWH}bQtE{r47>PQtX%E>m9vpvJlMUD?`3{3qsvOWr(XU^9fkD=@<}(A86$d91 zxXTp3K<}m}5MI?nT$ialO`OYvR60XgT4}dPDGfrHgb}B}6rT83J?y5S00*a$2Jj-e zAt?M`_R%gIj}8pg-G;2QH!2Dg5w$3EuG-M3iTt1ud7&Mc46+y>0jXZgB2Av9Sk`MZ zz6fG$XoCbY_~>IPa<3h$aKm4o<5EI}Zy|N!tR=YEKRuJ?9_0up#YEkrVDHJEmTpX;aNW>hCKI-JWOqBD|{-mz~O^=3{pd$P{WPn?HDNp3y{J_`; z_2?2s->x*?x##IOq7e`I!3Kmzsk-R+q0L(n{e5J|%cD~W#kH}CV%zW?6Z?XP`P;}q z;}s?U@HsQNZ*%zn(*m zYAky@<;F1T)8UP zy=azl&qW4^B@9@K!tX0{Np{MLL*!6GvK}2>J*mh8RyrJa*3@w=Ram(`6-`)rio&qS z$#`RA#Ory9>4LaxE3w?IZCt1DV($W^N_r7x1;Zi`ofs;d0cI-RfR)&&&- z1)NV8^F8{qo0CdW!XDh2caao}0*&09DnDXej*-WQ$bdu=K;b8rsURREM&Q4#IDi%6t6rwKHB7!TesK;Zu5L32=T-!LYNF7w(la^+=n# zeXo(0_KGWFGq(aw;L*a^cqsdG1aQ@C6a>G%=gw*ub|z+m3JiY#bN48@6h}S;SzCe%5sS zn4X4E8p%Ngl=NPe1bo56va_arxKZ>l762ERwsdC&#w=2n+|a0b-WgMT^4u0t&WjdB zw7%uV~*FK5G2q*oKr5hBVL1& zCZ65ZGujX&a-Vfo<^w7*qQ#o?bnG>D2-R*>lt-adP_5*rWA`WjJPfR z>Grj|01m}Vp5g~a3^^ONoE}5X4Vp~{jKJ2T#2j!IBtuKJxH`vNO7*?RjTYCU{B?*t zJ23MDW>wUB6qm@c%19HR+sNQ3nN1Y~#y}4`OpI7;&WPp!&oEntuK+2`1v3!&7@U!o z7gCDwcDi@=aBdUqL=m2rB;}1Y_V;J9L^iR3)--g87isARgyo-+A+;r*AX6Ink3+X^ z8rDe_1;ao5jhL*{Beg`XSekO;IpG3LtX}OXvzguUVx@A@Y0`*I>NV<(mxx*)v7^Si zfb`Bq+ztql*PihJxbz%{J_RkJn!3e=tqkLq&-Yz2V}zxE0r1j1EzFo~>M=Mm7TY+U z^1TPpGC9aqPyy~L9k&vio9d4&@DI8;w4thjRwbHWBjvQrV~1`7Uxs8u@O-V7*Gf+h zHZR205M}qnmn^d>w8S?QBQ|W%7yELDphUZ_5K|r2OLA)=nw&e`5%8G0n>Ddz$Y2VJ zN(mfRZmhV-diGc+HR=n519{$cjCMLeWAh8^Kn@jXWZ&IFLQC>~2bIaNzvYXSsBaXs z7RZkL;MF(&bcL>H6YkFuEaz6EA_!XYktfI{nmu^r9rDkmlh4z!G*vjB^^ z5hmn^MRvB%;s5O^cZ@JzYi(qHTpGg{3qUYXth#|Ab5zH_E%#|0 zC#y!R@`U8HKKfx+FhX?R%u9FyYY*Jp9M4v0F3JavAzI~Hir#{m|(7+!l$p^WCb#Oj@%SV?DO1 zLnouRE3C2i$SL*d{+`v=*8pnpP1okT+LvT05N223(uen`-N-n}==PCJWn2(N6BQ z%eU6M)^~oL*S}^*LzGmYJ+`E&!P8-Loo9`ERTgo_>guW~`uiWG0qh4jX$Pq%HRSd` z09igiUWcl<%^)ja4Lmr&xR;Nc7){O%P0}h>a2QaJHy2mak|y}EKaX2qS<9iXPd*o} z+fb|LKNNqn$Gq`RO3+3~HAy=*MEMf`x=|t+-Rv$c>!4mvU*9cvsCL-qdsPBfWnF)G z=g#qX@@#r6jK=~RiJU&sXA6Y`ag$zpMr(@dogB(e3crj_d^!^A6@6LB?yp3 z!Ln`Jw(TyPUAAr8wr$(CZQHh|XE}@cgFLxOyc=;*Y`x4szrbAwqVY7!LAnlK-?iK# z1Y)Y7SK%(N<}pL60_G_^ZZ}o|N!x8;{taAZ(cIz)a5xUReoaIJ5r5(63Tg1Qe^53C zxqiiX)QO}koDtuePvOB=32xp#Gw0J(5D)v#E<9||=~D&tcTmH#c+{e~X@(yRw|JCs zjXwG>o+&EKWBI_~J|E5|3Dp!8Wf$W4zaJV)<60&S314wjf&;`MX@!6hrSa$PkNWTy zRymXLIPlLSi|g0Mhz~#zZkO6GpV|d;XU`u!KfDN4?cafm5wVs)Mn#=1_AA7l;c#Eh zb?SRw*ITFW9?Bd#gbsF>jzyq-X;k-+nS9ntYyaG)c%<-E&8Io@(4d{^Pt_Fbh01@8 zJBFw97mfBP_KjQlVFcpt=`55o8<;PZ#a-mifiFMI7d&0&OO7i?~W8R=4st1{t!=bQJmke4-Tknv<6dKTe?swAqWg4?KW)`b=C=mj)tO zad}oZ7tBknoOaV?e5jkc2?M$o8cjb6c_+72noCz%_b#O>rH;AJuiXpTZA< z{zce@HmAeplPk@VtRokOxKCY^!XHWram#NGAkBwJ>^tB&`S)k7_c^qDfqYO)=048# z=uTDPLP#wv15@XYE0}1sr$j)mx z95_~y*?0MQ8r6Avxya&OHI`qp`Q}2BTN!IewRx5sEw}i=mPQSDFJPSSEkYxZzf_~J z7+6oqeY1K;d2|L;*7m%j)f;#CgD=t6M}+Q$hVWN1(kqq_UUxZX_yoqQryNeeB&O=< zbqy;fq{W%8h+sRoA4X{A+*Myf%5t@uSl#Km8Lm}!$KZbZ8~JM>c<}~~57!p|8ZSqr zHdaf#?n^r7rFEzIY;~NU7-K?n(Ro1|zzMRivicBh1+^d1S83p7<^E-5=H@5!Fz~f* zh_8gDrDX?unC>U2`LR=&5Vqq4V`Wun;diG&AVO)f=L zFVU_>Pp`JtXF0_IsHk`>MuiP~8i+I~Umv*=V$Z!dH+Lbb?lXE?TGUMT7@X~9Gn;h4 z_E~!Ij75*eARVR^+U!5-F;X#xgHiuL8l0qd_$~!2wvAE~@IVo)GpJu0ZFGXXMzI{s zLJC9A!$&q%+|_39ciFk2`ak<-3t`&@8BS-xErPQTU(pW%g42>x;MRW(%VRfJ0Jg9m ziC!OqM!!QT4*c(bw18hNasPh{^HU4y;I4WJ^4k+n5q=vAn{n0bL z)U-B%ZqDpBX!LQ_?da_7)TLaS;a!^u86Ieo{~UDOGN_`P@$e2al}f(ym!m|zql$zh zEN^n;qetKpuWc<2y>)SwbiMzx?wt^^js1VG}#Re5jYsm4FUrEb57h524#6 zsVnA7M_c41(82;@^19X|1?PF2sMl`p3R<5m!t0K-ju@~Q-&smhIQL6{w(K@T2#d)m zgagz1hl~$TlbaibRT^wZr=}L+i6Hia0YZTjOXq3{z;)AX0 zdedo*B+`mSK`q?sPuOvFw4E_A=^}O1HB<52LfuojB~!dtcjqmJ(qLC$Z}i0D=jY4$ z_Vza~jPv794tf7)s$AuA3ufBrU4iB{izGEx8UWI8u!2Y*0ll(=c3fY$NEyZ{sDc2E zlux{vO&1BFdRBgG4%)9#t<{N`U$>$t^ng%l5mZl6M$moiJFdJrS&9;85}&+o0H9yF ze!E~O_6`UbmDpIU>c|2SDP6?3h_=XA{u#-{GX-2cof1Mt&M=YH$HXDda)e|ZS;hUn zT9JSaC@S-PZ!ZmWi7TG_)W7Q452*$?%b};=G*iG61&BtB1)c_)*>wlTokcknildvG zec$b+KpPfl*N-UcRU+Q8b1J%Wah#rwD>@G5U1-*^(HME7k;aLiReOo#f~y39C?p2?=%+U^K(sl>&fUJRSa4 z!HMd>O|AHCo1i2@hZt^lpddU2meZ3+bQz$Q_+CE-t9orVArc| z+MC2bZ2VK@kKLEhlTEB#q|MBPB*e6e1PLb`Ry1)LZzW;ArQA zSkdwMA)3niI!E>UX>A@M2ds^rLH056{vgbwu*2s@DO7{esxKQcMM!1UHb&|HtVU5p zOfbStOCk@N)uynG$K!ubO~c;k_Fza_(BE*5sj%w|hK;86Od-Pn%NOKou|g4Lv?{73 zHvM4-(|asA%L?mNW5mqVtWH_z`FyM&u7cJoetxZbHWolPJYs*5>ip-Pz7>y{rv;crCK#<|%x7k|m>=x3=UvG7LaL$7q9l>aVN{KJ z!UKh4N}1GA(hkCe)Ji=8H^@5-6fh*NIL>MX|b>hg>hjmKHps4hBzm zI`K(*<<~$y10|f-eorWj?~^f?b`U%g4Zfp>qN`hyH`5e1>Z2P#cEqi1oDX+;=3BI5 zQS>qJMn_dCZUmHfh*F!8aloqd*a-a5R*!(?(8M{uLEa!Z4_BysXoQ?9j(Io~!XX;+ zPn2y!rNKVujgJJ2L9=qYrGm_)uXnUs2Dxnf1>-JKO<;kdxx60G=G z7&5YfC!BGHpk4;vz@~xq;hPEAm&t92arWm8p(ZzX-;w+9zMElJ4j}EZu%rt#`d6fn93YX%_+7ewKj#h^TRkCZvEBbK_Qxz4WBluA5Aojn?ou+YUB_#tO+l^ajNP;gnF43DLA$;sD? zmN}1Dd!-YuTk=Ixbc>FGQH~67;onzV_bM0T*MfLJBT{eLYk2l^gWO?Y)zZ4C8jgwE z#?;D8$MDQDPG+pQ+wKPfQ5WfqT9*u>h@q=96-hG$Vnr0`yXCBw`n82ljv_!^D3ZF5qeD|+>j-u%_q43YM)P`lzHh@1 z8#z+G4204y87WY-Wg(abhbci%5EqLlp>J+U>AcU9m_1R+BLg**kWqA^%0Xtgm}mQg$RUolZIY2{GeG^eL-MGy|%c>6Z- z!Tp{ns7GR#VqToVJV@Gi13bJhc=$V08Qb<)naSa-uT;G+^`U}&*a2jSm3u>x>bOW! z%M_A`+{)4g3bxZ^^Nb2$N77jj!zxv&%ouLdGwGYN;kaEbvQ5o7$?X1Rwq4P2K3l=v z#L6RDhp_|Wd^LOF^70H*dyk4BO>3sBj*U4r&1+uDI9W6(fnw8k9@F()~_r;fYZG})KM?}A*MucI-QaZBS-qhh4KW8&E` z$^4t8m(+KcHeir%IQJ$%S?Xjh+?i?$4HqPe!~XQ>QNtVVMR*rgO=w8mgywT+gV|A` zk;*-pMf69X+R98>%rOB!)@nJst}o_eQo)Tcpn=;Bzs7%V?y3AtpC zK#Gw}&o0C*)wktw$3^oNPwm_DV)u!GJ+zAv5A#g(vw)gPn8287FPDd~DB6>h+Y(te zF>%r!lB(itmeJQ=?y1-R6}OL*5H~5SPsqPS7NESU@|j>$Dbk`cT|F5yMye|_>q<(h zfreDNA+hC(abS)Dq>o%yWh5`MZ0kdOa%Is4{qZP)e6K*8?=bj_*5lTw#|k+Km6?yz zM~32Bj7IY1&;FWVg1mwo6N|2;d~Z^D*489cu|oaH)??lhYoQ3p(A6tS^pw4GxAMlm zp`@+5Q3J@?4Y?!RZ*Cy+q~DTe9`S2RYUZIGM=eWigioVrx{^G$s;nKGTu-T|D>P#Q zz^A^lz>(u><)zS{m48bd^vG@|+(R;3cT8OL-fA`>JwIEBzI6PVSSY0+nT~{EE0B!4 z*yyh04{$LN8L{YEZv<$hZN4jjsYN}|H7n2@=~yz6tOD>N^Gm7-GA0CQZ*!~ zEnYz)-eTva8(CuZ#izyK@Vt@~sUtN!MysV3rehMBCtg^%2nfVbs& zT~D33UXxfnY1Ax^#lgcdkY#6ui{fd6@vi`rwMm-6&-WjWbkiz_YeQ1f>?~YQYm(xb z4zyJy-yC_8k{SWTT&a*Mb2=)+G;9BR^b7pIL>L0FQ@I!*06;6y|5Fw@nf>Fzyya!& zf*26GAF0XQv>|PlfWt$-nQlya8_xc~QZfk0nAg*G{QbIa2)2>lVo04C^?W>8cGd>p zD9h*)sYxEa3sZC_oe%zZ0z-^St2QtAjY>+u;zBzKvUnAWNPw&cIY|3Iw8)?)5mjS) z;Id1IEK6jdbNQvq=wE%lt{VD!f!i-8AiR952ng#sT2UxV* znO>b-H^8PNKt@uuvelb#+*&Di1Z|jU0BCz70p!{FV3#EmeV_GMJ`=LbE4FW`p7d_y zlc6izj5Q!>Ig2ib@mfMBq!I?(JmNAf2n53PjS>ZC^Sh@4_+=mODD^PtN|GW z>}lE>)U~4C>J4=o%aw>%$l1hQnwjwbO(1W z2xCAKy#uJdouU8VttWm(+8@>d04M|p`adw9la;=~e__0CZ_ll^c>0fDFQ^joaFX68 zYtqEb=b`juQunK9T_>99DR%>RE-R#~kuY1bv6Kykm&RNw03-Hw>#w{H&;*$tKbg z(G(ttWZ2$WXk_Z`PHwCWYk8DN&CufB`+NAf@At*q)Af6~jRehwe96-cqW7j66Ln-o z37Pk^iX!Uzo-lA`b?>3zozB!yw2WM8V}+UIvmcf809iFY$f_vn;;qznL|dAKg^0=| zQE4bs>;f5DVCvFv^4MnT-!SZoqqZ@QTfw^;6cu&!r;6TiQnhaaH*AvwK|38arrhRR z_e~=un^}!SacAU*OFWR>$iK==9tdA7sm=1n7}mdh%RQb2M1KaE)7hA+%eQsCS<{2e zxza&q1K>SPuAV|~c_FT59rIDTt2{sGDh7kkd~zezMavO`%8pSPo@=OmoJw7*U9uxt!3Sp|_@q1vkczH3HtF$L6_ z?s1)U0cK^E3wDtG>>V}CG?xVTzbj8-amum)Am6Z5n3q?&BMKd=_8*pc^Exh|Krq6DyDKMhJ#mxw}*IbIWW%6z<(=xy{Z z)7J}IL_Y`Jd5(gy>;^w|(HF#?xGe}{cBG%b2NV1in97L(E&y8dr9Dj{8yc}l)E;$4 z6a)Mzpr$(TV?dP9&^Q&B7F~wZz@vF8(3c03zvErJZkJb45OaZu0?Fc-*h9aUzv7== zqqFzx;E-mLA@Q{@2~}y>_#-^;fp>_N8AZ_QI;?7n87e8vs`4yPVRZ|1qFMPF_vF zW2 z{_B3v;r3@I)s!<|{IhWWIy6@tAO0r#MRTJQbd=esdZimReTR8^ck?P2et&RQ9R8Q@ z-Ob(Gc2@qCHn6NMn3}6697Bok7T1y!Uv0d*g5rXTKL9cc@C>LC-k{)z>Z2$r>Ci{d^fcE>z)13Q@b{5cYdRi zjsMs+^|wp;mF}_I1|O$4zgoq?u$;Wq98p`8%uver$HwcnwK5^|{ZsjM_Nh1UJIIto zPQ06IquR@Nd2+kNQ=qmOHV4|u{jfr815$G5Dwr!T^z}T~TXW4DXwPHz1EDiTrEYe5 zzm;;wm=kfux%dT+M{88~YM18o{lq&nQmqliQ&Ai92~WwW^AYry?Z>e0_vh|Dkv|lC zu-TIbeI7ObsLrHA`Z2sjTL+{`OH@e7l4ug%4Q?aCoC(QH5Zg;Svt8nF-()EBQ*Yv5 zl=9P4i3m|`OHCSaYL_eH5~Pvj$(c+0s0jg!@e1d2SEsAUb*${tEk0x7t#*kujOT>E z3G{;BCB(bDM?+J3Fe$!-g^}oyXhj8mi*$8iNe(-)#@>^b^p&+*K@%>hT4YowQ)+cNL2%X4AT}g%^GBRGVJy=$q(?i(-Dc6LK(XSFG?-kOzQ$?P20;$? zEh}+r^E7@7-|=q`+)Vhqtwu9nCL6lBxki)tlOW6fgGAAieTKfH^84SSVP}L;XR{J6 zT;UuZFR!<|@xw@RUmuUR(bCx8FIVyVW=HsJcaZ5`KeVC!!*le>sGP|3Dcxh}gA=jM zP?^D)-EMo6e?Wpn2{;K~eymc{D`$VHd(@_r2XR4(i;zwbVz4x+l`(ba={hO}`f&iHyq z#adx+zH~8Sv9wFc0$YxI%rE5He*XDxY%xUK7;;iMe*CWTb$neCNEv(lqR}9AVA(Ls zTI@Pynw&BUdOK2n&*fP1t?*H4u6v9J}qUa3bIh&^Q;h3iJJq@A=wS?gXylcUH z9k-igu+(A1dR%v+?5%CCj2I;3uEpxRp4?EignraT-Lhd=A~ou|sD4S0erP8=@?8S; zlSfVIR9yt^7s2xEkG{Z z$7we$AJ&xi)zG=9MFSM}o98qYdj!KUINJl28@}M`{&T*0}Jyhv=K zC_T=emn`+#PQn+16j+=rMa+}oaaRWZn@`sH7|#BZ!xqG~;x&{iX&%sLnd`axGdqnqr}0}oSvIjII$zmMoy?%DJo zo2&_(JMmkdu`BN1!ej34PdA^ojqgF;l8;X9l8^PY<)bFnxeDb(J4DI4nOU!rFi@GM zAS|JME3U^4_1TdFrNFVJbi!f?5!{3fQOCK3?aXff37cbQND z=_aby?*83tC44KV00xpI`ecj6sW26^@(#~7yaMo6I{UD*>hYJsY)$lN%$meHplNGw zH4mu94_7q zbM)Xls$KNMmSkb#JFtb9P&bX2D{wK1nyH10yf6#k&PEZ+w>c1>EjbYBq^q1R7|ECE zsw65y?)62NYNCQSg6q%ve4<3b<)CXJ_hy5f%t#A_eeKc<5|6n$Vn{i>wVvp<;x>Y@ zFZy)0;vP;ySeQ^*WqL2*b-r7X>5Fa#pv!UH2#=?s%eCDJo!NEDs9UI2o==09Zo1+& z!qE+!_@1>0H4V??^q0E+AQc%j{S)w$9Z$lScRmPDq~a_7e9d{%o6mkS%5=HR6PD%5 z9(tF43?LtLFZ)gr%(~nVGjPAg1!Q|ZPr8?S#c70-Ex)g+Ge~>w)VS;Rhp6f-6{z|B zOrN>QBka~WGHOFnhj@a*`Y$J5%aS^3JEfM~{MfrCXtd!;SXwfL8=t%;=#+oO47Wt04+P5UxXq74vxOh%ymH9sMuB2;zN=K`pIj87 zC%qN}opsYHGk4&rSAn%N6!`M4q_I>L@=}Zk*A{&nh;QOYhFT@}hnK^2W98Wzxe*%c z5DoT4svdI{TMu=@)99Sk13-=7;XShOw!uqnQ2UL$!Qa>;bk=oY>;^s6@UcI8J{##M zr^CuGe;%_C92PPuRXxYMb;vWPMLKzEIz; z*M5;Tvt7RIL;mo!Ot8W~M)cNFAI1ewR&I?$f637J8bw?=i&lW<67OX$W=bTnV)Ub0 zNPFfvmUKhf8i@hBS=F5Xh-B*7U(*mY1l-wsvQC62zuP4D@^%BP9}*Xv@jFPlVVZWU z;}Dbi)!T|*FV{xF9x@D%MfW zRjPPIx~pKjqRQqegnGg(#4@3YkPG#dn61fq)lWb|m(HXHD=ZliYTD`_sRB9nBpxyJ zxzC0G%-A1#J%wO(335_$p9Zp*L3HK!yz4C8X7ZGl`_fVA6_2#>);~Je5d8H!kzpUUZDj`Y1l%+I*7>qv=VJN>_NB_B{hcgFz&sGg>HwpCYQFGeG=(oR0&V_Ch5xcLA?} zz9aykKkW$Xw~rA#gGSHSQbirn5zey+QmzL_^PXmy(fhv=r;5JY^9!YW}Pb>pG zeVn5`WiCq)E3*(tgmhE*>Rr2!s-Vio<8Z+23Vx#B zakOZ@Wh%REOq37sDS``v296uK9&QY^Rb}Su+Exm9J++OmoKml+0H0@MOzGfiy8t+l z&2O1kBj@NE%KG4Lpnv5*4}CF=XKA$1_Nm#8uzHJ>+g`dOKkF$(e7-}JWY-v&bc=@# z)Yb}5Du_&q;SJ=+PC&^*XoeNwvP3xLsu^wW_bCMd2zVn$s+wpv0rUq!>o2JvhdOY# z1@)pGBpTCTTqG=XMBbMMaQ&1p5gr10Hfx+N!%%qx)4%jLH>%A&-s$8PFANybBdAdE4I%n- zv)$!+NjQ=|D42lQOd{oJ<%V_JqCr*`qkja?DfWSlGTdqX6kz8c2-NfI&1fNPFhZiG zEv&!#q#`#dT!;hHDM$cU%u#7lt(zbNTYj$y&rVvmfGN=p@thMZ-4VrFQXDeaTwXwe zTHpcnASXI}2jOUs0W-NaY;B@cPHu|=>9ZxlO3i*UV)~%$Ugjg)q8rD21f-~U`YbB! zexSK*N*M%gtkh|qsmpE3Q+@X%2V12cz(P>=FNyf0qDEyFS0&*ir9H?E6u)OBPLU7T zAOc6G3KLt^qC^lx^pOeCgxT9T!R9!0+5gxg0pyn4vYk_AqAknxDF zoWc@ci!?;X$TA2q&xXtpBoj)%GHR&CHi_*j2>m7@2!~k^X?K5O816OkyaJ$9P6(wA z4HdCARNkmEC~|i3xC%R|0|=omI(0k{2TN#(Iy(n1khi}mvIj`PQ|7+DFTh~g!Kl;f z@j3H#9rG&Q>K=N^=S3R|o^5RuIkj-OH%O9Q6Uk%AMwTM*l#r<$yRh&hFvgJ@g1Y-H zC)WBmhn7JzwKn%MDCD6kS0!%tL=ywkg~(u#h!6p1f3pbA1xdC-RO11pVjkd`8F-=; zoR`~dXDl$15Sge_Cb|P4fjN*vlKR@24E@}Tb$AJpL@br}dm02rDNSE+@~i8W1Y z!#Q6x>}hlM^qwD&4)_ee9}`mKlneq7>P$NfRKa`!JlY%n`=c)Gn-n$BM-2@25L)?%aZ&7)}f^ za05p^yax7tDnK}Ftfi*Z5oMw8o3kGFvg%gpAk#0DQx<_ZoTB3LIRUrG9o^(4f#aM5 z!Bh?skrRWD_hF)~RU@c?^Z&WL#*fJv`a>-yUjwEow?!56kc_px)dPO06dCG3%I3@i z7!{aCGf+>OcbR_XI?OP#S7C@I{ou4Y_n>`?s^PR1^h_HG6{Lbq~> z8G~6csYRt$!68c``<2Ix!GI&dym5v5a8@zc@l66KXl;lj0@TR|ra126q6ENGCJmq@1XM(WHvVWiTaQukv^iI3k$XlVhU8rlwIXNsIdS#Z&+};9CXx8&1Vfz7){EmclI>Xq+7RgX}CbA^QCz0Y%zn{y+-C z((!jPCVN_CMBx79sDK{Xq*Cr)yUv_&D3g|jbTey@DdI;7o6TCy<7hp1e=ZOcpBtj! zI{S&xyf7Lr&72e`z6vri&&+NUk0DTiFA}`YhY~goWqB3#xBY=NM6(HuUGruTw-|MU z64e8t+#2G{c_@ZvCvwD>KR741%cak$z^XgJ0y^+x9I;LrhIrNXcZ=AP9#U#Op@|vJ zPc$b3n41~)5^_+a5z+&23!ihLw~=t-0A=EcwpY>%nd-+ujMoB?1Mr>+7UBZyfCpIm z4o%h~4EmNEA#%neqTsi3sPm&8`HT?K0`M0I!wBSXaohw8!oUdx;%No^mdTqTnDewA zxb&y4*)H;vKa*6ZaYB(rd`x+b7=CJIgcoH!oD{(gJ3RZ)CXWz3;aQsZAUml6AqvN+ zcIV13+523ZU549MQ0o&S6BD#Fgm;uN;t_xEW2PhAqTy|r-}kaZXGvk!`1}jAAdrhK z2y&EIhJBEOG?G6MF7fX_nA^k-aH&~2N*0H;4n|ngY(@?cHA#|COx0nFc#oDn1?SIJ zw0f9D2wlwxCQ{dEuxF(O?C}=CKZoD!k^MXXw0|4WW$**_(zS|Pc3`33yW%U#OP!h) zJjd-oxYC?Lb6z! zVUA@O5E~8=^>)(mo8@{kUbPGOV3bTLB&#6fV20wF#9QQN-c?D z4t#A?0|=-20RVLo6Vr0$0CO;w(pep;0VrSKsv$sD3ldO%)mJ-FggLv0X_^!t-$gVX!mkgDg#DSZ805hc`qM zq0C_n<({0#1@^A`ntst7&Qh}pD1Yf>%?yD6(&YdN$JTPW62!^IUO79jop$Xc=gyUZ zw?wL|%-*)I+URM9+qTT*Op)(6)I!k)XI{Mm%gPuWG>*`fd#wqLu}K6I{{+31L7b%jFt# zP4BnE0+}9$#hOt(S=&=Ov=O;5L?2T>vrG`Y17i#HQ?dxKQ|8?mS+{~Fcs8cBrU;=M zZDd)rjKEMHY@m+~pF$vABZWu`V$uu_hZ_fgy;9YaD>F@2s-9Vcy^dV|fUmM6ej*~_ z0+UY|??jGQ$s2_BMqi2c=g~oxjemq17(#kRWD*s29XCzhbP0MStZu@1wJ?oi-~&`A z4DP8Ez?INSZn?LI2N|r(TazGtw}94@!cc2vRLYWQwtKaVrm!K_??oQs4a72!dAi}V zc$l@h--bDw?d{iJI@u-Qy!gT<5zY1u9FhPN3?*lWm_VfgQ4y(~@Oy=%Q6d~l))M1~ zUxu#M9Oje4h)@R}=vyNRx_7A*QjhH;3VLwu7Sdz+gbZj)anEWCM*=)Jx{F0n{Kgc> zVo7`4b_@czxgqdV0^lF@X@^K?H`T~WD&yY&h(soLoqU(&iv}!XMwfM96PGDcfdvtE z^zTBM2_F`H?<+H++p0SLyG8Ij2m`1X{0_6ZJ7vl&0cWDqxs&zk_8L)`6U;|&%H(`8 za;muOcB`A8g{2pQIz;9V-$>x9V$;6bWru%BvD>XQ3oWn~dBJ@5hr_i|#ZgaX@I5cQ zPk;PxJ7kb1jwx9^QWGyc2Q6Hi2^`V z{bKn@2(UMb=y62r1;DFVzgoCRA+|2Cm#s=Bt20Zxy_;Gp7X{~<3 zA<$9Ir0}RI<5n@xkPsBjYyk8^F#k*UP8 z=NaxKJ=a;AM&5G2TA*?rbp5E2duK##!l}%{$W`&#>dH-5ocC9Q8#JOnqSzYM%i3d7 zZOW&QWPi$RV3zRCW(pCuw&N}e=5MV{hh~IZ9bAch&X z30rq7b7p*TYBsR7HDfCa`0fOS3%$zVhSXh?0;OwI1NhTAcmt!Vs17rUHOB!j=drB7xebn@Jq^luKhwj|#)!Y2%cKG0MQaLpM)_RS6;i3ga?k-aF;S@3PPi1lILIIuH^{hG( z;9#_TL1}jx!XRBSg4J^eDTrRUK#pPz71pztvWr$}h`Z^B^xv*iYaK&^W&FX%8de+kp$j(|)n%ZQ!fR(H(IUYma4J5qy52(vIwA3+!|C(Btgo!D*(~ zklu;ZJ0b~ZfCBgMlRgm0>$xKn&-fh*0|8F*c1M64hdZfkE4~8!Oq(lOu9HXF;%~L1 zmbkl@Dyj&@_EcQ;kl3o%d>wS}s}O5>BJpkNL-hDMJu0SUe7O6C?#!3q58Q|Gn)mVy zeZqv-!aGk@*I5W{DMqiQ?VOof=hLtZNfUF$y>C$DUDJ27(s<)B-LYo(QQGo>Gy01C z6ybNroS#{&lSha~kS_ChA@@m1A%D%N` zQ*~Ww{d~n4kL(YJODN-ozi9qQG>RV7L ziDBxMw>rPVjcztRi0{0?AWsk#T(qzFXOf&wO){Gi+o9>n0qgn20$P3A)EuQ%G&8de zQ*w^JKebCzf~TnggR|&1yJ+2tB5^&h1%EE@7Bd+tc8Ox)x&t!WFWyb>DJ5BayprfU z*Dtk9X#a~*;kk}8;?}Nj#@fP)y$%WH=SI4>0h};hl5XC#TXJHhccSs`cAo{yKQls{ z*Tt}BSB!{Dj%(cPJD6$hCk9B&;EZLoqIxCML%BjESQqa=A51(t?gSaQt$J#KcARBP z>P%H0c%ZW>q{ggqs-rd(rr&o;jDkL|udq%0pi~r6ER3AG()d)x2D#E)jglP~;8_;m zL!a^3R1f`Hh%}S91dXA~H&&s3?0-|sk$hf$F5=e@I2A&y6y9(5E{sT5k@Hs|{3t6#xoZzPMT(r1F26CO!!{ZoL`Agn(l&o)an8y=Zn zK@Yo-c$F;G$Sl2_8Sbo1c$^lSLf4j+zCEkV9T_&fQBzS-y=$yGj&iAc3<<`mDnG8O zJZWY+e@$>zrR}dS4?G>su-=6@j}DN#=3?$P#%PQ6);Nu>jm^^Fb}FfKm48Rvx*6Nx zY}G&2;`nrLcArvk|1FBeE`@FL*lAv|jYR&~H0 z(P)ye4y&@;oq$WiTgl|=i=BM85}2aWe_}ATNGyCK7(O=EAQ#l5U&fB3N6B6#P&wi1 zTMx&e+-yxEQNiT_Mr9Q7&8+G2y%oVW6?_LBcDdD}%rBQ-+P@h4DKtGa#FW%Z!v^*pqgx*ykk z6jgM~0@{+8jJn@9-}WE}T_$*H$VsCmA45(<>NBKFhwn37So{5K%BlYf&e)4)P1TH^ z-_IJbL`sPoAHTQ%mBhMT3@LqFvIn!iru84vSFR8!<8AFzy38Mk&2s(oShup|e# zFo$x!(?PxS+PO06W{dKzD08!Zd09MzZHMfwZl+!KomEx}@pyPMuTA~&%=%Yesse!n z9$U3f7Vbf4+3LC8V&Ko~Q;qX>b%@ALU3-$jOBom-OQk|^;Wqh_ zhxe~Bs`((64CO#XElMsR|J7i}L{s|s-{|O!>%ZLU9a8RSm8_TA;Nbc#QIm3~Hh8f{z{>K@Wg#*BIH z33%_Jg&o*2^X&L>^CASD=>T+rH7`YiJ4jR4dq&#LN&4RG;Kt!mdgL0bmjZX_FYrL zei4!r&8W3qYFDg+anI?-ggA8Jp18@Lg@ROwj0_4;QZJrxL(`X7pWS$Y8LSvH z$rO)fYwab5G;AOJJt07oE~zCrVZD2lWY}ApBk1c?hWcK0e=8WTF*rZjua#`gBO0!$ zWroB03m7Ql#<7qH`mQ?DcOcHC$R*w3N^fK=^R5JWg%L44uiVY^4(J!Qr-#5Og8`D! zg%CGNFjZyWp=b(m1RDX$;7N&={^?RF)<%g;h;TWKFy@U25f^137=+ELwv4dv!Y&>L zilCJIPcZ`am%%i7lu4D@$yqX#%8@8z5XFhA5@T(h*g;|kFu(V1O8BW@I6xc&DA)%r z80SU<$DzYif9^h~-@7)}Vov-V>`mJ_0tc%_X!mvO-bjus%D*;P(dXJaeka-+n73C$ zqaiJrX8KV8?OiLk1PQJq{VI;H^Pex-)x4r!>>kAMRs|PDbo~Y}nsE~J0ChF^W2et` z+0rY0H;Lo%Ci1n9FhhmFY#C5Sv{(P-(?`txjD%%nSFX1I@=(V!7!1!?8u6?lLc=vz z>jDd1l^OK5f_b5>8L-n!cDbD$>EAo900RUENCToOtaMdBQOd|KOZ?5p4ln=B?MYuW zKd4g%(IE8^;<^D0P-yoUaVwB^sFr9LZCSQo9FG`L0Z+;mpyY25s(hXLm{tq`8>m#j zfClyrAIE76(_9NzM4~+hXumcMog_5-wT-=KI}Z@CUc6fbmfnBpItSj&!Z3lhQ(IHJ zo!WLgwQbwB`_;B>+y1Jl?M`jmZg)1>Y<4% z`~^Cx43Io{?eA#n0Z*{m;lCcg*ay%?j>$Fo9H{>$i5((_&!J0xiU}XwMyy+wLp1&! zs}lQh7R!T3>R#W^#!FNEyLefawR@+PdXM9b`THLU|7*(MAtH+Xjkr zwvZP*gz2TGp_TbXXx7hwQx{?gn~mc6kjVpi^%_xkn9M-*60j|1Fme(va8NNRxAQeO z0HnCGI~f0YL!l0NEm=$Jf$3x#`7*77@Oau+=pt@YU*CB9%NG0rH(u%mWl1Wk{8P`a z-j8P6;B&3gt#~E4&C1iB`nM?Waje$Sw6GR^Z33ZCV1D@t-b@>9ZU1CIUd=^jM*<4xHaTu}W(=kg6X(J+E~U6*1ZedN&@0!d)-duv zBZ32ypaQb-m6Dr*1dyPZ7T~9LqrK&k7u^}uCuV`LKui`F#{>oC{TBHjdf|F;Mj*_P zkJ#SpqAAHW3@G3SRENVpkmjydQ%Mxl5*E=f!xeh7(B>@Aj^TPvFHkxm&V2|2r+eW< zfDcFzf2pJ}BoaHT9isBlxK*ubTMHc!fjSHKqYt!-+IXF8M}90M7VA*5`7uEzlOr9n zXb`QG!fG-lertFf8s8Q!89s1RQFOQ))8O(4GOFi#)mw_=R&fyYsmpvJgQSvo^yY3Q62# zX{P1=!-XdDRUjbuDGrIgq4g`G{(=vc@VJZ%r3KWWd<>)%Lo`JlE&#a1_B3wB1T+-5 zHq;OhI7TayZl^jq-xxR#X<}p#>BCehk3ku_oM^Sv*4ca;iX_z^ES0E(A#@9jW&7#g z10ZpF*>wz4AY5?8TFKX68b4&XJdm+e0wXjA+)Wv?v5T6UR5(U2J7q@1EbPO(0!|Ho zcCsM{7w~~_LM-Q7vgDKstse+qWCKH|sM|V#VgpPj$WX^3a7fu1TzTwF9F`Q5$E8OYy1p4U(hHrW%ey` zkq=kC&uM~NpJWn#tL|^*gnNSuqIq<73f6M42GUfxf$PB{Tk?)5 z=6U&bMYe9$jxqp@0gPwo?I51qR4}h}ZDM6Cw4PtQ$g3(BGHT{!GV;4TL6Vxu<%S}o z|J7>-qpsiKXS@3M(;V&rLyS`sqfI4$Z`wh8KmWd@sf%-)O8|T4*EzfaJBC5Fuu_LA z0ebtXifllM_0S4JHi*O7!ADV)Q~+BBumSQNFn=!W|6A_-~g45Yd`LALM}u;@|cK#C8d!!cuK=vBYOd-er}% zM@XbTaxE8+xMEogwf&pCB!a0Zl_5Ezzdcg|=pvg1lgTRjE*0UJ5Ly2|3Wrc;4i{0n zq@+jSmc@?|@gkBzo@{}#6+20vz)M6qCwz*RsKx^rA}5t?v<-8}gnz0+*6E_oZnTrz z!Piryq=#8}EX-0$r(+{pJJ46%#5lqUii7U|9g1;a`SoM7i8;NL4Lo zKi*s}!x~pwyBoMycoGQ|d^Nbih@fA_0k=ab;@!en$S#kUD@D}`h|sSzoQ(~zpsca4 zXLL1yM;dij)Q4+p&5G|!k95n5w?Ob9tnx=5tWjA8ky*-jh$CWl<6$u9a(mjezZi?Z zinEuV`GCY~0oZOELgXrBvl_;#97@K_6kOE@aE5(AyLp1(x|abQjNv$Blu!%bV`gw!H<5vVVwO5x4T6buv#?td_lEV z{DCQGghDAq(!{63T^5IMl67H_8u8s!Kf4fa z&~8aD!O9M)h(I|$$3vG*ahmc-f*p2wYZ66NhectLF=t?~lqrlFWxrRE*m1&~ z=>)?0(fk|Wi%3l+Tl1R7M6b4bliMz&s`E->{@ zte8I`2n+7X;?ju&jrTibe2#2jS&*TRwc}uO%{Y6586CmjJH9+ZkxGg)Ik@sxsmqZj zWT`rpExq`j5wf~96Hak|D>q2WMCg72SgElg`~kATyW%j~fMHmt4MxRLoA|1V`a>DP zbgzV)*y$+uIPnO&eXeK&77`*h*T+LBW^z-P*14@+DW06Pg>Z>a?*wP`19Gx8R*n?%@RZ8WzSle`1YEDO{{R z1S8DhD?b|1MG$w&V9h{`AeIv9s@&ViGazzD* zYQ+=UDbQI3Hdf0Gv9U@KHBv~>&}Q_w?{uw1@g%uQprU&NI0P8=G7sZ8D*M*w4*+fyq`N z?iYfxlSGqLUTsZSQBspi1c$aqGA|kkwPRCWa&1{w+7KBaWB+hRj+u8)AtCbNP^eC4 zNm7hUowL9S_7RDiRt)SVjh4**jjdF^!1h)}`4&Vy!tc3fW z5K#m%!eN1ChI|v)8zNZ57$h961IFeGo}p@ZN>B@Fbq`%A1c3+`4gk?W1TQ4-(#2ye z2j{jKCQ6WEdBkcni-dp)Hk|s=XAs1&Btp=};P0(7KJDL=MEtvNE4grfYYBA9X2S|` zS|AwGF*?ZK9OchO7ZK+ zWN%^JQn-TnBh!qKjQ8`jRaU_&!wl1k1$%7b@Hhw3I}1H@uEb$lRi;k}eILeu8xrV$ zS1orKtNmj_0t#~a@)-k|IT@sle$+qsZ&rHYKU(ECRs?`@9D_~yUG{kD2N^RT@?a`n zrlu?^V)tY5v(iWz5DXxj=$5tsDgUZZeZeXWrfY2!kNPw~BD!&XcZ8Y3@4v`IF|MR& zFDg^)K&tSSy~5G@u@27+<_8(E1R=sI;1MM-wDdxi^`PHTxV!|p>Ek#Q$F5cccfD`e z6|PcfWQR)ctYK0Ga&Aw_vJ^@f{RSc1;b#Tm`ZY)7;&%2sT*bAMEqFks6~d4{w;(Tp z7o{<;MLSd%v->e137`wjPszpbXNg>mrZ7a*Gvweqm@3{Vkjv`X{9YC0=>bmY(R1kF zp;LY5P+BbL0RgG_k<8@u>P8AWPE?xGg}>;YhOUe}gdP$4M33UFw#D1KvRK@Cwm*O< zhG*NhSh`D1Py`Jr(1;8-wp5h{MQv_FS6x{)XL$>J)Y3UT;1qV=&)M&1AU8jzS!%iGi@K9c|1Qk)xJiWHy(S5=F_~Ek0K0I=mFA z9hfl25)sEznBeAUcu@B)XeYXW_{ixdVhum=8>07k)uW_)*%)3#l*;rerCk?_Dtl=TyQSDJSK?-AaAeE41Q9~td){P}CqGdjYbT|LBBLj7?*f1(wu{LI` zK4(yX@kOi0Oo7<`?qI#S)Fti6Ep(l57qTEoBm;2dB>9C={u2ijvL)Qap zA8P+lLrroE_j&am5oN$2m+8S#s7tIuwBr2w*;r2ky|)MRBc{GKgvSBz7%b!IaRUfn zcRc@3H{w4^m&NvSVyb3=gh{*NT&5D)xVcYc{Cwp>x4;QYlQ$X$4ovKDQf{%%C_U9XOi3ny>u!-S|omsk%GTWLx{bV(9I_BfN+s z0PkPvKb1{pZ-8#DC7b2oYQs>MHHswEkrT%a51Mj!>KnPe7b|U{)U}eVTr1N}=!k&| z*jz_q*B^JC@8H=H^yD7J1=7|AafoBq#pOa>R5%Z_yGW@Ii%zF`;%`O&F#V8ImXn>d zMLrTd54KatQxRNJLxW<5F`1eH_bn+ex2Cx~32rQ%_^X-_YSov& za@{t|@qZzg2UXK8@LiTH$lU?U`ong4zccs`{=trG=STN1vcl()-dQ-euA0f-LFl6) z!~XzCTT%7`V-Rh7Yg*JEiKd#2(%xS0PJpN1w%ZrW{qs)OCzHV1inQzZd~A$*Zxh>* zSFZr{w8p3f*D`Gc3!R` z0PFrPhdv4#)sv>D|j0z8S9S$bv*?6YjBD}$z9+e$=RgEl_z z^_Q8u^ldEHf`EFaS^g9M-o$xNGf@QOtLrGs-Gp=;l-qg!*VSF#qWwDNM-XVp zs?Y89@UveRb%&`x{0!0Ei<4Blo%iNj#X8LPlRWd@JslzI+s*yfh`0A(ake#d`ptQ; zhsNqke^oGtdDBQe9ORZR9Pqi7Ctxe~zTBWpIVIc^DcMRPO68T{@{aD95m9eRklg&9 zI~~@m8=7VeYJy1R1MzCqAP6nLvAIi9wYghJoSvDWCNXlI&W}2r1x5SmZ@;1N$6YdFy~dAzJREF!D_{H)`!jsx z#WYj;34AJgV#n?0{ktXb{NP=&g!_p1u(~T%uc~)r(!FeovC4-+)e_#*{}%+2e-HGy zdsXjz|8`faG%O0s&MOaybxwXge_?){Sb8?5?` z7*8=j$KY8@JIYee*NPe+P6p`=9F7ebLNb}nAZsrb=bHpxFw9)+j*aF_)JJ+_q{@ku z%!uXoz7}W&*IsV~FK675tZvi}gZg;!7oSpr%t%J{2#MVn_Z2XE#_R#x`xElww3{)7 zm@oeV$-BncHpzBbNV!DEMQ^3V>+NLj@sUlwjSJ{IjI?EHuko%zFAC~qOyoJ;hwb*e@WJSY zRwh$Zy**n#PsEJOkv48krMOc^pn~}>lk8_vHmmMP{tW7GAezhHGWtB%4+08`ezM2? zNksS&eiZJr?YEuE@|^n3e<9$9?8D%sIA-A^M$wn3hQP7&XFw9wN z>*m4u8}1=-;|}4Q&}X1vn*#66cfftycVxT%73CHo=G3c;gGv8fx7*^OKNny6c2|40 zzi01H=Uz+IvDn|Mi~ZY|F1nV#Fsz-E;{Cp#)?mfz2!_jDhk*=WGJj6SA12MxY_Or< zs-gO!bl2!T9FT(XV2b7qI(@ETY)f22hw%Z~r+XrEK9HBTYACBB+jx znWXyd(_yTZ4&|1G4KOsf@}Aiw1In7eM@>RDgEHH4@Pu7Wd%z@ zNNhBN*Z0~~{255q-Z!|u40oB*}VT@@Ambs^Y? z;dP0tlAmEj!^ldbK?+#8Ze9vrvAImIVqM6s+-Qw~cRP)$yv?Ay0{Yt0-yYv#$EVNu z(S6;Y(%+XK@^6c}wlj9`eR-#L`9(Z`d|e`8$gqZhtyw5F=g6|Q0~s0sP2Jd-OBV^$ ziEo*s%SA%_gqGW=Su!PKEgQeG3an)VraL*NFYcpVt0w08&96=r>oy>1UQIrxax6CzDrm8nz`0E9xNOcy_TkD|d(({J0NV#*B%4$);lf#=k>!`!X3eKvb#yNw7}_JU zYIiu6?jY+)E~&?0Y$LVBElQ;Yd}F6G7b|Gsy$GBJq6Yjl9xn=P%J|MC9)J0LEY8kT zyrn0c)FPO7&MT*)+sEE4-!~_Xr0hOSpHXpS=oY;s|TiXc9-c}ZGV4xYV; z=(*}V4GF%w3JT!%n9}QRiZDsQkM!&28~h#M@o|&J2=I7$8AF@Ekc;^}?=lpr6m79~ z|A6u46!>LXZ6!M|cE2d#4T(YOJCygGK51q4S#$Ov^u2^ymfOD>o;r=N9-T+s1?Xm2a z$TFNKwB)K_aa*jXX4;8Ly%z|VPcdzlHoRA&K$JxvzTA%9-qUt znITv?+QQ23;Se9;#c_+J=4~iopp0VQ`FgxoR6*&_VANGaVPD-t1-mts8rq;%p z<;Rt?^2-fbO2Ov*Dham4wp~KY%vy)r4EHG9ZP-qw#D;Pf^X}&`2sVU~T$7 zvc;V8YIA)pOCS4tWN~~*zT|@B>Q*PhRV9_B=~nNQVlz<+)J`hOxa!j+rO4JI;ANX1j4bq3UvnvHO?yt||`$|;60 z*B=DKL(~)H@v-UEmn2?GzHK#2WZm~VYJ6a%Gn>rB5wV^xUs#6VFS{ql@20h9=3mXY z5Q3s&Iu$Hey5dB<7Cv7E=h1XD2`oU>@ z`Hg6Q09YtmGX|xA}+CIrH}er%hRyvHCvkV z1s%u^I!JJU;6n*}cC50zPnr}Q5~C5gf9j%H2W=lFv*I7@xXqTso2#^ej?$tHLeA+G zp9W7Wp==Abg!>69bA7k0+henbnp5&z8Nw|YL*;zfTY^+%n3{IIuVvJS9@DuOR$yLU zW{j$?BjJAK(Jfa$M7UL!a*#TQL^Qp_m+8{8+C`DfGbc#$e6CmTwkt6FwQbq_p0Jjk zl{t9r3l%Nvz2wzPF%|2l$HmoutCW>r^IurR(3FR4TT>56UHL&>XVh;8qp~XI$|{SJ z#L#2RBC|)mR}f+I(jn{1<-bdfXNZYuPuICgZIvCt*Q%$$a_nrCR?hD3mDAz6tb}+! zeg9-8$avOH^U2h2c9EKZ5hznA(-w{xc3-;TCZ|b5WG-^rbQy*{RNR~{Dco5dqFQ6QEh{%T9fmBs$vL>`Nw;tXv^{M3pul&*P3ndn2@U} zY7p*6QE7>?Y|kUs+QFqc{Izh~bKWbev&A%;EXOg%QJ)0BLKi{f{%Whd;~%C4Wes&@ zP2TRw02$UpVfG8HO4ozRf=w08xf~y-hb3=nMn6NqvmENf_QlOaRI;Z1Pk5r(rkwF0 z*VA4mFyAc7Mdmy7g(|W1A4K=3lUcvmkqNlv4elgpanITn*reQJ%^Uk-ps0Zd=+bECTgb6W{bg3Wfs$40a7Ofys=$-SA%s!`Xb4RTy%8h3tZeU8H z@@9o8E2~o`nW1y4W~C%0_gkW1UQSD}ait-8Qq2tp)Hn_%K zCn!q(J54BfF&~Pm@+2EE%fB>M3`Gv(K8h2~%7JnjELp{;lwKlPy71}KM*3h2Y6=<5 zrkge4X+VtNCzR?~NCA#j^ zXl&|X!230g^4fGYV;b1rSumM{BJNs1QdIvGx8Gt!3GMA1s?K>q{6dN);Clo4maYEp zxDMUpMYe2aajHc%jDm1akp($&*(Eu;ymDqDjKu?%V^g<+Q6+aB4({rpo|)B9E|*;c*FqAXG4wcMotk&%d(k9xgzBRN$~iMk0eL2q9l z6F#%4rTcE!mlva*$BzCu;TrQg8DMDZxs=jkcc58KfZztW7tva6e~O<9bX^z}3AS;s zNs{C%hJFZ7^|5v4iCpJ?W!baeA2p>UIR+uGvDxfyc(k-m{A}-jOGqIjZxu*dw6TVf zg_yu{WWEI8@mWTPue*88KdB6)L1>3a2@QTZ*AWS%SJ&5McT0G0VT(nq280VVl-}d0 z;YJ9cOr-uIpOk+3wPZPc_ZGJEHs4it`VOx->?<(V} zC&Te+>dt#VKSJs0nm7874G~mMnC5cic0_&VRE`qexj-JmXG_e{1N)y>WPM&Xz*O!a z;87?8vsB=SZ>H`eDFO-KVoldZPs4Uz=M*uQLp4t^m%)4hk&WV#)1gs^4G)d5pmZAx839QOs-qr-v%tf{^@4RcR%hsm3SkFGS)q z&X}iT10FlVG~Z-0sfzT7M-_kumxihx(u&&BkWfhLxh4wVobZBi@+%$zd&6Q<4&cOg z_=MwUgX15+)7*8@^-sI3jV}UPeLp>Ie?Fyjm8vo-vy9Iy`yOU}Ae}yt(*C_9-F9x~ zr+UQFTe+IGy`S%R6fb5cj0m%r-s5S1_K~@s%v8B)4pH9*lmVL>62W2h+Eu9-`DEr^ z?^%-MP^0g%Gd_=12IfR(x=0tP!&{T&#=|~uG_&^>8v&fApz#tz$SvrfiV&>GVipo( zF~;~5F@;f=$&=3N@=SD=x$d%$hEAN%WIvQ^%VP$L%X9}PhmAvVH=BQpwArz_j zi6Pf6bEU?*c`7Fw6K1cRr;z04oXN5N&Ry|C*uOpN@vqq-n>A+T@=%eZh#Cp4m67+aPsuCQmP?!i6qDS*$>(nwh;fssW z*&Dd1Qs*W=GXB{XoB6*umsa}2Z^joVgqGTsnHyg-KdrZEpgIYhVM!~K7O)oCffvw& zedk?YK<=EIx{^6jbeYpB|C0xU?uRBPy6sqvk&64$b%uEQKffAWWWO5+*yR=&M@8RB zPphsTp72{*7)r2-<+K90C|XzmOtfY-#y0a02K@mYwzezX{m=SpYemf3G?{A?`Wl3W ziI>?769Fg+uN68dC5@<;G?i*M7$|gB-8G#f_1JP#b-h-rE(mZ~Ik!bnd1ai*o zOx>1>W?5Ppp(kCuSbpDJI8w)Qp$z$&xA5}HW|)OUK}N05^;Bk94!1cIb`ZJLRst!+ zfv(uUyCSAPm6}Vep~#XRI={7bk)#$|w`meo3YS_>590fmzlg_M%8|9;(pqX%nAu6b zVk{CKirLByI*s_JBU>~njm<#l^89|kD4Mzwr6fQu1IN3g z9&L81muh9dm^exAiqulX!AlRfwG^9BGl~mNtXnUd)cQ^Q=vNjBk2NDz-K*HcuRh9` zPZdUgw0wB__DE;>?b1Ua|^$|X$x5oG9-5!pwDvijBn0m(h6E1uiRMz{?* zk1ngt^As}%bn%|`Y~#;fD%S?+z^tUekdF@8rV0~onD;D$=a1pp4pkHT9E=9p14Tg2 z>dBTM?hwxji6c~1_nubbqMBf7P)&bvSN&AB(PJT<-SE0Xd<^5D7E(n9 zZxDN-)LW8uS)p*OV~Qbm=n_~LSTzBB`mkD2*;(;|ZK$r*RwV-ziu9|ZIe9s0cxY>g zZTYML%dZn^@N!vpx7wOQhr75?#IeWU|Bf#+@XN;9kk(AJV7TE<3Rbc$Rz+-(MQXp7 zhNR)ye-N~SA_lem4h@Krz_LN0ZjXF#BbP4DsIPBwwDeuNi7T+~P_(Ak#dWZ4d!#p8 zTmEZNluBv`f%(yfAFevcv^@O81%?Z2$@lM5x)f;u4PxIvd)Pav9+|M*-}_r*4D{Ez zB(hb=ZrR~rAB|yTKG*AL$y}dfF0gs2yR4d0U@l7Ws@WnZX5o)$S-JP94}3|`<;iZ& z>6d)xmLkvk*YC(MQ6<+*VPg1a9_F{2S*=lbk!zu)eaFfhH(4#M@)O+G@4hFMi@2)l zJ+be+C$I&IoC{PL4bUEda?E?m+3}&*d4g=6^}1J4==yb9l#!8zO+$*WEG$mDORatt ztE}=h0Xt6lQwd6PV%uE_tbv_w5tem?wWlAm%M=Uc?NoQFuPN$Fbo}x>?lCpm!zp0D zrP!!!V=7vDue?jodFK?zyC_ovqZTL*9}j7QgqyhZ3sUb z$AMa8HT>=L?|wJp%C5##b-;vyv2Z!Qr^9N2OYe8~G-i)Fj?RV-)$xXl#s$n<2Iut} zbtM7aX8)AdWWQ#;>lUddLcEl`V$mx0#k0QdPeWAPTNTY~vzd)-<~;PU%I-5%ZTsKB z=U?KY~(cOVNbMV zNde%iKE)W==K@mtsg@zUJ~n1^pk;NHUk$-#hUu;LCa-DOl`5-^+zqx)#a`D^xD$G1 zQ&ye0Zie%S-+l1Ya+?>GzSdQ%EX-H(s$ECRatW$kcbuDajq*!pIe+NV0Ch7@@|hGC zTsmLt1ibqMT$h~sa>=>_a$}WjaE!-&Q%MCnrfSQB44E8HDvQ%aTTJCy?c_HF{L0PE zhHN&K#U{(?_kDC3HN;9fI zkMpC3HXxxdHPQ*779dKBe1THyrgD-^_G>kyEbeR^?y(h2GLey2*Bj(#Pj$Yt>iakL zNpm<>p(*6QY*?4R*g)9Yg(P}qLsg4I$g>^t=~CBvy|(JERa`ABI4Axc%%+>>7NaeH z6}Q}v0+;|$p%h0&7aT2l#k~^WOt98fAq0$1AMe9v)h-9A^_o4zOx>D7aNTt)Q0^~U&v7WcL@MrVp5qn67(r z^LDF{i_g>b_KFXa^K-jwB&n$=D(z!GhmqIQM*<{ZGRA);4yFblvrEx}HSu>*v*=_wcdw%x!&l z_VIe&+cLJFcD;TuRz?51w-n_LY<#$QT(6ydU#P9TcOQAXl^?m@((Cqm|8wfaQpY^p+g=w3Rq?e)2faT$3B)E9dCSIp5z_68*v#M{u4dnHD z_pM_zUq|{p-Fx`VV*mWJ9`!6fqB0#@>X`n!=w+t%hF}8bdgng=h&v17$4Z!f7tO9T zRDLK@@B16+P2T9$tk$260H~qmytC3UBVQz6`AV9gO=EGP`&CT(!sqQ#kCxuk5+i$N z&)c)dQxhTC{D*UOd)rK~2pz{jH%h~=6&=zYm@BefOEQtRJJ(6_P(0yDsD!~>kG>(@1Wy>%22)C`cK zi$(R`C-U$vl;d(DGqvJsJBtAQw7e~5Q}b2CfaN4*ciGhvYN4LIGXJ16Chpu*QDc3| zG27m{u9n5ju+%stQFajq?!9>aQC3hOC0n>XmX-2|g2w-j3^i>ki-gB}vJr4SXG7q1 zYn4-hNZy>2Y`ukYRkz-@eXUu$eT`pmUFRs8b%_ut==?T4BFHn@%Je~^4CZ}Xyyvy0Jpi^nmtKdS{sm$eN2R*x7Sw5)%@d_;SoiL>n${=YAA0jn z0S`=RX%D^LY0oFSho9>cOl6;$d!NHhE7J@+XMmo=Z$f>&JApX6*KdaRYg|ON4kxls z#pkdIq%^{3q_5YyvTuIB8x*d0U$^&%r*(u`WVV0H-IIOb+*T%FfjF3tUisemt^9?6 z;)?JBfLt~B|c}1CW^cTV-;Ur-__s^bEbFcK(geTZ$(M2i-w98;y`imaC{C`gpuD|Xs z{v1B`ACBCP1Wf?n^>c-m32$*c?5CvbA1S*_W>$Rdt~1cT$S9`3C??sF+Re9y?LUy* zk9ztyeF}j3-$<6%fRwJ2@AJVgKr5QJ{PtAQ+~}355`+(hCUQrGk=vQXK_s`1cTF`I z#&HEvihAMZg%G+<6f1t4-oi`ovZKRs+9hvdPMc<#K0DYT|0?`d6#r`S^s?P5J70aq zB!h=Rk&M|4C$UVS+G859@$t;b!*uKME3eqvd8=*aUZ_BgGkr7jk<_v)k3O>fm_X8IV@M{m`Q@^#?#u=ll_nJLQDXa>$| zV>57tBFWaoy-4ExX}nhi;|_t1|H_6hy*-W{x22^3rbnk>MDn=zko9ic4g92c_8j~- zl)t%r?-Jfa-Tw7AV2tk@Nuc3kc>6n&h_*KrLf8Axx2dxC_h;{DX>HGj&t)v{>56b$ z6ZL3_!E!|Vbjyx4o$@H{$+NgQsIy6?y|P|)jnOvEpeG-JfS&9_-MW zNFoZB0a1k=Y`S{2&-GUgE%%aH3mvClW#A>Iic6^M4*bOOc%)32i9b)Yc;KSPxON}l z#%9yghL|tTg;wK$bBVB6u1d}%b`aa!3hD09n4R22=w7` zSHH?^_KGy_dn=+S&`(;;Dykcb$(pL?ppRcIg5?xUvQVwL-HbFXV?%z|vFD)caxBt& zPmSfVGZr_&x#cOYTZx_ z;LC!gu^uTkHiZTLS!5W4umq%t?duD;sZoDU{G=?0pg`|es<9%s+%g4)p$1kKAeP{p z9Kx~Cq(_e{?6h3sgj$XT4_3yZ5BZ6r=huNfRaLxl^oLZ)=DCTe`Y@N2T+XC3Qy))h zt*CsvX-MgvD;RI66q(Jyzu;^|H`7j(nzKpULFFI2l^;f=dh1?w_~xgCX?bZOHv61} ze%Ge-TE14L!qEYg8qJ^C|3<5sN6B85epM&aIr4?W-nnAW9Fd&}|Eyv1uZC%@|GK<(N&o$LRpf-JHHBA}{6unD^s>9{KSuoCXtg!xbmmy= zQ+yl@`_3l#;#-q0d@*IX#fY2xW4y;_XBcMKj(gD|NBdhuZcnH>#?QLWUR?b=hni7k zWyRg~##4$61zm2;>k_~krcU+1R`V4{-hlA@)7`lgU?_L1EXR#=cteEE?RYdKn2pal zrz|$yA@RU&)BY!9O-*VB-_9}x@Yx8nbx5{ll`+gmq~0e@zI)iX1T(uYRKQ-@jsS-< zTZuCU+l1E^iycGIE$B4MR=O_eHgm718f7`cId^58nrQ{ue zKXw)rL!B3oO`KZSb6O1|EA4!oq_~UD9=s8c`<`b#>8Qw9J8!bMm#wS#2Lzy5skStfh75C z{MS>h2`=>L!AJk`0)Lm*Wz(AO+NE36PEy@Ie7l+-EKU6Q%x@SwvhR-ERCHiXa=0o9 z9C3u4e>Z!liC#S`iB%1&37HR8qz|2D_6c!Eg}~UM!Hx5`lf!HJ+-yo`t0&RQGZxN`~fT0tYLY59aCymWTF@0;UQ}zxM$k zcD?svDAfJn>afvTomMYZhi5B@xMajewk3G} z!uSlx>UIN0ZU6cL=Tnr1nP<)d=*KbU%yIEB2D#6P#ql0bW= zyWcS07!n$>u7^>)h>+hVE_|SX2yt!CkX10347Kk7tge46?0Rkx^^n4Ss`Jv7U$ss` zY(ETACWH&P2rj?9H@|t(mk8IBNIFLTF=Sg;mE#0Pt?$)2d}gJ5w`Tf&Z);2dNTBq` zi#hkBS=MRuzI|!;`eklPI2{w>Rs6>E(lt`)f}MG!{PV@QbN@5`TIlg8S=Yv?Pm_h} z-+^=dI-#}Z+P;oy(zYKM4Ecuh`OZSz;$u0|$gs8842YX21j4t9Lq`2-q$V!0nu;8; zp(m>wjyg?<`?%vpG4Z(cE|{{HT8ZDc{|a8?4>rvmShjw6wx!2SbEN`n(PCLpN4>HD zma{8OhhOI|%sIZV+!rpNTT7GgrPu_8G_{$I*&>>Y;Ddj5 zYvBiKNx(9N>N<=Lzx{Pix(>&AC0+`(WpU0>j zE_vmXt<0u4qP~+;ufby(l&PZkgzv6)wkxPTDkDE_ z?ChrZRBa9nmz0s8OU9m5;r)=D9rUD=_ZQ<#&k5X_;wMuZ!t-?`QE`Y={+Y*pLUFeE z_6rNisS>ShhJy1?Vog1ByU3ERd&G@E|`KB5g8Rjzd|l|R@d2<9(Z)MxW8m?GGrWJhrZmpv7f_k+&rXj zx?y~KS7GuKYR6bwQqqS=&OC=2-JWbR{rcMfH&#_Sc{_Cv4FZyj|Nq3QENxByP4ODd zQ~UK%^zYhUqf~mL`P38AtX2a9aLu|XocbK2%X~r`IzAHdc>_;6a|`Y8`Cgzsy_1K8 z(>yFFZNl{j(0XgoF&iaNzCo|_*Z!pZC=FFc-hA>X<~R0B^g))jU<8I|gk5FuLLqfj zHi}D%qmo&05v~lH%j>x0uthG@!IsXZju+)H-n*0#^5IZt@dP^3Y{+Zmw4B@*Hy;m| zcifTnpgQ=c^m-^~|0y_=&}+9-lZRtv#8x8cM z9>w9^xQZ;0FmO?iSu|7LH}7-K*8xbS@2K&!i84hc(UXXs%udpR!|o)ZafCq$^f7fY za=!(c|2WF)y@GGGg!7}=g&Y;G0g}bi{`=-cz}Z><8z@y8PE-}?i=cX+=>Qb zcTK$t+)+n~7cmY>qK2N}@p1m}K8<6+jS)*A5_x24b6kP@RU9i9^-VT>;wOqdU(ZZM zPaaLrf=ZfOR5(fcL(L@kpXpxe6-%cP^&pazu~`ac&@ZH8GL+00@+SmJaoYgUx zegS$vQ~7`1CIcgB%XtAIie(%rc1{MCQiX3n^k9%Z1^E3=-85IS5n+bq*rt*H>SlNbUOx z9BCA_Aj6MloN+A$DeNRk#H|~0rsx#w4Al1EgT%2~aE;XEz+6t1X~%@$J9*KU9ORy* zz2T#~PMaPM;?18@@hKM0vT$`jnMrmOX_~Pk++~fjucW!k~sMyq05O}Cz2Jp zs+s;+8>X3t!@wgf?a``=Ci1PD&QT7MHi z7?1y%HT|V2GrL=G1*(@cP#q|C?d2!ChRnv=ce#S?Wyc+Xgs|=U_Il;Q>1r9OL-8MU zozs>eQMN_Xwr$(CQEA(@ZQHhO+eW2r+vcs;?$P%T;w9qj6Kk(I#mtGeQA?O9STIpC zVK%4`3!79Rx}C_zEu@;MK|OWLGd0RPH#>L-wP{@nnM%&%37T(|fs6-=92t_8QY@O4 zri)TSo&pm^h9&r3(@bULiOGn!;{j-i9^$e_x@v5UA}%$Y`DZ(m0KPgV|prOc>tp_|l5&ZSK3- zKrzZuz_dYn)^4M^QxObNEhc3Q*CMb9)+(KFx{=wG^U$9yXE@5KyBjOBj3$A^ntRRN zRTsrw3z0t!)@}!YW22-bYXKLwu&_-G;KBKu7Wn#OGy~uD8sm&!M{#hiVH!hvaAXuX znM1~{yN@rossIO|M{k?AeV_4g2$j)(vmIeQ@2@E#gc=0&FEvP-IY!wzd#a_6{oIuC054-O|WM*li15nTDPGqG?=?=fk%AT$$ag7sV_ z_S?0DfF|@LQOe$7gw^lLM^J4+q@rgGYiqgKhcqOn$? zxhNcho(a_l!zVw%k03~G9(+PdB^q;YoFOy|qgsTxujKW+!Yrep&1zJ_IRk?|UWJ{> zUu}7oT6@u@O?oO=C}Adk{HYqlu;yaSB`=oA$;{EDT|QzZTHz}j%dW>k>MKTgUm>C6s&#ar>Jypi(&F(wz^Hb%-$twdj zrbFN1+U?z>lW#d@fy=q>D!~npKC^D^=KfsP_2@k=oxV1(ak)`*a9ilz`aq1Mtlp2G z(h%7a*}huN>@w4M_cd3;6tjT}t5+DqAT^xo(9~Q-K{;eN5Z?TxE!!v%wcfNWfj?!v9y zaY=>5{u{6InBz54uU6^gO4N{ZOv}BC?fp3nnBp-a`M2+Sn&0b%wQ1(%k*@Vs*Gbxj zxySBy!G`&w1gkUB=p=iwmgO?Eh<26b5RJ8*0mgLEar9$UjXyUx^Yj}zq@=o(AJ)G} zbiK!;g8Zq~^V;=^9rd=q=TBE9jHj2SlIfr*9XIx^YAI8PTN+?zFV{vQ<*FOovz=3l zIgL{Lc8t&K0*1C7UJ1t2f6fr~Wq8jmVg9@7Q9GPOh>BRuBK4_#1h@PD;B zt!c}FZC2S|^dzh3+sYdgCV;kKmMtb4cuLHJ&s=tUKtHx}@HATT)~CknfFUR!=eSY#;w+mfQ#{ zBbx}>Ha z^14eIhzwl-X$NAhKA=+c+1E>NTT~PW#iT1&4Nro^)u!8Z7rI4@COL?2;Ps@5E(uJS zPLn(a-D^7Pf(eP5pWO#%!W8`xUgMMS1Jky+(8*l*`f$MxRVg(Y45;_}#e`N64;fnw zMT7Th46p5c_$MBby@iM&;AVT;A=A+)Y~O)!V{D-S40uqqAaEW$8?XSk5(G+4nq1yk zZiqYVgX@gMM+{BJkW{xW0X<+#El5GG4Blqzjh9fubDFfR!^KfCu9z@9i*ZQrT>v10 zVLuNq8mAI4wK2=XP{6fTgYuPr#vJ-ql+bSMahef`G^X50XY6c2j&Au&`hS7a)$%cO1%H-k zGPnH??Okk=$ypcaFTob<7XuN06KYrL%5Ip9G#0l4KFzA5Vb*g%hRWGk0EI#L7xhUA zMs8r_o<-hQkPW-9`RfEUUPC;W&qHLV|B=cS9m4k|KyKi}=WsGnw_sIx$?nsH35uiL zGP&J4Yrbb67z)sIMo#v$>5_vfGZPUqEUz3>)pJB@m-v1{BjK~Yr=z1Gr!qw1wbki^ zYiJ_0eEb*HG8qZF_r!pYfpng{NrsGVj-kBxWk7qsDRf0eI!YKEAIIP!Dvt{X>Yw~- z+I`x>6=gm#Kk*~W0xOhM?VCn(Y!Ycib31wUZwNuyCuZP9kXdmNFJm6>VoKX%!h=@q zcNHn558oxMM9CKrKiLT{G4H?YNMo;c#kq4r-vHM7h(hBAh#EnQ2;tCZUq7s+ccdDW z21MEv%J_oH71Hv6!!_oxp5)gvnOauC1;Z^1_QLHm=3{F^OK@5zGF5$c!nAiQ{c*fq z%janY<0)mrT39y@&+a`9P; z+cmlq(dD2Y)A9?3>jaFU7Au|DPXIx60u@EQxPyS15j`wx5#VOqTal`!KbyP&Ws(-!2o_C9_ z)90ec5U+HH6*SsOtXiP~^C1fn%rga;ZtE)R$sTgR8q`+D^sqU39P#o0=6h?#*%OjUm4|hJ}y=9|%e+PPzY{ zG8gqiOdj*pq6{)2@n|SR|06=1WB0Kti-NX~Zj=K#l?!o^-pY3*z^5<{!%i!e^N*{@ zzX;hD8t=cYq*x;vgLdBy&F%K1Y0&)=xRgf(7LsmcY_Uqa&H$NnzOi1II{MmebQ(FN zdiXDi6#0=1X9C(j4-gBfA`}*hZPMQv4d9rj5WXD5CN^b4@7&}?mrO@AV`c+k`~=pt zpB9|;7urMKHWDAnGG7QmvVgxFZqBnhRk21Y&Kuqe=wGM+>v(yy2DvyFc+b+e zOnFhB>~y>J?JOUj961X3QLCiKzx-R)^Wt8P-sfi9`)N$4IQ&~oShmbX-jo;*$sbx@ zT*R1cbB^N3tTkQ>FP6Fe-3_LOv;s009<40^GbS7=$kAOe` zpaN%PL9R=yONJbkxuI?JqFrA?WM>ak9)Y>TR4!C>n)65vnc13}$TH=^vk3S?u0S(r zkgDyF5VWCnM$Tdbm&yC)T-i3EZ(va-CNS?QJ?stYG4zcixFTeH0W|(ugH;sh9vxud zP{VZCqld<4@-(>9!RVf9KQ>3SzrLs@l}Esh*0VN|+FUV!)=B=`HV%&Q6a3kGgWU1zAAm5vNeLaJQOzTC-(>>K#QN{h>su=5^4wJa6f!u@(?I;E;l`PKf(@ z)nO0VoK?wC9aNE5@#W(kpa#^&x8{R!PU-V53dwOL*2#|izZWGV=@zw#2}+v*;#=J_ zIZ4GymNJg2aSISL*S5emp7GgA>~ScFSn2nB6aLVv=Vrab0?zeS9LyE}T>jZ9R04sj zp-pbIg*Btkcl=4Le(j|SD^QyKOQGvFJP%#R7R7P>kqPp zil1CV7**0JE&Nm8AN2R(^NH34TOwt-yvhMIxT*~fIBfF2o;ml!x`SmviJ_8eokb9t za_uhxaTyZldQ1rmw=4+@zjkDbAHR1ZRCbD%?Voti3FiS)XbL8JvT_sf(?*1C<5|UC zf!6z?D2XEJ!Uk^0RJ%xXO5;C5X%u@zO!0I9#VPVq zq55<6SG?zk$$X)n(T?W>9zhP80{T`lsb8&!bbAbuP@L6&v5@x>4yFtNg zK#3Iw^)YVx@}yGj^L$&NPWio0Oyxgc4081OO5FV@_K%!rd50>9Zopw$JV2}!Fut}a^13;^Lsa;?UV; z$-d~!$WMUS{w}fZJo{$yR?QmRcE4>0EcSq%zwB>>@%V{7OAmW(@KLmXrAZ}?n$b17 zdIZsfM*{~mLP>sfPL$^{xeNCDcos80g$2k<2*HM?{&SWrTdMj1z4@(3 z723yTx+w(J%hIhKH^bdMz*8JmYV|zSDOQ#V8KJILi4~Y6tmeYUML|OqL~c|Xewc19 zz#XMciR<5c@pa?4uDveWKJm|&w)NayWtZ>sYB3#zdu*GIl;_5)cNlC<|3*lXtPN7L z*qR-RYOLdRa|y0`@hS21Q7>=k{89hE0@<^}TLP*;^BeB$?aV!Jk*Twz^cWx?^=y!N zNc679#q={QCQ$HXQRIsgL1C4ZQef-d-Xm0i->Zua*&gTudRof{tp2*cR+`-wfA{^2 zS5@38L|Zn?Hf-K4&)~xL?h2ytxg4|k=Dj9ZbgH7~r8gkF2|3_K0=w-i%0cQV2xwwv zTS`rtK(8`-a%yjVcQcHfc(8`@w(ZR_&4YB!)L@Ox%(@;QAMb5J>RyWoQ?%9ydNM$p&nS`3^EvsyXdwfrhmDm#i^d@s_Kj5HBDiwBQP zkJCAt(f4AZ;k*>ePCCQH;gaeWl-twjn823dXR;W}`Ne%ZT+n^rD=WgKXG|>-Q)Ml@Y}mA1dp=Xqt|^5#uPS8hrl$06*RHZPVGyCwMjGdeAl0afcVo6MYn zrF1sGzn`7eoE?%d4M!a!Y-$F7{Q-A<#06sPC;{mRVNC)kRlc{UwK)vLO_^Tk4k-QC z#8>(d11Jbn_7k(?{q5s0+&>)zyZ{bjCZXt`ujuU`8B@cNFx)UtsuqJL|Kt(JA5`W| zd_<~A%ZSnCg~?PA^H4J@+3EmoSyP3kAR)n3L0`gB%wE*55G#z|UwY69NE#G5`sqcn zQ=dUaeh3{#9xQb+qdouw4Rah46`3v_bcoc~{}T`i0ItA+AX#c{iMqKac7)4beW!~{ z5SqdsFq>K(Wjs@ch#8MACv%qT~1Erz;hB(ebVqB^15hrkvFq z;}LXKMe~e+IwS{jSh@Krx`;2Z`XS+6y^#u|{#G5YN~Ee&q?r+sY&CM?9B=-7zZqQlwB6NL(tCo3X?CE@Ls zYiiPqw?2ht|JL|1ATG%wD;!G!21%m?frGx&>7<24sp*{ll*IcI{MnU5zYlFqS}s`1qs`+_qB{8ZfOzgW%8;pf$5n`sA+R-JzU)`y* zGOJ#yrzy)3CN8EKayCtn9NAi`#ed;2@FTa+TWBgwN>ej5Q&$_igFvM4sv5>Le;tTP z%KiM9+U^>2y%)yms8zk@;%iE##5$c9qp2RpC$vznLbgG@zqd$d8*u zXFUwHmvnLCu;v6xHk`E}6@(|4;;Jf^#Zl*$aEC%nO=zwas-)bO_@&u7^Wg^*&eEZ3@4SebXrs=%%#&%2t*OYBy1-{cM@&#_%! zo11Owk0tgjD71IQ)}&S&q<0&oH=BM-xVt!OdDA+ptfYID8mkxU_a&lZ*`e4rJYc$_ zTmGtZ^E8hMa|1Mk3v`WE+jRZ-dFF@Mw|gYEXFkf*Qr*mKMD2%SuSc1-oqP~(Ri3(I z`$stpdkh%>1~;V)6bpHc*hkZR4~E&Zn@-BeS_fWlBhH3o8_a}IqPjyv+*%R8zUao_ z^xDK%iN4f!BJxrbcw62D?ndHWD8Fe|)gcykYAmgSD{RwPLZY^)IvINCy?r%O_|$%b zNJ$|_Yvx|>=0cHufNGCqv2Xn7D;;XMBP`LkUGMwOb)Nvm{Hd>D<-x{>?aS)rfWBbY zgZNg#?wmR?_F#b8%3|;a&Y5@3l%e&8yJD>$bZY&B^kWaTXoetoeFEjR(!0&W3H+9( zypIft$77u~agEnzf(85u+?wfgbp==LSd31IBcIMs*#;g?f=GonzR~S9`I(4oI{Ejy4l=CO5rO?W9CGO9}cfM86 z=&h*=Pl8bsu^X{-2DYtg_*aTys~uxa3D@SdVKZKJpPn3F9l7308Gp5(wp?zVT-9?Q z1*i6HYw9jqM2+?*KiHyl?K^4<>3oe;~&fk5~EwDDHU5aB(y)cTZ=6)PrQwQW-P zvR7|1u1?cYP28WV$(zX#E$O8a8RAtbs-JCkDu5({2+2#^v#dsa!Nn5!@)Rf+0F`q! zs+WjCepNFgiD4E@5O1t;NdrvztG8Cac=MljU7I&X?$hycs*I5w@3^)NxH5)%tYFo zNr1b76c5Y=uIHSL;Kx|;=L8iJ@r2Hh_RJ_nX+QSBdb93=?d4Jdo8#HW!h-p~&Wazh%g7W5vRn-Wcw@^9`s^Tu68c->ln{(*2+`XC`Ra0qOpSae8>_T$5X zTKvxI+E2TJD0(pt>I`DvJ!0Qc?U@01+<4=>)PK+S=vRo>E0zagn@VnT#(4JcQBm6q z?B42Nhv2VLf19@~0(l9Ph2?g(2qMlH2QFdCgWdkhYuLW6a$)kJl`myM_14dQN^1=4d=tnN;yL}jR%Enfp@f2o8!A}Z_LXN=u z{zB`tKvRqVv%rz1z(0cQBL?EGaHAS++o&N+yjLbH?gnCSIJHjHj@+4j&?2|`ax)tv zTV(@$-eO9NJ>Y!DG&%)B&+gtaA`MTUcWA$77aWGSX-Oi5tOFD1`AO8Z~@wH);bn2(Sy`MLE>-g zolwAX`-IUaB*~xV*iapb3yldLfcro()u0hLd8Y<=n+Afc4tWnLCHk|ob{UeT`%yq{ z!tfJS5a%#HIs*M=Hn#F6GvyEuAIZhr6Mh(tTXrnFcxQ=cq!4kbHZiT^P{YIIoIa|M zpATsy_<+KlO4=V;u4a}1Zupk|eVzd)5MB3p!pY5PFq=d!cLBj^H#+;Am-m!&*?0Gi z(GzXTFn$A8Opy03&90y8w_g@va(15de(tUJSPLrK;w8q^~|+(H$g2E)rDoH{E^6dcTkmGXHMgPCmcdqSUR zg&kek%QOJ}#AHrNpu^!~f09H$N}Q5FAvx7E%a-es$I5&(QOilwY*8Q`gM%M-BTEpA zd19E-hwc+?d2^l+L41Bm=>5H<{&tS(!VB&ffwam*Ss+ikDU_18{3IY#ptd8Gs>&}v zx-ZpR_Q~0w^mhgW!LN1O`sl>wASt4qcnKQT@6K}bmFH_XIvPfao5=D!6FRr zC@#`F;aJrocQr)j9&wW*ym&Xr>ru&?`PR9jL2xOfo5HphfXJpsBxzLw?*JV(;+mVO zMi_fkf+3{4D-LtdVkj!{kC+W3#G3Pw#IAz53oM$iH&-hI4UBsVfrQmDJzuB~OxVK$ z1`EtSUBQUpVSsSiM)2i*^g9T`ykzwDBz1mHBerL>AK($y{HI15$Sn`ZHMu6`pp&xH zv!S-#Vu%F#E$S=skL)uAf8hq0J!lW4Jtd#lJo@((IqMAP3*iozQ7!P9w#YwCsdVb+ zp#`_rITI@#H7Tnp(z?EuL5zLRr&>%W+;^W>4Z}%(Zmx)`CHjoxtodAM41Ts;Zx%BwMOx z4McAt)`MHvdUxBE^xro)I49=cDATKr+X{`zKO>-4H!_tRK1&PD?YO~}FE-dwRXV$n z7>t^aGAAG{2m1bbza*&w@rXN!TEG}=q;fUf2V|Ia=}fGTAI5UeZ}tP@HC$1waJ>ZX zOc5Nf^{LU}TH7-lXwjfrM>72vgj~q5Go>@LihQF|qIfyVI1y}keD}~b{9{ch%RG%g z7r)}>3Uodtb8gVwUt?7smZjkekRylLHWp5Z7kBUOTh56q1zb2Gn`_j6!ZTn? z=TS2X+aU;9!04LDgy0L8h1f7N^$SHTa?~AAN?0c+N$o7ioT(&|IlN;c(a??Ob}k$0 z<|PiBk^^=?pfjgyDCs~z1)ISJfY36uFl?Kxt83#3t55NH0!Rv>6Fbiv6 zH>|ob2CF=582C_*np-`m?WNTStj--=mYBDiw#+@aE`zVnqdRxD4q9~i<@r+g`YO-_ z5``R5HO4niPV%m=6dAs&GIpc(cWCHPlV91k1g)TW-%U|eO&L5%%*y&l@1-6Mv%Haq zwqWGvZ^dq%|0&?U1zVHe$}C|0qrC>Oyg!=&BfZfAuu3WU-Z)FR0LNv2f)$10Q}^#y z6In~gX{irA@*FRF_yKTT&WK$a9dGhh-@?TSDFrosD;?0KVNEb)*~GA|O=#8Ig?+9o z3Axpttlw&rd0~lkdKraKivNTUF=64+rTo4(5~xAFBCfut=eQDRH{ENh(|TTy%!UD@ z**9l@AZE;&9qQ2iQ50PSt7|aE2WXHVr#Ry8vD<6{X z5gjS%pIE0XNTRF{?=KxHpr0g-P38zuagSh>E!{$9j!8W#+Q0= z6X}K&Rp)U9++-Mxw?VG909Lz3sAzx*jLLT;wE3@{7*bkjj~kI&(t*`SP60_pt>uLo8;XvM+!0zBWobEs>CWazC`=Fb@?6P5F=&(0sELW*x5mJ!5q@U2=lg#dY zNPQ>v_`5kUgBh*i3dO=VHm$92E2ewGQp*E5MnP-EOByN8HeuRF@FU>N<7=WU57#9p z0EWh|$!9k&{DGroc6otqnry_YkG4FHrl%VU8?C;JL z8O65XyQ8Bw6o&VTOKfPKcjh4=Pdi9z-vLk<&GaeYHU(MOi5m*2Dke*k_y&QYuMp`L zhby~k8Dsn$Qh~@fkN!@%C1vR5KNn#qd+_IMz?^0B>`ar6$`Y^pA>V!OohXu`R(6)w zoP7@MlV#%N2R$?Xk{rjI_#iVq6IQrUMuR_@5csQLB5^~9-Zsz~Dj?8j!P!fJF3uB! zdhy|`$b8L-{!}C2MplL0o^n|^)U*Nq7+y;-m05-ml8&aj0E?Jij$h|)9=#HV!KalEvsiZ5 z?#GD)UWC`#jF^4x!`j6>6nu*60S86?P5iaIe42FH!f9ar>p7D z3=eEzioP-lX#;nt>wa2!x1k*qg-Ov)7*NIrW8~5s)wZkJJo}Z`tmVKE_9|-Pq(S>Z-Z`w% z$-QVaTyvQ74P1EswbR3jgsv(Oi?dpRl zV6nZy`{;2PRfQii?dfP~a7smX$z^A{^g)j}k(P69q6;t#Lny)Q(JE|tO3Bk&p5D|n zHd~vC>&2-yQ`L))5lwtR$B(XGC)Y=3wB^}b;^R@V;N-HuZ_Te?-(^EL>+@!Ic&DnI zb0rkUZgc)uK5(|+2S(fD0;O6}j`kEupz5%Yx-P4 zx;~A>q!2+=4GegY{!I>I&;$o00mg_X0bz*twWqa;r1$5xcmm#^o1OqtZl`o~Byapm ztwb|h7*3g9E<#WX4%5w0>@=W)g`;v4)09emVVPX89zgmZGhlD_92{Jq#&v_!n^(YN zPmWkdJyXc;Z1=ilLs1ju&}`O z5A!_tN*Q}F=}U%qd0N$4QcZM8B5Yo@SY$+YVW1c#XDdkpA7*D2e>zdwOS*Q21=Y&9 z?YoP-RH~Ocs@R*^)5`1vk`udf@JsZAcAMBa%V||Skqb6Wz-@n)J}<9z(JOPrz1akL zc)TiQ%FvGdl}!MsoLwX4=R(GllMDkmIrJZEBy!B>DkqPFID9t;_*UyegZfdLIF+(9 z@HF*V1~}tWrg{D>7JmU}%#;S?U#VyIaEPAr&1x*{au^h$dPJtHgJ`CcQLFn{(O_LR z{lm6T&mMG#U**7rA2*c<3t>R0v+Ba@B;382^Zj|4Y?!O zi?HxO=;05enCOgq!)RNR1q8osZJmsYUW7JzqWz62$;DZ^lH0jBj#jXaLdh5%{(J1- zIrqE{WbNqTI7)-Mq~`2j58;9r_M^}Abe~)k(PJ_Sjw)vHr{ILiAK0vF1qu1P!Hw7SUFC9GHKB!oN4PLblH64y=UNQ>w}< zaaC_9gpBE6V+=RK_hJPdL%>Ov@yD*9hqOq2LwFi1FUSJw$_!S=Q~3#V9+Rt*9CPYW z@klZJi*!@*A)`f81A-^nl#s!^9#Chr8zRCPQ_L#i00(j!L^C)U43NQ_>pUWZYz<&S zf)JLWl7ZEGuHA(xHZPAiH+x;!Z2Q{T#?1TXAHbGdRa3;MiblqXi1aM=#21JA<}lyD zzW&Tnjv<9_Tj>bsh?Z$qvdRcu5UC+;ae(V(aYKabX45RLhK>_+d#zH`_utIOrMe+V zOh6lmGW_73-tYHMAmJQc0>SW6GZbDzA~_aGe#$#9`aP0$xWm) z8sSYL+(JiG=uwedNP-Cg37BEf`#mLZAvZ@Y3C$4~2bWPBjj=A`EdJ3%@A@0^Pgn6Y zj0xi(<eHC1>U&vUYv9dNMOA*eHL=29wML@oM&32R>HuEs{)-05}Z5eY;w@-0eb| zc}WLFqo~ql4{hx-$F`iC+G8gga@=LjiOi9~Lya0W|K)P3nF-E1nWE5gSV;|8wx)cD z50jDjesn^-U>WV~D}g&_N;L@_6bOlNw3f_`)U2?xSM@1JEsk-;1>jTYwz7??)a8p4 zwx=6Vzw~`w3z`<_09iBMX1lq{_L>a0RG@C6>C4H+7#=^Rqz^KdX>ni)lXG3s-aDya zfTmnXhugvJ5~FePqnxvYU_lfr$As0y6O%v^dd4ss`dk14SKwlzqRoaSeGMK%sePKLUw_ zv`oCmEg`t-%T{Lh_Pkws66yLs{hfX-?yHQ0q}HrmYQrCoB2B;$My2Hy+`lvs3i7Kc zN(cW^$h1-YrkSmf5aT!r=>1{Yo3=*1I{=T>1>_pE{;kM*YufNm9A*Q*5ya2qOI|=3 zD^%2+8rbs@+Cp*v`ibUA%Si!+y=ZZIU-JO~GZkHI_G(gMtspI#E|PHc2$pwQQ=p3P zt*my4th8bZ8=TtWp3xB?w3(nDR`y{CZbs_N3jueYwx@?uW=cje!w>AEk(ofrH_2}U z`f>*Z66+kWP`-pVD=&wYy$_I-?o75pxCz9CtxM-kmdmLf?U{t}fxuoZ0UWrR!d##} zi6jGqGNu~AXDO)O=QH)yG&*Xvk2mafK9pvnDxF{M35D_?MB0kn!KF0~=WzK0_7^#e zx1w|=GHvispv-0%Fl0fq+{#lMC1Jw>74Q)Ck`gZrEX`diU42TS3@}~V7DwDFrbnMH zY>lB@Fw>QjB0heP(pdC6wmD+*0V==SG*PT7cGO&m(c8SFIIc!gA`o+9ky*!*N06mr zd8aKz+f{DKZw(FvDv}`59}vBLPVOZ?D%L{LabT<0lsK7*w>1D9DuLJtq*+}&2;33h z4W7jfA*j_s<7CF%(RqcxMya&6lfR$CyB!lvY~eVve?X|p;e?#jYY~l?s_N^{Ukwwv zvHEZGMkdFMxJ&$qK5dm~8O<1yDO5|ttdV;NR2rpR`m=#AnuUhd_e7RiLLLx${w>w3 zwDKfKQ#^M#|I7pY2$Hon_MA41JbE<=`vNlXuEOtw$WMS+zt+N{g>U7&6rnnPd)3(8 zwVZv#Yr3mhGHl_Bo)DN(iAT@s0IyeB6$=F3_C@wK{Q5-~X4_Yyryp5rnmLx0Yi^fr zAN0zoL0DNJg9UT6=4C=FD-}L~j~-?idCME%T&%+IH7m-HaWIhTt4jtZtUVbP!DvZ$%q!>Y`6rpg}@Q(EFE$ZRuEyJ9VA|PPP{{UO~@E>VE5H=dPTEk{g08o?(dCNL8jwn^!?xhOO zE3$sYaE85~YL=it2c;I)jk|KMDK-_T;3FA^B_ma~?5d^5YUnF3=bO zsR7z!LrR{sv9?zEhY0~VpFRtvvuZhtm+60GyhPe&GgLqm=^Bge#nce%H&NCy=+(|c z4T#`E$OZ^L7HL${CMM^5jE)>;xyWU>#~_;FrMQ?!(XHxg_N?7?(OOIjtvI|CJ%*y$Xt;Z z%D2puW9=>uo03}G(oK9c;n(c`^f-JZ-L&o+%kS5~Osf5quKduLE4csePEAf>Iu#gw zw8Ma*_mi$W?}>*NpLNGWypEoR#{1sC#@^lBLS57Eu@;N)V}H50Nj(8O7EIxsA?@Hg z6+GjZ=gtX3-qK0Ve}C%%b>3dNeAbVxviRn~-`ML^2mrIQp{pkLTsyC;5P+H8$bGCTJ&S%!Q6I(W~^qdaCcL2gR>qKaXXnj zIvX@max9jX2it`;bANY+>z@rUU)#qK!f~R{dS9HxCj^uJ7$g^;inp=3J9zu(zG6nW z(Ogh=IHV%i5v6|3xtoh*uDh$Qx(m#U3fwFF>!T(NrDr?5>H!dQtUgSj^`#!IIk@PA zDxW9=y@Yb!3MrcDcFtBhh*<3&TQU~D{}Q0acV4bzVOu;+1plaSzwfv=aeGRIf5OM? z*-{8`Z2MqIcpr-~rv*h{q^*bRBG)7gg0Bp_Vr(hF2u6<=`#KZ6sObAV?hXL_(h79S_$)Lz;6f zcXRsg`=sEv%6CES`E?(6Si5tmUsu44<2(Znoy@^A%x_a$J0TP-l=E5nX5{V$pqkKA zY4q0k<%9$8n(OD&_g>WWX?J$DbIigB8~eCIroou*%ssoQYLgK%NJ&YLy_VcjrLy;T z%M*x5a6w_vGdvcrXB?EjT(+Zmf! zTiBZYOSiW*Z`C$AR(@vn3Lx#qAW7h^bGLb?1e`SUxvv%$fSpgTMlCAWqzVUG6{pA; z${*WqdlIiW8j$mIZo=4UTdVQa7M6R|{t|c1n_DBk&D@Y7Mk$?#UZs*Q#EA3@;mJAg z-&*-ONdTT!>{$_>uYz8Nss@eNwv=F?&V=SHY*b+zonCNeg%}KgfI9DHf|RE0lG&2H)i0V!)9kV$8fwmJxj3EclRaYS2kUfx1dF z3#!ww*$kT$RbMSLYerL00Z{R4?Bd&@&TOq7?F9^tqE4!K94QYGKXi(rr&_aEsuM&dW?_)yb*6(obUSUrSdUvzd!4?8LKp+l_uG{4*&ji z8oJ8&BloG15jzgNt6ttby5_|)#}?f)7Mxpm%Lz)!g?<1>SXF6r)4|Pi zN48D*m6ih#9Nw&^@I`-qPBK`6I;cr^nMN*uLu+V4XNVVVodQJ8q1d;!V#}_zyO+KF zci~4;uI-3-c`|qN^49z6##P|3mdm~31DyU99z3i(XiM5k6SPal{2Y|q$Cc^K%@sU; z2NW`XF)(%1VVk`mpEWbIGK#Cr0ZLY3Q9LQOxUjb$8iqgkU3g20Me;1C5W!=QNF%XQ z?E|cd+lZY;8KqmEQ+h$-K>r4sweaVeG{%%rsV)QpbLq(f=?-IyxF^WD(tA|+SdLJH z*#!max0C8pVwmxcsh->8bzwz;y}_1Vc0nD~Iv zlSWCqS1vu+RK-GnU;|Etg6!?pTTT=u`Wnsl4m>gpBmcl(TybRZi~^0Ey=72jx)=%b z>4y(d5Nt7<6z12qeVl~k&k1Ny18MXaPl#6WoSM2kD*4QTWL_wXLZavw%ol8u5})k# zf)rl_Di*|-%>Dzmu}+gmOfy7h02Z@792TCFbA>`Ebk&hrJq^SeWtTBwZL9-7X;Z>= z2^DQrK880L{wQuRsQEEE+C4V#r=?39^{XX_ssn6jB`o@_xi(vZyu1MeP!p(dMKsCYf*b}+ zvVvcwI)v~bH2&qL_oU&7Vcx7UnxZxnni*m!>{Sa^Ml}E<6ld)gzn&QCb`yQ=`a5%RZ7r|r7{AmKXmkMhtbl>h zk?CfSm2}%w!!Ar2fqivgMNKLU8z{@*iCF@)p^pn&WHg4g$E3{VKWj7~VGGP5cugQz zQ*RcLA6Z$@)P=Ohrs{rbkrYFMqPzr(PUM}tR3g@wwFBK>0g{&o#tqU*XnPx2$ zI9Hitmt2iSR|3wMkDF3)SQZ$s6}$N}0)4h&wrG{iR28dYGTilMESIR*nh9{Jr~hdX8lC{^oqhC?C%P#e1_s_#w!AxUE<4Ch5gw9;5cJ)Pv_@3WEZjq z3eO?0NSd(RSFi3@0?6IsrTecSYtC8p>AnoEJHKLWa;wdB82j9$Aec7EIr9lG}a(RB{N zp#Gs#*1FYEGX%{TWBv{Q0?iouUfI z5Y#^6amG89j6fsPm8UKkPfLM_p2n0ua2I28_QmR4a&Nw@LpgMTTWEJGhfEjgr5 zbAj3Q@o;x>_4>I%ddLn1YM&?#gCTKO?rdP}&*bpQdILB@tE`kA#G+I3Bnv4QwYs1I zMPv;8)vy{lZu+7b zgJEPkaF_$m%3wk_tvNDaHjl(IbqsK=X>yj8MzrE6TMQU~;x+a9aD!BPhToLDN@r@D z=zIE6YkvTM0q{T3k~Xum=xQfeuOc@Psh_26lDs{zz@ebzN75&0Ioz>fXV_$IH{z@#8}nQ8xzwqNXj_w>H7i;&XXnpkyt!H-1LB z62RaJOc6t}qivNSMieO%1q9NPEX~|D%}6YHWXpx$+@}N?5ic<#^L{u#i-*{`wq_h6 z+Zm!bn}(a>zJ069*_hnptGxLKKwG~pnmk{Om8WyFB1ec8P|>+SWu$2HT{BrH@!Y1H z`?vjk?P4Ifg0%>GUOmPVbeIPS)h>B;oC6}@gQS*#7M4#-Rsl-qh8ly4-`4IVq636_ zrqb!%r~0pRTt1{zVbn@imf~ow*==|)yE}MoI3u?V8G8qNDlRFy_b;dj9uU)07VTK5 z98s1~uhW)3XnZzlLQ^-Ax|(@I!yc9w%lI#!En*rNaPJN)$*+r08tBRK@B z<_XWg`H~9Arvhp~xc5iz)IJ!Ohk1C|u9@STd9GH^u}&RPU#k>~@0J?Az*X6kwzYoa zqC7S{ZSyE~7#XicEQwjAQ|b6`!A^u8@DE55t>#TRCuiEt1p^${#p`o<-sTYtm|Dpz zsc^!IOC0isD(>-h#SNL`204{h?=O=E_h%639_$)&OlX*U670Q2ZzTE@;N94m#~K}` zz!K$fPP{?P#^RnI`~vxMUGUg_Dcc+KDYD7jw{>DHIIlI;HEK$U2l4}n)I+@b`-|>A zu7=7nZr|Tl5a1|>;a`W%Z4rha3$#WdTqmEtFiaJR zevDO{Op>*tAP}1SAXO%X{@-_~s(R21(zhgF|W@TCu-M^vg6a{XK*+oulce(yldZgEaMQdv$^O$>mF*6h8Y25b^PtNTed z%6wYspfQbMram^o>%WL&>g+}{Q-o(}$>BMW%0(7~F( zrBC(eM+L1h?EL(Y4jY_UHkt&|C88t3ffz%vXR}gT{l^}Fe$Ew@iDVl^h!u?{)aqB9 zAg?BvAC}KiG&8I=!fu5o#WRo9`SNsC8GemH)fi~Ysm33z9lJ&ENhwgXGle$ZfVSur z7)Oy7fRE z-kFYQNdB-`b}uH@kJuZ_+i$;=b!vbiO}8L*g~%EoTxUKt`XNif?YE%go`3!U6&cS+ zQ8$c)2mCpE{lo`Iji_^HH7(tGc=g5*Ea9XGk{B}nPTh&C`(&&4HoLS^#3u;gO#0$I z1YE2*3X0VnG|ESG{#YVor{6mDcYSq>(e%jzHFio^MDIUYv^g4#TNrsQ&JEZ;6Ja+b z6b@U09p-@ph3!&{sZ2lWagnKmR?_#wf?GwP&k;!VC@f(eu2xYY)WM}0zMaQA()gWn zZsA8EV(Hdx-TqL^38-iJ#K#WK9f_V9i`^3>X_OVSHW*|#hS?$-SVh#{TD?w`E#G6n zs#coexH_fP$OdhQ;7#sxqj!mtXE2eyblUeq-0*a@E}+E7bJZ@;dfP`C%dKxp_B)&) zrmz`Qa-ANETE;eMyQC!{EC z01PF(@#Gn?=#h$T2LvR0tHh`54XMI<4{W`+0sHKN8o{YE;z^@X&FYvDg3FGobnd`2 zqVM(8^BK_j`032)#B^cbJO(+4O^0bJ*O1s_gCQuxaSV&OP2uH&;Pg=f0ps&EohqSnp=X84d3}i zD1k(*6I3H<6{RqG%@~jj7X^Q6yVA3ME2IKAcE0 z!6u48*dZ7I)@YA8%x5mYMy6{r@w4eOd)!q4RXY&>aEx8FVzpQKR{K`#g5DiZrPE`4 z(dOZwY{3R+4dr*>FP-eToNJIjUGX>l>dg;Tmm+V3%WgWok(l6K&6wrQ)j{;{OUklv z=Xyk&@A`Sb$ZHDcHYFDq+r!Dn$^DFv&e8ZBU1=o^2-(e4)9mFe+lQt1{1lM&bG(Ff z1M-`m2A6s4Sh8zv3q~i%d0MQ?DR|!_t$7*Fq37W<<@b3oKK6J9iId^Kz61LcbyZKh zkRQiMU&UqMg!EqX_4ge6#Hi@fZ%cNx2yHr7*BpQ5{T_^1m~hTo49~9F zibi~dcvNildE6hU8mTO4MC^MqI@S;vc-8l4Kzx>q3`tj_qi-lQ^=p41&yww=Oc~-1 zW{AdzY6KqxdBPbu@BwGTWkXQ~VHuy!F^DijM*wSvPZM+A0UXKAoAocQW%(Et_!7Y8 z%l`pig%iLi!2w6M=WEmD<28uGCAun@L~C@=xZ}b1mJ?{8ai`k%_Ur$H3A`4?#upn= zr8ETMk`KQ13sdW0W8`pH3ajJ_x#9XrW%OrqBUq0H%mz#MDL%gHYO46!;lztDFl(b3 z^HmVhrT|L=`|%DMT-Wic(f*x;kR*6r*@n4!0jIQu7lA6aVk8JOzkew500cv^yq`VN z*p3(2p)FwHL>UMm7_%L!7JNA*xJb4RsMN zE{nQiF|>GoSp}T?P6-$0os|1+OuQ1V%Lf_v&;8tEvP$^Qn>oi?z9(3oL~^1X4ll`A zifIAK^K0qxUPgf3+CFi_WaCzN=`#S657Yp&pf+SfF<#43u!L28EqHi1CDx?}dl@}Y zgwzM-R~9Q6El~@J=)~=1IE>q_o4piJ(7kd1+88BA^x^ZlXBHSG2YyztD*z;?Zty`- zpa$f%alUQzcE<~{rwRrG_*H1Dt^*{q^`LC1Y)n1_7?@TZ8$Hho(*pzlgyL_c@4;M9 ze%>VyGyER-)I@Pp&EA4G*CK{qY^%9V9G^R24vFem5QGS0+J+kB;`B>!f-jh97FB^8 zK+2x;lrNagVZg61B%T2H&=vUVBVhhszf9QV^&b|@-eFJuq<4wDYe(A$RFJ`>(xb^taSU=#2y>j*>??EQz6$Fd=ijHZ;p zW&#F(j;$~>C9FV{DpmBq$xKL*98~;=PPrYS=sJr}!KLc-O?}2QYx@q&gw#2TuZlSk zXjNO459i=7Gi$RDj0|!jNkEcDq0<8bVe%~Rep>k6l=kU%v~ndBirE<=A7|`QI8@Ib zodaERBANRYXZH)?2~Q=R5)$kfbQ0`u4ha1w76#RkD%BM_PjF$H`QF%h3=T^UX;*<< zf_};jAV^mm28>0F8=ziUQ6e!XhiU5Wd3$q4D4RF9S22xP;lhC5QpKeqOt_3G;14J) zDpGW(>|tpUI~MD)PaxwK$C6XczrQOJQ7aEw+e{BSllrr&q&N@*kgQDwDv|(^ZJKjA zB(0B;uwn2e6kljj_Qjq*^I*vcqPQvbfRp7otu0v+cr5g10Xia4hM^*78T@N>QF;=@ zxe80$YYr|2cqFlWC482t;alV+tNe(1`GuQv_aXe(xL#Fsh zi*2+>6yFgfDqwLO^_>HvqpfzHz_ znhcGiPX91SV50~SGg%tSeBs^@C@gL#oN=$Bpz8aFB6wWI+`d2SVznel@O_+7&SiXK zJh-ay5zP#)&Es2nlrcaufDk~;K%WPJYHtbc^gQWbUI={L6~Yo3OZcVOO(Q~~YZuJ_{a zwG!?%re^~J{clD`fdfN!&>AKJc}I!So>-McR*8&kWt3uR8JSS? zcR7hN4A2e*zsiw31ZaBt@b^?}X4PC93S(Fg@J!$eVZ;IjO$7r5Sh@;ID+T9{J)9za zPmZ$@s6i-_l1vPVDkpR{+!0F^WCNRpaIS2{ zcaJtMBX$K%ZP6SgbU^8O9X~4LtMWGTYM^q2+dJ{3wj`qk>_wx*<~K~jNX%YGi8Z?D zegAkH9zVTzJ#zxt7S_=gIY?-N-$H>i!C*h=*tAj5fVdk}_f3|{T;ncVc#Q_9K+*Q`)4`NX3ZjdNu%kd7J_MN z4R@XcsJ^XO$#A9*5I1@9^78AF*BXHJ%3=bP^Mz0`N;#~>OvhTzv$GnakiA0g>Ak`1 z_(KkkA15>E^!h5K%_r~h_EbzM>wD`!Rk;)8sZ=MasosiVK<+{FqS3)v5xDvI1QhEg zc5|GJqOuHIJ)T@D4I;=BB;&}KKv{9t`syJN^DB+zbJptYxrW_~H*EyXp;Kn|8n-SY z)S0q*jsEc<`jO4^la$@d%oc5@>Ni`f0u*T2uhlU_|54~82 zslTrf{zF)r3K` zDIsgvzdo)_NJKL(QbH&r`5kI1N&rTaDoEfbY1W4QghM55Bt97t^>h6P9V9L_gthQ^ zCy~_qmsxf1TQta<-K}uHiu_;g3+k`!?)tr5`~u?|f93$&3p^?Fg-*gpz7nZ)%4;bG zt~om>XlVUCpBmg`dXk56V(sJOZuM`sRnI#76%?bvTAep#p%|bC?%Amsw&?P2?Bn73 zs`2$CL>_d+-6rxM*4z!Ox-f4wIKn%QBqoQiBKQS^>dk7gawXJ6@DaluY}k+BJ3ty6 ze43$wNh?j=v)(fnyzJ!KwRgJue0@EB(_QSHN-|NX1a&#`P}8u|VNM9Wzgyqy>rnd1 zVo)eG&|s0k(OmF5@{2VGoZ3PSeY=yY__zD`;w{f*r5vFR2rC8O2Qv1VG}-n>Hv$&q z9zi9*JT*u0fVH%y(a`{WI^Dtj`?j{Xe>~jXf2XeF>GvD_?~wYVLnpKrgE7(M6M9bj zYDm1*D$I{fmz)89IX&x~tb*(4@WdCYbj8XRo`l@4h4Vx~wHN4)zu|AJQaut`aV>p! zy}~T^25}G{c(RO+2ED$60jXxm^rCSHGCMnR)FK8?brN>V35|Lmv6?GDTHxS=hVuN^ zgfa+3Xf{e^em|Ru%yx~e=6r#nR;>Rz_)&xqmkom2Lo~*0xsmgcQ>afdXw%>5)r)2XZ5>e);Nzu*RA8Z=f%q{X2OW0sLZh8$l)!y9JV!C!7 zOA4gQFKkxd{Z)`~L^j_R!ANs)lKgb^#V81>JLQ{8ZuqfI1D43@eGLHRQoChSAN=%% z!jl`i7B8P5{Pb_?KEOmn*%_wy7Tp^(8V1XpWTV*1AqfJ(f$)2j`6wZe7i!$If> zCVIGma{@HunlZv1vLuet5A3jieYv%}z4`slU*0&+_wUpnkOi*3z_p&nb%;(dKwsax zZ5nkHF75b~PE9(S;61^Q;GX^JiC0l%LJ)&yfcSu|3C>( zCwKp9=l^wqS8z*C_SW(aTdhaA4$mnt{(wv>hX$7$aG|G@V8khmwWed5XIW* z_b5-VTn0vHNrayAq*ak<+8giOOb@sxD|wa7olFzunFYxiil-WdCoDmec%qbzF?b5y zextM23fny1la|@{WO6jqK{WL%ze78fHDm1LZqAr)%bF>A;RIrIA0xub<0#_Os^^Kf zHl3|@Ua2_~vK6BebvaQ(k=NK8@vr*TCI5t(W;f4$JrE^&+UMoB-#2kyPWw(_>s(@pGXp0iBmG*ORxp$1odX#LKd4 zx$E);+$9wF*7(6W77w`))j^u&`T_eOiY# z&g!I4ysqXCxz2Af?}ZJ;O1(p)(k_C`Hrk8NQAAwvJaF_#sX{RIq_4%Kz+>s8`rEGm zz%N5QqOz!>OU0kj%8m?DczS81?8^B_At%`xV#Zpzvzx-_fLPDg z`@qJFul)pZ86M>P!*35ucHd_fV zu&*n#aXqSG?#pAR<=MM2)G!~J^}s01{l-N#A3y1Kzt->$`1n2!@A2W{UtNuDTI#}u zs#7)f3>40NQWidabvYI_OSh-vaE$!_|3a@WWtLc;Cce>U3cI_t5^zC1|#?$2PxiP;Hs_ul@w*|k8 zMCXZK)Zh%t&Pl`Q#!lpzaFupEO1G4Ol`7zgEQd@kgelWehnRKIPxSv9pT|j?L8qub z+t-C(J5LS}_*=N{Vj1h4i^w55L&uVxf2FJjA2uv2Xy%X4g@7~ug%%u>RWIUGuV5w5 z-0y4($*5O`orn3y?$9pK<^8+8vi$GHYa_UNWw~>l47_$$R5O-hKMVOo@hU6+H({pG zVf0gT2h;VqYxx!;atN?qC60t%wMry5rUWXP{OJ8eCy#X#{o$(WXElwdG}ug`r!eJs zugRD2S!J3uEOH+-t#{T8vgv7OreWI__DM0y4ryb*VW$JdivD(SM7D4DVzM&Y$`P_T zFD?^EJd5XS`w5#!@n)nDstzbnup>kcTf|BRNgx|`xMCJ_U+(T*mo0nnIS3R+=?B9e zDLd|PUuh?RjRiSe8h^JQxw^2VLLOPp8*-TzHq8sg-G540a~w@a%AqQdtdQWsma~?e zPxG%2hHdK(UdCYta}0q|hRpo5oUuYTwFJh&Cm3B1y+pAoXnGp*LuDY5_-ftG^7R>Z*CO$tAztZ&T(wz~so!<3{OqWesmSf?`}q+(aHPr`7+U5`y$&?&*fZU;+AS*h>2OZ{Bj zI^rhP^&)$@68pww+da(HD@*bR;rVeaZA0WR=x%J%WTjqjva{Op8K|UbbH`yaQ!~YQ zp9ql3ZJ+3`2wb1@V%N=-#rB!5GJyByh1T}=rcn+iw zCdX2nMAApCoOD_Q!9;a?cUUZ10<9ZIq!p~szy;U4w^(Y zVZ3UlYd7gEcuI}6q5oR4=F&{(Gy*s~Y}D6~+^5vhkg(7HCjGBu>`W941^ukRU8s!N8ruWtbsfyS$FpLHa|2N3%jIkM0x<0KyKFae z%3mwL6eVk2h=IwAIA?a~D?|>!&UTZ_JEB+Fr`A>!Hvr zw;@Y4y)RceBN{W#&yR>hSYAV5exP2Hc`~)j+l!#pt_p~tw2A7lKP(Sp(JKf~;w+`k z>wXUFh~SAllf)$QxSla}*pV$w@*cCh?xIDmKYA-obGy++x9^|u7bV!d7cVg_@K*;q zSU@Hp4=v0I-cE%7`0KP$$|);(CKuGXBt6Jx zk{o{(|2q`nf6+cFDDr~sV7jX&`e7kxX~A~Yv+yQ?a$w#c7q}gEqC5@Yc_Rrcmr(xMQf5$RPFNageabY z+R=RC#%U(OriWKrOypcVf8>0BUm~C&38lVe`UXv|cvRe@i@o-Z&HyU<2Gul%Hr2Xi zCN1gCohG_&P4UL*_E4!xRjb@vLtJ~!f2CYAeqzNR2!lWGGYza*k=RU-s_As_u6YgXbj*o9gn@w&gx&$h;t+tX$_aZKX%nQJCz0o(4CPPzxx4xBv zROT_&VK%hNIt_=eFbf{FcFeE696N*U=KJ*5Ei%)jN67nIoxd!SnvC(Y46HMMv1Exg ze*lEpHOwVu+%eg)7t1XbcG^$6)O-(K+e-#ak~Fn;Ac5GU&sYGFk}(yTtlH=uB27!& zunmu8>%?8MjWBDC>x73(NpnCnR5Z=vxy?uZb95QTEU( z?3W{9KpfD#pY4<7^jQX*$C0V^ao;VE9qET-UUC)TgFQA9o23}j5ki5o95?ME!b_uX zPxrb&t3h7)-y&}OUal-nRipZx%C@7?n6>=75v|FyBCC95ftke7r z6LCM0BxrG^+VX|KKW;d9 zvzpat4B>FZmKzT1>IaV%pNhweh$aa)x-u?1u)YV=S8rN9N^`TFLDeTtxBX&D z=sH^n_a`!(G{;i(V7CjI|JQg^-GnjW&X}GS!U~fruA$tIx{zZwF3kkAc zR4YCg8x8$=&c%W!*)Xu{%k@wm3OC~zTAeDLz|x<_4~ShbTCiXg`e*0-32Ol^{>=(k z*vtH#&2N%|I&iGNfvOmyUZxS?4-Db^rCXX`TD8zEEzL5Iq3!7l-z7}g#XtxV+)P{5k0gZAW z?5$~8c1XR(4lk2-!ZyNDSUttl>|Ho?=tS{xSc|80O#K`k zgwl3mNFl`5$m?4j>hlXe9b9HV7* zfTBnV!!=raI`RsJauLPnbavvn5%j=y>>=kc;~4#@&H9}3qTl*pqZ>RT507Zb3)X8^ zb;3$=c_#C!JkT@Jd&=CS?M2Mo-Av;&G7p8a+S3w)l{BXPnp3Z<*aKH3Qv}AMX?+cr zb%wODI(?9xsiM*i$h_%F`B_8t6H!zKPl!suXMewsy!2t}N%%tbRP}}GtKbXOm)+-M z+wR%tt3Sx+t2fa9J;CQ)H;P3ZXp#p~yBR~CMhW<>GB%tUv*Kdw<*^cu+F-UO2-n$q z9qh*sRgjZFma6)!8~07U`z`zpo;>=xkaVpm#gJW)lux|FmR%c4FFNiON*hdjWm?hN zmcoD{VO7Z@Y0cqpwcecA_7T(y^CfFIKb=5NZLiXd6ZuQ(7L~2Qr z4M*F~74_OAh(+CNFo*o9_!Az0^iFsu7s?yXPcyBxv0Gv%m*2}v8OL>zs>;*pHeE*6 zU4)HBd8_xj-Nx>unx1y#QxmBK)g>0{NvWsa&3_Cbo7~uAF1u%RlfwO}*@)@}r0>m2 zH4vS?9&VNZf7NV+wS#F7(HojFC!N2#pP56o3swv&48%j&BrqLK+lf_Ll1*UkqSDq_ z-I|J4?A|H2L=*9$W8GPcC@NN}7Vqh;`ecuk_n=)FyA@}xSEzXux-vN%z|^sa+h8%N z2#j)FdJ-a1fi?pof;yh9Va7)lt&ENC=-u(%DYM*>W7`@6w_?Xj$9MTIYzz-MyCtL% zdu?6Tg$~xaL8KLXd{NeeZgQ?|-x-VdZ2KsJNZLZbr2^OhXLpyRAmHz~D^4}ca@at($UBZmQ_h9w~ zr7@qpSFA$PbBp!-Sc$xV74eA9js5YfS)=`tRikjD`EpICc8l@rX^qAu)^{f?=JDb= z;wJNT7XQm0>5JP(`C2l*ZKBHJ-Y#WnywF(RE!Cxch5cDMZEK+UZVKt89Lw%1#AUAi zF`)a<4z`?KcYBi-CA;G7EL33ZonD|QI0CP^;!^VWVO1X`)hF@I#& z_!k(}cxT@82Hz(AV46&R|`L6N(^x6yk8@P<}Lnb7T@rm|tSsS%XsM{2BJ`2n~@9 zk#ExxPwrD|jFjq4L2S0GW8ThfBE*QNyN|fiLOef?2#xrSI(i8CB+lR1#5TEyw zf2m-4{>)0H>h9T{3kXO`wiPQv_ZyTN;AJIco3e<4SyuVnJX{<)^lnbyi}2eb96%({ z?^;wb4#EK7ZhFp(bm3n%ymG!&F2kXx8N<{jkWzq|eK{&mHHO%N?2x>sKJm&N#;R2! z5nCXW9_@+{TeJS+`M2Z6U@4FX>m@lB0tWinU-`N(KLst8pK~7jLDdeL<=R2NIgEFn zkB%Lz=7i594iJwuyrV|jsh_Wa9-dd*!s(R0{;|~3C|Wk+bhEh!#wQ+h8x8wOLXQ!5 zJoR~JmGG0A!?`3>_JQCTmPfR#W*r!ZdV+5%v>3zN6JNk;8F8Wu0elcAYZf;E!kP>^ z1_}#EqcJu|`i=6;=f5NOS)Hms{z#-)2y!6tQ8?SjE`a6}OlM<#AbyhQdVw;K627*K zQ*T;`mwoqmXTY+gz2m>{UVI}Zya8r{?Revbp7Q4~Ma30|>0fvmN@Qv3;3Y$H0fFMruHV)aU*kVJvOom^?|1p>Hp+} zqFtBw8C5#P#fy9emTu1RaxB4_s$4~Mep`mrN}d^4JcAMYbdzS{SZVhJQnobdOVt(`wQuzB>>1>iLcv&YR zvZ>|LsgOfK8Dk+oofgF4zIF^Dh}cFwdz^JMQ;hnl_;UVbdmPTRqJI+Rn1p+X0>p-3v9FoxrIeF+1Z7^Q)>;o0#Rjv)q$ z9W~5??#VR(Tg)ko7Zvtt-pOH2{=|gL{28} zIZ)iYLTD`WHa_`@-4WWHQGIFwg&U?hq%j-zXq$^!Rp&a*N)xj(AX_dS?|=zBigGrc zDib!Al>jp3x0qmh3f6G0ay(RjMitT!!!0_10u_S0ET(anvXJjVZzMPf)gOT(rVDTUu@$GOu<|&LQXGk#H3*= zAqZqN9w8PGu&fUroo~7@HN4ilegj{1MRWPalfE^!syLoeG!CKp0_a^TMS6d3?Io$` zSratH9RYb=)xFXy4qx$!K&0_p<6WcsUH96|+si+O^80`vY5D8;J{>5!`m-Nr9aC6$2!?_URFhimsGJsTx4M;pH(Yi7vkN3-IPt z!sI2*?CW}N9%})hMS<0@UBF>Bin$~Ip3ak06pP{%TfNg0V`Xv|XmX}1RG`!+Z9*ZV zmDV>5!WLQB&l1M=C$0l>Cwfz!%|TK^XkHBk&jJ1g-1ieJS;ABFs`r?x5=JplXJ39J zQBtpRzhY9gMUUv+aDG*dQ~o4$khYhbJ`iWYO4u0|>IT{xq@(v)^H2HDJw+{LL+4 z>r4?tMJHD?R??PdA3{N=kQ$e3qm3X85&ZB=cTIm}e4Q%2aQ+Hn3(3s<9@e=S!@)+P znS#6<+Q{)c(XmAy!3g7LqYmGH7R{X0|JgRKXJWq zKG754$s$G}{xTgIz#@jUFwt+O9D3m_Fxe4F;+yDlC?RB z{=bDRRFU>3qOM5+)2ZH&5O4+sCr;S-vEsHl@nZ}|+=E|Gj&%q()2WfBX>*?Fb?Y1> z9*P!Tl)=6>yfZazkpIKa-h#hxtUvpPp8ENE%4YzZXf}i}Q1>Xut8^5%m*EJ+9 zzp79hg0(0uzdwD{^fJrpqO=lm&kxu?yqjB@W~eoKBFpqstvROw@eMSs=@qg)KJQJj zlfJf~$WxpySMq-*N2eI-P!7^AqqFUR{WCLg@eWv7#`JCMcAm5m321not+jvyKDZH# zw3)j+whs0@cF4ab7s%-l)sAHl9JJbocm$?tcRk|IpK{H~sLBvm?m>)G{7HfXwFSkL7)hi9)-C?w|KDNE$s6aKX+!{kO1=N5V&C4% z{y)cx;gFH)|=%v@`$ORBm|L+qK<${XqO*; zqF%lEpMQJ!3z-RA4?znKrM;_~&5e_S2Mq2E7??XW(prW0(C5aK(a8QnuEju}%M{Zm ze%xt~*o*smR#Qw+ARUnxkg?z(4FF{|5JFmT$em|Mk{9#76noIUc_iT^qfhZq9R5pT za>vKx{led$kYqAEIMabjV&vi>OkjAv+z-Xy`g6K6u>rnnQmiw2`h75c{=CdYtKZ7$ zQ!<9p6!HXc(ON9S<&P#xjALf&ix_n@Gb1gyEXyVN%L>yW4-7FDBH_1HGTG8t+L8{@ z+f18PU4vh&XUB`Cj52vY6jzQ3vrXrP*@!F>OyGH(fuARs!cT;pk}5&An)?^!+_mW< z5Jk>vbkGh*TSVg{&BT+ovG;WG;+n+E1D&dV!h-BC>I z0)|4HhT^B6Rd@l``(ko9Otk(bK=&*jhfJDktZ+02YFPk;e44tcH_fn3T} z-PhcF2(BX4Js>R8nPftiztnvnv!ecVb2mTZ|e{|12YkaD>sfnuSA2RP! zp{$zeC@C61PbP^pbvQI+ctkuMZYdB7yzLf|z&|Zv*Q1YzMHZE+$9k)pKE)bjuwI^{WuMOk`TFf0Q?Ten(4Gt@&D3&A0P5@Me(E;zJH18(|jWJ%R zBI#|zy}h$|E59Z0yY$X??(W_#UU0VW*XXqw=Avtf_E_FX&P>oH!Ui^~@ zvI%%{9PpJTW{|#g)C@>7F;+;js6sYuOj0H^_!c1~W+6l&=^;qjX7P_q-fqhK$7Sfp z?=rx|fuAA~F!u7r8p6$RNmd{w0G4__P6+i)6Ta|~jvgPEr`Jl`l?PA(aGG*n0uUAg zQrKGlN`=WNG!T?^Ag3FT{ri~$&vypGkIa-y)H6Bqui9mmXot%DOE82v^#NoCI0UYs z`!OsDlNlfQG-P@IO$uD_dtlbxc*=r;nGZdNjJqNAfMoW&f*&54OIqs;;oM@#T?r}l zpE5)Xf@}+mj0BR&F0?!9VX)*c#%eT_G!alfWCwnrvfA^^Xl^S z#NCyR0xD@;O7-exttn9WYp6r_Y^Ol6{A)gk7k_ZA8DEi8aRBr#6VnY{bNl2wlnaPS z2Ed4p!nX7T5%C1?-nP`wgC-6bH4r8&(p;`8-3PMiFK7sS4+-FhM|$wRYS4?)v7oac z84Ll|5kRBh1Km!5f5ujyZX2L**dY{v4Bfwqp}>7lVV|fp%V9&DzYcP1F>=NkbH$2g zqdaOkU5+&jPXiug`%NHTQ3&NC1~*y+83v95zBr3GGz9WwN6aw+b8(0|Z4VNg0FA!1 zA!$4QzWRG}T;)Kxz-P<}Uqheiw~{|%kolnI#Ca~T8qI}EN@T$>mt53q{VK~DGZ2HR zTztiMCpJ|n^+2iG%Do3^m3;;Jp~Vyfm~;jlB!CLeX(=0!&XO18v71?mtCX~eDZ6O& zFwoYToEWhuLpAf53N+cv)fr~`y+O9DQ>nEB=bn$o{=_#*yHStJbcjzuFF7_k(%p(CvnwCEqRy zO*S;1sG~ZBLerr$EYcA*26(hwdtsw;_=grqqbo0;*sH1NO$LXAKZ7W9o$!U7fh9R> zsjX;bh7z=5?4l|+qn=OfkW$!d>o0q5$7#`1%a>;yM*b~HAz?qokz>ftsGWt$OplA~ z(dh^QKOc5Bz~M^7z;~ilemUB_#rrPQ9ih-Z$7I#I==?c|`FmoAGg4|I7E1PHwKhVc ziGYBw9@i{_lC6Yi)Fo3Zcr?+6qK;dFKXS#d5>S>;r9$~mK^Ilkf-2;0xYiDEMh0`2P-F`Xg8Az_{*0}RZd zntQyuJWL%~nG?&F}%apoX(xXo};F^EgC75Li(j+0onk#OQy)m^|-iNQ_{ zgf?z8=yKWpSNhwW@2@ZH*%G!A?g)NNz~663D4|@G_VVB%Q~C!GLX31ZYjBy{pDnymQe-V03I(vD7F1r?R*=xMCqRju^vk==)sFU@PpFCPKE zPXf+IG(@CxE?4-wQx92P)ohW!2P-2 zpFT(6`&zeuRQ(H~ZmHc}gg~E>x8kCuG%tuYErYY|np~~vhSZdd{zw~T39KhntW@}V zcEH5tdPeB_RD7Vn0*IQc&fX6?^@{Mc%q@{^F}o}|L*_I`2TXt}PD_o`@to|$Av+9W0SRQ^!a@KtYqG)SPWgagyUKCz6&n4T@YStEmJ zY0C-IlD>AV&3!GBn8HowR)MIh2kr^RAgF(3W{@v=CM+mjTAZW)5vitbO*fb%pjeRO z+bbFigM^U;Vq<<|HlS#Efa$DXo=Cus%9nqpH8kWJ;YrzH=34CrCdMH5f{6=?xX`6e zotRgOIJ@;w6O?8#ilb+O8*_7`i4Iyk&$;P~BS)<<(sbx)mkOO?vP50!Bb?N2!_NT0cQnL2kWPyTy{yiXBt7l`_V>J}b%XueH|>VPH;fmQ zT0+**o80SnaU?)5|HsO(pgbUmJYKfO8D+rUro9Z!6Uv-^F>_&h3uUC9hG-G`$D+BBZxP#Rb!j}Qo8NcW$SsyxlWR_(RzzJi-h@R=DX?;3&BLN^A6CTJK#*(0Xru*zcd7(gD8P_NxWQF0lCyYL0%J&P0nl%D)+9$~ zUn&|lMb<&w&t-`PfD9@gI{^^j#C+^?6$;xTEbM34Mi<~DvhVCfc zn0Z;2KY`!-urNUO8|*v@$b3wbvxUIrm5v_wU=$~^Zz00XntO$qV^b~vIHf=TL)SUQ zhyrX`v~SzCZQZtQ+qP}nwr$(CZQHhc?)!X`nWR!5m6O!(sk3YEm7d=Ahc^skS}7jE zA6vkPDB5g=mEk#x35FU>R%jxF&&0$=M-N-&4-{5hRN}baLY{6$J{!FPQ{a1&q|`aq z_*7s%zj3~NG67=*A_LJzMjQM|Of1nM(>B2SUfPHfIuD9cj`Ole)sV z*y<8rq{$rWbR1O%QzHZ(Bcjs9l-LpgbU`|Xm4q=%N{R4eSv6-LQB4BdkJD8y7-)eLbhC#$#rr^>yLJmvZ6;`KI!) zC`MYpWwt{PUCiglS_+tqyFsp`c+#*bjW)d}sX7mxR*r&S0gzFneP)rThoWG2u0x%e z{B3Srr)_=DS?4>A@Mp))PV%0<&A64={`#|0FL3uilT~!c|e@_b|QQ-%~+iwYv+xLW@dDM4D9s9bYGS*K`85zz( z)9h6@@l}&@ymhI~DrSbdr(D8O*V%=5J;Vh!WMg}y9n}sQmnvMKNe7~X`8BnAW_g=Qr170jZr|S zR~IPyUp8L&HNQ2M&dg%D=-Z&p*Ss=Q*}4UF{PIY}AS4Z%YL#GW}qeNDSUd4Hy(RZBY>H0^zq!)gf|2Ky(O3%-uFoN!Gn| zdc@$01_mm;hhb`*Va(UYqeqnYZ$c|3n`v3;NIi?sD3v^$-N)6#Eh;ygnQN{j`5QVO z-JOLrH8ELdG*UFz9JD)6JC?A53E5k>RDAbCBG*T*9Nyn7@=qEjMZsii?wY<`I4`ZD zbZTq9XS$tUI&PK;x2vu)D~Cs?Pj8oseK@{rV})#(lxn@#<1eBp#ag_(b+;i_ocr@R zyfszpoLE*VmDPGsnpzeVvYU-3pAV{0P`a&~Ay30BuR#C?ozZG6_n&q3%U6?&CzW}p zYM=+RC!Jq$t2oNb#yZ$56uc^pQDiVIr?g3ZEaP>%T_z7sztxF&{&{X>mwXhDRmFdA zRecz@$(-j}Gb(x^)KX!#&9u$oZN!yB&r}qfS^D*Y|Cw48=9HP`YHKSsm6e%)YHBMq z=a-g07}PKq^k~-sYBO5yeX&&gPfj_FF-I*p_diZ}?blhY?{MF3`Y~GYeA=x-RclOAjjLEG&MI1$A zKc5PR1d5Lf${QAscQ^VozdzqQ?i9_&T2B6tA9;l^jLz=v_<32mN;EnUv}GD^0OY{I(RxW&GecV= z`fpC#$8?qbBsmfT|D(F3Gq_gr)$wPV{Sh2}Tfz+g>6GYTb~< zkmiHJDBet97ym7d_|B1^5m_cGqA|eldDIHBDmA=hiV{;c6Lo0GVoIf?5qmcVmk98N zo>Q^>z=au-BN}3~B_d6z?YUsl>@xo8Nx7Q{5*-i+Vt0@hAV#f~NdnvC0Gfpb(dGA^ zkKxw_*9xxu{vyEf%z5^p~I(8;WhO*$*7hj}(NX}Zu zWK?$O1T2ODKgvkMsK1B4e{IzKHO}gDGDdS$>WSjC)a-MP0$F#c-Kf0S~VwmAt^8RkR$m6NvPsqey zYrqf5A*5v0xrVFvnVns)xNF$i|FS(-07vshKN>0r5W6oSgz6rKysPG5b%knmbyXL4 z6JlY6h0Yufsab8XT;+ zynPXJ)Sb$W?*YW~f*fLS;7lM$IRHaCP+Ss8i{iLltc0L)K}Q%~SgTN1ckwdI1DP>* zF>WnQHe468V)aF04vGSo8mUcX0DO7(!3r3)-WQR8>CFV7qK{TAYL{6nq{MK^u;(;t zI90(*58YX1i@Jw4r9SzdazJt)MD@*E+%PFiJqVcAFa=R ziiq3wNak%`5M_Ey?GTm3y+5P_=n*INyomqois#_?N5yehJUF2OL+mpNJM~)GY!gkX zuyoVX?2y4TrV88!yO6pA6GEdr~3y6uM(hA4SGfj>OAg15Ju+X5^}5xX8VzL zmcL@e3#N74lqK{^5Br*w)omnYtx_^{{)VJ(UiAz;sWAI7Gf_AG7s$TmB~cv;&>()T z1xMjI5rQ?2i{(SQTeM~zeUiV35x^>Im$=Y~e&%%6E$JR92w7U{jr4isrkG+>D43B! zpg^inHsq-+F#Z^4q0T0STbp)%=2ehJ2|PyD#bWXhP7w$CwtiJE3k%nim30p_;K-0a zHkL4+2&4>5j8N5v%xr9Zh;#l+xBxczUdML!dKCQXO7%@*e0~r!vwWqXYc)>U#WHAp zttd+PxG;cavVP<7j!P}d1S>`+gQD04h{XWfFbjH9KQ4_C($>e*VRSRgEurVEX?m`IGd3*{oi3Y#X+E$`t|c zDVo|qwK>qJ3_Fza9ms6D<6LB+EtN3%Th{$)Rr=Kw(DE-=|Gnxx{&$tpJ}!Y>W#Cb=kVv#(B+A3^N0|`8&H6t>0*~gT`07 z)VsLV3slBC0Gf7%vuf4Z)wD~cT<4v zLaXZ?!MjVI0(>BU)6!sRWS!bGRhYf$JcS#ntz0)mJwNt$HdfbB?a|0p1_^hdvXQo5 z*(SZWHQoDV^))5uvstpa(f*a?W;7ZqHu8QX+X^%j7hGdQVU+4yp$=q~lyo3sIRtp5 zF2{7}r=)MH7IoQ;UMG~-GzpoBqJstdGW=Nrsb#45D$k5Ck1u9@X4>Rt5E{q^7T!G)=Z2#Q=gy6ki0irH( zv00a`7_=<7u-mqyMc@%d>D$r=)PIXglf#54c~AfV^2GmNPvL0%AAn`GhNk0Y8;Z|s ztzH*gf|Yi|xf7@Jp;1Holu3uU3(LBBX^9F@d|pr*FgidJvFN$&Ejzw2HlKt^X`>JP zJ(0k!Hl-7dG;E!tX6K3k;8Hw zN^S?KrR8Pix>e)46jXe#c4_^XV3Ye8%}YJN;qRTt&oMd{~9b25IjD%5H~rY-036dpM@1^W<^< zJy=h(p_+5B^Wl;9G;<@Q$4}yI-6Na(Bqfi`jwa2KMOKkbxoxIJ_~7rt*c1y66)KcU zN~Ewu_7?OGe&RMJ=woA}YHzpq&;D~~>utWx8R#4Bc+!cu|95_){ZPLFy$zI=`A5MF zbs{%|9cj8~h@~f=q<*Br60k%M;h5dVQk5`&{c*?urwXpzNHATckAC{e)B`TF1`ybz zA)*)*fx3=H0AxPQu;q*cpd{*20RT4s)KE(CgpDU49iz_6vlmqqGqAdDjW|LOfLaHa zTtrn+;ibGhi^(<-fjY$vd%t_#AL2O@v|@!0u!@s!c2s;^u1A>Z>M}`!?>Q_Qs~}VY zYU8L+P%1eTUbZ}``+~4~f}pPZQeNW(j6np@bCSzUum6#W+ z2-O0^=3()T-AM%}aMl<(iU~pk>Wp-#;i)(XPzZxMGhVt4J5D3%-(gwE7-aV{0$8a4`dl9V*>%j#m&bg_k;F;&;zIf0 zBP^xuhWQ2nA(MO4y#4IS_9P$TnhLbbp!>?7h2_TivT*z-V&9LZYwJ78xGYlm*zlYD zM;aCe@RCmT(mi93;t1*x zndf^+7zY5zQUKZ7FQqxD7|nn`3Op3X(?fdN=Lp2-&+%o-(^AQz1of~h<^Fc(DvXft znTVq$JOW~aZr*EEMiVFv_S^bTnFOgVaPx;|7_h1R5g4Z`Q9Z-|h!C(+E!Oa`Jk1g} zR>=+P{zIYfUkvMn4B;%qr&zrnBNs$Zz<@3iTr@WIr;>3cEo=w_r%X!${w*;9?T-@m zDSMW#Ni%zaDl>CMrJVjeklCJY6N%F}f+=9JaFjf0^Fy6EjcO8NJIFIk=kDCJ$mKO! ziwUTk5kHTIqE7$4-R)l^T81b*ZvHhnaG5p;{XP;8$*7pYhvXOUxL6rSa{meZ1HP_! zcne$wvpk?3zyY@d9PFOb9l|Qtwbj+V68$$$+jaLQF=D?D&@^f>$GcoEt)de=0?L4E zu%-+M90{^w8VA^L@srBFJkd*f{}?tYQu?bHFq}B}8YqgtO~gyqh3p%Eg!I|fjCH$t zo?0fH3P@rsAADn2Da1zn3y3kQAUXCVe~-RXDxV-DeHiMp#lWdehX5-8z#EL|)0Lew zU)V*OghTer^)Kmm= zLAEb139hse1ZggS!sHK`VHnLYifIfO&5$3GrY0y=L_oR(?^py&is*v`)yw{G7w?29 z-VREz66j~nPs>Q~6t-IAau~76==*Bv+4sxK06kwzE(%adwm7sdR+iRA5<^`UBixK8 zF7T#kgwy;zUnTa>xb7B~L0qzW|O6CB{vCl!n+r)WtcagzE-Z zU7_qo1)4-8fFw5nYZ;cieFIKo25_&@)DeSNa}yRV2BlhAX=l{*3X}c1Mg@EtdoE88 zfPNRU<>_tVHt4SWDFjj2wL{5jj6q!o8fxEUftGX#+6{OX0^B-Z9N)wdrt{Z+v{ImA zH)!_+xHtBMalS>m&YBOa*w$Ri8l`jJE6iQm=D(etkVXadC*3 z0Qx&}NO*8F^~-eb!N z;BH1ohDa(KC=X`(+wz$uop7U;v55+Umq}W?`Nl3tjblo!^)doJjnD!z#Wi4OBt7(M ze_#JlYzZ&42IpChgPngaj@7UT1!tRtjR)ip%ricSxx*}X;k|rv`_)CsuZAI}d|)TH z&5NGL{ZjM-H=6V=R6C_ClT;{N$f7r;D4Hx@5z)fFG*AjkxA)d)w<1jFpZV*NZJT{b zxF1^n0o(ex_C365n?TASP_!TsJSqIWKnAeD=hgK%hNJ?s-fJ;@@5Gx+p0WcQHX#tI zPdEJxRu_WChH-aIWEDR@%-%+z z^PB{$+UEW|luZL~k}XWW&34Vo#ET7KCF{Afo{I0ZWKb|~3np{u(IVVTicv1)I~$m? z8T3!%R;%Gn!Xd>yu8$;UbST|>w;S4oj!IWo%D24;-LZ`18wqC978-Ahz6hnER}eu9 zdesf6Nh?F+PVNL(xMVD|)?B*EOR?mmXY5bfD-F@#t38O-p!Z zu|A|lWbO#!&$}3o?|KCM&j&atAJ^?rxV#&SId#3SCVO~nG99WGkS;KC_vICd+>_8z z)s6@iEtM^lXwPNq_DO8U6bb&O#fR!dXhdO9pcVw@PwP!=SpUvKPsg~{l{MU@=cY27 zN(NOK&>;1uTCUK8!Wyv^7&8ju#m>YEBh)i|0$kPTa}BhN$HB0!0gG6#A0#*+hNx8g z-(7!SKeL}TCfLPEZ;tg+w9oY)kKZ?s-y@IQo7eN(&(Tk^&r#Wzb0&~Q_HAA5+uE8J zRZoJ~ytYrRt>^<~T-?i8({_xQg-oZP_sKUrN2cqe?4Krp7rLjzqfAZbUl@AtW%U-t z%}Kh&bNHzrr=$}!Pt}owV>7ow-5Fb2vd7Y}oPoJ(Uj!*#{xiD?isMw>>xT}MU?i;3 zk=kO<^sR6I7dYFN-<(}OjXu)h#&N3As+L*%&By-zUHlR{XyrEjTzS?C6;8mo2oxei`>~MXI-&kg;pvM;}?R7$E-7S}E+Suc4tq zChpx*in|xv$JyQ3-v<(J|2P_`^6PEviE=2Ij&_}lsO*K*C*;-Mj-N-SVk|RGY}>ub z^Rm9oZ?_004g90RT|{k4d48;eUGiY;W^SEY^e**WU=t*_(;d4&mo1p2O`O zopK3d%Tg!k$b{h*u6Luq{0!Re6mSW}_gA*N;J`5eh+{>ncuIX({`fO|0O0t&Uk{UX zw0pXbbyUL3uCY+|h}N%d8_0;iy@$TKVgu1foog%C!#no3Fe8k&rsnKVPT5KR)-T!@SyC+*(hb_V+5^U&3??51o-RVUeb{D!kq5 zJ3OR+g?kW`$$Z+}d_0ApGE-CObas2<*vEu&PlR9JBI)wkbdQV`Tfc>=dfp5+SBh0z zpoDcIvf0KJyjP{_%m%NN%5%g1Q%WKB=^LtelUkh?zQ_W{W_TY0Tl|-x@ae~r|-aPP;&pW-ae`pJ4q-0fzId8Y(brSvbyf6(z_kl;t$-x>`>uWU-4E71o! z7YDPaxUK9nIP{H5Y2 zhqzS7m$$STDprq+x|^bSw*a>x)e+9S0DI0@)sg@&A4AF;rD#;x@un}~`#1iW-ecvM zcelL1!XXhE7U3fyS8NIH)W(gnn@&L5ejHPFrI?r*qL3izd4(D%#P`D8lSI)bK5lTB z$|l1>z|chV3b}gyGNIoY0lvSy%%A9&c3RyUnXZTOU%J2FnYO$>zoozLy*}PX%|B`a zhA3Zfh6KG{U(c1bzHhs~$9O-Rx3RsSZ}+@ENWt~&1bI9^AA@P5pVO%*EPAQ7GuO!H z*;k$CB_+M;C80Hth+1R&u#4{Msp1={$e!TMdKdkxA&56$kemq5dg4D;FUo4G84U8~C7~H9Eflg=q@Wop z!VM)1LyX_s9~g)q4}?k>PEN)rW2PZS9M23Gl)}W&fYfQ6hNQ9*dSo`XwJWNC>j)%K zgGedMdzU|Ed8qBmK8(s9WCO*nirUKIu!!V96BmKj@y~&tbmdOy>p?*ff#jyk5=$q1aV0~L3x4zQqdyqy6f>DI+5=a1iGQ%L5jJGQhbk%*Nq5AP zvPBgU2OqQu3UkN;GToQyAvPs3-?v?|(+O2$Yeuu6rh-QJW;QN$c6Qx56-mibo25&{ zADb%?%VW6C1ftej(N>mItLO+b=W-O^SVhTia$vD|OQjKN{UQU zL&VQ~WnF>X5P-VUuzX^kfTRZvD8!VtLS<(i)Tbrw}dB-B*h29 zxLbjZqb9M>Jp>A{r7CKQ?PyS$-Hp;wa7O zfo_w@C0PA3sYAOtr+&RY-ga%-LI3mZPJ#bY(yxHG?DXZbc~{rV~|Omd=Kc zt3tGCz}7v+U@l=}xu|NnseYM14MIEFvcR;{VlEzhmQg3%aLpLs3*uMpeg?ePy zhCv)n*@@Ql+Hw6n)qH(B4Y#w2*Hx+uA)ZJ)yn;coJvZS~zqnk-*j!ggfVD=?B8Wb< z%E0`mIl56YROVMY;)523rL@?>(W6y-p=o2D5Qi}H;3v(MBZWeNZvhmu0=FrO@ER(^ z5MM!+;#>v%@T+q%UbgKTC4L17#&v{hO9^?LP;vovr62eOLbF-aWeFuxXI8QZBY~$h zCg{GaGO|ly^9H__MYv$b#6X}^I+L|8U4sI$07o^Iu=O#4W%_|&Z7|NfNge>dhW|oh z5}_&Z4?*!lL4@ExymO+cDz$Rr0;Zg>c&GF@{l&}|HraVKJX-oVs@?lKpcrm3#h8)g z=$R`ZV!jL~CJXw8yx0|LFAPXj-$h?a<;0B1Khh~j!0F7Es zKBsNHwx|Zsq4*0iCNBbUHX~r_xcr{t>=f6zFyyvLec(k6b&=+`7$dME|KwuW)CX7- zDfAqUgj7<$ZLvWM)xX3MxtT}VJ2R#5GYc2+QGSBdf-YTbQ`EUG;ALzyBLM?R<(aNl zkF>WA*+`BG#DVqPbJ$mrP4igX@Qb4o$ocDRw-TObsjev0%Z>zu%EwDeg_}^ZU=&83 z;-?QIuHKI}b=TZ!h_F>*fzy|ADDDs?dUG*C4ym_Ny69>yuu2g0cW7n*xT~?Bh}sD* zvN=DFVq4zYPR9YDOaFcixXd0M#AD7Qua)=l6NT!Dn-6G0J7FqWMFd(w*8OTUD;460 zPg0fnlTo9$T!y#Rc|{iSZ&oUXX=uoYO#tq<)G%iG5ow}l2ui=X6=bC~(jWy=Huv>_ z&uS-x;Q)6w8Lyh#;y+ZGFBxDMrnb@HCIk)P3CEVkX1tO}Ur{@3uN+%4w-UawcnwjD z17Rqj_cS})uVtNBMD+X3V|h^~gklwxMAW#-{z&;-m@<=^Z>cC3`}0}^$aBV$yx@Uk0l&n8)m0z=*vM)5Y`_^WJaIq_Yp9nW z+d4OxU)KEN<@B1p@FUGII17A0jVC>muKRaWAfnR(D}D12W!6Y|ZcxIxF~U>}8=yFXqm?|s3n%6c6ttk`B3V8zwlb4BryKNSzOpWvLm6_|6Y0b<0Xr7PWGpJei zdA5T^YJX|9zGf#zNbYR6WXr@9MGTI=EL1Y;Ya5oT=N@m?1X;y8fYxn*3vo? zde-^u@16)u0I*w3qyP7d-Hp9(QYT41f6Q)k zt|b~+g=Q-8BTbR;*3q6R{+*}a1A(b@lM}g*BUFI&6ZEd?f~uJ=+}{338DPG$X|or^ zQmQE#L|1y;D-)H;tce+7O(E%-KkA|*oG%9YB^;$Qf?pwm zEkbkF{`vWv%Lm$X&1N-_O!jp>&@=pLq4TUNkj@;yCUH!7RwasZVMvkNT+)2RHku@J$(>*hAJz*IuKfFjO-NYqOGPffVlm_rZRMLQAxgRa z7#GASu4;&Kc*yD^QC#YVDHdkh4AMc@Q5o1Ym=VJLQyMqZ|BZdnUREp!|QSwQr97& zqwxi%xooXT1O63z^*4)22oyEd9*Y1Qiguqxup%@e@X7YReqZhVqD!VT*T3a%9C_$VVy7mJwWA z%FKV;J^eT;T$(6b?n0h=fq7sA=Jwm87T|=8U|k_BTW``PKZ@7pm`eWnfuInRk~}{N zQJZY?F~e6zWK(7`WM)tD_4L3-7qo)1gV}Pw!g(0U2i=tp1DFF%&ZC@Ktn?i zSg6HA*q^yt6_b+so`~8Sa=tN7XE*2wN7#pTt_6T2X+n}uTYZ(u(Ne5BQ_Tqhx#Y}x zo&5`4*A9S`1!;A%p^mLbY4Gh2HOskG^`>BebH9Vw-;D@xC=l2jVU*=eG|`T4V1G|# z6j{`Vf4uvhV(Yu59C8cMV>J^RoK-wY?o^yff$b}fl~*ruCV9?KLVmR0LN75 zX3er74Y277HhhUhbC+Q=msfSLU<(uQ8{v4{hadjY$6dR>y1=Sz2{Ef78Ass=m43n{ zv`O?=<|$q=;{@M9y~)2Ni73HXshBfuVorVlc)Z<4?@RHVGE{+yAJFydy4YCe;SyF)vOGMr72HID^uwXJS*C2CfK5%P-( z-Xg8L3kHtKJY1C*4$h8+@v6?jW7E^ql*B{&>h9qq1_15h;TVuke2JmS!$7@r!JYwU zUj?=~%AEG;F7=%Q5gz4Nz0&dm`A#xULZkSaP=J98#DmeaX%Y2f5l{^V`LB~mA9)u2 zWE6;JG_y%SDKWYQxsm|lIa1>CTu?bT$r)_)HC^UM1on?i0%UK-h%CF^8wNrUy zigW{1Vwqi1k*#gfm@oIKXxVNM)6zs#al2#@E{-$DrAPMO&t?j71_$Ry%0y`WmV$`} zXD2aY~k$28i3lH_lS~n{w7o#TU!{#kDK!?qG}1< z2ZzLrb$`}z(%bknp!R!G9`F)DkDHSN=JBMFYb})O*dT0B%P!HQ&{`?fkZqO9maaO4 znH+C$pglj`gzkL^DQnMvXvZcKXsObIE^Yv0u`bkLw=MLhI7OFc2o|n`Hw}fBaBM_; z1Rjv4YYr6L{XNu;PRv7n5MEu0UY$QXn={9=1-I=F)K+FF=4WW)?IS{Ngj4nFf)JWb z)4Wwg6sY#iDQb_`HldK8APR7QnXe;JZ6HwhiEn<6TJvTNK#BKAi5R&;n!%) zXoX^D*}q99gA5J%uv@j09d+hJcEvx>C{uHcH?Gm)Rb?^{NzHU3gk^RP4f!Wb*O=m+ ze!E=?U7!tHDAls|t*W-fMhX4TTWoDZJ7vs%wH6WnR-JGkV@(+ua;?-pz(NOf)KXpd zoPPl;Eyteg3G&`TNBL;35`>4SQLPlOojYxG&l|}Q)nB#8#OO2xM_cZ4w}V55&EI%g z!~K`-1B9R<a*ProE~44C>Q>*O-9g1ZxOviSla@wkyjPcvpS3dB+5)k7_5qKJg6&c^H@RNA4Mp1Rq3%T*qB*Uq=t{> zqiMq;Yxy(4He)YxTCgMnpOtMn#Z*VY$kRneTlz`ERbCl=()qINBbM|qwK?OXt2+lSb`*EJo;)6+z7b}20=XWpih z$1-3XLAR~QI)yr%PUfcdF4X)1p*eOY%2S~jJ}M^H{bd@O>FC+*ks%J!C9V(E#0?K1nasKM-AaGsyF#; zecNrzNkUD819z!U$svPQMTfU%4gH6`C@xR0#X-?(W8D{@jywSvlgsN6aBklt<$)_Gg^xQTjm;tsTrukJ>hwW0-CxxRM)c!FhDXI3+-Sn z0MvOrj$fj*TQJ_9(IJaaNJw@WmflX{Y;+8&xh$#I`ku$ed|w`kENh36-F_vsX(WVSz7?y0N2;)5 zjOExYp2)FJ$cWM93DkSfYGKod5p+=u1tx9FMgcb%8Ti@Bc_ry=R5%@`wrc$|+SkE> zXrANwJq!%Q^Pjf+tyS}@OeG%j+himPj=(E@lUi46jf*6WUb2g1o^WyFnfP!XiA+fj zL6N}8!1KfIa8`3+TkG+$C*+zjE});Nn!_O^G{uZhG1>v+R=_J^`#ohG)HO&C>b~0V zdIm1pMPqcz*bBs&RS;yp0OyUT(!=8J5r4k#h56*kNLB791VY_3WK zKA&^i$JG&Cml`PRMQ{0Grf2#HKBr9IcT&7@K>L$QIO(N$I>n{09BHa7$ z4J#@u8xtCu>X+i`%f;l7TypO5*Fo(Q=K9%e$I$%j&cT~aYjGDiIzb&jAHueiOxnjxxN znHT@$eOgxwp{57h(#s*aVq-(|#00$$=i)tl7}%EaLmiPD>QCkE-2lK|6Yw=V+on)J z{@M$*Y@%e-OPHAJAOCPqOFZgjzm7_r#8PxMQd|4dn~g}U5AhXIg!Z0}av!V5EA|!S z${YlqbgIxq(-HU+(D2DKZS~jjdfa`7cf%vHIg`eV4SJBz*!-N*3XIP@=H~vY2bVnY z9p#Bmb}6m7_LoZqq&``%tsY11K#qoYaf`sQuVn1O;-n^<>I5W9C9 z=Q^|7bZ^Ml2sWRgcNePm?u}(bI>p=F<3z?9Gb9wW#zev`Rtw1&8R8=I5M>h+Fk`OgC2AFQWyN8 zi>rZA=&OM#d&tLSa)L}Wj0(HE1O#n~YYWP}7akBMCY7WwAbe@?n6jE*F`({@ z?m~ZOUL)GZOk2gaMaiJF6K{)`1iuMx7EY-ojQ*M=B{y>PGhFYH*=43#Kf;3_Qbh%> zL@fWc-Vq1?xoP4MYUQ}bF@~R z>bv%J(m>G+Ug>AVx6^VlQZRm7WnAWV>xk97pOH-c6xd$-6prODH)l4cX(iPIo3-z# zS(CeTY>4<;M8I(_8rc2ZCP}oXkle?r-9YaS1P@Y!qpk%`*7}_Rcir__TQ}r)B1gm{ zsFSN#F3L1k^Eb_X9xOHf3chpN3mfkE17~XqhHT`hJ!vYBKO|nZyr#%{al)Q>-}LOV z4~p9-x%uCV185?wetWG2ps~j!g4ffi?cVGaMXm2PD$v%Xy)&!>wy9Y4*JPnR#&=*n zlIgfFQL8vlDRwJytylhkeks4JZEXKxIfaGbRkbl_gI!)(DO6Zwyf@EU&k3t+&z!BT zSq*_CMO)Ltx*QxQyt5Ojm^2wc)_KAWyb&7X&i zDhWv`*(ioB?y4p<{k=n=%4W>a&8K^b zW$*NJz-8e<{E^We+ajGxn_t@BmACIvpo_a>vk@H~`dgxPo2za&{lsVgYISz0Y7PRG z99MS`BgK=UbiRz{Cv!rsI>I-vQxfeECBMeZqnxw4??cX>Lb@Z@8E(GMOrnV?dqudr9$w_5+%@hlu56Q1@loZV?FC7wlNh9Of%1qL>I%|lJ5~c77&Kdbx@oh2^u&WStpRv!wgL>KFgXf zc-X{J#ALG<+L*>6W|8910JaqR8OzkgQnlWUmlwsmaoc5JoO(^7UDZsxTt0tlS%{zn zs4&Atge_-@NYtxoAAy@u$(iGKZ(RUMVu=Bxu%_)X>gmAdgIs4`bYn2@ik?5;>5znS zf(lIYp?s`Kh3@7|Z|P$kC_sC>{7iWVjj#uz27TaLH=*l?u3mhZl>Mn7Q8*hxe=huk zg==cSxQvkynnTreXzaxV*+?`fii0R1PHZ+o$hA#utR@zq3~PtHGo8YoH3#tzhdZUu z>5QE|dq0LPTE;y(k76jN?Z{@d2^S0pqm~%rNgwr>y3#Z>=a)u7Gq90TyLw0w!TQ_C zI&z17n$+XCIuewucIW+4(_<-V$P31)IQH7y*rht|fp>k)f>eb-Ka>(aABrk zX(G2RM9?gXl4?!f{8z>p8uWw{kFD}IGaa1G0D1Kf7N*FLgI&9FM{hTy_iOCQS%6cm zUd5`K<5ou+-oT+}<++E=_mtc~>581Z<5nmQpA`CXA+1SU)`l;akq5^3^Y!u8FPEBQ zRhC(e_#bclwfNA+5J*4$%`QDqxO zjq}9plk0lW?IV~Z;7wG3Y1kd;ZhM||OpV+vybtFInWH|*?d_@tz86MK_ZV|-P~p@? z8QGOk%PG}+H)a+n?G0Jj*W@k$`K#7*B*s1m1?J}=K#6^8mivd-<1kr_yp1pwf^TOe zAupr4$^9`vCUKGK19qSO;d7`oKRijTr?YtY(8*Db_%Z8B9Ki7&hOlW}GU=6F=u40% z@9G=3O90G&Q0SKC(MdHlK6B|vL5HYAG%8v{EHwOFhY zeP{?`yM>AwuF2(kD&>@2r7G+%r!72C0{hWGy+HlWrGmp#wS{k@sN&w$H5yx+NNw#1 z-|xO=2*}T|RXk>UfPqV+l{Z?hZF{=EO0uG;55<7Ub(pH7m9v}6Q*u>Dm6J`A=&jm} zln21EETUD@Wv%ryGp$jnj_f>OX2}_icaNhJu{28og)T&=&e0Yo3_vH5h_bwqfZuhk zwgyi^jE3=t1rHJZ;Wm3J#W4diAaz6s4Q~^)<|Hz_FmIBy&L5v=-XNvVT6L7&J2N22 zeX=bdq;jkBmZw#g1h8Z3uRF_2k-) z*~(4rkPkc~CZ4v>|C1sYQN!1wb!MCEki!VaU)k&oHxqoqJPoY;;~(ETqC~l z%S-~IipN7~GKA8KK72C*nq)qN0I55aIs6>KU#8||XM5+b!s^bP+1b_Ji+j%gtoh<5 zDZHXj-QV}AP)hshJ%UIwywTN&hK0V*)LfhJOv>(Q9S&2t+OgQeV#9Pg=N$yGRKMz?not~ho6v9MuImE zp)M9|J55Ek<09MsMoOg9DJmS4gk92)dM1Iob=04c%>AvHf4vi1#wl= zB5FiLw%}_?REw`rGV?4vOMW4DQ0Q#1#NYJ6)6`4~>mPG?m^|+do!?I@EmNUr!6!{~ z($Xb1l4O2#3bgf_ubceve>IaWb2Q!2vsb^m@m!6|1**$aQ3fn_Umg#C0_^&DvOu|` zFVZR~)Z}%(YfzsyRkfaOkF^$b5$s-QxQ8U1ng#lV7XcXV+sSPSxgZ_-gR^ z!fCnRi}1`=Torq}D-6i)M|?$!b`N~;O-#b1VQBTy_uV@HyEM(=>kMa)G=QgbwvhV? zfOi`b?~$e;+9_>P-dcZgaf6dIY^GTAoJeV_OjW1W-0~yLAI=6+gza`JG00Sc8Xx27 z1bZ3@>=zvpkrzyK7N6o-!QpY&GvH2~x4VtT+2a`G&NpR&HWyNhhE*#3>K64S3&J2V zbiPA_O7>808{@h2{1U7TV#RnHa;)ZSH1q>@TtW1xHd(U#wXEgLDQvsz)(P${Y9a%x z+7vibEfZeLlrlm)-0Gm}2K?rUA|5}=>e1?kmTDCAU;H=D(W7 zMHm?5FPLmeYC{jh6jd3s2ZNgJh(w$ZtZ?I<+efvRy4oO{uF~gxJfk9Z`5{g|9Wi@L zK}Zs5u2{MQ^Uy~DZ8l;8RnT$uX*R!~hY}6*tvDRh$``44{ro_4mV?+>vJ~NtoULNS zN+D>SFuIqM%n$ z4L+Qum-w|($If49VDDB1~i8KasR^^}T;LCuU>dS(28_UnH~N`|yQ{Dkv+QH|wiq(XXS!hkjx z463T!eXH!7qYtVA`6Z58MwML=YyleTI8J@l#fc`x{)Dj)OQj6yr20=pY(hlvlnQlt zDkUOR+oWKwjoAP$mf3*wG8`^PnLK(HKo0jui9B-|Kn~M1Ap)$OK}op~#J21jSkFAa z*2E0ZCNe3I`0++^m&=gAFOea{S0+dK10c^FRUwBlLAS4A22}l=`gv7O6JYA#^*msu7jLe9Mv-X;w_KVOH}$fw$9W%VsiL+`Tex%qEZ?Q{Fes^`X=<{ulM>a?F9 z{DgMmsB1va8`AFl-*61SY zzx{XQ55zx{Fz5Y*sisV{4d{Pn@c4n)6UGS(kUseX@qaG;fjGC8@jsskM>BQ&12J<* zd((d*PWyovOCf#o2V!m*)LvF}59(?PmqJ`tPmQ#Pdw~xjW`xGE9W|O4HJc8^w0|C| zB_JD!iqSzLJ8~c}ye&e4E?Q85(r>u6k@K5pnkz+`maDcmD}`It;_X`EPOUBNTAN$d zIsfONaB?%WyhrnPLxk>s3|xEy^h@4~e>*?kc)=<8%JuyHElYrw{jq(P^sUK8q@=Z2 z+E?>54sfL?JeFIulc(gCOVftU5o$$3$3AbSxYY&#`?Bw@Y(wI}@&8%>Gb&#a^ka>%W8lDG^GJHK z$o28bT%i7tQYhyg&@Zc)g%5QeR<9jBbhOWwH3QVANhYvS8U3~3guFIwuI=Al0%3B< z+8a%}brF1BI{YVgQBGfxJ%>9r-7QL6S~UPv#zCK4^JG5{ymE%H3(b z;!f(@Fqt8K{o(Yj*{&GLw?*4Vp|#VU_ytZ=a$**}-Fn&$X$j&MR+s^$kBZnoM_RwC zaDM~Z{zc>OSItx%B$8kAr#+|esUjMae{GYd#dUmfGMIS-Pu_)+iwZAErK`OK)AIh1 zRI|wNy!otlvd{NWQHuvTu;C(yN@PJtiW4lq?u zdt60d9VEPk=e%h_aIx__pdz0v!{kS?6w!4h7siodN=@-J4$G^N-Ok49SdE}1OMtbt827562jxL`P? z*mQ4VVR%Own6-db2cN-2K!aLJTjPMGLYq!#>$b|w8%&lZhNpmT{B^}pJJ1*hC!5b< z;^@g8uaM;^#%Il6R{ZAP0h~HN_ z>z%XlcqoRh=9`PJKIN3TW8{GA<}{n8%;7uzo$It!^>qE_l+TKhpb?w>!uKZW^$y5D;RwUD~gkZglVtRlSBniKEzU!QDtNv)W;-E8fu~BSOG*WB0TJ&jXEX@>P!0 zC*lFS*6l%t=WVg4ol^HIfh$klrWfSxUtNh3zpOY$U$0xnj&3)`)|Ua4?MmCQzU%HF zC$HN++uIP*8jXmLNzbr8lkUN(pYyC?CqQTAcQpMa-b`DSoBhvg_D?ueN6|L}xrkpC z3u*a|&Fye!nzxKW8y`EHWPS7=LH5J(Iv?n5ENdPlSyp7ln#fkiv)lE6mv zL1N0RI(32+O~X{d?6m;5-|sj0>HdA_^WwC<(+h|HNN{_dW)b^xp(30M4;{tV%$4JD zeCPiC_hN$%2ZS~W>DVqu@@%C9DzeK3Y%6wQ)9y-}<%s=3mP)i`^AHLyXX2*9)_hlB z_mqF9QeR@>y$e({Z+0y9|F~?k%w%%xjYI|r{#uYz_2JxLgki|wj55OM z?a=6<#P?(9>xs~!!&4q{MhrB{m7VV~8iq-)4CdI_DNp02_d$hk?Xs(BPRp zW%x|FxPobB`uBqBZY)161Vg(Na*2J+d^4IPC&W3;G1opuvMKITbBMN^lQDtjJ8=ri z#R(&+NcYEHt^&+chvn~^jm*hefL0*rvITxc4D4g>@NwQUb3cZiFbY3stQ_|JXC@F^ z4R&eRbF>qhL$pbiNYuLXyjM>W$B#ULP zJ2k?yd$meeq+Y#JQHq|gf0(}Ws7Dj|v(AC0ZSJ6tFuopSq>A2e-6*{npSL0Zd{)1E zPFCE{q}jP$B*Cp`DUjs{_vg^ci)6tC0CnDc+9Zt`+d2~XS(a{9nCz%udOeAgeoYDdejWY>{h!^1qU38?pO8R6@K``VKga*2LuPAX<7j96(_mPv zeyO%8_Urqm*N}KE1x$=*VWFm#$_8rPF;z%TLY~hZ&5Sn22HipoOR}G~z|ZTJ$N17D zPm2Hndv&$n7ljUMp0Bd{(gR(qaeDxA+@*8silzM^)O{+1|d(&Ds8mE^~(} z{}zrpSt$_?AV&4}+KSk#-QD}QDV-CBa`FtSyT`Y1mSIMImcDitttS42G?K+mKcF~> zQnGz4Aup(3=B~(uoq6>3JexknLKv-{kY0!zOprx!=oQI+8k}LFij3%#BUuH7Z;nhM z_dz5VeqKcie4H!%ouRw)m>9P}I?;dbQd`8DPeeV#QG-;e{86Vru#3=RXdNjBx_?-o zF(aNcS7wfewWXN=U0{;{4nk0-VWQJ=RYPkpHrae~=#{QbB$Vnf1d0;IU7{O3K$8&& zLNgjg>l&RnRV#`|Ui|$%9>yX3mO9sdMfM1SoqJBP8Yq^bnc z^VAAT#i2d`k9{_pC$5OfBp;P_Yc#fcJYt!&3`rHO<@$4mGvB)47b^ppu&PD1 zdNQTj?^7|e(gvSlqoCf9=wAlZ6{L&bDqy7vJJ};VZ2(xo2u-kfq`&rb;6107e;WNu zmK~eZ#H;d5;R9yM6`|*uqbOm+p)NrTVABJ&3H($L>Ax2_GtbRRNEP7|8iWK(q^tet zBIu8cQ_LE}``J2`q-)Tu0LV$-x5U_CZ!~*Z33&$kMIR_8(KG-+&25+VsIm>&ydtcW z`yk37Frv>?X|IwEhT29ml5SZwtJ*GDMCKQ|^85wVq)JQpv+%qb*HV%|jiZCw>0}hE zs%;{NJYN)?RHF74J>}x8gc2=qxZzXQ`2&5S=vbEIV1YW#A|MADE1H`{p8s!ACPE7P zTAN9EFWWQ|u)Mo^Na?=Z^{A)FMTHt$$uFMNtUh16+^|v8`-d^w*j_;DQ1($}2%oaF zN7paTW9@zTUqeGt(1yM_`C{a#*=q7Njv6kZ}XH}Oi8b(@3 zuBz#HS}NHOyY~Y+NCor8ZQRyO?p#}NaCup!N&x4dFtSXMBx1|k+mvciY%=yc%zHV6 zkHOA&hhA6J^V9WnAt}HLnJRla#{W9)eKIr3wKw<*{;l8eSI@_cLK?>Yv&)!-@{g zPmNheTEA+n)-TUTe(k4KQBdihl$B~S#h9QR`wY1X=jNMtBG+gpO*T}gczamh7!fo& zXl&+XTyJrVfGf3ZL1M^TtsjnW*M!JXyDQ8P{v(TKhs95Yh6ZCGTCjk-SN&ly>PF+G z#M=J0Q6BBtkAXa~mq|8-0%p-m49P>9_dP>zUWXJH8{;gim2gkiz zc9Pmmx~D?=RmS5(8PC^<%WsXMV|P1~H0%K##cN9j8q-Gec1t}k8Nm8=UBE!lar)JC zvg#+;FiO*}cUYZ^FLlW)WJWv!T#%zdh{MZp-tn>incJfJT{iXFzz?dGjm-lPhV62<){T66$651O$RBF!Lu`7r zMfbkW@|8=r2@v%7+AiNbAV3{AL>H7mS`~9$@k~8rg<3J1@s&IsepB#w^=;DcxZ1aA z*{1w+UES6>Ov=Teit{3VhVH1hu`xQSJ=ZeUI+O12-@L=0F`|UP3Irg_O4=v&Pu_xz zv=@!#xhN(N5Ov`@iXo393vzWSyRrsyM}JOFz_b9pGL6kckmx`ZNI$d=wvNpJoU=$O?Ft4xVak$wI88E+d+`Zv*efqYJ`~2zNp2aEms;_V3*jAT6w@6pMz3~Cz@sHzOeVVwh+F@9pngf>~?ezMR z^ozPSV&c!`j>}k0mwL0WAI_pI_9?$&)jgyFKZ{Ie0F&JZ*i)Osy23;>_|V}HT~^4 zPbgYUAkKZ!lp3jI=08-|3tLRn4g@*?J8~u~CAEWTNw1H%qARwREr(2C@#1~^sjfus zx^1f#RL$|AwO|N6s+^i~N`(H#IsdI59mn0ufdUcwf>R`f42j%O6){Vuy8wUEB;>HE z)3$gT4syz_Ck4a&Q@=(}Egp&NctMbq)co)*fnN7JqYMwc85TO(b)jBp@7sNA zEkiBDAk8UHX(ppCc{Mxf0(G+XRZ-xYAytYP!C7>T_8guwgNp8re{)x{A?TNcS_deA z;YuB{TT4;232c(%sUO>I&Y#vC6295ca7Az}guCpQlJs{EN=8k%sbZ6667;?|vmg~V zxB?l5Afs)wCKiQB1bf9DRZrRN+Q##gws0yVPFZrwgn0v=Q=mVi{{}Op9DvHU_f(VN zeRNd6H3z-i6kmvwgp5u{1jt{%szpCrlf5=BsWjoYI%51p@DbNE-K*rnqfCwIivBg< zH994g%ba*&^@wghj4tW@>q~!2**Fr+sKt^;tP=}r=xlPVt7L=L7zk8g$P~>DDZiGg z<8Tv5Myh7Q6jjEYGCC>f6`uiFx@@$4R8fKEKQ~)FQL-y(e%fvvEm(J&%*BF;sFb1D z(}h9_xe~rUui+=L9uiBOv33Wp30`bv=ImxAn2f$sVicKK(-YOj#A&=_enuJYBN}t>vy;Ls%oMar zjB()3@7q|p$SLGx^eYw$g_|VB+;&2jUj+wStBnCDBV9_f+KL#Ud*zGrlcF`oTtFm z=V8aBK;}XF5J3wg(mj0kaK>-hJk^Z`-ia;LUsNYlcmjd9Ii@;LoKiS{>eQ|46xJV# zDDUR=V0JC@Aa&2JuJFL_2V8A?vv8X;zVR;uKZH@Yp{O&pBzC{hW+(|88_Xv8n zy<>kjJw$?Z0XQIK%t=8@XuTgf$|(=#Ub*;T6RhK%fBh7S<}A&gkQn7$?^m@^y5nv_ zXdcas`!Tky7U+6dJw(;upNigptauCv82ewuvqtVA0aGok;VnyG-1~eyveRzmT(XNS zPN|tJ?gL?5ylkZq1Q|7w)AjxJufRQui{N#yT<+XWSC9L1`s&prpkDPxc}@0NZw7BZ zi-@^727jS#?>wK*O>uYk#Qir06?qG92;gPgr);G;O(Fyesa(W|w{l+~tZSK;rZL(J zJWJ?|uZBlg{(rVuZfIT+PY|_DeX)7ZUT2R>OE7k<_?UksmQhit&HrdmQ$-9qCx)Tk;PlKWOr0=i%BX~O zY5dSXIzOErn|$6-v8I|o!VLWi+|%7j53{@89vAIa2|9ZMA_|G*S$2qOJ26Z@KRY?n z)V8Io2f`~EBUyS)19+P$GoZ`EkD52sBaLS{6>0sBP`~n~?7Jz&m%uk z{S@%i%R=+_?Nk#f_bvHo zRMFh4p69k`cLhiMLErYxhs4VU%Xdq}CbbSlNGM;C+`8hH(7ugo?ii%ta~!eGq3w~= zOf6)1;Tt~{@Xc`5`UN$YUOt5q1o4NdrKWAkd$8>l;2`w*#2!7P?FuC_1(S7-9jdTk z{3~pd>_cf2DGp+L6{SqYps8-21=8W>$Mh$1?CCj2*T}`t_>OQbcIi@jI~<6vpNkR3sTNWbV#Fj2B17VbiHkLv~B&9UmX`_9I*O zmR|Of2nG>_$+ZD50h(3yA&;Yx1YHUebIcc?XfW(qA;LXzBh1kVEujrd)WvQlW!>)1 zWr(sg&?&NMOW)!m<#g?{jRu^D>}7kO6LrWwoU3; zs+gZ~Eb9Zo=v+||rAPd1R9~Kj%;2rboO*D>eyZMRQscnZiyWUE@s89oG#{L?qUy0m zk6WfBX|L8x>clR1oudpvx{L5L_wcYIta{&cAjuAe%7R)c=wsckk7<3|KwT`^CgB=o zC1kJooiE#_b9EY(m>md#=sJG07KC@N=5l|%hg<0u^ch@(|0Zj{B-cs@}t=8h{^ut0ZiW_2hYlQQ-|4x@!TNH>a;~(tF^;o?DAWJZ@!L z|0UM{((xhqhe1p*wB{=2d1ptDR$3gcm8@x%aBbRb>x>>c2&^YJQTMuAd9o{bj|e@Z zy<+S;r;jI5bgOlmlQuiE)iIj>y?qK%INj$%T|V{7Fmgnn41j`AP60kSf5n=_Hk%o_ zyWznKkKCeXy@VtT-{e4GRkeqyY#bh;Q_pcEH{|J4LQmnrvJSn(@AX64|Kg|)ZE1IA zPj=z;4$q!?u-1$vF^O3bgl4KlI%Cu9$4Fjhw@yMdStN|GR@GY|Gotz+{XqN3j)7%x z*&z7i1L-`{HdlD4cs1V%+}>BcRyH+x(T%f4%o&MA>`6*dDV(WzYO!NuXSZrCO3h@- z($bW=_vB3Ti0IzlK;tcp7(_rjGB^7EuDo9uRK15Ocb9RD+dr1Ox9)l|5*&$U8u!zW zd|ZP1aQ=4CA$-F%UP2rvZcIK5h|ETc~;yO z=t+jrMZ86)xMqrm-p0{<(>O9g5+nSf$H==oaAN{mtC;r@o_AX7=U-35WBN9e<^pG4 zyqNv3D>I(tK`N1X?L_aF50Bdu(Y*u6QG9#g8{)UPF#W{MsOJ>Oo*WxfhIK>e7v57E z;C*L@q)j)YN1`t_C~H7A5B?-b8n0{a$g9DjWQ}9SfY7zy|4bl%%<<|`K!JdEAb^0d z|F@!(i;bayh2zgwAERt#x5$Y0nX6|2lbA#&DNYvut2$`}h=~b)=|0|E>WI9xy1$-e zb6r#a!_}I?C*=MvtoX6}F40`IqZh#$UqReX)k$3Z z>Qzm4{D_p_$~NnA-4=GnRz<5CuDJS@1w%SNfNSXjdhVeihB?X1Dsq^+_+WUnaUsjY zU?xJ<)Ol(5j(`%jW(C=iBmFyJUN=1pe;R8IyfF&dwX@mSw6S%rLs7E#Rt8C$Fa%4K z$Pry>9&$%5QwHk^whfpZE{HFo{n*Z2DjcK37*Tctw?L2;(iTg|7>PIx1F+sCfJp|E zY!azMT8I(`)w(i>n#NWLXWb~fBiJ==JwQe$SX;qfWl;V;X^vszW~Jm}L~A>ekWk=1 zFu@p3KyWyc-o&_A%woBOLmaUvZiM3VJBklabku>J%Q`r7nFp1*lSnpk_ayxvyO$$U zQ~V57%i!&eUS}PekWq5pz3C+Vu09hp_^*Fd*jn*}ee!|goLkb3RyJ~eMpT!b{}xok zI!j;!=PvP?Sr@9hF|LeSZ(i=NM;89cQ|Vf2v z_j`@RgkKSLU4!lJwh!fd`=c-L|6ECJq45T2KdZ3=WZ+8LTS+S&drq-Hf6rwvY| zubJApgefd2P7Bi;)7Z|f7|;^;A3i9PjZz+#0hqfXE%iFQ_TtxHEyZW1CdjCC*suCv@3hlVDM8=v$WCPYz;a$f+fw+xi*s_$uHOT^~Jcxzpk@h^Y z_sRA;z?Z-}RqC8MWG+kUZu8H}%eT#!m6yA7<{^n#f;lqq$w)l4#Bxx~p1zyU9UTP` z1O7wVOY=6ZMQ#$45Tfr{Ir z$At9Kq!Gcy15z*y419chG+v%OoTXKNdqJbTD}DSuiHc%)*bpHkgBie%y@i!PJG~i| zsnrHZoHc1YiePpLEg@x|QNeA#(Hw^J90D0DS7afQ(?D@zKX$!>KlMT&xH*-55+iY- zBoPfaTwk5JM`(JOQ)1OGEMj1OU3x2=3C)r6I;yB(sYX(bySL+G%7OmFaC!QZ zi)7gp-#4eN0+t3ZiuiEZfdbe6AEMS#b`~{TJ(ENc9 zK4+<%ZaXnqkEV3+)Wu(;kNLWun zm}eDi?-2K56FC`oqviDyfI$f=w+q7GflhV{M&0moXk) ztfJN54oFsmKYEx1s1uq9yo8_0ipbuPdkUKc@pK3@qBeT2GK=$)>pI;n(EAFFEXH+= z8G?xjLDTK7O~S*G7Mo7D`k>GP#}D9rSh$@Da8~vQ=aRNFUvKZCspJO6Am0u^p|<+s zTziMoxa>Kh2ga<~Cz7Fhe!+bXpGjc5y2Ol}h^6&oOl6&!apumHI=vqADfX+x@k@s4 zx2$80dIV<-^ecNje(oHGu8)3l1ys)t*LsD2_xHWm2zYW3X^7_|f6qa8OnW#Dk=^PR zFYJg@8FIeh?V`RwpLv)HKiTT6vcrqJ-ufst#k3F!rXFV1%C0x-oU(OUE*mM%vRVV5 zXZBd{)tuL29YrNizM*&z0c*vTCP}t+0n5kk*=y^8xDecMh3)X!?AWMnxNbi477|gg z@yxcR<=ASm=`1FZw31_z{!ZOhSvl2MW#nEU6XS4-X{t2nT<^SeCeu;!Xtd}~iJ|e8 z)$)(!+V5OW^zV#iXlBVNvGkUH-SpnAQ?Yujc;1h--QI9qNxeM2Z-LheL~nOlbx0|n zkejqHjxla2ns1>RwZ5Erm^Q^PSNA{%u6Y{Dxz}Zne44Q_m6b0)Nxv02WIuLmO%Xqr zLe-HM^i&*;r74g_aOP<>VE4*B3Zk6$&s}GN*L(iRJw0NY#Ps5VsYXm(wWFE*e2?`^ zqge?8uX_yNr7m%^ks(Q_6c&!dbwHl7*lo5UF$*AEb*YsO7~^ zF4Q|X08>)7bBf9<>`h7Bg%3oO!EzX5%8n&<%?cK0qKm{y0T5b$Q9{S_7iJSqK}i`Q z$u3}0fSr6L571c&Cj<8U0hE(`M!y!43lkI~`@&?OqZlE7h(3C18V3f!gTts&F_qI4 zb1a(+_WNQ3;Lp7BX2ckTg?il^sHj=Dr6DbXZ{O3cMs|q}oH+C!kpJ^m5(q`vU4Q`u z1gr%Fg!{iFWp;*^CPvN{uK#f*>gMtC+hkAtepjkX7~XVNvbDy8HSNg6RE5TyEQXUcN8MWOxH+~{fex9%-k5mnAdziNXJI=O7?wwU*zKQc0@saa2pE3Q*NTFJ&2m%3iiYlMWL(OnpXa z&GbPFhX`u9*5ifpG}SX&IwZkKq%j~BlZJ?6g@unm^&@$_hMo9{ z;&74T(bR#%QCposdFo8Mi4d2hK&&L_!bF^tJyQut(?RSrU_}CJW~-?rvm>7MjsmhF zW^OF*I0**hBgknnwKtOQ{2_5CkQ?jRG~2C?xM|M}k%5;~R* zlV)QzZ;Y<4=4h>tl1QG8aU1e6Elh@%5Z?Llg7SJAs`>yWXvURLBA|D6g)C8ft;Mw64PHy3rB?Crqu`pSZqm7`$J46%c_*MG9UYtrZp%KslMs%~PbHiEan!#Xf*rfY z$l@R5?<8albufr#-~hrrGI7KVVvD|a0hOjc22KFG-DrgAKA-2Kzn#_*o84exb^teF z)K|k1BWm^SwW6c#cA+2)#ck9+P13n?+5D?}=W}P~ z+D~(<_x9P-b@q1kS~f(ul9$M|Zq9qw=1G&+%`3nU2lk&{Z>ibm`-t2(67kE9$+yDc z3ik_0Vin?ruk%jr;U6}^D_)@&UnHLkjvIzM~2`EY5xl#)z&wgPaCI z%}^!~rRN4v<}D~kVli9Vs#6OcPmCf#F|wcJFU5_ABPi+_!@T*^v!EwG8^D(hOi)p8S^KEf*c2DplJR55;1fqQB9Qs z`U{|bN5b`h6=Q)3hA)o_^NW%CjHM9S%Nw3%dTyvy>RYRSf!zuGjnGN(xDVf`t6HJA z&2tjHwo&>c+nGBJ-Q8)6L5|zv0PrJ)oMgmTI#k94NqWShV7m=Mu42d-q1b_m%uoVL zwdu76N8MhXlLPGFzM-X>B_gbS1ysKV^#KMEj^gJ4 z*>JO5F!^ox@7Rq8gdp0dan@+}D$4ZFS*@*Wc=~l8CstNZcl&RxYyv&k*_8`8)$yl> zVE|yiVpTxC|5(vR%F>e{U=wDsZ27915z3)V5a+-f5l;TvQ)RPj>_EcGRd0>hdf1fI z@sGL3Bsa+QR;#}Q2n`rrPQc$lZoD!&q?mR~JC7^S=2*TK&1!jVpK8G}+aux5IwNM* zGIr~bXkR8x(saq~^@b;j{h7E|xGYr5R_p#LG)1Bh&$-^)>7E-ql9X1&J-_a37xlyQ z?l_mlkCnlJ*GSl~k}bnjCDp|uPWw7lF$c0No=VU^F&t220f`AF(M_Y({J^>(DdgcN zmxl;`Qdr%dpHb$1lA%lQS>}2H-)4o~nNLg@((TA=%|9K<8sYp|9pIY2$mdrFtO+{4 zQq;*In!~1JTgT+C?;1KaYA}mtbz^a1vv$tBo=^o@=7iPeY+faF%H-KD^;=&HOKk6> z6lA|`Z}>N)f3cj2W6M*QEx#L6>YjWY`-%(3D_6lNWy*l8Qoil3-ne1=^uPXf^o0&) zBb$fBPJx6UX;ZVZAs9g8*9Nzm1hL_X3AQx`Yn-Bn93#_sCc-P!u_{6ms?q8f24{{O zpQtWjW0lM}%KbV9o8tm0k56jp8=Mlcsyc6Y!??Rhb=l;)R^)OX~NS^xw34h7%-387%l#_UynJ8AD2_UqEc z0t+YQ$37&T9MoWmpoq`<2G2aOgN?Nriq#-TD*SQZMV69%`b{xv#ArAGl@k~XE`*C@ z1u{Cv_0-+o!N<$h$#;wIvSM-SLq6HVYq33~j|>d}jiH9oig7#Bfirw!pf~EB#>YOq zF0$xJs2iz;1Zw?pUYNAfp~Df6IU=09J>uOi)F@0Ocs6+)+0ZcJkU~aCaGaT>i{nQjTMtOg)YDtc_YM3K-X37FWf=!8uU2I>&5J&4Gvtep}GE zdrf>4z%5D$GOMI-Hl@wT(~v;G&%J4sNY>`N6a7Scvpf}uKHg)&fq=}18o#?a>%kO;J%|pf2ZI9-}A&m)RYSitemfi za{ZU1XJtkvgsCtADhYA(F&!T^TZ{r5GMsYR-q>oS!S1Rb>($LFOJfCT><72a=5Ret z!ohirQO3aX$qEBs;=h8!@Y#CWQ!le7QeSAjwDniJtYWc#tbZizCx`P%ijFILfW7IJ zzeG~9Ns`_sP7`}YC@GTw^spWQ{`ijwVX}Yd5KU51D0cm&EIz*qf48{SIx9aLhlC%; zbmjHXz>l}j&;qk06iTI!!~YYomrtSDi`v)+KE)$F136wC=x397YvOHo3DW_TOf{u$ zXx1OHS?@LG)GQ@3Q!HtoOjs$zxpS2xHN&(^ZSIprSMzpi&N9IyRA-wpMn_iigdUp29_V71q)1k4Vr4rzk-h`~11P0oUtf`E;K8sKvXIjP7A z`qDgU`_r)L2DgcjbLkxvdk|NvMe@R<``rf;-B)cUUP>05W%^8AXZX)J>_}h6cFOA% zL%Z(T^u?*R8zClD@_UkRD;&YR#m;B+OCg%H`^`o;i(AqhdLn3dq9*PxdV|(E^%^@A zC{5_kha_@Vv|$V_iT>$73kWv_nMH+i6~oixqyqOH7_o>SX=pUbE6s?h&xn&;47^*P zOk|iN?oSFtgE9_uXNOPn2!;D~A_*RD#%Fy#@0=jb=C4@Xfff$_+qcSdhUd(AMPQjS zeZt~N)D)O3p-8hLII^a&ixWck`ke3%f#7$~R|B$B_(ivEF$=Tcl*y#QBGs#HFVRjJ zsv=e$OuZHk@; zkfO1^uFa*BA__ptY2bu2EKxx7`hdO0pI!9E1w~!rnciJ6&xE;D4IujdZ88%kt$qM* zj+)LR3nLU^7J+;k0J4X}{5u1-bOr}o(GY^J4_!=No**N`PWP0`GNT=Dr5F+KhG}Cs zN|?vr$1S*n>;T#6{2C0c{Jto;&auvToKTI1(wW!lSkCc2$Mo>`_^3&pTW0+&n%VAl z?xbexY2Ygnp`&-`NRASWUw$Fg1)T-a&%IQg+}3`>)t)Ln4$%x8VYPO4T!a z170muO^;{P^Y&oBdV4)z7bo%s8x6Z+NTXNGQ9k^UfHd+aynDp~HPjylP~`S83SC^Q zJ{@STwEA#`(#fLMc{yz5ee|%h_PG$GSb_l%QJ0eJ^|=f<+V$1wd>f21xvev}!j{2n z>YUs;BVI4Ks*HrFUjx}{^z8a=9a}|(cj!x<}6K7~M-? zt@C976>Vfk57JG@{rpEE%1*qy;xdbe2(`EB=GL?Rm2c6L& zYf4mMQ+qDekXyFLDN2_GQW;bv@njkFz7#k>Vspi>v|GYaUg9}*u_Z7|@|8dIvdok$ z{>by`)MyG5aY%`g)cv~PEt4ZX3X*KqDg|u^iN&u3&hF^XU>?rz+Igo&U+IWCnC{-D z{QU_9p&pz?w6=qOr4@eQ9y;LU^r-o87JcAZ@gtE)L6*?$duF;IpHlHAU$|%{GkKnd zIDHwW7kdM4#QWM6`l?gnpnYULn8v3R6d&9%5E}hw5b@-E_-O5{X^dsTM#ftK5l)a? zuEXlFB(nDN(e~TNK|Un6RnGo8b&A;mQz;c<0JH(GjZEl?Fg2RdUiZYMgg4&H); znUdj4`-0|;S^*F+N7LH^x(A^Pso=$$7p@`n$GVb^T@z`krjf3B|DoB$=}C2RsCzzfyq{#xm#(9kGC3C!dKH8S$=@IPMr+mw~_|Lzk+J zh0?7TJRuA9pHMoD`7MaUx-=TI{g6leo9zr@f8Czjtb)f_`NXOY(M2*bZK|RN(L=<$ zE{!b}F!MrHQ%!#q99J63OTbtR6XVMo#>HW8$&bOZ0i&Cip%I-uqcu@(kHIRUr2V;S4*n9H@TNUV zg33=q0FaNw7=fs)hZ)0;bwd4618N)$;Nu}fs`;@(b*Y(Nvv8?%RW&|xg%$Zruv!DP zGZ^=^2bbC5!_)$KALy+JcM}nM{z-wqp)XzxkXyhiy7&A{688w!bzM?OzvI;*}+H5lv8bVc?(*^B5xMT6AJJ4#Y;tT%jdX1 zKlKXVN{_q4%fjmK-&YK-zPelAPjQ2MZ~vw)es1^j|GK|lD+kHMoj`S5^icxa*4r=i!Mx^n;4? zLg(e$2&4qQ3WZKe6EmBq3ea=B9WWKC@A5K?(~gnU&%`MaH}RM3()K-XbS*#)a{3t) z#kh}GfBgQ%6iyyc?&}oDFJusJ;Qk#R)dhV>QAcyEUVOTlz@uhTCpPVMCcb26|BdG& z_!Sxm^B^Ftq6R00%L%u35diHoENB%vyVNdu1As_t*SWU@_LCi4nr2$&RwPaHJedSn z;QxxZ5Fp7PybD4+l*ag{3@#4Zqrs$L;V4cKuqERdLK=>`lPyV>kKm%x%8W-|n|Ml) z;!iZn#Kr=_mo$PU@T8W~WZfJk;iGO1PWYTX8ts8!s)wa(pk7bwH$QM6s9vEh?whAR zr|{n4g?o@W5;Bl}kKi?(hZr(@f;ZQI5hn;n}S+qP}9 zW81cEcaod)>(=?|RPFtDSM91*Yt6al7|)m>S9HT5SA5f;CUlf`(APA*_Xol7#XXJ+ zk$4}9e0u$x5J}Q0z8`ENMcVAW6EL)9jHdf(G!)R|&xF~;lj^3i5-4Oz70~q%Inme$ z1nRbOwGlbZ7OSJD^RVa`UHwzFb)?bG@)-f!B+7pV--}jY5L8+~{_92@U?#sdLxmo2 zgvnGq^5=ltTs$pPLa`IxRX5Mq*9P=@CUK7yyef*)F-+t-`%n|dwWskODTVla!>lj3 zA;HP<$5{YRn*4Tg5fRIhu2$d5(jnqsQqDW1bXmG!d^f~NEig}6&%dw5=AxHXsmC9O zw2u0aCKsckWL7>Odod1RCFQV2n6Df?wDj;lKBqD{9CmRaEYkOOuPe*e|2MO=wtxP5yaK-cRZT$aCf6zO z2i9ni#vpG1u*_ytTn6Zu;*v_w3UTPVmo;`4TB?z*NQ7{?@|iiI)}B+bCaW$o>xf#` zuC~2F*IxM2opAtT*$xsdu(?rRW_&-{)-W=@{EHf#%ZF=LynG5C7eGd^`K`_wiovc_a?Jjn0U<_okoWk? zXY~e>&rLn~>_w;oM+@ry?D@j}@2;?DMxS#MAM9V-clj-l(HENJ*TA_V@zly2#d7;Z zX;*&)c4OlzQE_W^P}!;N{TnefHn{=AgpL@=whBt{QrxRVLPp8^8{%_BVbqzpeZS&Xv(bdsq5MpKi zD0Vx3(t-b%uFlTR_~%wX`u{6-xh=@J8V*fXLpq$cSS@D_b(LAS_+?e$v`cIXn5p6x zFKOj{KVp(>MiLrig-q_<8D_SeNSCf$d4A>mbNySSQ%P88O|3>KYIKA!RcegEuf>-H z9sU?lU>%dg3B?#Ipfi6m!NeRM+2JfB4CQ$#6d z?&tmK{&_@Bxn)9|uQNyF>F1PI@%GBHi*X`@VWjqhpzWS3iS|K=(uynXY%l?r|CO<1~V{Joa~! ziT5>$jgUUtL^;%B`))XWm|1`ucOukS!x&h!;-O$SEjq=uGq;p@k}x2?Q{o)1y7jl% zJlUZ&bhRiEN(3@fA>;*aoWI+_W_^ILnc3f2ZjAv(Fy)VsB7^D2f;0%(pkn+nOeIw< zlRU7zC{|M}HhLKkhSgfywS@DHF6}wR0{g8UhHtoRU@V1E8*y`pUMc&ZNwA@{_2l*C zB5w?;Fpmi873r}99ZQWDVbghwc1$epFk3Q}R9QCdLlkz*WRPrf3TeGDki^?0Hf9E1 z&zTZKnyvjHW3i=&J&Z*|LXVB{a1Qt<-Y3_~iU~UU0ARTV zciI^aG?{}#RCQoH!*+OW?(Ty7v^`{83-8zYw51#67`jqorUH3Vve1<>OJ!p{<1!+G z^E;5~7$3|hO!7NdDoUgK8*h2IowXDUwTBEfCqpwfQh9+(SVMSV@6|D=ZI;qldLRb{ z_j9BqH(WGW9R0VKbnsKJA?9P&R^N(pVEw)DCXA=?Cd^Aj-;G#F`q>M5;+#cS-^lH` zL}4palouoXo*(CWmw#@mnNOF8m)rigJ;H&03n=cc!2zziDtE13AN0SzkBw2rZk}v_ z(LtV74E{e5IW!qBqi2dv(~`Nz8U7h1tn8=nh$-t?KfdtTY}>?ik&ve3Rcs(x4$qZ9 z@<{1d&gbE2p}_ZS@sPzW zHZf;oaT*$w4&F0}_C_%vY9^a0HXw7&*dIk8mpVFT@qamLwMxE zf0|d^mgVmT*l_ar*4^e^Wacu&@ZE+8Ea*o*_OACWy;J?U^nyD-Hy*D!)?4WQE$nVz<`00oCq-^r8gUb0)`{!vw>&-l4(qGOy+s@g-g!imNB9e}_--ik zmZfQo{V76RT8?B~kl_nE+sP)Np*}@SQDiq(JDuw;x~&JjiFPSJ^t}+&+WI)68Ky)g zY1sQ6$eftIzf7p|s70dmfFt;2MCfDh&*V~V$4=IZ^MdZNW?x$T@I-377HBM4*$Wz6G!DvJ& zN1gi`ejH1#2ALuaSujUkBQjm=9WR29%D2R+d?zuHVCelu+N7`kdUt?Ht9JzVYEJ{n z5(mDM5`%i`IO43QCIm2LT+W_hoT-MBE)lE{w|KgyUI7hJ!}LN-13Lp6#49covR?pK z?gy>{XMQ*F>BXNjk`q&`TdQ;-oGM|y9rDGk1x)0Cl<^BXtM5Yd=YGa8nu!L~M=$^o zIw1@r#AaWtN*@|1lnqXVG4Y4*52X;YOm!BZa4TJ?ILK&fnYeQgpPB?qJ+n(VpEr~c zTOV%u6O=xeC>@ANLMi}-^Ex}Ehxfq-IvvOYUw^8!Uc$a+GoNKk`0)V7?VTw@)W&uH z+KcBD5BC8A)DCCJ9^|VO5t*UI|1T?0Z7}Wm^;|by8F5mb&%EqFM$+};MXLq&6tMub zXAhkDb@!qr*S62&>q(g(Gd9BG8h2v`WNz&E$&rOBKXx+1u0FqqviL2JG|J~;sS4>@ zmFm%nH?=A=%>i0|crPud^ce4 z>BKL5tStXi2qE19Upi?FRW+dw)IjoZfI}+erFPYryRxDpNy1id=wy7nfOUm~Z0Kji z@-7~|MSSXc+=csh*&qz;6GMe{3DHA^nHS}}j`iOgY9)hTshPyo^*E1NI=e|#5^>Pn zQ4Nx<`hCx4p8i@3JI!j^0Z(PyYD%+!U+wYJVlqK{gkg!Rb!nL_Kr&A_Dl(LH)L>3H zMLR4-_uF%{?z0tXVXNB2-&ApoommT0^Vk)HZ?MezhLKT`;J+IgwtLay4U||k_e52B z!Hg1!O`{Jh*88d@0bEeDV;D}PB*H(Aqz`hDS>a`INO`)n((*y^J~bgroW`g+fG>qz zcsOF^erg;mxR<8RkUa+WXEb`Kn5m{z1FJ-bi-QbRE4v_6 z-5}#&(0MYp@ObKSvKu>Re_Hs8N96>fOwm)?l!EG8xDZLh=BsTyj{LY}v_`cSH52lk zG(j2n(zfe#gsdMk{UkOFJpeCR1fPj*2MsDlW0P|o7Nc{ESF6+(7KnER4lP?|q*e(-X48&b$?O1T15T3GUpZMl3$equZ+F1^*{pS;~2FVJCg?;Wu?KpRbui5b+OWI zuj0h?OagtsH%5f~Zj)`PF~Drrc8$^9eTRPeN8=Ox#;9y_lP)as4=@-VzI!_JamA3k zK7f*$s{_M@5bKs-xx4Zv5-x;6X)1Ih;{5t|rp+bbzv^#F-W5rc?s2@o|vwu)vitE(0mm~=z~ zkqs_%#h0lLUI(r1(m-@|t;`sY>iU>T+*S-%retm%$pwYlGz;`xn)axM!QUr|TC9~| zoYxl80elYAp~u@5>_lu>8IywFn}u9dqtd05C3KH+(@9s)>Y3mvC0k!WY-*;y+PI^Y zHBgN;2dz^V-dU^-(zBu_l6D61iOJTqAI(iV9em7+x?3`xEN@8w`n0Zj&kGyGIs*}h zgBIIiAn0Ny78xpqW||hU=`)!q;{Wtal%CNpLdc&)hOxiBmI=TKThAcS6=TYf7@X1K zM`7%63C;a}B!%vyo)Y(*k)T-c_`0Dro=)b*H-MSK!8UoZCxqj5&TAmV$jdrS+#GSM z|0bY4sfR%NN7St;S!8?S^Xm)l{-i}3YbIB|Ci`ifNV9g$?fF9Cgy6r|>_Ug7tMgC4 zll$XP{9grvy`iI_lcD8*f|GsJmHuNU@_nh@&4OHuAuR=FH_*$vglk;% zPnm8}j8K)Lc3r;sX?hyyR#KRWN+`Ft+(tTY;Llt;@x&GGnns1RQJ^&g*$w;HICxQ# zZ+Fad2W#8_uD-6%GWNc%Uo4q?Or2oQz84u< zyX`ZuKwfWW2ODk#JI0Fvgvh;6OV%DQ!G!N zkF}2^UW@>Dkea<65gX>|vZCqz@f~XVg4c@z{p;_}!JA{=&ow|()x;Y+dzQt{=Z#)D zfivXl=8k)fWf32LC(Wn(K9tE@4%t=a&E3+;jOWgHdF_et@mJI2x^=fJ!n)cP&PW4r z2H>Z?KJq>a$A#Pgog4y1Iaq4vV44NWzpSgRNT|deHM|}w7L;M!+L~H0;X*5yD(HO( z+jN4Y8Q3FqOg84-rz=rPaQYYx?UfK(Tqs=Zxneox9sA8{3zl?0>pzY-jhMmBA#&eNP zPkf=JcOW1_oCXRiio~=_&+#ytu3wl!6gT)NrxqiAS{&o`f7jk5;>pMXBv}F<8#5@- z!{aznsNVw6J{gsmseTfF+N#UdGsloKh(HulekCDlnx~-4c!}7p@I#c#Y3P)4mdf`A zE3(kXY0C&MS<5xaZ5WcD#jcx9qX>@w*?pruIwcN5C9Bd=pDhMk!d06CQx3l1zE?lJ zj0g1Ak|mi`8p;a*tfUwucmzRXqKk}Fw;hE>bZO7tK?vH1)yz^yq$xzPEt`=C)tjw0 zEn{kKPob~T-{s44b$;777eN$wl@}DZY@7!Z^!ob-p<6w?wcXK)|frdF&CDA0iaDqJ5u8n zs}LogZ_W)IwRX%{2j0JfFz;f47L3D@zbxA^4~~ja_7s}LHBRA&<;J;i=E+S-gXS;` z7nmR(iAf<*iQu=0Ae@5+(ES;q$4UWPj`eH$tekb7$}o+iJmM-*Mp2%aPuVj_Um^U+ zha6FVjqTU`KZq#kqT=WlIJWIBkXxc|D6@0Gd-w+fQC^Pw{+ZsAZ`)<}s#Syh+UZ|s zItdzo95h?B9I-H8c+$Uf5`@dreX9kH_W^DRazU|sju6yq+X3#cl)XlR&&9S&VS82C zNr`CLkl!QE$$@b2*?p66XZ36|WpzieZh$+APHaIf9$Mo6s#YR-9A9cw^!fl~NWi?8 z9;M@Qme~|K%7)!!|KEUj9V$PjD--Vdh@_W7QVTEVJU z$#bFNGhHt=l+K!~qIRe3s4L%WE~OMA=oCy9(g_KZCFk#F0FXCI5hc&p{$v(;y{Kco z<|X|zFl~KzcX#maL6sSd(OZsH0?xi?MW8Oi-?he`!CboM+qAwynhmioxD2Gow$)Z( z7)Y$smPW_uwM~M&z*h>-xBVfv?w_FcFHkx4(Oi^AB4Xe-Foup%60oap*RP3>1(xJO zCb(=(e5}0O8Thw{12;$s3Ui677pl8@L*d*F#v(f7z15mnorY$YqxH!CSH;KIq{&e@ z93c!mb$2bYoj3x0kD*7NdJkJj#IuU zxVr2_oF13~eMR@S3w!*bt$&$SDJ?u#1t?wq)6?D<@BWnrO!?J*xj6iHxK$`~_PkFA zrDB_JJ+~B-;)EfKBOM2Wb&zw42)Lys;e2nB&O5K7`(`%>IFM}BB}|WocyU*NL4{GS zOykxVWDX$Ra8t-0Yr*VffmW69x=L>Q{R}9^f5&-%119#VJIrozku%hpUp)*iNIl#j zXR@Ux89iQcJ-ku$=~;yeV3_rMz;ul4?YU8EL1ajr;5)6}3sd^0!+I9e1S5rV*e0Hl z*F`}o8|FbNy{W@Wztu}T^-EEXW!>OLFxYt9pzf}kN(;kXh8V;F06<4LA3`y6!a{pB zEGB*(%*QB{P5?a?rf0Jbz1kK#2KA}Ij%(P`!)5Q1)K2k?8^cJLWoX2#9X>FQslAtE z^%R=Fy|tWmWGKSyNx$<=okAB z4YHmMCQoX{DeSin7WaJsl!zEF8FTIA_8b9v8vLwk4H|MMn!ox$cKZ!apl^?Fm(Mm> zCIrN#?9rN!0Rs*jY}j*e&MhF<;Mjk3I1N7J`62o}deGAFIWAzEoH;2a94QgPA}yIN zL09U5TX}hw3aXcK2~_F?ViCp2qWtPa*~lch262F`wMiJ3dD&&(gUGA`(`X_6gC>-6 z9c)&dT>C6Vdnyb;bD$WeDHNt^1}-uv{y=jD9Z-qr)nOR4OBlvxn!RstH3>`L3+N+e zy&`iM=mo}UhTnj*D+Tm-d8PQpObRoie>4ZRX3B&>dbk{! zxfi*O@X^iD=Ehdgf3Pk68N8V!h2!d3$OoMp>}~$WD6<7EmEJ3lk=g&Z)fA3DzXXnf z)fN?=p=D3T(oUxSMit@8z$gKkXhXTacu-qObj&V?d>0(p~iqlpnY9uC%hkwY*?{zZbMC! z*{pZeh`aEl(Z}|05imOEL-!CUP#(AXBP?R0!!1QG{Oh3F{@i|z&WKHfaGzqiBxMZ$ z%TJs^9~&v)NX?w!L6;%T?Tlxz%+pbqJ^-6&scujSB9Z;*GRM%f#_0y}I?zV@7SDsw z?`SCNP*!YIq-Tih6_O2bQd9t_soW6;iE+n!ADhiB76jEEM1NDa;ziIpU1Rzc{rCK} zbWjd*=&SISJGvD{v}$#*yyiJc8na^=6epeVU5c3mD^rA>X>^G7hVPi3D88~hoKN2| zIZ*m&J~Sy5Yy}WO%Bx>^{Y;nhjpFP55mOXKroe3IC9pu9 z`6=k*c}>=CF})1tE%;4d5$^*v>j2k1!6_jE5}3{DT^9J@aV-Z*OuFExbfIBnW5 z>3;wBDiKO83wMoQu8Sc|O`U(U785nmZ)(>TBAK-c1uo++m z&!8>iPCD3BxbBWj7A%bY9j>H*#{}X_xhhPZvl`l`X_?QRnKpy-Rbk0(p4udKQmec$ z)BQ`yx}`8H+=4jMJjZl2m)WSl)PI0L-i8@a(n9v<3zFaBkl@c1s3E!*^Dy&9@=+7e z1XmOcFS^P)cpo|@buOR@s;$oYIDT>daC1$`e`a7KJD#|*_v&OXw%5@*RzAj^MQ|GX z5?KPNvB>rsl;I{wYw~o+c5*ACJJ~yxur0FR zu(ZDQgJ`@EQem9WBKth28cu*qR29H!W0c9ljSSQSBU3bp4=nvX$`^+pCf3`&nKcXC z^atHtifq|;*%DV#A!C%G+k)9el5A7c(*#`PnARSbQCRrELAKPkPbzNcx<*f!G`#-_ zhI<(%U)Ez>uic>t_pktt5G@Q?Vei8`*m}(YyK2 zaErmG`(SVLoNKFRdz-^(zsFr~UGd<0bYL2B2pf?cyCmhCu#&V(Yhxq7N(2hoI~$os z9@eUkhskB2ane~XC6Cb|ihw;G>8Ep--@Pl4v;bL*s7r4DkyEvawJLoQOqE=VW|nZ~ z!)nfnhC`x`UScoHss1Uqyx4CME*UrB!pgC4yAqK#ZhV}Sz1WckiOEWOt5(}koY$mI zJs=i<`XU(Xo}orukD0@!jAs;u(HnpV>rC4+$@6J`*SV&WqG?kPuxnBS*NOf5_$qpZshVL@TC@0K>rT})v z_FDhO(NY?UvvIHBQLPtKDr4?|^MKkYi*Cu5s$i?m#^wZEoW~FvxEHVQzHR}nF(^O# zudf{mii3qhj+*qv;G>}vv+UvddA5K0-vQmi8h;eItDa@E(cjm|5tFYH#6vQV~rR zmb+SrXLPzW4YW7#k7oq=tF3_#+&&jhQYAir7NP8Js_CpNPsK=TNxmCiXCU(>)f?PLsNM~y zHQ3fi-EglnUQJqcq6Et7(&Yr&NbAn16=VHja6XoVZsXg}50!0Tgzbr`fep$r0UGKc zpz?g3k2VK1=t7!BAD*y2K&}baF)YY{o0>KZ;@r^>QH+efCMz?JqO9E<4%|(>AuSMn znaoa0hol9Dd)frKT~Cn`?cVtpHj|}8tA;n_?x;RgL;w9FjiiKt(a&i`4|*zhl3!)5&AMr-JRYcBF7?_EW*KeICgm~LdMQvwbD+LvlL+P{2o4A zdC5N?tGPuk@v!g<T}HGdsl_*j!75cOtrfMSYgciJ?r;p|1Nn-8>3Hs-AGQ zg$XGUdo75uWM1zpLL;#)%tWBVQ!kHI%3(D=#~O{C((`WVXxGgX{gp9UK)QwV2FWL@ zZcbj+OB}3>h`YgV#Bz8$j3Z$E9KWPg(NxR*Cevegf0D-6yNv1Aeb-pqn(8uC%;Q!` z)js5$vDQ@6Q#B058yIP-8Gkt#WLYSG_O{dX#@hu-J3CMgm=`Y1W8^yA4oGoY$_qKP z-O7Z(bx{@vWH>dEa>IIc?peQLG~6F5s<4eFs;V<>!8b?VYLzwj_uRWoh{0|FOdM$FenRhaqG)pc zW@>2&^aI}0k^6nCwbc;}&EL}FSbtSrI{o$aJ<{L7E>?2FiCRG8J4vZu#uvfqpgLz# zLX0)Bca4SAuw-H3ELM@!f#$5TLsL9j+|!X4cz4O#ycW&VRTwj2xaRXHSwFHoJmlit zs4r*>%?ey+3Nf%HSQ(UFP+Ou1eEZD=tx}|*#c5YCyq^1tk-~eB)AL8lZc7P-n(p4B zQzZ=yx|{OZ3!X3hmVX}W>?h9*^y0AygCkzN>g-aa!&FAjdV4|UTXRXbwvb0U=Xkzu z1*|L?)mQ_KT#+9r?bi?vuO0<9^#)fUB&*k*)+~gaubTf6UDDd496Q8^%)$Dp0^AJK zUBwJ~^4}{0f8F!0BMR`oI-V1oqlt;DBKB%d!wFT{R?B3lktD%7g7zPLHV^3MEfGld zDtL7{vxf}E;Kq4X*IxHq&FO%Q%)IW2mz{Mf&_F?R6HA;=%)#k?S*p{aT$|cj z-F{hd9&44FYbB|a51h)}Y962X0NpP8H;QoCOd5;UTnnaS+|<)v@Vkg!Mzp?*Wbv4u zrUm5`tW7fO!3s`6WyK0kPJju^4J&W=noKyry8J)|h@)=dadfamojbU?X>Je=pY27I zG>h?+N;(bsCDw(QKD%~RXK}4WPK@3{h3x&bS`i{c-u;pf&jm3DD}#)OpQmZ?&?=hx z55L&dII>ou!Q5JwK%l`2z6f)^-DBy{d8!iNky@Mg_Kl6_5)H^Kh>m!6fwQnU%;M}k zsGIUaoNc~|@?An53y+{VosWT)%Y=n{K~&AS*npb}GtX}E=AL@%l`z(XLZNMDW`BL}h%zsRV#=1<7lhxXv~XEW}-K0xCLbRt2VA)N3RBDjmWK+61MrDp1VzauhqZ z-ca8DGtR{cw%Pss#d|mOOJn6sYFT@Hp$3X(d;~FqBafn*X`_)Zy0k8sV1u%z8G7%f z%lc;z--op-V5|Uc4F|!h3=VP~ys`ny1tR@His$D&U=f`VF+YKs7qgL5+pjzazubJQ z{yEBFC_14%J&WtcOJay`)a~v}+3e)-1U*2f>rUY(+vdfstk?wZS*Rl=ROh{E$RN99 zXm#6HYV>KZBR4wZ?Xd(n7V=VEiWaX0O9Wp^3+{(7&kx%ID=YUPdeL2i3SHl|Wjxf6 zsVDj*e{}?IkRdc4Pd3?T*7fnYI3y*+3P}UogQ>?j7nYR}7e(hb^e(>)#aDkh$kz#I zm?P>zqAb-ND3&vie_lTnHxK<{AjlK%l8Z5^jrpE$U83K~?!M{a;TGKa_ZfQABOxJ> z+hvGYyZ;m2T?~+b=Nl&5lH@iTF-bgoHLuc#+@7AB$oBiY`nI=$g4=e$(bDNJ8Na7R zLba~1B?~65NK`mk7+g>IXmBgpgoyZJ=b3&2-nZpMh-Tzd{&)4Us@-d`FnW#MDR4k> zB}*R~`>M@A&|1H6RM6XSzLjVelP!Xoum+P2n!2fbZ)m%$x;}6qO`j*=>od9caXx!< zgGcKu#M(Ge&sJ7ZO^ez!c*$uS|ET%B`t zuFB@AKEfUDH2!`w*JGXZ@{_p}SrM+M9BqMwq4hWXzD>vVPWcQAqbpbSsbnmr^4a%s zRJ5?u%WG$|qtor>!s2iEdUAvwG?r53z7fZ~v>3w9-?z7+$`uyzMDuX!<_?0-$9JX^ z0k3z9@3#w|0wOirtfSdum(%6usKW(A@;qgLk79B8QyDd42TpgH8cM5LBhIIm?Xp_pwfM||@RfM&202Y?=Nz!h zrH95O8mi}7ved?qt$K3jobQ6cP)}8Z`Wd&p>MSv17wtz**YRPqHAg`awUr?A%zf3C zJB-`ih#mDVd}m@TB}c)}Z;rvFjEP#)k9`|ClqR+~zb(4jg!* z4@gufFn`2?YO&-ux6g3&~N7l@?Y3VOZ^-?XJ(1uwp^2lT7HOJ zBrUcR*Wx8R2-~dL>t|zvToiwnC3K-arILQSuNG~Wb?!&jimQxy%iruunde%2_EgZK zfx<}$^eDF9s%&{m^QS}q_{Je3$EWF&y%yLW#w8KZJ?H(kbr@$<~NvJe9T}zJ_$+Bd%XD+daLz z2F5Qqi?^flSw9oMeBtN8A6+xlV=;37G=r|q^Yw=!QFoVVizWegaQ`OhuIS*6J|5jx zHZQjl%sA0aX~m*s*jl300g4a;#StADspaTU8@jurAK3}0SjzGYGRh|uDs=b$C$m`- z6KmdbZA{KSp7%N40=2*=(ckNhTJn@mAt^4y=oHEw5DNYTbgKlii!n@{z#$b|iCFIB zl(IaP_Pi=w-s9srA<{o?GNDFZkv1ZSED~FRRc5D7ELCY&3cse3aWJ)^A<)g~`F8tM z7&;U4)%D%l)8+p3%64c6IJO;f(mm3xjMLm@0GC6tqbz`dJ+?|uMzvJ^2TTIRXldOl z5g=UadMEpo~*oT_%Q zH;VoXyfK-E)uNs%b?io(&467+-XBCA*IKO+NaoyV{Gh#H4gwG+l&(}!wXO$7DNg4| zlbq&)fv^DU`VPt=6TxtWO%`tY!H9wfnkb4D9%8~5b ztT=TwoT%~faWWIny5K?v;xR4@7$nXw)@D(TbHun%n9<{XZ;ZSl;N$1Xr}Zi!j}x%8 zV1KlLxO6HRS_Ql1;vzn!SE(jcfU%iz)DVwG{NP0}okn7-_@{ss**Zhe?eV_(`AC0xOCbZPpH68QYaH@&*eFz#-EG?L}ADutxKttYm9Dnd9Cborf#}Tl`~C zAix}c!^60#39_QW$!oziM_wQA{qJ#!KWfnLIuvv_Ba-Wv`J{9m1s-H+&D3oeMW4#B zN@QHuLp0@+%VwCk{1R2Teh&TlAgt(?35A38U@dreWmmYOo71g`dVwBOygTl{v#SMj zyBvN2&xh93z%T#SxGMJ|#%&q}oy~N`A)9Y@S+J0-t2Dt=VfA}XoS?SNf;e%?wYnkD zcTM$WRuxZFO8j1+J<*RTK9UI;GwbDK&Mr}Dy<`~?qT&2Bj$+`ZlyJ>XpQ=y!We5q| z&a=9u$5i5FnzF5$W%T+po2{2+q2!Fa0m)3Xgu2PHAKLpwh$lF|)b$KtB4d3|&NiLU zA+m9mN2iKqP=wW0jg@s(nA)P<}3!T0_4E8*$j0yxz1O?I%2)%rA!C@ z6_)Y9h=-e9X@}qMojk3qi+{D_X4SMX&f=P!pc59I?a<8M=VA;jcXj%@<$Qg}4L(1r zpeFgwJ29n{`@C`3qQyq5fzvT-ZU2BUahg1qRaB3Rplc9&OI})w!ZX=SE{wvEfPQRf zCoc>OGRw&J$OYZzkP~!?H@ZiNBWwve7(Li$tpymp0 ziaZ2|3sX3C{h_&!P2*i*?wT3xn2z6u|Uwpxo${gV_a z<7CV4q9XT<*EZ2J6EsHgD8CXuZqvO4%Unv3kG4D*89m=aqu1BeRW)zKA7F;KA?3Dv zatRTB-S51h^@P)jfz|8epxV1yqvvl0Uy4Cv+`M;o|%l^Cwkf zj!iZ7JLh4-ecKp;zc|Zzwtk1aOT~`T_u{IQXg3Kz{AN;SOEPQl#`R`qJtzkf<7Y>P z6<38w#HmIXr?T&;))d!~v|9ra)fH9|t2Le%uCqC47eTu7ot+tu-m7Ddh};mI0o7#O zBQIvs0ef9t0|Cz9M&gSo{xeP8OjU;|Ur{}a3xBQ>!#bA2hXJa95BeHHsHY{^|xBPh8g4B}%uC!TDus?VhXT~dTM@h?XQzT@7|wAr#Y z82s6y3MQiiU{B~!P0=euXTjiMT}6R0XM{OHrUS(^7pXY-F)>68tf3Tyb7Q-#CK1je zY+mY2!67a7wLgg#NJ_-3V9hSJB^-l?#@H=!I8@>7%7NQ?#nXMtzGC6G5TCiVnL z7ZcLWHFJg1^O5G!TB_c(jV4a5xbEo+lIutkY6RBSf#^e#sXQ`yXix8d2qc= z!V;^P?4l%-gJrs}N&)1-QuT^Y){jK^B+ywPsRF+@hNVj2TVxAswUW18JBL)4Dp<*) z?W7~A(&EP|6zUb7bhL4Z7knq;E;G&Inpe!2LQ2U7llP1!_7F)UQ0Z#*%&UoDz>D{$ zoMn?pN<-p=@1dwX6o6&g8grjlOu&|A+P2b7VDq+PQ}bGU$s;T39Vo$+6a%3!4sNUWsEn-lpI?6WPFvf+MDm^W*e%OToi%_0Z?BY)_a9v~bFDg@4GN$Pd zq%wJ!PB*^e`-dp>l=Of$bk+uGYj7EpYs6#)MKXB0;rHPWw&2bj|F$*f0Lm*KxLGXV zBwdfGK+z>;oGV!{fA`J<|@fEEJVQ z$HvBdZAN?kr85;FHV!VSc!g-)rm_P238(qfEqY1 zNj+BK3}Cf1AX@sg{3h7r8&4kzjXRJUy)+_~mz&R3_VtK*>$ymH6hB=gX)6~DyFsxj^JoQ zqa`|xag6U>8PF?W_!UA-53v1BA@nWieUyRWNl@4k@{h?0YSKFPACUuqQS7%YUxdg5 zxs6~G(^wOs<-u?}IfTidPxi=wUeJ8x<6q_aoQzq`@GA)UCDXe}AmP^%2pE8r(U9y~ z#E=H04r1T%Z7(KsIB&~g_(I4yFK!~~(x7{q^c3E*teb`GGASEV@aJ~t%49XEhXC*& zYr(}he$YT~TO*kwfYI?Za8Rrkxx;aG1&EdR`4R@q$G*`$pXJY=l)*q?%q4htB3Ds1 z*6^%5M;p^ZTY^Tx>_;ZhMR#^yL1;yQzd6nbY+dw3r%zH|`1pN(4$&uU<7Ac$s8PG1 zVHuh-QJ86AJQTSVsI#zZgVb#rx0(z^vyD5XNJ2sYv%wA`>t+PtIhoRMe=S@G;^@J? z>qvB^QV<~u557{_4bDPNtLqQb&qynptMYn%+7P6sAv#uX(9?j-J{tywaUj{1$eb1E zC#E_RLff`SHsKZNTxl{crkdS_%(M0`C4-UR-2I@60BSL+N|&-03;9Zf=%ChRsKk;o$V^prP~7HS zu&uL~4N3+>avBs~jVy{2Kkg`ZAH;n%-y6KWsSMY{&_xYBmloCdj0JgiYB}-Yvv`%E ziF=$(Aow39EL|l!R!x=q?6_r0c6gJ-_cqYN{(ZA1qO@1eo6xDvTxp#vd8T&Mr`sn6 ziQ4Dy$H%A4Jr_;&4wGY#pDYCClgvm~ujBpaIfhQBsT?m28p~_OM|wzo2m4qyx=-W? z=OOAi-#5q7SV~h-DiUcdrSY)&V{7=od%FPW)kJ5XSh}H4McFQ+b2OE_Hiu-#2s>oU z5X+SmJ34okc;85)*gziK&7)l!_-lt-Hprbe$%uVG3gP<^9*|bwje~t?U?;Bk(4Q)! z9#^HaXPgk^VHKh?RSRyk0{P$y6{3rz)ARZ)j~3 zPfYsTF&-=+S-kL^Gdh8lN9WcjJDj=J+9*?O1hN^O@DC||DDa$#-FQyH9O<=u=`sQh zyyUob9@8xFMOJ<3VlK!VVfS@FCchhw#|Lei4<8f5*C$QG3@jTRaq9D!uZ{Wa@%Je| zhGjN9*9KC2svwU+Y2`bUqnFrAY$shSp&eG|n}^f@I_Q5EQ~R6Yxm7O~539%=GpbVL zAGh6hILmL`5v|tUoDi$Arnr-h*U~_5_KEx;iy{Kd8{1Pm$Gbm@(!*3z!i+|XjMSZ3 zZH^!C+iOZrkUhbKJg4jrrY^Ub;%iBbfXHr^n$E^8fEe^?Xg37}cNp6h4ivRZH zKGahDy$N4^AueV=gdQhN{Wme2<>$l78*`(jz7v`Em*E6iG69E%tU0N7M3vbWY4C+M z_(rC$mdl`lGt$Y500w5$i(bC?!M%Am&h0%~eRK`vkUC{|aDX&b%xU@fjoFqfiNT+g z>T||P3S(V`tmU5}SwkI~ejPtuAQI3{AlUol;BZG=A^Q-Za1c$oNQn(?-T zSeQtX>>ni&${(b}f%%B|1)w2{XE&{98P^@@k~#0(VGit0b&%m0B!vVd!RkdFoPgDz zKT89FLCMGp9x2wQ?=r&NkhC(kMggpbO+7NPU6!#z5#$yHr$|n*=Sv*{TFGraj@ECr zAbg2Wh?JRSG2cj$piFTJbtOl=emOpEe+Sjh*^uMV2`}HK#%mvHH^M*w^Zac%5_oy6 zkT=2yiz*^FqL&{gXO+L%4AKj|)sbKp6|z{$pgvwcoCk753E`Ap(6QX0Ibk;vDk+~A zP9Y-RP*2|o16Lzg&J1IZR{$nfSxl;eOISZltJt(wf_OYD473=k33@|y9OCe7K80rS z?ieMd1mKiA+L@gS4~t`NkEbKA1U+qUiYlYT-{+j3#!O__ft{g<^oR3B*X?B5HZ}WU zmReuImqerI59w0Fc0R&G9`d;sXx|1hLxtzLTpno6tYz)tC(2g2kihbwN!+SlRx^E+ zdR*@HvOZ1tqEP3YPGHn-cYQ8Ij@3pBjT}Tn%-DrLkom9oL$VWePQKU$%%~aMzg{b- zB-fGJ;e2NrvEA_IKiQWJ(o{)nE+*RoSsN4NB~nX(pZu}MFb6x7Iw6)f63G70)!emB zL^1>Nj$#Ev2nVM0g~ARMySx8^nKAGny#XFca!8wL6MzgFtuVVu_)nEG;v?65c0X+h~ONsy#=o#J&??^W3eU`fquT6*Zq~!E1%~w^EQuGiVvmmZAe5kjQ0aLNX4=c z(MbR1!Oy9WjpFrp?!-GB7r`6YwAt%!YqHg^ILp0Xi_Y1Vc!M_#6gS87@slQ@$ce~$4g|6#(Fe1o7cbnsR8qu0+y)nd-Qd$=f^jGirlZWw@8hT zSp2`+Zser=nSu&Sqh#V9MIquPyv((7xN|4f>e|`J$39`<$QPOf#cDkM2h~*9*GRIE z3zOhOvVcG%$K1Uz{Px*E8{+Dj$K+|;#X|0cM!v(r?cYsqs|u2yL}UTrt$XjXYU>29 zzMogXJ!|Y~7d=1ptx0zJhs>kd3M3KVOd4qWzoX*)Mzj#MF>cqHbl* zx8cQg<$jaDzua$U8UwEaGch|o>R6VVL@2hU^rlYRvTH1C>^^^2q{uY!#&LM9&7M#f z9))&1@XGc$Gj83=4|8G!OmF{%)6FVk9sJDre{{WLa4%7}?i<^-?G@X$ZQEI~S8Ur( zR&3k0Z9DnLXT5vhs(tUd=TtpCt9$g9{?K#u?3!adKmL|HK_BQ9dw+jNn^W75Fvz<{ zzR%Cm^;Zjl>}_ryz{Azv$$CF%z;~U?Z^w@-!{5JuY+p$lriy}k?Ynlj-ycr;uOWAT z9RC$THVP1Ye%(z>AMC;-Jygb8hKkWX0%70L2XV!?aCUr$U8)SA$?B(RCNgHS`(tG= zo!kbk-y&-ELBzl21iM3qg<;OHp1O@d6U;$|BrKZ&B;g$Q03J$elF5Co7LDsO_xDt> zGS*_fT4&3dJJCWs7_Zlj!up3hM6PCVr5M~ahykv-T$RFtON4nN)|HiNJZ=TO=2^E@##h^?hX*NL7 z)x7R98=(q0EVRqJD6(K=B%TL_JQH>zSkx;gyZ<#8wONEd9o_qTx0L(z)#aZ%c*|bN zkRemLMtai>%!4#<;r)HCG@d8ilgoG4D_*xdac$^~r_4iZ35-MbY4J!PwUEAoXXg_E zCr0R72tsMQWwr99#f4TcREF@PczvvO!_5eTyw2OW+H!m#ChZXPFs~2#9Xl@nEZo5ucJi!X0PFG=qZ#UQyyu)jvaQFIy`K*;hBVaq~c z{E-vhCRk=mvp4hJ#9*8tuf}Z4qG*#Lp+*QBP_p%Quc*iKiQV)ZeR4;1FCF-oucd+He zt`&vZG3V{7yHR%<#_==@O^=2%vr*qJZ(G>0Pk(*Pu#+M3BaK&EFH(`H&k zb@0<VBlGI*y^p+_*q+KE>N{7W4A9n+T#Ct zkh3Ac%ah*y@o{xG;3oyD@CLdr!vTVT`DrY2+tcL+ASL_l$NQL=W($Dkmt$JES`w|F z;UkPLtF+fRN~Tsz%EC@lQlo(R@mA(*|M)v$SJ{jDg|wZKn?UCmU}>B;#spCdv4={! z87MQf^%$Eb7CNes#&}!~4{`Lqa)dV6D;V@zu;7`xzi?q1{xRW$X9L*YMXVv*`Amt) zt96tJy`|wxw?FcCFXsn+?)4~c@%sRE3ygg^IRj8!_8z}W1R&?>bpRcDo+_QnoJ&G} z|0s{6-MBEOeC~2`3i*EnMrV2Ix`(XKIeydF^4{GIZWy>*dhgWNy{5c6^zZlXDbIzx zpMFkC{^U-w7iFIObnh4(KI<<4b-eG=26#JDPS_6k-4A2Bf6@}CY4u_?@0teqNt*j7 zxO!NawHzlkj;r3#Mb9M)W_hKpH{lwV<9upF{_g~1aW91XzfeFxd_+J%|Ln{Ee*|Mk zBlrIRFs`*OT{hd0e`BXDLW8w|aW7J@+Q;ro`w_xF~=6k+e8-o;T-e=h?WNK*DKF zECt2=!|CJy!|7O^9{>JFYC#M$Scb>nWxKNjS2yDrw2%qjXqx7#lo^$GV-vAZX(~iG z&QxAq{FaYqSO7Uy1%6z*S#Vo&wt3E17EE2YX%4bM-x;}S4nI~B7Ct%Oc1JjO{+=PC z6fB$o<)jLXc4B>pB%;u$XC%9PJo(Xaz(lP=mVy(G67>loCL37ck3th$Orbn1SJq6M zas_}m73Df%nO&#KJ%)dIs?nEeE)+{@QIsDMN?W;Eu%4)Xk3T_KClTg=_Wdt@8~vaD zP@zuxV?35vkx23caVP9xFE2-NkwWY%$3c1u5**~_*hXk@bNgEPzrx%*%$1^lyedMJ z>#tfiT@;x!I{4!@49TQ}!Xbgd^u43z@Aw9ZuzZ7zxaUxII(B*3;dWh-*RUy^(m&Aq z?ftbC#Z04pjms%ys?!~Qy+kZ>ZGUHzlx{=P!V5>uaLac$PlDi;G$Fr$@0R%$!CQfxr$fvn1!u01$kmOt z+B6Q|_yn8Rxy1pu^Usj0GVY#J}=z)>%Ln?*KUncg}z*<_aaLHKKFyQo+Ocjoar zz*WKU=-^dwu#4f7J6@`MR&YM!oL)Y@sfw zS?*5Yuy6N&t1PnNH$6rJQLki$)k$Rjt0I1Jg)|rlHwGu7TS5j#Fpx+U!oyb5n4DlX z!QzIneqajI>P!_IVgHCr)27E7K88xI)@#kod@nUC*nqdwl5&)&%u}= ztpu-@{ehP^=T|{)RW6IXd+fB;Wt^jJq>36NS%=2LsG0Y&#a|eSzgqDd*|Aydu*A;KS_UROqHw!vZ7_CIJ~SaKGr7p$ln}! zkb9Uw%`snuTKI0NUSw1&a=rjFj~6^aaVTj@DSvzI7<^Atur8rtz}skyWb*smCD51( zQQN83v-T-zpBn;NtEPu@(KIlO2pNLV6$jbv$78+SP4CH#iXoTacsps~&)Rp|FoUrKlno;%)g zT85+#gOB9unsJlNp6+-g@U#v0TlBF7mc}lF)gqSpglL|ppXDELRv2U(%y%GNC_z~q zhkFkqh&AE5JZC#|6@7(Ief@tCP{Z@r#R=f-odo)d9#i{t2Emp|`Mq{R9WfJVQDqIy z%na%{RA~ktivIoh&;sX*N z%IcQtb{82hbm!^>$T%k`IANt<*C3Tf6u~jUl2S$vx5?t+w-o$CIX0Dwv=DdQ_WprXY}daujS*16r_bI4>!*EKq{}ULz^I5C%k#!lgT*gx1)L z1=_v$T)#gOpu>n&O`a!i5|gx2-Jd?xQPiq1A6?Z;)!Z%KP>>`#l8ooaSp(M!iFgN9 z-U)BL1}{H4`(u;&| z+XgfLw8qyTrms1-F9m3355a1#8k%ITeYU}(;{y-LZ_oIHBH^-)kUE@yv9^S-&@#M4 zMt$8n)$<{MoY#lQpyr(oU`d>5Kd6;lr&WH5OiszT{K2PQWQ2r6U)4^GVED&o~pe3l4FaAmc|s+se%H0q>{=UN5|#G`k_yX8 zN~#&~_s_h)$b2=Y?4aehz~y$=%Q9i|NvxDqEJiuoR`rGeO=S5F}&Yo_G{F(NI8c3d}L9*1g=NyUE418*y|50CYYb0giMU zHXwrSM2A4C&4+;E8+ZM_-oC;%>C%yTA*~GMceeEZd{s?8wyw51_{zf39S@i$@l820 zQ#V!Sl1q80(b|neQtgjZvXJ7IWMm*SxXd$ogf?6NA0fCcj%~3ok#myv_N@?G9L2G< z8d_%#M(ZxZ_|K{x&#^e)7Wo+jm#P<3U6yJGj?AUX&tp2`@-qO|8Mb1;#IhK)`(q;=nUR>foD5X#bA9iTa(6Z=*MF}2;LTSwb zpwn=&8Tx3(!eb@qFth=c;jtq$^si{tIN(637@8VKK*-u&a;tF7(h{+@8C1boD?mZo zG&$z3uB@2*xq>lo-a+bv9vziw zC-wv~-1UA|d81S(%bgt6=s3-)Qof!rVZXTezvv!*Td-a&-h8B37tu-P09{EzVk;=S98zaQb%3VnRyo zU21-Pm;Hd5_Nq*uON++4m$!3v%?|iCnE1TNQ&}4*-!3zhh;<;MYCahoOe$S|KDfRQ z7luu-`B=9>_@~b+Ft={v1SQj2(LG-x8&)Sob2V2x?`H3enBGZhle36&}R0NP}K{Y$Qf(}Ffx)L3#6>@M}B?0qP62^x4^;f^k zC^{>yN6N%b<4OS@ful12S;SxGJ-pN;-= zGEafpIg)q=)?NPLNR-o_qLr#i4^$DkI5(L)n4Kp$HR5w$;l3&|ei&@_3$s7)SzNMs z&-y->l#AI$NM*lKT91!8juw%)6(OmA@f&-xLlaiN^wZFmXI1>x2#LFd^ao&i4h`I& zdziR&AfdB!__YYMENlA5e~-ysA`rGiz@s2T9^}SehEi$VSZYrLE|`*6KV7pX`jzXzs2miY~8DnLbY2-H!ZJ)64E$el4 zO^P9PF0p{o2*u~`%K)-iDfFK!dbOvDlu@cGI)apV6pXpzI9+I7eLC!4_Fyj@7g>Ex z!0_Equci--|KBQQI@(?635I_@K^gy-RAuJ;Up30Um)f}PEqgz_V#pnxR{{78&=Blt z6oTBclzk8j=Ppe5sa2*bSKh8uUlvZC)%Az*D)C2{>o|R9&BF0`mLs&Y{I+OSBa0X(bKS@Xj;_$xA+s6u}u~SR)iujO|KVQ zq>v(Iar);sS%hPJGs}7IE*aB3X^yx@kWmG_b|^Z8A}0LNDD^ZCDHBF429QmHd|fW2 ziaa{)BqrxWWiKfYQqY{8kgj0EkA0|KdTmO@EfQ$r0)t~das8&)VW>nN1(L$q2>h}l zerK0V2>HawA=qHC%f;rjF%LRCEomk>UV}1BgGqi>aR)M#Q|*!2+-I+WS8qO>&5ow} z7plCMLa|)77d<}jm#KR(JRJ&AQrhGbiu)&B=s6v72z9*D1X6lRSF_DOL_`5}LP5!Y zrb{6Aft_;54m}=^C&x{NgH{oWh#`@BD2=gEvjE^+cprV(es~@|qM4s>V&1ZO{rL4+ zRE#5sSfmbQNgwr*34-KZT{c!7#=H>yh04pCv2L)&2xGa9LNxIPS&uAt!${O(4Ze=f z^{?OEY+``H(N|1~+qhUn>bWjACuhe7EBJ05DPv>u0&+Ss2lz#A8)tlN4k6Bn>PyF< z-4T*Va$kqmb|5KdrUHTSDF5FEf``tkTC*P zpPwN^-|q;e0ke9nR@#VHeI`4a2Nc%KG@ z@T*zx%A$%{bWA>(M##a38yIrDAMT2t>oFJ;y%Ck3?l zF#|w$okUlHF=`s-5+Ow7^{V3ZZ--mG`)k5z7pqam< z>}CjZh0NRQ>0Cf8WrohEh`BDUZTI+atzLh(gNHmk#?7t!N(8dwbn0ww8f2-scE(Vc zJDaP_{(6lOM(|oBtCRqk87)DM0R{Ol(cCMDMms4M6;gR82BbIz zFNRZ%9%JRDaJEhOF!8{=Ch$fM7#tOfDGJe1Sfz$o0F(+Mt|VoJ9O>#CH?a%GwoB;N z_gFkAC!UNn>WKf`bSK<9^Zsq_U((ysDNgszJ%}5o#I0ZuAN&!i6Ijz3me>I@GKUlz zzg$xxlYi}D29m!6y3m*G2BBt*R-=+Mh4AiT0XUpO2dm^lMGVZC5PoM{RF@*1Yw$Es zas+?En4OYsoF(t(TrN@a9hhoV_Up*rUzuJaVW98;$4J-TFeZslXJq>Swx`sG!^C^#Y_Z(2{X@QvC=L*Ad#S>riK3@CO&FZK&w`43G-zd~A@ah? zfZ4gCP#4l$DF_GN-Pym(*3$gYiMMcF;nP}3!$!i_kj2-bdT07LbZO%G1Fu$4-6;Cb(4Pr?Y8|*Zp3@UfrAz*LV_&IH7ZzYfO%+=P)nI zY~DzJM$A|X<_;qJ#n~v-zmOhzB!ffYOol47(rn0Ylx?_$y@4)pK2o)Z3vV+*G7cUO zj}LJ*h0-s1z#19J1013F78mU(N*WT0f5ilq_#ocNk{%;8?`$;cTx<tqKk7nrXWcRNWb72``4=gbi zt}H@kTMypNUbCLDUCvlJ5=+RWU_zU19_zHHI227f0yn^*@SWGP;$Y>E{e^RV@fYVl z+Kq&Ez=8K)ZG*s$Rj_y$4Ur#yb`nQKHa6vW>L5{OF4vSNvRu{J$s0-QG#n*mT5&0` zBA$Lxl`#L9R4EA(3JT@;WnfTl1*`l`Tby+1p?N1eT1hsZIry@vS$g*&)c%s#nhSXf z5=l|C2=5nBRqgC&35an{54&C|zBw*MiG z|Nf@C#H~PfcCLZX`EQXGB%Nuu}TS!MMbN3`_hFjJ*(Fh zyHolAE%Y@iWkrM&BrwN|wu6{Wej&C2|3hbo(4g{L@ z^y-G5yHNik+*DB(yl4Ad#|F5pODmlOO+~98l{Ir{gxMcB?BiL^9ArIkdeaLkNy9rd zwaXobxKZ28xw<@?@#+PC{n&@>Lf%#ji8l$Z2?44J0omjiTt)ea)2$^rACY1YNE)kA zKijPh0P`2f55*$m{L`dfv}uno0W5f0yb=vmtG**%BoZl(BZ z5~Tzr{cDB$vTajY{-|I@b#nPKQXhgaTuElel+EbK&B=+mkX{cU)@p+NdrH|ShBb9j z(od=z>XhJz@fYw3P?;?wzho4JN&#h!nmG$4o(E@TWCp$L1f$#VeNp%E6}6e__To*{ zP1O0xiK~rGM z_ORu{n zVY5JE%u=ehIsa1g*v)aCf1uf;rN-tpKfw{MUD}Po@*PVGl2g0lNdf!dJSjHB*SEvO z9f!8g_6dh++q{qQ??Z`(sGRj)+QD5*LUm=|x-4M+y!7L#^1rhWD>MHUp8xIMLp-R% zqcBy2Ew*~p(yhpIcOuGS4fZu0`f@!xs z0X{6Sqid)qS}}$3N`hVHXKbe*l`bx)I8-izt7%5A@7#nD zpy^#CIJ2|WGOs%zIn4REa?)~KV3n%bI`FF7oY0U*k(#FYZ>`C09GwM`8_dm$P#q!4 z!|=R~Qqki&kkI6aJ{?pCIy)>6ypT87)3^r_D3lJ79m#f|Xd~InfXdifo#I2SG=JYZ zkSe%CxmIqU{&yW$GD`%v(038T&fKv2ZKylQhJ^!GT+sE!yq^sRy)KwIug#7Lz^oq5 z_Ip(}oDH66z7k!|NP_z>gm`51MyzBuR`|*<8|rKPp0{queFiL{+`7}c2c3WT41HOP34 z8MGTU$l|z5kdtF@G@)om+u+Z~%L{(`Mc~)GOL+4Y+7yfEFL8SgZKN=Z^!gfywtslv zFBLn}g@bW|g)Rg|3a3-tG>DPUZ~xR#A(UTrB?+`FZ}>+z`$P@5)p;Y4!!dA0R!H6h zH;oUiTp#()-pPK8K-`#hpacnb0?v;*QeR^-+!HW8dhWZ4=+r$(0IjPEc^ZnRd*+A? zwY@|McOxb8O5g`V{U~f1hI|155R6i8YdKPc(l-OaL~9#*`?k%l=5RQM#~WxWjChOG z6(=qs{Ds?QD`>xQYA{G$LpUZ(ak!*pJ9LhfmPi%W#DTF5uYc^$_0&Ki#M>BHsJ{y<||oR%AuTyD3+Q{NC(xK1jbKSQL(Q%C>eIa z3P}9?9TfppyhN!u49j@j?X8s(g$d$4lJc$y?$uIAgcxGr^`nUH_Eq*JGL;vd0^E&FHS4W)x`v}7k_s5UeG;y$x z22KML^K8XSWA{{YlPt9fDt4mQaoh22V;TkaQMb*_@;V zO=;CoV_ku7*RFku#(C(b9Sk3Mg{8emAY4-#IaIcFDx1c2=61K0AFtBEsp2nE*;2U( z%IO{kdmixWV6n+D^txRL2i%{2O|+^xWD^8lb34;p&}8u`On)iMdQ4z0IMQ20t-B^i zRLKNmWX%ssb`$86$?Pb_p zUh-qikxVbn@ohYvmrNG@*gnk`tQct=mb?97c3>yzs6}gKR%$aPAC=*<$>a$-Qv^i6mF4o;p`#M*H7Z#SO?XhxtSS zvSP*?^oOtg;R5BJ36Tp91=@fb(mpfPNE{xtC%jOkpMEGbOsVUslvW{W^We*fAfLb) z>+LWr8?p)zD^$wwku7u8doSx?BH?N*_l-43iFD=RHlpUT5s)(~<}KxUt9{M5l4ZFZ zTpFUyVSHZ#^L$PZp!&Y20uY}#VuM)A^k0QD3J_~68905qpm1p1l)~qR3_m0EC1c}% z2T4OlS|;JT^|c1g^Ni8sbsG{9{5J>taqao+%*4XiS|x4~VO5#J5mz!`NN?v3%%niD z;tWTs6WVA!q_Xh85sv+4f)eIcG;i`fgK4=sL>;SWzk`&C?m>{mXY>ztMQDK8dV=m} zX><>KB&KwYz9E7ju+N72_~6?iG_61_tG67SZj*G#6kU*>sK>NMxD0M(Xnb^ml|ABA zBR7&w2LrklfzK=oumJ=-#LWSLc?a_3t{Hvdg;muZFp`nZzru@)`O&KkxH?9S(!(;q zf(Kse#37avA7F`lb}yo+jFC&aZ2>kz5gZVMhG}OABJVMx23!+IGf)rT%2+Uv49~}K zZ4i;rky{br0GF-91e6reO+V~rn++Fn1lr)s%N5Hdk{GY+b_Oj)ASNvW7=tj*woZuW zsGflrv&~0eF0#x1K@!r6q3)Ps&1DJ+SH3VRuP$IT-M<^g{!wi_&)4=YHb#`eI&g!1 z>#xn(mY_G13Ia#e{dEx!p2Yx3?GeW($B3s?Ed*`O(6`{{1k5>GQ(A_r6p^Or^a=Xt z`tZGZ=r3D)f-vYYnpy5+R) zVZV3a!w|F4wBDZe0E%E1#(aco+>m$fHSR__G(ybTzmF>qzfY|f{1!qEC4~M$QJ0Li zU1E!a{r%|;0{+vXk+*X>Tfv|=yw#9WWfr=C6uN0(65^2?!LZf<7ldz}PokoEgT@E$cifp7hL zm;oOd;ke~kqpp*#^7{G^W@|E-Mqf-CiFdxW5PXf!?#K+}fDUr} z;Uqi+*x<~2K^>}x5xUUmD5ECP2C*>qdyVD^JSi@#gxuNn2#B)FfYJho+{6 znZ-^S;}mS5lzOMn-+6wBXRjJBpOHO%4ocHt)?>LvVB}MJ`hqB!buTc&ytZKS1HI8j zS0$?CqnyW?q)D-|hrHtHa11mlj!geRxZwF2QG7zWeRZm6m4R!}R;vFtyZDeyUVxBA zN~KagVOhCUtB9{(spa}ZOo5ej3I{;|GI*k#WDSFDf;u_fjJtg$n3YbG%be6(}m!jcoy zv7*rHPoYZ*U2{_o^Dt2;eD;Ayu-mWrcVOP>#ewT@9Cem<#RGyWkRC{o9Hq=@&{(VH zLcvASkkD8R|BX7kBr>-9t0N|eM%bMASk`I-_? zc%Ps&{t1;>8xW_xlgv_&sHm++3@KX(BYKRabqx22R81c}%AvdpA+TXSIsl?L1c=Qp zOQ^fBa2gd39|mLK41zQ}=>4uO7Z}IIcuv#micfqoAq$q!d>k3XKZ=JP0`+RDU3;GK z_D|AWU1nDkNIuWElajyX2e&n~%#n;VDOl#*X~rOw*C>y3MTjUTJ=hI)tzWp5Q$MJa zty_N4X#|owsV~?(Uy^F$-n4nuJSygSMEVF;!qQ`qVuf-o&)yp@uOFx$a<%^>^PkB} z7TqyV?QfV`x}z&BAaV{iUt5Xlt2^#cGEV~ZxLTr2Tm*d#r0vRl!#3)~UsTHA_u&M< zVz|)_b*iQpit%z9zzJ1!W^OoRbXb9)7$$3zrLjQA=w_saBVuWSXJbYKRr1ya(3o`q zk%@oMBea*O#_NfW#ed076&BSLw!-LkSc4Ub6UE~_L2xuke!YTj|8rpGogf7&-JCQ6EOA=}bE@DW=MM9WnoF*=saDU!BZi?#I1wI=>xEzK0={f59{j%ow07e;MVhoNKFi+Uc0nQalShC^Vl^6FPN-Tyv#7TBadt z+j!-qkEb?b)vHnd(S%qU1?&NgZAianWsm{yGxApy<|50>;gOG7P{wcEP>xOMIfZ)( zBr0{^le=`@{CDW|NFVg56{d3Mo--5rsp4-rW!I(|(n@Kr73H#I3ufm|QuQq}tx=oa2h^%9vd& zQE^{gkcCO=mJP%|3H)cwZNtL^Fmfgo?~IY}e`ZNGP|XA|%)LG81F{#aK?=1~TyCS) zZJ_gZuJ`5fngtFQJ4H6~L~b@w!3pOH3KVYTVMP0`Qt206{Jg?czS_Jco)=Ns%t2no z&t?o#8n|UMNkgGh)vg|HL!wYsBZs{e7TRtIz8!nwS2}MSJRCS|l=#Gr*2yKSiZndQErZvP>1u$beGLAV=tVI>b+&5*6;^un@{g;1HHPwcwLaRMJDhm|xjNIXyks>QHe89gSDPXU$+_fVMEELc zoougi{hF+O;g>5zOGBE3jm9Y?fZ!dP@Wa{ncnqQdn`Jqq9mS768>gA7j(7S=)1l?L zbKHgsT%xnb&pb}rN))jlKcFuco0-!|xJ%#1Y%G#V+2zTW{dR?i-c~lZrf`~Gx_-dv ztfIZD={9!HIUvurV4x$u@sZuN}IrfYJm3|e;J)TnZ z7x?bp`@xz6P>c9H1TCk)aCN8!^trex@|_0$9eQmuDGuP_=I@#-cXGm3Ka|E}ra6rN z;zGbynk4pnklZEMS7n%dP-3t85Lg060P;^kj=J)mw`gUxde6;*;nrIDjJGiASK5Qz zkMFyGIt_D%d=a7mImPnLFW7wKuhmK-I_|lx_2cBBE>H~a;(22YJ|I4`de4;Tp^pV| zHsGHt7fAD>S28?+FE>dOAuJEHi9nh07RKd=F%-Qd|F`j>$b(_sse<5+vv1^>z4~|0 z>Gy$^ZEhe&u2MXvWk!!z>T#5DQhgpSc&mcQ{H-R<6VF3%hRTh2wliuuwqwgc0UL|; zE>uBPA)m5-&w+QqEw!aVN!IMcA|G}?i%~jIe3i)b*&K$4>-76vd&ksJ&9&Ez6UaxV zh3t%#&%ofjtt0}H9U3nqT-;AQAB?!kJRVmH^a-?I{pta`=_51(i5Sn-j7z7TRsLa$ zrV6JmAM5gb#5{?WAnRS5FzDCYKI}Txbq*l80wY4NS7YFN^zmR^aeW%mefta(@-kZ~oQL^g z@-I@pZ+~T%Nl+C&qWJq26Ta&KBK%s<+`|q~PiT#LP|RP;23A1B9?p1*NMv9=2}-%- zOV#|seWk;K=GK8_uAv*Qs_mp(woeRT9x{CM#vPcq?iR4F4=nE&?#L+$^bjn|4E`6RC?_pNqYCF#)v`-}epkQt;;z;J zJG-FM+|j0Xl67Tesx=nc;i0v~$dQ84<_iYist4W`Fq1gqmI^;}x2Z+Hc!ov56M7rF zUv1)XiM7LRhOKk{FH)+{T`J4-PocXr`LZV4cir{?eGZ@6EN6X=&l-&h*Z+wy>-~2` zZ}NF!BoU#v-qw*8qU_xMX?2qI&%O>;PsoW}+rl0zJh6eArER(cc1%xyM#n{#^Ce@N z)e}6SuSk)@L~Y(Y)A7n}vK42kFPCtb`#XfiMz#&Wk=Cazw8!RRnC)5}S&yDi1b zDTcRQJXe=gp(erK3J{OZUrZhMSGMH8hJ6k72@#y0l6{?j!+&n4pnv^h;zfx5iwNfc zsxkvw6bzB_fQ8iqS&9G=dK*YxBs97=@$R~Owzy25lcrY0qY)FUr#H`B*P*;}StXvU z9>ePvO>l|DxAM#Dugbhh4%?`Yi=2m*V~0JDc&i)tamO_l{_6SP%)45+nY)uS=4;Ml z_EjhSEQgM`i#CW;bI#Sh3A2q~%q#zIRg7tl{*Ts1I%XzEyto)~FH7@u3|RJgAX@Dm zaUnEnHe;T9-cHi*YWk&u7r9+P-OLk0=W}qy-R*6$dN}m4v-jOis)Yydx{Uz8o9Jvg zCm~)Z_9lD>LrpwfoAq}c;T>xBr`cGyVBm`y*`sGRn7C~Q*`Ar1_92dYoK5!99kf@5 zBj+z^WBc17G*W-nS9+pVY3Kf_H~KFLhz@@v;V<^OHuWHn5TO1OzMj2tw{EMk@4C<8 zF=e!Duk~NI*#!C_? zf)x|PgeigznGYHGWyLMdFWZS?PtI8C3%%I$$7aD*?mIdb-M++LPJpJET41p(BCz82 zpMuObjuThLozGMHyV={|gR2%Cr#Pm=SWPIbMg*P3F3H=KFN)ODT>W9=7YKo% zgD-LaduFokLU4aRB#zKC;%?Cfd+U|fcbWIN$$AWKHtgH2-?ePj&N)tVki&wytKCDy zAJ`G4sp$c$HY7nhKR183bd(O(X5Lj@LbMt=777dBnG%wQU0%wYYAZ}XY^mqb zPp`BSq31t)V@dHaRHJmS8tV{Wb12A*hWQjD7UwX+e zSYib2tGhE|y;!=?{W-BrYw_F>-&QU90YP0z6DmULluWM;9+PYdW40=nAu;aM@|h3$ zeI$Xz-Z@;ZmyK9_7IMUpRuf5WObi6AhG+LlT?&#|gR&45la!`!{1~G!faBTbWYDn= zcUs8S{*rA28<#V(a&(`?eOm2b>SVl+&H(Tr%B><68gOU5_Xt7fMz9NTJZgmbh27Te z6mBNY)z=s{%v;6nu(;DBUX!TE9o%7^k0I(KXUeA`4|O$U!n+;g(&;15rUBBdmD_sP zg@#motYJ$=YS@CJs>H%v6!{Z!(u^X6d^NvM8D6JLPu2q?o3i-#pj^CY(oFiv{cp+i z#+8xWbKMbPRV*G03r)UY;)6kf7FCHBt4&0ZXxh2GPWAD-?6PxUb$Zh%s8Xa_gDNgc zq@N;93zuogP9FV1tF$ifkRAUyI9iyN5UjG2qyt3fuhwtnKwCay`;^9&RwwgBZ%I&I zG?Ykf$Sn|Limk2=?V0DTK6yGcx;e|OJq#^sJ?}xAyj(!Y?;kL^Q13D)c&KJDYK<;d z#5AvR7q5U<+42e7OMM}|hA?T(ZpCc9`__yYS3#Omb=TPYVytm%H#E$cS!k19^e~K_Ga71*Ab3p6RW0a4fwKK*`+hg2+@5~r;RaOgc_DyWkx26 zl((!v=|wG~uf*(`QLQ8v1|USGsW7`4zZ^kg@(kKadA2d`zoqSluz31lv%?Y4j!7B| z2DZ6vI9SoB);0CqBW6i;TqD;9a{AlHP8ekMMB<^W5D&l@Gom)%0wE_bQruAA56~wt zJQ;OpHP%upE|1tROTg@g38Hl66pT!T-D`j1#{Yz*31fL*@6uqpswn8AIE%LZAx7eV zoRK}ZbaRuh_;g?LEI|~v`I*;@*N=gg{1kMHc0)IN7fI zZ(y7ueK6|GPqD{0RVyrXV6-OIu449t%z|?lyQn_V_0wf8JE~NG+W~QkakgVX*Y5>P z3<{k+RwY$J0;cJ=Jd0)j4C1J!X10J25)^hK?_-fb}?aXl*y2Lz5gp{=~`Yd9TpqL z=;>?tebKsgbH!lV+Fry!VC=EvoopaUAFNi=)~%<>1ODM4+BxX&tVfVJ#_+dM8S@$Azx@RxB`dV{n{TL-<-CwCZ{?oT$CnwJ2(DO`0G<|)4Ls)qcPN=X@64SJ zZ(U0phRj*pfJ!Y-iHcH-igVeeN6TGG^PiuIR~KSi8o8`$g+98M}vghnUSLAbQse zW28I9sR+$VM)0b!Gd^ie5^r%6sc2azjc4%a%~ZBEHa+NU$v54&N+NUb>(Z&CtFNa^ zJ)_K}8}U@Wc7|9Jitd-vq)@dfdgfCKr&t{?x;7}yi==V77I62UhN7{rA6$8e?sFM< z>QOpe*P2Cb?~{a`#2A4r(yH6HLtfCe7dp3PRFO5}x@fichjUvWP3-rioHWct^;q>z zj<{+T+&iZw)cQg3VMqK`8K)uz5V8FyJW?{3qOo(GEtd<;$$Zq-TkhwoqyYG|PFuKP z_Q4G9bk*)2ZAc*wgLB)UAqr?taG%P&K3*r4oDy-3Q~>y@*a6kD4uB{54TfVMTq5Y; zV?`3h8t4L6yIgm={`QDp3u#Q=J#eFv>iOcCvZB>5{71FCV{- z=ht2TR>1$F>l}gvQJO8>w(aieY1_7K+qP}nwr$(CZQI71?Tz@~rq;Entf)Gf=X?W5 zIn)&rZ8T0f{DYQW5Q&*Oi+?6it$@mKpc%2qa#a{pw2~QFad+W#DTE?&*Xhl*E9{5sqg2j+OUilC-Z!b;1l+UDy4!$WyO&V>0pr;Q z4LdbcO?vfIl8)MlXel^_W>c?|G;A7Ke*9qVa}r6}xExMVobau8!ZT-i3P*=n^Z{0e zabeSFnRJ|cXUS9fHfaLPihd4gYN9pDF}11QZ0_afC3Lj$Bo74eA<`$agODH*sYYCG?7A7Cpu z{i^YVqg1%g!r#cJ{?Agl(z-3kUk1=6PSpvzJ7Bm|?cDi(R+X+I6!d{wGH_KhueoRyu1)+O~0h=VcDbxH6ttTVBG3WyY8sDd8W- zPdd)Wa$~oLF%1wxPGZy1<-0!%69E360R!eADVN9VEf3uHc-w^ZK?BeDKt<`HY^$V| zVjmy;2d^xteqvn{UD+0_Jyz_4hGwFZV%zG{wY!JJbMzlc`0e0r`To|kvQwuUwHvdy zRyES>wTm?P%zbir-3bo|Q7UDGF!ezDmIy=c!)@u6)W23IEsR5g(fw)f&(be{2W4c% zjq*W%ptJH7NA$V^NO!JBF23P!87uG4!(M+NTV&A{ z=6Yg_f$T1MS|}G;NupB0`S}jVbkG$yJ5npFYa9-CP&{Xc_xFXFt7RWb zykKxXO9ex0a#1-0$`yLi5t-y)NZJ83JwheC;Ww9|f*?~hXqok5PWj#BECoUIo-S|w zNS2ckfEcZo4$;)#kdz|Zi5`&;`CnW2yS#~s6X z@k}(a3M*Osxl2-EYB}TLGvo50{xFL9$czrRdPH^$P$tl%j;?T>lu=z`7x1%HG$lwO z^o9spIUJpY9X^lS@G`^b7+-oNH{cE>+enkfYg9`!amD^mqsD}Tihh>;DvxkjsmoO3 zMB_Q+IK2ws0PQ4J`IW?GGGg*zJgCZbg>uRkX7D((S~=9 z%2ibE+B17wK|pLYWiwWnR>Ct1X)5+78s5A_!{-4kCpLfn=KF|cRm9a6jk9?tPgV^; zClK~}XuAYO-6Q7=Rw%8l0{cG|Xzqm9@pZduD(B4=AaT=BR{0esxXT$8SRSxFv*eK4 zMPQXk^TmWpx}$XQ=jx-HHOW=5h{q8!Llx_5dUTitB!x`^vj~iqDEzg^bAwC&knTEx zA}gB}_5Erg95oFnsJu1>68=@0TIb*knN@1S64z(k<~J7-Sy2|)I(R_~w6x5cf z(8Yf4fZZhy#sb8r!G&&0@Jxt=<`n%f8tjoU^0fcb{W{8(P>q|=_GQXjnEP!MZuIFe zDKxd`*vjqu$jfu}yHM+~3c)POHlVoUDVMDPZCkVu0d_7-=gUI21eU3bgK9vfFJ<1% zhAF@|%JIfmMOO4HPbEwTlrjmQ7uDcfs_l^A6(DkstdY@>gv#Nluv zm>FLXYMBZv{obr$;8B0vM=NZS^@3#L31mJ5nO;hSEr@7@OJ`J4R4Fh~(5=Y@QbSsV zhmn)NTrfmMMuB-agG)AvH(FuRtd<2Z$@=jp*HLHw+k|Oxyz)I@Aen5Pg-cK3!HUV} zW`NTU!~%tjw4XR&q2{{!_ZB((6jr-4h;XSs6))cHlB=Mx}na8uLwD;1^K@Evp1Gyx{U6)8vf%a^tWt@sD{sB2|F0h7!Mb zsx!3VY~7chy3}tP$#aWh4P%lumh&Ayl~jfbIZ9ZQ@C{5?Vn?jK8-(ULLAB6VEwroh z5wIF+ekf{6n}%8Gpa>?!6I)l+mPk@Eh_^Kmj(_s>Y_$Um0kPUKwZHNYcT~qQ2r>TH zu5-N_+gylH+&L+!T>9~6!0B8IncKPuHo`yxNl>JU(KOAfH^|N(jaw;Me!GE!iMYMk z;4Tmh7W-$p$jaz!Xs)2zX%W>`2;OHS-JUY5)Nc89mG4+Jq4wsse3lg9S7 zwyWtsQC(#&E-ejhe?sM;E3EhH2q)n7?~bfNyPnjqR`&Z7FEQk@9h8@4W=xD5_jr~N zInDN(5sxq2QS0|ZpK-Wdbxxlva)ey<>)~vJ+M3oln={mCbeu@@5BuZC#~p|)EP^QS zFwDtK{5SFtX(td~{|*_&z!BHURw%5oC&&P4st#+1<@8Pm_HMIb_JK!d2Lnk4wM{Fv z$HhmHH>h(s;!PQVUH1Xb{^ zl9L1R*jMf@yl3a@U#1H;)%zVZ+jZc1G}V-|T*2gP7PM+#!8i*AM*?Wzw)bpd{cG*3 zs~b%PS|kaJ@C=ctIPZjUC{3%Ed-Fq^Auas~S(s``KD)66AVkYxE^h|%CCwq05wfh& zbOeVlJWRE@NswhzF-%o42$NIDHCOy)S89!T6^|h4#Wd;fm}9s@yqs&B5dkU*V*#Vd zkf1CSfg*ivE>j7Xy5#)ZE&HrF&3NudZai=Zw5@PAdYfkhS?wvY=b(3NDW^$wsAs&_ zBXrNQCIY$MA?*T0ESvGXabll+>m^$&=j_*OgLUe!J^!BnWQTIIh>6K17h_Ems(2H`IQ%8(^V1C*2u_vdo$D=#RcyG zp$thuR}^mf5Awy>1#9^AB<)hcJ!y|9dL4s4ousjeaEi!uuHxC9X{E8;RASTM4wyrj zU2n~x3_q39eq8RuKd(gp7lgbfcn3f4@DM0gQw(OXnuTFBa81|`WU(vhaX=^)b8|#` z#M-OdWqoN_pPu<_`Xhr;a{nplMff_QC1nbVYvDltJuU*<`=t>zKI2={J$j{kqb!(L zyEgkTocB$>42OU>j{g^B=qmQLQubk6D8C$_SK!X3C7NqrM5>9t%seC=xSg3wp~I}I zpVWwi? zA4e1Bv5jpE+jqH@CFXNFx^FUI5C)})Ya2|twBqB6y+pEM^b)X$oRglY!iBLu-HHe) z8Oc@RP$H*mQ*zvE0!f;flRlq?A|a9V+_i(Zy}n`Kc3TescFWlviK}lUNpsd7@DNYE zHESam73)R+XIn`%+~6CUtw|TY5W2#Vc?ly9=!fzuY8ODCF#Yw1Kia~bHpw!a*w-;; zNjL~I7Z&)(MHGG3H5*so))~aOE$|Tvnhac}**H~6j7lN5dT>!B$h=g2xVxhq%%|xr zybYEHLNQK9H+WHZYvTs5^5}q41#94#t=P(u#MUPB?oQ+%(#qjz)e#fc;OADxB-`p4 zXE}0U8zrZuq&X*+^_OWuuLm{olmi86b&3${R-l8%V{sz0GldO`>5lQN9Ef5b$c+ju z$_6%Z=lja0vJ+lyZXaS1jnaB4FpBypdI4~5DH4TF;8XonkvAgl>t`CX5e?+b+ToW? zW<}BEXqaWY;!Dc+?ny^5~)EMI1P*#!HkzdD(=w(w1+-ZO?p=*OyDxE6*=-R5vlS!m_ezn-Eg!PPrSm?ST z_Y*1m3I+`~hSR|t_Z`M{hkrS;3F2=aZo&y8`b6gnJIPqC3j(2oQi+h$UKWt@3x}>6 zu5^a{?*x;WQ)+zlE+3oo95K8?oKB0yy1%<%`pNC9;MkC``Adi*lam2R^#$A4uLcjF zDM@0^-48itI$8C#P!$HUq~Ao~gv?`Dw*%PMnshxl%tOJ79m|0ZvaZIuX@HZHiSgoe zEW@o-E{XzXM2IDA#~W|*w#~`y>Hvqjt6cz@eFtCRIlcdsL2!)L)g{ew4EcF;@Dgcv zDbx-Iz6Z!nIxll;D=~O)h)>ltoy(O_N46&DuR|^I8-T+YxnLUI!j75}XBgTF(YGP* z)@CFuc$2!d)XgrJ;(L~V$XMJy%pPf#wi4P;;E6O@vII$S?D_T zf{&b*WVGn&ps})8_6p)%QN7D@O8#tS6T`biZxWrmBpGPm&W>&FrYN)o3X_BjYq%ef(LOMph+7S6H%Is!pS>$<3jYhE@*)f8GOvmhlt{az z8hhyOG%>Co{1zJj6Mxp4X@hSlA1(e<>Krqf=_ zbm+PpGwRKW7ok+^e8?w;#ZsO*-SxGDDSA*f<-P!_)nBntWSxQ`dNX0?QFfc1-!poT z+A6Q*y5F?rK^~^Zqh}@4PHVBCB4N5tv%?WVO~y0A0!S-yqLjv_(hY|~tZpVD$6htb ze2yp9(SW$HTtB$7kY$7I&mMM0fQjvc|G>_Yh(;&_gpKfiyI1}mfDT?=vD=w=wv!?B zx|=Ta+7jPWTD*&*KN=ujnPGwJtpB7CVHB@cq%9>&DU651ltk)Mq5?WC1%d=@52ZL; z(5bY0m<;G97T$yYcRpGfeb0r(n$%hLHkyvMY^g|PW((=PkK-sOkp7Px(gsAgi z!w!k+o-)(jI_lQrq~(`qH?Tdgz*UL8ta+ za3iNMrOZLNLC4uU%+r`Q%f6Gi6~a~*$?+>S5Y(s1 zOF&6Ddmss6^m)eBpmUTh?mLkvanpBT@OGU?R7drsud(;_f-}X1godP_Y9Ef!UgLI> z_avYVjZt9k;1@1Lwt7lS>C+}>pbDkb9yEfmHjRjZo3OOy*9r{MN$~3Ic8mQzcIxHo z?fW_A%!0Q(_}rzp5Sos?^F6X}i%XfRnlKii2>>Mww0yRFnYQg|IH_;4Wz*zpTmlKo zKP{0qbKwQBvFpogW~$N}^Da$DCh)4~l{!4-wNoYK8^lj0GPfcxX_>;wQ6pl{*^a_7 zFx?6np5tizH`~Hk;q8%4Ms-?DH~D_z46BJE2U@Ze~9EvKb?lS?U zmqX^zABL*3s$D(r2{e~eMh;U=cK#%5qOcAK0M$c=6N#{j(Boi)!*r7JU0l~4aR&ge z8r-u0lHqpy9ibwdlvbYKn5ZA@_%T{pkT^BgzJcLFay%16!109f+~gT|kl}+97;j9> zL^5RpgolVD+niC0YXrAJK`(BXNUsF^som(q&3h%8R6fH1IdTsnG#{A-xfk0qJoCAt zxsWK*EM9dWvWy;+WRkl+TSnB3nxa09B@Wh<~?i9=W)5+3IEtEen;b z0wTFnu%|JK0TfOe?bK{XJ&x8Q#>!zt@R9&hi%z#Gok2|R@otH1Gy~K)q~FGF4Mati zln&6msf|YQ8(&b&Ov#WBEF({}JaW}n9U#GEvX+C~TJ5`L5__N)rwFnUoewooay_fc z=Y*VbOVIV+4r~~j)vMe>V^i3>IcFO+Jl~3VuQH6*OhDn+>m3}ifK!ArlA1mzO<;tm z3-#D8JP)}~J;BKcP!DBhP7Y(LK-%M)0&^0%AOKxh-XJIRKr+`a1(Es z38Ychz0R#acHb%j_g6D%7cgKSful%R{Jb-!TchDkFPg~51+_&NPfY;zYD9&mBvS(S z7qrUhFt$&eU$+X^Qus1#FdTWQDlJP z*L_?GJwo=I5De=6Yq%wxau=wta4fU%OF8NA>3*x((1}eMMP(B*TiF+TksPwrS7`Y7 z#T7~{=h zf}kz(*B$Ms&tc)lgmib#CCAlOme9_L!-p}4jdonU)ly=lw9t1x1$fs0QQc%C06pV| zL42r+FWL9D%`UN;vVLyzlBJ?nmnJ#}UDHi4H|grORyP<55KlhC$YfW$3^&8G!{J3e zZnt1Tjb3z|J`EmU^*@)3IP|xZvvKo3Fxro>C3t}S!B<=N;?TT8AgfUs zbEod6T&wNI>mS~lv>=UiB?Xx?I9M7C|9bxI^s|-P&`kWTj+X|KJ|;=7d7C^~nD;q~ zbsoT$?phuaLUtLB?`F5V2;THjiB1@lP79$9ITmfq6IK#w??j^bAfIPM*s@lDJUV!gNoHKDqZ5FvxJNVP_SzCPg%G}0+~^8U09G3y5Q3UoU#=Pj zGMX@B-^6Ehnc^=z2uuH$zIooLbMcRGS3K!l;AJ+$(9|{1HleV*RSG>`>D$&lT-yzmp8tiOnCGlfuPD;kDnR^7mn? zV}`5qt=H=#$J?b`sxDwQM_;|;Pd63~x|yd!ptI1x*u{6hv3Vo@AuFgN-q!n%eFnD7 z8TLSOMu_f5WZ#)4^WuW?2{D;bVDLzxeY5>&QSZca^pducRH=@bJGwpLzZybip%L8+af{ zUQf|_9C+S$oO+Atyz`8HAda>U6`4!bb5nu#xvtKhEcoh*RxMexQmwNiJ7xxEjG28E z)1BQQ9WF2&qwE^ba3Y8BU|fDIX+h@;Ol$lFAeH<51B?KX^HtS4b!o9XpK;xRby+Z9 z>@QHKjdfA;f*wXx40iY;i}kz03o2t7xs&f}@TIrR+}ZiVxY2|AUKdwBSwrr=Ep?Jp zT?BL{iN=n**f3o&J!AUpvp6oLv{)(Z;R$R|nSbc{G zL;v)Lk#S5vJonQc>cG}4KVhDrJA1h9ooZgae}Li^8mlDxfGkdex(X`2JlPcJ1&5{b z-}~^7qeP27Bj7||kt07VWR2+KEn$vOW&M6*;NIC5kQ9(JJ#E&snY<*OUijt3wZM71 zrBfWVor3DVfr|bCCH;QqJeccA^pk724L>KCM}dx!42!BoM`pht2Pwvh&|sTAWYCVo zre*kNkh{Tm1A!ns2K;CaMLm?6umOHyi`yx0jC=T6m~Q*L2w2fU^UE!QH6SL0tH-Td z+LA}vn53QMf$WdSxm-AU_yoZ;f(*Rufb~aabSig*ss_U+dJ(-v0o#Df4ZF4?3jGP3 zD%Z6PcZBEBo%br>JfnRBD9c%|y%V8wUbf|bXV;CxwC~QS=?qAE)^sbDI`qVXzXU0G zl0FTr_SQ85?h6xx9eW*#c1_2>@7H)=7HEzWmJP>q+X6-(@A7M~-8BM@7iOlb#ab_u zgd_`}!9b-JX# zj5A-3EB9xZ)ddKImzKNXiBU4olHqlJsWU}3+_dK$>1PQA zU6zO1i8i2=g-#m=jm?Y`j?IrXLXW@x(S$2)QJ4)+0e{wPpJ(_Quqt#!CW-`qBylxn z!%tF}r%ggMq|0Uy{wu!~hnC6m1$d~96PH+%25oWNGEum`m+{QxZyuWt&61}E=?-X9 zm%C(MrY+1pt0?TFtja(xT-sSH-a0aK}Pfz`F2X`6ob8=6=vhEx&fJPWoX5Y(o>-J7X zJzm&xrAUx1TCD}&k!ynR?W#j+-`%HXckdrzdD|NX@wJQg@zs7(a(|2Ob<73+_L`hE zlD~tL>G?;pyo7G0J*QN)NUF&<@eJmjYxP6`y;kPAxlKIDlw*PLQ7~GmNM6WnrBfs?f>cV-a z7PAB6ba>g(1Vk(1h_`b0xoOW7ju`QC`p!xWf1iU7{Q2jzZ#0oO}8jRmkPJJ#zbcI9P8*r@i|vjGT>~wynb} zm!-c)xnuucLWH~aIVbh?d3_*G`h7VIEft;hF-(S6_&UIp;Cg8I*bWrUZtV%$l7su@ z{n+Ou=DhWIJ9?~zr={uf>TF~4__*3P{I;wURs4c-=lvZ%obA=&`gyoIxbL0%Se@=KUY72Hg6V-%-s%KLr3*_qqp?Qg(fr|9&uOOGESXIiEUuJ&SltIddABiLUDdt$wFCRZ8+4 ziEJy7XzJ?{;KSqjE{3|B=^RRW@mhNJ7Zut|4~feYvbe{qRY@L?>Fc>j@muNWynbzlD*ztJqg4P z#=yBW2q2o~UBNQL7PBp`#nv)Q3xYSb@}2Ri|0*eSVbkmCSKi&xVMXm@-dvQjl%Vl& zb#rrr@D>a|$^4wC_Ik!0ZNRM+eD$4sy$grUW@=}ID0Fo3h+6LzDzwyybb&DYkXqhc zR6oNH?t;K0bmNc=GbZjMxpRB_EBX8V^O-yThbCcMYgINhUVf2A+nlazV^(K*zJN_MFPCw;o(Cjld3pltxpBMr^Y;Sa|xe z1hV10AZ{o5o%ICU4IV|>k2UIX#84EU4AR%Pdk*}sO$coXYZa9k7Jt}8yYP(D-X-D9 zDNaoU?#S#E-R}wB|1u3=0XtN`ls<6Iufv(gGw%h}T`H!@lgWx}>j|;Fq3^D^Eg- zL^URkUj)Yzq%k3_O53K-$|HumLX}cw7cGj{qd_drwzm06ihcf97*}omkS^s^gFZUF zjSU0>2pp}Cg78=n7Z+@J4A+WaZ>R|e91R#`=kgQkG{BI~ww?N}92zQvUtG}>b#FP^ zFryCrlFl7LY=MmYy=MH8p;_65`jQ`Es8Mg(a#E+ExS+Fc5-$wv3eA) zfW2X++?9aLoe-KDRsFw#JQs$uOHk()9SV>FA!`nXA`;jHHZAE%T{=39sgX#O2Im^b zTgP+sF@BgBNe+VPjcJ=^lrEth$$a7qWKwQCHkeP+@8uOAaH|_(0kPbX!SbWh6r7=66&1OZLT)7)WO3w6c!fhySTAU`S|^Qimn&=+AUpwU2DHs+g9MhT?TlE z8|xW7lwl?6dMXz0yXrE(iZhmY7+PRUftn_@`+yDHnVo>F<8-Z9qsGJcTU;>@59BpM^&@l3J~iwdnz*BrlrH^aqrJjPOtxMb6E|$%virV5Fo7}@9T9Ubgy}Q^@?-G!x`vJi( zYu!6R3|!YYE3lp-fb4_v3XtZj32lLLxd7==RmEUq6$`m3`-^CX0k0ue0Kc@2<+5r!jB&K^{_X0apd0M5 z712*yoAA!f>HG661KT=_jZ`YXZiE=Lxuj0YPI^kdY2Hd}7m_8*bZY4-wjBCitf7_E zw)%%R_R}|QF}$8W%A7Q}|A7Va%jZ|O^h$PgZoRou<}9rD!eVfIAiV(jus!bwvUz)? z!8i6!#n*o8m+3lXFZcW#Ede!;k*>Wss#QD84XjNMI3R~Uc&-&1y}D5|q*<{AQG?Mn z`9vysTFE6cy?SjHIir*?+p0MGqQwh|U^F*;nqNGQL_dO@<_g_l7&vG7tV4{Eei8>p zENsmc>HtA#Z1W$Z?aT?XUG_$HsrBrw32?u&F#-A9FMb`&al`19@wqEc6mLDKbX(Wb zPxkn5ACdh&fGye#`Ze(%#XHI$Uyu;L+>;2}%m2+WGzn59V*Ste(D1r)@c zlBB?0WJ|)saLgm!?TJ7`;z66Gg}eM)n5emKHXQPlh?QO%^;Zc%%;x~MNaDZL&b?$nCs*YW<|>T&Yy)Ug~LD~z};LNOQ!qF%T4UzFlytR zpiGBgM&G&)w+=_k;2dW{kO0a=qf2zWArY1Ds8kTQD<=?6#M;iEHwX~LDrN?*ph5-M z4jo41bRSwUry|ds3=2@B`cD&kF@NDqolWqcAxh=o!w3qr)wK~#A7tlVHt!2nyN{tq zA_qFmFQ(;^j1_@|$r{hGeIG0-OS+}WRP=onzve{2I7TXgcU~=^!dHqP-ln(M@}PNP zo_g+(Uhq^8IQ`LA4Q~GJjsT14Kdsu{k^QF+X0;_91sTkf;>*;XGhD5zCTE$ipz^X&qWOU3z&?oPD({#uK1DZT{vG7T z*!`*!I?(TiJ%sECsHipELS9}j=OOUn1IZ`eC2_d z%$#n+F~K0GpHUL;?Q^X?N}O`XQwGsOYdkRMp2(e-Z?R)|b$=^%m;CO$Fn*<#wAhip?E5y?B58o;D zNv}DG+n_1)VNq20U4dum0TMu$BBrDqkDapbR~bjd7A3SITf)$V;poaLcz z=K!`0{EN2;(FzhLIS}#?$($gSlhee8oYqFNktGv$Wyd*}q(R7X zcm*I8{@^3A?We1FgUqBZb38QHC+C5*j)GqQx`rI@{T*aV=;h{<%~Fzs)S;dA2uLPz z0Azft!FF9A>)XFhjjg#5OW~*+N`K!_o(o~j088U4Z+O_a18xl7t9RCcp6wA99C?s# zpCDKi{Lsg@*LLf-Ap%Vx+j?*tey9dc88i#GXYbZOQ5g8?&TKh9}K!kPEODZKCJOU)wp z6xc`q{7&!ldkvBqkqbi=Y@u0ppK2d3hs=DGhC1RbQ?P6UoTaz4)@vHNC|w zdbD>3zqk`~^8V3qnL)qseNwF|mjoJx38R!IVqX9XSZIxA%VA^yn(zZ*TW(P55zz{_^09#d~%;WDzyE~6%PDgZ1w5N34L6TAKJBZ84^Kh!d=!=vK>?b$dKMKcbQ`2_v{2x zI~vMA#|}T05pTlScGlF|n`nIXRllz*Zc~%Ty_)fn;9P6#t_Li=5~ePho*n+B?4m#& z#K8)%rIu$Qi{pE~U4Y;Mq}xAEX{%?tRI`Iw7g4E`?*Olos5c8K=-<^p1^)6H_uwD3 zWUsfgq@TPh_TEWMY%ZwDs*jv$4PV;k`#Pv zufz;P!=q4~1PJ z*;&H_-Mk>m)0rIm5uZc@IX|GrEyHM7JEyg0rF{jVh^XIFw+kpaaMFpj&x`Bs_P!o2 z+Fd^$_71@O+`c^Qym+zq{2{j|A8%JAp}0Y%+&#m17O%r_ z6l}`Q2cqLQnk&4+c6V=vw}Xl)tRDCG--^f3h4NOK`7raE>;787g`;hpAY0#m*ZVEj zn3(uM14U85a;T(u48B&B*!K7{zKyuLotkLlp5IhcD1OHls+9WR>(4ry%NLKtCs6Jl zRO+671OAZwjr(?GSPGM)Rm|w}23a)6%vfrhBn+YAYvB#vFxUy{=;jJQ?R~Cfxg>7} zUeEtTy>AtlZo#@%u5`~?TXb!1Qx{cVA^j8D9P4zMFR+J5x_P2hC7(~c63&!C2;}>- ze#`O|t?uWc@I+by3zb|wchn=KW_|n$afxnj{QV7!82azwU0*<4nb+9kzP{*yT%;!D zh>kE*kyd|9Vz#0gfg}Q&1(%_6sb&$i?tj`}9@%FfX@8dFjZtI*cu_%5T zl*e-M+sAi&5Y~2dc@;@pX0iJzbGkzEP*&Mo~^}r81bxRlk z&k0L0zdiE{@Lv@}+Qt0|5f}iV1snhX=YOdfw&pfYR>qEw|Fn!^Woz3-I)s1o?D_*G zPsGzD;(lbCL>R&ZU{WY=e2NsT?JX-=Y-=+s4gI#GJf+gBDd_w>S+zZ<_uQtr+PkmZ z20aG!(*&y&2Xx@&1p;+ij&e|iygtwF7(AkI=@8;HDP3d19pPkX3DTw=k~1kZ>>ONP zYyTDxzL9@CcRremCn4ZzfdYWB#aMiu9?#v_Eru-OILfmG)9+InpP!4W*OuMEDsr=l zw(;{!hpupxbrJp`Ch+E&f|qK1P`IJ8pybuKZrlz17|9y0 zLx*nF78_0gxSd|B&2#l|tS8qp-1bg1*scn@xtr0pk(*xU@H8DeX1GV*6+?v<@JK9G z^Yj~6xlM>BF?wBV){_m?mfU3*rjH#(!#j3WwYhLi(E3^1T-?Ct-(Vi<&Ce0 zL@Y9bYKar`v}DL@=HsEI_?;Ol6HYISK!Ib_Ck`SWjIB>zk`f^^bH@nqz?F;FUR!1| zQMuy;&hjKLGOkzS`u&F7sY0*&oShlBK%6+_EoX%)KaroOVfX5y8F8_eSc%3YYSKSZ zuK;QVmuuw$-4Nihr!#N2N%fhSlA0&m!dYR9 zYU)doG``w3Z_o?gtTV6Teu-neuQsWNI@F$XIerC6=7bz~ zHHsxbh}R%UljR~pk}uQ!B|m-)nBfnG+l&Iw=^Y-!`s(-drL3J(&KnA-AO!FFl@KXm=2OCgl$g0 zs;j4SmR;qFuMCjgh@min!-ru8G=qrQG6E&nP`Nk-6-7|VI%~%1Y0OKIQvv`vMJT%l zDU2+XsG4S!Yd5uNQd2A$a2WQRcJ%(_X(DrE>OUZQo67k7^uG3YEho-e zWL4_um!Ux?$d_4n@b_9I{%&z^$NA3x$%z$I}#Ers>w@<%zBrljwX* zS@RLDn$&8DGIs$cF2OPqFzlO-WYE5FhLgtMKDwm;a-fwVy%ILpggH?-u~ zJguWj3`kw{%<#>l0`5j=)NqDJd)*5T-){bf%i(F|wZZsuvHIPjbNiY4Y1hAddK1>} zsJZ5P*oCx-Jn|5P3Au`tK6(ymvx}gO8u6Ak6tvc!B?*sW6xJYX<|vY~fhQ0OMC;P6QqF zEDGHQdj71mMXadiZLrPFq~4Qhi0r(fxkO^wg0?jGxVm(H@?qZslPVXtX8f=W@UF`y z1h z@aDdN0{{r3|Hp>;U#|%VeSN2Yci=F`t?MS6!|u20T~bq$m2jrC?x)W7ba}Xe>Gw)Z zbG*&Aioz2aQi7af5E>D+W$bEiHzTkg2pL6vYisQXfdPOW9b;N9fkyR8IC$^1PqV&w z%8S9vn9;`O1i_fuk#B|%DU(yQV|Q`fPF#uNsXJMV=eX+S)fRSI$?M7UrLX(*$#bUf z3z{dbCZlZ``V#}1Rfo<*+UpJNA-*kdtOu0zlweBH)s`pkN8U(nEf{*P2zry=P=nv7 z`E3y>Ocqt~+LGGcn+c`SiMI*mi&&6LcXpio1}Kz@!jY!gJyFzS*z(To5Vz#iXKn!O zrHA25nzMg4?97=5Z$0p*5mP8Ne!lb?h z5g%t&9@DY>{xOJ!Vs63x z0k*&F8V!lLMMn5F&Da1lP&mXYRhR0tHUaW7kbp`;F`NY2eN|Bt&_Eh!p04DevM8io z_ypL1^yxpZ73G*P!cJ+Pz$Lv}4ou6X6V*?`ot)5+MH?ESF@77!P0*IAcCI>tb^;fL zmc@}wz>bcEyfFU=%%dLI5=z7aIE`=sn82PPNg!S4NqC^V>#jdgnYNeFlAM-@&s56O zM~EWpnoJquL;}x7EymQCNLA+vQgrZ1-fp>z0(Bus2}_0$1KKV8k+al>c#nQUXiZxn z=(#}M_P`3}2Zh+Ztw#-0D(XMP0v$?>Oj%Ado|L;nvy>?sk@BVVa?!mo!R_dQ=-kw< zRTFf}!8qMmvNs|VkVeQz7J6O=RZm3*Z@CFt+9@_TuIx|<_=A()*|Y>Y4LE}R742x%01f@91bxlJ?AXwrU-oQbqU%C@h?w~SyFW#R)%rW80 z^w=OVd83z2{x>9%?GKz)Q@xt^Z^p<_1XmQ=XvpwNj^(!D-*$Nq^{poSwuhD9i)>y{kL|?85M+4t;SMdywUJgL(dUOgsq!v>L8JBRgXt1`d@tsS+JQ zvnEg=x=vV#XT7@^JcLqJ0f!SKR^S%t>voV6naqXf{tW7uy(>HsseC`0s@OLqDg1Y= zN;!Rfu;7rP!XYYyB^_#iK=+GLj$%Uf2h_I;3yp>;#TyOE5#KQbzRp)eTv^TU~{ztl5$}3g{2JjeRC{=F8mSKCy0hU5xs~*w|RIjwQvxNmfa;5uRS%$0DQP5 z62L4V3}EVY9akg&7G*2Ro-uJ=+$s$e5mFi&o`u6lIo=6CF}&nYZ$O~-!K zimpyK7f#bNB$tiQs5`; zNg-Cs?PXkiV%H~a*Hi^s&%Q*?v?!)AFa25Dx4#MD+Pwb~e#ii`c4_i7s;nATm<@VQ zMp6&v$QGYfAe9)Yy8O2w$jqKdbfa79Yba}^>iA3(icA6kVb0e8QHG#u5PqM|k>P$p zG|G<){T&8tX~CPsR|TIT(4qv^nA>WI5~kLygC8@xc1%T->rw`q)%+Lk>UJ`#L`c%K zO##bCqjl~^=+yOlAoCj3$vgd7c2AP|cR=zki6F&3D!Re@?7`3&ejrwCx=HZZ>|rJ# zYT^k^P?K#(@rh4l6aaaFST!DPMp@mn?6IHa+(-2Limd}H;OAf*${nKrh-H)q2L4IR z>}tS-h`7-ojk`1iw@{OitXnr#*v?1!qv&r!cJM7>7*sNDPSs>EYBa5`={;fiS0G)G zgP-k>Z>%ITe_oAJt~wNwc(e2W(RGg9nMK>Sj&0kvZB?99{KmGe3M;m4+qO}$ZQI7p zxgTz8`?j49;L1 zP4f00ln19nDbiJ#S`-$k{sv_+$U@w^EnpV3#&v3%HT0!+1%vx@h`KMp3G)yEBo$rv z=9u?pN^KV5bTe^;v#=<~D@)>X;BJiiS16ug%}u2o<=7*?GCUgD_jgDTm@>uaLSpC- zb`M{x_Vl45nh~~iD-ez(I%__lX=P;uk^`c+B4ELT*noFmh1hNj27B}mDM&64&}VfF zx5XMBT*gH9A__+-xoRJo>f?8*OhuQEz}cHsO-|UGIwa9TwB=9707vzmdWQ&7n(vj| zIKf9#(ROyRS=Lj;mc_RTqq2R5!6 z;JzMSGOq`nJh!rnKwDH0$iOe9*(W*SS-vVc%5xj1!L7$RJjbt*mQf&C4r9*IGJYL> zIVK(FM3kBAOc(EzIB8O8;Fa>c5A4eQ2(EPHsM*E^3{#WKFYYyEV2rg>%pGBIe>1Tl z9yv257=Yrcu8z*yg7NhFQZd44tVn}@^4Jm`%}a+ zvi+;ExJqU;x`^ZQOKZoBhbMyqUxkPyji$Cl%CtST)9X=>b^Yc<1@mkbg49C6%lEUS z2R%W)gIDwxkh54h#`<2Vv2W>Us_VPpJB(6q=6Dmyw_aT(*y#94nEL z*Zgfrd%}m>BoJGX*-iQeUwzy8%D-nwU}SIOZV`?=zl8g^g&ekkCl#ol`0d>}2u8`TIHQF2q85`eWNlTO`y|j`Yueqt5?Q%5FEGNgx;>Hq~LgrOQ4?#hgEZvo^=ca(%mm?nsr>&HAB< zuAY925`tn?ImkYhYtYfofhu(;bu5%xC||7=-a9KkhlP z=!&V+0CRAX<>6u~3v%QYlIvE{TC|e;Okpg>stvc$^rqHzJ=?%5*n2VM(F*mMtlYAk z=B?uK>7<=%Sr)3YIU$|H1wHI*q$K?HDU( zHm&%Di#x-|?QI}u`|EV^!loznW8Z?gpvQ;#Q|aU8SRQbhaqbI0yz;%PFlbi-k1h** zXiHL0vWGY%A%?{En5(Ble6d*;J1=H~Y9tr^FK0q(Mo%$VxyL1XFI&h5+r`XG9Cqoi z8Eig#YsEb@6@hWqXPKS?SCa3g&go&dyb6Gb!xYCige`ER?rqf$&{OcmZR39btwG%U zq1M*Sja;UX=6!qX4NgfQ%bmDGNYR@7hsMP4C~xLVcaDn&F>=@kmVD1whU9Q;j?mR- zHCB{>uCO`Ejc)!(F*-2D1~!iNAbid2NEgA8+DIc1m~h<^w%` z_p41l1Tos)p!BCpRat-0#4Agk(D-R;|KAxDNUB|9 z1OWmH{ILmf|5tr%XJl$&ZSo_Lqm+KztuZ0_OxJLkB*57iV$D&|D-pI=qb4DV2HDJs z2vOikEs#Z9mjubJ+vQxwHVe^CBHFjTUtM)vPr2woK2*3$`4`2rjnKnL(f0i;!4|-O z{W;|3;9RLTykgjLG&rS7VEFqQq&Sd=yCbk01CKYVqB z#zbnb!fTGf<&H)@mX+qZamy}s2IVd2&>QnT<&n{XG)A!uvr@zp(+OB5k`6<0hZaTq zsD{v~5Vz>-vG)h3ZCXyoxnq%UlRuW|`}qKWS9WM4U+Ci!3FnkbnKVS#7S8pfB^pfn znkn0ncnoyX5cjyQmx7(*NB?2_HP1p#sce=Bw^H@&9UJ7)ecxi1tm1bv@tSPRYw+aI*g8pKOZ+I z&ODSD-t~vp!phWq5_T30opi^M)6VmZO*yEWPa?`%%+pnz?IGfGrQR2H2jQBLg$)4{ zr8lxM2BQro*%L>O=KKJD&1cBKp64sXD_Qv%ONpCc=htnO z*-9;0VOg%|c}I(a+o(GcCSIK|Tmc~Aan_!oXxlFp2Qzm3khi-Z#ixAPaK&ejdznP$ysfNNC?i^3dW z`L1cFCVU9(m%DOd9$l!cyQd=x*xx#$T87Fxo5+&G+K%MUfN3%V^`qy5g99>SF(|8u z@l~iRgh#9_N2SM(%ds%jVAMLfgn&pUTkguLZRBJm;UJ6=E|ADb_!UCl)MD0{U^o*# z+D(&BZGqbQbrOJL+r+pn%EeO8orwwvV z-({~ea2N@Wf%%llVaSGLf#NuP0lJeui*d|52p zSy1uVk3p#OeP-kK2loEo3Bh6+`jnKs2()r9$_s^c&cEIkey>BKEmWeKXtva7$Y+XU zAydAQcLph|(iiucchigu7b>e!TmK8Kxd_7NY#Y>XAv@Z+6c#PcuuLt0JZpM>S;-Fe zyY=U2j6^p-MORFW*S9}%vxlGp`HNYUX*6Nhdp2Pf^~*qr(=zt_zpuMcyb+g^>dz(twv^)7RHG%L>ASG3e0y;q;Q$D?`m zoZU%LBxXe}dtI~8W{o5i z$O``uJrYvl&BrPn;kYn8IEdZcW6bA0uBQk_NWI;d6WkABNYD@D5w>iV4~vWKk0P66 z)|R==+ac`Gnu1)z4V|FuJd9kCmdX$-C!B;K6IPW2u20aj6JEE48jmF(!Jwz2YA~K* zD>vo9LEV$?49fmJ>ZW)$V{x#i*NAYI5?@~4XD?mfmz}-clep`iZ}1_~XU~^uG_rQV z))16G+%GJ{iQCfo0Qq_kl<;wLq^Z8ap1%Ha`fn#}U3PB0%nIh{>Pz77LGz>)EXrcq z_jKy@Gl5mUtCSZU_nn+a#NNR9NHg}_9j0uqNf_puQ)VCErRApjxw2pk$r#1tw?7<; zP$WMY76Tb!&_QhkpCl+`K|dgWfCTwO7sK6jPWQ~s=dl=?;)~1#R9yx0CJfUaWf*;F z9LGy^kZg!ochf&VroJkW+k@H`(Ntx?K^gl^w{A}{9MK>8S)5l}(7l0~XKcT%_&Jo$ zQBlhvzI@z^4p1K%X)nSE5Hyowse&0A9tJJMUL%eK2`#}lrc4ZXYh_cTa$gZo3YVgZ5 zo?=OLOp>{5#cii0%JHL32|2??N-L|M_PK(Du2~S?H&g36k>)loFoi2#IYXWp@fkKT z-&XE!cz%U3APsNiU=dyyrWT8%WJUS>4##vPU6ey#^?ns0EHA&5I;eK6`cxr0!WZ``qdB~Y^DZxj@%#W(0J2R~8ARl~m}3f-7I?*V^d zBeqP*&|L;jld0@YGns-KtJgdcmhe?Mrita6uwGe`!nDNik;hNa&I-K?| z4?X&@iU~3Vj>U?L*wZmTfDA-eUEGSQK}DR@Da|Bq#I}Y}ti>@JDoEMBCJd5TJ`MOy z32l2kji>PKVyF7i%p^7|2)+T}2EwLjDx z6_1sm?n2fEx)B(<50P;Ic(;AtLn0*SIrwpD%)yy`F1^20p-!ec5e_J)kZ`EUvQF`g zRgt3}CL4=LNFOU1G)Gu>Kx61PeLcCd@YR~Jv}a)P2dp%p`THCAO=)Pp?Km)ilgEQT zAj zE;nM;qmSYwx8XslRFD^lth<16dbc5KvSgqZX2C=5N_bqjpw4jT7S;fA^jmN#C2>2{ zoy6c+Q{r86k>mDEkty^YgWzmkR8t`MMyv-&5TSD$>Rty{ep5JmQd-OVh<({HSzgu{ zRk-v3-a`g??Fh}#)-FibqP4Utm%7qOLv!;3u)|2dxd~M~7X?%5L|rzB12XBF1$h6G z#D8SE?N_C8Fj>OhEm-`y=_o-=b#OA`_POMxF^{OUv_9b6a(jOzEYlP*$|7OQ48sK$ zB+}h9?)8_95{s(O!VGC(fR=#OxPub^0WH%3BNDd=A!N?qlm|UN3TYy5B=qv3{GLG8 zCTrS1yM)cA9fgBIWSDp<8`fVY!)!l;sEEqt@VHZYr$Evozb<#%eNvIfwA_u=0%w>e zRga?!o5i|=3Ppg9JA@sxBC zvCX@%;x1_eUy5_`)0J}9IF5xCe`o=UB7VBP?uI4Erk1Re_TxAM9j&${UeJ_hYi$A( zrVZ@#rrwCOkDtLWYf783267{ccJM)K2QNJ@U)ZH~WggX^G&J3|h^o|*<(2hFpbV6kk0h*+l3f`J(k zYdKN^aRCk8aUp3EHcjqmcrfWKx&Ky86*fgs6EiVxDT_Sc$6-5lrgpG{8$87=qcJRm zi{+%hXAH;ZT7l++!V;cQdUoeuLLxDV>J|2qPasPz#orIHA$f=BkK-XkOHN8GiFN!p zB-+tg3le!1YUWlj(j_^leRw1hK?27NBPCiBs^2WCA+hwARl*Snh&OAjjk#NKqc1Y3 zseF@+tH>G|D98|p=pVlR3epdP*4sHGm>xRvoFcMr<><(*{!8ZFx8o)RkO zk|wbErM^?Yv6C#Z_`3@q5q*@7d)u;Jk}H>r=wZ+EHfJWpE(t*~#&|q9gNBsuzCsQq ze0}E*t9?}Qfx02tpv>hS(lnP;LLSk3e|=STzV>VpFUpMjM(8&A9R)J{;rFiSjd|k(DXCoLf~EOwLyWa z+n^O@_p#Rx;}Ccfw)RIF#^JCn&@n8K&RipZAo2)Us?~#&T|6(gy7?>#I<}uF;A>!# zzxkme@&Jo-5IAlD=e;U_!+lv9KRt{+tLZLSi=+=c1b%8(N{#4!IOTgzy7Hx?cz<`6 zeanpQggC9aKbX&DPCWFZEmcmuDeml{FNr3>%YXm*$?oXDtPyTG+#-lBE3-pt6>}#4 z8&_))h}c^Fyu@4ib!~kG>rB7CfjX1)!-LhYCVUw00Mp2!JzCRN>lFo;BTVFr^T{|a z^s4vS!fx@TTCU?1A^5%1OPqDhYt*TWT+-Hg_3|Xf-D1?hT2qrr)sT0c-l&yrpc@`* z%)U&2V;X!}l5jsD(&y(Yw;w&lepJRCbK!i3wV3wqrntG3qNMl;k+)XbEnr)ott+e< ziqc@;G$El2oraKY$(`6%Kvs>mYwIDR2KrGt0HoO`R0kZc8{u<~e*z(;rqy9-ciaji zu8&l{TJ_vubvP*Z9VmNE^ajBciWIMtfPmdXGHp9pn2t(d@!i(bfU=2EeRVq>)%kyguHMJW(SH3ObA)SHLN)3^v)n4t9_ z_v#B8;KAA{=J(0*N?`3SC`B!$B|0|*8TEz$D@Ad;sNOXisREJd;s_$pGyWa8a!OhB z03H$WqsJ_=I#k2SnyUkw2S1NSJsG%@u10sI?I11^s|I%;wU^4`~ylD#ylluzjKRe(&+ z8Y*8@s83}b6y_8g&8-xclx#X>&n(cCobhw%=gz)2b`z6n)uu1@Kk7-})qkAMpr!BZ zquqRN`eP8_YyU1FfK4)PzQL+*wPo%)enI}H29(PZ1`vP)0R^D^R}E<6VEv;3 zn>7C4LJ~EuHDz=r#7Z!>Q$)2@ogv`9xvG~=A=gCHR3;2Yiu?BLsO4I6td1@#%T_$7 zH~BgWF^896%sqRV*?FEi0{+e&_B*)M*MJU)%N(~c#mSQ9f5*2CVofSMU~}`DDUaHrf=9Z4U3qM@#Lk}Ut2SO`^>UsttajysV&Jm z!4aOtvn41>g8w!Cs*c*T*dVn#-eAg_4QH);c(59f7ftaAo0YtIXI7mQ$VXGkC04lq{H^0j6S0lIKlQ3mzg3${eVH^T%g{h*9h73D|1VYrZ zD6ua77Bonm;8PUCCer?5EuMi0H>wkD$c32~V^R3Ml|`bh7{?lg5y~b(am}0*hfUKn zW?`8-+9R}N+8nm}c#F+lfa;u2WXdHTM3Gp_Av%!o)(Mj#-nmU=h^ON>i|0z6>1p0B zg>gldt`vr;8Mn~oUhIrL4oDqcn0P+1!xnXdv&6r1juAu+-h&yM-7|4BZ<_v_Xf8F~ z#xZA0xEmcnTUp4oORr#%A|w_36d-ZTOW8Xw$!tDSiqNdBLVWT!>?UjQUez4^AK1lU zDex7=J|!hDS_v=C%vZUer{@CEh*@g`3d>M2A4}d2^+3ivkMDcPmUxw6ZBcGp5JbBC ztN`|Jf6-a{#x-)p64_s+ccvi;RtxWDs*M-FtnZ9_`UKiIW~i7B1H>Wp0RK$0zs&BR zYKXWk`F9Lb|Ni3%>S;6a5B)uYSU@s@-L`3+}5{~SJOv2tk`Eo$vm zfi5>gIy$d#m|rM8B2RBJOD_9IXa6V(f~?+=T}jIO(@V#Ts_8TMg(2=9Sig~el2YX(LyIO!&E$8T?;hfxL02mZb<7gqL+k6-M)=#PAOeA<#YUs*fm z=y){&1FAE6;MvuqcWg- zc5toH=^oIhw}R6#-#9Vhc006mP=Q9B-;}DE^0Cl*iYu9CeQQ~9n2jTw2(e4_+v^#) z`jB14sue9`wAvCoU&a?Eg>_s!x{nkmRkZV>OS3+6fr#rBrRrCjII!%ND>Po?W$DqJ zHRnPrXWuKK&aq=~xc4&8&?!8XNY(_PHi99~%}jYb*UKn&=#BqWI+k6l)&&YTRp-(U z^TO-WJHi`h>mrVO66KS=NI0?H_Z3XOQF+x}7DAK;A+BswE~_)$&P>~~moLY^OPz6q zrI`q|zCr%;QU!HQc&~#30-`1PFFT%&|MyCr#IbT)Z+`6X0T;;uE0j}ep65=(2$I7y zy9A^g*R?#Vs5ld$#IGun!TU9Jnm;I|urb!I;O6 z#%dDf8~M8i-3EQwt3Fgi`fCeXk|_*Ye;LdBgm3Fb7aFQsN7;FCI*E48#;^N|B2dsQ%shwmVQF!z+pkK@d%K<-eR_SAXj! zn61~@(;3V6lA1?U==y6P#f5?_d(jxAk1h-vH0K6wn*5sLW^a`%wAL=5BlTqgO!y$E zEJki>A#u<`#41(e1N;|w@x42Rr`!#LeroOvnC}3r|0vk?Y$T~i&eU+gM95#nhE(4+ zrNkJEgF3W2qG(zBF-asjIs2evyv3esZiaLsL^qxSWSD_jsE0lvpM$4Z*#U!+`CAb& zYO41KX^H|$YV$V0w&SGazZouaf+LXkAV4wLq-p3wFS`-+rAx=WpH7KU#(5!av2zft zGj?Aquc+|C|OGuxTG$gLph*9@*{`(Ie%pD5~}2WK~pI%SM*RG1RNT6S!H)(->>2gP?q1 zEcmQyG=Rsy`X4d%Ay>5MWaXOhpylLoI`uZC+3U#s{Kx)9hvCV?>rtT;RslU-B(6_` z9P>Qf)y}Cg%EUWZyOFu^#hZISKieLNR1S6=+n|b>2y@%!P>g>}sHSxi~`Ibbd0H-qvW ziXL z`lNmeiaJ9TJZIo%dq~4$HwLwJQny9(4C>y1w623c(9v3GVd}N!&8yw5%h)Xw@k~m~ zX8?|wg?|#q0;^4bgI1Qct=rB86P@7L+0ySGK*E+i&wWKG8(_qytLjjPCep4Wo(FCv zp~@$jU~A%cH~9>x+?E&s_Px6{I-f4sH5ghh23593hixG$rd!I1w~6rzAIlKD|A0Ya85?wwbx!>{7%Aq|R6AV1A`b ztFkMvJ*F9zm!(k(1KWv7Ngufp5)}~f7CdZ)>PJHO63VU|dadL8y1BMDJ^$fBp5v*| z4OjApbjmPIGd=jsSkV{9o_L9;R-8fdN`Ef*4o|rmN3 z9mBj)2J{~h--sBpC6c+Z_Rf*xUX&rN%hFi<#Sr1J?BF1z529wu4xyL%S{K&y& zmm}&|uJ$0*HVQa2uw4Jr6Yoe@q!%bS&}}+xwy!zd5FP7!+SxchQgLEr3I6`Id0pm- ze*f*YcwOYNbo5kypgiW=QNR&}w&>8zw0JaGuI;K&3`2imOcO(8ldi;j`pNzn8?N-e z^@;wNb7t`#`P27OKC`Z0wu~z9m{(37=$=*Z6Ys7)=Q4Ta=E;SA3Xt-XBIQ1h7!Y$1 z)0Ka`<;nu-YUSF6x1@xPW{6j(1TLcX@qGF?NUIB~JbdX8Z^91g_`zUc1ry&O-B%xW z0r%~}NNynZvL779WogUG0zK{(I_$jV&OJV1$a^f%h88eIULRck8kOlto>9qEw|$vr zcpl^5+0BwvBOGoI+wGxs^yprXt@V*plZPh3h!9y7Wvf`n1ng#JDou-uKFFG4@l>s|oo_e;WRCK$~3M z61UfqA>ohrzR^m5hcCC^+x_P&$05M2UKaugC<*z$plzK0r`{whMav0?6V303Z9^kv z))zcPZd+U=#b?W)tiT#ZDL*gkk5X3w9_3aPRXDW6^?Aw9Gv}COHh0&b&%fPvqD_`Q zX&TsDqCE*i5gF~&Y%D!e1lW5M49S7$cDFf{q(F}|MNmk&7L3mf2Z)=|xsn8mCDRU> zIk6H=pZ}T@u&Q5J5n~Kns(VR7om!h9W!La>W@&O5#6f8*2-VU?-!3(r#_FxEmP|Hh3zpZ zCdZ!=q;na&y*du#Hj1kWk+kD9xI@ssYz)(bl)7{wBTS6Oi z5y69JnXEG%2Hzo@yz#6|dE%3Fc`w1g4%2EZxsF_>vP$1BvmA<2g@4lIbX#@AL=rbl zAfLz#Dl{S6liwEO)RXR0m4-G(G11mh;zr)0PV!B9+m#rd`lhOHE(6KMEf==BjJfi{ zuYZfFWvUuJUqVLWXuBXmNo1_jk6a+2t=B0?85}2wb7?FQzHOsprqUVKWRKc9XDdLq z;Vqt@I^UP2o{z@{>QHy6%@43c&gxOB43@|EA}RNGPfQp7V8z5-kvM9vb5ZHQ%MrG2^~J z;wp8K0)Ju?HMl8B*kS{hLBSnaupANd(=<@%t=`lUWmCCWRBv|%+=di8L{pcI!dT>1 zwWCzg)DS9tsuH+Mb;y&Ki%S4&s}uIX*PsWR3ig|_*S`szwzS*z1!OkBldWde2T6Wb zN%QG7YRgyP9I@T2Ww9Af_coGxR*AaP%P^b7`(f;2L&C%Kaavh@*Gx;AZb9xy--asWb8c)tL`UzMZ~U^U~s>(KEBm}g*_)f5HPEVY*> zZu~Z*6To-jH5{jo7oE=%*H0C8N(0;MM_P+1;AwNelx|TsIH3ou9c)gHJ zb#Motaso-){2~y7WYQR#ZLLvVA4@r&S_P(c=xgA9Q8HmSKt}s{Fyovg-7## z!fnl(#FeNCY~Q(2s&JWdb=|V8m)fWvaFbBU=#Vb!vKF7Ag^@np@$>h4t_vkC$O1+NJ4@jm;ob-2eOfyn@vlm=U!`%uG~BZ8gp z>lK`K9VoI`FLWpuD;{@GX&+yAL-PsO+u4fZ!*sU+NaGv+&706iumqG-+#Q;C?%u~! zWJmW(R`^Lr*s!Y$G6Ck&lV@CZD~USg_DHV=Jf`?13NBbWaz@z=(0wunRYU=|*YmGX>-RiSA7nvGccuFum=PLy zFF;^7h=dAzSiqlLuqYH5+0)^VGwhNDLP+)_s90@IW?T>b_FT>cJSFr}NPoN4Mt;wh z@S*S-7qwg)QW+h*9-r|qs(h-%7bqnc4D%^Om<>S_h7zZuvS?l4#$c3CAJ_j zl;Q^?L>UEAwn2KiK(fGy2i}lYD%UkTbpC2ZGDcy#Wx1)tLX1lO%1@=CDtx=nbSXm< z)Imq-3<}T4F%g{rlPn5C>jp5tWe%3*sBFDC)JlfePSqfRK5x;eaM{xMv4d221!$ac zDJXq+idx?{K^j%qB=f!@rHk{u?w)Q(++s53sOmbSzuhkjc6;7Gx6}{{@>3f+K^I__ z-?`TT1xsfFwQi6-8L!W^`V|h&bH!$HAs*5Bf5kS&0Wr3_wdEv)rhu|b||Cf&KEerH=z z4Fnw~Y{b*ZJun^Mf={zYwa(rHeu==|A!DIZ%SVP}LL3oFQD+)f%K}F z-xTgRe(Eso+_XZUUE|3LVub9PPx${N29?j<}95qearZ zVDtJ?wwRPuW1-M?_8C*uK1h$bJ{R1h<6XV+b$a^#rC+-ol720+S+^q93*^7zmbCF{ zCog^hPnGxRhS_RgIGUgSzH<2_Psr!F;wva*eZ?KPwYdeHYu~wi<${!iMxq_{RDz7@ zSE>tMG*hpj_KSF zKKC!E5Y0}-w%w)odmSPCQ!Yk4186{>Z@8 zY>6vG0~m@VIT$wN8kXoMX-M9R5Zu&iSU)}$)L<{)JQ1Ny!g*xlh)AJRr7a7oag#uk zy{lOMqiD2YDyv9d?=8iQaQUBT4JJV#n1~K?d%G?Nck&@tv#Cx=$yON%672-MGzcVP zWX4mrCWWh~b(v>I?CACyu5F>ufs$ZG%{89+s(C5vg@-XY0x`)ozud|A_q`vPb5;ze z5|$jo7Lc5{A`b8Dn!>L)yH{*FN^UbALS_^Ce&JCvglgGdw|1QUy4)lXTRWsy_~{16 zb6ygTp-k9eiQwP&=}QNJ@c;dzzX|UON-)ewTB*cB`^(n*GF`0k2jrx43q7 zoBV4BY-uh`7z=EOwanBU^Q2%2f7HBSDW6!zAw%>+wpt?tPTt!4S6}H2|9J?eoMg>w z^niuWgqA7+By$ODNtR1y6=+Vx&?`2v3W0*5eJSu_k43@7l72|!5IFTfu{r-B1($Y0 zS8<7d)NdjNR8*tL?hx5yIjpOQbv0$8eDse_?pN;RaKa*4F|mf=nWeb^zN#dfR!>Kj zMx7*ux1BI;w(-Y(p(-{duPHlDNL~}pQrw(2yg`KoGHy$B1iYfdd1&n}FaA|ZI>V(;zi4jViiS*fvET6S3VtH1}%=nSSmI&%4e^eJ=`y7ZgY7ADD#O+x_E0~Lo zi}WgWf6v!ViEU*+#r(JNOaZqO-W?OuDDfjM?izjSW>1TJ=P@^5o$ zR=eZL7DqW!yY8@NjH6Fk(z-d%ST(!S1n4h_KeFUol9(wnmFEVl!?<<^PM~(39_=p4 zP3u}s)>iS8q&)hy(3{(Q+Wg+Fo;UiBC$ncSyJybdDP3M2+XCJ~C)8>(;M64$o&oo~PW zfW&cLS~WBh4*Dzhs7H(}ek&%&{eyM?NAmPgpf#NG=dnkjBS&?q3oag8Az@ri8vGd1 z0bqGCyXi7)ohXm5PRJl6hE)Rl3_`5r`W^G_@lvu^tkiwXImYps-uzf@=hlvSL`%*yrEd z#dILE23&lZbWXo@(dcr2AZ4BRSq8I9%f`Kry%=BXE;n<_5Q z5z?F5yM?|9U_p$q$tBTlS-I7KRL~X|P~ynSq3a^*1A+x-c77B{-PxYWh#Ag4q$+j^ z>bOb$3xaTkDU^X9w`Y8Zbi+wQs+z-=;!t(A>WNvR;(wf7KK9IKsG5o+n=-Z7NmbV8 zOdp?tM5-Y>N`!_!!y2CUi!$L&DD+6v!+uO;Pz^m>g1X~jH6T|u07~*?y!W2%fRu3q zY#+9ipg6QkC;%Lp=Ae6mjV(s30ADvB&jf@I}RMxpa zT<2V70&rQZn)6xk(|8jlk6dCwm)(LcZ)5r`Iy{u^5ni*T9N-86e3=U7sk~lO0Rv0} z0}n(v4OGNizYTg+642E^gP>5&v-PZ!6Yrn8)w{;7$_j;C-Ap zAQo++rS;Ss8SAHo5C6CYf0H56{b{HhpfAfP86JY}W94rEqIQ*{rE^!|{1jn1dDN zOt;FT4Coc$?FAqEy$T1);*3WqD1nP$pakHDO75F>UpWYLMF`N zvVi=EL~YHLok_ktWnp^>1;C*n4Z2taTO9DKU-hrxV=b-7z0FW0F$+~WQHf7K(7)NE zkJNn|TB9RH@GF^<9^NjVrfysmttCdc#+-x&&mR+^Lf(5|U~~3%%f0N02mC?NaB0@GlOiV$Br2fr zqm8R=yUog9=up+a?);sM^7vjqiZ~0kIw5lqIs6YOi6MviLo0$k>5Pm;>CjH*)~)vp z?2NVs)U5Fa5-V5YflxQ~$o~6wb6rzWqL@(l=0GOS_Ef13? z&Ge(YnD*8kxH`S$92h}Zgbw~fgr{wrZPO95|j58skGFWn}weN^6&R%%96An&{T zjdPxO&ptSu$np`xAc}qzy?4$~J?FByt6++x;kdkc>!S@s1JVYrq3PKGq*wFI5x)ZC zIH5mvA`y;ARJ10@8_+YzN`1=)nRGwnqQYGUR2mD87!XP@fZ~X7KYIWb$|7*kG@XHj zq4_jl1cU>D*qFu$i}#99eNKZhM82d52gwX%pEw3(SsTI*TLob{-GT5X1F3YqG=Jux z^(I?jP=7N5q(l#7IcODajw5b>;zWQH1gjs!tF>|&74Uh#N>|WF?n@naxssDWFVV(s(*-nHrz=@>9QbTjb(R8d|_x3D`L_Ro@~fN#v&<690vz?WGG&yF4NsGP`}a-4g}>D3v-B> zA$(tb!HK=VybEG#7`|Iu86S|r?v=s@&p-0%n;O)P*Fw|-m3lE*qKu# zAvN;YsNxO^(?Xl~K{0&7gYOdj&`Es942sZ6&}dBfQ*zjYeLvrkSM=9(iph&pJf0<_ z{X~NSB-CW^=1GajyN`UAqHl{U7hOsN0iuWQTfE0Hk*@Vh*7waGkXtHFhjI?97DDpI z*UBY6qSvQ`SIf3|(lFoFRn<7-3=~m}8ae!C6o$n0oK|rvSRyzZ|H~q1khb8Xc zoF!2%uEQFL#N7w|Z-u}o-R%7&J2^qm{CPr*c)>awN4!kkFJqLbX|$FSy6STpzVYUVc|8Ypt&Gc38#L{;MJnKrCX;fglSQPPRjyWNzYq#x+tESQ zc9KX#0DIG9qxV1dtsU`X;%2`t*e(| z9N?Pl?;9Narm%6&TIKR--8VS0i^T=um(82<8QCt6ZIBnnuD^Osq{c-gD~9vGDKi1qO;r?@g=L-fWjT@>E&i zO*yhzanj`?Eu6`zb%9EgmNqDsDCqLOaH@1#u;SxqJ<2D|{%w?;KCY4mXO*G(YHX>HARf&L%7EK1r-&!->^EA|O9JR17Rd1hJ zNh}rAF2D}3VXw@$&^#4Y^J9~5rIDjJXLUq3)&dSFf$ZKRiXMT$NOU0m@}OmbNzG3!ZOQm5F;)E~)jXr%$Qv zQevwXc3xteVA%qP#Y4IWLFxhUvvqzEzU4pD5=z>{a!^|Svc#5)w&yXX&ZyuW8%U~d zz0&yjJp3{Q7!QRQL_`FHnXJ&$zyS~QH+1S(9PYJ^)zRP2bIx4#K@rG+oe5N;)jBB6 znEF%|q=pw9>Mr5ypOIsy+|3Nh?zmtKU*k^)K^6gdEatRvVGMGS4B5}@(+mjp(+Q|H z%`Gl0K_!j5$FCe{5DrJhX{^6?)&ZJipEWZDZ`XEk?`-ra)R@SnAz6MeF9Ba3g8J1??Xkf2&Dn$Go)+>)Hrbf%2t1ho6OzSb_CTNpx<1+%Mf-|FR zBbxmk1*_)3Tv4I-3_+6IjX9|V19kLc(ahax=|RsRj+SNE%~kJ(neLQr*Nps@+76Mt*Qx~ z&ZiQ`cq>(P(i1nmd{P#j0BYGdy`hahsyVByf4nh`wIIF5QqqaG{1ZINN83_z(MS72 zQfnq!>7fQ>0_|IhKUJg+x)($GOLg?in;6=ENNJz90c1t#d|QN2l@>^f4HZ0xdeH#C zMgvU6pt=cG!wgap98d?_Yl4WJZ9zO^j;si-*@Ohuc408Dl)HqH3sc+2@dJeG)k0`! z9 zV_dZ~K*y$QwKNyTwJNU9vmk(GhKDpTDiVfn$1)O*gPEpO zq(k-puyszsnMK(Yp!n$ zlE6>Vz@%O^H^`;S3@q$kDU8{8ZRk%kR|D{tt|fm%Q-$F6?epFAVikQgR37yxno{E# z%r+{v2%@J5kCoXCx^v5o-sGod!TX`IDTVECCa+C&SzFvq*7OZFS=(f^b>^yITPvp- zI8m!eCH7!hGP4FoHUHMi-K+!0Du@j!JSh;ePoa&4%6?^F(Kxl?Xy^Sm&=PE^a}#v~FLpaw#pr6~Iq#G8Bg2!d=MRu3bV2=sIHc{ZxG_pv_=xT0*twa=q2s0jM}HuxjO%b%Lk+K!&5M7uZ} z`$E_I*=MXaj4sc_8hRDF1u;IVihio~bU1FNDXITfoLJHmGJk6yQK2vE_{8oyNnaM@ zS=h*!$KQx@RO^+2cg?5%L@9k9*gd%mb!IYGOmTX_JL{ft-g2Gdw_9=q5L?9u)UUnJd37Hx( zL#KEk4N@CG)Na^f4iTdPX2kdr*Xm|-M{!m3v%THV*Mb}UK-Yr%i4Lj2EuYGEAFnjz zDaTpMUrUyx_VPgSB`Tevj5_4&*&{O{uI3_da_J|9^fvC28+QorUg*&`qDnwBH7bzd zN=tntaYs!U+Xd-~P>(j|AKg!T>4;)BplnJ z5n59&;$D9-iUh@7nQ$5ztNVOa$OU; zwK@Nm6OQRjvY@o(hocawIzE*Ze%LYbx44{Hb>VcHZAH-Ti`K*+M8x1tXLC%e929l) zTyNnKa1SmxyB~*1A)W+5pH%g0&*M9G`EX@m&9jnoxLk8Wu4VXmY>2loE}S{O3XmQU zdKHTA6ODZ#9>(u0Cbv5Q0^^BjKeFn)qOVG9@RAlCdsNwO?>z7TJ<0QZMPlP=?Gkr1 z)JKqprS~VyXlqy;)6A|EmI4d%nf1)P@|0-QtZws0kfa1|h&1|D&CstO+4N6XBpm^r za9dycw(rAw1ccRLghX%E`oi+oU^&l35qQjv!R&#E<e1~Q&?o6HbB42(Ta$YW?3oI(vL8K0B;5r#2bljQ*@6fh<6AObA3l}8{Q{f^I{a6u2LmsQ2Q~NkvK<+J%ez13?gmMMetMbPq_`CS;vjF1Vh`S(?Ii%+giV)6>4l{!!TB-BA=l4 zvl#YKsUREYx(G4nyG`9e?njC3P)tW}>!}}-*;D%zkQP^~OTRGmsi~=S;?avz=N~G1 z3Y6YR31l z60yZ94vVB1KFunG6CWuFkna;2NDw&1-mo{GdfD0)+JjbHS3UHXkK-X~6jK*cWN%_O z3p#c>>_bDB@frn_N{=r2%9zC{X|w0xgZ8&T#6AB4kVQTMb4>|TMOu*hF|LFb={`=U zqALXFzR@Rt2d?*Ms~qO;0*3m83*3Xn=n*h_;5R_EZ!+^v9rnn1XjzEPhreiB5!cf( z@h*SbHiQu#OS93@u4xz3Wi9?H*{rku#`l`SETc++VyWQq0O2bp zbnM6;*qT--ev>-eS{xMf0o2xw>du=Ux%dDWy!Y01L;wO5^&w*QD=zQVW`@rA`nUs5 zlFne^2nn8mmsH)xilkgBybbDE8I1YxVJT9_bBYXh<1W}}S;vi2;(;4W3T)hf4;>0D z_-|^jk@Xv)zKi}`TdO;lb9(QZ2|KDym-8E^8rm%y?~VMafU2YU=-nX*yQ?XG12pZn zCVL3YztL86D&f)Pw-=8YvktgtVNo@*e_;5ea7V4UR$6mZLcEwXPR#tTkNbLAYPW0E z-3D?Ug8DW|ga;|ze^-?`4Jm-Pv{{GXwhp0T%`_yORAA1i{}XP{Y@l)`S(nJw%yEUS ztTFg_mcT#O?q97>_i1WKiST{`-JroyF;sCXHY9@7&Mn zv#5<<)JRqW^;({h2IwLl<8yRa&wF1vci;_UcTW1o!6-*CWu#U0p3- zq^4MmZmu)};##Pp^Xk*XltsJ&A&u?tJZ)+5>Gd@+7xi zR9o@-)g8qWY_B0cHXtr9d|X!K$qSSl>9~{#GD(YX`R(AYP?!JQagDs@-(hu{cX{Me zC4pZB&XMz30mxFa?U!jP0cT{>5Dbh}x@v}2#G17dP%V`24Fs);&MJpCDjo^TuKQVI`vmmsT`#lU)r=`GLr9@Vq z0>y-zd7D~Eqi>_SfP<(0iUe1#=Zad0>xpe>{0R~@YSo5bg;98hy9nlw8ZM>l2SEQI zrt?qFRYEz9hw}%1PTRxnn>SESPi&2T^nlm#Q<6?>hFRJW`kS$BESWCT;N-~y;MMQz zo;QLDSZ16nT+Oz8uA$_<((0M;k*0doueKee#NPG&8gfh6cHTF|ZZ!(@(Gz^)%SSWY z;hqpnq}L4eajo~*NnRbtP8IX6xr4_Po8rgk<@oOQhT`Yppn`p78$i`lySjPRYZehgHd*-Np#uTpKQoDDFg9mZ^O)+Jl!!S(Y zi!iuehhrNZa6wYwH%RKYQq|I(GWRb=J14}5Q!uke|4wEoS(F81Z|`Rnf2r)0w67yk zZ4w39D`+VZSdNhoXx)w{E1+?VKdl4!O!!OmG{8##l^UK@wCiK38bN|aA_<=v{Q=53&wNYHR{;0MXmF{|+TE@oPK zXzWh>Us9i?tBOQ!`op=QylWLC^da)$l$8nD1|jLlP_0&F2Sc30gy+KQq*Rk9g2-D1 zgo170_VTfkCM^5m*hh8P9V&B3X7O$6grRX-l~G35(Y{Ssy+aSNXw*S5$Noh+P_M<_ z=NJ;F!dV;UO3%;cnbPoz`m_l?SVa#|@A_@&exe(A9Maf3t;r8Y5WMh#U>(w@L;90J zFyeesw-V7#Vc&&0XN|3~11Xq1oeI#HQjF{|zH~{BjFbIv@d5NH27EmLzlb&}f>+~8 zHZ&84Bx=I2GW4)c6QPhoHMu;sgWdwc;2FtNGOj4?Y9^kz^8P*+i2y_t1;WF9)>f!t zQH%PQku?EabBdUk9T_mG zxcC!+SvHbzCe(vrRvvVR$zjaDO*W!6;8Eo3A1bzqfDi~w#TP*;UlWs&aYskf7El71juy|xp zb{T106<)OJ1sZ!e2J5@ZIIrU^tKeW8fjlyjz3nh1gCgxeLj$WbxLf9ffzV;?R6#WvGZ8cg$IojkRP z=;%gEA5(_Fj%uV^qCtYr5=d33Pl;zXSEK})tgmC77#_Z~B7&52FaUv3C}N2HaQ4ri zG@u))=!_GY0~G<$uNTXec-jS5u%M*lx2OQ_Od3SxjF1%fj0*;PNEmQL7JJ6=QXubC zK!jJu`q0-5;_CM!x%XtAUC4_%U5}LL314OLh7L^ZCkLD|T1p8JEgktYMjL4~Wcy%R zX#|1Pj0B5_Wd}_?Dcg7MXZz2GA=1I-PCKH@A7zg?RR$X`NfKp{4n!pdSwZxf%h~(U zP@)|n37;}98s^SO{l&=d0<&RgvA~qT-B;(KS<4 z#RSdtLx=vYg6csxc}mNL#CPIEr_`4>Y=>YPv%d;Qq21;wTb4!&a&J{EWW*MI1aZg% zi*_UxW;O&oB*dwerR}Kbfy6u2^a`!yg#v($CEfjq{^=@`eh z)`VRa7mzL6$fMdY5HR$>GL~qK3H-38w6;OBc&nc&PiA=BJ4%7;d%%#XqDQ9~H8W6$ zP!@~~p~?-wLOaCwVU3eP4Yb<82%L52f+|phTvB@rphq7~#jHqym5L042w!gu8QDY~ zOfcrc?Y;1?r#2{H%t1`LnFP6Bq}ZEUii7M-EiD0q0O3@XY@qf*2pp@e{}hInt$`!N z%Vmt4j>lD*LCsPl~*aN2yro*R+lsV$V`)Mwg>UW z5Z{3;Oh+bVh)i*t%?OL9Repc24<%BN*JC3hH`l?A-y)#q{I!ksFQ^E;iRi#B$S2@#>av2zw zuv6R{CrI#GaDw%-EG9y&u3{NGt)&3+*!L1>3TGrxv+YS@%?Z$FVSzw#Bw7d#R|I&4 zo-z^RW?;~zkC;tEdSenlX#nw*-6Ts32y0k-7$q?G67#@df8;cO&D)22!Y$#&fy$83 z`s;h7W9y9eIp+;OgTtZ!qK270>BGb`11Oq&V`{eOnHal`1t?Nc@@`~lPFwsQTbBps zv_Ab3^~tp(i+EJ>H%%2A9TJ!|G(O^MJZvX${axydU-mHXNH;41?0fP^Jz_NMX6l|H zbc-ot0({DWDo6;2OE%qk2<(&t|Wx?Mt3-Z%U*POruKBicV;jz}|6AIN16S ztL5S48cYf?p%{Cch z$-acXdc0gx`=_BZ~52RGv2$d}~0p0cu7HS6H@lIndAXN`E&o0A_ z68ru~`-F6jK~V6Oh_HEWQfZL*I|9n}4hCP7#%9$4wp<-l`;+q ziz#@0wl3CeiX#%u^@Nndv5L${QyzpSBU;>ntWa;EF{f;75AdW=A_nV``1IYLo;^od z#mNNQCGQR$0g1zE)4cIQ?*6wtv?>uWN~4^F!%)oBAc!2^gEbGq7ujL^v%U^k1k-iNo=yscw z>{oO*S|_n?v`tR^A^O<}Kg_OuFXN@c?GS^JMn3_5`w-@>^BEiq2{3jbda^Gcynw>5 zZIe?U@B7n|t)9=P1NV;k8cGA$`%8!3sJzoBnGp+GaF;kZVQk#B;?K0`FanCptVyR%OCk@+qK>=G9 z845F`57ljLq3x)k=4$2Xgyu@*^M0!zi`Ii38tGUq4Vwp+6Q4psxyQZv;bxK+K)-R+ z{YXUcl46gtCA%iKnn5e4+CJh=MLaOewkeF?UuPRTFGR2jKnunJ4Ij5-eS9X@@_)O1 zKkY6B&ad09SLVi!7PrTRN&c`stb=`~Y&Cs1S3HojbI<(pU}A)=@wPr(vP&=2rHH)R z^sO!X4wP9s`!e8OfLtp>8QNUT4o@9pC6!5SA|2nGdo(ape?fsv)J9=}TRK(63)8SL zG{TeHZ>d0B2p7kIx6c$9HH<4|`U%&oWF`01lwA2x`eMk8;MtG|Vwu#ok|)5s2U71O z4Qk`>x}BWchO#=H9vt`=t*RuPy+fpovDTJobni9oj zM39h>PhpGpVg=I=i~z|rr04k@$|S;qq|lSVZ@5rpT$<6l={1tn zXoL1a&~?<#F%}*>ePVTi#5;Sz;pZ++ilOGffnsILqexI0uLiZ-hEM$nLR=5rNcXhTua!A=@#2#CA_FSNKHVN9){L{y3EgxShM zd#HkZXJR9%e1LG=YnC5b3d#d#IG%18Dhl=yi>!Qo;cPE*n>W)um8lY(s#dHE0NlSY zUW{0Wq1cWz>$s+~~a&OuS-u^f5(VLoXUsD%o% zN|)ce1P|u?Z+tj)PR0l&4cd5gMw5qBkV-U!oO1T&+`@=!Rgi}O{{q%3sbU$THenqc z6_Z83EfCSOrfKd3?HXk~1vgRo$$gsY6>SXRqCv6rc@=ZYyU;6-iT!PbQd)3*RTN*X zednfKl%AUE({ScyE!VxHBc~~ImUX;yeKKT*nq^15kIVy@t6pl!=}xJ3yoRbQCL5;^ z30$5Du*%I>JIKJ+;h$5>HMIXmDKlpgF1Gx`DBJ$i4PpMTqQl10`agAty0-EL2iiZ3 z@+_3uKG;F+qzl2kA(q4dn=%f-tp(AMtzcDZc6U`Q9@@tJrR65hN79HhvFo)DQ}TrS zUrhBB%Xm}Bp@@D0J-Wz&_+6&W;TOqsF2%Q1tCrWXR+UI(5Q(RJ9Et$(E=>}RUVhl% zL==r>cOwDm8Lat)_nLU%J$hDl=BJO3m(x4Kls?O_tssb`lo}gKVt!=J_DBZz%9tTB z#;nFHU%Lzc;loNnpTi{rYjmtsKk;581J;sJ1Ak>#((l;H#nNcLIn1V2G2?oWFvsR}>VLBhCt_5OhbHJ2 zO}x=4iR3{xCo)L=wb6o^`tU*rv+=41&tc-*azN7dCc+IJkHf^J*P)l+fv(5Ezve{U z=cIhIZ|ih9L)RA{v2$)}wP_CMdRw%vaw&aE0&U9d6i)UQbGBg1ucS(iIxGyS6sLhr z@BERKjc(p70hH@oFn=@UL($a#;gxhYKvYFV>75LMz+bD16_Oa4BmibgRL~Ij>;$#o z;$F9ORh@uq$*(G2jS9_-Zb#K>iR-DJF4=@f3FoFrMzm)|8A&d-(&+%PRxNowp`Fv6Ypc|mjeC#_*oi5AgV*(pW7^rYY9f+l{RDW+au?9LnP(0U_HuQ@;V_gp28vc8=yx;hl4h?{_G>EQIE0)}xS?d;BuVdzSF~t;!2Ax?t{5B=$G z$lR>{XD-uhQWW-zL0GW@hvIxb?@1I%+{Grx4w#^i`t%BhFKdD~j}1xi?TT)umCHT; z6H<@dsSF>j&<}cYnICgh6;h(8h}QUjDp5P${EdI}U%NNpcefc*bV_JI0I#9iaD0UV zM49p2nnWu3)8}=zmL|q3L|XR3(*Z3bCLYRCA0ivNMZ&u)fI9ykQIJdp-LbG!B`Gc+ z+ksDkL(+4vd1Y#C%@`uX-Y+CRKc%!y`US9TC(qF5g(h#C=+tg7V_uF1DjG*yUc(hw zukM)078pPsUI;%ozOL6K?;p;D9!yLub96TCe4|ERBK}Z6e}z zM_d81KOv-@))sctu?}|I)Mye7(P-0;ofX+>q>U`{SYP|Zy=*9Q&?^1?H>g@k!h3^d zo|DV|hWqScpC~t`&wi;vaWGd<`vg9{Tt4KI44FpYtoRdmWAXQm2L8^CtovW+2Zy(Y zC-(oJepK}snY;bhV6FVy=rI1*Ug!27sag@%hSNstldish(gqosw95XKYV?&Xo_D$} zin%dGT3M;rKju)Z0uZQQvI_GDt?z{}zv0e2!38Ifq$FCFtT}Hu6v#=4GrvvC1}(^+ zGp}ZY{!>E3e=MEbyXDodoaY~{=UvYJxH6T)k(PfbH`53lI&fJDp&^cK#R4DF}LQ&^s#$L}>MH8mjOk$^Fy zWs*1@nN|ZwM1Y)Sa_TQ=+Mgnox@u&$dr(xeJ?mzWUFNpl2E9CFXr1rj<*-eX8Ybkl z$?~$o_jo$Lvm`~GNv}iG=2V6ytF6sh(zZ6D5tBoIw<7cv+!n1WK%Q2D)y&RjF>@o> zMldhE07dC?yQAJL3%H&uNtc;&acWqqD)}bB&;ve*#ko54@Zpc1U&z?eA#WsHTQ0JE zt0wBt(2-~;9ylUR5&?7Za$rj)E#yiB416O-GGm*R=3og=y?xG~>$sL(6OUU44~h3f zNGhiCAlbA%yn<_$NGsH?3)?oNY%;3ns)# z_3!hC+O_F&ZO-aO1@LX(0``=HGwYLvm}l9@DJ7YY9)kZ|M@iB%*Rw&&@BxNkf>a55 ze|J%hpyW$<4CN?2w&7`lxW`oj5`T|<;iYJw46?Jt2;%!4b9%zeaFPETR|`-`NgcqC zm9E#>Rd*`**`#uFb-mA>66~o0hC^RKUcY|xVYz%NxO4xWL=PfwovC#`huoUB@6>&z z+IUlqDW?xmk0_E7TM{c7e(>u=j^kc`a_KKIp=GWw^dOd;WF`L_b-|g&&|o672;~`Q zw?`akhiUh^^G8sQ_&+ehD63kBrM^<`Ztj2o+Y`XGrR;(0f!n6<_g7E|#YSH|?DfYu zcn|Y(KXA1zmOO<@zwak7a6yjaj*XMw19ETMkG<1-mHXW#^Y44;!9k7i8{eNt06(W= zuPB-SBBazD1XDB11)5Q7$kUYL_bDkq-}K)H`*#Y?Pi}wwUEHXf?vKR32#{MlzxcV{ z*bsl?*pJQb`X6-23->I1OhD~5pQ{HJikZ^n%!@Seph6cjL1>M~n}|OH-8ZuBlF6bG zh99PKJEgGV83`L%^#b>dXZ-Cyqk%qMeB)RJb?;AW72_z7WYti-p>FIn21a?Fww z%b(Kl2a=9lCf%3tFufH;5Z&)no841GxcgO8j-*_`wHJSKn($57zX%=iRuk!2`$TLn zyo~c1$|KcW*`^QH3qSGNx~IlijMw%K5db&EY&@^)2Ug%ZQ1{iUEeEgTtrxq>TVynS zAz%GGIR4zQA2>jhS#exmUVFG^j*CsdvGr`!M|)Sv>Tbczhn%11Ac&K)9N?jPoqS7a zNwUJE^tEcgJ0-zeVGY^rdPA1UbbW|yw3-K+4D5Y9k)2*!j<>A%nV;?u&}buTiy4*; zWNpgAv}Fv@<>yIeNSA{_Fa(Ck3f8l`uy?2@_Egd}vdD1!v*6B5k*3X_3O6vQ%3OvU zxA^{1M_2oX=3tM8YiErnE^g8dZ)$;d)sRIo;EJQ}P>cFqT7hmUIx9*naOl*#d`}h# zf@g256JaX>FvjTiKV%aK2Kxo~Wz{w&5`Fdxa9Qdq7I5 zcEXa5IUxj#HU28H{(8P@WtF88UEcQS=XnAw+mB_hd? zdppM(j)Cr0v;A!>#Y4CR^$(w`?cPp&)2)Zs@ zuN(mljz`q2A84b$C&qe-!q(O{XjzKJ&>oVvLRYLlVpAjdnkqTq^4IT0^m&2czE>$X zFC%NfrF*o|e`A*d4-MioKtUM5=+pPiLEPc6IcU3CF`@IwD|_H^`nTV~Xi%j|*gn&| z0P^c*`1U~@kN=9g(jn?Xr82--x_V(0DGAc&w!2Q%ZAXurTrQsL=Sr@oF1dY=B`Bo0 zi=vkoU(2$@4ERY*z?SzY66Cru;RR}y4s6r!n+n+WW$n;d`%7`ImKFEQ4~_SH_P3pK z8Bxf7%EtMJ(s?({n!S{EROEDX87j=d00Y#49m0f}Nc?9poXPXl1%S(ec#f#>f~=ThixsLyAcfqwdcKucTb&;B%w< zg=IJ`ukSV{Zx{Xg&d!eT?cs!_kka&JDh-q=n`*(IOFnO_SFyQVNipOc3}#i#{K z|B|4!^g;rYw|KDy?3;p$C2B5S&bKo6e|?}ji(Yw=JDK(%v|M`_asXY==AX|?I9qsB zabQX8vjX>l7q9Q*g{7yfTP;UK=n+LYo)^LBF2ES`I5L^HH)=W=H&>pM7W9rNLR!4& z+pi))ouIi_7xJQb~_4r8Hym2nzPVyL?un}ZV)scKBK}OlYi!&yiGZw3&EgZnT zL?wz`SH$X;&f28LV>i_jpn3*lMUmfk=@h8avCw9u{$r0cprRR;v@rYAcpm+ z0K;fHrjQGEip84k?QK20qQMfdJ{+k_Lg{+|d9k58gq@X!0McM8V`H#cd5@aDWQ%@# zKO{qCAc$-T#}Fl~MdAQ?S$Kv1c7c1!i#ma#+HL2q1&B{-SBL^Vr3GR@r78O&AckZT zN?B_z-d`mU1m%=ak~CGPGMuYfSr(loIhJ|ad6}Otd*+1g>WJk75xO;NuI%^@E*skI z$q_4FVgAA}hsGg(M5*)({la*iddDj%Il{P1RD#?>scWm$=!HDLqiX>IeICK$qr&XTIJJ>&) zENSVcqc=(sQ7Zzg`(WJPLSVS2;p~7s=Y@6RF!G3DGd$(#W8|E);b6nDuIgw4Z0^&0 zPCsKLbC&n4QdD?yDW@0);Yuj8EeNMvsW=xrDJ=1{W5oB{%09ZLrzA_8%o9?0SyZ0k z%9_UWPTcG`&w9Rq=dAdy3NV}}dESYJN*sxSZTzeYvJzkW$jm{cHXNGUq+IOZ5rPA&3 zYWIiOnL*fjxEYg3a>6WuqrWOn>GIAy4;!K+@QBgnTZAgt+C|Z2TX@vmCM0cE-`TxJ2sYZmTGo7=1Cf$F3F+j4=mW=TjG(@&`YRgK zl`-WuR0wQbC1fk^4`-~Zo)4+KROCqIV_Yye_yN*=z|_oyp38kif>ACJw?Pw&V=LcH zz|Cb)hsm_s`5#J=i6Y~+5;DacM?f-j-Y5EDw#0RsYJN<{0Rh-(D5RB(UkK`7zi-|X z^qR)=Og(C(!b0GfN--Wf~4;AEusuTH} zK|?}cYZnP};|iAvNPS~0l9K_FiAdd((r9ViF-?Q<8h@&HFnHG&U~CYu;fdT>&upr; zVl8=_QLf}oUr;};!thB)J9SBPW+u)p=t%VgKv3sPFP7S#8= zj%p}%q(60CqO0DU!9OKa#-iO=3c_R4lm((N9H(C;R(4 zt@}cmeLn<$@WR;d7tlfX5TeQzS1IKRdHe2Sf<6dihHxD@yf_}YLicq(i8cua`9%n# zK2uiA{A4vpMA04qPisur=8g@X_C1@MVH?s5?0LrlNMe2Q92oV+ez5>T5M>Ezq(vWV zLaXH{=ul2_yXu>pnykFnM7hDPha)YY)=q4_NNpeLd#LGH>x^nzF^aApPx0lSI?8&- z-B^b6WTu-T$O-(K&DSiowQl1=s0NIV!t#gmjRT;mQINAWT|SC}p_*SJ5o4w|fl32- zNA7xMa|SM`o=)Pf^5%a?Bk?8wT|S1&2dnpH{yZ%`ftP`YEY{kKSfcLKI1bbCj&KOQ zYMyF@qb(0a=d?u7GTpIOGps5e@YzbNPaml#F>bK7y|o-D)iOX5>3+RzAwDokNhsX5 z&qcs3TM4*%_r~4g+HOR)5no^{kFMPktG{n-ew(UY7gH9^Oy(`p9LDwG>~)-cpGw8G zQ`j(3-MOSm#JwlpTv>LgPN8BlK9L$w@PpU30WIp}vQ)b;XmU_)2)vmUnoE@pz&ixb zV2$f$i#2OAv_-Nb$d}!*jVLdGT=9bu7N|0GZN}FpX~~+FWw(de)<3xZ19@rf>h7pP~@G$2Mo8i@@WYM8K?TpO5_tM^?3f=*P zuVDIGDK+UMs9XpVi7#OgE+7fCBm3~^hQga9DIFth7eLD0Ui2>zx z+zIGZXr;hIXKuk>aOZdw9Kj(m_}shVx55tPMA!wN~|hChGATOH&Si68dS2-nBnU>XrBfUa%Xx zfi*Iy)q&O-ZOj8==HR<2dY<=B!+S%|Gan&Ch=Y?VPRg|l4_k=y!H(7o-K^)facGZt zaXdh$1$!CMc;yQ_7yN9YdIyjqi!EuX+DTS?J$WCO&F8w+u$op|$ICmQLbap+NtP$_ z?102q=sK)M`hXVlk*8hr3ZP5FAPtAp9Kc45bqRb6lDPnyEEXEi;^G+$RRPBe*%!l_ zTXJlId;%|yKt^H-E zYx;|XzjR(^Wr0sivx}wDi^0^yw*DJyJ(Ee<_yY?B)XMk2515=x%>I>6S6aT_>tZNB zJ2^r_wn4q4t9E))2S_7ive`~CGaR-#!l`ugFpnYf$Z+cFC+)RYE=Pm6<8^dBGdwI1x z@<5}(V%+Wqq_2Es)~b#Z46qW2q7Ppf4R7|t%fMJJ9h@dQ_rlm47uE{gG-~+E@s|+^ ztOB*8|CYLC1BxA`hP`Jom*lqfy1ks|7?ZGe8Z43KLZ0&<0~T1%tuW3IEJ$WZ>OA}d#C@#X1-o{vo>E$&n|D- zo}P}mN&w6&(HsTWi3yi0OPK|IlKZZ5^0&R3b2H~3qc{Sc2e7phFCDS(dy@vliIb76 zk(BzuSDz@GD9EBGbcxGSAEV5z;wt~WhUhOCslO-~nHCemghJeuMKNq!XZ745ysM?_;N$VAm(15q38LE)@KC8C0Lpvt zNQk+1SIWOw&49vJM`^Dvz)ds!&5e8T*{jall~XXIMEz}L_#@L|^*-M0b=9F7?Q{wG zcZHQ;X(^1~X+;5l`L7OV#LlvLNT=0T`l=sH=AvGFdc};bO%Erp<%@RK0-1GM_0u2r zs&#dJen3*Rlw~#_8a5I(qPzo8Uo!E5!Bn+1lLLLXWO! zCdCv|z?pM;8gBBD$)vnA&Am*-+tbiR-$uFB_+brW9>FG z)eOo#=2b#N&bzt&Z;Pi53VJQu%d(x*a!N%L>T#OdDJ|9;*9LXTg#x64w%SL$M&Iab zjbHe=T5}6p{`@&QnOUF0@Yhzj5hIM&Aac^IUnyp9^Cl3gDAHP(u_!u5BRix6<<)OQ zNyf&SLj!5CEk|I$h!W)@Y&cYAJM$oQM9l7oW^oU@o8ZxdoL=%00nas35`8dYl+>a5 zb(8v%P|ia2QP_M~J)~5Pzj^;S7vqgR(%C+>FKN|SaxT50sA~p}M?^f!t_66%kF;TM z!3tNiH(tl3=jj}$29?N6NWRzhETw7>>rUFT- zWu<{~%$wl%(Lro{g3V2EG$nA*s3%D^PqASu@KYEAyxBBG!liYy9M4{z`Mfx9Nq+DTX805tUm~z1wV~k;emhfmW5aNS5a-8@;fl|Ff zYM`3^Aa%f@iVFtT{a?Z;jx9 z2_r@v`E%ueseLotA@*PwL3hH!LCY%n@ywfI&Kbhc#!_xhePjxxwdP|TV z^WtzKa}BfiF$4oXF%E(lXcBLgP~&-Pi>?%gm^p%V}ilgmx5H$ zKlUq(?;bUHxH0s@OZKtLxd(*q6>X9AFi!4-f+vPYxF0^chLe_>ym);nnT`hE^#oY8 z{V7G~-8$6OfGiL29LFTXNYj-EwW*RG!Pg50@X0-7Wee6nI1U%fDD8NAQfTY;lfCVi z){bT{oUC{Z#8pAk%RgCWSe?DM8%r&jxCnec-HZUJ~|O=tEe5MB=c zR3rC?J>_b^rw+ym1z*=zUrx#v>r#&;4~=vZbT-=D!t0gu-+TeCMQ;m%K7#nvQarx1 zuI|P>Sjof;T*i~D{(z=P%^I#PX(9`^zO&ArF9T}fYo)h~7SUeyg4O;>GQ98(?92&e z`=iO>0zcdh9oWsE{YDA(^#{z>UU1PnpcJ*ANX%St)_KtD-5h_l#xQVLE*`$!dTY@O zy}*+7winE>`=D=Z&6#%<2m?8%uUQOp9<95`gqQX^DGvm)cC9(0Wk$@58L|+w*3#+& znkH$62WbVh)hdcff{~9(_GY&TM1Qh4FQeAocd;cfGjS!FY~wV~M{ow!P*7|{U<%{-7bzKk-!gvM(c zd#Yqa%_Pj9{v-kSDe=*!x@B0dWvxm$|ugh9LYs@wgj;?*LX9 z@JDcQL?-gHp)aM-Xx1j0qQLZ$Yy~@?e)t{!<$)XEX9L54+j%w3fHlcLmxtRK)x=n* z3AjsyVnXUKrFk19s5lgdN0AwoB>E;7f?!+$)QPW$iQvp(H4PZ<`u55QN@ag3)rhtL zQz!@D`6bUvd~Sq2Bga!73~BVfgdsUj!S1pUH7LXP;doXLgpz(F%t-3x;ZCgfoCJ`Y zi9k1qn;xI5$AbSpgt|5LO8;UUH^#7VeUSA1Kf2y2I+S4R8jWq+wr$(CZQHhO+qP|I z$Jnu*9o?L_d;W30e(N#%rMs(YRn0Y_n9NDP(xWa(rc^vlc1EJW7)3I}EZm-klR&*B zauS3!iwpi4;>R71UrHQ2&1`aJYOFMAbQ{G+>S-pImQJ3fDfQ($w!5F>}u@hQ%N1YHC_r@r5(o z5N#qR=oiC)Wcd<}7HDtH@Nd?ViF_l)$f_wq%2;Ep!J1ixn!SuW3DpzCnBfzmVmc+j1Oomw>!?ABGnn5={xwLGL-M&iQUA zg3A(HODp~9U#I=)R2t?lP z&s21#)EKfeUQb2T;O6-`{&4ehGe5{MCi1^YW@HS?GV;JQCXesO@cusE*S^z<5=jzf z+WBDo-Ji|Xe!u6#zuB+!ndur>i6t1#p_8~I%S7s6G`Ki1Dtx-xmcNFblZ`#6?>Pa{ zjZ@P?^{wphu`^XK_{Y7RgE(a>?nphfB}ioAr~b*!-8!S>gV>Czm-bR!_PegBaxJuR2-6Lnbi6W=p)Unc{|H&W~4r-)-Rvx5%VnWN~waKFeqeWx4frHAQzUN9B6)E#O}r?sVF(CJFghWIeg^y zPZ!#A$d)b5KLqEGG&3M%Vq%+p9Fj|>$xK^jz~8Lw$VY!$P{ zY1-buokK%F(TBl^{!J}rrgxHkg)yQbuA(SmV6~D;nv|*rRVg4&MF9(8B`YGIgjnwm z5wRkJI?CW7fy4@3jk~N7i$#O)O2=FS(B)4@YFDv})h=`rW~OH*(l;ZMw2dsZ7u8a# zcQtZHuF=kmBm!#8K_?tglFUYDYF3nb3{f-c3z5_#FWE1u2!d{-G3s!G-em~pPQs0Tv}8cb-(B+QM;E@1`nx4@HX3Z`qQizVyYA!|ITL^ZXbpCvpWVv=|* zHJw}k)T)tTNoPA@;~h`1D}jgx{`O%3w$ekE&4ESIB9zh*oH^*=>=+$0iNDoFvPseI zaWMj8xBxLklU>G7dbO_5HHI3p)$&L~<%p0q<4il6pXyLHHz5l(EGKei%nV5HNRk7U zovA00nJRwLdvWEHSyS^If8BCai{^QB*y^-Ojog@nm1&U-NW)>$qHRTjR$28@#hMxo zwg4wf^|kxuMlY+#^ERAGlJKYZli4V9^>2b&Z{sr3S2RkptexbzLkJKVfHcKKz(Mzs zzHOaf>#K|A>CQ*@)aKx1tR-dAgz{va7L042(5G6U5??Ed0u8A!tRKRO2;lrhc1gpX zt!PlVVo?hrd?K+%hGdZu$C(o5o8=pGP24fwx~mbE9mBA<k~<;R(`1vv}K?=2FQV?Sn~FFO(Kdm2zUf&@M7iZc5SpqbMn>{8@ z2*umqJ!c3NN&lk_op?_i!$MTiJW&c#GZE&}hA}8$W1H)5PvyPEL4;z^x6M;uia@?7>AHf^lHWIH5t*7#qR*~rVNSZGW=XyRkyV^?_pB;8LC8hmJd#cCE z!8(<$Dc!(VQk5VllQxHnk}*s~U%(T!)+fQKI_k8u(3~N(gwNz3(*@FSKgQLl%vRCP=f0^t}5Uey>PUnLK-lcG6XGz{RhB9}4Ay8}1P*zJf= zM|rbApyde5gu5>Uvx_}PopzN>*~t&vmY{#SLX)%*!ESgHNe2)8WUOqqh7fW25le4` zhtjL;8GcXtoleO(N`17wdZ*nYg)>W$Y21|t z#?~ohN0bq#>;4{?+FTW&(858XW{(kYGH2WE4m z8KN&pqLpeUp~x=Bi^t|5?tLvC;&Ccw5k>^8;U@csIRgDHu_jN3P{&TXZ-lxxsQ->( zaY1NIpL@%=*oe=m6||y+mxGOPbV;rKhwrwcP+eq6s*{>`pvdu%p3X-3{FQQaKlH~{ z&~ZD{=V3dwee_qW?M$nec<{9Ei=JB+lj2wTo;NIhw>dLOPppDiKp$bu0qqhrbI^N|o+X46uQahmXT3qsNGH1e z=2F{1I%guAElw=H*y?M$Xr|o`>Gu#RK3^3hx{0)IjO%3a{k->gE zJYzzW>-ReMTVKLrkrG&0aPdo9{~>fBr2VkAYS>HIzH-YN?da1-aSUnT>OFnmE1%D$ z3%I9!D+ga+9b$!BCs!OjeYZLS-9NUHyK2Ud{aF`ilCohA8m#76Vv_ z=i{tTzgZm}F%*Cn213lgGVHHIoxKyAe;DoH9`VVP(JYY*fBsbZi-z)&2LU!f*UVa$9NqHe=RL3ayzijzL91>{%BOPwKYHc1OkA-&E z^Gha$#m&xHDLPVB9>=HX++Blx#I=CgG9C%k$$sJp+9^CPy@8R=OgMAZZ(O6Z6d49qS=__Gj3Bic5hys9K z8&Kl~)1!}4MX`j)7a4{>TWRK-`_!?!@zs0C4pUuh+=3X)_PJ`sFKI&hg1h(K-Q_B8 zmQDWm@a(@_qgMQEsCXf%Ec4T~cgz`Qr+)DQ&_xm!PE?zHB3z3V{M{Dmz^9fybja|= zVX>av;Ti;u=5E(5JZ;Km32JxZv{hxdn)dJ^OPQ~O2K=*`5))3A?4P_`PO%K zr<6}F)wyBcdtFAzQd!n&DP0zB*0u3lTgY2tI+d)Y%L=_9|8xmFN63^g)ABFES%lL` zXCt9PMfuS~+z0{Saj8L^>By2~zD!WM`wS$o<$^~-p7qGWENdqGCG!~f#j32k$@Cp@ zG7aiz-heG0S8I^5_9hD-43Bi-xn~Q9J8N+WtxbKgTW0cD;;yxwx1dC6zUMzI(qx)& z1?K7s_9#n&1QX_BOF>XW_Al?{rxDyZ3_%>sjbK}k zgZARih{dMM8$#`peq-YW@8JCQ?!>Ym=*WGR!)LT>)jcs$5zOa_i!rpdw8E{MJjllY zU4xj_V(M`V<$4@n@R3pDSA12doi)KwZUI9ep1{Hl;(o#ao;*HPDq)qGXCmx~Yp@bM zcW?11fs_s%%D5M@`lgyuePj4-KXx{T<$Bj-ho^G}ESnL0b( zLu7KoEEZa0U=B|}`;tS2{}fKEZ5dvH0$by#%?bna@)>1~yWmY`2zCXe2m^SX5e&n# zA-@&~hB5dM^T5rRhn2!q&tl+(ccv`GfegxH0#?{?5$RA23*`d0X3eXzf{230u~A!_ z4@@^tI6cLnOpJiwiXJ&|e9$64Oc%Z;rzaZS? znz@{P9TW1L76P;F@9^%i%bzcF^A-^`BwP{V)-X?oh9XMxVYYL|Jtzt3~LTtVbVhIC7 z^uW5yNXP)n-}rEuvba1#hXO^o0Fa~jCsOU zONx9)00L@Po`O~t(z;>gB^hP^rCpbfKi_}#8jRe^Fydg@wZjFh6g+eIgL96teO>8* zssvvNp58y2hQBnsx7*`&|M_g~fZgmW2J%LVAm{W0n?Bdh^XC-#YXSG_^&>6r%^Lg@ z-Dl($?)^N#KiV#j&(|uSSD3pCk!1g)_qge!^XCZii~lUd&otz*S8U@+Q=nKy&eF2v zr&}W+H}0~Z5zFA)*JJxyrYm<&z@0W%aB*xZy}nlNwQEav{9NS+xvqI;B`XB|C$#nh zLQuoy-5*x)%v(gTycyA-!2^VV2M7TV5MmD{;O?HC0D+HB2my}}A|4?`9Ae1rocI5C zM+{%fF(Z+j<5}0$K9( z%|x_FAq>W0RRRH&kQUkVj>s*Iqm4xRWw5F57hbm|OYp(0%*XUrl6S`1n25r%zC9-=DSU2)Ee>`8S} zykh9wh`(|lo+vqaxgU|}F+rb;B!R4bDv>CJr_alDc6@&MLo~1fNOhB#x!<`z7Y$Sqqp3qVxF~Fzl8XD3P)k+l3Y%(^<9WCRtx4exhqa^V+uXa5iW+uNJZqu` zbcClk7B*TDdmWAXy)OV(pAt66zo^>!QDv!vXR8?e^qa0Xpj$x@z@%Giu9FogtE;QDIYX$=topGVMB%xy} zLyVVWKj#N$BzCo&m|HC)4iT}7yivOAb8n`b7-zc_0k(Ly*U z9hA}riTn{i-Ulr;iK2*JveAthz*wR3){*hw@QHcQ@I)tabPeNIQfZrvHX_5Q4*KmS z4a3M>GkD>2HQCM&*T)tlnIovEHX?MnVOZmEM=6!Lno`F-)>YH3qJmjpLyF~u&Dxsy zg&!2j8QBVY}KI-oIkc9bv~~`HuwCm z9ZCZ9;}%}Ujph_zGj9GrFhLEc1#v&aM-+IuJK*^@v}4E>sn4G{v0?3+pxQ$;=`P7b ztkC=&p56WNJF>0lrubeyyILRTvW`zjo=@vOYmnrNNAEVr?S{VY_8EhlW|{4|ey?VO z6OZoFBz}Z5Mf#af^yPc>=Ec*!N1P5C%b!)ZSh=R7_YBbo>>Q8o)|I#HGRxP!E*~sA zeA4YlEO)QebPg)PcI!g`nIh$Cy+h08^Ks1xjT{KXhpaxDd`%AI>TV!?iDs! zvwq)LV1_(?MvZW7XM#7^YVWKVGp8Ai@7uMx*|AX4;B)^EvC@h{GkD{7&e{L+1AYbm z01lq!wx)J2&h$2>9;Qyt^iHPE_O4FGroZ!URLwTNm<{1OPye7S6{bDpcyylG~yEaC5M`qf)5KaQuP{ZQ_B8r( zRjBcfLEMnU1}@Y-m_ZKmZ_pouePtwf9MIG^H1Ib`P9q33KvK;qn7c%Uc-J<^2XMIy zTW6$cRp1&9*yI6#btG;3Qenji46`U% z--J>mivTlGg=7QIDy7VzbZN}kk_+5Z%I2{n8c5y41H{5W%n+Y++u4K|VRmeDm}LTJ zUY@v!Xd<3ES7zt&qY9HpiX?@Sc`#4Bmy+E|&OlQ%+c!SdFjfo47TR~s2&EQv=CU~< z*>=OC-M@~h%OU(m>Nv10DQ#{Pnw8RFKo_N){6c{(VvX<9U#mM0R-U~8p`SY69=^`q z8U6aSC<>-wA}6s%5|9TAKrfM$u|xM8KCbVn3~Htcb}op*8)m3AI$m*Dx?H=UHIH*T zWI84gzH!a5pABKQdVJy*FTukP+gkc(S;KngT7DfXg_I9eVmiT~EdJz%MX7`9+#lIl ziza64lKQ3I=rl2$GD&F_buy_swxz0zMQhhBaMJQVtmB?diE`v94{K>G+}1G=+G}y_{b3%Scc=H%|R)3+Vw~g#8#UhC{?&L*ynR1guMOC83$a zfCMi=vr{|9Pcrqm!tp%LuOF{is%lym&*rD7PYc|wI~<#?`*DF~p>M$Uai+bRC4BM? z^?Wr}wNXuR-{yty&cx1LYQ1^@*!@3XrsTS%<*&AJ5!D|$7eDfY9^*r%=Uk$0YlEeP z_0MOc+DE$&uWLNmsrxc;PlJ>Fqn3ZLZpuWwyWSLikI7y2N9TIB%6XsZz0b+lEViho zlHTD3Q^Y>7CNXmlX3B`S4*ylCwzsTYDeS%<{-73shuf7!`Ka{kIEC;fP?H@+ zIK3!-?pj<_)qjh}B{(U}FlPm@mhxE>+KDPFi13pD(aWF}jE{Xde|nrb&D2g0Yp zHMOqm{JQ_>N^Wj$T6+9TuQ`#9NWEh403+lM3JXB~*}AjdpILD5LPXEg3SeIY(j$7) zN%)&I<{Y2sk(b2j!gP}yKu#@q3hjEBd!l=Lx6^&}a?m|YY~Lotx{elJ9zdg=2XfcN z{+ot(BuTD3Yyr?1f?UteIeik2Uj9`riJur!0Abp+ys_xUN;H^`I?dB1)1k17adH=f z*3wh26Vmi|k;9;}%}ZD6UOd9NQn`<1CGFK-1G6&CT{Nv=%Y~CUt8SS?fl@JLoYp=n zfLzMLzNKgWt_CYM9DHbFIQVt4`!RMLA<%!%f(1ydiaID&~`W^*Jvj=8l&TA_MGjxyHor}KKI zJSdYCO^Wov#Gn5-@XCi&zJM2O&wA8<_qxaW=hEChzsfNSfV4{~sEZ5*+*~BluLcB*3(0US=!%44^2+KR*(gD!Zr1CYRWox+p)f8`Dsu=; zwpV2YHBed?#3>`S^4qy#fz&m-iB<_&!Qz4hfkjc#T#3zF zUGYqCWF|PlGMTX>tDTYg_ zS#xjdEOrAPY3ZjSqMVCQAMy%M_DD1!lU{@oGmuri-Gvw;McbZ|KVtCIc2Iz2-9+&c zP@wV;_Z;;WVHV(}LZj1Nf`2hZSi%-yo(#>`rlLys^<~ZBs^$se@qKM-Z>(RI zo*GxGJ8E@V;0`=d{{DR~!kG%GML!IIfK;Qo-7|ys`X&Ye^lbUyG5#|#oQW%m%~(n% zdD>23(bCkn)aiK~z|5XYMSH~couI7qW4`@2=Y+1gzP$^Ay2P1Bu_;n1{-p$#@lw6K zCL2+#5_d&0p=JP{CF9bi%s%)Zm01R^7vwx)9AN?E=(Tf7=I1q{*+uHXICKm%T_M^x z{r#?*aRI1hIYrCM^7w!ZjU|^Dh&ejqzsUCRwZhNR$?~=aCV1g<@=G!UdJpc4EULh6%l=K zfj)vI%HpZ~Q2yNKCzDa#C)iZGQuRywgM3eSkda^ad$2%VV)oG9bq01(?zK!yGsaab z)sQ~gznCOQ$O0<9)RQ#JJQ&u(ndeqecNYKel{?`&X5jB^Cs04pKr>m_z^s3C{e?D1 zd6wnu@qB*#-*;d89`?F&MY`Ce?vLb3sFmGYquf=@ZO8xqE!*2(&+NHxE#rS%zkwGY zY}PN|UE9+;-+JZGwmRzyqxp^cy40^2Wmsd;xgx1b$Y!2t32VHt_#h6qE`ZNBvr7pe z_~!0Fl>5B3JYi$o46ymK`@BA(?K`(HH<(-4Qs$P|>z{+#ruh#GpSlKFgMUR&G=slz z8GY#4|8r#1gdfB~qpHu)=HajP{{!>2%F6)Zo3008q}hl>AsP_=Y1{H>9p$%giS7!bnxPt|pn;pGS*Q7n_gRv5>q zYO$=S&sc0RM9CNjvyjCbnF~rkuAOBH1{O}_4|d0Ib8q3c+GrQZ^~ot`Q(!oB7lpt; zl}8}@=W(VfgoTH6el=pgSHR?&N=YGht9B$0u7gRT7efZGkhyLjwR@KCy8||`R)PnG zbI?&qE5MSJPyKv;efY7#wXo)hvn1C;K5G4W#X%?=TFRC6$nFwbTh7q zR~?~3m|Zp`*FT^WW`7XU<_qxk7I!9e;rEgHl+Op92suFKn4m;} zCIwbzDr2wr|ALKU)3C#hwRLA2pcwI6cy|0+?$|Y(7k}O4cp`Ci#V4{g#NQe}(F2<= zz50hNmqouod&3cZD{uL$JFoKP2jcwn&iO4`v$3r!zgEzfUx9FMJl`?k@qRC|_}VrbkbAYaQ)^~t>s7WU>fSKN`f4y8t;4b9mq)|$Dg zUY>pD+-$S0zoePk{*}1U2t`G=7RtI7GbZLy?wX~sxj2>#lZg1Tl-ES%Yj_1CP>?!UHcJ2u22ntD3ucOHI$o*ofV%Qz^n9TY+v4Mr+}& zHO|pdxjkAE^V!dqlUltE4jT}Tea(_`z;&H?WtG;_X>tiKY3;1%%T=na&waFKvvee7 z%1@+yraIBqc&Rg$KkTVDW}~wo!MPDN{a`gi&!dx|6qg`)Oofunrt}p1k^mJxc|@)M zgu^1`ouXx`83Yi5QN;tF`+Kj7MMz>65b~X56a2~?vBHUA^irdyz9&p4COfr@y2>ce zpUR0*S^lsu>{RvMImxmXTr%RnHi}0Ol~CQUyuk6>-N@&6F9!p=Gun|3#Qr(1luke| zZi_PlVL<*IT6)eo?1sX5XxWdWJ|mX(Ggu~jXHrJ&$hCHbX7OkT{n!$HMVm@-ariJ> z+Ho?k2l?OyuS33;NmC{qtg4jQS&xKg;Wgy}EPhY#S%*J4u7!Whk^(`>dk6}>f<825 zrJO7b+17Tw;_1zt@0L~>0i3;E{1#sj#vcZ`aqJO?pCpcF+Gg(d>Z=AehRH*#?8_v1 zzV5T7(5@`T>6Y{O$@hKli~NCy%FHJ}rT_2All6_HnESghWBuA`{7*E)-pJnA-tIqg zhgs#l|18da^ZJfO@P#ZPaF*T?DzduKY%F5r1@a4p0(6n8<8e04K`%%4`PU%~Nv&HY zR$6&EAD_PNeckn|R-6U>KItYCDNmGkiW2&FsQS)m+IJts`NoThrUEmdGE!(d^5TAh zVAgrJAd*U`1FPQQH1D1ZitGg2U^GECC0y1|7amyo6xd zrGm3_MkzSTdc|`OKN{eeot=}Lh8=TEn@XHJL83FbzuJ0G5An24>cCf>B}k$+u}%SX zI^&qj3R4zJTH{ag!rI_#{q3>2>Nt%o6qKQm`Z=RpPQB==0vaH5phZ(cl1gd2`kLuh zlSDV8!{8-W&^dqOH^vr>wX0I03)Zure?=}@nw=u=39o5Sr0ALq6WM6qwhLi26hWN( zR55L-IBM1-#MKN_3`s=1{-EDB3O2$bBkLCMe{=I;#?;ZhsPt$3!M77{KCs}x+lY3v zJ2^rBu;%+BtJ?-RQ62FiXXqv8)an^Mc4`(m$~L2x1Y;p{-f}^m zsl!<#*D=Det1f!p%agl@&#Vk8EeD~f)84g*!^g$GsOghLsUJIbcT9A??CG;SwZ?R%=3^?BZQx|xsQUVq$&;s1M7-!o~1 zn;`-KAV>fJ5dJTW?srtXm^m5Rn!4LN{l^xo7u(u-YwVuee^gZtc%jTdnpu+DY9&m~ zL=-vvA;M)lrpAEafrJnY5DkE~x$R@C+natH4-Vn4OBFm@;MVW;dV}lc<^~4yt(Kil zI)6zu^#-n|rl_fd`s|uX^RBDdhfMHu{)G*Zk_?AiDy<$TuG zcOdRbzU}8~v%ULc{-xhg9b0YB=|xI$*eZ`kyyh?`R)5})hdW;kNj7cHQ|})5#phoz zj=!gu3r{9qbSrv~+)_;x7`}{QLNfz*lEDXYpCEjmZNpc zG0fkFbLySh3ZgX*SWc)dEa`R+&PeHNd{Z09D;Ighz%7rd{DWKbI;2Hw2ctNZ*nFa(++NP{W?;$Q+4xG zX5awewX-XV?{irm%pad^Eu*f!F4gK!m(R=W=0|*rB=k$4xEidOd3n7bA0ErKjG!yS zHTt8+<@?9MD>m(tqD_xFIOh)pJi^B zhogH=3TbQN@#)p;{kPBUArGB2QOGOZJ8Omwr%aRceoqa^E^&fAQj1AL1Q_^`$HZI1 zrcXrEQC#}rLJVQq>^(BlKT$d3c?|)ePGsMu>tF(N_F+B~(tT3`OoVfoj0J)M={}ZZL6dRhvbX$fp$L*Qr(Jq5wB&k^cmm4XT|(T}9L< zF$r3O9U2g`{wN%+r9+;vX;OiIHeiYu0yDPJG2<@;?sANtVudgsufHAoUT(f?yhh!Mn% zJYtc*f<^Mz~y5&c`vj zPRr?@bP`hln=IPyG`Nj&Bc|e`kE4U+h*4+w3gyt_ESWbP_ z-h^A%JjR{5Oc)_``-xq}S`;5hM$N&1BIKjOf1RGYOC2zgY2ej>+p-BBkiA!Y!t1#c zJJtZZKYOeAnAdY7n5=TEMX|q=w#+&VsPlu9+5R$ZhoS(O=Et%Y!Ft!Z{IiU~kuKRZ zp{PfqIXW9LwW@u)2ui(-nunyzFuyJx_pT!$D=21*=!Q%F8PO{tErF5RJUWsfJl6ua zlYxGhptD}2C3S(wl`mo*BZ{eLPDjaDEO0~Q+fRd5h_ShY-ROIv+47NI*D=(^34;fk zXJPkfrdG!ZxQ|JoWC?vtYXolF6YuL$DX)k#Ecu|Tj<*0&A{ujvY|L(`Xv4&PvBr|F zGP<6qw|hyla4<&jcVUWRGOgr-5K38kKVRx5J4#ZB1$Eh&dU}C0`foOogto8sWK`dFS4qjy=-+4;U6_PLZM^2v;1-ynJACkdUnBW2d;UTCm*Zu1;T#JR zvfQwFZB2~DN(AMPk)Xl`p=r?34r@V~zp5U0=(HA%vs1GPBp)s^;#J!y7q)im?zADh zUlo;r|As3k@%CwluhMGPb^2GvYrHy1@ibsg_ zKWK=K%_0%EoHLCXGG5)=HzuDB4ppz#0BXcKI`+io#Zv(YW{sPN%l8b{NwCS3pF5;%$o z=s;kmuN_q;sH$4Z?h!^Kiw>A5$5|-EzBKi6q4??N*x2cgX%N#~#EYyyTk|eueO>wN}u);&7 z(^u+hSfs&1-e=MZJd9d{)qF-v>5A0(%X^2i#+y2z4Y1nnn0#1}43 zvAZv*8u!p>edGjv{#%~3ZnbzRu}&_l!WsDDQ_refexRtEi0 zOg1$n;d=c&gA>gZUq|PxhzAn)#}^o=%&6p*`f=KDVv<&BwXd(>DY9VddT)z6gdoJ&~s@V3&&r-qVlitl04-dKZsXp43?nv^Qz>RDo9krIHEnQ^TtIG+Q`$JSH zlE5{>mi$2lZ|ZPuGOaR+%99jKH(?cpbMVXJVc4?KiJ0cscZ)mPc~$@pCy_`^`M;C+2aq{Pfam0*EHPcsY4; z^ZkC&!;9yNo_kB&e7YoWFDoiJ2{BSo9xCJy(O3^qN5N@jY&*(xbWTL8%tlm))yXJe zOwg?=J1bUx4O{@rE}7L-LhHs-Mk}m#>q2yqk0uH=)8%fIzqB@(9~QdJzZ>{gMNc}f zGvP=!E4p=*($^7qYb^)-isHj3HjdnaBmUfm$NS^sjatXvkPjIF+;S?qu_AfvN9<1I z=|OKRNQKVVS)BHi0@%c+RU*ac7PxK-wG1)+w*uA2qKU@m*x2~#51XkY`CoppKw&;` zTLW;Dy)|yiA^tU^vZMVn<;!ct?U86o08rrO#cnX7N%O2;+r}&*%s)XC(ZB^g#kMmF z;4vNhp{|cHkS}ESw=AzB*_VVDe-bbMk8lNw8IaiH>n(Z-?X7DH`$@xvi()i;Wku~M zV`8|6vFt=;WXqIcx#|`4i$0aSsr*iLwKlD-?FVNsrL9@Dfvq<3n@2_`&Q~qe+{d45 z-LsaA7o?Eu1h*AmIMH-0y)11-jmJWMf4_u7FLb-fV&(kvR;VhpsuxD7gA`>c_x(~$ zJ#iw&!`)5-!;249O@Wyjc3Z-WfFql&4!d;lFmFOe?O^CC;kH5y*q<-eu+9Xf2vKIcz4)JL+Qrx|)$&07~ z%OjqE-RwfQ4jc`&F+mwoE8;zFA_-l!-0Ybfct2{NfoHhp86NKFsV`hyvU^bNF;Bd( zAj%ak&#rhUkhY$C`T7aUH_)yV*LZxoU~qoL%wD!l2dj3z4wgrJQ|+8Xfx(owVaq3y14C>fP77B1k^&P$CU06MV?k>Xq_9);ivqaFeL&E%f48f84q52Xar#uHK7(kM zlc%n2iK>wwvx9z5>1Wv!vlQ9h)VaCsFQsPceQU1TM)7|)$DKp$|1ry^dsqFPHX)`xY zLbq*DU#W!DPOr_x56ZOAVMb}Arv4?9=&7HM{n-9 znqIWxR?XCnSc9irpuV)PzbTC}vL9LP}b9BP+Da+^J9KdMGT}Uq3-_7>G1WO(Q zA#yAuoQnX8j~*iedFvUHzfrLhN_$Kv-H|&w43KF*iY&bxxz+#<3*$|P5$cRRFWgQT z`};8LlX~GHDI7p;AjBehJu(5&dDBT#aZEEBldq(?;Gx-|%i?$HhJ;et?QLtQB~oY* z?LI5+vh(}LduR>vJw}$3eI?y5Pk0Ff%6f5WEh@Q`jBC32WG^I-NH{~?SJ%}au`=|m zP*kWrQ`N&PfhWl{#f>RKef<2^d9Cc$24gG8)5-t2Hp62vuj<+C-JnIW#9cHa`VgFu z2w^P!aGo91p;CUpMb_XHwAvS!yA>^5q)S82*q{Jco)O#oWXqsVy4% zL)~6fwJ#7-w@N{payX$Db$;i*g7&MC5?DXq1}omNUavkWHimlHrY>&&dtUU0YKTqq zdaia2HO_8I$w6>OjgKo${H#3K>;1PhEj*j?Yo1GeC(6(dW}kib5A!@(r@fEe`>Tof zn^DVlwR1a2_y_dAK^pa3sLYe!X3&Ag|Irh2HS+w0W_EbI{kPbXem;~J+;EcG@Z^?u z^;*+X-Lv;fCD|mmx{)nIi3lUDqf{W5;=}HLT~+`k?g16s+WUKZm2nV>0tXHBFkry8 ze=(ZH{pPMbdd-YGy@5K78E;3f-ZF6ZsChVEPifXxO%!YV!nTu;JhbKk$&7MPIS%<; z&Ldt$_kFSI%Z+^cq`(&T%lQ1hSt5xB2uNC>iXU zT#?-dKIZ03jzI=Yi(!8>#QByvp5w&$4SIN>G#qbRk-M70dVl?j@6%6zSUrCZR>#fr z1H%D^2yC8(YrG})bU(R@mXxWM-0}P2?k)-aOy69?^!4v9#cm-ujIsUWt_k(zSpl`H)t5Hf`Z=GKD`>GsovL6y^hkmiC)W zK1&q^Ano;EoIr3_qgGH(8u+QVbOAU6vXjSwuZS9*1!*)7WD=KTJnQxS{08J`(+)q( z_xF5#o?K_U*Nf09u-Fb7orvCyfF|Z}-E7{%X5$+wX*Pa1jHeUQzRYDV$W9arklxWZu@?i zmCzI&V1gIi3uxFoIr6l7(q!p~Cb4Z}Vfxv?!Pr~&iURa19Z}pbM6pP=n1W&IS=Q8^ zH^Hg_DopVx3f~G$Bd?A@fW1qTJjSwaQQSuzv0ifM)vOfgNl9e@PdffJlz~3ingK|G z7AP|gn#GVWhQ$be(racrJe3BDfKP%Px>|Tmu0?!&JUGe}A*VPJs1@c)oT)G~ zN`A=!Rh;f9$1{xDtcQ?Fe=!e~L7cik5~u}QAQsOwh>%z6DWnkPs>ZEdU(vftWrBQC z1w6yQQa9+{O-MEl?2>K}3>;PAOoN1Byb$ur0y=&LZ$po4zrav=5g@^}1v?&r02aJZ zBEBH}Q>2&4VPYsWA(%aT;O~ZW%?fISAGBDM5?MW@5&DONYMYgl9d}O$EWtjdHZ18q zTd?>P2kO$@Z0bF6$rPn^@=zBCstY@iO;%N{n>{l_>#qpF6+kz zjG?};HuCIR8urxzjnsizdo|+|o_i|n?Fcz*e2M*^p{!O~5$Duu(bl-}peVdJ@RvWD z8O;tsyWr$#P88rV*rcvQQhpgV0zU)G-KolrcnYp`^dO-*V<+3a^|G5dW=%{ucPsS< zE`MEv6?3_{{xHhuaqUkWs)|ro(;ZZ8nGK7*?_apIKp#)oUb$}IWOzBq&k#d;L8JCr zmkzm?p>*6Y^#frZ&G6aWb&sy0&mdiWI(IzdkuPwye}j-g;{2?3mSFe}mUcIQ0g$t7 z`hKZ(v#*stoU_e^wo+zseriPzc&OZ9H`>AG`SITHe6zWA`SVM??G52Z3*dF z@iyGcp7cpi?3~4b^y$oPy?9GjIDRaC};j2NgH65 zZ)&YLRXmPTz8CAox+rx1t7kI=Gk~Y}1 zaT}Q%R^1b&?L=}2m*9uJ-)^AJXacH8wv`D{ae?I}>jU4vmC?WIW=FSQ3qOf(9@uD7 z_`f+gB&VNfI7xCJYH&nv35zw<09$&=x5Mu-#2Yi7s*BwJ!decxFrRih=s2{aK#W4e z9E&NG@+|Cu?PKYNNf&#;UfjC1U70mcii@dy;j9;h`);^HBWoYFMwums(O`QEd^3HW zO?a8S$U3r@juDNpamw`^HStdP-m<{tD`stviL-qh#kn*ra~WJ*LdhgzNcrNybHqYp zUQ0DiEJO{J*M&;(#Vsy2(V(y5KI8(uhU1J3O1Mb|3M9Z!3WPPy;Pyu238F`qYavBI z^^NsIfbJd~H+awR@S)_T@b0q>nX!N<^&EMA0LIcRZ>(71{k-}8yoe}X4W5Q|JA0Md z+wOH!OxTB_;9e6tg?@6>@bI62ClcFdLFck&6y5!hF#}%cd zd=QN7I)BE7+`tm5yK0HR{a$AA%n_Q#ER)En6h)tAhB~g=K7)pTSv09R>lmdMIFhjF z?4k?+=t`oryl_M#_{z?C15w$In4emtu4_jrc=TvfW$k5>;*%hMnLsr1!R zyzju3W=p`MJrF*MMas%-N7aFeaUU<|DiS^og_JV#0zwW)c~E^ouB!@ZSHlFb2#)Ug zl`L_%`2gZ`pzBWp)&1T2b|Gc!7eqN1!vgA|^035Um=Q(IBSoV#&72I`aw#5zt%301 zN}P|fG=50p1e6s{8H1g{908-i6u5FK9HA$s2h}a1Jqn$R3bP20os!ZJs(pz@vthtc z0YaaKESg&*^31R5NE@3Qr2o)W*B$~pBUb>7Z0cuRa@bS32yPBL< zY_*S(mkVP9qBUq{We<026$`ei6sRW0V3gicQ~a8f&rALVM;?&c{4&T>7YZC7C&DZp z3W-ePl0+JWx$&S#mmPb0V;YHxDQGz~f?=YvgZP%-3PHH$D7aCt_X8A@22}~4BSGh_ z$d&`%MZGpCz8Nse&Knl0G>cu@Ku8kYKiw7r^}^?n?w089Vvu4&_@s%Hg=QvwfPLAq zmz`dp0w$L&AaH`XwnS(@O>FlgR6!%-&ygvzB;xzR3zxo7Fhmpg3$FleRFMHIhSg<@ z2sS)G4y%diLz_Hhz@8ix@}I~iI!7@e%_W-#10~0rMNtU%gCa1%RA+dVM`G%U_}Bku zEMu{1hvk`0xDh|*4A%>fe6ZI2rTND0 zbEXT{<3P~Jknyo+LL`ODAeoq9i9{}b0BGgl3KUW%3#UPyEC}4}0OIljs>w`U`p`JT zoEJbFLXDXS)`f^b>ZQwCb1z{?rVs?(gJKg9=Hf3nf_YAORlMfU?eUPB~!(ZL;`qCpZxKivD`1F~YI@BjObtvDCFTAU&{&ES2%8`& zs}^(0IZ2Xw-)VaOY}{xdW!NS^5TZv#$&oEVLMGJVc);Ev*lG-)`drE$2qJERmJZip zZF-?6Z!2?QGyD3*(wd1;uh;XFYA{&TnTb%na7*Yq@5-7GtIHKFh)Ax^S~nw9<^wp_ z=QgTXIlsXiVp1VU%?1{~D~EZ#&!1NPpGi)LquNsF>^b>)-2L> zY8+wpndh@IqLLxXz8qy(yVC#3zRLKykK90=c2?1R6R&jfrL>ba?)19)R6Aqw=XYiQ zv$0y$!ndb>MU$F%W&_e=MXI-v4JfNQ;j}85!~D95+nRh^6Pwk^h&5-z->|oV29+G5 z;@&=O)LC}IcwSqz8gCjzHwe**{Y70$75JT@LC!2jc50{g8gZ-JTWYT;jc5el&W$I%^~RRFH(;R`yK$#u2fl;CH|5qUao^zPiUrv=~^Ok$se_w23%%ya^& zDXM1_hmXk`p0haUsFb(qa7Ulv9(83aek4o8xv6_zZ@xL&X|jHqP&63OaY&cr(xB}` zG+OzGVD0!Z=LHDQ2@8A2`7gy(CE#EeKyT2a^VsKO?3VfCAO%BhwPWL3(!>?$W&GLm}h9X8{w)4uZ~QTvU3K0 z4c!9$O4skmu>bfzRG&j#@0h^j$1*|xn5xN#Pnh+KMN@rNzBnxDZ8G%!`Lr&L)AlLT zK5m*Qp>j&9P}=G(GpN0+c|}q2b`RC#u8G)0IisaWBnN+f)|&-slq%g)Tf<5DyA{%| zWX7LxcC8qCuP2;nl~kOG)RByb|NH5<8fbcFV-^>6JuCSP$mC%W^-~s&gadaxCiNrg(tm)lg{_LdvevLF ztI;<%3kYrl_5+G>`{Na|26ZA02e=)_5`1hEX*)+SHZ=h|9blYE&cVcEfixlhd3Iyn zE*QAeP8CwZF-?X?MQYQ%`V<$5k)5M!Ts9=Ju{5DWIqG(cc!WWD$4;}D)MBmjazHJt zMg1g}6#V-Wo6<#8q_U`Gt>;!8$gL=b1b8*K6`$n|D+!v-*lC*HDt@8s58z)jdc$AFN0%x?H zyu*{pWm1Vn-qCu}ax*+@iuKF6Zu4*Ka$k*JZM3ehrmsDl_KI4ArW!BIeH=sH?gXG_ zyBS859d`9q?S0*{!1B~(rb1?$aC=njSnf-%8}iQ-qbyJ>nmg{}<%dwCNN zpj!kYUrwI)lOV`Dr*-4APSa;vW<+#tHEOziMmm_&@kVlUpa_-W%s%7;JZ>q{(Wa;N&O`u?SfL)wQ1cXT`MD2p zwCQ7LzaNbo3dkcd?E875*`MP?*#+m9^4X8h^pOWYHoBcZEt#0dnSH30Rd?TbZi6RL zCuzb?+hz~tSpO2VwWXd*pL)A#NuBUR+V3(Gli+CJLlw;;=gQ=LYx`vQh~FU}n2UVb zjMs7NTl!*O7cF75084hbOLPGn-QW}s1(0ZhOz?n5yaqnihmW!VpKkC5k z9x<17G};7ayWkm$VP;SUC%pc=)q`GAbvZl0OMvk3T0hP(*EJ`U&xrt0p- zOy0ajE9>Pd>#I&71FL-Dt&Rq5s>LC!yMTF*(@2(%!w^;znIqA#I9W zx8M$(RvVRncA>_NrgfVcDEbVj-0-IsSCKIu7lXi?L_1N_Gib)16^xCI@ST-psFzxC zT-H{H;xA8TLhbs_1h5z@!U+vO!-1knNs_^TUs<^@=1KWgv#5Ttv!&2tF1Gu=*7n$$ z%t7EIbFg%`fb4KD@B(ZD%K4fJ6KBd3W++%$f#J6y123av`Av*$Dd>?Wc&Ek? zt-)^ZLBv;hr$bALpuHEzLx}pD0^qo~Fth8&g!soO%y|Z3oSd_*n+?NCakgh$LSJ81 z#fjNY(0a-3p|fL%Cf~@@oj(@h_7zEHBDzvexhV81QXgV735H83RyYl;`X3Fd+y9sR2F&=dZ1cwC^vSJCOW?WuUu;q!*v$FHj5y4i=?I-hOmQ;{W)o z?GYK>6Sqa#Y|baA*mQg!tVsz&9YwCORY`A2{abRtf0~z)gA*J8cO$CT3eq-Ea~&)l zC-v(i=(f15@^W46y2#0e!Mlrtn_GZ;!BZeCHN=s);lj609{IlJ+%_)4#jxMU90Wr_H9y`OqO78c3bEYP{X3scTrk0HWN7gzn(S z6~4akv+ElBJpV_GH!|Y!vGwzt+7!TC>24CePVMcpMEd6kfOv9`3=BR~-r`fkMWmE1 z>2Af9glXF2?7t1_6}3TgKu`bxA_V`}uL0-(82+07BT;Wd{KrxhoLnU?A4|HH_ATGY znyt@nmr3L>KHZQL4f!WgAA*!*0I-lmKmOy54jp-rS+yno+)om)4c@sn+|$>sciZQ$ z-;zPfk~|P`p5U~9rz4T|)_myh&s-b70D;MVl!fKqXE;QXeQY9zE6ent*>W&%Lnl76 ze6q~J&B@8n$Iba!$*4B*KkBLCnBKUfX824 zKsaHv0gz-}Jj2d^XR#Q^Uo0X68>1ETScFO7n_QPNDd0YZ#NU_RGie4aG|H-LI3tZtmr3w zStQE@i9`_zB|+qQdK7Xt`uq}j)^2`s9o~?k(WLq|gVz4+M~3K>+4gfe*-jV2Q{FK< zYu6G_lu3$bAqSlMaHWfI$}jH>us0CHr%9^k<;PPNU||)I-V+*SWP0WAS6RB}N!yD; zJU}EHOuf3|e<4LX^zxCONKL6ql3R#W#0Y!7^7Vtm6yj>U!8qFQ;`FR#{?+uwp8xy8!7{jRwYyEHyRKJ#rgav*j{Adq9H&y#ui z~$bH|*euRe20u&~v_tjp|2%?spq-^T)h5J8s+%0Lj8jM=W~3hlBB4?fHpRAzfpN}Kk~^alWR`XlKr!fX!U!~Q&_fTXb*>UB(MyTs(msfM z53JSLd*I>cf0k9pcO@>P=Onz8u%}|F{k)LJ&?%x`j!F%pGUg9Y=o!q!lQE1`;tO|^ zG8V9M1~X_wgsIFvTL{Sc1q~4=xF93BQV~@6MI9`~@OPf*HmZg`GFYA?l#J&yhpO;= zjrbA&>@O~>YcY2K*mE; zGHXWhLy<{ZUl+6m%!zNzGH(L7Zjkz~JHLUt`OAqbDaL*DiPN&X!g`bQi59RK6pFsU zfW5{S>qf9$-J+ufnj!N2*{1=X^FL!)$wqs1@P1b_+oFCpn!(VmS3~)Ops`P0-zVlv zx-1RQC1D$qXEa5-h0sV6I*hc?+9lM3&id1|)Rv) z8_9Xz7ck{*M=a^GvEk-#t(7a>E83GiM@5Q(&oD^2k_@j!5Mo(bVU zg=nR#=1-KI_=Iz?aE5nATHt(i*|xnRG)L0GA6TP)=;`w;b?cVIi|5WsPgs#uK_t9_vkFL z$gfysuKt|?XbQ+Pb0b7?TyG)Z*g!K12JI_3ZI&vsJ=^L~0&ZMUX->h8`W#=ie~bsb z7fjX-C5^DEs$U-oRmKSX9OM8BgAGkD34~jyxp#LF36cZSg=<0!QyM~05#=iT$UjA& znzK~}Anj%GaXnB)lGSQ=WD!E8=JkW_Vdb_4PLwdKK`WuB3bL|C4m{RL@o$Seko15i@ ztU7omO$3h7hpqH==U;sd-`YcRZzod1)(Hj)<4xdKAdRHJFNFt;NB^r_Z)gj;x7#@I zjOPLxeYOKaMJnFQi@BIN{ubfQ(@p`lVm7FC653*~i_;O?ZCAgh&8R;TJr??m7vDj+ zf_93^g2wEcziME2R)uR^I zn<#EvO$0{#8s_)z_8o`Mhx7As_&7gY1h<3ml-w~IK2WB+ld(L}GG;^z-a`ycyaC%E z*CB{Sk(#+u3F_eD0a#z5DpslGG01#Vu|>SkPCKb&H3jMN$0%5r03|hGLdt`(>eGE& z#|XdTltGuxA0qd7Eg3dC7|>b*DJ9Ayq;)FSiR|1WI8J!duDo&4XZgfoF`ng9*fbIT zn=mv@jr>ttE2oeK{V}%i;3bTkGY|jl?w`6SjzM*p3U|Y}gLlKAK&iRs*3lNLD6V|I z-A?sKc7K!b>nFDJkc8txUQy3wLT!hXg}hQJ+4M1&Y~_UOG)_WQ&8}Cv!k$`oD0bwi z-a@B%OB?%^@`(H8ROJ3Dxtko z&H_)i%KUQi@!8${^0Cxk98v#Ukg<{C)eWYz3s*+LV>{Zsg)6`N~oQW@T^-g7XsdqGi9ahXcFs;r2^ zHRy}BrkUugD(9yX>}Y+j6p@mKdnwPgx|@v0_p}wkU30ws;;Tn^k{a~;UkvUOn4@tt z7yy7flK(vk>TKxzuWKCE`qwqGA^nHIyOY@28u#n=aCD#8D|XM;AdUAudW)(# z6KQF}5vTWa_oHu59{`syE;>-WZ$7_z`y+X1uh+}f>){^zS-iy*pSj*oc=v}OQppnA z^3PvORhSvX_j&1_J?tY7EXedD)rr!1t@9-qHBHw^Q;laS)Eajz-_cgP2bK;4P?~>; z?=?j)^X9)%mQOkM1`9Zze$e2-`S$HcS|s4$?CdnN_I3s4^UDw*Aw1DF@NjSmS(wNn zHI9?}M=+4MUc)}ZNU==u&disvxT3c$3k|Zt<_zKTf{u;ciX581*ln!y?nJVMwSQbZ z+@Rm^BRY(k_8Ub|XC>{Cz=*$iqi|A*uT0wwz#v4%c;-NRbbAj)h%8tAc$}e25dGbx z2%+p+i*tH*Sg$jXliL^G%hOBnOi$jOXWbZDSJ6y zoQN14?W_p(IIpiT23g7AB+-7Vp9 zrZ8t5E#{o1X?8M94Yo;@szKTyQNIjEB=-0Nl0#bl9kJ=EgdIix+9>yiJ#IcnCIGSn~{)8;2IDUeY+ywM)ISYJ3)nJ=C;RU&K9J-RH+)O5( z3lq(x$b>p$slmrE-NYE1k{=+;YBQ(C&CXy@o@L|`#}h?J*U>Cv&7^I%UsNy5$OoQI z4ijh%+;fG8__ZAZ?*wlkrpD7IPDk=C`*=sn=UAt56Q5{(+E*j2A&0l(NS^ekLiIcj zt{*ghUuy7l~IL>NZ(@_VU1OehYrPo zJo~{MPEp2FQL!#u3X27G%$R>lc=$aZ2qCuiWYhCQn88|{V@Y2W+_-Qj^swg4jget(UieE9)*K?PYYc{RBEO z%aA4$(VeotDa>E7v}Z|23PNQ@UeXg&0(%NL7+w#NAOL!JF49^zL-H9Wnpdop!xdO( zC5F!SZ;m%8AwLpxYTEJxhI z?V3uPjx3T3ZRX=^#t#e|f}1P?bEJaw`z~j2cVEWI zVfhL1=Ug_T`-7a-HZz$IPJI!Vcbkt5pqnlX)DDB0Dj9UpQpic?O7#Y!{U6DwqyXqy zh+~mlDw3Lr0*Rpr>fc>JgaojEaGeCL2B2z~cMUk_EHdrA=4pd~R!g_55JkW~Mp|I0 zFk8nP0L%K};zPYoqL5QPH)iY(fT}!KX07#r(%BznlCE8X+eg!$5l<`Z-!zRYH8at> z2^wPw8VMx)ivuNAh?Nq?&G@cO1cv~0e8|(hD?%U4)V6nD!>gJJzyyweMD*SP+KQh} z@oz9xtwFq#51GhLfkaB0f?+z3IET$u^PPp$*h7a7fuHk^5hE>9{O{cpheZcb+sk;ts=%XzSu_ym5 zvP@A#<%9^vK^kX!1Lu@WP%=A4j)p+UF7wL@KDz=w?iY!UTtwLwI&#hWK>bf>qa%XGkY9!q7N+cOlZakU-mcT=!mBXiL{ zc5VPIPE&Af3uD!#5io@>5NR9jNO@WLse!SY?qY30--slEo^O!`!D}4&02~vCwX+k2 zODyaV;B}B&6=mwOzR3<|!wYR--zo*IA;TRax1F3?}Sh+495z zb7Mf}THHx$>J@s)W#gu-@=jdwg?c(V-qMWb{Zl82gs0?QXAAQ0(CzGdkk2d5}4%7*s>hDc5fc{vAb)WonL44@t`x#A#mDkp6?+R`l zl9#o$SmOB&gm-_PJ6KO6hvkmoRrV>4IHg2aQ1&!y$&Lf|#=pxP^3cR(K!7ZHGIP+2ZoPnR8 z?HAz_Imd~wx2IJy#Pa*9oJq;Joh$FrrbAEgY}yo8 z@?+t!yx+f!TvV#tQBLiC zn2`pfM6Xy*vlaSL1lz!&``z((DM!uxB7!F@(Kl^+y9-iy>ym|6@GZx$~HvWG2;(-;;epn9# z+IP+2(WS+SU7Oz^Is$|JyZ+}k+TM>HKMxw;Mp!9*+e-@iehpsZ{JMDmLim|BW%hG} zQy+4o2rfK3W~bNpI}MUkC(RI7S|9VJs3NE;Lf9jEbtOGTbTS8X$$|<+%H?W4CIW5X ze(U6-6{53zs%A23up1mn52wFUVT#8I1h!Ko|r*hoK=EXd$@2pE90pF^M z1Tfq+6q*2CieTeOeZ|$ZiN56*RA8;C)9tZYkyM#h6&A!p8Rxjho7yPLKuvQt31KGyOCDELn}yGqwzuG!HaC{uloLf13^o^SL4R zkLd^Ze=z-=4c#pNjS|!9w)UI<0H6Ox2?}k};g&40O@B|>Z5G*01JO;^b+9QHSgitE zA{NTON;tK4exI>PG|~^!UA|ffV?>^>Tib)3#K#`C`R_2Vq(1(W~8>D;{E<-TrSFT_~13QWak#&&R|NmE)9 z%8_(`dVN3l7iItUJD@4$TcUS&c_vr)cJ_36`8k`yim?zzU*_%mf5y}&G-?5aPaBmZ zJw?c!1-UkXmXJ%8)Y}n~3FU}2m86YvQ>jJz$!Q9?$_Q-;^)3wSRFsA6Bx1tYBMlFM zh@IP*I52vA5tc-fAc^5CBa%@{rlHUp7?{;*pTVN4)Ypa7BqdbLW57Y44ia*S`>xiR z>$-r)h#g{V1Mw3gD%JgUaDndx)R2e`;3%0-bL@;3Ln)IG1)0~78clHK1fTs|P`?Dz zrP%~v$5$LlltQ4;Q|M@P^ehx?F_S<^WE8-Gn3yC3;09%n0zrf|O?i5Zv$ z64(FWl+Tqu5|*8maE zV1md2^gsv#32CyG%m$>oc5EP!o{hW*Cv^xh0$~2QyjGPiWW45StC5RribUjYD?QiW zAH6*IPH#iWb;818a-640XERCM_mIa!u3+%Wm4x`UK=Kry;u$K)wSh%}L{ATtxk&g% z_GXm4ITtM?*CZxt>v<`+=WZM2_%QRzWp7|WsY=KI)bePC22%MztSC}Q#|52}gEsLZ zDj@Ao)gpSek*?y%d~cCpz7Whzpt>n$U=HrKVG|rMgC-ZapmZ}o*I`1_o=)E4i}_CP z2Ej)~iCh}>@s5#i`~)}Qyqp|qDc_zg=dkcOL(8S+0L;a#wfh?XmOJz2o`I>;~S`X}7l1(U}2vwc0Hup4H+fZ7TG*qa!x*m~&W@ z;r17@^>1OLqsgcoD&u3XFDI!L@|gWmT7z0)ywGr2Wb&*Y{=Nx-T8EkSmRf!I zN(x(_{rC{QG#4Vlsz-gP!6&leMpKS-ne8^?LN&0NKQ_z9-n0BBf8xtq!S{wi z@?Ri@nu2bVK!Ma9XbzyhQdbg%Dxz+-PpcZ>Q0AlcN7o{MpKJBR9>|d3ZlYc_ueid0 zGY}Z>$Zn3kCJ|1CMacT-wbjHZcXY4E&=sFDK4B>+4pxRLRWq+L32lc3`D7SQ|=zQw7&+OBm8R$!&N#V575>oEcB>`49eo) zTyBTledjM9$T<5G#~v2sy>j2p%lnW(mE3udj6S%ts!i9tecJn1Z~rDUWJ9@4U4hfi zWd1ejHMnkKVE)vG-DE_43Sv=I+xDg@Jy?kyDnWc65x@*2xQ{z3nh-|?|IvrG~t*=a<%NCm* zkk{MDZXmm>@!ueIaFVQUpiK5nSvB0j3T_KE(O=lo)nlCCjk!BsAB6Vhqw0n$gWdI% z*RYg=@9&Mb|3c+WbO=Af{KMpJLH_R|IcH-N`~LvVlGXm-+f$?8&k0Jh9(jN@7V0$? z$_Bk*psBq?$s~|qs>!t(T%H=gghVX;@9io2qAC@A>qqb$J-x}{-f3Sv*8ov(&Nx+& z*k|?tp!110Gp!)qjir^2vP+^%PtLM%jUy46>oO8}C?Zg00%r;X@*tAF$)p}^s#43E zL3fg%Tg^3@yZ!ayI)k4XcM&}T8ao}qB_bT>B?1`RtZl{T`&T8vLg!JS0EHjyp2FgH zb@cUoada)+nf^kW4zzM%A>mb3UupXNJL7=uA(l>yE=OT)mBQkXb!|Px^HPwZT zTM9&ThR-fM01!=)rdXk9y(8d_rcq~#`+}_eOfbj8l&rb!Rri@&YB=C9q=_8|g}jP4 zlFw`*o@`J`fe>s7n^|FXFB`mN(af~iArFqcRjj3a|FU%XW=EtZBt$>TQKn2@1#S0v z9N7h?O~9F>)i&PzC_9nKrYWi#y1TNGE2eMj{cFApa%`!$jc(4p0a+LUkpE8+ks~&^ zm*gRaNr5evUnwO^=^{JLD)noGxX=)uNE0Npj`wy7<~x$m-Z9btdwDs1{XG3VO0BHT zg(&q@)@rrhMN`a6(tjdSjhv7Ha-Ge!g~#3d(Ou_oZF`bUmRfVE2~Gel!6ZY~Ug{)R z(Ll$fJrk$;0sFwaek+l_NYUhq_C~do zue+q@N((OQf8I-)nX)b7j+*~g6aw^*(E4!=!HyVWM3%W{X{c}rc_t40^P2@>N=61( zgA7d+75>Rk%tLd#;f}ThJXbn1tMwSdUT3o`A#*Lgu#&t=Y?VpEhUts#WOCMVw?Ibi zA=uFw1Pp2_N~liahe*%I6UPO;tqlv)!V#G!dD;I-Vdolzv#<0l#KF8#oI88H3*7bV z_9OqZpN-7hg9BJTbibdFH)ue1h#3pGhr4yK3-%SM+Bg(Mv9Yz(Ka^1kZ97-X;8tOP z!%btuerBJ{D$nTlNs`MHUI|i4Nx!7hF#3nxC2+O0SA?*oMvmb|q&xO=?G;SE_Nsf& zArMl+qr)L~(36|5eKyR`?wWz#+GS9p`oGuU77&r*PPza9Dg1x{IR6X%%Gub-(!s^j z&ir2$+3wc#-jrydc+=|zvQU?Tqa5fKwH1Jmnuy@>_+RhDjs zH0yX5w?A97>y2Y~zF67P+n-`XjGO-Ur$%6|GK@MENAjt7LwcYHa4}0NNH=I90)wpNK?Ohr&i0g z4jG)htZ8|z)Vr-zKYV&s8?khP5i-}Wl<RQ)4PqQnpY7Ti-jj!Tf zu6Pyw;8?sk@$AY}0Slb2xUx~-0tm=WPy6`h#bghU>1ibb+X|HNaNG<<0qSwPuP6cD zXvQg2AJE)w2drpm*0j6jCBLp9tMSccoB(ZH)aVlA0&Q*qK;|oWH{#T?pKQ3LfM@J? zXxv+w)zaFV#^$R*j~$O>>{ea3TYax+!HdT-<8x74;M|_Hw2pMz_h75Z+5@k@g)ThC8Vh7Y|QaLC9183#w$lgiKz$Xg4xsb zAnpuLZ3*%dFENkXFHoV{ev6hs=K#pVR=`C$>mSKJ{I==V^BffyhlhtJ6=*$bYZce& z%i1gfe-aJ>vWJ3ku$raCGxX^RbFJI&UJzC=Go-J2UX)7t=ac90;zLxU;cXkO2-y#kk#1bG>!|q%oY$KfGI@MUjt#hX-FBUVPc)w1|K2Oyu=-@a6Tu z(AmDqhO^rnhU}gLXZIq68V6RI5jK$W-mM@i3R}($Xw~QW9_(BEZhbS0tmPx0{~+QU z)#@V$xr+iK01wF1|K(}JlCcAW_wKl&l4V&NG533dt~soN9^G&a#@9=bBjf`g&m#<& z$Y^B?=*7EwhQaerrBtdvL(}`oZ&JTAuT$D=R;yuNA->R*G(dbGrwZ2F0Z9!3 zrw`CzluXGh5A{N^uEHSJRw$HyjW}>>fFYZK0fBtyh+M#;Ao@_Cm&-XQukQ8g!jqXA zJfGHcR8UI&uyl0o_VaM&O1ru&6%fo>lulhS#t{rv!jX9aBc-6-7wB8aQ-O%CPNE9B?4LfR7wvX& z_)5+EwxvZe96*VsHu#wbFGM(6KuvRqpi5BHh;Bgqp^ny>>=G_!BRsaA=9T{5Q0~x@ zy92)hSDJsC(i5{K1Oqj)NVoQxRfE*)=MnAAB#p+<%gUllHKb}3ZS0JuIF4OZmvMn5VS*!ST#@xFNZ8bkmgov z1X`Kd(5YEGlGlXsLJQRZc(zdT^2|zLGZH9351NP0mVH{l?2Rkr1GS^V|1rV)VH4Tb z;o$`&{V~1maDjOYulneBv2BV2$n;)R9lHWDQ+Nuf10kw%0FJ3Fs~0^V`M^fDw>UhW&@4t7 z5?mwNTpTdZFzqdZ)A=_g#6K^-SWyBehoGxQv;I|L6QanC$Q3W&mGy`Ep!LLETUbx^ zCy>RHl0i1sl#u~z4<_(^#32ffuRdgHtOm`A$`Li1TXbn_?36EAiEg&0Co!YM^c4UA=e zex(q0xZ8F;80s1Yxa>Xw;Pyl8mki3T%mf+GbA{_jW@E9d23P#!;gd%TCLJIFyoh2z z5VdX{{m&NJ*K7#@&@f1z2bEz$vJ#i>Mitq~29Yq)nI}R%rD?2RACbhLmyys+(hu>J z(|(;X)VW;P{W#v{ROHKhf7~|USR_MoFi0O^3!YL~Nqo$RCn7KyQNOlvhHId}qJ$0w zsyD!Dmxc`R4G4^IsU#kUjecAj)TgBe`4-k(3MvPHr7bmu&DE2Y28#lDLbf;)~y8M5-3|Enp-tNrk4;)I$ge#c22D#!F%-uGQM8(iKl6A0z zaY=hX=wF}SHkjH1O99Tb_V0_bZ8Nxa{%HC9cuE)58MPp@^+`45Am5{tA5D8w+B%&JS3m?3%abqr=o ziMVNl4e#HA1HkPh?;GQI0ZUMWU2t0cX(%MamyXwwPhoTYfa4PYU2>y(A$hw(EeCx& z`2IKxVfki!Y5#Mr-`(E+2lhvX zP5zylIdeP3j7`6ljrbey}1kDNp9+Z^8@8l!eYTH%J z`Skv^b(fGMjKo#hJ1cI2lj;!weUV%Z4$pE@R8X{+H`?e^V|cY72b>pxlbMl%FSnGI z{T)XTJ)bo}_JRm9_Jyz}(40QereRf*O;Z}zUaI!@u-B|znD_fjbnf?A2$BR^gM>99 z7h)->;^_tYgE^Hxp%Njrgne*15}3CB!2kClg*dR~r*efj4PyNg5ioRC@9)}VF0Lff z;-MvqV$?_kQUFnw6&SIl)Ya#Te-55|afBt(kWb>w8w$CA1B82UddG)TbXim(v182r zZ~weIf3L5*Q(FJSwq=z}PAsPrdF{^sz>jvvIa(5Y2I3@rXk}Osnc|{2P$$^(3#qLu zxo+0LL&FFs{Z6OLzEM6AJ}qp>KvKiK47|UG=Plyq0?9 zk;udxkmk<5veVul^L4EE{4#NpO(aSgnm zWu4f!QpR@v<&vSLZ$lLnzY`3iR5Mpr)~-xSq)2(x-P?*x=H{Il7Km^JSE_mwEuCL! zGUg&gw(0h%z$!n~qfJ+|KKEJew(+MR=5(S;=X%B0$J9lF8|451s)=?+4VWI|mRGd9&)o zX=s`g0uHIg9VCJAV+*T=>1eoHlvEhtNm%fcmyC7#ZL<|hJ0t%4blkv5w$}};1Xw^j zu6_b6E?EtJqcHg61aM$Fi|>89yxXQOuweX*Q)MH z8mu$Rh}vx7t-p3=6Z)ik@jD^a4Qg}^1SA+VkbS33+Zwj1uAu90%&TZQ{j{5@IzOY% z(`bzw`t6=IEn@l7VE+)J{Mw@Qia)QH+1p*c9t3yozPo}#y&d2R=SP)j8%Dg}k>5h) z(n8#J=u*|0J=zq;*Qe>Q!&bt|I5M#<$qg-u-JF!gbz8Y80qGjE>xP}mFnTBk7zlRm z69YW$uvEzx+&6(6BrSr=pT|FURr-~$R?}x@WfU1!K}tnFY)1R*;;v@hEOm;5_RI?%s;Q2x5a{;^RNa@tHka75ieG-nB}-H z_W#BrE4ymi-jQe7f~f^BjE9*@wHd6nHLzHNM!-}t>5zq8TyCJ_4-HnA#2#!|id2AO`zUGirMuptle#{#h@k0U zLx*5J70{IiWYm`Mor!(6gv)IR0q0K zwh9GBqZ@`)Il53ANP@zNZ6wD>M+}te=c-fEe`kOMEv1I z{NqHtJrVCt1TUG=kGMYJL!UptefRPYrNMv$lR3ve{}Nw}Pbs60hTeo09(zw|iwEJ3 z9B=DM=Gr7_7LFvtlt1x;Z5(RV(%Y;!%>^&wYU0%t^h z#@Tift`Iv5SGZBjb2_|>ufC->k#?$V_fu>yu73`%qouczdU9>|gKMC#!%JbIHsaBz z=eUh8G5(>IG}H4X1FRo!ZJ86?5*yN!yLhw882Wn86^!pto1ueB@Zv@3MbfsG)qL!C zwdSNb7fWT7zxNvp?`+Toh0_eWvEOb9WFlo>a8Q8a*>IO((S=^6T69rBQHw70E_0vp zD{0Y9=jJVK{I zGfeJDIxUnL%B|&rNc9b#O-jG9udsFp`-6RPcqslMzW-hfPQ|ep!2bkp4>=hub~96{EbV9QY`*IC$8fyQ8R8_asvk6-OO|%JAyFRYgYhEi|HmKT}ZCwxwF-2!7L+)21#2oS^SqlcmJ2IrLMnC@tP>0 zn^G2B$s76ujx7D|7wL>9UFxUT)5*^K{E@WC1b^oDsWnvS~=RU5n3an|tg%vwKKFkElo+Sts2l;;lDh z@E?~SEl+;~mBiJ|_Fz;5_30LQm3wj8HJ?Ox=4{rYPCNhwDstF6Kj8*_2tnj=IVE}YE@8d~1*oTNjyEP4tZz5;E4 z_%SJUxGFF01qoIQ*u}LcmS{lB?o-`f*g!ApLIJFZ60Tq&{hYOa+9SasYDzL<#kI52PaAOGbGU%ORJ~7J z$>ddb{j|1NZBUPO2>_6L3SGA*4TsbZ4jkwPqCJl?;!lLU-GE`HpTcN3V|)E1&Nj}I zxg0-JdQw~!2?ivtMl{B2`xEYyG< zb7F5i9Y%jbV*wr ze0y!*TXOI#;6&e*NtTJfu+ub>JgTwch=&RCbgFtGnGGIrdx98CCr)3>e97f822I)^ z(QWey00KkQ@h}n&Dznhp)456gx}f19@qH8xupQ~Od|8=0FxRe|=6&0-k+yW^rwvLA zy1l?VPZ-PlL>FbZ;*VL9;6-cN1upcU4=9EtkFpn+yx031I1G&9dD5sciY%`Y)|w(FymQtC0{Cq zfh<rWt;PrE~5AgAoIS2jI z=;>o%{5vEt2m*}4rQ@y9wh?E@y1=q$&La$)19lCJY{f#0Q&hmk!VzN##Uc@~xs)GPRD?)vi6RbTgM&gn z6}~bjsa0f99HP^eoZndu;Us}`IPs)~v_lFYA*|&sFt&jf3r~qv)FCv!$Rze}79cQn zgfji8qK>x{>P&eZp|;dSihoEW@r-wY8W$BaN*^dOy4Hz3+H>G-ldpW_V`K^iL4 z!zZM;!nqt2mFLmS0pchdK$VVM7Vx)TMXRv~WW%L6#e$b0^WtLejLdn(L9YP_)Nm&D z^a*(|pgxa2aTW8Rkn}?YLFKhK-b{`oV&m8IzzVayGJ3>hiNSwpx8XxUPeF6Q{TTFu z2QvjfQx;ws^F#X`sJg(7z-~f>d55wode>6PQoC!#$0 z`^H-n?LIj34Xc$mtovEU=p9*fKtu!Sha60_m)!!T;#qQ$g{^Hgxf(n^T>!;qWq%1) zgES^d&>hfSXghHRsN);Lfsr%uGXpA;~#xR@6bM@B=g|P&cn&$jLm=xdbw71^lNaU<*e0 zBzP)6nYd|?SJ?UZhKY+hBW}g7(A8`ia%Iwwck?w0$|0KGX<=3LRS*%Df z)r=76q^>oFiC8~H5)f9>s37xBCvpg%;Q-Y&GG+G(Ca6U)%dw-2IC&Oj>qYkvMDU5- z95#Z$vjf^xU~cPx>b|g+v;~i5b##iSmdo*%DrhPgaUk>(w3$Is2DWuWidoKez>F?f zvA|@;qc!&LDhm+ZdoVCT>sUN8VBXrdP!mu9(-!6@GIqgm+_)b27JYKh18`R`c$Rx* zuV+Bih>Jd!L0xZ1ftbKSuNoJ_UTfGXS4F#qp^dXo!`@X3cpDz!8z>;Uf9v$#2zWef z*D}D$$W8$G$8M`REVtW6vkD_y?)AGEqiA1#>eeb<(P_aWV|e}X(@!^|eOYeyS`E=H zH>+o@kK(M}s=OawbjoLlQ0weRqtauGqFOUV$LRGtO?_DI4UOuBAu8o=ZwL%iZC!Om z>%C}{>*uXb!>A71 z>hz2%SaZ#u(P@_Jve?ySt!LD0(9U_eR>uP63!~YSkszQL@|tl`u6!Dbe%I*e&2qyK z=k-su=7lKtdgaPx1Il+rx6-M#p#{-sLAiPh(CyZGJuC*);+~4~Rr!;+D%aoR;~EgL z*1gB=fvg>32oD1t=pN+5O&S z$LLQ#Zm27SZg&;*qpw$9JH2K1@Yt9>wlZp~U_*lI%2RvYm7V05~Mz^)D3E$CIZ z(Q5TBC3gU|%FuebULSUARig_%Yre-8J5WdvRCt9Nzjq061Q4cP?HEl^K9{wwF549C za<_}WKJ6>5X0Owtn&($laaOLpCxr;Sz|)j*-V>eL#U<#Yep>)vHAyek;dTA2-|Io& zMce2AOLc(`gc0$i9Kf!q8fUehXhH8ffcOwQT}F}t0|5IpfpNRU{*7`I7ztTJ^vdsz zp~MVW7Mi~>IzdgzMbJkW5@U#z=#paL8c+dxp$~aj1u(0Dtc;BO0id@!#G*apBcKVC zAa)ETB*arC#+#fJE2Ix@M0J;9BEv{n`iU14@Y;?>s33CT^Yf#0QY_O0D-v_U{4&xD8@#lz~8UP*GF0K;uBo zW)+aTGLZLxZ+n;LAd4s*sKSUx`^_efO%Jy!=tCYm=xe{J1`=^<7!9O8b`rcgQhzio zJp*_O)KaV5LsB82_r@ncv4Y5u(Ba_1_?7FmDv3OBa{ofK`)Bo91&9h1YZ`TlHz|k@ zWf-Sn#XyA#1O~nkmquAiRoke4sC7}<^^kRIP0=*qakJG$MN_YT64&+44bJD6t$rQ) zE}(%bXbzBXqude3CqrEK>sX@PX$VlUy11;h3A2!t4%o+$*r8WsC zkyE0G)pTHKkm80^VEmq5fDx~gD6mF{!GmeA2G}B_c9eQ*w(t>YFMw4A3Xm3R;`#}5 z&RMI|1X*tohX4cd1PHKD)r0aV!_08rq6?J%A>J{J;W=msoJ4^B9T-Q@5~=}3{SI&< z(z#=_>z@Ed_fuDLB8~ppIsX00=(J!?F|dyy5j^_>LnF74P`3KOxWi((4r0K9N)yAj zA}~1vmk!VS&C0Ndg=H}v_^NzCI*BA#Kq;NDCRhxEb-)D>6AaVx#~~;`;_BMj8Svc| zFgwq@?Vuu6(P`EOxjS#6%D}hXPQ{lk;PGBZTGz{pbm+L zY#9U$XL&)B98nG=w@HY@^xENBw1X27k(Q+ltfCr>WVdawvIE*;S1@c>0Y%^|IEfL; zSE~W0fohG00rj2nv0ZOf4Q?GrsUk-Uv=DUZ6j@P+E?(9uj9te#H#(?&hUXpRF(uci ze?_;B6B~%k845R83~k_I;6qfgsEm-Ei7MD$zYf%)>913(pvaRxA>?4D=rDvF=<9l|QDbfZ{(~t36c=OG2VLY7 z1O9(iMjZqHf!1h(@t|{I;CL8`AyZ2aj6Y)Jo&i-5KV4a6&{>!z(9YC(>cP1iMqAmN zU?NvO0Yy<^!uu{*&TX)OfF)WT;LTqD473w!_im+MZ=+TNB$~ZI6&V{a zRiOlU2dJIFRX9Vd3TH`B+$eKgvR|pRz@n!i6jw$WjZctGFx$gqKq)~@!GuzDrk;l> z%EIK+oVAJKmPUBU9K^s)`Iz0AA&*yQg`vI zpQK8V)E{f0a|M|QVE7@?!j_6}#YnRpi1CvVOy)`-(X_;xh|C(YrCCWyGmS=vMo^sh z>vcIkOf%BiY=P0NUxz8H0XlSOG|PybZ@6Md`<)J2w3pDDK12^4eG{5c(dY+Sev%Wx zPpwu1=vYHlTY)aWACmS5S*AsYS{3G6DKp@Bel!5S?}?G7q{e}QXpDE2jYIV%LO8>M zvIPDA(Z`Yb8s>PkZ9#q4*qHYjGi<1po1e(K#VG>=1)cg3roRioD{u)B9Fm;o2R#dO zROZMxB*5ivv>CJ{upuyyKK251K&>^q=g4raYV91&={72!4`g>!*I>qBgAWI$(rPrM z=>n1hlrHztxk7pq?-D!=oDX1< zMcD(Rk30%OGjJ6moC6&dwJOvD|BsC1E}E9eFlY=)J4BHAV3eCITOhnWaEFW!pt;C8 zW&PX*7H{?&(gL|KK(MfjL#o(8zl$tMnvlWY6jHnC(iIs-hfR*eIk07Q*zGrExFr}W z6@~#ti}EFHF+iO>SgGE0Sq(Wr!X=ww==*Ca1Hp@^%(1=87K$HSAh)59DQO}FtOveM zrA*P7N+0;SF+{(DX1wx6qdbI3x`NIV+5uFe*Hhe$(*0TtOAzLQ6g{ z(AYa;*9=C6EK#!T*=`1gLh31PF0EP8(lGQeZFCyIml*j#o02sWs34d&4CN0aA-gQp zq==zi{gj6bs-ocN*f5MU4Jn4rPz?-?j}vSFloS}MRWcl_=q+@vP|5SW;|G~k*a@j0nkZ4(*2`@4N*V#`z)KaS zX@i(gHyLFFb|q?t3n`YY>Y>;Jc7xGJMNJ+hs!r_1kA2oHT@WQIK~V;HiQb0qdh|fl zuWA^i88Tu*icG5mX07ZF3>T?1Kyui9Uu(AkAo&fzAsJ0Hn>ZDHj$-^aNhsvCzJ$?EL^leMfbo0?NEr3Ch3aN zCI$7kKJ*piSq3;4!!FqzfHNQ`^$CVSd|Mxkicak(n8@Lhx*`2yCM`(OC(y>A8Nn#* zmBGO6%IH+PPnQ1&P#{uE0s{j2FgiFm7}bTDk$Nj3grhG~5mr@Xgjl16b{fT~pbee} zJ~YrQ<|61C*P4=@cnXm&;RULLqKX{B==Pg1S3oDwmJvQ^kg+Pou4H6ZA5tkqw}-ZG zr;9`l#_k;CSx8?H{A-Ny0-aFylu27g#X=hXve&2!4A6^uy$g)h?Xt1?kzRakfN^nA zX>#zyXF3=Fo$;oK<5d}LVK8q{@-P^GRsKXaY8NzRFNh5VoDVK4#Rm53{C> znNu_qct$xN>b0DCscC;mDKgHqLVZP(X&d+io9V)>fvldtNI|)QEHkuC@w{B`q8rpV zGHR;LsX-MCund*<$f)IK?~rwj);W=#r~}jOCy@0zx;NauWG|TOF|68efLG;5X*pa+ z7}9%+EJ%xvY)a7Y@72uD)&$H@pUAT!)duWlpnZxqtc)PBMFbNLur|m;4P`mKBfJ+k z=ty+JD^h5=(rSUO%5Yp!Z~D-f!YA;3F8a2OSUq1HWSe>?{JTc@wmeL zUbq7R_ivf@QgjzUF>!9YmL6KS?pkZP6;x95xlJCAj13M4hk}lQiSNHZr9)%z_h<1Bas7wy z-o888+aDYr-JG0mO^o5+C(fw^XCE{F?QkeEuX-3nBlzx#I(V0nICRUu^t_@8Sm7fS znJ_c*0{G`A@{q|(;Sd~_$bpcZZ!kDFv8{T>v98op0dZKl=`>O{EhKT-c4RGe_K$=4 zDj>>)IQS|c%9@Cw@;o4({Knt!>JCrrUj-yOlcz9>;kwB)nCY|k!82>hc`DiH)|!!N z>tL6faSfeGtgN$^aZQ{mQ`--0g6@u_czYV3C;&Z0 zMQ%B8N#F2{p1Dj1kdqiF_h0_u%*Yg%1u6T`@G%{D&i$9lJZ|Co;4heP2NUnyWPr#$ z(>z#i?Rj1*Bu+fAYxUG~rfR0lnZ~bw-aASg($K#h@)dmLO#X7U@n&g9<|$9YSAa|Z zwl<$De-^a7q%V2o#H26F^$R==oYmx(Pd|Kx4U^ID0e!+J_g@z0|BOAz<%+GJDVOJY z@dTIE-8Gp>sC4-3E`#UH3+R&n5-!(W!$XHkyp@SeCin?K)EOmAEyAa5C(d|-W6Wtr zLI94aD)F#I%8muHs~;op0x6E;)0`~l^5o6sl7E#{Qc{~9Aj(m zGEN(Znoa#MV;bUfd*qt#(~;?pXZF37(-FWBzwWs)=i$eAs^ns0Ppl&w4@pcmW3Je( zcZT$&G+Qs`$wSbDK_n4N{4jK~L`w3#yyME8t%WnUaH-k^(~C}ROL{YM&ZO)DT;iqn z@#Dv1ZR1%QXP)AZsG9RR<_5yF>YSgd4yw^L=Ln`4p*wmI8y8csYoWw5h4AN{Q3FgZ zYm$~Rj81F-4^qk=EWEPu@2?wh6FH9=p0JH(G~?7DDZtj|h%%~Kk82ITD6dVdIUNIV zDc_p6%c!J^BPKkGe{=MTzv5{CDkS z!h)`;!uL!+kPys;BlRX7YmkE2t<-8-(I>WGZqlhQpqTDWmZFhm@i!A@jsue3@6^-8 zN4|}Tv3yF56}m?@Es`e2e0Z6_rp|m~Me1R0a^(J;N_-?gfymrmjU3a(Y+utf^3i7i z04bA>8c>=(kkHi=|8ABMWi^qcncy+BG++}tpHSe0^LP?ME1GjWpERNKI7H)#qj@t6 zbfwC^l>%nNosgdqf{v50-CAqS;!(!Sdpg{MiP|u0g_oN65jw3&=9-&eDbIX8QUna| zJi)Xk#4zZYgB0LT0%_h^ESNwXi38~*i`d_xgPm#hVNP8ZwIaI>2@kGm(GA`%(|@AOi}?-a^YeFt#NPCnayJb#Gp-TaiDk+ zs4rlC(*m|yJA!GTv*vSLD~E>O1_aTyf8*?kbeu->)nMpZ7PCg$rs>HB4p=+dMWaWG z5M!?XTsXm@Uq~I`TT`qBiuMvT_Q)LHNxi`jK+={la-B)GeLEa`98>%T!1pL{O6kDAYdKsPy^tC4PEf8#3IH z>?#TuzeElq%myJ$!X4NhndjC*1Y^`n=|-G5yHF!CIM_$Oi#fv@DVHGMZe&PMFW|YY zOc6!cXO0MZ{!>oXcB)}A<5XxP>lW-IRR9DBESumCvjw49cN(J5bjS%$jnU&zPg za)w4N1DG9YFH|Yig8y`4-{a51nWVxZk%!?jk(Y{KxAOprz=DTchORpMLK%E?7@ssf z`Ws+|y0Ik97RDWCO1 z4J=1y4I4L4QSXr_MAMTRJ(OUOXVY6cxX(t^m@hC{a}i9xC&rD8BC^~3`IzI>RDZ;t z$!ZhwS|zoCszbyFM50U%Qo4B^%MDdvu{H~N_Einz`Lpi?AJjPPi)l=B`S zw|17{cidnGEVNIwdz;Ncb}b(yf^>JdGM83Pb!RbhG(fgU?zpnZiJvnoVInTDEfFs< zMssW_u(H5plN`j7Hl|E)rcv&)Y;a_)2b5It9=ZmA^I)ZYA6@d+y{t3^{EOUi z#sdvAF#K}r+L+!*^$E!-djQp3`;l37xd{^quQI`5mNNiHfT%Q@<7z0#f5eE3KW_PB zw8Bg!z74vXZ*Ey9ten+mr7^!`S9YCS{C*zhUq<$Hr2UQu-=*8(P03N+AtirZu00+0~odlN?HW%4MMM%mg6U)=FmAUz;O(98fz0)EZZiaL||sXO!)T zFj3bCMh<5-6zBzdO5uKlSzzrr^O5Op#byH{eJDAe6te2t1CZW%h^Q4qW%)y`f+0*z zvBbpXRKW5vbusji)k`V{emG$gXeUs&^7a87N%riQ_J|GjF~(@`tSrG)B^D;L8yNzc zg^Y@O&yh|Q!a#wR z@|!pW{r z5uDWFK|eM+5DUzbA{$DFCQAU7Z^Kn2y+aR(@cB?ADPPA((L7&l0cG)^@9p?HA* zhKRZyfU1a7o`#0!dd|q;1HPikMsa02yp>8S=4q10g2_)q%F72pdec?Xk^xGV2~@0k z8?GBwz4F-co%$jSiJS$h}%d#A>9jqh@sQ6M$PkqLsyNT`LDrznqwS=McuW~aZFLPYM zXhE75Q$Z+gw-wYBDB#UBo&_b7(~B7`v&owfopPZqoDg-8_5B-gW282r+l)1c01|7! zIKf#B?xY2&d^+Q*mow+VzhS1#0EH_6xFVpmIu-_oiG&cs%Yi1PN3>+XKt@BxxRwum zJ9GG!=KRc$q_fefDQ8&S`_82t=diEm|YSjNF`#R!9z>z<(zF*jwT>lV{)D^FZ5>G9kd%!5VT;*srd zvX-Rc&^zWqxgWrE*UF{VGS_Sv1Qr3@A7pUBTeGq5J%P--&Qh+OZI)q*WdUQ|8sYsm zeM4<+TVY?<(@YnX99uccsSWYJlI)Cj(9;DrTlyOtFkL1LuN~~9$*(&{qh;ip(8XXf z@*#Pe@02;N3gm7D7k1<%lSSWz8rlwQ1LLr;f6H_B!sVqE!toe?fpZrsSc}#ADU}Ed z`;kP3l+{)^3*1Jynp>%$o<&HIT8HbdW@eE+JC4JcmDTaE&u6`-koyK%R3Kh^QI zCf{P;Xm6YZ?kmP^aHt|0MT8Hk&0E^t5NxX%S-*kTx>Y1pu_qbN$6NM}MkEKCE4e`4 z;_j|ocvMV5z>r%}QztUy;(_zH9SVhsHD2tH!1&R&nqw0!bqbkHD7U!d+RU_}Hz*6a z3J1&LG^*;&a`tLzGHh7Xit%v}I8FjlmeE;%V^x0Z>f5dq(9(jfMjYQNQW!@+_1&X9 zNE{ucxKAv4=z}32DGvsaj6+-$0>o7KJZDfJq1$+HL+}DdX5dpOsQuNHr_7i+aSZ*zMB&mCj zrEje_rRgNsz-? zhAY^)U@h_#_B4>=b3t)AkUZtps#rD#r6&}OMzF!kO%5y$Zu=I2410yfNBLG8W4`>5+61j7Mv{VcTS7X6%~t0kIu~)u!fS2oVx*U$*7u@;f_! z{6n^+yy2e6?S5Kl!8Q*F7M1!h#Ld&Wy;@0B@yOd2OPrR^we;n4+_(Z*J~-~3x0Ob; z@o(YqZpB-)`#8>nY#x|ZGt�ESIgZJh+r}l{W(FZZH7mBX7M>k1c%q77OfVF_Ap1u>$lR2~p!oDlu(*nWNLrQ013hX( z8V2)b76*+II_e?v~C-TLWG0mKK}vb92NU z#8r@@Qg=H|X>*VqoNd&$d3rMCUrDtHcjrLzS0F?2MTg^VYNI`0JFBg>4KC%OuLCxM zB~W*SNrSK5I~bLzBcnrGEKlCDBpb!sM|-hZ6g6q?1EIDV;j&6>Cdriz-W^3z(CA{g z?1H~auuY;93+HgZAoBWR9z5C_z@ehh6B&eiuqV8wT|R#nQJe{HHIy72Th`wm7m$0x zih-*laH1!xRgLsU>g8>bl%-`&ydYkL4<6SF1$np4X$w~D*XR=?m+`v&611=Xh?PO^ z*gGrpSJoF4EDv_Mtw<>aZeR`w8H)Pklc z^8q+(2V3+FoE3h2DA=H|MXp@VUo_!~IM|WJ@z8v?#O1H*KxH}uMVp+tE=*ajit%o9 zDnh1;zaJyS#PEbqc5kznIXH&~C>|g6xbDWYyl|w{wa2rqvP~|QI-7`VEo#2TlUj4! zRO3fUvVl;f)|X~JW~GdNwOqALoilIpKFv@RLG99(J3+QhX{un%d-bAF0w*HZqDTP4 z1@|ZuY9WoZ?;iPMKWdixfQzO104LJW!C5D zX5>OcJIN13$U8nXvh={`3FVi`PH@^S|Hh>&1Wm z_r#LW=m{4djSOc{6!y(UX(aIUo411 z>FP=_)00r(5{B2-E~ZLm$EZ1k&!OkQUl0Bw7g{WV${+K74%>?`jrh z{2ZDY=;%dzDF7k%IK{ySFTXGg+5jIkSOg?jSdAmVt(Yyzr zft)g6B`R)W$lcOhI<@Ic!4sGgPl10hH`mB;9;K4d>@`@@U~G9%6*qg~hG6ijDBwVl zc>j)9(rRmLAJ0C6Qldh5 zdYs690hv%f^S#P)oHj}d|D>N4R>(dI3C84HnZ*PZ3+8A9?lei?i1!{(CazptHeEj1 z{)Pl?WP`CbkymmLu)Tgv-4Rk4zv06>TveUn;SH*`pKwtoMZ`0)U)|cW=cz_)GtZcjBM_2?gK}{0Fb$PZHig|MMM|!~evfBn~n{7s=G( zYEDn-F!~TnvUYiDPq7hRYr?YpM;$MNbcM*0U6& znTxY6CdKs*B3Zfl#Nx1At#jtE#p;t%^~_&#tA^ z+|8nukc@iQNWox$YdBW2-Es#%wAZ;vYnSvdR4HyS=@UlxPK!7Z@NqxCNm;Wy#NoK2 z4-|LtAgMpJ#jHU)&|a0^gm|PFA0I66>Q9jp@j<~NuOlU56i60Gj#IKsVnDD+6q5w) z!rjpWMmJ$Ddh)CxrLe+3A+A^6D*mhBx-j|5_7Y4=6HmvHJduaIqP^WqN>90_BM+bh zHhA`*ynR}Ma?lM9;2qBGJIi;YL;51zsy#(1^DNBc&tBkQZx04vI<{+4P7&ubgg*Xz zsa^6`XD~&_Gf+>B4_(Gd?$259H=ovW@qC>&R4`Q)8S?`iHrLf9H+F{OhQ4Qcu)sQ$hb^690X6tYP+sqAT_X~N@m1% zUX}j5^$T~g+54pocX$bd^Od-MefVbzrVelR^t>1;1CrluOju&uV#5hQ)+6Q#KubM= z6o0W;;P-Yv=qgTZsPYL&lhL7TO_N}XFRu?d{3D8kZwC+idH@VF@;oIVNfSN+YA`B- zQ^Fn#{K0GR+o$BUz?F1f?&S5I;B2dL0~%k`8!-Y`^iHT0s>%hy{F%u*)NQN^uf-H> zxcU4^*vp8|k?4s0yj_we>{1K%GktQThx|cZt*kN5B=<5#pM*!}^<47o;0Kj7I(wcE zSw;7OpZJtrvW+`?r9Jo#XL{r@eHGrfe-l7yD%1QAY$YRSfB4JE!RIgEi{lc03G1&# z{q4KcoBf>C*q53LLuuIA1Vm&naCtR>nWxue-GfLT-Qf96T;bxkgKlAB-i+vzLDszup3L~pS|80q z*rj}(=a0|&I?pq-DCaLBA|-}Ot+FdN5nNa(CPV{x|1@RgJ@N&;9zys=Fk&Nl2;0BoB zW?r#!>7<>&9qzQTSgbZ9U~Yk19^r?;dLqsh3G5V`L}N+e1%MJm!ywJW|MfD%Vq{b4Q+hUO4iIoXom) z`}!3J^JU(u^ccOpy5z^Cd2*@RiE)l>}^irt9)@g6%EL z6<^Q{Mh(E6#F^)DLpfiq9og90$%cy#^5cS&@;mZk1d4|f9IPk-7oQj;V%_sGGMLwg zG+Cs9#Fy&e6>m}vzE4c6H&4sx16Ca}D@*EO3VJ>avTQBxmg-_^bAI)ON{h0gt2eRtkqz084(H;m(!^k$Btga7>@vNCOrv;UlOXD}d@A_tk{& zxuFypSgcn=Po@V%duMpdd$8$xC6Cn84IZCRyu~Jv;p(S z3o~TNal;Q@E!4^#_$5oyd?`5N%7)BoH@@epGQXjsum7Pbx|lEhOAg zYjzfx!}NXJW$1P~25_N!=)({bp=IEyirC=9$Ws!*Lnr>w+YZc21Lmax^S>D-@A*qw znSg!LfnFJ^Va98UpfbZtQcOL|BG#X-(~na@37OAts=nVf_0gH$TD%-Q^_vVb6@&e3 zr*?k@d_TkwH>ibF*}#g^2yT{A&CXcG-dRta89(bxX=veiDnBC?EX0P}!NZ%RXk1$^ z0&{}Vv`LD;P=lpN{KBiB7-^2x$=m0aX%jR^t@<10u&vGpB!%#XnNc{mgB!qr1L{hptHS)O`}zxg1Ee)*Qa9hI8=v;gfB&QC3v%DfL6X z&y4fn#vS}*uP9Ij_#F&3^1ItNpa1wJ*ws_2!1SK>rSFm+bCI&_zhH+=Y?EgC&JzXj zxF9Rg_L%Il-^!gTST?29KYrf(!*}0fGy(V2zkMh6Pmcd`P&zst9R8IYuI#UQ)2<}? zv!`|}73p*6a`V#lbaJzOpW~}XMc2?)$JDQvGu(OzWX4y@(?cwg^+K}T(^d;Z6fT5! zS0AWFz*KnBP)`GhZsY!7?^HnxnfVzIzwEqIFeKcPh_ zj4g`TJIm0rf@jeO>q_A@59O>5ol|#Z(Y8i+l8SBHX2n)jY}?L` zZ95g)wr$(0*tVVAy0_hz^Ke>Q5AzRF?ZXq3B3k=nu^6xlc19$Re zYM5S*26W!~g#_D-Hzr}Wibkd@w|`GvCt)0MGKM7s&|p({QHDmrNkRV<*RDqN#9ZRP zcJadmZzCLzv@v@*qV@6{&lBD`U|y+8#5Q$Ncg0+>dKYz@Z8|Og1L~RV%I_Sn-U}socKy}JBiD)>O{7^6jc^kczPxNqP6rm zNeg#rIoe$hmx8lL*glteH8pW>P;Rp;tP;*@9+3H znZaqlEHcawUvM{0gNkTYyU8U*ct^FGHo$fDqX>8%APx&R~u{+%5#!mj1>|a z*+$gD*QVo0=hOa!xU@bpY*-{>^&ET-iWaWYI#$k9nYiBNGtwI|WG+TXv6#XymmRxQ;)M}%+xL0qTX|^X-N zizc2iRTt*a(a+9pnR)zL0rVAP->grkWu7^TVsi!OVWDzfu4V-~jWn2bzjGr%06yhy=)+jc3PuiJB5wI_5_w{7Km-e!~P*KGNs zR=r0BgRdrjw>co~wS4uww8h4@WBb&h&Js&~sdRB-W31d>_iDxSPSFK^c=}|l;z%Wp z!E)23OIonvQS*m-B*D}Aqh`@$(~Df!`8UDx^l7{VfyLZFk9)H>_W@t*ql^_~ec*vq$ZLo}w?Qfp>j?M%kJ?|MYRLKm zOx*%ZlZ3|DA+KNqAN~e+tUiNsUwok9>wCpfx`*)$md>Pu$#z=DdJ=>8nkDOzf_;#o zz3>moWitaV2D3=ojW57hUyDqJR`nmVMGDo+>6kBj^Dn;znj$9#5Dz`JiL@=3XujEA zlGTfs4|;S8Q2n2bde$qlP~W>5IeaVeRmm$LTtgo*I7VVU{%P`<_=7?E=x++=E^!v$ z-k0C`pHIAV(yLPfuUI3OqY)wI%5^;n4cv=<-#pjOpEU^Qc9{0q4|~D4T)(zV)M%;k z!wctzMLCoFfORa;K5cx-f6w+c_>FxUk~je26lhEXSTy%0%EVzh!UrAXWRIpL_nArt zUA1oFQH=hnh3qg;Mfcu!^1Gb_dc-t`t5j@Y=wt3o2Pdlnry|mYi-hj-4%g=m>{~nN zx0owP+1jD(eyx!62b|)EaJaXb}ISdJnZu&_LXhTb`Mt( z!Mp*glBc!%*^vh*uDob%o2!;K*>IgZxaL9cnQ5L;=a*ruByR_!gfP_4ejs4Tea++XQ879<{HD)h8FKXH(J{(@= zBfPsneA;EEHP6d6;mWAR&r{%mHL7KW&XQ)B7xWrmAlQNJrO>Ie?k=5hO5kBKnD4iU zEQz=qIc+P|ek$7Agg}eu z|0w?iGOmARyt`r&G+!1xOT@PvfwjO{l<%ixBPt}F)IRC8sM4|0%b!7gxvU~0aDLGl zaGWW+nc`P9JVf(~>lBMl9kW1Yg=%Y0G#mW1Y(9k0btY!Sq^ZB}go8~YJm;r>OV;9K zrQF1{MWGDuSr=H7Q5WOo&+s@HgY|A*JjX08!0obVFJEJ)sV?nRd2CS4_*+Ksk@W zGm{mQbn0y?JqIaF(G10qqrOBt^gOBwxHUIGQaa2@`GpjKT(W7G7cj&rz5&AP8lQ{U z6=7g42feOLH5ZRfH|NNq5T2!;SxX@FLnp)q1uQPH?nQ7g)^`#AczWT4EDD=IBO1-@ zBQu;4(KpJKYb(}RZeW5-$azuIKC`{G9huuEpnqb2|JeV)3zr5u8*@l0X(Rz(nK2x0 zq*RX0lrr9sPXVf3B~wd^ecb7CGyI9dCb9nBU+u)O}k#KSTL@W2V3*ovf6?O?wZv^U(G1 zkJ$yyY7IpP{A~C?1WN=kTM3c3)&!XZ_*&qxbyn8`1g7M{m|wf|1J%fOTZJe*)G@#7 z_=Qjsgr6iFP=xu#P+S#7UKlGQbj|$6FyLPa>_#aDq%JcR0eOjd=DjM5sCZeVKc(H1 zn2`ZK3n*Q5&P+_RQ$>^MyD^`?EXxYJrvy7Th!T;XjkE~}{T;s&114o#Kd2RGa1nz8 zNzyatat_ikn<6|Wj}h%F`Z0`23bOEQKU!=?u^DymOxAvThPWW;f9ZpHLGap)k6_#g zax2k5W?F5gGZZN;VYm@E1VMjo>Z=yG)3K4(m>sb|;zgh)gq<|tCUlMw|N1hHuj3(w zHeDA@0fBNH^@s2`TU$p}fvWB=(cae*a-$rL-G5D{Co0u2<`$}|cJV~>tz2A%7?p%i zZlMPAqSx|3wQiR$X9R`(X*vy-Y{wx5pWx{)R{V)?=Xaab@eiCTrA^+=6(}gmVb0dB z*|d~oColRY^{#Ml`VfsfZI{#S59X1&(RTZM%WE_te)Eq&Eq=q@EHo9A*m%VANXi|^X zR)&FTSBhqhiVeP%0L~^MTaPkc<6v8`VU_z^l^DljM zf);&Jtxw+rnWNaN=8i6qIc=wHg<1R7UW@yRdn9Ah1O^C2zs}VpD&{am^WCErMba7f!!3V9 zni%_y%?NnlTUUQh^eqQ1t%SaN{-WIW4Fz{>-U?BiEdpMsy#8ttucziZeNuc4|J|m{ zwo0tzLWM|K^dN*=2}G%K08;L<8#cRn%g2+z;PGz>oy8Vjk{8|zaAlc{}(OF zDA2?G^30f6f((MQjKR;FGC|iIyiBFd- zMFR*bvqdt*wLP*=oHN4fy`RGO?@N@?nS}#5AD(zk!{6}N<+dnRQs6I)a{2h z;slX82p0mBp|*K$4kTZ80pWn{tg!Qh&%WKQ|HCS(ad(e>A_?ujF+y^_`g|!awn%rT zjg*`)3EZVgdIqeR&&@IBfMtxc4$ z{+n_oDY5UHH~HpkW~(I+9mJiIak@LLspthcyqcNp>rFe*ASyya79ZMkSu#W0n|l0A zQcb*zY(9XLUotbnstV_Q8yPu6iVIzw*+8QT%pkMhb0hevFE;%#rsSq>+3ADUo3s|876s_B;*Oq&Wp>TJA6r zlA^xI^eNQ-YuGmX#CSY0i_jAh2)_RZElBDlE})-&A3RQ8Z54yAK2sHB-0<2l6I<)r ztyxKmLyMy-N537cEgy-`0z^V$4j@F``h2Q`MfpTPF=N_@w&?FNzxw{! zy@F4sQ)MPgKIUM6`%@*OZl}^s72kL$oqCK}XFj$)CL~|ip;;r+>^A;WW~dk=o2b#CUidg10(Rt8y#~XV_Ty&WpwWnW^%WLWOUw<#)sKaEQUA z1IZ65kHtgKQoD8f)(-Cp}OBOsZD(-$;(T+Ti;SHGS1zV?V#5heH5 zph;mSosIz4qv2$*fwxS5ymI@`sH|3DkMeW`J?EaS?HYPH!rcB(jC3M?GJHazzUpNS z>4AE_-Nh_6BX*6@1=WQ$h03;LS`wHA;30NVymIFbzH8aC8DQ7?ppbbDqv3VUB92lYL zZw$4%#$bkGZ5yE&?itJv6~>0w2>6bW4T-jW^(&JW5ftl4gGn28KHu>7g^oCOLT;ix z(A_s~zU{4E$|O%*c%dB@rz-mJN6gKZBLdTV{_fLO@Z;q86WDrv9(?g{+Ir?H$RYM& zKW6@ZW4V1<=$M)MEOy@b@Jo)#e~pR%IU)HPfHwc}oHU53+YCF@Fs01LlK$E+!&&6= z#fZ-s(Ltx*h#$`3d|u9Mxi$@~;fDUsgN96xtKIGfeSIOk<$H)LCCE4Ge5BX!jQzCs z0{_pd3VtaWlOO>A@MHdqL&e<0+WKEz9r9VZuCv!~-_s%qkj&Q~k{w@NaaE&}+br13 z6wAnEZC-WdU&xb@*~`Sy32C^<{Nv!JfxJLaQ`dAMw!~4PV8KBI_D$+@{#{@99Nb}S z&tP=_6Xrhybf<$b?DNja{>Xdx6ph`I*ISuzPB-%pl*;C1CC2@_c-D-R}83G$iVK$sPdfoB6wfxjlXi z^m#dYl*6nkvq+(A4z;-w(j#!6po&7pxc?PvYTsJv991?F^ve7mJo8%B?lM4&h!c*= z2sg;Q!yY@J@TD4QckTO_GrW!f0*IpuFFqPXnmlu~2$?}&7ctvG)Pa+Esym<>ZdVd-h&v5 z(tto!vXba`4ew6f@uH?Kd3MfEI1$#Sz}KaD*-UWESOOiL)ox$sp%Ckh_caVy64C3_ zxmZ|+AGxX9p2x~C0I0{vT23>Wpu)$z3&6J*Hrz1PXVoj#5ojSjHT1EsP04Xy6SztTe z!P}wq_5D&sy9qYyyk{`9k%CLyFmIA_R_@N9F9-yOBANDtE&%!Jo=2~E2IFtPz%-uB z?Xq-?iDDtV+V8y7bd)eFg5(eULR(-79PRIvle!+e$~wuE*_=z!d!b>_)tv6ns5+=? zd}eaWY4ID+D`%ap(sI6^R%*V#1~P!?*vVV*l&*0kf%sxX2>$OOxZbmH48nETbD_A1m#X(por;h^*SjX z>Yt}OGkjbhY!k?1@11g8oNx(S*tlvQWuoE!h4-1+8b$ud&s)#+uY50SF4d?rZ4|;2 zQVKP31CE!2w%#Y;(6I$~`x5qPd{qT7qw0{!2H0KkfHVIe8s@XZ@ktJG`1gW(H$J|~ zPIr)&ybpO1UYNKEBR(dTOXv(HkL(b~lLX&KG4fQ%1v2LTHX{NV`=O03#>U zwoj){UnUPP=Z8CE!?WITlSZt$O7OB0QT@H@~cXXR~7PkQ)wc4#c04Q+S z?~;oSG{s5gWT~?7jxd4Eh>KOKgG$5cmxGTtJax*J?dp|5w*dUyKDIw=_3Lx6fPBE8 z%0q?sy7iwy&2Fd4FnoD31R*5RhNO1IFo1DDILW!g#q|KN;PQ7CLWgbdNhVKQ7w%0; znB6-~Cm@O1w*#0H`+1`k)sqeSI;FnRWyL%$UwvR~I1FBiABY|Y+*>8*siHR6Kux7h zEgDf^U+P>8*sNcXA-;5I01@hLtnhSo06azHYQxYHLvq204(eGp2~;HA2jY3MX3`hG zIAIHMLgCQAz{l`!6(nAhlrk!gsVZWlc zNpl%suCXKNIAtEKXu@>DIsN9m-3=6G|$9|thS zc^+a#81%>VP=wp+mn6$CUofC`+t-A%T8jgp`N}DV@r*1CxF9R&xsT(R4Z^?xHp({u ztFFglafTcAz?<@CWzn&s#N6XXB^7Hs0dNUmA*K1E)#!H^Q%Q+}gr~mR`QR~BmHVrk4Zqt|(=Ay)pNnZzxR z%Br+_rAolqj<9g;`VPCnAgLSm1I7j(Odye7Lf_PRB0~~yKr%LjJT0k9(fuRAlE_mP zA|MHZq;6ldvdh@u0YP1=sU+~YEl=y`X!M6Atx*aMkjbS?dSMOS2^TYu+NFIe|IwL! zOB86qNB9wHef%h}{dsqG2CVtMUuo&_^7MyLlN+qaRl`A=g{djzb7_Lf=X7dC+W0AG zpN-$J1~8}*syl2MnGfQDSiV=oyF#}O=wkphfF$S=3eolQDZRy{S8Zaq>~Ot(tZ?1-+8#J6dkcX`QpfqAdl`p+l!=fL9DkrwuRmA? z0hQX0%G;Op8vn@rnRG3-Jn0n`C^SqxNM&Yr@w%gtq*7B%v{#duC3^njX42JMWORgs zx7DAAnJ_mGln%yl55;=xwup4kt+(kHrLZ{{jzoP~QE9AF41Pq}L>za>RYR(|aUGSYQf+>?%?yn(d zg%DqLh+g?D5bA4jXtNIy2T6sLXDlH6NwWXr)SEJOIUBZQw5)O>Xt_ zy=hfI5Ezf}#1xaDBzD9`lUUv%(3*Ejr#3LhNWZ}N29Xh>Y0wvc0}qwpXmF8@zrtv< zu(w2*<(KkXs@0|CoKiuw-aZO_o%g-9BL1@p|3;h}#tg6n!fW;7klz!Nr7mJr4W8Y- zQSx8|a}wzZ-?3{|dXpp8Vq5TQ768NP3JyicerS!8AfLHa7ZxH``v^J=Z7dEQV*MLz zOry7bYH3aFTm*@t$Y2gJBXc0$ywPQa61MM+7tU=m`Bixi!Inb)ZJ@Kz)T2=nMFl@i z$bwA2@JAQT56zSZqI9&t{84%ndP<|x<0H&e%%5*IT57Thx7h63J3Mpqc;AvvB@s&= zW(nftpOO_Ik6aLPhRfQ*3T1yy8u>9(fE<$Rv9q!T1%%kSYZQHeGcd^zVIK1ZCe}Ce zygz3!{YXJv)!Rw*(gmFv;ZseML>~!~xKI_{0n7QzDnB%sQGgO*1m_yHG9qOQ9p%f7 zN_Ok~KmpGd!XNO^8HKdh%*hEoW0ogjovhd@e)eZG1np#?K`|FFad z%Fx)b9m2nIoP0}$D|J?5Xf86I2*uq{-q}FaY^qGn7Mgp|#6J^_^$Nu7zw$OwK!6^9B1{^T|JrQXq1D0+A!rw~u2{l2-; z^i*nH++{IBx>#iG5;}C)^F8xK&2#<@f*#(CUd&{am-UOXQ4nZ^h-j6v+ah%7w`Jp5 z4jCqcVaWGT)VuBipsz8FL98J>&v(G1WM#(XHLyV-#$&Ms$SCUz-axEfu{MbEf@M6M z`%kORNflD$dTy^8h~ zxrUbg`lLNt_kDW}-1@vZ3eC}NK=b{&S;^V@x_iq|!>mEa1^?@?(Exh>lvee^B)mErMvY=f>-6-<|^&hamovwEv0a=z@dHqmQ50}k3xDKE%3d6Dn7oS+?Q?Znj`s)>l| z_yf}l1hZGU{-fyuRNJ3S!vmxpJFi}-DfFZsGQp1R}{ZZ!T&vA1Hcb3r$91y$b8Fr-{pWsq3Bs3w%eV+ z*)!rm6md=_wpsa{?`Fy4j~1uw!7{|{ZX}CdU)s?SgqhAUzVO?Dtph0EqSk`a6Khiy zmS@`Nma`LLgjZykQi0iUsirD0GWYMv9~iY8Z?+U;g&IIFcQ*=L!M5PsuiR`S!n*g7 zAx6hU6bZyK-l#AF5+pVBwAB5HyiRT$)rw?TeBu+41#ACblmEO<-Cmuz`8J>4+QYl{ zu=#7kK~#H&N2tTlLa)iwUCCvU_ty?J!y!fP+JdtVH{$Z9(5mQhUiqLCTPbVqrD(wB zHC0B&S9zuy%USc~pq%H34nlzRa3kVRe%_&(T83~@njq@MKFCs6lm8#KsAKcTX-}2tB0lAL`ZbdFz6MHV|Uvuyx3UXA} zK#>r5TvzU0lR)mtnizX{y#Aheega2-SaOz8KDci%+dl+b0z1B~IedT8!{U(rll;*| z*ZVB}dI(N*^~(-AqnC-)nN7jVD(qZ}Hq4vQ_w%f}%G9pa-3ho=aX(P?7ZWjkHiKFi z1l1~+HK$CsYzF8iW=l1NJ;aCy9HGkTCPz)vARA}|v{V8?f3FAp6@6zDX_A&-BdV~0 zG40NgNV^RXT~U6q4lVrNxYi~OytGV`b0CYSr-=DRw|DKt-1E$qU0(LfzYm>O8K6y zB{S0$4I_umtJK$!9~Urr*vmrlsFeAEI!$5tq|SL&{WiS#c%Z1Pyg3?=wCpxwBXpXB-9?$ z$7>T9x*MTx9^yhF#&OTnZz{pbP@@7nVTA-1^uf|;5uBWdyh74qV!ELUj^_F-|tOGKYf&imy}Ko0LCSm1m2RrLU1k z3)1|8{N<7KtXs3h;S?UR^*7gC1-PMT9k8)-f8}jU!i!b%WR_OkK=4;<%y0~K#^RUa z9uyQG(l7B|)w}vy4F^PUb(a*8m({3TwV~Wuq)&`^;KWj>T@4fTPg55IRuzV#qQABb z%~(NVL)n3$J41$TJYI5551MxR!dFhc<1G zv;`KTj#A$E)lT{b-~N(8T5bz|bcnTire)e;&9FCF*chw-JXk zZx$Xs5cb3n;(dN_@^DM|ZR&Da2A|4n?R~R6$H^9PnU=rv{L#z zgV5`?|IQTPe8tqnAOt*jh4{psf44|XmIPa#6_Kh=omYTD)c?2V`VI z%g&er={d-WmYa#kVkC9CUXF19+DftllV@d#be7*DdK{SYDOYy}{L3^K-s8LSdyAG< zdiu_bVZnC9q}rVFEe zj`4P^^VPZv`#ig#1(S_f!lt}N$4IeGiRKhx216U)`a*VrnOs8nkN$t;(tl&W>f z3zQmEqXQlzRu2TA_fcirK+C^U$@3y zl2bNyA*176LAKqDr{FCclr>8WuSP8lGsw5YJNne$6m|qL-2=V~8LXFN(G<4#^b7Tu z`h5gE0V5UgSzY|y;ZcJ(9X-9`8`4XmAr{OZ{KXN6vGomuyi%GhIQzIr;9 z78v%N@1?S^=@r&jj(W1hNnYLR!=bc#%eZe=Zq?+>PBykAn+^qI87GLS!kyfv}L)?~1u;dG4M zHp#&1!U92~GmBuo<%yrEWA~Y!Ug4jc`sPiL51#}qjVgjg&UI{ZnjtFJbbB}9-L`=2 z0Dio6c-Q>>vD7KKv7?6}TqA|mZIbuNr=%(5+L?%eGQaQhyC_xLdzF}?gD)adzk*_BWQj?1q zpzN2Me@%h+m@6gp{!yS}Z2ztnCDk`HV9otE99nM|W>(*B52Z z`Ii15Cgw;ABgHR9%`M6g=BU@w2U&w`LAMGu7<{kSCe~Y|eyeyKTE&|Zb59{AT6;FC zai=a~^~Cb%F^?_^eNKX^YY=|Vx3PwSt8CLIsr%sLJEci+fFM;HNcY>?WWZW>0mFo=W#Hq1v4&jDt4{KZ za^V7x?zB(D?T8MS9P>(7%#buQ82`S2e3}wfi`2@Wm>X9fSK9EfdS4jmn2` zP7Z*u#rOscq2St27&W}0pQc1&Tdo)$1edbM?E<|Q?w0#f10#gVsZFf36>J186C-Q* zMRq=4La!?8sNuF5+RiofNk^&{swVRbvEv!#C`o5}sdGg{pO!u6@VG6&s1UlQuL@e{ z0_R#eQCZ0J#1%~xmlCke)cQN6Vvx_CxE~J_hlO-yT7-EbJ*40lQYX{@h`(o1Fw4Kr zrnOEiG<7t7c1dVVKdw!+JM_q?F`SPzH88I=9bW6F)+geMk&9bo(SWl`Vg{!>#oTE4 zz=<^`b334w18=2$=yRIPm&$+W9+;GCHgSjME&g$7^s*i4(?8|1r|Q=TVEb`?Y)mdI z_4Bdd#rmeibU7Nh-R=2!B*S+Y@{whB@TW-`k5iQnrSai6*AQAVK{p$Om6s2`k!|Y- z<_WbO?Wix9_fhYAF#GI8HmCj|Wz_s5t2E3^VS73_|De~f@)4oIC@QoCS6+ zhfyi!+81`gbKA7Kiyc78gKg>%8|$RoH`FMEsWpPlZ;fTSxY1#&FV!Q)dI_F2GRq{L z-7%S=RJas*zmqa4V@|i#m;$209eay-Jw+UE9TNQ(V!!YSTB(^4*^EkFUaXat8=8zF z2K?y0lu-qxnv=jD)t_D(sKRk2eeLH)lwkY9m!Uv%mJ#@(|2DoMZ>LdN_iSi5d#u z2XK9G90__(<569x$vb;p;}8&?no!6jIvW2 zu%@L!t_7QRb;dw~^R}Mu>bPAY+wG*n449M&Ge_D$V>5~jPsH8Uhw!=4{UHD(jwkMb zIBDPj4C7cP;fxWGsjv&^jIj?;a+{7jV(>d2>p-!Yt7eY>g0YhwX2dZ}%X5rmE?A@Q zJ_MyfU-}t3=wRULLCXLv!3uDE1^moHC7e2*cCN_Pw_Rv2p-Cs4e_%L~ z@)5Y?g`NCYIbOKuWZYzc=pZ8t6JkWfC+o%g^`a+A)VrU$dDhY-j7L3HwSjU<= zZQ%D*g_04P%%NK&8@mLXO*$dVy4afHQ&Gved$mwv(53%a@H-53?!4g8>V{KLAMt~S z@gw&zhEleX=uZB>WkRst)t~}KVM_~E!+Lq;c#&iQs#$?eH`+ViY?@xHcYXz;--@^Wiaj=Rls&8Okar0(o9itB)qzhIIEL z$2r~oH4?}wCw1KurXLhn=)fSiSs8T|AJn6)#7s-(!A}~9_=Ob`XnHZ>F078ZK^6VO zNcQqoyLP^g`_*8We=TsoTmlTW;z@PiR=7f z;5ESuxOAT!sCEnL!342GRt z(LJe0wWLJp@`ld907qtx;n>1=eJjHr!qwDwA-+PHW9l6HLwNedi6Jgwa7T%9g#;BT zA8-Ef5asxJ$PXM-3Hl=_+VwjJ^^*KRBPOEOs+0K$p32K)5Rx|Z3!$OYDlnFhgh(tA zE99rfJB<`l$>(M?->2B2LTFtKm9H9;hM$cT0ycV=LrS-XIK$N#VBB&!h(hPh8wvVU zvlU672Wq4N7JAdcao{O%sMLa;(b()QBEsV(!-aWwp_p~SS!CcUaT*na<8DN|QnYyL zpI6hBXiO&u>6~@7c>$r=q@K7`!AFqgrLtH?P{h&!ul*xRBvyZNC9iw`4Ju9LVCaP` z9>z>D;4`|?Tn06EZrwV#j2Pxfn+3L7ehg{DQJrnE!An-SImr+HeV*R}q!1Jg0-|{_ z+vMm4)|g#?as}hc?H?sfMI*FldJm%tywdP-JVy!#5(?Qr%M%Y9quNEMv%j&`D z>SekD?GWyq$orMW+X5$qO-7MRNy((OFtzV<5-;vva=*mp*%-U`=W^p1 zeOjAEb2nQw@=u#4N7_L)j!p#AZ2fwEZd)KsxCk(?edK8|yaAE)K9_qVQGh@6aMDhn4 zzUkKcV`U2Y6z&S}{tjNGBQ6La-CND~Yr_g{>BLC?M)3XQLG z4GC-lNzI;rH@|$IEo#^C_|`!TBK@$+Si3_^pi=6pRE9a8V7wDAtO$4jnoQM1Oxp-Y zwCd@a%A=LqO3{-!Cr!V+D%Bu{eo}q%=8Xau-&T65N|s#M#&`9;vc`9>+IPK~!zCeQ zTEytIAc^|)#{afv#UK^8Z;})!Ni}9!Y1hjG73(yy!0EOxB{yfEZBk)#1;)wS+8h=V z=$b(u_~;_~y6_s)c%le{eMhl1!fNr1qq&933pP@v&8n3}%@}On`Z%us;W=%$hn>im zo5g!TWM|s#rTSLNfp*j=wQ;w57)!8_PRgPij!K6?vs6;M@u95BIbGP)vs-j|h`jc3 zJ_`tRrP)4M;)Ocf?0LBp6G{(-R`SK7d*rl!N#O3YVx6*aF)fra6uGJOD3gMliXFE! zn(CISxA@2$kq;`Og;<*qL{2}DsI3jg@Ao^pGHad)5#LOLEpGUzOx(Wi#h1#Rv4`md zOt?B4vI8W?kCmj@2OBl_zIOHH@Q=5?Ts|oEN(EluHt^Bc6LoAVN}INP%TTp-Xtn%j z3#<(WYGOIPc`@m$N@Bk`{xg2ksTrFA{sRD*hyE||6AKd;lYjA(XcZgaIy=(Wm7Wk4 zm~lt>MY~jr2Ufr%`1&jx?b$`bB^H}=@~V8Hq@rwR{re@4qD#VW1by#6vYiW#Bc9m> z`BY~}o#QsTevK69c93g@Ap;Nr$2}ryl&uV2bQ;t$6#SW$t$=S$`HW>`D=R5h0=P=W z{zAM98WnEB%75BT@Ur3$lmGs6mGHcUfoqZt0{(1&x4qq8&m=U4AZPlsF;M_1 zt}VZZ^nlxXvRjaEEVYS7#92^>vlv;k0~8hoKxc6!0Nl ztRSy3N-)Kfg@C$<<23Yr2U5nk_C+R?G;43^M@jjc~2NmrJQnvNuFAUvJbgLq3UWx@m+OM-SH`DAP=Mhhd1&+%n zADM3N92v8F_m8kY;um8vtM;gQp3|N<-KyFmA1?`fDP=2u)YO3(#YJsgVSn+T+gokm zA+#>!p<}4SHIpc?sOU%7j&9veL>R!nK2u9(z#^6~h+1^9F<4@)=pyvypVCR(9+Kn-5T zxBL7z{N%|Sp8+`i2)i-3ksI=aSCsKm6?i^9#T11aV2>SLAXMvAiW=Q(3^27;N%UCs zr}411E!8eBkn9{HODsUIGD7fqzcaq!3;uOGiUI>78u1N zHm%iBu@+5y4=*yoA~P8x%~n)L9+rt)AOt>dH-6*g>H)S&Z20`XS<^X>?F$(s-#53h z?$;j@)E5Y64=nnhoUN|cHX5~WnLu(riFbjY2^U|`r$0K#)6GD9SW_PtzA6RW&N^bz zo2Bhm`Nb_EG{r<0&ds+3V$=N|q>?Mxs505anL(mJdS9L@i= zj;VAT#|(D7-tBj#(gT##f-6&P+2Q&ra4^aPZj0YNf4c|x z7s&i=tba6bbW(K{|2Td2T^t}(7E`eF)1BKl4gp2U;P(*ZWH%xlR@%B#aawX)_<0UH zOblTh(Tgu47&HJK(vXHJ?S~jRh9~~$D@DSEDme|<0en`*B!}nR{?&9%8ipB51 z)sjZ-N5si07=^_7eINTw$iJ0VhPCaefK;=B zKD=u^!L>7$kA!M>D=5z-To7v>K+sAH@zWq+z#R>D2mfyF2g(@)a1BMY-xG}ve6tsu zrFY7U-I(MGs`!r(UuIo5L~9l#LzQ^|y74%lMa+grRTA(UIgEyAd6(!AXA4`~N z=c~d&#|sg)2>c4Mz)|HM!GKqGF%jwS914>~O%a$0*4|B!opUarI$WwqN(KILFkbc? z(P6NlK74MIEk3?s;`N#a1*jhe%IcZSUKOWHR0S} zny-MWH+C2@(`A50tjR?}a~EU^71VZP?7Kt-$F1|lH-r;ojnj9?y0Uj{tzmSIFqGhv zS0zv1=(Dm*Jxv=8Nh?@1te92&27AQ)M-ZNQEJ@0BKMnnpXw;ncRdMLvRf6Mr7mQ_! zwH{7^n_&$BkzK>Cj;#)`__XX!^JEkC*Z1soBs(atV}39qR*N1QB_zz1*$1zyUbPol zd2?bza448pg>>ewv(i7;#CTsTN6=H-7B9b{$JAOJSqi_H6or0b)=h;?RTM5bbUH$g zgz2bSAT)Iuz;kNeZ1j*QNpR{2HzCKbkou;0P@_Gia3QrQcJLTmIgHkk(;TzEqvqn0 zkVWBa*9ZV?Nd6R_laHupN%Q7+%PDB>Z?Zt^3lPk>+ z<&(I$TmP(Ere2iu!?+UWM3RaFZm|oR8MacPdbu)tEkY42Odmw0z9Bxymjz+mE`VRcK8F< z5D~#!G?Juv1zjQzIUyNrXwZ_}V%)|DR8k=9j8GIFAzu>a5r5GUn*4=Lz0j8$_cn<; zI_OQBa01Rju~oPQA?AcPqh}y?KI2 zfq(a4(vQm`W}rqIi5z##gT&$ZSbaycgn42x-3IMnN^rpz&kY98S~y)T={T2Tr@(O- zVh^?x(sHub8N=OV9VpfZo_x3F}aHfU5OeEjwn*o|yJ{8^bPUF2Y2Uz-;v#{knbl2`txhZN{EUJh?u zDwp;7EC2Jyuco)EbI0Q3?J_=d^kMVNYZ2XAQRZ7d+2qmZ0O!3bXvcZ@E0U-uW*uN_ zr0spq$=lgk4(5v5%D3P>-yUwjmYFcIZ098sv+(Tsx7WuCkZ@Rbe7FP_g=sH8X9PI^ z=n4e-wh?~k8*89kDG*Mx363?W;Mw9@lC($nrxqX@%VZ~_h=>LT=g^HI8LN*NpwtoH zXt;;Z;NrQx{0_(*_;oiJnmevq4L=7&4(B5P?eIFWw(Y7_w(Ym}efNIYt)0{IFJ$%^GcraDojad+Qg23*sUb-At*?ZH=8AGOlhM&Wda zrYM`Xg6#TnZG=@xV6pWOX-Fe11;_@nl4LySt<1l;W=p-a_?3%k2B{csKbIrRS)aus zo@HqJU&xwwsTK|dehBM5h#qEY>&q3h`>Rt^pks)u+x!E{ms9GeW4p`KG@5p~sr;;y z?7r@?H^Q+3A*g>M(ewlcg#11+_ia%m>lTSeW zwT4Hh&}ft4yQlm4{Wyp_2e|zJz@Gkq5;XY0vOEVj{pQA`5{SbQWldwj1svni@X+Ih zoLegDt@6?s(wwq#Vm#9~`no=@!oC-_XZbq+JkKZcoPFDPXp5`;aJmVP8k@tUwB~$B zmL|ORORSNe&NDNknYLPHpDSXp!pki|q}ERkAmk&4z-xi}%oo3Uw^glKXP{j|NdkAa zdX+7jrY>y@;yv$K((Ss%VJ3I8W7qR>8RKl7#6A3W!~3Wwe=r(%#@>N!&DgqvFo&iG zZ{Y3381RSYzO#*^Da5-*WSB@YM$f7&G0PBWg1sMroEsgLBx-qGxf$u`#PB$k1DBm@ zA=0_oPwMs#;%5E1w8n){j2>o=!LCx9N!ucPJPczR=2&)v@bj>cBjg)(xo;PTvDtx& zB7v^X%l{Ma5IJP%Y(SesJ1Lx$G5D2EJ-d$D4hKXJ~b*v$S}m#ZUcxYd4@AeM4e zq%=cRL!<49&?LwWXl5N|iUB|bEMSr@>bOa$lw0X)mY%HMi5ScppPlRVd%tZotPhzQ zPy!N%Or++3`Uj$%o`latkP3C^(JWnU;q5f!BO7eiTP)GkW6%*f1DFt0t0=cBz0}9f z^xJ3BZM-n^+VqACO|qs^ID{@OG6D*_rg?ezZ(uMK1=>W>(7q}jR@tk9d@6+KjFzmH zreo(i$pEnlYo6V)<{V^u0AS|_M6Elh?vCVb7m(W|huqxn! zqxjMiP-C$DxgfW@L949rE{y`bW38$)=f91yoVLc!d0}{3(FthW{`)c+EL61S?8kAg zLb3-#%AZxGAoQu^t7%dK7I?}d#zlvg;UkVQ0&H*E<&!lGiG`lMJ5=gbYc=Xd9i*lA z3bJjyMOjEvfA!QZP7pn(du)#itFMI>Wro&8@8*XDUmDkfs4G=6TL1;|q9fBhb=S^t zodH=pO^~#R(JLkbXh)>B2;9g`n`n{Iq(f7ZdAHuWOqSURC(URws56QYPAh+E^C+nq z*zy5Uz!XG|5>)Xfosg~FDN4(d#0}5-3NBL1PaYWT%N17*yXk#=)^osVREBTmZd=#U z^3dR7d(Iw&VA02-zkPJ!Ck!j90?Xi74CBo%YR)PI>`BVq6yCL>N3DpPhLRexo7pg| z<(i#{b_A%$S04gT12fe6YgvVBvH{E83?fw(TwxmYIOua3hk~XlT$LSPxM&@j9=Yrx z)|pn#0f&+kjvR`Z+LY!c$5q2XbH3Y7wza5p3u=v(0LN_4kzqwj`a{OQFl!tCBzDN7 z-(fgW_>MEtvPWGgE8Du60at*}jzMbT<;nj3V9Ht2U!Fi*uJL5;pLBPbx%9t~*88fm zfz5IdL0_{ah~$=r6f1fE40&sXbkSCn(pLAHimh{LZMdl0Dr zyQQB+_3LYk*v{aCfGz1q+5^3gNZ(6s|H&JE1%kd0?k8SA7Iq{|HsvB~uUth5kCLFZ z%iUPSNl#l5Bc1zWFzT5k!oQf1Tq(vMILl?H=RJSmY?7b9@;x6ffHvOpF2J4e1Nq;F z`{ zP%2KaT7{-Y$>j+u*zB`@-f_uDQ?0FL!fykqH#Z&LH=)HkYpw{@RypX+%Tz~CVN$4- zLFX%*l|MJSu`9Zo*3BqL=cuB}Q9@~_>V`#eN{Th4Q|eqo=C5&Q9!#E$T)8;Ef6%O> zytl-}GZTpfB9)#lUSI3>y0gOm4OuH`QDGJ2m7T`?Ts&MIczGELo2wv?s-CsS$227u zu1|qC6e1}VxwDk9r#b@^I8$Cofu2RhXKV&r-jmbb8s47Azym_BI_!Cksi{fZiBQ*2qsELKMID$F4C2sSh>j1pTfVh7 z=rPH~pqpyKZCV0Eib--1SE^o9@C{K2L^>y$vM?{i4xJ)z!wBjAs6yM0AdS_D;ZK}h znmjq>hRnr^H5nqOnkwkGn#?mWIK$6uk7fk*I&i5|D{128reua%i#i6}8P#0(X$k&L z15>lAC}oP%O>TeDcMzzhMKrVJj0~*GlU;z?>oj~hL5H1g*3oK~(u@VEta7i|#$%C5X$84Lin7BeQYlU) z+XWjdDq5b_0#41*v|!LeF11UK!F3ZlcA3(J&T|Oe2M^Q8AKZK3z{TDTr9${ub(m$R zEW^Xuy34@B`b8ED%$r%Txp=Z*|1~oXRWJuU3gsF0Apaq*=JT9TM=6u}=&!XZ%Ce0S$?V5WQ zD~sPTLFUKr$tJ}l0qY2pWP6I?DMfGc;K|SFKUAi<_yoCa&Z|zD2bAy6x)l4aw@d?R2#FZ+}G6$SU3@ zVNe4m9(dJ9-tUm`7SJ6Yg(2zh#BespeZg=y!1(v9`v%DFmw|^JQUpYKo?e!atD4HR z25&Z4@Il>V+y6HQn*5`y1^!vr&;K}p_CIpK?LWq~nyvG}KjV6~ZvL2ym1KRw^kfrr zlf}57#BRXMCaHM=S7Oz1}JvvEVy|$&0M`|MT?xA9Cmort1D6> zTalCmJ5ecAOuqlOjSGtaBM486jB*kWvQgZfV)&kFj9Lh_VDUcjnGIGgP80kY3Q1Q_ z=Zi>RPsdLziZ(sgy;f1oVOcC0p;#_kd#Mo>5E%?Fg7Q4Sb_e~TV0Q^d8~B!Q!U2IP zkKBej#xzqzJWEipS#A+V(?paF!3~?2yh1)Mx#-GT~}QySqd)htGA2%53PV$SBq z*xD7~*2JBzpSlSK%``Vf3HeVt=y*1yK2p5{hJh4C2*T-=OosccJ{aNtg^*L_6{#XZ zf=w41(!k_W{1BhnMyA{B9H6)+T`vU*x?ZjZVZHBM!!2@Nl^#`E;vYz-$%ftma(tys zl8j}+rNp7JF`~n#fE}5V6IR=f)7q=a$#Yiy*%G&cZ7Y21DVeDlc$U~NjgZ8rh_~=m z1%#PW7N*ang3+szUbcDId@=7vbF9`Ptk@a;8TMA+*f?43YE5( zrGONzljyVUj2H0)#Hp3o^{m?XF}L8sJ9=)*^~D;?0xNn%+xu3AHA}<9$-Tpr zSwjtOiZUFs_o?97H1FLKiSY({*-8lJSEFUwGLUfZlQC2U5sFPwOb+3?>zx=ps{kB4 z1?%3##J_E*y?#_|1L|1SMhDJRajz`aJ|r(AtrmAF%Q;AC5yeMydPqw!1g|Ct;6b87 z^ACwUvlsn$K~6ak_6Cp#?s+`7=ThwOgEZwho(;~hk@Cs;U?UZS)jI&N%sO~~GkrQM zqMlNv0nhRIcmaU&$Q(TLD+I#-Dx;@swY~I~>mo|1ix8!z;{p~FZs5XInI$Ol*(Wx` z*&;J``EkW&YVN7uNO`f2ZWt1AJ@gotdkAa85URucG3gHs`KZmi#JD z;m!v0Kun~PLv#nNawfh0y<7ofNb&#ibMyN7#K*5S3y&0s3lkg_F$CR9snh+P)|D8c z7FrNRFaPLFU-V(qbuq8?=gQOb$();gQQNVqK|_n^Xo~meWnk=4_e&WZzJ*3!qVgBr)tzqa{F&PZJrQKLzcKnl25?b3ZhNpP>Z49};9$p4}iLOa%oLuywYoY`~7y^S=9W#-|Z38qmT#rAb!aI-^_6@@$l&R$DGaoO?%kc8`=M7Ts!3- zcCy3}eroj{M=TReLU835Gb$x@NgM`2h*2HPkpy|jT}EXP%ACr^rRu(~b>W`El1(Nh zfOh!0uK2pzydw*n5p<$lnaDZKy~J`Q23%5?C})b#_L9`4CvhUg;F|@$5(xHiVa4Dk z9dc`Y9-ZPj zk)Br0Y}0yu<(Rs;U9e_R^qr`G5{}WTQOhNLG%G zRO%r3L+bz|jnO`BZUNOf(pf>Wq{MgBkH(O%QlN&oPcs5T1HdSy>Lo=23Qy?)2BlQ& zcG;;yz0Z?_+^e@N1n1f5ae+&aDhkH*obU*GsjpM5M%_aoDTpD>kDP-xAqwhWAoXTW zVJ5N6n>#aJt>RPLr$Sw4Dw$0Gq5VT!$E>125a>8f`0WsE1PIo|Lhha3PK8rCpggY+}<@{V&Sr%d%0Mtlp(CRJ#l7 z?5f?4TN9Cq}7&L9+Gk7$7=G)&x zKUXN(?uF)>6}k67W^xKp=XPpoP<*}6CTfGEW(Kei4yS!iz+7KLGQl0|J}Qr>3$DZM z)S4e<&*e>Ht>5Rx`;JljRKDZ}w68JvA>5wo0h}kWZfDTbz+;$|iVd;U+Y!hQ2jTM> zOu7tgT&}=teT5+`4U+3iJ>Atn%yYT=6k-WdZ?PA2&krk(aoF&v7d`spt@eJi4M#f_!Z79@h_M5DOk3Zry~Ua>_vtkh?IoeDN)~LIDR|JUrDm%7FMn^< zKP$)HZ~M6QQWuuG?xttWz;eBx3kdrSVEPXfL0T%bw1Q%_B{Hur?@ zccqt|6W+5oB5(M{rQ~jW%AscZpcv5^*_k0A4k4Q3qY?;uObtOm6gv?_>~8~p{AG5T zeMo-pAQ%v#Guuc8KOd(?GBx>08w=f;fhOM3e)Wx%GbM9@#%OLzcX?mF#TicMPE@W% zxqK|#=Dd@uL6#w!wCm}j3(394&CQNEGLz#PwyPkQ74|8POAq+Z8d|>T{*e7-T8bXcbF&r3`mq~Rn2cTMQ~J@LYe($uhXVtDbj&0x&xQYp?V1_p5Y*=l zsZ?6sQ|sqlJ6hte`33lI=4~}|D=z+Ho;C~s0QUc+v+QjwO)OphbK`1J`~CN;59N2q zKlEe=j3b?c#ZkAdqE3JD(oT2Ny3>_i*q3R%l`f#11j$?Vv+Y*WAX>N4wF}#P=+M_+ z9-nvN+Kn5c)~88Ek}NbWo^^i^nWlg-1)KQGX>5gYQGFDM7DHS;B}H9o6mg0@n@-&+ zI@eb62p(*xm;Y{V-Y57dGM`=5h6GYhmjpsk-p}{-;dzU?&LwI7k4aGU!cx^~$=8bf z935R9oy>f?gbeb&Yo~&=`GBUE0&Jn>Y`cj4L}ImUD_w}5liJasC`(i#5gA>lDGwU> zOQLR)jrDJAbqRE6vDQi98t1+Slfw)1hJ`V@?g0vpuB2R3Nfc1hw`szx>oB`0+60jF zBolg%zM`Z^Eb-#`v>At_8M8}l>1t~rnKbp43zuS0sDQVRxB4GuHpb2ss$Pp~K&0xw zh=Pbw5$e{MM8PbCI=C}TI@t;|e;7P%((sk$3qfW7uu+Bb#>7e`E+&u5>Pbnd4Ip3p zkebqsj*V%nL(Zn+9ZoT=!wP_LF%CJ(7}eZ(o2zB$Mw7161zKZ=^|qr<+bW!hK(LzA zD887$^u2?9l&fh}qs$Z|l|EMjYyzH33Dj2*ik8 z3@GHl$gNC7)X9ATehs|g-N^;IgLM<$h4pvAcO%~6y+Osp{4V73>x+lP^9?@{mcuK4 zHx)2DW*g{Pahc|TX?NP^S$Rh>lhrV0uHyKrLbTLuVX>wjwpNLda*Y%O)nKPPcgr{u z#I{-kVY1F!r#z@Ba~owGaRjvn^%%=L{?3tq^P3quY&lo^275Vgv|QruPRSnVnjjr; zhZkJ5@CWIoCKzBE?jh#xiyV*Om<8&cU*s%d(?k@ZlB02L$^-L!Rk8HABY5(!_2gWC zhjETPhWMnko+C^|sw7~a^=h5BS(EMo6X(hgg;V3nzJcfv_N!8}`HH`3=lwItRf5rXwJkH`4;+@l=DZIR~vxhrYhIrr*M|M(t#9XD)?JseuR|9XXzwr9`!@MdCV5w{I4Te8eI9 zLnt{(MHY1)grtn&qQTGPv~@x`hTX%Fx*-FW6~Guix$$VS)ih2;oml6_H`1tA>svR;NxtXvk#p^(6>U3WqWW10a=P&K;+9z z5yajwNb`>0zg@$(OF?7-(Oqo4pYDR#%C#VRTj|y}Oyh$1o3v8;y`TF-aXm%OteE^> zKHKm$JD$6R*Sud(8*bYMSRq+6DcJ0htvin1daI>*-*96K-f7v|nU&xl=1rfk7 zql2%*L)}5&q2M@U%utd>u`M(;^tO}eUYC2f@OGaZS9B;+(0)S%UE#&;Aw&9 zWvsq(PSv~|+n_8r<~kT?&eYR>&%B_z3Bo*gJfiV#Kdvb%Uy~~P@wDHH=4EdT+eL{=}=*EctOgDR1-@U zXu_HG3~JLEaMPJ%xu|@>BBs|OX;aMrPT|ElZeH_VV?W2{3|hgOMBbVQfWa)$g2Y0W z>hyQ@eDTCgF=mlau)rPId~`49_j2@oaP&MlD6uBnXEs`&TTZ0RTA%_=J5<2<4uRBr zW9LitC^V0_K**5viZz^B9N6D6rp@2gm^PXyua&i0R?u=TjoeE&h; zSub6$R?aL<=h|_6dJ?UmLU9i&VeG&gZb%K4Vkpe2C1r|I&N$O3qI}TU$3cw*IgA;g zp)XRHEf>4KzfmEKcY19yUf@<<^8#;R*^LROo;~L+lk;pd{X;PfPnyV1H3c&Ynji`=rAWJT=0C`SuN>G-s) zgj~hlZ*8)B=n;&=l!1nW5A6L4o@~F(@vd1}SPioQ$YB<^Q4BjycNJ1Svp7pgu=ZKc zR)im4jN_oM3Hz#npWv&fqY~Jb4Y_6MDZ8efHZj{JrQXxC)!aU*sw^e=LQO93l%Wn= z6=8FpeLh6!#WKZTxjhO9Md!Q2NfO68Et(g<*>zt=#hl5$i$Wc4t-}tptw@f+vmN=3 zSXEMsqM4WfD_&G{c&pj{tBv-}%2|BxTL4ROg^I;n;6)3g%ZlE0K4R}=Q&CRd;Bd%s zq1B)DOMB5Z=tc9-C?NT{LgtB}2jOG+&+rW|XQyd1?)=h(4~=GNUvUripD)Uiq!9&pRYbjb6qyACO} zgo;ubY1lh`e*C#ReEIyw3nHhlQ%;oJCp8gmzXO>{L%A%m&ykm}%b~PxHo%c(?Pp#h z4_V?yDgre>Z@Nryh0zhPr(cvv*JaLwM$2KsPC{*noJO^Y?jt98a$WS40~G#L6pJb? z&v~%HgyDhpn*RUE4O!B%jg~07aj6$C!GI(ZhA}Z9BvJo`SO-7|q zxfEm4l+KKJF-B;R01|`DRMPZD=MaBmbZ_%n3BUnkjV*8l@H}pJHN%c$BugIbE+L;a z5T{7S5}I{PR6~`9Le4N9UEaL!( zqKFg^and+snkDr@<~?QA75wy0qRlcT;KTs~!c1XU=TThnoN-hi;m8ka1!CjwitDvwi#v>3$0hc79D_HhQNCr)k>w?cbszaczqmNw? z3rSaLpe%@!YIN2vW6N#bxwMQ5yanekVWxsEry0(2iy5Kppa-Y8*nBHhCE#NwBOy*r zCcI|w0rp|GGd6&mO&jpmrN4t41&SiakPYG^$L@z611Lk`1Nn;JclQw7sZ8-pz^=3>Kt zr_FzQ>}SAZeO=C?0&h7}6NK#xpQLj{$VhSitnn`Q!>F|Yo|qUtwp*2LyZRA$)%DJB5;w};?7X6#nGyex#``s3A>XF zkB$<^!@cgLzURApEJsBv>N#S#6M&txfR&;gy$cv zR5=WtSd*G3xiY?gL%za=X6i))K6RJbZupnDMz*+EukWxIdtKSDQs2m` zZF{;*Dm{bIY{Ffi+4pu+lFN4e{~LZi|LWz||G^Is4gdh_f4XKlm>OFe{>NJ?QJ1zq zWJBovR7cniKB^|>2HXq|#A)4X)-`CLJZMg1iO00GK`x;PEN)bbd}(tJOFjxeYqkI- zN#v0CV}4Y`iyI>jQ(v)xQdlKHa?6)Bq8bNDlY3Er!{Xduq;6cDrKf9bv}#_rxeoHZ z!rOTE`MExR{`$av*=FbheR0N4*8Dq?j0y0vpBA2XbCVX~PAm#G*&pQF_O(03@BiWP zY3ez(T4KhjK}Rx*b4VsuST1FY$?nFYb=kco6j`^?v>G~ zT_ern5s;J9h%Xq3wh^Ie8+ujEX0T+VZFFF%X-HMXa;#KJ*<~1q$ms|vqKvk)(pvdi z_xe>8b%vfh&y=){vAQ*U^J}u7kJ;vm?O#!XPEdx#@VRU z4IndqJS0V$HBReH*$#N9)Iu>xW8q0l5+91LL(-atQV6i60#4Q>5>u17(WJfabK%m)fRcE5_W94_6s|^HnFnJEnzsO+gN^uJN;$*#jnV}s}TN6(XxzVIod=! z@^v_ew_IcfKj;OY0cIEIcv7~g;X`k62W>}dcU@JlElylDy9Hk(Q>TUZ5q~;JgfTA9 z5&dR>F~?Ls@3YUUrm3ld_1b>n{%jc|W6VgkYTfOLTc-vy0hX7iUw02$2E#>+{zJhl zRrxBi7~>&mX&yjnfXTjWxD~_;d^6et$0`wh_ozi=YD5FW0@(&?*9x2AJO!SEh_NwW z{ZJ|7bR05aT1hZ}wuR5PUXMQKlxXTo59= z)Y?Huw~2-8f+3lRcb$do7NZUPuMUko^QHCIfmdv}vrgNkO(2G#GLl0G z!~Abo+>{AQ0nPl| zYCV4zS_>afJ49}~13%M1Sb;ZNs6B$aEEZ?Z8ll70u#+dr;=!ZE0}2b#j!zb7Ryz!t zLzcN@$>#Tk!(S?-q6%ssu6|~}q@&;LD`g66yLnh~fHVhi{OU)fkeD2+&?OSoC{jG+ zrCn}IWY8?U#p4RV>!Kek{q5nfBuLkB2A! ztagn5k%yMX|Cvk){l9 zLl>Wb>|o%lK=L8C(@HU5m?>wJo%@3f|!zYfZbGSHf=pAT+ub^J9w zborBtiVo#V#@yI!j?ur$Ja8!mHy;!w!?ummM{aZY5=yfI@Rnv~8&FG4RDTd~Pp*l1 z3i$Cynn{3NN&p6oqP3WR^?2Flslu$HvY{{=c7BahRYVO#Xs4rVNYv3ioQ&eZhJ==x z{@jJjGO$EMKIsSf(ZFV*>yo86^s$RYLN6*ggu4pH1F$#v^-6^nikVrxGVyT1qCum5 zIcqwUl0HXO4%#TY#B|M>Okk#ZKgg&<&+!rX$d++cjFdJ)OVyeg6cL9ud`_`9FrI+K zonuPE3?nC`PR~=0jqPLfaR6>+4mq3JKDzKa`u%+{D&v)>dLZAF;PclUQ5zxYWb^=0 zN_W+C$(-e3XxGyKcWn~Fp5Su??<>LH4Pv%P@4?~WEF24rmsx8rt~CE#J(JfJU>yp7mG2qBmFqzu)V*~oK7t2?~jjOK=0gS!`H4*{Ye8NJ)pF|6> zDBTX7OaQ~x>Pgz;XkmjtQ7OI<$Ig?UhNjI%;Z2x+ zfYfEyK`nC$#hASA+L#BIy(Yxa1 z1y1|B2jp63gnaYlQaCqD_VI**WebLk`%3NwA!}bKKK}*c&$~ch&=Gn<8f{b~Thyc5 z1K8!ay*5P15!evxe>Ne~2hhySi)W26Jv0~haTdzAsK}4^B1VsVdmdb#6C;tAE35Y= zh%+)VWX$nqjFX2QcM4>Omt=OAypTQkDhPeU_Jd_E*?s1DlSSPfHaB+q*=Zj&Hrw~# zSQ~f^s-GtX001fcUs4N>|B-mtytdw3Y>EHcp6i`D3vHtOEW37;FF9gsNP1mu>S}CS zQ*bg84KmClRV1idyRLp-vM>QiB!ZH=uU~SzqCQwsB@7t;4N^aX8AZK6yh%bDE2I59#om z7BO6$7pKX5Z16se%@D{Z*3++gUEXHp9hw7V6%&OFeM_ZQ zz(Ql=cCxqg82@H$cCZg;;~6YeAOj^0;2=zsJt&*Gc{}t)kQ3o2le?;abA%J%VUOI{ zh9o{)mBpAAnX+Fe!4RK@Lg(X?*cSi~bKVHvx7hb;xD;QrL6bqlGk5Gqa}Jnkn5};9 zS^Y8uAbB)K+pa&orhwD+$Eq+xNBlC93pC1#-cnxZYACNZ1T?^L^J{ed!Po=i>(YE< zz_=Ts1t^*Q!vvxt9sy;fzT{SbD_{#)v}33ndTbVUN9O%I_S6+NBLWBIVh`to2wp(& zh_j%@@h77uY>zt+h&olQkKut_`;n3{*F!fs6glc@6mQ*nH16NxE>}vdnlx4mV&U zKx4n`EjJrq#8cO?sVPPpr8Xr zCnMAtg#q#w95RF!hoIKDj1GJ;inwqwoFA*VcJ+Lhz1+Oqv^Ap1ND<~1(3-bGRu63q zeC}>FMM-N1+;20!67L$;(ZvmaCLIploSUYRLEc3kdd-^w4}UYev%wpzj|PSdA%!-0 zp*)z+S+G1#{aeN%AClkkSNVsaPdd=+r4Ivv;XeS9vAc%1NF#=Wc?L^XB}*!`se;3JDZC)`bY-9@f> zp+k~yyKCUc$t(uJ(;KkNps;~odGo|uOVN& z&O=$ZgF=RfTwZTBoFEtsxDC8Z_PzN49VySPZIhYp2Wfbg!0qKwFEusvC@vUZvx7%z z=5)QYH^gyC+!*(`5BfK_GxV$u1=8AFUSUTz5JRvYs1DddBe})*Uay>)Qr272nd-bf=b4f=v!i4&TS0zq07PY_Bx~-(poA zd+WnSWA<`$be-YgX}SPFO{Lox1}-ja7_3tzYYwBdJmesPpa+^~@+@_@d-r(6e9?i_ z)vL@Vjr)JbUb8=k5NX)lx0iba`)-^4`7C}}y0NzF;A=BSG8Guz#@xdy*Suz|%O#c! zt!`#n7+k}xWZp17j0uQTl5PV+=&7xW2G65;l%LZ&YGX1*dW~Bu!^9d*oyJ$7ZWyE9 z770piN(2Zuh^Sf&Y08?%Gl`1TfKunkzQ|Ve!;I6DIbc$kug&CmdTm`H;KuKq7e`oh z|KLscaLW0e z?DjL^0XF2XE@YHy{5j?UcD&5*1)|-_?{7l=cv|xYK*|H_)gUpy6VB3NbJcj-9W}gI z0>);+ty)>=wZF2l*KdIUF<>2g;LeDQlZDh6QlZjj&RV6iE+f4N-Yr-m3=!C9 zbE&q|@efK|S|mlAGJH3%o!V~+5k=73hd|gL6e< zMwO@wdb??3^~iFp8||a)052m+)iUN*xOyks3gt(>>0lOdO7e}7uNW?x>Va%>fpOE? zC?v~0g)}N}%u;KJYgBuh>{?0btF^eb)-;Xr1wNb9BtJDOZeQ>ow;(oHioE4rmq69} z7Z(`}H-(s1RzRff3AK;bu@tz^fdr>wWkj7)0VB|r!9H>R~PTE zzOd^~ixOk3lBMMOol-;rR0!hFaF17#PPL>~8=l>`(*=6}H!2lSbHJ;4D%S1*Dw<6x(Ih*z#=v(T^{P#R{WxymEC(tB)01xq*+IR(gAib8dYsv~q4>+O9a{yR#hF1L% zB=%)zbC}*1C5=t^5$+w|52Ngy1t3?;N@ZpdSK{2FpQV%B6KKOc*M4Nn&P$Lt$n49D z2C}K4o<|suUWng@1HcC5J65H@;pK9ift;8+^&@^i$U%AgQ$tOa(jm(V5sRWC2D|qA zXhq}o^l5vo(fg0xX{9OK_fg za;o!d-sFmE^&@y=zPQH|9n3GB-Bp?;>uPs5msP-%7(taFCbpbP%QV!CV43KRDD(wm z5SOew1QRRa296nSr5Y~S0VgsyBHStp5lJk~PKlN451FNDpB`2}wd@oM;YFz@P)f+@ z_2o%x*G!esdmW2R54O3(^SaW%hZ)d?={)rsDCGOA!rE1YQnW(w+TAgSp357Xoqc_6 zO-ybGa-vmvf_4M>a5)noAz&)&4Slvy#r8brSdc|T%-kXU=!>3l-^8R&%rHUi8;0LF z8w-B`ALZM&w6jK#u3LNf?q0JhXzld}|9XcdY6rX4EOQ^v0wa~w_Pl6x$7CK~s3F6| znJqT|QakukHwwe4y`^izhPhF0R>!DCI6LFxxIb*A(8vB!8rhKlS`P=B(vPf-f(X86kDwJvr z&nx3r;Dlq1@BOY5SRG+>X_PN$Xc6h>V^!!khUz&l*LKeKl_*3h)Z-t_8!i;#a~CV+ zc#j}x*P3;%N*@l^cS>LO*>TRHYIdgD8&11gcS9`PZ6>B%Z+V=2N2>b8^zi+BcG-bn zL!Pm(E7?W6sSASk|G0(xYa8n*Bwa+J^YdCIN<6Pjb+{F0vjAxtxB4#hA((fL-X)w^ zif8_?1IYwf-2vf6LKTT72-;~%-_IPP1+1f2iLdIeAf$eBubrQdkJ1ywpH;o>Qm^P! z{-x11q)zi)^74230n;$nXyHmDs9528H5%kSImwOgf9Pg@lbQ#DGj4^2VrH_F+EuZ% zD;MNQmuaMMw89UEjy}^Qh8yZv-2}im)WHG=wT3~q<0QLu(sr+4RKF(3p-MCf4qIz z5M2V@Hboq_kh4EZP%`SC0F^g|79ugI(D8>icpbnS%Jf31%aq1Qwz|d#pO{=}wj{B< z3kf*NViuh^|L`Jp#GXX$A-f{_1cx8y10NibR@PHbUXY5ejYC%lgk1QOZ~eA9V2qc3WG3Fg>Gc5@OeN0C0MYA>X{d% zX?`?BEtekLO3+au4g3YnYHmSHdbJ=sMrr?MWMo710ZXPYNlhFj`<_B5T%?vlQKWB?;8mx5BIg}@2<35 ztbCV@mJsgZBRC(ZNr}UtMym5vjwF>YU&w4WlD!cjkL68pIqZ$0o>&`&_dHg-s9Ebx z84jH8UQD>~`Hc7cs`>*Iq0AXgs4AY;ySp&H8Gw+=JuNPkO1uJV{Gwv8eA6J}tmZ+ZH`Pl(7AJz-=@b#iu4WUs?4)v&Z5 zHSdM}%4g*hNtJ~wroCZLs$d{3N2vsY66&|*hYe|)s`n4}d+ptff*Iw&5ZK^OAtpVf z!5F+Fv2nB(KTjng*O;%+FjSO(3A~m`;6dY6mKZ+DyH!f zbo`PkX%P8FJ_zj2?H%8#pFdsCEyt@V{iMZw{bKVb^9Wg^rot(cIzBIg6qXDq^$9NE zhxpm&^Dt2%h8Pwb+Z&ZLu-Vfo#+b{Wk#wjxQ?Z6>X@%|;;O-+iq8i|blQg2 zONUHPvc~wlMHNbnI{rv3wl~XP-{E|XHfPmJtyl#W0G5THyVN12i`_pGCD`Xn;#nzS zng0T*5--b%YAN=SoSK!?jkSRBGVYS*JejM6vmT2%ud;_)3Q;83U=n4Rf~LdOSce^9>= zKG-bnU`cinI`<#-r7&WsV$K{Dau*R0O&E(d zrQcJynI~5>hv6kDoU*fLw+4H}*^K5`ATDl!-XJG%L&C|y%lUm(XahyrxWiAG+-0~S zb~M+4u1PjT1fg{!4X!&(ZEnqt?2VfXdBTMLG@1ccQUZ`?r_n~O46Zz`rL`%ZU2(Hx zHR<5lMh4u;Y0Lgggho7SBJef7Qt;isCSdQF_6=))y;ucSEFk)h?C&cw-c);#L+n&s z76rFT%GG$khhOx%g2MDOR)Q^jUq$MoX*|KCzm6t7&+(3piT>i4BvVH=U-TT5m;^4y z1JodE7?4d2gXa4rZimyAtMPpwLpl)3Gaz6M#2<^%z-v8nVvq86x&~+>GrXNfpaZn0 zPFSmBOu%Cfi5=aE(v67XQ$^Hp0^F=dZ0C8Nz{vHypp}^?8?y}J&{i2D9uML_z$>@9{+`FS}}4F z%1i(jBDF^#`~Yh=XaWHD?-}Wly#rMkvs68*NbyQUm+Bu1!PK-amGn952gx5DX>%{X zsW_3Q(Zlf526ukH>Ng1$I-M7nrm#k_24*ePjOOW zWsTS@1hB^YME;5Y>HKTVWVw9Xww=+W-hbtug52ZoldP)$_LNh~D(hqPpiUyDSM{A1 z-anf%PqCD*fYJcs`={r1M;;o8gt{(qcI%aOuw4@}Zn5f=0l4-=I;E+jB`*>tix~ZlhuX zZ%B11CVc#@B(@k?G%1KfBGgkD8vgnq#IO$QxLF|%YCtH}E-6}C-itMW4`d&=lEtZG zDDq+<0{F|9I=4s50Lk_EO6kmG<$g;a9S7~pCK#TtYUucR@Nu*GD;@-G%3*v<31ljq zRE&;EIU}(1d=vF_SZCx-CjctH7)`2o5CB!biXHW_K|xD`iE@C-6d$U2O~;DA`rKJT z_DD?w1EHy|QUG=Fp)(oAS}ifIZ8o0O0leK%azedMPm`*BJKmU}CVLVGm=?I6!7m+8 zv%uU4;u7ffn2h|3OfEh&QJT*=3D}?$4K}Ct_o*L`DIN&4<(o3+8x_IN8>liPUH8A^a5>y{i z&+&O|2c{bF4Ytt_%VdXj_y^kd)E_dYU%uR8fQZ=wJM>>UGcagCRag)xOkRjn1pekH z!NGt=w#7s$DrW{-ifu+B6SQqiQ2${VDHse%?lymu~QN;UDu0^WL-vJNJ-H?{Pi3eS!o6X61<#ac4|=c_RrZl{T!MzSwFkMfY;@f#gP z$aLOPUurUo;>D}NhJFD@HVtO~k`_(U>(Bs&1)S>l#p%v88@b}H_&AKcyCf!qCC zNBeX!SWh9@QI>XIhHmP+P-N7A>F0eBw2AwRQind?_J$2*GXuPu)`+<^yo`>8BHV76U^sF` zKoUoNz0k9M{cn$|L0NRHs9bv@dGfYi)YHVZ_)BfeHhI4l%wYv!EKbV9SFmtFEl^kP z^$wi$whLX}^=+9GytEItNW>gjqdn$?c=dMe38$;2u-U2dYLx!=$ov_=@Wh?bL^Jo_ zW#M6nwXMS?5~B3VnU zu{hd=yeMF4e#8M%&GW;hlbpuuX=XGA+K9zb=#UZ?3E!2Cv^-I1X;s% zjkY5$5CmZXZXe8m4mQ%f@O=DE#axJa)rXm!E<<;$?~sbsl8k~mZvJ)+Ab{2R2h|)K z%=1`RNzwy7JentK>M9;mzz!YNsSQwqgM+mt4?I(pPsfn zHTQ3OQF3mljv7dj_P92XLTm7PyMtP{o@mORc~q?=lov|w>sV0(a5keJfE$W!YQ>&P z5u!fsuqr1jiM;i;T<0n_$K()>rQtuKLwdQ&Qi1X%rrf9S*Eu>yoTd~cG4)qD3)gl| z;S|$r4YL?21qElMtM{m?1rT38b->EvBA)?i8*L2}^$rUEIw0lM)VXA8)QijGttQdx z-Cs+B&Tlzgu)*YP|F90RSEa2Fnrp?4}6_Kmj3jlP)}!tn&o7<(80NzprurqGpK6VnI<1WC4c3q=br#BhdLDa z>4Dt8f1T;wyk0f}#oBXXHRhWkfa^1%r~?PcNz<^1`Zi2uQ0+ZG8qikqHhm7ZDUfV)Ny}?+fyVh zt6fl~yakEkPguiN{qaKF+~;rf$&>|%VV?a3c8(6q((QCBYbiS+>Uqg`2SiMPO%a9G zgc!FuLwdca1YYTudi_j3ohG^TCJq?=N4%cF1ljYLWMy?CB|V@*e#4_J9xbx+M_`j8 zV`D-<+Tz3FLybwEs$eM`V${T8j|$a-L6wHV%4jh20EstcQu!l`$!P@W6q_l;wbFUQ z&n#5jYt1Y>?cptAzKBXo*lB`bc++6xjA#=qf%i!H3Xx)Q2!8Z;tmYfnorJ?cbUr?C z`*d&S!RfrL&wjA!nNRn&f95FNwc=Oht? zW!f2YkK)<&6F>c0KU}t$FR%JsJkXD6O}8-&iNK*O8VH#Rz`TQ;RoAe2ErPpwV)>3d zjGV8|Sw@tZW4r4CeN50kE#jPE_}l@(u$|lOOul_FYDxrF+#%Kv2yEB)430o+eO``9 zzTV@nY1%U>RI}6Vqy%_OaHq=%b{WUO-YHQn)u&t3PW2TMz}N^Dz1`*K{V4aT22Q)o zfTE4*OBk%xO27t1wAh{N`DSO$PGti&El$H%1;O~#%D#~~i7{Cv)|DXhUf^6b$r1_1 zdmnb1jnpfGzGm5}?R46~yc6UhAo_@-JvcKDP#zYqamL#=!BVs~#SYe_k`coCWp9K{ zz|f;IOY|pA1S^qKH|4cF5udMAWCy@uT2UM_G84EH2x?m{-mcRhS5p>Mi)4#m{+K(95CA`jSeYS}K|K9Hgn*Yp0LSh8At#-o1q zY0b5pN|GxUb)n~X^X4=Blnmm`2J|pK@d#Llm8Z??6o+usk)9stQdQ&UM4-vS!@Los65o0ZY#V95@KVB!Dz`4E+` zkBMMHR2UX_@d8Y}{6{7)bAI?ncDqgwC|GDFsWN*LYfpEUySrp`4)NslVkl+jbTES^ z|GNDRi>ktUj@{)6QqR<9l1@EkT;tmd2)(rY7iq6U-(*}52kH$>+!qNXF%@a%R+L;S zYP^Y63?*c6ndmnj<@kE*PpWWb8Zd5X&!x3@D!_Q?0W+C_Rb8S@gi|s;oD$cW)4VUWS4S$KaoOju0V&*zTGQ* z+^Q;S_iTm!x=r+%c+w1Wx!es=Q@QJ!J0Ziz2a~fwg>2Jli3&eODJN1^54PS(eMFQL zs-V?zcfUT{Jb$QS+I6{(0Ce+r%lU6X#&+Plg}o*6@V_^X3Huj#oi95tgNlcN?TDB$ zUI`N*wBAC7`%{lLNTER{v95ju2X`sokYB;51;7^e9fIglWK|r>U8~X658z2y5eX8O z7-PMBlK@9{9WBnQ%8jcszJzzDL5_9kpEGwCZ(F+Ptgu_LA?q6c6VK6UxqsIkcz3=d z%*o~%Emns&clE+W0}Trk+kIv?_6pkkWSwhIoa5*5dlniC{aqC;$}N2b)pe6HKZniP z^O30KrIwUbfG%cmV{n$5##ahRh#Z0J6Wl@M4ulJ5- zx-8E_TivoF2~aG7pA92e1@u!4AU+lp>zW za33oQFKif*@-)|q_RgG38P<%MBuk>dwMwqS7e#WAOgT5)=v+^MH?$;^i4)A20sKB4 zYv^x!p9{LTOSj?Ea9Z*fLk@in-55OJv)KvADmDR+E|1=boD@kE5rpFnVZ1U$ zY9kDSoj3pSKJdjKTz7SB0t1AvPxdA%#~wS#13sfI{!?4KaUWk*_K_x_^nT6LJ8%F$ zz7fe9Cq-)68LHnI_Yc77_HhhCE40}Un^ag2g9t)ksg6H7v7(e}=>=&w$A=*KeUR-~ z!5~#8LWhUJ8L~2F@)}fOmfWs@$vpBL>@qvl5T9W)68@X}uD@gu>H?Dq4C)DWD(@;B zJuehyUj~tedAJSwj$d%J+IN3^HSNHkin&$`) zXheNfukr8D&kyeKGi+qK{4B++zR0y(e%vTri#<7M+crWxg%=oi$t^?_{m{O|MW1JzvkJ#M#~N0w$?#ep2sB&Wui4(vxwlS`1bzVZD045tA; zoUYbSA-#deGL)BuygRBTq0SiC1aEEKhnZGhLLpA`C4>&}9JukbXykeh#8>wh#)bok zO-F^T`ACBvB9g3qx8s}%_B zoZC;%9c^d1$E|Jd;1~J*H5rOzLBwBZ^xY7}Zf-fzEu&;Ad|DGss9X>G)}tgjh|Y6q z5)L5g3iXb<^On}5Dc1okt(E1s>v7kn4V#5?7dN8HvGn2LaF@}qHe0>Sw{W6d{VN*( zo~ZPp2K0-w7;gPuhk_;g9yDH*UZ$pV!omv~nfS|%@3fIb zPvv%PoK9xf(^5Q9OC|6RgI&QO$;ehJd_czGtst6~)vd~1oTMjJacXqDRN3`Mwx_Md~h(eBIo)dqpepNju|g2eXLyJ_7B}=!tq@cT3))jd>?YRL%9KP! zi+P`gJ)(@~j3d5x_NSbcdM9h=;d_9PmAes_6gDrzaOXjDUdY2lVN2JGd0nk#R3YnC z*_!QjQVe++B53{eVpZmM(LL8o)}=O-x;Ah2C0Geay-$j^t<6;?#qTVh+mDYrJWGws zAe+yr=pd%yhx@Q3gN53HC2yNQUYR0+<5mg*#kqwKz)c{-ru`69NY8CgdnEqC31t{&er6A!*L%WBwDL+tKKTlsrqV=n=FW4CytIz$X) zr!DjNNFZchgOL1mK0mkbTOU}W6?t}i^3v0DGOWB7?@JjfDJ!fg@bhm2y->v zVX0#%z6=ndbNr276f$8Fg|rI7s}w=0IXo0p75J0_%*#~VM2<2TjWOzy<3yN8$W$d& zQz+mBdm&vzreziAqX0SKTvTyYPl>&^SEmQw-B8$#LpNWcOw)E(DxjH-@zaydC-<`PNX=`hl0VqyO z7`&XzX)6z>)=tm4K8R==9jz+oLI-Gcb*K4Nq??wIv5l&7vMYfuOZO~1txorjQ(B(* zM#t;U5pSd&skSG&kSAPS@h>5>^<(g(10uVF;=_6Zxq*?6BRz=wI!GJdXcPdH-0F+U z{A*3@Gb6_+q%E=$NzP@G$O{9dj7g5&>GcZ=u!dmYnWN|9Z1ZU;9vo2vbCX4pQUhc- zVMm^YC;KZWtdKNbKMKlxA9&$oE!4|VtnP_0%r)93`5eBRot?#vmz`zRV3Ytq6$!|7 z6;rLTT7ZUzd(&n;gUjvU0bmDN-~j9Ij05M|;_SP7GTTX%;ER%$e_c>ZIJ(Ab`?66d2w53a#dQ~4UyoHiFvznluSs+jV zLmzYq+k_3lJ|gy>O>s!A&E^~==FL)5I?1DW$x-XX^swk7OyfXtlcV%Gkn!YXLHJpP zda2^30yWMchPHE2^|M&XveNh13+fwFHrS|?uD95JO6?J;jrN|&|AtLpmO1upLqA?a zK8>q>npk*Sg7_H?clm3AaK)Mj#og*?U0@|+k#6V%V9WVv*MOZFll!7WdCe(5b05ZO zqUA?mmfA6CEDYKS+`!rfietm}*BO(*cZ0T7<*Hd?nExoa%q&KmtK%vy@E=_js#r#O z&g0N(tY5{2niC7rIj&~ViPPQ*PJfPuTY@!6mIUsuaDy68T@odlf(5-yE*NU@+*VT7 zU_&mj%G<{eMg+BzDMu<{ssa1Vo`hMKVB$(uODFG{kITF6U&s?#B>KO4xRehbkI?n% zd~p*vv$$57cJb1d_<}@-UEJ+no?PgU_{MyyJ2$nL^gYHBsYSQqZ4D@e$RhRFyR(|f z#M8O0_LQuF^a0}NSITkg6db2gP(y)&B(XyR(%pN!h-#pQc|3$48)2m-JJ$RfNSu{= zj*-&cLWrJ{X0d;g&z4h$=%w~1QkiV!y_6XLt;v6!=US+JQCHD~FamNWC2g{e_*d&K zAjy*Qvcccrjc46^n~*0AEUv*>-Dn@5zTP^|Czw(8EM)!cQV-7}W%*FGd=!QIMeFCn zrrN*ODhVop(qkZRvK+A%E1g%Z&H`{1rvW8%J+?KCA|C;dN_4%d` zQMt_V+`)8G&}}y>=3=)B7;^1r zA#>aVk>gs=+_&so!@We0w~^JrWn)@P&bj0w`!c5jEF995Rb>#;0Px*+Sm{&vRvMk zjFsl&)Y|}FMrG*irDo4XiOWZetLL1Xgb{s>kbJO0ATUT3m5^f42bzIyupL2erXx<{ z*A=GO3ZM0BSK6yk6R z9!Fz5og!2y?ByHhW_roIN#HiUhLmYn{SA_upHipDAvRA83DvI=dFT70Ts0nRKti8D ze5W?2X6?$D75O2_Frn-^{^2Pe8yEX?orjCR7;>Fi$?LDD}!s{Qe4_-mj>s zOBfwKf!QC`4T({nCJ}9cdCucfoeM~(fwxz_Wwt17%_5HTG1J_y{@XVxZl18kF&ODX zosq{ZuJ_$8V|rL*&ud&TOMVKZMp1+)owxvzofO!HEO42_Mzu__Kaf&5Jfo=!?WuZ6 zb}FxHeI$T9ZX7|3obZ;l(7A6;ezCh^vq7Uwa&lVn!mVXtZ4BP0|B)HzL(&nOMhaai zw-S#tQ9HtwQb-g=X%dMZ3P)>T1a^~UhP))T4&!u*t0eb<-QPxVi9&2Cw0~QL%)sv~ z5}OE97y$zw?yDvHfkpasFf@5rXOlZ~doy;ja>u9Bt+l(YyQ@(M)k5%Ts5^ZE8fP{9 z{?u%oxs)BY?;^f0cu@228EARfa2iZm%kxXLw`oh4)IRbHCr-~C^lS3qjlnmutN;cx zj9I}$G5xfELwgGkN%xd-;L}K3Y6^M9Q#9ujS3QlqgP|Bt zahyX*_ees{fja;0L#}1(MC>7!Q(L4wqCAAw@HZq4oZ|8r%(=!y1l#4>SC1Yj^iV z*-jH04Z3F;Xuyt!piN0@SNw|wH~PCr<%eN2^NM*#bn31FFuiy$j3TZG1z#EQX$WO@M9vEaA+AOnVLj+R!bmJRGwg;VhJEgB_=DyfPu_hu@-EP(I6(`` z)7ngp09Y^xUo{1Gw<7HypxR#%K*2X*=3fqF_4(O#W>i8zb194(&B^|h{PQv||d)Gwb-ZXsu@Ka4I4b z%BY6Z6vh{6x|{$cKvgb?>upwO7ugj_d?kNf+b=)|;kvF$o<2Ng6%ea-u8Sb}lYk>P z;1#GtvG#BBEJ#&pxj&%NaY{!cf1YiA)?FsPz z85QTKr*COwVr2ar)`L{69M;+Xi_0r3ZUPg=U}k)6oZg{LCe_-!kZ1^GRRbP~*xZ}G z8c#I_xPI*YQDsPtwXu0epeEi?;kD3k;oPw$a4XSVkxEa}6>1q8o82D~S_#1v@`L@_ zGCg`G*tl_(KlbghA8NTFP&L>=Q1o zARrbj%-Xh2sijs`0c4k$=7J$1K+C*yf&8(%Y%MsxSxeR_;Mm+9NiH~DExERC! zQF|6GZ&r^7B_vG#W3rxUgN^I0yw#2!#2qy!Hl#tSM}ovy*C6@Qu7%W$1r0&^xE^{(Ti(KQrd&g z&cAeMZcMYJ#hd(mNv!l0m{DB7v@l4sp6%m-q}uUx#Liha&qy9 zd`{~yH~NmHObM|N+!yRl>%_*O*+cLQ}a(_FeLy!KP2j{ zt@dY7tx#qh3_ydO%ndPQiZ5@t%Hm#@Kmi<9+4?}*V=-C73aM!k_#N*1Q8;rUeg=#j zh!~0PXqmYM=g{pb$c`Z}xfIy~o-~ho`kf7;-A`ZH&QsjK9YJx1)MYDPK&u8knXKy@ zA4%?)mIsxZm`JQ5vu4A-N_-k~mfX9;?^c_}#o1;+efa)+x%lzgv7du)en`^;CqGEl z1yjiaoi$lK|eJuU$it{|7Wk7S^6aItIq z7`ebXBfTWrAh3w%@uIP>E!FKhYaB*{QN+3I)BQrp{s>?!M>_%qt=+vD+!+z$b@Svs1HVoAx1=_#tYsS-hs zu=c^ko~l6Iwl$hrZw;p72MdnMow>kD3u&9YQ_|kjIixQQYAR@DZ|#?dJJDP2#5w9e zK>s@-tZ#rLHUGBt*eL-3(EdZQ@2F?s_?rxt)Ia{)UG$e58ef%kxN488J8&#C*4`)_ z@f1}2JIuzKMY@7WA`vXS=Ue^!br+4V0;iY=Td=aeoSz>#xa&B$bG^U5wiXRo**9hx zAs|uqt+A7&4%ei#-W1oyu{(+4GhZNI<^f#)R|B_YYWL@(0h(h``k~mlw&zFTNB3t} z)XKhK4!Y87ba;VWJ?Ve8b=$$f2=2_W8>lr=<)aLsd8%QgK z@KjK{z|JemZ|hyXwk~$k2)XrWSF&ISEEL+@XeUuR6Lkgo6AUs!J$xKpN!tdQ^s|3ny*kSk+%e47a;~4O#@vhz;|a{rY^NN{6Th&PirzBC3N??mrr7b82lamtO%;cJ$$0DujSZW=+%|C4lL0p5F=& z+k+U{PoJ-^ELCcT^vn$-jZ_oVe;U{f`Uf3z1s$U%9aMl=nK?3nXru0xH-X>7#L$bE zMao3KaLaXqKubY~hOU4X~5;UA+H+=Mq!e7A;B1X%g9D=^}OhJbDZI9ifc z+l0tx(H458yuUGeb!9>9Lb9{`T=G^`c^x8_CnXxX8$#1^E@_-g1MGhN9LsV#JhNT( zt1OqIMhRE%GQ;nlscujn(;A-E4>ybG$Cr+*&wJ&MeW!VLOvm&fGq%FjsM%?Y_TGhk zp*dz8JpDHCSd~x@9K|JMoqT6t$0k_)tA~aLWsvj~ZK8moK+nIGo!i}SD%;<69E(t6 z;j2PBmja~}3Eg~;_VZ+da@w9I!Wa7;svPappr@`n4^~UD7xgTR{F2ScjHtRa*71)P z)-TpaIo?MNn^p$rI-jtOelkSUb<8hSwf=VDRD})NT8W}DYNf{;csr0XoUSwR_lnGX zxKA(&;4ZaY?~cN2q#>SkkXAbEQikd!E0PEcGM2&G1cJ{gdLsf|I);H22MlLevBaWR_YCq?S@k0IJG`O~dj|@Keg0z>1>^BLa4@ zOfWdcu=w0rY&J#?CsUU1ooNic8SIdqS@*pf>?pZmBJpR7J31izSpN&S;UC(S@KNGRsJpUw4Q3$&mou6-^Op5#d*z6O(^%B`&(d#=c zl#=J|c8AL&&161rfI(PE=SG1`xz1C29%@FVyiG_{+jKb=Np6*oS(7>~aw@L$r%TdF zh2ZsZv#vHbW0_iDjx+u%IA$-hmbx$njfdQJW|KwDX2DOV!|)HF$h2cJkxV)MUB<+I zn4xy7)s6@ZN~x4sAwhw80<>jH5XZ^gI;&k)LYcQ{eHQHpN0x#P{itbUxiEknUWq{e zqn)lospO2R=KHn8j34&bEvWpIOf9m32O-l{j-WXuGndV`9}~0}6;)1;=WV~!=QS2G z;mlPj;&fSq`9UjaHDB%e1`-QoiLT6Qv0flm-5)ledY`vhZcPg&=W2~eu%veZsmHQUSmD8CT_??$nrT)SWvB^% zp;%MSe6cnSZei>hCQn=)3;*0K)kTf5(d@OafAaRFDtj+B#C&M}mu7WM zi!Afd_3;P7Gqnq9T}+I#+)nYbF#MV%Qt14S)xaAo(J0iJpd~lF@!1Mr-oxx-L2j_C{T6ucmhd|*a1o7U2c3>S-zQpaY$xKzp>Z}Pi6VxzVZMnzAB@0R&B0p7k0tuj zz}d2aDC25%5-h-*hsW6EHn^vN=oluuw}|XA`sV~@6{j zm?xTVlR>QO08cZ2DqA4nj_wmSit}_LyOt*u&<_U8KnAccc#@sM+P1u&`|{C$kYc~s zg55_gs4Je=(5-Ut%H?TEriIZ^5NQ0S`cHl-AA59@kQ!ZRgdw!<kba2Yu;PYjax zeKJ)d+3oz&=DrJz``ri>5*_6iBY(~JC*NDINi}se445+7p&@T^!?bR&Uto}*f>4(f zkZMisv^&+_mz%fT*JfF8BIwFBA47b9c)^U8UbZ?KWFx0GaN?2u_()DCeFyLw%g>b9 zIs$m6eUJ!cyC;N2ybaYMokX(In-1sy#f0BUzVDm;@|skk|1;R+*KYB@x_SSf9ZOtu z@G^S;BZtB|Ryv6NUNm2H%BG|rV?Szcgq0g{8lOYkFDK8ic0gkN9oVAg*)4m?oFs1M z!~r7Pv;jhlJm?_d7Z4t|y7FxkA|nmosDz!)6dimaD{N77kc-kw0bO^3Sfebcbkb=) zY)%`&l1ZCfA6(sBQ>tF}DPOK!ZZl^Ya2TK4LHvgp63+6M7mha2C)Kjz0uv>pg#;xp znDP~jVbfBbp5?7za)20CMwh9HYn+_nfO?Aw2-&#)ArEr4__HrZ1&%lJL~Nje-k_>X3q;;x@Q8bee)-4+2@F zYX<&IfhuIG=Bv(9uP+6(Ztf}VmCr}IFG_|3Y07sepN|nv3$J@_oDtqu4vmD zHqp$fIaoe8J3Y;ippSrMpD32HH2V_|2?TFe4|?jkk5KaM-hTsTW=a$Lbw( zN5z1md+T+Q!*#>oS7=MwY! z!jEEOP*JREwPm9soAAc=!T4Cusk$#DE&p3+eesHbD~HZa5m%oEllr{oa)H%efw`X2r3Cyhc(eCokx-xJhns zEVFv&O3_zq(BYtb8SlI^yA<^b&2YnD`3#Z&HQZI5b;3(_aP0KVtv*BW7-V_Yk?0b5 z2CCrSnTlq*Om_LqP>Dmwd#3?INxOsKa#(lZLFx_D;fR-#l>^EtUo>9aY!3Jm4^8qW zET{h;^MaZkDQDu7=NXE=W@A{O6bPAv5e+E8=YJNax_6_(S zqCehMl$}Ypcag6jbG(N68K z+?k`PA*1i=>@|wi2>g_~iXOi$Zu|cmoObtIP>g@oHlZm0Kf%f1zk<`0>i-zV{i5XB zG>pia_fb}Z-RwHL^MxavP3xZmerF9rYnc@PNXk}4mcE~HbPOlz=i4=aBmw*XxOIA5 zyVB)vTscGcC0HFp?)`g@tPG3ohDuBO@qBHZ?57$C*;mY+h$WaLB8C_U7YfzmnYqgp z+o^{eVBUb>&?zNw*Ni{i*1dK3+1}ljF>U5I=by!c1Cj?J<0gjWhyUsH{64-4vKgYs zM4}x-;BI%-U(xAu{d{~HQV1XIg{ii7HCYvfUxWkC9Z#E}0Kpd~Y&zE#3RAQj~bg-xDw~mXTHp~)?0EDGa z3ZZRIxEKo?XD{p~%`BBl%>3-mDd#?12@!`GclZmUt7dw@Eji+=MPV~Jo4mh_*x$7} zknKtgV_q!4cVCyMMAKA=CugcG4S=W{u82>xmAAikeW3Q|Yeptg2~@j>!m_28nXb}LK!~pnn>GB5FJ>xguDYLwDv@8_i4HRC zm<>%ty}V`xVeXYLYKz(Op+<{#E%pamUplcee1WX;1Wi#u07DMCh|%GY3)wkRfm47y zJ*h{*ywbQR#`)gSdbUlgh>OrdRGcu8qzDIN&%vm-4q5cbwQchcDLzq@nq+LGDm3$$ z<}830Xwd9@yslnAWSgu$BZ>zgkhJO`JFE1vqxqz`*uJF9;UJv9O2wlsgm7@0>-7DX!KiYaRA$+xZP3ayXvbJ93{#$k z)G_&Op)hRYq=}Lp;~?!6yhd7!;6MPZA><)Aro^7!z8e>^Wfb+wy7qbv$_H z)Gmtf$SdlE+8K18XZG<@#&cmHpI3AnNax)o6B==#eKF@~xnQC;bG@Q;#MaaSY3(Dk zxzL5Ma|Gx1d3`-X;BXf}J%g%(07-^f$h@GV8>MDmxRSh{T(bZ{-Mifx6nl}C&AIQE zbIBa!-iUTR@4K)1*wFC2qVru*I#b^LaRH_z1sK1G^vw+sx%=S;46uRkoCtkp;Wfd zqd=M7NhZ^Eu5^GfuFF0A^{V%?miu!d#I*4d_LEtOl{ymeO{Y}Nswc<|o4?Q-;=|c& zG8>bD;lf3itJz0p1-$ao-Uat&Ba&{Uat9K(B)E%?Rx4!(AVTvDusxK)q};LOe1G@_&$i4ys}CP zKyD+5D%5uiC3vK4GN`V=fcCE@|GzXRjs)drNGVUx9cD&QI9h0BO-oz)@i1Z8=(rZL zNKRa}xl4kLeuI+|LYb>|Ui3$X5uo% z!7>w!ox%nLO(^RANI}2`z3Oc?F+4sR6>|`@ zbfmKD?f`EjG>UIZwVM)oWc^x#&Bbf@0@?#iCbKAfN19&@q;Jre-_KKTSS~@z2(}Fi zn%r0s(C!%>s7EK;O2>B@SvT718!ljCAYj#*uQGMDxH?|_!q(I+&C$thC`w!hS~0W^ z<^yTAE}6*LqYXgSRbaB~N3Ix~Q+5(ba>6->=D=ex(Vk+m`c^miu+rNDa896Vm=^NA zD$Fr`v%?lx9-P@;kT+Q7({*SpNCwQUt%j{wtR6~@iuC0{yAGuz{z~6~Kj(Y48D0n; z1wq>^468;W_G{pNyc5W6PoW{D%TB(yi-zbSfPLL06`BG7U9MbsHsF40FNfZ!=ALuaZt4t(&3(Fw0J;|q@#G#C3yr7Rbg!wZ zt~Xj{C_rtmA!wE)pFeOBp%_vEq#I%1%29hpzI^ckF`rLi@39rNQNgT8p+$kjF2Nv{0$Mz_m? zI}(d?I;gLu)cl4LPhiR8C50Uui|Y>~yd$8aA00b2N&p4WLcc~`kE^@~9b+1I%7v8q zx_I+RFLk7KyFeO!rjJgD9r{Z};d=>3zFJ4RrDac`-m24zXxj_g;VVWv*ibg#>G|+` z*^k8P6AB1KneRe*22Khu#}DBDj_KnMMv@AQ000YQfB@M4Y3gBQWoxPDXyovF=eY9v zFVbuM>B}dSNIlVbb1AWv%gx;Sal7{#{%)kEEMV?E;M`}KGQUK9-UfJ0#e`pWP5U{2=otD}WZ=ga;v z9G|IyXu?kLKn~jdZB&i=3x6KyW3Iq$mkvscCjzy47^=(frclLWo#BJD=5dIlrQd9q znn?TUX6z5xZWcieO87TqQNJe`_T31PxKxr*_;c;(A8h4Abbz_Ot*AR%QbFu}?eb4$>ZZ{MV z9)6FG+kz?R)%$k#^F|slw6U3#)%L9D2hwiWBN~LB%m<3ddU);jPg4!;r*3n}mp)DI zqZ~4VkhdBQY3OYY-zQu3syn&+{j zey_*V!$JG4gg;k&SWdCk>{pJ@jnj6|WV1T&36H*B9280sM>?PhhaVwK-NyDxd9ZcZ zjRfD4fF&|aOEJE7LMq~Jm;BdrcGu#?9Lyen6on~y*3Nk3pqCPkcTnRM+*OR?5&efXYP3J z`=Kx%@_M@eg@EUnWQ2f^wC#YNf+7%#_`rj<^EB`yXCC?V$jMTB&@|8BigUc4Z~ipb z8|j{ANVc5*KXjc_lVHJu2HUo6+qP}nwr$(CZQJf?+qRAA-r2|9xHqEy!#Pov`DHPT z!CvlcAUtVs$|#9vHefQJ-^ZuvA;@pzRb2hd{;^=G=e74A25%TYHrOVt9C(iEiSwx|8&P zfc&e}^bw9GzCwpZ!RRql!{HI2TOXWl50wrzXn#rrGCw$o%A|-^P|!yHUFQvW zxw%jffs1|n$Sff?L1`302n}6pqjtNbc>!%@SA}o7+-pZt5G|i zX)xf{^~`>!j6(!aNHaTn&TRs~FDZw*E?1dN?;|}D83W8fyib>-t;swv=qaD+#re$& zNwyT)TIoZ+Mn;7^2#1GPp(G{!zEAY3!=N045K#K4w&<+FU}8JsI2g#i$k&_o)VBr+ zO$8cV-XR{dnJt8qd-Ngs6%+@h1PuyxGW!NP>TYVhDr+M8Jthe)+`o@-oh&#Ahr4ED~vdl` z`-QcDM)YOFSR#D~TW&oRL6|&#FvKoWI%0SPZ<5ej zKK2&bNy_2;y!=tl&~0(BEUjzi5ZZE`MTUOE;^?iGp@Y*Ob0LD5q9u=xjt(|Dd4FA@ z-;#4KZ3Z3KlYzmDOEwjVyh6_UrfdaMRu1-?M1+{jpS!uixV> zE=?-+;mQ=8JMdQgM*!RO6lJDiq5SH>E@D{H=`I*3`2NXVx|3~LzSe~6%a4P} z%g;;E+gx&^XCN`+ZA~C;n z&$Eg6-7Vh8=>!h^Pm}pI{%g4ZO)U2XT7ZceaY2j++-W?nH}%?0RT8ht$~-=#C#@jCQ)|hpnu)hJ+>t+ zc=+I^sKJ~cMW9%FHUJ6J5in3YC(HbWjkQ;rXaau^sE%_rn323PJRtJmpeIov^Us+T z=1lZj^3Yk23zcPLoWmb9qJ8Fe;H_*sDIn+P5~d7eV-0Bxx2^kNbFiBEOzsi=3~ip+5=ZZ96})y($L&61%WXSk!Jbw=4@#CNEEBKrjP^;5 zK*~iyU!YQ74wjeGp_Y5Sb%HA7-CuJN^L4MA|Kh5 zOPfgr5DS+%Wuk1+Ed+{t2RoL>*z?2?wY9aaGm(pc?y!=RQke>us*mBPNVjQQPdYsn zlg)ei1kl%I35=%-uveWGIcRDu4S-LY7sar4-^25@_G0oFSMBRnnNqY2ok_JK1r|QW z+`D5Bt5?@T&@A}gRUx(wHJ=vneJvFH2wwqvhmjhg?u>)hN0lTakObWLijqhH_qw4h7yyZ&i`@mseAbrTyyw+9Z5erhO|kVPTfiymj9h z$$bH^BA|WJMLZlKHWOx(VarmIU$eb8v>tbz~6dPgbZv zSwjE5q@h)3ElvoGM#NGG_^rQL=kz_|JvhX#f?%yd|+|(mPqg0#=`x>LWEH~zKEH8m@8F(RSG7b#l2{&tP=26MD_{6>GzZB5b-f3q@^{xT3<%U}i=$NbKu z8%V)XPDhlm(uhAIhr#`XF$`&h6^Vt8Mq3s4=>YKiL5%^4@l<(-SZjh2uIvw4kgE)D z!+=o~ihQ_QKJt~4APguw@1Dj;n3zFf%((9SUX$&W;-@o6>Q_I zvN8%bd`g!#*zjSO&ezqq(A{C(Vg2g%x#bwo#dEIwJj8d&Mx6_Y|gFAPy_ZU6RXTqmJL5!kXnm>=%LK&%3KhNd`x2E8ZhG# zGg{bPqK6w)iq*#{4b_pPQLR8jyYMv=XN@^RKsh&}fB{N7N~X0ugMtnDLV8r~@Q{iq zY=jD^B$9lv&M?MA01YOzd@LW!1}m+uo)BW?uPp+pzz*ZS>*-vVOiZ#yw(zTaf4?nX z_;|P#p)|+4_>~Hk+J?MOt-^st1RUP?by=wCVn$I+BYwL;Bcjt#=MEIkTYe1RsBzdj z-VgV3K8v8uPT1r=kD%&xU-s@DKj$?BsS5I&$ArUHUFh;A@V^#j{>sb-myp`Y`J*i;MIsV;gUaQ~?fQr+P$L&!j-XUgUNcvD$ zDgw>x^V8ZeMYOWQ7PRFSwU>Wz`;l~iY@FZh`}*!&>$Gw9>oFdi3x!wc><`RAv!@Z zx=NhyOd(bi%SfNCfN6aU+R9;m@0`!R)58uUsyJ!0!zLf1$GA`a0T-NlPVCO+@i$h= zPBT^K;L$h<#uJp%N%8mlVsfq;URz$s03z@rYrEQ^jiQB(;YCqGVrkyTee;#OXgUvX zsvr+fmy(hYOAW3o6D259JTo>Ac54~oB(9<`#BkGSwr$1Z;6jl({iRcsRa`(ZER-1T z6cw?u&XKs`Bq?S%{XAb{Z)PO!I2F|lM;CjSL_1gdvg`Wcn{MVLfIt(njE6f`8Q#K` zp%OmkK6{&sQc3$1uJ@UIC^h5{RpEmZy!3o^yW;{O)p5D421`n_=T2>4WpvsZF|smw zF^TtdO&Fd+Wm;vsX18N4S^LGPnzj3#%>-9&|2?#lfVJPv_^3Jmy1x>U1dKXesAVKS zmCqtugbn9T_ACudxezZH>;5rW0&fjr)#<{1WCCR4F3b)27~S$5ER28Z!S>6?nv3u= zdr(>u>Rpd4&=gw2@G+$nJ$KflVGx$5#?|=+1y7{LKA~Vo9<42rAps)8-p|0MKvI{t zM2vt%fvW`qH70lF->f5)x*D)^uluMnbT^%3!?bpevFf;1y}Et3#xCNBT*IJ2Q4Z5C zO$e4U_*N4uw3+ysWWGmzbi$|*9(RMJPuJ#vyoj>n+5^Z z0z2$vkP|70SXmJxiIgfMO(-EJs`J#d398oBP}skhK&cbyNW2vcq4mqOiKC6P-wVnB z)Z|c*KO(kS^t8&|+Igr*BLOBFIT6);NRtC)E(dw=8J1c>t9w-ox5>u~Kh!8-nG#`) z@QdByO26e?y-KatX)Suz)8PiYlgZjrhwWVpG_cT$#ehqUJtT-^Q%NzE*fD!+KTg5b zKW~k$h!=LFu<bCb+cp%f`@BM@L@UX*&M9cK=- zTHQEHmgHaQr7%tEMuO!+8fFVK0^7u%oRgiIY5G}SFhdHOYk>qFpuoS(6B;-7eEiMs zHH||-WA(C(GkAzO_b2!OXQ=BpAX^QuTJB1^&Wa! ziL$u4Ns4s{eQAk0pM?X0Dz7EUq_T?Q0u{;fsPp26maQQ`k~kN8XxK(OJB9OYWpx zK3^@&Yp@wHoVbPGH|e%`EP_MXK9|X~VTT2|caz@vNLq#5QEA$(G!=Lz(YhQoq;+%I zR2JHMx{D@`QZ$#M{=^Ig=EV2^JX1efB-v!=LC!F6!_H}2X*=)XyON>pDWH0r*+Gi` z`{uIjGSaWQ4M}*FZXyb+3pPD159~FM0obi#x@haq2X5fCVq)Hy=n9(V<*?{)y=1IZ z-Pf?J%O!F!s+i+T)0}1ZbXeGwM;)HR$*jQS(`i%#Ws~$H-%jA&zny^90+9M<7`I$S zK%gq`1{6}qwW;L((ZzAHZ`K??RQ+>e>V`xYJ*;mBUR`p6Z;chdmN%-sITcIp?CY$_ z*%}l`HG_mPJ6X>Yj$9pag?YJ59+g+KqxI@$dXZ(CdS2pN6)NxQsCf`x{u+c`kaX^> zE?sGxj=n5Lsqjq6Y9{%3x;&P)M91vY$oc`wc-(6^CwBuq?TQKpKpP6XN+7A#Aue@C zsw|str;rO?>OKuKokZkwsNDyq_<;%zOUi{&uW{#MlmtGql+B7o zsx}geY!+{vzC7%sl2Myk*fFf2Vw9TUI5i8$1bgWdZ6bcfh(%{}=JDp|7Pq7ZO{QFE zyCQh~ZHP9-=tou;4+Zu@%m6V~JYPHF0sE3=kZy69KG&K3%2ym5(Iqib&I$G}xMrDm zkI&?nO8krxnV=;AW*(elcjkn~>V;~bfR$`Zq2SpZ0 zdf#f#XZ#f#RcBOqvJ$WWF^NVoVhpr1%pnmh+F+z*ZE$=gX|-bH&A^N*6K~XV+Noaf zSJ1vE`l|~^-f#)cAW3&Q>91CywP{TNAeG(tC9RlrEzIyib|q7mVfMsSt#3LX;~o`bk103U9V^at2~ zoivGo&Bod)O$~OYj#0i%Kuv26Uq_mr$+T%qCH{sG&Z#yCP#`F}F$9tF*a0z23KZRx zsgI|Zc+Xh1`cI%b=okZ5Sn!j2mw22i%i!yor3vM2V$&r_k9s9|0!CqbTNbkjgY;#N>TK*|X>s;H${B{|8e zjio#A#?y*3$Hy)UYN=|h1YerAS;AZV8Qd_aR6InLP>}56FY$&n^zhE|^>hhYH=LWu z-L_t&2~$Z$q8^hqtz`9G3^^=t^GsO%nf_d}JjBVP4<@$wvQEGK97ml|3`Q)A(67S{C9(FPW*rSy=V$jW%dhTPH@J!bM`oMy)zdd6#bVH@Hn zU`2uWNLqqaR!Bs;m4dYuNNO*fI0`46m?dcfKkYSaN_w323Y>IW6pS2`1n*}j$(5?~ z%x6iSn$ezz$(@m^`VcHS-y6?S_V)Pe0L_UCkwr%yMo*~=VV{}sy_lh}s(MZd*Sh@i zSo9I}CK3&*`okxQG|JgO>5gl3>}pcT&tfYd2#i{{z7en~=96-_*Q3rsk!~&>8rNo1 zRZJ*RS2UT*YKlW&%lmozg+Inhk;6hlxMD*|GOZa;SnBbz!{yLN5BMDXxe4bHz2<7K z&@FS9ZbvfX`7&#j&dNIKt)=L?EO>F)F;f$3#*RY@UC1YEu0=~)v^78fFq|+y_B^(Ube0ng3_fcp>MZ1 zE7m7AP35m$W(oAox}un&vK-PS>ue>|TA9U(y4j)3B(+kpm57+T%@snTT!t3_v$|HPB0`8) zS17u!STYv!SQ;d^3?8oW)rwMEH8fiECJ+yrdi67jG}`LHe3J!2sq)wOT>qqf7)h7n zAI1J%lHT2y>Ib1TSEFWj%D# zLVh0muR3tCeD)&SC{*O>Ls`B~PdVt<3ENwDHKy`Xc+J9prA(XK1LD`{0J_gQfEK{_ zZbh>W2;V?4d;-2#KgA*d_jrh?mxV@bg6Ajv1&*gad%{BuT00;|Joe2Q`1n@*hII-m zr<}>iM5e-n0bu6;hi=+c?&;P(TYPxt;cha4pq03r zLmcO5jWQz?_HPuA9t{~cQ$f_I#9qqdqC%tfu!fYyy7HQ47|R`m|9bn)U4H`FF^ymH zhvlg0o`R+<*~6`hYDF02FymBS$Xc;Az+@`F#xK{U`q@u=s_x*89x&~}NO+2lB>+cd zMX7j8%RN^F`(wWodR1Xbtz09D{%vwL%4|+a=T2xI?Q6AiX93@G|DB!tbQ}=$+w9rh^hDoK_Ha}c%*Q`x_HgH=bf*xd zx%@)_sLnMNUt()jU~uxZJeOM#Loui=Xen@2{@{hO!w+tT4xtvDK;ROP+R?z(W2qWY zlb1GRX~@(XyF9FI^uA+k9J~;DgB)3dnBB8SpsRy91Ny?zlINnWh2f3n5;Ftqs3LBZKCi~Sf-|?zm22fhTFDiz-;2f z-|-!9iGGJL?)a%yOsR*mL9A^4dLp_VHVuZP=>Uo}k=d9soPGSp5mrxY>Sk+N)r0Kl z#b|6>2i7wViIfc~7wK1}l9gbz*(lEfpsj6mgB;evX?|gU05~|>_q(8O2&(ECxhj@t z`jRLXIS#5y=7<)fF>uB$N-D6>d|Kphu}O82kM<#4ONJ~CinMsEr;ImLF%3u*nsruT z%f+*F8_dWCKV9bcw3oB~TI(-3QNU4n#&3EbcXoXPHR`K`7D9b%&;fuQx|F#+mpqij z_Q{iinWsdV)CESf*-r>*8vOSLRP3r{kdD0_EzVn1o$rSXYHe)0|JTVHe7#;DhgWOg z=)#>bA837YzqfqS_v8r66%BiL=TbfS@QkiA`|S(YkgJ9$;qunbg};~6y@!sL%`X8T zPsH2N@cVv2%l=S)B1zaKQrBNvx8iq}Ab;C;aXF0OQ#|3&s7ted4}HnxC#GSg z^-JGMWAZ;%LFu?|$Cv0A|fa1W({~$f3KEz#(Fsbjf%`x*wLdhkYS2CFL|}3tsu$-6 z_!=z_P|z~9Izuv(8Ul_3^Fs&sYIWI^c0C|t7@CXKiJ)ru;VSNN<#oXra=yRS#^WKN z;~tdmGfy^vpy1^r|4_NGre2ayJIUk>_$!lW^7ax)E!p0~p|6y3xw_*6@? zL8fYXPu-&I9n-aLo7LXiH8ZGF*cRS|a~CpX*hepS66C@t6x z!Qo396mI2nNj8D9@Jtr}yiBk%&E_n7EzJsvB$qY?-|?H zV!iKS3tYOpQPYGYIjXPDDWzEEV-y`!AXsJd&@BFO_drFbG!ZC$%aDU|qnTVzM#5l+ zWOfmW7l@D2Vo6E#`u#R6!hcC(agTPLu5>EEX(!99s&c9-ZGvuXu=23w?XfM?O1^q~ z#)Y}C#jkEZ@nvaUR#w;F(^F{Vgetbb;H_`|hFA6eCbZc#)ZoO;hMnjB-DQJMnvOsElQqn>2;C0?iK zye_U?Uh^rCL47-Oe-s^4@Y}=`Pprxc{~`LfYoN@q!2MqSG5T7xa@>LSItzM z5@Jl2T{N4%s_P#m%GL1`+Vt@g-FjKgR(}z6qP~M3tsVf0NIb`eWRG zmFy5qEn1sHHUO)zw$;cxPkbw9A>IBet3f@Y{<$RKd6>ST@Y+#Uu!1e{kfc+PS-=TztS5GC`(>pr(23-@v|Yjro3gamDh>2CpX8>G6kV{~Fy%C&7sj%ej|* zIc6Ku>0VnK0e_;lZJ_-*PsXR$#W%}n;Oau6QIE!PH?!`ScD%eSe!cVU`9!&M?M~GB zQ`(0aGgcbGzIl=lf%^|l#EL74-F9$*Fg7$y!wBbz$ioNu0fl(@ z2fK|J?|6!U52x7;k+aDfh@NtcVKdA+x|efN)v#i?J-Tw2*trszT8F%W3=bsqbMF+| z^<&+u_`UNs|YhQjq$c-G3zuW5~-W>VSGijGyPR2S@>`ltLjDWMz_@|Qcuc9dNS&^&^19R+q*yPaxKigeSx!!V5+&{K3AER)rZOQ&{1m~19 zzGPYBcP+E&?w>a_$0HVo5{7vW!6)I;$H!xcO&Z_!ZxDvj%S=>$+Kt&3qF@H0|6?Ox z83&~>@-YPmPIDT0w87kK?A@HlZ@v>8MxVs>F3p;EToM(XOVHe*>xR$7Jd4zzefuP=g38hjkSzazloS6ZDBE7{*zScQ865 zGq6t>+r2~lyG+(1=aQz{A%qU0LX+p8uP3W3cQp)nu>>$T26OMjWqNcaQx}v)dTMC3 zXeo7(VBF`k@|=*lSRr*nL_Q3Xq}@2dJl(aQF!>S{O}Sf=GfaIrj7%+O5Q+34xTC8y z9@Kzifek7jjER03R41_g31#weHEUR{?1J3F?9qG*(M$CyP)35cp92%NgH@8AH%+xb zXEBV*oK8#P41qe{&hnfqDI5vRMmZ!0{W_5t2KB8(S(j56^R3#NnX7yOg2ThNEnl(| z@y&(dw6~*3%cxPOXsU_t?!KBBWIdoNK+WH2E{B(n?~&`_m%_dyau*W5=HWI_I2;O!uTmLeELa4AJX6t&5uL+KMe?)Lx|L>dl9Z zwUJepQiDGde>CjT@2xQETwn2kQ}Id4qOAqRx_m2^@$%$&syfn-Y|~ z#fUMdEzDFwXl|~0k%cZ^2D8QH+=YPccjEi<4ef_RY|B0zh&?gfCPvV}?4!Il%Dq>} z`-N?x>oj|0kZ`{q?4dw-7P3Q#DJtQN*!gEUvgLT*rS)P)^IP^tN5tLYh=MbLCs5#V zlQ_@t#5xUJu8HN%X}s{K*Xm({n!-$2hr7Fbse zeMie$F7r>p3a0O6jPVH07_O;%V2`iRHPXPQm422C&}irh_s z4&l(L1Q@sX<;=QCK6!R0q4fKB8M-O-s1q$)J0V$?6QF1O)aaW8bFrgZ&{~0a=x`kk ztnOu)nEEVM9%E^Y_V`ZX%Q;GCNR0nu1o=lpQL+uMy{{pH)#vrC0e}it2P&peqZ7*~o!S^>O)_D3Rr zQ98=pd5Xa#fIc`jO3v$0`RZWVP@_&-QXk_G_2|w!-IYx>WG1V>HKXrm5q%cHn8Av% z9UqpK6y3rY-bo_~wcbtnk$-c(c<_~74uTJ{McoFJ^x9hA?02n6?#duWhVo6;pdiJd zZBx7o+LO4*No1EMpJ$*D8vjhNFgg)|l|DaSEaiEAoPBf-U%9w5h2p|@Frv)0v{DCj`9_!+f+nykK=cW-X}2&FNp(lFv}>cO zU@PagIzJ?9A`E~~V1(+YIK=EGG(i4g@DiS4l7Ndgy47xs8Zh&y{Zx$2j9~d;)|VP| zn0S{0Qvy>4B*MQbKP<1QZkQ))kU)nGlhAtMH_h3c*`O|O;Diffz-pm!j?s<2x?d$u zil><#Py!z>>xkU6RW@06>@Yb<$ISpk6Id?`sokd-lgU`aZCN%a#m)orn+JhcTju3A z$`)+ZP7cppwZZOaU79^ked5)0>U=Bg&H8A1_Q9KUkC=ob=goRV=u7aYJJ2Vmp;1f| zj3ifbLX+l?AmMzjs0``+aoFR|<2;i)58&eIj1mNl+~0HFihd9QrJ8Ky&mQ&IT%c;3 zh!Y(n;o=K%6njG~i$2>U{kU0{e3zIaZ#ra(n}wS&fK588P3XBAKYC zi9ZAeC3SmVc!8DeY5!j)jY5D)(X?*Lb7V|Soe zjMH3gu-k6hO&0BvHVv0DMzr=s3l)}rE&V`DS zXnGgXrL|p!0|ScNlWk?fzlgHrkKuYt*IK`=;Ao3JY8IGm3Ws@7W_O*T%OuMxkDaPI zd2ry7p{(T1P2coST(9>w4x7}%oXA^kL;XkdHY&p-;FQqGay z)N|UJCTHeBK~i3uB%K8{F}9^Ij6tNSA9Q{@fI}&@f$d`+RH5#Nv!{AiA6%~6!lLaJ3!d{w z&>~SOq^+^suin}q+Z9xbIBq_+eF@uAjpa=yZGQj#zcR`vfh9mX1OUJ*>3`W|G`2+~P6Qg&^Alp~*YTq#mqIv`{z?&b5W|jgR?3}uSe$1i&|C5ycBRUn^!O5qrkihqFx2zFB3f%FmpELz4wV^ zYtBFFofaIjap=x?8KGeR^i7d#47*!>nCPw|+vhgdt#tLLk8STo(~mL;eieQg87DdZ zbD3kG^{##QHrm>~kyYE(jo;yoAbM(bszj`qS=W*23St#3ofHTn8=Sy4ctRHUbo9dZ z|9*QJ_^QG9QPEptTDCk-jfd6umZ~ivXVV|v4GQQtKzstPYNK=E+ep?qqo$pR?GWkE zeST?K4~jd;*+zS3noA~xfwT6#23WD~J{2r1!u7(|p!LaY`o3l>jlGKL(A>YBH zH8>W*#54BPH0c>AStL6z`lZNCL(YJyF*!;70%7caWYjCFa5{0>WXGCaoAP6TV8rzwuR>cqK` z`r8aNjioyvB0VrvErnDD%Z(&LEq0b3bK1O^CXycRewKa&T?Io3T~tD{F`qh)7<=8b zy3_qCu}g1&PU~F_xE0z-H_#i z1SCss;3N{dlPv=QZWo;%C zsfmXU;=B&Gn_4Y~`IgMrSXL`N=apI7Hx>}`4ibo$R51|pX0VnVhRhpwpE;@-YE1nB zK?#)^g~4b_6Txg}8V!rBKr02IPeo9i?m5Cjx2ILK%3_tZ4xJ}NBmI5VV6_S`8(S$b z`H)Umc+x}k%V(@J=W&Bj;Sq?C1)J>&D~qh@Mkc}=CPL|=C$O4NKBAmSu%eFz24#lm z?54GxI6MkOosjzeml|rU@awUQ@PUM8H=1U0I@*Q`_nd3*zodl2JOa~Oj_kFvHPcB} z4hIxVrz465ma}yCy-a9KH7aKE1fU}`E<% zD(9!g31{LM4e)k4)ym?s+et{!=@){F!6Drc*)U%c>JXQ$S@2)jD;6@tbIY;;b<~}g z6}E6gQ`8QP^5}s&xd2nH@y1E;oZiV?svJI^>7dt?R)8xyIP=0{UaV9m%hCPFRGOf6+?T;Y zI25X_5%>tB>5&FDYiX?Nk;l4(Q+`?vYd~X|1<3bmOBqAG(qxHhih(;a1iq489igdM zfmpdUn`qn_ax7Ch$&iRzj_vz#w2W*OCUho9(8P?Cv^igiOafTRA}6j7PyQ|9U|NA8 zLRLEPM69n=aB(}s06Uo38qAM~5cn;SHR8mJ44X&?&AD!p{lsq{v> zXU0lo1v=nn3$Sy7RQ{%1mns$}(?)qx?JKLmN#o(`X22jq3uI6U95Hm48d&Esu)!e^ z6SdAkHPcCk^)N@*2XQ3T$v;`*kh>A(<|MvJF@gv^DPeVRgE|(p{m-(Y?E}U^fc^~>9bY;4nJG^PY0#!!mUNpb6lUk{>M$=Q8W;b?%lrO z6SJL9S8gFj=mY3?xA%H~y^r%_f2Zmv(ke4e+HiCxFLS|vxeYyM>$5USwIXe4```N)D zDuXj<#KOK`Pb8!VDHiqTTopm__qkSv(w;x;0@yO14)raMO+-8f*d7|7P~X)eqs_DT zGU3;fhj16wW2_bfEq712*{AQSr>}DRjb}NKjrrd{M)ziPJBq=ZllOFyZ__ui1n&LY zc)P8NExjurt+iz0Mdt5|#Q-#Zbu=TI4<*AmJi-!z;<(rZHi-Ex9zK?V?0rCK@j?g9 z@cbM$=HS>|e0OeQqV2uLU++@K6gA3ej?uw+1$Q=m^6A-oIxpC@~_ez?5xbYBbjtELbSa_&8>e{Y!bHHKN9_GN)wOYi8D zj`Z*?zwH0I9LxAr;9Tp2xFq_Gl6>Qc+cW3Amjc{W4V+O>lOj6+kuG&+NW<>}L_=^> zKa1z2DLR+&js&qlteb##R^$>WD+esn^mdJh)8l?OdTj*b^MZ;7e^zuizadrl#U6)m zNq;}wL;XIF1j2nUn^PQGIlL6T{J+5j-)m?T00aO4u78uF|0$mJPpfsYv@`$rJ2T2< z?TjLxwD&@NX(Ap&ay&?O%Vvb?euc~Hp%LH`WAn#MK=G1pHl?iGyAB_zxVATe1q^20 zTq77TnAWf%lmWv~Pl)H+Pz?Rc@!7H@euQ8Z`r3be=JM7vK48bxo~Y|laU)YzGgEbg zcV~NhTVy|7Mig15y|!TNm3awjbEc^iFV*+oyyzMUce}JFlTF4*cesU0JQvGUO#$^{ zA;x6NWHDoRu*(O<#`Eh|znAsWR`yRUOlbtJamgMw(0O5`n-Qa!&>g(Z%@!(LmlaDL z!;TSd3G?>NtNq^UN;C5|ac%WXnGSIFEJJCT;gW=kY}CETB7<#HwCIK(=e66AtEjWe zDx7gdR#Vs8X`&M%+S0~0HwkhU@2gf! zvujBt>e4M2|Ar#{@WepTJp%!z%CX8(8qTDOHNap69ML!&ut23@9b=Df>NPYRvt>7# z(-$(2ppD9sHS1v*_FMmH_0uShvamAaw%isjHoNB)@Xkg>RupUF79W{n1m5j=@X9JC zRC{mo4$A$pSaT1C+b(Qcu-8@|-3))XW9|nNyU+Az;6^u*AfZna!%HDT-~NO=niBem zRD5+ab-P+T%}nHv>5t}*77KTK~34IkUROlEe2=0aZ%o=%=s zcH|7s(9_{gPegCCk0cKD!h66ZbAXP4UZ-P{(k9t<5NV24s)Wi7qq@z`Mb6@+Y1_D|>pqJRZXB^( z5DOPgA0t2&MQ5orW@dImZN&JCtb84dsCq}HSZa#jW zcbY4g=llAfd|X5yUwn6->E|Mx@I2V~{anbykI-Mp=jU{it(w^ZX>@aZm2a)_3f1Il zKYK*;dxC@<7m}aVy*ni#ScjYQ;!Q~c8NC`1`W7efsY(Ddx|S#KDM{$l$>`GEB@q6V z9KM~3tF(t!-V631si-K|39J1Top%fdV5_ zVm+c8v;jO|DcM9ZnFn!^IYqNNPEaHbZ%@eQY_rgNlrcmAM|gJA=f&mRXXN`l26$DQ|I+LJ6(j8;hii-mXQ z|LC3t&VPq$C*};vcg%O1q$EH;8GAz1&0g*kqk%~?>IOh;(;3vo{X)yIIo(lMFL0QD#3IhqD(r_ zI8q6*g@!0W3dokjc})65BA&+|CxIK+MJ2HtK+LCEB!bBna#Ko7nvlXADN7S;UfQ_g z6Aww9$;spFcNrVq#!khMR-k}pnZ2JSOuyKWWUxszd4&!oF=TcuYKfs=nDJB=vckd} znM6^9T}~>s-ftnA6Bom&jtCl`T=c9=P3zzr8lf|5 zGQ7`R7quLsdUEBb>x2}a;80~&*Np&=h-O%TH^oxKb%?!V%vXLi%rC_*UPcvsV~Qts zGy7xCADl#?b4l4)@`cj~3?G2IHKe__CppwTw8j4nvG8@it4JyqhDWP6FIrzHEP!#& zJqr&l>{Qs|V2D$uI${ASDq<)Zzz7bsUI$^5j!w_s%Idru<~)X?&EUjZp)iV;mMME* zYRM5jL($62ewexCWt9?R_9x!j!$`H^VT4H}{$(#Nh9=l$vhvXiuhxJuw!!E-wL>Ht z#D}jz8VIjCTRS1vghtGk2}~U09B)aKy|sd}OXm%PZ%AjviIYmdYoIfFBjUaIIec|A ztZ*77cb`fQ#YIlW#Y*8yt0;MlvC$P@5is&I0%@EIbHN{#kfDMBb^4eIhb4q1wuA_f z!b?;la!?TVZGg6raGK3>8y0~RCcQX?L)DyPOCDJ=R`Uh^ZYe)v96r8Imp2Rmu1ou; z@EthM09?&4Kp6(EoW9`oOZX@In=v50`VTS(I2zUK4N@nBXu)=TW0iJ-=`OFy!LFaw zuD`_P;0^avX=QT<*yL-VVTN#w^H+4Y{D| zB6EJ=@E})oLs1c9x-pi3<*TLuz{E;%W3r>+VYc$P^T>EXZ=n0KvT>Xu^v zFY%(UghvS@7EcEz@B+qRulY$p}lwr$(CZL@P9`+nVj;JoaIz0TTmj?s}5rbov6x6@wm z^X$;3?s*rww4c}O{*9rI)#EfpN{#gOjxeC*?u21>9JQf52s9$PR`umnXT96H$vUUu#KPzgsz^omN@Q8mCA>89Ql5-~XfM5dOt^7tF$a5hw z{2i^aN{!y=9?Y~UHV)pLgb%9)WbE00Z!2hoI<`FbK0{awwq@Utd+-TF>44Cy!GBc? z;G1?4Sl!8wD{-vg1WMOHK4=y2mHJ^?v73I>X$gvYW7_F23_&0rjxQ+-ByL=8(Cj6U zrpbv9*HjA7s3?*-L5ZE@Z{++kYZRp5qzqgN6ZnN&2XSS{1Z@zn45^V^8AjP~K`kVM z3~~fu=#>-ra!y2yp)^3CsN8+_L5$#M;?I5oiIO%0fs?`QVi%i%!rtcWXv7X%7=d4z zXC$O&r3PxweNIG9A48#>!6EIR7~0}~ozd1ru3Q8S8>W^}e##UtTCoTnX)!i4vP8V2 zll;KMUV3VEgz|SL)%selDJ1q8`>Vrs3T4*e5N6$h!y!V_L9i4Ec?5wZem&qU3c#N^Mu5BBrWsh4nKoaW%urr=Gr;lwgx%GPa%^2l05&Ddnw zL%c|kVY-i&H3KSZg2)3QCVV5y$ED?JGBkbEo@WrOt28`Ik=dTH0$o*rH$1!6z<)Nx z)y&DcGkk+eSD`j(yisfb^qZk2BX<=6tS=1}juZS~01X-BJ28uavv}4lOcU7YJ*}yr zfpumk2|-hS$65>|LkDAtC`mAFd&a&%HpcK%R~TniHn`G_**!oWuZcK{b$&ml3sE$= zzD}fhL$-y2R=BMr1?ne1@(vSP=Yr4<$`Jh)l%n!Osq z;7a z@-IonaE?>KE!U%`Ow|kZN++ChFoj<1RM0kGbi9?=zhpYCY8X#>sC%<&+yi{;j$blJ>kcL(U9@f>jJQNeyLgX zye{_?^)H#gmAfzK+0h)BMQv86&w4+c3JHnKKHbnBV5*Kn0?s$&(IZe5if?Yvxm zOl`yVSip8}wx8W!`TkWwbP*zXwH3BOwio#&S2kZ7@OZsgyIb9Kcy0S!ta!0Cc4OcE z7+RoU)Hl?TPZTR1r21W?qNj-$gJG)!qE1as`JWeBrYdHqhbU2QB+CrcUKBanK-Cyk z72XtCvyPT{imt4&=`LC8)PR)-piY*D1WKK~*XBJ9U+AFR8Qm{2m7&nn=92=kx!kFbm$-bn7K$1Vfm|v}u3shScC@WaW!;oT zc;c1m%^B0b_q@{r`*{d?>M_F5PnHq%9RI>E-(j{EVql&Ja%({vsQVYA^K$@(0I~S7 z53m!YB&Rn^Ux01J=UZ!0-**C6l$|bz%nG`raS~sBE&2oonTPJJLn5GL9ESL3425?x z4Xe`Q^-s$1eQE2Ife~q0fQxEUbggnqKuFl;fDeb5(;+BJ9jUvXg=!xY)h6`K_-1cZ|HJf!Cn{Z z1URD6HOvG$VFek|4Edb+q4p!Kb(S*j>2qAsk|t4^>&&MCSJdyVr$EN_47USJ4(||9&pgz^rAL#?NWA)8;+DlL6mm9$KW^R`%^4 zOsNhp22=c|7< zYvLPAoZ*WH2gFaN4H^K~$}+U_noM+cqBh59RvSE@Uo(*HDStyjs+q+}>7qH$PzY5p zfl842SW(kGVquk3SkmwNf4zcg>a=Kjv?-BCX6}}~Xb^R3fM{v00D%x{vG5yqFJql0 zl$0VC^g$~1*E=sAr|-0F*T%-?m`T+qlQO8@0Dt$no%gdfAlP!y@PVt7+?g!PNlG>W z!X+)B)lj9NqqOU>Z-K5I6P`3aJN_kb=TLA>TyKP1yJW9R+_!j5G7w)Bs8>KWkuuX^ z;FG%8P~6PlM6it0tVBjDt9t}anI_MC}qQzh`7vd0)- zLO;Q)f4CQ*o*?sxV@-_LKgHZM43itqPuVt6@kr%JaVo)jVG-ST?PTn+U!N>*X)$>Q zpfbHup*G!tX!V`P+h|2P0$Dop5+JvTcNhTKC8AkN+K75rY9ixzHkfj>&s7z2} zTl&^Z*i^x`)*g(zak_M4qIC#t0C_jM@Lrb%4D@|1?g8lSt>oboC@6ky+)E=f+&c zop*A3zL7h!hS4&-$Blt1Una#s44+;eJax>u8;Udf-|WZj;(0Q)#XYD6K78BHqXT)Q zuKVyuEd!pe-K#ky%Ewi)$mOFtdI#u4)GNGIkd`!0#Ro>+0qsD$xWw7T3tLD#?=rM3 z_??f!ZxKAz!UcTLXFJ+tj3N?m&zhhd8}o(f1zf|iN8368y5d#6md1P!XSgJ~b7HOk z*=3bhgG(_-IT$QPe3UjwhqpMLUo7@PIX4$XjuRDIsm0)Y?|h^rtGn7)wPI}u)MphU zH!shuvJ1wDh)INej`;pX$b6_)EI9&aAS4Es7b!Gm?|Q|xEt0O{=OQ>D^MRQ}NX$VB zxbBW=P#N~Zd1v^Y*mGE8fHX%|>X_J!YDw|Wd;vHiA$8j%GX3!lJ|ml^9Aa<}qZfSX zRuU|d=xw8;jFza`7|N@Aw)?`+-e%1=o;&JkH_Q1+uHWj*2MkWT^4MMQdkMWOA$mfP z8ua*!%=cE%>iD~S^StPvmxPyrM))obH%t%`sSYo`ndV_zr_*QDaxns$K-aJGtCs`` z?v0>^Y(kbq>c(PGO*1R^BB@}yDKECDJ)3H@N;frUa7u?wO~#mAPj+2ZX`V$meGnbf znL_XZjC|&f*$0C=Ka)^Xatu`_#qi^5CU=>2wXe5H7GTBB=eMnQQMdOB6Gjqgg!}>n zVTG(1)n@yM8SJrkd6?xk&Vpmw8}RwbddoYbe*pDzeGg14sT{EcQ=4jDGhLXz(+j=Y znasFuKsL1?FOg)y`c#(P<^C#8Us!1e4bw6>P-xv@SzvUP$41+fsQo*!Nm1yNYIrOH z6NRWBa`ShbK%L2V+MeHDX98q_8Iol|fvWF@A%CikpFjq6p|uyGFF{XUzNrki8)OV} z6l@Lb%hJtwi1jta}ycYE5UI>0?CCR_FDOakwzO2I-Bp-QqxYp=2 z(}CI@QU)VBz&zR`W4Jb95;mV~V)8^lGPySy%qL6(B8IZt-Q75VYQkc{SeB5Pv7Ru$ zYVO|NuPk$qq$ z|Le}3-7qEaduiyTZ|JE*LMLRRPsP{$xqF5-)X0^?Ah2a%uGQCv^{io!G_VeloHHP8 z4x@V*;OqGv7n>~(ORjq6S622-hQA~Ij$bIsZGvT|$Ok}fZGVLBda~0n{bQQkW_vPA zTq=|RwpuXlOLMox3Ok2auC$pV6uN%qe68V4JpO2{4HmzJa1?vZ(YckD<3d|Cfb@ym zy)#e`#0){%Y8_||j!-2NMnnc;y%13Q7a)3>m!jZ7B#|>t)|CN1QAb)tp9&n*G&xLC zI!(R@JxM^9{jvsTFtv0BfVBc%ShNNOvUAD`#IJk>0tCz>I<6_GWvWR4Y>u@U+Av9f zBN(VgF;9@91Yd;^C1HsQgo7LY60{^o|KMw+;-YU!e>KHr!09nt3_#)VK9wS3vkAV) z2%?PkRxOS^x15Z!cNN4X)+x8A9mw}MbnxX?x)%y9qCER~SFTcLbVZhWi&AcNHVm@> zV+pTR@A-&dqdIZuz$h}ggsc87KVdw5e)(nPy&C-eSxRH?`o_zJr4^nA&C`I^TNT_QLpD^WW zfxK4pV?QVSYC%?Kv<`+PKao42BgQH0SS(n-M-Z>(rqaM>PcO4H`|QY|2b6JDyEW;A z5&VkIVaL4tiEuzZs+`(9MlcVI$FoAhe0sr~(XZ3B9rtuk>r40L3fV4#e zMXtN!SId)4Zk#?klNc+3jPs1>5fHe9Je1Y5{tx37h6HSzton5Cqq1Ri_z;PE6E~@tv1%NSa9t3^&UlwmkyW4bnEmwmer- z4tb$Jwo8QcdSGt>RLbP5z24#=N)87v?T4rqdnqO(aHk@R-qnk=gGK&5G)k4?#RRaL zM`3)=&n8iweux2=Z8YvN)ne!c zcu3$Fw+t(AbMxH;6#GRL)a2qg89HC+< z**wbfA%G;e?W$SbP`*2(_~}+oKKyBIBDnkZtgsZKmgVW!%n?QK#$&ceb^7CHf)mak z8-0)+V932-8yUEG4q8sgZLG{_ToE`+iq+8QU`6Cr=e2fp|Ncmb_%3`{T;*HZxAJ)m zKjLR;<_*v=_&wm6hXCcs0|%I*Y$4fV>H02>Q;G#M={n2~=cTyh`8P05%uINb-9$Pq zG+*3QW%_;FGhZ^1LvbG~h4s@)?O|kwEtHv&w@~lCbjET-rp2`JQt#C(rP@D};)um( z>7$4RnX*mng2RO{JiC6GuO(_Qiw#BSpQ5={5qb9s7xx-5- z2}mPmA$a4d%9#Ug6YSbAaKh%r{FQSNVo16`ZZY_oPTCzD;VoDF!@Bd3U@a^oZw&9 zTbrc=Ye1PN+&J9+sNm@Lo~^}9_W;1ab_?CaWbv8Tmc@pB*80nvrL!0F=aa!vHJ&vd_y zI7?z#MGYPyyeHWB(?MhgWspt^qgs^`iYvyc^`a7pR3qmE#ZYy4?Q|TOKoTDz11IwE~NU&FP(%wwOR+OzoC9 zoWNUo8t45(Zp?npg;R1fWtJ+nxF&m~p?^hgwUSXhzY-+&Pz$U6hhhiq7DXl9WHSg; zrr#ttA9`46coVU{&1N)(f2P7mo^J$wHDuGOY_&Wi-K0ZDM^^&hqt1^$LBF~ej zGm;m`t$8Og@V?$8&p_gL9UvIvl^ix?fH_P<2p&WxL1PS6`lFL)FsoBH0Jc%$jUBzc zO>7^%9YHg0#92#>fscXb+N-e|Bqpf&Vt+q>A9NAV2KkE-i-I4ne5@FLUOyjCb$qa8 z0_K?7A^=S(#a)o7lRBiwCuV0(F1V|pCyd^S$YN&PyRmA_N^{GaWmp_mm%=sreTt($ zp2-4CvqmycBbx#vJF&qtWq`z^AN$}6q4CX^_R@pMtmVd`#;hIw6;xu21l-a)HN`FI zme~e`XKkCAC(v9qo@~0M)tibt&an_)=xrF&i38L({|E1JXwsGe>Yz-VVJ4t1-vy~N zygjtG!5;hh@88)X?7&%~c6_`;hNs(IfI}Wk2FAc@Y(vSfB|(BySc}5%&nWLf}B0Gl&pIsYI5x0?deKGg9m#gO$VNybh81Z&Go^ zs}K{5&_U_eff*glWi0%h5XWlbBnqa902JdB|3i=V@-qJnqbIk}Z zCM?c$GZF1Q?~m@JvY(N(+Uh@$k!Bt+ z^K?fr#L)+n7;y?if@Z0q1Zr&-U~hdjA>x+YQC#~uXU>)TZ#_|tQ?-{^&7|rz5r>a^ zDC*lqfwU5^PLrH`v3Ndd&OZL+(1>Wzsgw*{_=@R;b>zteW~a23V092HuW^;D0l0-_ z#kr1n^?T0M9LQArZvHP}?UhK}A!6!Qez*mFmiRC)Lf$c)$hV+UfB%AA)Mr?l?zkCH zjk3g~D*{3igebrYr*-7+keF$pk2dYJ1zS=CThB31^BoYJ3qhFik+GQ$!v)PgpImI8 z!uz(r)3;rJIod!*Lc*p_?n~G;qz@S|@k#tue|k~*h0QCO>O{ic#{NLspgpM|s<4=( zpPd1l9WEY|W-xNE@i>lDjIj9yyaGSG!+&+VP~kqQfIXUei1z6Sfb5##aWC~{%1Xdz zzEg)I7f8HVX^58%tuh2BX3BB1IWd45v<+mAALeUP6>o=?P32Si4}x)%IY^z6jNA!m zakWVR1@q8ph17+bQ#PvS6}b{FfnBu;EX=)lcqA33aYZ)8VjF$CsleZJX!k5Y^2 z;AVQ=p39HKd03_usCD_acKJS$ZTW*bxJFFyOC6%z^>w`nVF~3V*hJJ?h(_K4YxCXI_+(*qldpF3@Q+E!Uq)yFY)I@miK3iYi^CtOIf zXhgH;u-9qcj90S*ueY@ozA}4%Kdm7=!yHTB1yr%fh6#JZx`S9-H-3j#BaJBEOe&I; zz>gPNjtnhB2i7I30sjEpZ@j30PCj@#%Av%2644o1u(cjPPc27W#N`%qG_@YyP*^Ru zPzGLg2P1m8nUvun6?~<IM$|HwirfdBC!|K%|z-rh;H3?@)#HLzRG@42*|GV%n4l z{#y|TYwn`5My&vAcz0oz@8NhLxR>d*kZA;Q@l@?e{>KIg`c1h|p$v;Ig0@8{W)Gs%$k?l$PaY zv)J7BRKKGyg5!N-sUg*_1Plk(G9X8d;#sLMdGzfy9^$r{FikRl-z;;x#ft^1ASrr0 z=HnUV(FXW$mo69+>ldEb;#Jz0n>G;!4UsD^S+tG?0DZ~1SzC)i4NRSL=YqPvz^h!T zKEKNb23dLPnhnc+mbgSqUg_1uA+z5o)BF&aM6@J+6{RNdx&4b~~bFyup(yUau?7MYJ|)8y4T zgQU^VaENDak3uoH@bgNgnIixTm1=3Mlr2e4z6SA?&y(POuGA5fUma4zbcc~{c>46y zu)b@McbTK5x@9K!P%1BFBj(58itcrL->9*uNMP9el$3I-Hg6rqLOfrw}jSvOXhY zm5(z+^$S#ZBi7yH8}(!UPpTu8qAE2IBE_|jk+ zq_Ck}Mdbst>6f54X!Elm@z;_Sb;3we9q-~a?&Zw!#Y zsyG47&Q?f03OAFv3=@M0i8~cO4SD|x>PxH|tx~4j{5Xhl#inVj&kYL#Q3K(nkloP* zRntyB`1!^Lq_`0v7{jfPcon(ROfb8PX`$&McjM!+>(${pIOGd4pI%9XMcwNNin-FS z@}w2fxP(>K8pW*HLFEuY)oKb2Rpp@w#uRxSW~p)y+K32RAVty1$m@=`2P;TGszYhL z23J8HMeYyGh$m7Bt1$wq#MH)+emC^2p2}$Dv=E_+H|&4rw3EC> zdIUIulO#CZk@~5Gi4Hp?Ekt?P%=YOa{+8#Gc?ae6UES`SG!&62j3fOZYX@?{%GvSr z7m5F{xe3)5yL0ikE=QrW_|WGFC1K^Q^|RBEAdR|$>=asLN(Mq5sySj8>+hne0Qpd= zxrR~ALX7qWj)3Xa)CP7S!W5|32az_BP|XpBCL~4$aaPEVZPEwl13s{8I6>8}_7lSj zmFUNuq!JB8{6THthVVc+hLmu^OXFQ1BMYxJBgp1Zih3Z3mv{;=rwj&wg=BxAx;;F- z{knJozE^|Lh=E@QBI$}SxW4-Gy5FIjp$qZ?54LewSC;PRKi~ej-~3)bFf-mS8RbZ* zisnqZVU`_MWe8-dkUfpL=`fUctb+LtkCkXuR5SAT?1JK3_Z4iwzz`RCN%4eTqv$2+_ zZ544R@-Yp#30p6VdB052L2W(u;teikQr|`EkHyqB4W?Io_bQ=GqKa6&mPDaz9V0IP z`+Q$}cMGi@{Vrf`%Vi(boVn?B9rPS#ZE(Nv9iRGzN54abO2Yb=ILK7geN43oTmlO$ zaL-eudI_Ck*F-4J%I|nP3ObMP(DN9pwz5`Wl1A0xklLZyh?>vt;$_eCA1@=eqIcZM zt`zZua(~aZhAsvl=%b2E$ge=`pz|($jfd|0FaE_nl4EQUeF*ps#&h_=D7l8!ov^88 zIZ`qCA-aH>R`Rx>cJ9NbHb=+N51vomO-KQp`u^Ss-aK7;_|xn9q|VRCj;ei5dRCxb z@yz>6XrEAf94Uue{c)>q;QKRV+eGhoq^YZe#*wP|y> zH2O#)7fxu)$_D1a24o0P>uU3H2_HD?+}D2Go)&BUh|^ij0bN`j%+Yha#VFz?P<$*8 zHxyH4a@O29$<{N4Jj5rU?tK4-TC!(PjSu~Xr-tD6_r1Wt1kFn_%_(+G<43aal^eWu z-XG`k6f(_IR4pN(o&{X75#(8DqJu^24uPU`r=GR$dzIqGbXSW#0K*K7N$o5;T`NL* zUaW_^qwgvQo}Pry_{((t{SWVuilt|o*Kb!0JG)MrnzulEd7QUd?C-nj;7N zXU9L3QbwJA%Gs(Hv_<7m`{N7c;5EXm&e3#D$qWw0&<*sP*_o)NL&?lLCG=2gOF98NmCf5mP=mSVc%WY28i*!n*stsq#XtY@| z)KI8E|FP@oa)7H@Ev!Z}S4UC2^l6?YlG;AqBt4_Ixd|llNkI5y>~KE`Ob(Ta>~l-q zOZ!&1g3ngIc1E;uY|GyyqfpMM{ovVfacQ0V#0=&a5S=+6gh8YI&} zd3BGyBIE&X3q>l&#hum!BEz`QEgF)@I2nh(_x<6{=nWrP+(nOHMUFqvW zMGv<4Es-D38KKk^G$*RSTu=j*W>$wA3J7jO)wS5d!LFc5wkH~PTxp4!^Aqe-+bIy3 zFAiMcXeM9Py!CUEsHAyb3|Cez@4l@sa$#k?e4oZixN#GT)+L|aZ_cRE4>-231nTH6 z85aXN0j{a1GTK43f$8WjMbN96y>ZkwCfIhy>ZiZ;cyVT&W?dh;H^+H{e&XCge7 zLJ&nFnoOz{Mm<*C3D=L=WYunEI(~j_`4oJZN<6^tN8kZ_4+u>na36<1dQ5}A8+)5_ zs9oM(;hB!o_6Ot-|AGsTip%-tdFDV>RTVbi=Y(3MvZ)G(>|q-9xI0mQ?`wC5F_FkKPb_zsb2rtZ zp$0ApybVKZL(k9IcGeeH4oCwkEm|i%LV#q1GaCiqcc7h~&)a840YjjKGD(9X9!L)E z-{~*89B%K&%TvTFw&Wyv$V*%3IC~@A)8E7je^DE4`7U-_oSE~ij=HD9DJPwq!MpaN zL=$%AIT0<_!=EorQ`c`P53m8ebCeI)z>_v=;Ud7*ol#WTZP@H^L(*j7haqRt*lhX2i6;6Z zFuzg-xKjwNkaj#KmtAowQb5IU>ftGNDq}xzb@UWmc_AGVbv;!TANvc{!ks7;RSJe| z^cHtN@k9y}iHu;|RkY@56(U5m7Q0wEX|HlcjVi|l-M+3(&JC-Qz7R=%CyIUb;nwS` zJ96OdHEu-c%qF=}C@7Ba^NkX^9PYpOmx%E0Q*IjDAp$~{eZFjgJ7Yx13Rz)5HI%5B zvk_OgIAahusC6GnF&DyyX9{QsC)jVrs-R8WdC&#jOj@1^A`-YtFE=2$#v%W-Mw_b! z0qKMRM@wI^GN4>|*p)bKpcWsGs8;Nyil}t_nv)257m{kZe9MXB_4kVW7KA%r6w$~N zrX4^gFKu8?bB=kWoh3egr0>G{QuyV)_h(G<6;&J@su@kty`0JBGw&XiRcTjCl5(5w zyuhJjQ3<3>Ko0Xr4!0A>gn6TzPa>Po|dPlLp{VCaO-%ieZJTM-3( z6aWIV+a72p6XIIloO%xPX(6Q7k3HfufGqkEfln&x|g}iD0b6>sc zn?gJbz@Tdy6bLCF3j!p_Qe(+NBiIP9NMJBoeT9jBqL|~dyDy84I@^3r$4W-{G&++0FSen~e14hYkNjMI1 zl|%J8zj346S>6QELC5=l19%nz?wB4arvGsV!-mYBL29DU6lf65PVHDgbkg&imQOR^ z3k*SxatRNk+(2SD=`7ZnS;F)oviar6b$84tD|^2r1cB(HEki-V2o?#?1ERzi=DYCA zDx^qD-b~h#TbrIw`EB~M)w!2bX$mHGnSGTyEfeROi-i@)*K!NJte{{49u0i70>>T| z12eD%)EyQSrQeVuZAR6{H)otcI+%l4X2~{BVnA!Cn7`~U@|SAl0s=}th>fWW(}a`O z;`5UtEDvs)N2uHCh8o5XF{`D7^g!*q>8?{KRDaG=we~Z*LQ~JL1*n!c2s8Z?p$SC? zd(|EzF@^9TO1(JdyaSQQa)1U<4_F&JEw9K&C7Uv#i!P z_2gG~c2EoC+{PSs8gl!)Sc`FzSG$wN$E!N5Isx*dyAVGV;nHFE10^Lgz|xqeu~G*&DXY_qrnh>f|#Bn2x!cV`_R4@PKeZt3jp=O;mD>}r7lUZ z45nKsXJeK@7(=_>4I5w$`oAVY&*#ekPRay~<`nK?`Q9rk^R@f=@t!lcyxJ(^5J2Bz zRD`e?s)z>XZ`s|8(L@&`mhyx1q`o!=LN1c#u4}PZvhXXYOv;H^lqS^{7#eW^`bz8( ze~>|oq%4{__7Snyr^nLiVRTJIGE~zDqu>_Ac*`IGSH+h*7 zb=0cbJLRErs`(YnKVm$r3d*k`>B?48&1RBds<)C@phy(h`w^#b+7cE<2WKy%Zfw{; zS?8Z(O17)EnHwD(6CkWhffD!?&)13!8d~g~cd5EDwC8HDs+kxli|%F8m&+;}z|JW# zy*Zb;dljxx*PV7!@hDIAUl-YK{9k>Bt&V0V1zke>=ygk@`zPZIVG~Edjh@b=>!23LGz_+-O9cY2O-E8rFUZ`E-795N zAw%o@GMpQ|Vpln0rB87gl-HK1b%Nm7OtxyAG@eGi?UcPv7S54t?HLJI0h>IyC8Uz9 z;mH`xC?xbHCQyQ9i^He(#cjc;?&SbQK1-$r9`NFCt5Kk-+{veR$1c!XO0y~XIz-&L zJGG~PsaG2A9H@k(>ngjuE3DM!KoI~AW>hh2d%wxY?zJ}T<`F+{`+jzuHErnQ-N))Q zx!R^IU(0(fZx<2uh2uJbQ87d@t*Nd)BN0sk4lUwfIu1$~$q`^{+*|00ome1)`X7A} z81Rd}SMaL(yt%FsaB#1dl3w#@Xm{XV9D+v0r>pnP_-;F~(39vz(cnY=Gru*xudbe4 ze7nzwQnBe9{2AWByq3Ejws)o^o0x5`#!!ryH!e;bya};|gREaq<{3_tJ|4NIG7mep z-ztVIe}y!eR$yg1YM*oCaLDhMGP8Xjlc~(qnb_HT9$cBRtTjfyUiTCqw}$jPIf}WA zC-$gGG!b(}x}dff+Y|h$faCLCnjpqiRKpwi$&i*-hJ$Tk)022tSXN7|wf3djPu^pF zIhm1C&fa$oi(DUNWw=~+N@M)Gi<8NDGcs-UWNdVOwy*2&rcoiXE>-DH8C%mXyzS={ zaEBbS3btWJc~-dh<(p{tNOR;XrOd`6t^`{@ka$kJG_QrjU7%`QlwAx}6FThvQbj~z z&BihjrFr--()On7;KPfVOWrC+3ataO+{mwF97 zu7{{~!%T)DO}?x!t`7V9Q|ytEYV~55@4qhDdogcm$Tho#gyKt!7RBK3FffICnlll_HGJfy_IR4nRu%w$r_d4U2J})`v%%y zLU`--&#_}hvDG^pSJvNOs#eYs-G@IXXu1>HG}ZiJ*4ef0KG}pT3P^@9L4{HXh8*T@ z1jOhrC*3J5MEKVZx<_LG%-MX^tCjL~*>^>@ej(z&gL~gJz_49m;AB#Eo2B*J>Teog z%kMertI`^GjuH4$cHxXMkwaL8YbQX`rcp~73gBvB1L}7h{s5-i#xAA`Eb!Dt?dztw z|E|N=P)^b>NnoF&DsE02H+1En6RJk$ATkaehh%n;+6($aj5|^LB$?MzyJUk2&WA4roKlv^i-E-K#{CC?|!D(ENLjxW#g=U ze61i>0!p85_;PTX!&TRQl2$C13E?sPq%n9ISs^{wmhd62fvHxN_KC7esTon%^(1&W z?VLTYY{{(x_&HwLoz6uyf`jX=mp(DtX{2CUJv>-0Q;!E$1`kL!IQ@=@(PM~_4txHd zYKigb8C-k{xen?8=n_VVTVs;Twt{STUPfB_YT|ps^tla%cugn%cZ2 zzKlu@hjbkJJIgC<#Ltexv_UXCaZ?VKKf_!!GQ?iS3-pD(f@`q%?`X#KtOl!_l?xjA z-v)YKbj2W&GujXxThzEKGrGpBZ+?Cr@##@J|2CIxrj0@%_Xm!nK^q4D<7;aKU=8)f z@D6TjS^-Iuct@=heUoRL9HBrRgz;l`cN_h*V;+Ln;SGOnH`98#ZHwBPC|G&^FS4QF z3vnRA)$uGcshX{`)e`GtY&erRE-{Xx9+ zS-UPwaKfpF^ps(64;Jp2i?x(oNHfEHY9FH1^{K@Eefyz?cum$&IE6`&ac!G}SP(6d zsrW6T+STG_n)U?#&e(9*ql&9c5%U6(=VT}$bCuJl3B)_hMwB82=OQt4PuIJi$5C!O zb26C~xQ13ffnC71Pr&O;JuU~@?s%;+6ldj71w@q^z4aqvbf%qo*goEqj_0gB)xaYm zV8m6Mn~Wt`Scr>M=XlYxcpWsMWQ%9ZxEo}}hdgWsVT4T_e~%%NlfmBPQ$TS@0|bpi zS=S)sAFg0?5lmTWgUc7o{*8=OfuR)3X3O_$AMR-?76v&}z*->}O4kVxXy)PM6{R*< zl^R#E>QSZmw3?oUS~>w|Q|SmAVyBtqS`t;HY*3G+SthC3(&golwYoMr?O?Y|HL5pN zehA^xDgOIN(yz`9y%e5`{}!bGK!@pVUi7d|L_0f94y`(OophRCx`Ht32{sni+UOJu zYDCE8b3&5ITc;8%b(W#AChE67Zq23dB7)QzpN_kwv>@3l?@%UnIu?g@-ovjXzTM>} zNEg;~6mg@n$G ziAPox^yf3diJzq^S3vgDVNasbTx(PTk&|wVHtu*&8VuVbsB}4wEJaX`2rRHVYRFSP zOo=sA)2mu)2|a?{J_Jb-+4K=qNCX3AggS5JOI zVT>3}%G7`Kwb(5i+v^u6v;H1$B&)B%Qg%fIKABV2*udcK6`b)b0J2$Nm8&X8tYLE~ zej(^tb2=~X{E^@M>-FfDi8nA79?+IYO|#Bs2;S^WKmHo!oQm_%+&WVq$#a4;YhH#X z#b!p#a|VCjuum0DRx@^LrpldDjZq~aF2^4hjEupBXw*Jt1hOKxgk_#)W;u?KxN|}t zFVOyHZ4Z+O^o71CRwPfuctzeeQ7Laf+^RAVpzsdHUVMv;>RjgOHO6Ba3KD}=b03AB4e17EUsgK+%u3|G0y-@f3M z`k8ob*7!82Wx-vj^^eAhH2oLwF0o0`vO~NznG`29`n2fRK#qUFFley!*mDxBh0LPk z8QLAMCC-wQg*e`Re4*c--eh)5fW>5R6v*+4e@2WV!!FELn|yg+OF$LK-RAuIab%FY zxipKIvfQ!S&DP(Hy}mAHet67(j_2@imr{7n4|uoc+U{f3zKm_#JKMRP1P(m39N4dKY?CBwGu4GJ=YbUU z<>)CYT2VjRet3Ue83R2J)$slFjmftszd7!4cuJ>0kH}t&zOr3^-Fn5sn|ng<{_F74;w2ruQ$-%m25-g8hHF<;{QL7f!~8zdO_-3}3g6<~v_s zsOc?HiDXi`nhUEgle7c&@D=52W3Mf;F&JS%BuglLfY}7a#+<)y;ED+{j21hO*~?|#nvs+1;qq~LJfG;d6W-rIivUfm)nj8!2R3YEw5SbS zWSjkqNR=9p3rtz5`TXpdc!=zK0=0;!Q6XHFwQtZNp zz%23T@5mj{uHa^pgQ#jhsVCu9lftp>4Cg`D8IceuMtAe+X}S8%s% z3wouL;HZ8{8m;m|ykJ?q<_Njf)wM>RSc@R9|A((oAsNcFm2j#OqC-HP6TCD8f+^mS z&X>8T?7aJj)3-~`Z0m|DU;;B>}wCurMb^m_D({CuBo zQF85;8N){GR?S-f&*6o%Q{ZR6t6L=Fn*Rqzw63T>I!DnWPhL@8w+r9M* zNZDhsCh?h5Wm46~jKZC&*{X}3C+O$&+J4>wzgfWw-?U4ODyO(JpV=@}7!gQhm6vl< zQG1}of1gL86FlPO1BFh*+fKw=PlXMXZI-&~+{$Ho3&yndACXOL$sViT<8%^p@>u$k z(&x~SbJE$xhTF~N)zi>y7iKoB?dID?Lth5;OKio2KlV1kYc5CF`t>x)-XbM}(iP=p#Nnpk4{SwRUrtl7+ljY3!Rf_-GHL}&L zcXo+1cn5n0Z6RO#MeN8wxKu9~+XJpES-sqjMAec#=>W{)rHD$K zPDR)9AgXuQim!#4gE{y4uL%l)vVk1aZ1U!Qj%iSfnlINQ#{ zk~FYr`s=1Yu69#FtjhxREpU!4&Nw^}f}{?UoUCSoqJn4uQTA0jlg5(^=Xk?qjBUr$ z_Si}E3r<47Af-u%>Fs#?_>Q;gp(oBsItgaeT3c1@JtcDethRwUv62Dlxppw2sZz?9 zuZn)2Ruvl7R?b=!>_Y;IKnnSfNXhlgP9qwf_Be1sunE7$UztRyJ3wH`LIgV%S0=#H z29#C;5H!ixb(&NM5D`nT1V@Xku@u?V1g$yc*00juL8;_DJ|sw%KRS$;TQ{W}uC!JG}&?06xc)f|L9#6f~J*wXF`S9PBU^o3o=1%}<>1X3M9VI>K~s zlj@yfwss88QQ4+l-G942-sG&^mH8R`Sa=c+dPy?XRAGQ zS3ssYP(5;|{L_k7RJJ0bNm0LC{XsDsh|o`1!*jqh5%e%QboASG&1l!&%hltb3JJwKL0|b>(H@!YH zii3R0J(4O79NS=FN&Wyk6h*0C(g~u3(L{;H4=YH!M|kG~D#lniHIB=UF95d`#F3Pa zxnmkAz#EvWfdbS_@(Tm;EM_rA0FkB^zGCD72N-Q%P4EY#kb&A=F3(3$9)V;DHBAF# z6@sY1vJCU3f@zXJkw2g*T`Nz%fOiY?J7&ymL>_ z=a!=?gkt}lgBS@|)Sm^%B|hg72(Tq}3$S3&8Z3`GPb}Fo zIuuU(DFqrJ5S$91tZ31#3GF_(0dh(qlFKvp3TH+@62u7qhpu;u5iQ!9M%%XCyKURH zZQH%uwr$(CZQHhO?AGn`a+CAr{K;BtChK{Qq-u;BRm$w>Ti06?`r5;7d4B}Dq2}0Np`Q|qCHN<4yi+P<)~Q;mC2-d?=iC%#3SENb+4)UByn#b~ zU<1{QDPDtbysO;TtcPS-fda-IAk`AXdsML}9KKR-${s(j?)UwNU>zqxotSUqwWW!4|{VVvI z@RY3wE}^aNqy6%y4VBlu$41w(6YTjO3wrt&IsTs|Od?qPQBxoQfC^9m0JQ%FTCD$r zSCW;i>=x-!zH;;oG>KaYVih5Y2RNIef5Rx)szt<$G>%AZt@elFudS&TUpDKrpVDO^ zeA^sqX4_BSbW4{lHTbIyo5$>FgT?NFPX-d5Odn>aP0|DN`~5i%^%v!xc^d{2ACk7~VacnO-UwlZe^T*#S-uXP4ja zHiwVRy(~bKs1frA8Z{@nSx%|B?%>5$>1q36b@#vzmjH?7%+n3Qx0rGwfK!Zn|Mr;b zpr#RxqnHITilkx56*6X?bMmDM8!F@pXKSar6G{<7PU1T6N~1Scv5O{+Fktv0$evA| z4URQb1rZYTaW1cST!M)fsK6geZ7cs#A4i5kN-?wxq|emmPy(vx;R+-TBo_dJF03di z1#xzfC0|*iwDa6dOiWpp#E-78~6Xc@L z3tj>(@_gKeW!}iQ5py;E?zy!n|EG69LaZk*EfJHsdSBW(c1w~h44fSe zu(dBNEA*4*<#xGwa?CvO9qZFcK8%Ei>37{braz1*X#O$JH?@TB+3*eY-|NA<)1}bs zzc&*E?EiH$|FdJ^t!!hrL67jAqqkqY0K<$S%C%JQtRkj`8ZD?&_Rqg)j9Auk+CBXm zMt<|_WgX6fBHa?wFc+Wib?bem#YT%i+8TG5I39?hfl&L;cu@UMHQb~%3@ls_1^9!% z-?=>yqioPYUqT6_K6|(W4x)0Or%y&TM>L)>{L>`GS<~jWY3cSm44RdYV|P3taeqB9 z2*KWyiw8$fdR{#z7?X;bPhuNXQi&sg)iK=T zdqHHq({MV+L=5mP9(f7j5N{G|-qEP((>;u-y(Gv&V2>1&cNl0uNu_HQlWw!L>*;Qp z-O27kBtkD)bYcKzKsI!5dndOw5jzZ|Vy*YkjAM^XlQsi?uoTA<1$af%f>leJRH-4M zUJ0Oz1wTc+<|W0SrcfRMujzSia=6eW#A~rLejcNNC_@E|JW5t|bt5d~UwaCFU*&X) z3Jy3R)R2TZf!vO|vWmEUaUl)?wbsFNeRhoOK(cObY&p5mK_HVF`9ZvCMfkBr@gF?~B&%25snSugp9EnNuMH7wXnYegH70Y>X z+^@xi0mS1*U#LOx0|%8|{T-Cw+%V@C>K3|vxc?HoM?VE~7lrXGXn7RYx28BlkFWBS zW;#YvOy68(DQzR}RmIP-WtMqV2UO<~WWD>812{~RnOoEvT^^xrLrSNi<@7}616U_4}qC%9P#e;8lx*dUMb$%Ni z?D_Q`2CZLcU(eQG!2ewg#M%N#v5)`&YvlirVzB+sn*6fnwk>u$YR{^Yy^TJ}@Pn{R z=5`Ijs6CF2p*3}9OMU*Q_OG2~O1> ziN-B5Qs}GZ@9sVy4++YDB{^9BCv+u_Cz1y2S22GL#MLTZ@gJ&7ICY6|)7`S3(0ka9 zmvYxh&p{sQKSa{XEM>a^ywPo0Xk9TFz*my`k&UgMy9)3+TEd6O@zm6UEy< z?*#3Z;%Dca8e!{k&IL`*=A!I68a3wB3># zEg2q|d5lb4Y%{C$`FZ4PaT2TjU>tlHRND{R?cun1mh0g=C-oNF{)pw<_5CC%I_u`} z&bhF;z5cmamFszbVcQM<>h*M9xYfoh(S zpvDTt&ATI=7Wz6`qqV{t5IU^r{@AuJ4UaOg&0xwBvY^eOBFN@8koo{S?$)QjSUOaS zm`7Qi9}%`Mk99!_i&tr$G2I9l151&?P~09Z*clG^FV-o=E&Rx8h-On+Yddh1VjM3 z{$3^Pw!O8c;Kw;wgxm&|yIUm;z10d0>Oiv~^`0iDiJg^|>+Jq!OzmG@jspLfU+2E^3fhGbHu$~%*tX&(v{gVURExdz>R&Mrv0b@JY0`=pyuyaO&7KJ88vhrm0(-&+%8?hp zETM1CO`McI&z{gtTw89$HWtwd_(SRMj)b8inmX*@JAet0<_U{-Z**(EOlAjsooSE% zWn9;U&97>2(jzp^#=!z(#02=81(N#Xty(9}ZxWC4L2=Bs zSi3ll0v%B1EQ4}!bHj;R8>C7WE)n4TT5hRUx|L*>raVzjcPVP+5+O`cDRDe$BA7A| zdqW@Q5iZ1oMu}Qi)Qu|~`3Cu6|C-AK*HfY7~ zeyG~>$Am3Pq{zRnJdCC4hhZQJd1JZYaN7XN1BfbZNW+YjQYtInZcr4jSN?ekhi?{l zib)8Sek2b|me%1(GN`)~z+z2_uL9{2`ysP#{rZC=Y?(e9`s!T4Tzd4ui+Hs29RPDV zmi$7sKHV*5_(0}V%2e`n(H53QZzT8Dj?WDEIbXN@8?63!ctyY3)-K(6X7d#R@4$(> zd;)b(4hih6K1N$VwmxIxS%8Ltck!-`eC!Jqq1$kj*{H-P0Ij&3(+1d2reoc`!}RdG zHe*$xdj|H*h{|p_YXt&_y5_&XF{J8v`W=K}p@*1>hW?Hsj~CRX#U9?vcnN0UMYyWg zd1ZJd=*(29rBwPXz@PDc$J*3K$cU$weOq(dN`!^?`Agk`<@L*+lqSte3SLZnL{v>% z*rZ^!s=rxs-3DYNt>TjmW>1{!Lj5-$-O)x0Fp|Q-jYP|tQ;6Jef4gB#A>jaGkViP2 zRc{_srQ(?LLPJgAw}5H@gNKC>c~V@rxjZL>r9gPv9HVwj?*4Ei{gVzuAh6{u2u6J{ zZL3_TV8NdL1e12HoU&=kRLPM(S3!wrMvlMYFFQ0gF8dPSh+!IF#wHk#nBp$}%&5a2 zLH9GeNh-7QcfCdV>U0(Mur!O%IF?x|oY13ug8V7XRWgtUv+V*usiIXp)la)@s|uTS zcWHjlPdw05X$h_A+b?(Q4=Ct9^D>z@65k#TQ`q~r#e`{WhBD>_LWcsKpC=*?qtv@f zmLWjOyB;|CX2Ylet!zt2RdrYj&a+&YLuxzSL5J>Tz4@kT7GGH+7gON0sS@N`{4S5C z?_A`+I1PDxc;76JzfhY9n28HZ(o(9VRU|7zCZ8=5DHN_%=f0H_Vu53!N?ktpT(6K! z5Zj2Np95Z%=F7g>(568BryxaV)aG-V+OYPNU>#X&sJ2RZoFkn|C$hAM!ln)4 z#7Vf3853HXAaA1Zm?6R3?_yVGbQ-zcHMFpXYMao|p`P=Anw&6AFt3jb6)#^7%GBc4 zlM;f%M6dNMM&oO%*o$=mO3?+qw5nMr5q8t;U|OqZq&ixG&|jB4pw+V%{d>o`BjF}A z^Ewe;)-=e`6}*ut1OBe%x?;42qW`q3K=HnbxeD;AJh2vkdp&r(&oVFpW zy6#H!i7!+JB^qj#GGd9Xy>0n1i4*Nkz<#Xg+UH}Z2L6V!#{JLv*1^^Er)Z&I<^?5p zd;wG)$5l8KRNX;m)eo(`=qRk!!3lcRtZGGAHS-Qhzyeec}QLNe- zs_ygst4*AI95W@Rt-<8g0WWr|+aA%Fz$ZEHh1O>7c_g%a`-|OwLpWg12FB4}K6CKP zwEz4srtM&0YHDHRXyjmF{~s_Htn?pdgYo&xY)YtP|1iKx{8br2Ed`!AZl!8MWdtMNITYML5icOhYuphyDZpIWLgkIKACA<28=xy{L5`0W~t9 zAUONBX^sc=IhpIQkc?b|>dCedmEjxVoUM$4ExOJ9Ih$Nq1tz=`;w6M_bgBD|y5=?8 zX+D7KOH*BDt;BJdrQ4j}Flr_+%6oA^o(qz~p!Tp+*L-QO^>X}E9jb&={|fLyLOV*J zDh`wvA(kqVtmX~nW}-7nA34-S#R*W7MJ0fbqa%-(sn-dolusSo61q*bqwXJ&Jh)(a zL=K@(| zK>v3Ubchckk^J(QsNV{}{a;1kZg2PhSHQK#rQ-oR($9?EKBQ-)48&1`b7%9Ev0@AZ zr02=smh^54X>~xLuKU1-A>wlK*H%5LkuGbqg!J|5uR3Lat(1;sy00Z9F)=YvNZ!fN z_E=!a3cQ|C#PWqu6LHe`&Z2@3-}c;qGd7S^W@w^uFj=Y-6rElhy)cE|@Yc`5o2cZW zu4R-+QE^}er#^}dB%ME$G+XZjfpjDZu?1Rl1%xsKe)1qI;^b+Ph&i1>4$n8*9rnN$ zg5n+pRu`VTUL~Lq$>bSikIQ8H4+IB{7YQkG3+w0Sv1DeYW=zyF*V+z)Y_lwRp}kaGyxg zKCtR!zLNmld~erqB4Hu(SJnQY8U*LTJb8@hxYKCSg?46g+#>06!sG@e!(KDGkYa;; z`82mM1d2?xosFucU{kTn;k)H?lIU)G2dN6|@&JfF3Vq#;q=YlkoJa`L0l>{;@P!hu zG5IJB5xms8JSUCX<&u43-EuzVL33T?r6DpF&q135%ouz6{${rrxjZsA&!7?|?&E}+ z3bSH$E<{=0k3-UTDFuuLMI9{bojbX|*Wn$?)M%R(?{;oiuyKV6+B8&*U9LAV`)d2s z3#_?EGKPL)zgFql5h9-aN1EY`M}#D+7HcIzkWEQVr!Z*PlUMsIr@5-jml` z8^ROLZ#P<)ADALNoJSapA%l{loDfL_q0}s; zFb1h~JM709KnkH^7J(Arif#CGJKgb-Qxx??B7h@=_76=0WK7RR?j8@1clV=YEt{YF z5(;2!vhp~1jhfL)gH5rGr&N*ze>|Pa2KTlDBbDQ|>;8A}ol(jZ4lL$KU(j;(JzF+Ubixjbi+}a9 zk|{~{d5&mcM5$Z^W|EPyW2HPUT-vdxGV=KjzUP9R?!sY2){2av6C8Z!gGuKe%o=K+ zXZESGveq#CJ@H7B!*NZmEC=klG;&M73%s;?B`a5$({upfi>>_I@JV8u+FlU6MxAZ;H<=|0-svDunWz565ds!1sgA3`uA~&i+&vGp5n_#1( z;Yz3NH-M|iIQg7e{wl(ii{fMV%kBjC{Vmvk+5!UA?8k$NhG>fkaSKKTdNA7<&+j&5?XC zFQ>^ZOh@lK$f(XQ7C|ZV>1pTW0WT*iQAD&VFJz=JzWuFT7$W@nCguDXd#j*Syee@x zC>R_RY0Q4S9C^`{ItYw>B?e6Dbi544h!MP&v%pkDWdrQcj<(!FTs6pv4>qe(;$!uL z-Xx=I`}6xZsr@(CIRaW&vE!!9%ZGPR>&4r>)3oh7Sgm-n?fYBn6GJ;FE3&k9$M3wY z8kzU049;4g#_lOo#84;cb&quHn>6}a+);||C)K!H>T*hTMFJ#I0BWzK_|&ez4isVJ z8>Aat-*IwWSlbJgP?$;SE5xgvS0ES$DYZI_bfW17QP}SC0a^Nnt5|tU+l=m&vp`Et zYhx#tWeiNqQ_I;84yuDZ_D=0~Jm)~C4Xb*t#69lB&kU%#Ld*_J-R2DG%Zg7fOJ{30 z-ET%k8Kp0XqIo-4YBv|_?nHwM)5=u+oE`IZp5{XAzXsF3)GN&Ysza(_VG}E)uUZ(^ zqn#y)Fwx)ev8xtXGSfR&Sg_H!X*#VgwA|6!1Uo}k|HNzH*Q7mLIZh`cwbx6MuebVT z|Jc8pR25$p%3y<2K|0{QY?=)eS-MnOJkw3nN{P!3$&@@#7qJm=08Tp<{Uenc83E?f zF@0vvXE#3O+xHeE9a?|c*3dNEWfbXsM-t-8jZ@=h6P>P^KYOu3YlV3(go_kmcB8#B zgU-sGvnckdXEicRSkFX%sga(4i#qAP+<`_3RVc)OIzp5Kj|WAcjo3ktyUL{t+}~z2 zE!F0tHhzFZ?Yt_GpVId_RsVcB`KtTuY3kWZalDa&BY95Q zqVfuauA5+NGHN?fM$`p{QpE>j_kEL8Z06)Ck;M4NIM@12G|V!6}tp zRnLShKl9mxpwA;mW_Hf^lMFykMcX;|HfCR2`pEhzZ#^UUjY)v|HfC2{~;IE zm==zi;xC;(B4QIDai9)e&8dd#hd4V$tfR_T+v}Qr7y?oSQ!I1rFClNOx1QU7{%$N> zRY=aw-CnppdrsT4;~=~ZzgGH%P%nFVh$Si;z5bnen-W^J>nr7TP;ZDratOa0Y8_ND ztt}({-K9X;CEKx_Ha(KzAQlS^^;L|8i~SZydkw0lcrV_Q!DtSIl%OCn`)Tv~Ilc@g z9ec|W8cdoQyNql#{5}k2yXDpEI;_%}u%4A`?nhW*R3}xKQ-{3|V3)*rmsER-ndBze zEf|s^`_oSu-hd&hd-j$th5bg|k#7+GLke|w9rT&KL-ZZw)dI{(fKGisupIQ65d1sm zIT6_p!q9%wTg5B1u;v1IVZ&M&awI*Hl19!H`7?Yl99vLnp}y9_W9#X(8D{cYIxtjEQccuKwv|#6spE6`9>$6Z){p z3n)u39)8WvY`T;U8Frbp4%vBP^r`iFt#i1pS@*V z&)kowC6x$E14$B(60HrCAeHKJ!gNrK4P(;?zzZSpkq4X)6yDq5Ju54Kwgls9AP?1} z=fNvt&%h-5N^1&4mv^Up;73s6YKh7sAlo#5^ICK=bS)#SKYBIs?)B5zJQvWwLmb zT+3qW+8%@!A4EYM=Z3lxf+r;upYuq04~Pcse(evSc1zTK_b^ZMxXy`&NJ??M9A=GS zQp&6I6KoKHMT@qF>!UCu+M7Bt(lt$Rn@q$MAdDAXQGkdEpfemHi~YfH0e*h0d&QU_ zm!h7-M4ayX*Cy6MXzUrT+2-s8m5hah)HWGA$->whEsI0R5;Wy46lJWF*4J$MKf6h4 zu+R%Vnq?fShJ308*2Bp%tEzrZSVwvHuUE+V8zv*j^RWvUvOq^ zj0^Xl{MdY{teaeCLoNilP5#gI<$BT63|mxt^4!8Yb16QcE%lUfw*hY3Axt@~f`PXc zHe*&_7aKTwQYQTc-7VSJCmd{h+;P7?w+}(9nk6uo%EqYL+&pFPZp}UAkoDQ}An(*L z?KJwSPu&Gg){XRV-{^Rq_6Riw=ACtCgi^K)(gVB?N(8-CdehQv(=R#_L};pSp*&!|d?|U!v!-6f+C83G;>HJcZSkGUPx3_25!{+H$mvYg}bs?~>T+)74SFz)B2K zhR5i96d8^+b~(9*kfRpY3fgNkjFh_F-kiA|E~rO`qPbxgSXA9wKPxO>>VRXy zaOH$>@}vc_%Hr>+E~-S0OfNfC+WA z8Ca6sqmkFGxA~`7(gpcmtc=5b*3F*ihM5Wkp&bbK^kvMO4LfxbTT zyFxOVsAx~U8Hz?Uc8U_}d4LcFwMn&vIUmq0I9jszd+yxXN3%zqkao`5XyCWl(|456 z_6%$=?v)_KZg-fQTwWxn>m93D^WE&TpW6IJC1s-7b$0R&^B7b{I+STV8qMy4%MrXT z(dtiunc?@K2I_sH`J7f5WieaUv+4Kfs?Ee;9yl})^q6VN*FW>!DbdwuR=#rRlT>c7 zk(YnQj8eD+Zh|>rA50Q{lv~yQc5G<6XP)W8-q?O!$6Q~3Z(lD?+e`xuD=7a(YW@&V z4%`XL@_kz^b!A#Zvm-aKq8ny<@&IMT`U?yRJ^_8JU<{PcJd|Q?evr4n?AOGi@qq$q z)hJp}RouUeyDf+B$=%$Ks^1uc6pc>`#Zx3->pv1QtkJ_m6iMJu;}cerhRFSQrN`Zx z#wvsHYOAN>)EpQ4jEa6gMoXfQ^%k(|D(X2&72;V4XF?EPk{Vh|VKPDKf{x6Xo3;lF zWf$u%lnLvHRzExz=yBu~rhTOAk6%p$`=46fcl|yR%2@U(iDRt+65^DDCJf)rSv<|! zs7;eR{)4FXi2?R-oQ#qs8iYB)Iv>e-Xz!(v1VxfCTq-^nhm z&Crg(y0AaR8EEXtpA!HgZCgc}8YkUi-fC`ICc|>HW`r2zfj|F8<&n-!iu32UAp}hE zzX#jS*37?fp`al-( z9m4jAf|J34e`gBdeSwRYJE;lwUd$nc-dqHb3rCx8fgRQHIB_~@oj<=oY|fZ26s=tM z2p5&2Jrb${79V)zEC|YD!?)Yf`*NrjF{EDn!%AZF^zx6|@>VA+M^87`k!NIRv_zyi z0f;qus{PW9_d&dz`98~ih98CPN3g`WU#7f}$uOQ`m9GYarZfk+ttsXP*A9klk+n)+ z+5yP>l&M|~NG_YBvbqqv!S9l9Ea84-zWLUFkFx{jgow1$?BCF!D(?AKlN2|ZvlA#e z|7J3@OwZvsO6djgMwzMe+UO?mfCR?5vZwE&=)}?N@rE>JKoT$Wie= zg4T+C7-yLt82ICc18(x}?&p?Ns$Jvo;^k%0SaqcDv2Z@K_tI&HsXS#e475^&s*2@7 zv1BH+@qmT~Yun-n`HET4(#2j&@FA9O`V1?aq@2Vlpfo z0HLW_x|;svP?EeYHw>73Bs3GvfOz4*ij*dhoZ&aKp4)|&Zy%6K({#PQ>l7vySR3h= zbQdOj&81=VzFns)0=(GqG=|ouKvj-~P;k`C<4Y`oVMiQ2i$WhPr6N8zJaYkkk%*))q1Z2(ehm(j+AmnykXIy;YOoV$W2;xbHn|`wU zA>rxr=&wGaWY{yn)WVp$RSs(v7lDRY>>u~Vw^#c*5A)wh=!)c<+!)%-z$RCJn>U0c z)OaPp>|{<1FW5RvCJ2S1xZq2~i2`*2NAQvnTwEk*iY7nm zym5#4qfv-7kf|-5B5H)TXG6RT824n$HJYcMwa zU8ov$2fCd^h-&7t!Vf_Nhzuv-BHABC2L0Lmq?nia)*IXgPB7NhjN*ck>y3@pNa^u= z(~h(VU%oTJe}Sp6AnxLs!c|-QOxp3OUH8xs|G@;&4)yjzywiZdhnx@-qxH%o?acn+ zfzp5p#d8aGBf7gZ+lZDz^?s58_)fGpLf+5J_R*D3@*seBs~5@GK?bM zMP?lf9c28+q-egdW!cMd! zA)d%$OFsB0Di^z)yn&-uJdq(o0Tr`_(FkTLHS*RrY9%%MB}7iGty5qU z9k4Q|_s$Mz>|`{KJo%R|2%qs^Iexww|Hi$>PiLYR#u^ym@+FRJ6JDtA!T_Qxwy+=I zGqQvuxVRQGuf82@(o3fVue)b$;ioJbnmyF;=JYzrl3O6)TJNiOh$Tmw03T?a6S9eB zlS!E8`s_@PWQX=D;@bP9u(v%C-6l;{lwxXx7Lb8txTJ3U4OQ@PFeFgo?a1!l#;h}) z>PvpwQ0s49MF3~*MHX*um5(E@xmp%5oiYOphfh4ELq*|Jrjm^r=|jy7Bc?Ar537we zEkgMm>G0JM#CaTC=5;+=e1L+?ulY@ZUUzzN2~Zz_ljQwDR$`h3_w|Hu#~Oa^!_g-0 zJ=CLaGH7nn5RO?1ou)eJi-Qyi?`Z5qwrE$2^*>^9-XkqN*6^UNyIIJjZ)DBO)~wjm zs)f%T4z8`a)$JVS&3ZU3dUzJKdS7l&4|RkWcZoF@hiA3*J!NcN4b!=WRE~G6g`tvc z{}!nx{w24JpddH#I8%pi_OPW(7i^{SbZ7XrBHY8$)Q;;yh1b^NDL%B5Gk7t=h;=e{ zmErS_0L4Yta3=+ck-E_lYBDSxq<%8bcK71DaPI4I?n{tyhx{Dg9T6{CW!KF^K3kG) zh5mS}!lUp%0564P-zkC9dUw~t_wX)lE_*i@e0{It4;&`S(Pt$cNej#o$(lJ@-chQk z{J8})9d3Ge9;mSv(05gzaVIfRn`#Awl2^N^%g>UY5&`$as4w6)2m!=NL!08i9IxRI zmKr_z>K^<%%BIU7(FwRJauD?W9nOOst6y_WcNfe;^C}5oXvAJdIgh3jmuw9Ia;kLl zxmA;I{;pMhkWpgN%9%2_0@|~pk!G{3V<`5~Z5(c_MYqqa-x@4`$Mi~nWxXD4r@Ss? zd6-VMP`#ZkPyzvGkLwB9>%kDe4P7E&X-5Ak9fmDKVci_~o4+Y&0C#C+=~TjIunWbu z(u_pcI1hZO958WG7^!o3GEhQFT+>Fc7q;px2z!M@$)u*>lt%N^5FpmWa}Oq8FS|Cw zktr8 z+3QtZx2l)dY)0=jqR;OsOBo`x$ojP}Up$ntq|kiPc16P{gG;ywu)rw3yi!z7$EoU9 z%(0p)y5$4RZwGzfMHApyk2pQd1pMg2MXj`3$&D!$r{G7@G*-_{zWsahQc)+?Mz9)E z`0PI#iz6$5j*)^m`1x)J-PX=R?`)N7#^z;x8D5M-h#5l7MPnV0bxgwkzGlJJ7m+PM z%0K^S$Tp3CHhcgN0MIWE0D$>Fnn2FB7Djfq|AAH9AD-B)@x7O|YZDZyN#y_j%Jgt8 zH&9I~7Sj2g36u0tsj2qs#0^>Kfw_nIf|#_MvDpy$5HhI4fc5G7xyNs8 zZEfA{CC`q8OG7V49yc$QsFhGoDO&y;e&5JzQ46|`X1qp4KF1Zpmn({Y<#B#BjE|9G zNGZ82XO5Z_FE*-$X%zS1>$+hG-=?25(~Kq?u5pzQ+({IR?OAt!I2&Kxu@^0py8`JJ zj3~Oqr0D*-o$c}V+_+^(R}W@naai@7_Z`2kje&>TeS)`eiKh;Vkg+pDM(fH?;9d+a zJsuemHye*$_H21qjyYzjuFL7U8gLrnq;gcBXI?}^DfSm}Hi((n`lD)`%2*wqJ$_jL z4K4lai_|?`j3xq>)SJjuwDi0a26FZaE%OAVxtz&}d2mN&f`wL0j&e0VG8`Os7zu4R zi~TnCH=|(p`SUiu@e%79P;Hitm32TJ$f6-Y_x|00!#87qPB{)C}>XAcw#mg&s*^ zmq|N_+371~sL&j!O1An}?@83}C5K5_7|>1(h+IIC`hm9_*4>2A<2#nI|BEgn=La*0 zaf_&2A!ed)kbF#t5Ig{>P{kRT4;4ha8)82jqDMNIKR6VRWI*9S0Gp^!sUP~T+(ydx z8;3j`>_@n>@Fa1rRn3s8M5E?%t>Mn?Lpy~XWc6LrY5k~h#$Y-WxvKTc!A z0aM`{(8mH_bb<5unwl1l_7h>TgWY<~g8?KD{Q*}zI#cK~QE!H#LFhFO+1Q-+=TV`T zOPg?B&Y*#KDQB)DVRXO=*{_-qRmnhAiKod8(c+9gw1&Wr5+khH8$+S28k*aqj_gDl z7jX->3Z3q)QYY!SZOYVFn~>x#4#v$i0RN0@vVBt_baaxtKkBfAeVICCBN9H;z}r%n8Cpfrc2B^&$a z{Ipgl@Q2VOGy)jqK-7eH+n-wV?vMH=3|c;HRzph0(82OCyhuLz||UC0^Ki?Y0Rnd0Ge`pLTgM4Xkv{aepVUGKb!p` zh^U;p%waAFm06S2CJNC^RNWdh7OJCd=D5axMUnsPfp_(`i(%=kQxqsJny4GMWXx^_ zJ*@N6`~QVWq<&n(%hi~{0hksc(UqEEVlu`R2PrTpGT3DqOm#5L9iapS9i`XqP&n7s zVy3=Jqc4bZKLm{PDw+4gN6M47)zY(M*j~+##A3xI$9;$>Q-wp<_hzY`+Ea|M-?3$+ zOKz|F8;jgFbTL7d%z7sQ7GbX&Ze~QUzly*eyp3v(xANvKfhVLiO{O@DJx-u5c1gWR z(ir99g$jAbxC_U!mH3>KSPn2{kk8T8P?$&&2cdmSSv*+D*T{~e`+TheRU(1X# zYYweU1X9`dktjv_6SeY1{k~@ylgf=+AqE2u8;=u0qd%I7Q>eCqqLyq#qmz?Zm4l4% zbRdH8@qy(H`ZH}UL)=Lq%ebs~UGqiXP8ri!Cr4Y;B*R#3=JJ3g&F1k-TGHN4+q~J* zo2Yd9-ZSxBcqpWQKofV5N1{4wgK#qaQayx0D5I9O)5dNxVA#C*G5k;JDZo*(Xw~Jj zvGo@hS}@Mlmhtt8KxKD?97u-(b0snbUAr`qBB)=fgXU9sM~duUPC(RlJ<*7(^2Doz zx<~3$OWxBh^|h^KuETREW#ne zQ|3b&f~R#1%))dKV%4qFC{CThC#A}I=YG)=TBW&FT4^FT$&~H8zKDo;YeB*xw&HmF zER;R*yNl{-Gb|k=tl$8I$>KC6i3Q@DJFA=4?9kiR|-0@5L##R)tt1D_muhN15W4HYw zjxLF5biOLqeWa%-)Oo>KOYZ;!KDF0`Ymq46&}e9d66aP%y0G+PPJAF7bQQY$(C63hdcRl$~;?-85=t7aMtB1vHVA4Je(pUp!rR;3d|d3|#k* zp1UPv&!8wm2?~Cl)|`ZP-i>DQ0W@=6-V++C#MT9A$n&E)Dya_lpZ!4U*|QMr_kNaP zuIxXL4Alt$bpMl(qERQ5o#$iC%O@RQFw#*n5oG$vd9DVu+h|ejF`EyZG3FSEhebXW zx&1Gg?>*e792KtFitg_23-nKjmVj!?HMo|_Y+Uo9(7T$P85d zjpoi)XuG0!l%0OoK7r%eMTSb`4Qsw#+%EWl=|@TkkL+rh-kJHTJ==e1tF3L{rqP>1 zm;6C_Q#cg<)upp7uXrs!ub*=KIiIl)t$xd^T_KCBLBzlAoFpCKX(DR{_*g8MsKE-gK~Vlai{fRqu+^G zjJYL1H`8C%)1G`5U6XV6R5_*($h8RSB3&W$y5F_-$IHbbJu33^^L>FhX1!j$@hO-( z#_HL|H@sD+_L{G*Wh;z&8@dh~+Sgg=i*M&@92eJ+*R9vRr(47S^KH8m*AU9BT0=gO zsAJI~MQy*NE$o1Bqy176UG*(cg{M+n=Z9!vYscg=Z=;#ks5#Q~D0^TZ|8Q?srs3u7 zuw2tal)_f{hNgKLbHm8Y=Vj%JnyRUK=e~cFtw#8|W9uqW*W9CT1;0`G$yk{;>oi2I zp()mLcgxJ;wz26|W5Y!8_wzx0eLB9d(_HJU+AGic!epjzH`(dau6$XWpx>ek{soUHQy=;$fHGq&=bHvk2yxkDBWI&pl{EJO-NXwPP;C)ead@ z>e#0vDUh$$dGAM*-m4z@;cKPoKkYJCnK6!kJ~_f$9b22|5^Y~6j5d`vujkD-b_Crm ztcfGDm(o}y%PB}^R0gn_5}uOfO`M5n%0GtZO?p&%XkERtY1?K~H%z9vJVvZPkViU7 zrZ2B$7|d)pPnZPkKD*+v+mKC2B(M>qEnW_#I2*Okk0UREJF0O7+-pN?*+faKav0l+)@Ll`)}OMnXxQuty0pUk#9da_yH^nl=03CJpsd zgZhFiZMK}h4{06rZ~)a=3P1#k&Nd@aY8{8rE(u6Tk+A%O4}hH$H)OI4VuSi%`16=& z2*^p#xPdXT&SH@D%?tSAZEbPrNlzWQ-BB9r2HW;8$?`-C-x}cE?|~Y!p;(BS~=HteWKvNPS{P}fD$|q}j>Wlg(FG54rC)NK{1e3rcIC?CN?5mo+tnU%d~ZvpyaAB`Xn8TCL8n)fm?gv28qjh)aGp6 zvuj~>;)Tq0d%II8SB#Bwx|c}xm89mMYvw0mGx!c>1J1g67kn`#nromAxSd(4LF<@7 zGFB)6uO}v>tAA1#n!fw^zuqqCVm3sz1*==4<2xi#M9wTXj2*Ok|LKq~_6{%5Eq>bq#TdW2WvmaKKA@7j!387=vbGLP zDs-pedJ;9VSK(PZQ{fR{glZ_35&D30DcUW)UuXdM9`q8PqxHiib)9SIXr1;rr9yY( ztoKlgBwrvae~NM=Ihc@pTe)$W9Bt^w1E3!u>qqnT(+MGE4zot~w2BnQ8l&~^6{E-8 zK@-r)$b9j;*xm0Frr3mO{I^mjkwKl7$&mcIzMWoQP7c>vt&z|KDGh{YM$lcp|82h? zhu#-!@BxWv5Pro3UO25$9tt7?X3q%TE$7%4W%X!Ni4%qK3c^?wO677pAdd484&M?~ zhu!E`|0^4loP8db0SlmvN#Yw2MSdLz?+D<99I>Hi}{Zl5Yo(nLSQjxEDU=hR2WZ z%#6rFBaA1NDuSHbyPz-m6+sUW!DY$-Nk!_Uo<2l4A!P`SqtlX(UOWBdL9B8LlCpk@ zutf(LQ1GT;u*~u=KInyYXRO_~cLK4f~#} zzBeoBVPz_)Bu*G-Bm4;7Afl+=7XH>kBB+4M8wAbZR>K5QGQ=i_HS0nj{ku$C6_KTY%&m8QB2u5paBa=wZ@#UX62~V8Li&$rUX%(C<6S15_7>qkrN$!!YV`C#W zaa1qyVuU8izOWWO-SjYZ2v6l@qMmUy)$0wKBX8N?SU<3F-UW1b5S@(;az^$WBFVK4 zqSwj|+CIDwzFQSFurAu#! zWHA^UY+)W%6CHBV;-zJ3^qyTFMgCSpMHSw5JXJKx+s0Mcxtq_I$5KE*g@(|x^*IAS z>VXoKy1a{%;1C`#W4Of}`8mCZZwu#A15P2JS+K-hF#mk#J`O!ft?3y4vu=^2vm5ORgjkIS0!ux(4>NJ6PZ(s0Fk4fI>MEy4UC!a5=BN7~hc3n86D=jz@U2Go1%XFqaW3 z)V$Qijp*`%np+P(af8hf1h#jKW`?q$*1mU*FsH(7eoXB0?eH<72JNw%8l&4VV@h~~ z=**s%MyN#5y+G`8D(`G+ z9<3gF$3-sH`}a#$%3L?xZik5ED4X4!`x|Wa=SR%E!UfkqXVk!V=$~R@inx(Zq&f zQNBlnF(%H-22uNYiod}3^Xa;@)(sQ{8}ie0qf*F5K+PBEUpw>y@XhWR?ABSR9g5Ot zJj0-;OFLM$TP-l*bB*P^5tWSkjQJa+WCld1xzM+56T`wq!?06upuh)5$y#X65e-=e z#jqAIrqc^`UV;q!4cZnmCn}v(Nc4=%!fNIq>$P8;@PyUkYl&=H@6oL17}dG$o64|j zqzKhZM?KQ$JswJ`@$K)Djm1drKBguG}pFG}XxU7c6-`G2WpYKKBf@x@TVz z12(c=vUxFBwVK{a?We4j$K9t79b%`3w&B+JV`{_CDj83EXBsi-GNW4ZHCeB71`pG* z5&3RgvbAo>pk^ z^V%9dBXl{sm184H1*Tb9!_a0J*MdHgPJlVgI#U(Zl9&*03s%aFLv^Q8DN!tn8tJa! zng^=wM{mlp+NfxWeIGjAb=kA_niUO)}oF=aq6q)P4 zz={6WR9AO7f76<46ttKoEG7Q@ON;B`Pc;Mx&Ybp{sFrY2FiW`n)`omo-GA`DU0>q; zzrH+Kuh9a(uTWw6XrFTVmRj^&)8?ts@3RLOz?DD>Tmk;&W8TcEunX& zE*mXEEeYSZt2Vr^Pi+RjWV&`$(k8Xl(QAS}4Xjjk8*|b3Jok#zKSB|5^$@>UvLNEe zQWNd#EA18EP58L5eB>dc*MezCqI<;u?hU^6Ba<^cP4Mu(8|~<4@%xs^+B-uv=KVwJ z@9S^V->t|En7SzDNY7O>-mU1e7AcnC6)GYQYL7s^gTd995JsnQnQV;UC2a~c8DS^J zhtk`nnVz{&YIU^SXuxkKj?;Em!6+raLK!Q3YZZ4vs6KC(lQnMh5iurl%QD*xxIs8- z)sxaD>@R}?DNFZpE8YEDu{N^GdF~cmeg>{LzY^Hv47~VDlq`PYj4<7b|3z4iuCr4D zYi_$P+S=6$d2Qp@+KB14nq!WAM8RKK#d2l?|0>!p;I{Zu@etfwt?j?bxL3~p#{MB&Ofxmars9j#_ zi$2|zc>*x}EV*Ga+wTRxmrw0ZEbPo6bqgin_`N^VA)VA!7*pYnK)RAO>r2!(4 zz@36CrJ58>3Hg#X8VEt5q)^RbTZ4Zfk}B~`mh5j8IJO#->9m>|R)R&HYRx%|uwoj* z_9yfJOwz@K$Xkr`zv+VxKjMcZXrPwRL>#e?kwu)u$R;Tm4AI9cv*y8 zaYqx@tu1UN2x&6C%{J8p%VD$zjfs3V~ zNHk%+AsWEM=9RUVK*>t6DnyEbHY=ccFD>5N(?uk>E?^D^A-BYYE;IzHq4U6QGbm!5 zRia-Aq1!ms%X#d7Oe*sv=Redh<3);yaylR|MjH-Kcx&Q~F-ecmy7zj#WAu82T)H^E z`D4l5?0w#Lkao6k454NRY{iN%2cg`*F7nUA!JGM>;TGuqpPB>$;CC=z{ z#u#ZN-`<|!_ISv()bwDTi;R4grBh_Igh@0}-z;=ATphGIpwx@Ic$fk;4ccGr6b_IA z8yqE(*%I}))y2~xE4K_>GinDq2{IW(53DyLUQb-nm}u1mUrE+2HpS=sg655G#!B+Ca4&TAP1i*H_Qeim!0xkJ0U86E89U6o6{QvaA#c5lvX zmBPd&{BvcKCjp{ruajCq#`;Z}s7bG>N_uNS{mB`VqqbupK`un?!!dZ9Ea@+8EW)Rw z>*5H4^sVTzeqx0pzvxxvVW$Zol5NgQP{Z|I@wt|Tl~CAE1I@Hp47-+H~N9x zmXZ_9CC+s@6+4gQXI<9zvuRzCOP4aTVxWl$A(>#9K+V;n9ecYYS)x+;DrGa_lg2ot z=v{O4ZWvwj7Z(@MLGmq`X(3b@sK<#=8N3Dqa3o)>d(YnFRvt}-O zu+^02fIrMfD^l~X!n%dl0-_j23j}iv;u!vdOK3{4L4IzO3B>KVgpyERug@I*?-x(D zMjNQCBr-|Tk5AVeZeOSGm-hNQe%3(8qJ~=I$H%D6+!@e%H|L=invMnY*31;~5U)Ur z05n%QhyX)+58VmiN;1<2V8X(?1nh|vx`=$M0D#W1Q+V6qbBO1z{WwX2?t{`C{JcCG zu!sZFBdF%Z2!JtTciKC8Yc+_MKUyf z)C5?Mp!>qlrn7joMZn_=V4;$DvKKoqJKF>t@XNE(qsr}e@i(7q*SD_mb8S{vxE*8G znunYMoQ{{IH5w=%P#6F=1EW1(rRq{OTJYjAX2`#AxVJ^c7me z+NbcWZ=k|f1M?UYN z5#yr8C-PWR2{oL;%D^SDdG*lAAe9+wpBV886NTbgz4T~Exj5q`nslXL6xgLShuo$q z{oP+jELgE2Q^6_Sx?~8g?BD8%qNxYUNION{Yf|a|99eWg*OYTWC3~cFfi*-NvP!0z z_i(0uV%%*--|p(g9_*>*NcvcX;m9-GhfEg~zRpaHaNrCnFc3g~EP^gQ} zQR}fG$F&VX1y}i#!)||?aq>sET3spYv=l(3Ot67w0e@SomT`g-j0=eAcIA!gWAj-3 zdrtnhG%cuALeBlAbyqQa63)R6wm(z`Wv{s{A(RGMw$ut+ro_X<>+2@)g|xn{MGx{j z?;JA07PUdH>k>x%npf6KKk#g4!l#2${m8<3n|!Z-GNI%nNeQ1K?1Qah|4#RE0Awl@ zPr>v{`4nPSn}Vf~Sz2i!s=BCkMr^qlZNCRMico}JU1lSct5|OQj;>V{ihgk0EMQX;D-i+0rR| zZm*B``|Q}qr}gB=2t2K+;@@F=tWp;S)Z$n}WebC7zrezqh-K?AjU`<9p0-NloMc?m zZ*D4=fkqtd%m(st#;obRrOE-}3Sj}uhI$-MZ6x?LXPfDu=n15~>U<*HNWzVDw6wIdhg0g6;Ppb0ki!9$!G5h@b=|8V2uP zo0ZY^@C);>d~^B_-C|{@{chgHam7=upx=Clyk!fLmb0~9J z!^w_~%!b7(TiQlP>*_vsHd9gsonaZY^NsH~Rg>!)gxJh3!8Q~$7G5hnGlnuwlLA*F z3~zA=@7}4oKEZveZYXv`LsMO+xkdNCyCEXb*q~?Tz8tSBkfDZk?v~UV6bv#vUeHIm zaW&nhnk)>Qy}u2smg`}8lEj3n={h@Hic^>*g08SHrDL?l|}4zqTKK zO-s=FXfS!E$Dlb&Wk86jw%%?tR@<=j9@X`OES9y-Y5 zceT5-?N*UImqUcj1R5=!Jd79j;L_OGvZXz_@C1En3Cw4FNyA8j{9|Gd;*vM{&N{MZ zA&fE`^-6bjtcQUS?7(YHKiC&yJnq?Y6PTMu!TI@dfadxFFyeTPr!wxNSm69jc)L43#_^Y*B~& z!PccOTZ92a86PrzSvq!GJ)AG~rO)UDg-fd_#}Fy1(at^_Ga;*sl!Sei<8y1S2+R+{ z3t1gp)jM+8xZRwxL-WP*Qpv2li%IW+LE+BrpE9-vvd(^7+D-JMjBrHsFA{>;w-fP2AB^Ovl3Vy3J*WYKn4u?W~GWs+;i*l1bJpjriF#w4I*KRYH z{C)+jC^!E-g{j7JW3p{ac^hR9xnSMgJ#)+4zqXv5253k{NIk*{m5SrW!2~vfhb>`c zN@1buS&4Nb13kwdLEQ3%l=|{VH3Und{66|34LCZj0?^^^D;(_=T3cZdu2LNTL*Nx8 zoQqQ|4vXx~ZVTgRYgUjQ8ZaqhEwA495$OKH;4sGt8s!^FesrP~V7GtGdj4+cd-8u% z?Q6C%IkW9cCJAo)@@t_7YmFbX`_NXYv4{K8?}eIQy9AMPFyX=djzwl<)<=K6Ua!7$ zU)$kW+f7P^3d2S|sT-7(Lmd6r^s*RXM?xUi>4T^GkL&DHz3uY4iU-u)5@W81%`X z`bh6bgGTVbvJq1!%OBE-p^N4Jr$STHRz6}!^_#0hDj1Q3VmB?#v6NcZq_@pb0swdB zSZ14{zLPE(NsA{*Dhd00$CI838v!ic%j9v)>S=S{x!(BIt5#qJ^x5%B`&;f&(gRe! zK!39vAg{T^#Ske62 zP-#5)-8MeqBZ=_uxo?YSUEl@qSSA`Dpo1p#3!c^X7r5Z=m)rY8*CozghNK3Vj!-o2 zUUuh0^h0vfR2B=48eC?h=xZu9xh~f4#M%;BjZEl@2{t%Gqb18458V_dfaGIrBW11#-kK$vHW^W`BbCHzazvPfD}GtRO({L3#Ew~)g5f7B43L+o8oCILHsrMiv)SkW-Pg$?T2GNy<+2#J?x>8sSo^|BPcgP9U4qi z8HR$NwILS+W8xmbxT4-Zo95nTjXW}gIVvD}z>IfD>24WBN=1aQSsYb+`fKBdyf&q- z0wO(vCLR)W;`PNwnShAYry|X=O)s665fb>^*s^8Wx;p`4W`*B|S+I&yIbn5O&c|aj zr**xd=Dli1FyrQCM)Wnq-6hn%#&3limQ%k~qn-w(!rT$hsd)K;plU_JB3&K2*tDjnBRh>8V*i4o z-C%ic8Td>?d*|5g9#nVKle{Hrd-88e5eIRmnd2kehQ4tWO{7%9b8oB`0uiq4SKTXQ zgiF5Kb}Hf50LJjQJ`w>D`jaeT%QRgzqD9~>uX}q=5PswfTlO5O-_3ypKB@+P$}+Pg ztQZ;i<&9gBgtH? zg@Sn_Ndj0u=Brj6%{h%C#SA)W>|aeo{2PcY?e2)$@>>`ut6XiWO;-U;2n=e-C(79r z(h=0R49l{!p&07Zg%HwiIZcj|lZqw}CX%>{mMf!49pa7$m9SX55*Wn03(XO)yE3AY z1GLhmYANFGFLVO-^2i3l?+K^_n#OtJGgAzK=K4faw-K_LY*&YQ!lY7M4t6iwOi!wb zw~j*lYX)8ZwzsUZ{2bBVIW=|e*j?8fS*r)gi>)O#w@u8gZIkP}{$^JWA1fHGU!tvF zkpEqjrB*emc7g%{nZo>UKM1$~JcAB~HvgHpsoB~eaw2`r*44KsXS*!TSGl|cS7>{~UfN<8agSpla5wAv>IiIm;MhF%GwC(IVnJ zTyX&|Ky?F>{vGlO1f4fbJ3mSQDK4gM-vXH-3f(c+^Sgi0aej!?@`bg%@JKya>?fP4UB_mPASrhAyhzSn;3r4Fa|nvI=Lct(eY47N#OS1v<2M3 z!9FaylvQGmdZjP)i$-vb;Y4V}GW3g*0t#hvPgJepmF2$ab}(cXMC#Oq2=D<5(b@3; zbgCtr^LnKU4`7ran~PXB!=~DUO3xj7j;0$i{`C?~SfcwuPE*Xv7TSH8fO)VeDPtSc z%sZeIWi#GJg2V0leb|I5PLAIilbiP*V_dQzXg1`gv_GRC!71~&~36P@r# z4B7q?-$Gx>I(x#{n8~LLJ*p2vBci}1@CTm1$J=e5|HE)vtl}Qb3rdC$+ckrYw!Uzf zP#KbN&@1|QAVmME6{Ki!GPHg!*)m?$+BIo6h@=JYP*d&QAy_x%W5=OzO`2ogne9G; z&Jf#9fw!K(P`7VOn2Wi07yTH4o}S{BRt^~1ElJCkL^vb%wYc@N6~~Hr%l3^~<5BZ_ zUW$w3)8zrczBsse z-%^hgxB8CkYwymj4t|hhEe-+C=YqaF-gR;xch6Sbu4>>Dx7@+s4sA1|VtgXRT#gO) zu$vfZ)$txzw-KIB-`>0$4;32LX+CLA9DCi>p%x97hg=^2eZLWO+}>TcA1rHtBI`;K|+*6s`k)zadV)l#@R?sv#vW=1^?8yZLIezmS@cLO2Cf z_E_OOIL(dq2K(Dk8SFo@y&ZZL4XuzKG()Fg>p1QCfGY@ku2=cf>F9jxJ`n%=%J+@R zes}-ys^x!B>i>(T?q=!mpB{|Hweu!B<`37ZAkhv2L=Lc}-Q+5}X{K$jn`zy7W0&p} zMHhpv?L@{C0{CaYS0urV$#{IyiTHr00mxOqSdbU!+OT1V(%E2@%NTFTtG6y~_8(!n z`fm1PbJz!;OiWZchF=4h9nVOLjG0hBRu@}Jp=LgWUg|{?fHP%R<;6b>KepT;g?xYC zdqVh!*h_|?oD`465Cuc@M8k*ti>RM#n&hW?NqnlXNfxQ7v84hH!Ye5M0c z_kJz?OvaowBp_TQlXQs4Q}iu2N1vw$vwV$AI>&DCdOd%^A_RIRQ$;LSVaB1mWxUNd zqM=yl_lT$q7qC4YgSxUYL22@WgH5K-aSKLL8j4#^0V!G{7J{YS39;puW}&7hr0kNC z6bw0REaKV^ygjK&&!7V)^q!>mT=yRSTWRBau{A_X5yg%mpRSu8qTMI8LiA&DTPgB@ z8j!iwC45R>^E-^*oiB_`42_mFW{sTd;FU9p>SU105ZPk>jq&-43Yr~J(w_qL=3qj!@^-oIV-s~XrVa5EbO5Iqa+-A4E6Q-UZk0Awo2%y$_8M3Le00*G#<=8q7B{Gi$I;RRqm+_Kw)_9r4#*19+~Y5 zDAOu4S7#)S+}p&1rrEF|x$8Ao4y?*dZxspd`&4^75hDUBO(>Ex+W4~H!xuNU@F%To zZ%sn+*H-k+?XBlaB<|dgaJN8AxQjR5NOAOq7WQuWn(tU%?q4UmV74rswvx#TWr*_7w&7|%Txt~~ z@lns*i=m#L2@?>Vo!?K)Dffnbx0;HX4tKSQB3f89ha0GHd>nDrf@9%ALoVEY>FYdg z90s&$wQO#c>VluZL*cAZ8YY8WcYR;GO)>%{-+pM*tdg8-7}0OyDv zO8}!N^H6KRu4Z8xmYumvLV8g#J@Zg^3z@SPxw=ix7jUq&hK#dWM&5qMNkRcCciEto zws5#CCo3_64B(<8g<4p_sR$`zvI9rfqQDkq5lC73YF+6lWDJuNZAp=sVqG*Ac08!h zJkEqURVxg;>)380^}x57{H#46X2VL0l&cosp*S?HhISJ-oYv{z*%C~X?E&JM?vRz} zI2^70X;dgzpAr29EIn?vjRzvo$t!GNBeYqZLtSyytB~-&5puJ^gqTdzR!i-QDErIO zMc|V%%PXnLc|_YX;RbhblLITug;qvorT_*kchfIv)^|JHt%}W_X2}-4ibh&z=HqJ{ zzd#SQ$+~oBZ#%pfhwF_91&z82vmspcI@2%QQ7+P868_-LrQvR##IrfQ0AUqBvVOBsg3%_%)L0c;9Nn#(>JG zO!l7vZJL`Ju#X`@9c5ZSlll=!VR{{w>H+3#c3a?0^VPeabzBll$oR&#rT40nn;GnU zCKqKE5#h`!C5JYlI!=s$g$P!4bxH^76fD_z@AqA}XWvHKlTwR>{u9pdr<~y7zrP=( z+dim(ymbDp#B&{cPOLf)bkLC9>noc0+LETaN8^GGpr+S91=jVrtBA9lq*M{V(KG|H zhz(B~z{_XAH2UDl8HP6b4z9jqB+Gu35dBKSbW$PU*}!+~9JYIChBshXjxBmMi?%>R zwGVXrgBy7*m8xu2tMfJsv1;bgi?1=`cMSg>@3v0s-q|35fNs(LH*Uzy)4|l~C*sv; zd^>NoB7E=ohvs&1Y(&GzJ?s!?Z>Lt6jY39ZOEV3{pmqjP_)!hBx1rw5N zr2nn%Zd-#T5#A9&1rTmsyG;e@cjk_;CJsiPN0|+3Rx4*7jOWSrp(FIDb+4wF1G1Gb zb25!~Mo5!a+2l>cK`g~cGfqsz@%W4}?_6>kHgpgI#}WYEZf+aybK9`ijpB;-aM;pH zzn*D@DxR*&-f#VUyw5C{kYO!|&*#1To`vK4{9Jt>hIZf!p({vff64tkAl9-h__sB4 zMvjxfr`p^_G&*94Ds#-aWrgN+p6F3MmPrEh#p$aZrwxB@G~JnvC#uaq1&ubFddsZY zFfu)PfiohW=5HnH(Gzi5`zDeFPeqw?AFI7PK{^EHGKz`GfUjG%dgTkppKEBWO^Onc z8~E&6w1h-M>MLjES_*HFR)T-#9ySU6@GyUHOxpiu^?M2CshJyNL}DVV5W+kn0Oh}x z(H2JO#%BkdrT@Tp^Q-vMn9+vD;vM6X6@P&M5+DZ+b!fB$h6Xl$e}%ECXz7~YmDJ!t zT-S>B{UEB%v!%T)V%?_SEvC$PKy#-54OjFSFX>s^-&g|_nxUm1#Up&N@a8Vvx=oHu zU&^f2@s}#DHFX33Vj4ct0Ys&6B!y{aNhVaR^|wgQj#=gn%(&e5O%j88`s!QX+D1II zp{_LDsJaQx6}u3m-|K|Q6u=8%JozL45w%>q9o#@5l+P!>fplGS^MlV!dj<*{#sPZS z%n=u*(crBhs}aDKKp9=Fy^$BGY{EFN-NcK($C|$CzSfb86wp&RhA!qbu?n!qEf!r=r4;{bkV$_m|)A2#{powh{1K*ZEBvaRf>l`w&V zkjy)GpHy7TUre*AJX7o|`pbA*UkRSRznBVj>G6Bc$^h*>BWB&%=JIdwduZV?&5nwyE=e#vKVqP0LKZV*t%jM zku$;T*1W;DxccK_<6x9vP;^U43=UgvFK(Jk78TSa;|p;9-4JEqC0yJaiyp=lnElg{ zrDP+Kale=mhY7W|XCpx#xpMh!bFml!!z&dHeUJ@}1|1hSuErctBqbfrifvG}MaepI zo>EhrA{MBBywtcthq%}&w#PI;p0Iqp8A6h{+&M$GkTVk|M|4mRB%9p;KCN&Zv_R%| zg4Ig&3ToE!V>N;_CKbCI5kf(wX$(7$uRU4I*hGH@9vaLu6kOUndd3xPl1HXWZlkvv zGC_7#gBYfe{&Oc!+?2Z%7PKx()0-{^QXw3~rT>i48Zt$GupU*lFmPc@B32O3o-Nhj zpP+n^PEu*S)Wcf>N$_$@0fi@%Gh;hY0;wDY1`YAHW29jdGofC|2o@KhpRdjwIxAcU zY2&D?oXm7YH>z|*7%~k&!BUD`FGW~;51W|se+BhZBUnbbp7)pSgF+rk3!EAPzq_Lt zg9ql*)2ljy>PQrTVx)t2(&?3^tO_B@gE3OXWjr6!7g~3D++2z}M*Va3&2vlS^i+=|Y!21^J5)Q;;?cNjvXCWrq9s zhmsv`Ybf~Up$_ji<~PgJ{JD>ITJ~)`v5{D8m)q8|p=v?ZnyeN@|Mla!BN0wI^i^dN zXk)z|m4um+qOEjB_9P~$i<=Zvlt#9fM)p)(oU67uOYP*Ka+n{)QhoB~zClgHf1%MG4$^}ETsm8ik6lGoeGVtBj^=&{X0;)=hSa&Ls`5-ywUwj& zDnM`R2+xy@6SaBxrw-+ZgY&uQ+4=Vk2N7m@EbCj}x91h+OJCRa%ekdv(<5uwB4%X%zoeAe!SDg%jJ_^z_ty>X5QvN;6#Kw#j**$o5T^YC{4LnKJ&phK;vMCQ z<%cU#cxR0jiWci`+dwE_m|sz(7EWKOLq2Ra&~6{>P}xS4mcW zh!L*mSzY4@W{I62A(gWq0(-?kFgygzLP|LeVv(HmqE+wZI)wc?Of7L~sp<3Ms(tO! z(;$y_k7YXc#(6t`@smv)?Yrl@n;Hpq6UeSu0it*i4JGun#F!DHOqt?cWn8J@dHIFW zjxYuyxT$Etzy(GTnp~{ioxGoL0zrsmaAe7d+aK<|1E)W%9thbjtWYa#`8YK+_dMNgUV*oxqt69N-4R#llyaWV_6GDc$>IQPiSn~?G_ zjg)))(dmB}2y;b<#iLeWahTDh13Mpf=ELZbA9ZcfQ9qktIGk>YTd`Y)ZM`z4B;I+G z{=crD%w(3xJ(UdJLlMC%8#P(YGlCJn+2bOKt{Yp%FP90k_f(-~+i3fRTw{<6@~jJu z;8jeLFwvQ2WuUFi7Ox8s-^+5_bbigINmt$9pMC=U_YvBD|A^Ipx&uE3lR)_YV=nS@ zgoe(}raxvqoDp%3e-hch#kEB+od_T(>o+rluARLM878 z3205ZY;y-j0+O7F+p_#LVEFe^kI^66*4hdI8n&t*{zcgAH#IS8JKlCUNgQ!fZhMD53j zU#3F)#}g^_wH=)O6fJV}p*lO>c^cf%SgZdw3R$v`kI(b|o^{-~1r|aucr}W`BEhm^ zpfD-1;n5Lpmue`uyjm@$6xJ~SALrNg_07%g)%DG1#3>EhQ5Ly565*^w1WJ&cACLdP zw`fJI{5 zd=+=zl>NtATcBRy#$#znu7VYJXl7gor>xbC zo96fRDfOQ#Rv8)?e1nG9VL8DdC1UY6Z;qfMNsNoN^gv*ejDt|rgjFK8fdSpsgA&#$ zjnQjTijuXvxdzvaOs#VH?k!117qI*r`xT6JE5e(;VjJ7_u2j-gyVccoK#NwBnFxS){AeDLmxP7&VYZ`O zR0y5va1~i<_1FZ5Ol~z1I!^C}F0NwS8*YlRRpZb~mo; zZVK<$Dq&wyG^0U=mR2)GL(35fGxqlL;NvjSGY^Zo2$gQuS%+ava+l9C(V(KIb(nx_ zaUs<=i8Ik9oN(AmA+}zaZZ$QFwjl;<#2|8z3Ax8*MFz1X23r(cr?qkHLS~1fVM7Md z)T#r-u_99^xd!Tl{8|w$w>0Ynp6eijP1;2HZLxY&19szSO`QJPOhp2>N3&)20ovq} zoHCrUGr$%p>Cyt9u{4=qY6k$utWY!{jhjVSI6oK+Wvs^skXq9l%Kvcr4{c&fVhgYGL zk)rCPjQ}gQHTM)YUEeU|L2EPj*D`JLhpc=^Qi*X*2ul>#fz?gV@LRc#5pYpm3OOsu zFa4=DP%2<>NrxiPixbA*bNh^m8g}g%ypPS+Rx)X=dCN+O-KJFCpsWqJuSZo_TO32| zyySmK3#Ld;U&kbp-965ZXyHml>@C9I4lGatuq?y9suA%o*xF+>wW`{J+cumI6}A7C z4$8K9D7RTVAQX8M1b7}?PfX0_?VsGt4&B=Ae0_z_Z}ppia?gM!4o3D!H~}D25?!2-AAlUk7htxia{sgAeBz9PN*-86Pdu zJXjrEon!$jf?c>9Y*fg*7}`2GlbP2u2Z<7g_iCIFKhA8g9QXUqRe(AE>OgX1m>d3t zM4gVkMRCH^O&Y@!z~;|*i=IvwQLXi{w(1N9jvF6bxVl!qwzJ6bErPGB+?oyhdC(sY z)-7Q6iValt^Qa5Tv>j#q11WuU&lzy7eb}J!UlPK881&N3gwm)>DN!xb#Bd|rvzM2{k_Ux)YOO_TU zei@|M5e5FNk9AxRjSVyY6F%?p^v`{DtDz2``2URBi3&T|*z*U~Y3RQ6$9B-a6ZdAm zF+N)=FAq+q_WzGX3VCP1b=6Nh0rdw4_`kq4KLOIxjGO8eL{QnHR{AmyPQIEL#|{JVIg>MK7({iqh@4z+1^C8qYxy zOUu!QZvu+Ec=>j6>o9Y6@9?>yGtck#_sg9<99~ec>yY}0&L`KP9D~elkCB4(d_2DY zTz6TopcpIRfhFU6dHVi|?d#>_>gwiC{(~7}Y&Dqes$pn5c%^9+8!>yzesqqPkyN51 zfs;BB~WpukG{ji1TFU6j3x zOe%4#ZQ6Y(YDqG(%5_l4)M}e-AZ>7n~GYIw7KOejq&;Q++P@-@) zEGrLT8MPrl7h2X7tT#grff$TAP))f2ZFy#KcYDBt@E_cFU+My?|d}Ucjf}Ku| zq?2?pR~ET>ctKe>5f6sVqdV(gthQVRvM)w~(ZU;ou=*T`p4D%EFl7z!&-a{v6}ybl zl9KIu!c}t?lE z$p9uK@G#WADF@g3F~`h(T}k`XunJa%pTFE5NWEN=F3Ie|ZAC5Q9(&r(Ea9^UgQUyI zWY&gHm;)y3r$w8qv?I|XUf=wPI~14{Kr430fhjYD%WHJybZEjj%XEaA1+ysuJ*Gus zoGBT`kb%n7qG&7&F&R<1C?g$2NNyA4szgguyaZNmS{KT-#z3ok4UdHt8PXQfWs041Vo$qp9sm z>M(m>qh<~*kKRbHa8GfVn&S>B4?9%*1G_`%&j_rMqiPhN@+g53BtWS2*I+Y`O{WPB zhwSLlkUG^tAG4tS@txwx$*AhHX}Mwu&X3cYqQJ(o)v$C$DXz7dr(|hhFb_<|J(9?&g3uPz0efr|<<@6#@Ye+0}W4gcBQankO7Y-6@^m;*(2&<0I zF(B07zz)YTbBGa0(b{O3ZMe7$Qe@C!6nq(7w9uRj&xcig>7W#5b+8gqY`h9KI_k-n z3%42ihyIcEq6%mp+z6B~c@v|K9=cJ^!QnaPfcQfyq{aUWul%sx@fPrN@1ug69k;08 z!+QEu_l3%CmjiO!)kso-wkuNc3|B0k-iLPZNDkM*P+*k>&pv*@gZt8v>cz-$-p+{m zCvu~1qB;F@m$Cmv;Op0{Ev#G7>2(VeHUJPg8ajJo;70tzK%BYWD(SvKb{a7`Oq) zBMcn)bi7_ATzZ?*B0Fw}CD4u)pe%ykFebzp7_C_c9g_B@OGG;Cz?Bb-ykoAbzu5G% zPp2kpbB3A|oNviJI_)m@@3y7Ee97n|Lv&J zZp^R@*u59oq}3THwp;IE<#)$lu}b5+xXEe!3n|+frsXgnh)VjW5!1*pX4?E?qr_KDi(|#f1Xv>DZSx?GyZ`+++ zra~II4Ej~K*-PHH`&CEX8sWs_cv#$h>(z9(Q4)iRcAHrRxJ*P7w&S3wSxbJ5*X&UF zC|Btrl>cA|pdW2uz&} zul@iC9=H^>5T-0S9>hGJXOs3Ge;-uWq@7PibzP0`nx4M)cNKHD>L9GN=NzkmW0JG%>#@vD5T!b@&Jvr3?tT8(8abBJhr zT3VSeyR2h5g9HTHXRGifeI<`}!c%zKM~{44cw7tXR=|;Gy%;k=03!vh(beIY?D`VX z!0X(zN~j*`sn{yc7CRuDwXDM?z`dQ>tJyj5jlWjgTCRW{mBcBk6(_WBKkAV{RH+E* zzVC6|;5eM^oGn$j8&qfx;Qj$z2b{&3^o;&ICWx^1dm5s2JE4dZR(J_9xT)YMmj1+) zVXy)P;|saAg7P@OFH_x&@%~y96XQj~sjN7iKqz{Q4rmR|<>G4ECA*2mRQaOHJ8^_s zO*_i#aDC}`j6pyQ;GgyJVaXl7)#58))CVCgEln~8#zdOl>^byDm--lD&BrVL4-yE? zcjFIQ^rGXe)tU->J6EgMx$5I+W}>22M)sRIlWNV)LnwpK@G8E$p+unoA_t74$DzPA zkZdwh>Q=Njn1g=MVO3WbZ^$=#b>;8JgDQj(cPo|#Fd8@rp?e=faDpFXE{|*Gi?v`N z1v_DY7Uqi-1%3VEt1AL)OPQJn&ygSz<;fpep0Hy9<$d*exqL?RQ%o-~rT4-Ij8rC8 z$7$!`L+zX%_cEUF@GEvzukP>Aui^MV-g@P3seVq70-1GIID3zMdpdrrcs7?>=HOpG zO(j(|a2ksztXBj%$`9G@>~v+u+?PU71($S^s-w<&lI{hQ=PUgG1U8+S`Iv61Pte-7 zx(D65^isE_Ty}#j8}Jk&qp;Ia^P}kW7DlPNRF!a&4Hp}nEza_YHP5D|pe|Vh0oOC| z`NAO>{X8>H^}t=-(w#1oU>1cdv#vLIX+ty!O&2UnW1Xldqm%^V*ds1uQ5d03lyHCf z5yW&dbf<)$5c#4mK*H$45p+57sD!T`b7V{%&CS_hcK8Cm{Cs_PI{YNx-x%x>)v5A} z*P5v7op*gs^{?uNdP~7`@?wNF0qdc~#avkf>C&$Jz1o)enA=!V8-okv zqJfW)R|k!a{&&&$2j>RQJ9_e5+g#Up5AQb-PJu;B=Yx+0F{63?WA=~ko8K&i=Ee4{ zKR>=CTH0H>)<GFVFm+J-vMbgt2hc?eF(!bRYLC_@I8TRs?>1>kbYC?*wx!&n3~Rn!VkR6@HX8)^3Ax6NbnU$E!xZl84h#cJ)4^*8 zS}q#&U7o~s-nVSYx)_&U=k?{*|D)@if&>efEnT*4+qP}nwr$(CZQHiGY}?gkPoIaG zhEWG3`i&rWu2(|?nWkZFkD#1tg%)l5#|sU_32N{X!d$@Xb>56d=k?i=5ZLgyvK1x43uQQ`3@X~I;h2xmE<2sKg< z>M?nH7OaxrhpjOIearhxOg5Hh80^szZQ9^kK`c!h*rKWkSx_4*n&3k^&>!r$P2zdj zhiGkjeO4i(w-JX-PA&6OtMB@4o?2$-XD1Ej4bsuZr=U{ zO?{-LU=9JP1#L?Z5sYNt=CLPddlny3&898f)jMn1TYK2P!wg{f^DxiI?{@@bP}plQ z!u@34^hPgwWY=lU=#T9`REgt4UsD&!3_8jB%SWJZx*eS8w(^|3(5w{3@ye~AM{|xf zB!P=y;D=#dJf{oYO(dI3Ns=9^14SVkaGsCup%l~X+)sCSUpMr70Wb$xT~a$p`?T*X;Rm^6+r*aGui+F}s+uKd;R>&^fUOVVo`E=vZ%|t?l4p4OdL z6f!-TN3&%HxBxLFB?$_I?-b}`ym*T472wViBkQAh2PAqY zniqKd0o)%Y)d1aJLSzxRih#Mra}CRYZyAxaRbV=@EW7#`irJOTN63QU_tR$;xV?8P z5{}5tj~Cs>2|d{4w{LVZT1c2fnGZAT>%w> zG?#13+m093*V3E}hceE^>BhhZ>IT#68BzOuk{bTzIWfKx@0=U=0DzT^E$X6vEo7V`c`f zL;=4+T`KToOiFgoXP|$UcM%Xi?% z&Mxpk(_jgG$AUAnDtBt>^WY#ab$T;=v$8ls!pE=oM4nx|8&CTK^iL{j0Cq15J&HCA zA-XVr{wfhsXr9$y>n9~TJ7&twZsa5RszffnBjPUXzvL77x+LZu6rC?)``nBQfzPi%?`6B$T8AlLVH2ldx@|d}O z=oXDtsd$f3|Mq76#gBUrZ}}n)S?S}_FHolItZ_*^hrYuW62Bm(VTAKh8RRC2fRHGBS1 zFRy%3t~uK9?T=6EZ)5cZu*}bi=e&_sL!s-+@+qV%=4g48gRb!WpeHk=20nGlDtm^0 z)taPx@<|;wpc2nr5~u`H=c(d7EvZ%)mLycGV>k}vbnXg~6d7?XDdGisl*%rs=<-46 z36hQ_@v5xFd));BGl}shlwW)e=B=2-HRcr<5nLBb0gV0CfRbx1q&2w7Wh|s9yev}> z>A>Pbs!;6)Yn@dg=V)=0qb{qHT=2FW~#*l}@xNt9i>%cQl@)8r+UCMhkTvSC;n z$i{?;tjxX8T7aW2Gw==u+O4aENdlhWyDUQ%l{DPwlY6EQL!Wuj>qPZmGWbd;JGp48 zAzO&}VFgXvj@w!4!Rrl~CR=62X!jTiYS|!CjUPSL$C+(!bzh60>0`viv!~h<6gOqT8}#9X*Bf8ODv}-#Q?t(&N!P@m_|fWvaDex?J$<8g9G_}AuHPO7 z+@=SQb&r}X*9$U^Ucg^0x0COXf13QyV)(2tCqC}dl*Wy3#$ghXLIMq$miINWL@szY~f;|p#% zIa#T0e9qdqPx1QF#?UKm4YOH1JX}+#g;0rhnizQ*=~CW-Xx8f1u&1PlO6EZ$<6)m_ z=w<7#Vv-`&VD49SvGDnf4MS_^rfi|Rl?gw*q0i4I+`o0%(9oIr_HOErc zQ6*GoDNwyNXNv#KjycBo>oW}1r5tiY^-@y3^DcuwFom-;@6a(p6THfHU{2dGb*KH{ zGm`FBlo0{`gjax`oo`z*I$O?V_3YM#xt>Uuc`6+W0{!7mh4s(^mTO-qw%A%H=v(on zq9R63o6UxcdT+U#RC%x5%^EFNQ_tf)Q~qg@wH~TB%xq{yak0Uw#aovH&qSgLNHBOl zl$CY!2KdYAKZi+v{_+@GHVH1qXK#~8$Z`Kei|JhqA-MHKmb=#qeC2HN5P@c_5i?HP z5=SFbvzPT#dXziNLXin?*lZ77GyizLS;^C7G-Smh;Fm_O4;V%^S;{aWyt5errp@x4OetsDE7qs`UrSK1^Zc~X8-c`SOfxq04&Q_=yy-51s%7N0X%bq>D~cwdoZpVJM*-j>Kxb_ z6Vcdsc@^ z3VgP!B0taEoQSE6+yqEPchRl&I@Bsj^U&!zew+4orW%RAYor_0_*=pB^jcT$R!%4t zAp#F6S;r(zc--1Cxjd`vd;{`+yX)XRMdCHbH%%IfS@S_V4560HGkItz*>@lb5x&y1 zWarb*ynLL|mR~|5Y^_t)Iv_IEl$dCuYgKLs+Jg-lA7mC#zKgn_(sD z4e>2TU7w>igv07=Oq~W5sv2DsfpQ;gVS3|C=DX;Qy@B9+Ex`29Xa%q zggX8Je@?wiMJ(CXoN+gB`ICmMu;({#Y$CBfuS`5*Zk$+9yERHnFu!+tbC+<7_2abq zz?H%papD_71sBjJ2^~S7%@ae7y^mD1$mFjz#5zFDsx>GK3w*4@P9c^pAmM7H$qY&x;mF(VJw|}GNoFQ);hL154P>JH}8F=$ql!d!llZ5 ze9-J;(;4u?9P3Pe^WZ40SN3R#GZRQpg8z;jt}~$^oG5X{Y*RUegd<9@MO$X8TzApqM4ToSYkh5m*&fGVR1n5?{4i)r?}2 zT)O@-ON{u}g5~eE#!oce#m^iGR5S<=o-Uh1oy&?DSn3+K<@!n5uCDn4c|I5$!WA{;Ss0xwZAIYSdKB!rW6v3Pg zYHhSK`nf}!Mo_Au<;tw`*Fe-*UQX`ojb4eVt(^^9Zc4S@t{(}jtiwX+^_y>4-FyScYHFE1(haC^NQ zoFcy5>z>Z*+=1fWQ$OLwJ4?at;^e4E6Q+OF95n|N2FqDiwA1i%jvn?~USFQ}?z|5p z>ZX@;PLo%;J{uP;QH4-n=aW#7NGvx=WFu;%;%(GQ#nJMNNQ?_Pwj2+o|JkRfa~>0N9|cl0)08c?js8#>?QboZ#hOIqj6Qz58zq02f_!1Ry|5aE4Oah8$o=cy-*y#h#s;B8;^xEgc23@KLhWQFMW;ZvJi&R}dyAKg-u%^YF)kAZ5zZjOxx>k$unjHuJmLi8nZ;@Lt!Xr&q1+AX+ZQv&Y+L4gysRB3;>v&o zS&-kl_c!1~M%t6$1S*E!!W=hu$Y99cZ|GXRS!FmujT^_g`#O(sGiXi_ykvmXF@iob zml`1$x+ZtM7Kxg$>KZ{7bG{u3=nrt_NTM!y> zjJ-!^`zgkCQ~I=lPo#Bh&pB84Y=7lDCmKuSq6@Y)*Jde6VXSx_CIKSctCqdgZKTWO zZ*Fk0Jp1%OHF;>-i?*hCmZp<2Cv|-%dNN@tmfSpMwLHM1nOwcg+0Dth#ksk`>Ephn zF++sG}!aP6F54<&1t&eGArpw!F`VD7>h z&eCm9+TEtrvQ}zT8=X7L>F@g`|AqUgb-;b}%B%l=aYN|a>BZnb{E3vz$JYo9nl3ul z0r6AqX$nTXmp915f1hH>(lt=*abIc-fu5C~RxRUWA7!s`#PO4uQuy55Uhk#>t74QL z1@$bkM*yEZJa^V$>|LEd{Ni4nUoLmadn0&8l*4uSc?<~E-sgu021*}y2H=nwV6q3B z?}_P?9eidNrqLBS54DH>VrM zpB^+szAc#T7slAyYBI32x;i)2b~rZ}#Vn{6iyESD2Giac9@1q^Io*m-7s^7W3SIg&7CiiDEJTEe~Wqy!|Cr4}s8zvKlz#u-Q9 zq$0$OVk)?13rY-m_TQN}-RF~*EW-7DP8%4~!Id$l*48GWwzk%`sk2iyZ?NCb+qk;( z?`-?=(WSuN_8eKP>Yr%s=spXvUsq`!n0mPl?i~!55#t9ZW3<++WoqdUCoK?RZx~}) z^q*1swp;AATtP2#(}(K{xXdwzXbkJ=lDsL0V~+y+i4e|!Wi!ajxaw#&yZ1K7(mi6~ zJtq32-h+%Bqwf9KCJWL+Z}rY0WbdK49n>{?I-k3&0C|ZDdGR+x+mY5fV|{ekHg>{? zQ^SrCVmgxwn*)^@ETfT7t}e4EFt5I>Jh2p5fn^~pIKwfsI3cDdo^d8!rk>Jf>Wb*N zjgT;!z)?dNxHtgAM-ijWKPejQ9e8w15!{PAjsqqlWEl&hbG!~K7l@uM#z<{xai6an zct4PHNI395?&uT!Ni6^_2hxw*K=gZkK^I_SX2RAWM;Muak}nY;=n;s9PB%=!9y;v9(jkhNpClWnH;B@x4 z;Or%jn-X?7qV3j-p@=YtT5w}b64`r5UlOMC?t+=d^{_mxg|@oYFk)shMZ{*TiSNSt z4-+BoGDm=bhK29U#MTV}iyRrhN3pg4O(GGTNn#9a^So%M!pi`0S5e9P)8UyG70{gb8j`cneqtH}%DqYn=a;r_X&5-lB;BLDY59Q;+@3yF( z0s+i3RW6_PbAil@jw}-xKj^+-`_iS)vf80$cEBBjstmc%Dh5s+UifDp!(%_1~h2eEWPH!37~ zl!}g!N}w4OvA2Sw_p|E5kN`gF4Cs=4Q*yu#JkU5lEP?xnHEwIGUG@5deC9vXIE+<3 z`>|}uC&@5(V>-;<%X9}xaJ>}k`;!6GP3xRskczmHzNQ(6g-VwdZB2OEf3BeUumpqq znF#sO={Hxa*M??rW_#LcCW+>alk@3v(C)V9&tPdF3!hU3c}^}>?6wDPsdrTRdmm;1 zu)lS-#_niJBcG}>seomTQ+d!%u}`AVp-mh;M;tPqcG<^pbD~MbdczJ$4!TFlM8dRX z)p#tIG=xn-w2c1B<9JWe_{D$Nu6D3+1@BKSKhYO!((xfIJx;h|xbwSpDN} zIRj$)EW*(`P(SoCo-sIy49DG6lAbc!w=Ccp>tW7 zgVwn${#g-1)>?M(9fcSik62MIbASMdpI17PZ~H9;q6~Bq$LL@fb8Y8UZ_>5Vu=ow4 zouB5sPhp%+omh^;!=-BuXt1qhT)|*wU|nsUi&AW2%FMi6JW%J33u~7-3z={ebje*^ zJ?;@;wv_@Nz(~m^uK@Mrkt_*zRHb+i>izQ=AxuyphJ@+3lY_PIWYO295q2}4s{6-7 zh(Saq758X1h9#{4HAgXCd>Tq)} z8Q6D}AuX2P%P~cDc)TEBA^5^O0m>zzTc0ER!90)n((tfwl%OftO?gMK>D_M5qz_tp zeW%7M46LImn2ZaMs8ZAoN!%WFQ%<*H+4Zs`m&uvx0vO1+N4v68-F39r!qD?@ZD;X+ zC$v`Fr0wQF0{~di{dZ`%ySej!Bw3?a{}&oQm&yH67*1QAIp%Je&E>eo$^Nh|>CD_V z94L_>B#XoY&{873v*!n$kRoYBx|O`0!zXdr*YFP;4)9O1>;=O2&LpQZVf|PWQNsPboRNf4KY$PU}H+r06<|t^)Mq?Qdlt@n+Z~hU1v*thP z;s5(?(8K=|CX=oJ7j32rT{6+_RMaC2x^j{;dSmEvkDLcOiO9s437tvB7Xj%H(H~qlT00H3 zX8h=7fzM;zIg%BilCNZ1SpWQnh79Xg;~*6a(r_g=G3Kt*6@l@ zbWC+EENmrHJTKn8a>;+MKWXO^Uyy#y;jU}FyB83{-DTDK-d?%hUHj{6ymGz!;sBW5 z-P_j>o;zQO*fD`fW>aTSd9mssM6P% znU=YgKNc+c^Wl%fg$WPm-|y!~^62Ho=2n@OT^RQWF?VMw)@00kU7g_r2I~`BeP1we zm~>pRc8bNFz$7E91bRVbRA4xfAw?r}C9HmaR(6C+klCJ3%$ zOx6*+(kc@p3pNZ=lH$m6sLg!OGPnT>%=;1KGcn3UZsD*hQ_5xOo}r8TUy8uq(4e3~ zd*y}`LdAm~9sq-8j@iC0(kB6C2m5Jcpv_P0rg`s#&@(1&wu^S&0uo4NkRF1TL0o3k~6vD70R%D3Zhv`oi73wmmN}56o&g0a0;cB!nn4sWg-o(K z4kC<)6wZP|Tnu=e1tJmKfKo4aLDHvKI!w9}?lcwcqWf7wvz>5ea2OFgm^0=$nY1B8 zR3MOe!QAnaz!}iV?KC*+5Z0tZv70#l_WU_{Tm#T)9bl!(4Pc@*fDC?Da)?CSa|!Eh z4uKQm5~@*yUlK~)EJPEgr(xtN)|2I-VlFf-4uQ^7x_^|>B9>DbPBJMg_)?ST8e>G0 zl8gq8(E&;1eRXgJY9ciUXgH;}ZD2-XBxkGwC8Ag!l|Q(Xv>2{b42nxQ(biFYdZP>e zg3yQ`D2zJJVxfhu$p)_Kz zFW_33^dX9aVp<%MDUoneLa5?gjL)Prc0wQ3F*6UMS?o-WVAW%S zaFC-%+7cu3qyVNe`#jG?fhYn6bh1 zOd45461q2-F^A>yC?S(l3xlb-^_;7JQv<0NB}jMx19{=sgpO+Q%)DD)799Q#UM5O~ zF4R4$l9gyz6jm8_MxX@Q4LR7;+Idcql{V-N)gL|`_Mn@QxMdiN%;3tYe$NJP zsOblCVJ^;5E_HrRP+H-+NPI}~+nU9PzFu8ON@_=2N?HiT&nWLev8mL2-L8Z*-CEpj?@m{JI|pdZFa;cFNjVk|iI$WR_I{8 zX_#iZXp-roNk)_QHdgFD0C2{>oQ6GciTG0=Sr_a0>>#2l4PWOTyFJlqh{q~v&DTzY z(T)$l6`xEGx!L@Q&{Df`2{*P46zG5)9UpO*BKGi$$NmSW9aze zt>og6IJ$8sxjZ;aGv_C7_RH)nI5Dgj!8@jx=JL?cYn6Bfqh{uaDiMh7e*yGr=7f)j;!nixMTUP4Y%hQnn@mxAX;j_Uby?)*si zW%v5{@V;}ll+3-*6txvQgp3tQJupYPtsUgh(< z;*!Z^A-`^B`a)&u)-QwU{?tcHz#Ke%&1+~T#a}0Ee$EQ)F7}g2+s6@LZqrEUVaq}H zV>%oHSEL@Cyy?Dz?7_hA%(CY>U=gw2o3_4FPllHKh%vke+v7eNT^3gm zgX~=2pXKBVs9nC#{1-8e*R1QOli+18DZ^R&aTAOdh4>RZggOaQ)~@h040~F}+o(Tf z%2-ZE*aGbwNNGkdDJdk2tX>=A@dIa&ME{6YLA-w6eT!Q?U48Al`d}|!=G_{hn+vw@ zN@>90>_s$DuoFLNryHDkVt2*B{b{ZL`yu|0@B3sv-7G~*ko|JE-~0PEq9;9Ov;A^r z7h~I)4xec$FV($v1kLS?%D=fjd$Fb_u($AB;%(HxV%Rcn#pK&m>>=#`vr7Or_mnBNr~q$fCP&2hhcCvj(#BllBT>&x7;&`fR# zR=a2Cj}~0}8rx+68$5+_=1JY_|I%Qle~VZ*f&u_E{=+l=SC@r_rH$!-9Ei2(TS-f7 zPrUksJ<2&I$mWm@G?o``ng#|^frKtxOB~k}Z8;XJX*#-Wl8`?C;W8ddW8htaYw~fu zR$qJSK@TR^pJ5tGfl+EA5(b|Hm6_Pa_p$QPdcUk16^B5Vpn-gwRw_ho=p1oE;w(}e zzH9x%L&q*WyG4hYo1c9jY}_+u)zUtM3xSqE%y~&51=an0J$-zSJMOQSun@^4A$sxo z9n9_Zb^5*AeJxJKM-H*13QhwjYenvBj#7v;=Rk+@C2?wMj(QV#(RIw3!qav}n1;$c zXb-iStnn}};CcZ0=rdVQS9lAu>Z(fAFFE&^e?9b_8?aBrz&r zi=Zd@LN5`q7KAFjTQE2b!gplAZ#x3*g~@28c8j_N^qnG_& zZ!&%ezAYto(!}c7|D$+~YnL1lg(zC6qGKR<;|6K5Baduwm8^pF*apaoJz@VC)u{=p zsKwB)UY;ghRTMfH*>HrA3=-^Q#*GZJx5+$*Vw%O`h;WQA3!10~P3GFDOBd1bRX`W1 zFe+yr%TrJX3)DO%m8U5*b#2wj*kn_r3_X~BsuS7TT*yF1#?0G3V`pi%11MV;DJ+Cp zQfWGfD)$ZNB9Tkt@(lCW#f&f70R$g%-BJ9|DkdI`jPW1JZH^&4O7DLC+KD$f%*H)v z*N_~$E1oEk$T2y}#1yOY87So*_dPC-9^y&R5V}_l8xKU0Y9=N8mUx~81Kv;&;RtF6 zjWV1?8uCezwy<6SY}jzTUM=y4wm zo5j7>kehtF4=p52N?9mbHmDGRO{TmuwTrliF2%MW+cD41%vzKUB_=RMo;{cl~i6RmNv>$Bl&h9v|h+z^VpdKZUdzc<_!NK*D=MmMcY2r z%wi*udz%bC&4n`jum3_7W!|1D_AX}(bsm3*=_4m>U=NLjz${YyVRU)n%QrQm9 zVSzayOrX((tvPP)BB{+tqJ_x&DFs|-yA;c zN>^LOW}#JhReXU^8=CB3{o4SgSH&8tHoJ@bEt$HoFyk$3o;8$k)C#Xn6lhQN{%M+8 zF&n2h121c$6ME0SCo=FkHOugYBi?uU(h`Hi9dEt51AnF3J>-Ctac2G~a;Qcx`S&^?NJq(vCA}Z#-b%$J(F;spn4F9%#-A6qfs-LXdb5EVt^VmZj zbS&Beq9GLyGBmp~{v}h+N5ht`a9v;(84AKCABr2tLoY)PWm<$KI6>IG1TKEV)`K4< zaZYkmMHvvv3)apx_59x6m2ao)02RF67n%gfXv|+<;cCyv=`}aR`jTm6D!vM|`gp$Ih)D+F%Lu9!fb(i9u*D4b)>r)h({8TF z`7#{!&u)JG|C{sn?k@KKk=mE4Tib83p?u%!I~Kx3q)u67)$SOuP;Zj#M3qS*c1s7+ zJYz&QbxM?oDY@oUe%^4Gd?h>zG|SVKTgRUfUv?buDFlqr`!rl^RO%aCW8riG1rr$pvtn_c})RkA~(rQ?Xz9E>cjOd_*m zdq!5BYr&AW$$vqMZ)xuRu4!-XeN3SF_o#oDzsXi2=u;}v{8vmN=#vY7nR8zII zyKcWsnjZb(Rqd!p#2uWB2>XJXzoX7vkJ2nO5dp3a#Swo>n=h047y4!~^4Eu@NS6y* zZskIOR8QYw_o9TSSh0w+*~etVBL#X=oU}R3dIMF-sn;V~BBpfFCxv0yUB7wzzTqyo zsI#`E#7-T22b~B;!sMoRq%qY6A;hn}8}Hy*+1;4;&l z4TKYOrw@H4wq5Tn{4S`0^f^2tPwa`N2%q?)(m8&KZ=zl2rQYwt@0T3_*O0=QQvfI+ zw`hj2o4gKC3Y?mPuzVyqB$3jSh$%54H+5SO`wCH(oG5ca<@jCaGQ<}wiV{_Gj8#c4 zT7V(5t;3xTnf&YfGnI(>;&(Lfo*)Zf?XH=hxy<0FB#nY{EdTmOnwf2=HSQ8{s{Kj z@vjB?5Z-ILo9g5buO@dDaQBHLB>NSEUb3s`5!KQs`%+~5j3y)cTh8hMi-6SXP}FE-Ra7PhK;)mU{g)*CH1m?O^l2v@N$* zYr}+x@g!o&^4Jfn?t(2E{l535+2|Lo_%!x;vG|?Ea}jDhSxN7u ziT}9E15Ejt91$-IuI-+=_Q%XER*G+a{3{fKb|Thb;1wjwfr0ABMA4W_c33h%ZbDKY zz!n_Yb#<*Hs51I$raE(n?^G@Ge?wmqG&G&c{+yv261`wNy2f{*3o-FRXEO>F)=8Oa z2f=cLbPCTgP!T9@dth3a)HP>}<|sLY74e<0sAdR$RRrl;Adt5AH0o-w{JH_qg{&;V z@B>7e&7md90KJV?+gR8QNIga0$XzkC5J@$lT*>lA8W4Nw6vmW>Sj+4#f{&^Rvm*!&q~u`yjYkS;Ob5PI@naiys2m_CiMUv99b}99@z87F zPyKo4d^_`Fh$cAXk{o!`HtIes>qZ*r-U|p+H|KP)=52Z71&S`bO%LmvB z#Z&23jdM4ukrr0%r5@<6>8po|6e&eU%Pm(%iI*-ZdXJr@-Lo@}Pg{Tb!jIJ%O9n$8l2ylF=-<+0pF7;ru2%rxa|$Nhn|XH5{%?5z4?bH9Y@VfKvM3J?n7bC1d=dgA}@g5~Gk*sr`+RnSPx9ws!#!Bo>1bSJa(~;^H zC23WZv}nQ~HU)Bn6%wZNfP;wMPI>_Lp}I+u-`-L!)um*JbV9LRh+e2gqUmPhU_P3Q z6F)V`H_MngryG2ee;X<*Kf~|eg0|=t(IQVqRc2^X(p33-u!)Q2vg65ZIH!9!DH8Ep zo1eE?!DRWh5D75pBesK;&LaYZM|Pd?-@NZR4z;EmQJ#s)!{ zEE2>fTES+E^V<;U7VJvcu-Psv=^hD6EwGF*2=C(q{`|qF2195*6tprCIhch_3I@@p z!yAk^3^dd2bCE@qY>Z^uNa^*@3-#ORh|pru{dxO+H!gU8|JT+adxa{8v^J{s2XB2% z@~Ymj{CR6&Z2pqUP&7?R&C3e9fr0@i&FOa%#hkcN`}$J}oeo=fYc_mhcECu3E1FNJ zarL;w^>!$cX;Hp`;zH{Pn7`ny4*%1a1N%uEJlOd4lgxvTNR1*~Yn3bd!%@HpYJW?6 z);wKg{c8jH&&9x8bd9F&Sz|PdbA3a9b7Yf6yHrUo5BdHp zKGw~k^um4o>*jMhn_rMvub)C#6JhU#z^QHIx3=JhVL)#8pkM2-7(?O|58Nu#lT0`b z82YG-W{&M(QGQvZZ136dn$iKhbdAb-Xo3w9PAc)Xq@BHTx8N!Wa&FJ(e|k%qV9yxP-`i(E9(>d|GA4p$YlcP{3Caz|D#*{H+q)` zBg4N*ma4KBzr_aQ_o`3$HY%CMVcCGD5m01NPs>)L9exT16R{-FXsI1dgPD|r-`)Ll zQ%u6`cv}a*B0LgtEdG~xmH`GhsGJWK6{6S}PF#VDuP>wX)m~nHA08 zG-hQg4N5H)8PrfE1(ue;buD^jjeHbpsb)g99LeLx^@<5!c0b8hvhaKLY!F7lN`?i% z-oY``8p@ZNFU z^DwSB9N=F>RJIl*Kq!t%lO=NZf3uX-_DKsYFGw{RTRH=5P#1DlOEr@FW7}PEyamg+ zCvqr44T}Q#gupsg4@9wViOd&&fL;SQkyG+P{j5aGrum@c(p6CwInI(Ue+4`dQ-%%c zV)AjwBm4M~Hx&|tp45f{rYtXUTmMo*ajZ+9a>gmntFIO$V5SSmF5$vOqM~MqK~h^5 zXDx%SfUhnZOOe9+q7>^`whh1Reipqh58_E7{ER)^8?!w1-jEfot@oO2q{14p#>YBs z{-wscC9>SFliFD>@CloFhEMK?)0~VsnW0O+^4v7aU0#E9D`MVZnj7@(p~GUefj7r; z%@5RSwcr4lx-oX zevIY?My>55n-;0@7h>r)4LceKTe9TwN?tE(`6SmnoIALWF=fIUd5y#<`nK3~egi(s zbtKDskZcR?4mWIyHNv@$$`W+JAN`dg1KVlm!O274jqm7oym%jLf4e6J@JsuEU|;m* zo_BciO?iZP4%`L4h51Tyh~^!I0*B1|BpI9|x%ymLL#RKXi0N|u9U+ZLA{mRA1H2=# zRBN83cS-8c_0ie;{z3@Nwf9k0dS)@(cKxl<@Xik3_y3>$xDy=nJ@t>Km;0|L|J!Qe z@sIfPuOVHPbfdQz5PV+iHGDufVyi<0J`E_u>0w(ObmAXTp(L$b+72UX4%Twlb=N{9Z*SjwY{^{&K{|LSodn&1AAT9*RVw_#OLve7Jd zGZMuB|drNGq+ixUErEFT;;7`##ILB_u5s;HsGq<%$-xc{ch>)kAAxW|PR)=wyvt4U@?I+4DM!B{S6asH9$mz2d5M09f*G zCXrT*-&uHThK96FAzZn<($cXd`<*}e!SiW0-)3-(I%y0=*+k$DVHeV{Q5dln66fhVq#NtIbQBITJWVLHu~U1R+t5Unbc+LmP>OI2yq zXJWpmMlRNj*U2V7ZI_dAmQq?L)hQzQBL_+40{XOFL{w_oa+F9i;UuY--01^wcVzQd ztXLg|SjnR8UAoV)Y+{tE=Pcn~#5vrjSfL?PPP}Iei`#WHX$9~%WZ$iRyp2|k_-0@x zU$|A9mVnrCFefWiS^h&OmVknvBxrR4%5vBg{xK(qa#-`tyAR=f@-ZUU$&&J%n&F`? zvq&2RR2(j~pr!qQBMO!m#v0$}_h>=yyLZ;*@tXg)6bo)f4@^iCp@MT6aju<#6NqLr)JyW%ijPVI!}8_e`f^>s?770lah+Lo zS#cUgqA25HT+}cw#Ka*|DccOy?@WXuqXedOu*KWyE`;4 zr{axD5bsGC2-yS)msv;2my$#k_zNB;cc%zO7YeiFpjp7BLf1ubs%+;n7m<;*;g*x4 z2To(CQJe@xr?zJb1LSC8FB~`oY>5V5%(*DBQ}@OqQ&`7T`7Uoly&Oig(x+LcF%Zhu zGNFfoSuK5H3*J;zu+J?^qcwkbs8!`r4MkM?Mj)SzS-0v)nSSH zu7#qP!_C8e^8zQgb_I!IlTxQ@r3iqOcEx5!z&#-lCQZij=M`ieG({e{WI4vXvwCfG z5)~8IBD?eB1fe}|Q(}4%dbFveVnysEMUyc}ZKPA5z`0d$z{I9cxJkKgAS(P4{v(Kr%fIYJy`z@ zDscNWR~AEmx@fV$@_K+l4Kzx3K;sHW8wgP3{%FwSrrdrTHWeF<5&(pvpYN~gue1v) z=#5};M{6Ucllspr&>H?3Y{JV{Nq>{J?)x`AcWq$5MH1HRGia^Fxc+6~mw4t1v%j-$ z%aWEhd2^KM_wS~i1N&1-%2>4n+hY9SYOB}5878|YvqIzeY){WnEK9atsJ#fYlP-=B)t;%WvA9=qi&>7(yaAb;5> z(|mpRSZ@!lHcwRWjMtX~eHF(`ieCCM=`~Hz1t!phm~`yCSNC9kFm`!=5kbC9ZLC_Z zoI7u!LgHEk#7&;Kp|8&WL)SZYS-M3_gK68gZQC|0ZQHhOJ1cG5m9~va+g9hf-S6nW zXPo{4`^$b-tQiqA)||Ex{M-ZV5yhylsyn7mFg{PoHj%KkVQ;3LTQpFADq}q7D8^_y z-dw~ThiE@vf(qDw-GvdUxoyCUZe8@^M`~=1@}amil6rIGonKCT@9n!HOjZa#S9W3< zG`_!jv`tlB9Ztnbi1&o04O$Xvc7)qM6g~ZG8veN|Okdg$q40wI|EnsF2LJdKs{eOY zsnHW`12m_GO7YNAaEFLwIh{4o`8Bi^Ip7HMtI);}rTmt(1>3CGhvd<4C7}*=0sP(h z`DSmhbD@6a9JG(`k81{U4XUBBaIugUO+Ea_)*T4n-86B(ki^aF*%egEwjxzzb@4@eBFeqU=)Ut?Sxmp@ ziN~ysTkKu~F_sdUf_$?_aA?ZAB4mI}bRkQ7@l@me6fu2TTOh_Doz|Og$*P&eAU!n26sXq%)28R+eSnyn(N~f18 zyQi-SI!A{aq4aRo=T`)pTThJq@*j)(hKNqO0eH+Z$YdI~Q8f7VwtI`Uqw~qY{dz>yoH%C`q}`)5y;!YPrg9dQtFE-J>^SI$K{eoVk5UE#^dtccNHV% z+JrsFSF<2_Sc%=xv1$QTQ8wm+AH96a038{9^FBP}{WJ3-|IHQP_-aSxY-4P)KS9Qc z%*mcYIk$r8D5v2p&`#S0VA49vA@TEi*=+%{n{ClIYsOP|peHr-0<9(8c(?qF{_vN5 z7G{lBs($We@TNn7Te_W_t96?6x!{j!>D;Z8mQ8ZL07yd*^8lWTGP6-4W#_n)70Roz z!*C*&uBHz}{tTHN3b|>-ZMI&NV=38$OB3W{>WYLy2P6S6VX`_LpXA`Du{&eEzUgEsbH2=a&&-eUC@b+sJ*D_h1G!ikU1CA0^Dab zqAzkVw|8vb`aJ)cgBaKOH;9QOIOy`&GglVYF7%Yp@A|@Qmsx)1ittGX39OhgkkV7* zK)Kmb;K(Zc?Y5t-Oq8ie)gsj~X_o<_wrHI7OxX2=un&rnnuQKRP#wqtjWAJwnUpCL zg0?q(3!kQ+6})g3 z9s|m$lx!iM)Wj!3LRNTFfT2zu3aG*{D4If5-=wb*!W`CQ%m(cqTMnyWnSqS9Bmha& zENv35DS_ATipbK2rS&47Nh81&5VojYEHYVBs^X|BT-17k+p;`<1pH(_dZl+04XUUyI3r1VD8iwb@uF=Lvej#`a(w4{L3! z3On6dYTNg_vhl5oJcU94*}bbu?=3Wi$=cl~7;Tg4@2c`@`QVSux5|_Hh3({h+`ez6 zU59q(d)*AS_OEVu|5_@3KliD`jSh4e`JOVjC2U>k*q>@i@p2LW_?9|GPO?y10||t0 zRcov<=5q1!bdxs_55acOsbQFEXn z$Di*tyD}ZU-Suo5M1TQU-9al=6&>4T0c!btd2AoEIYO<8IZO?k6$8mgF+p{*rnC= z{GL}+d(;}md7va18Z0==4;;upoPM2ZQc&lcN7? zN9taXP1O}?#U&%Ts>Vb0fimjdF7GI6?!*L!dDPvqtt#D8@Te3-wm{zHTA?1P+|NRJ z<-FdO?e`>BfJ62#3|Fr%UFEt$9Ltn}t5)rx-F1L1<|@>ZNqw~bATfJ_Qn}|$hh~zA71)?_)0b3yi0~7zCU(%@d%~c&k;M~yun%Fro5^gQ8O2gP2)Q#^ueKsD>Wo+O*Kn!2nY72@wPtt@9Q9uelk-I+oD>^3&F z^&Y02ToA$;vUPCnFNS$^a}ie1En|3Gk4NB3HUdmG`E+pcRqGbPZv)WBXsq5S8UC&!9D3>SxeP22|3EVK#8#qx5sE|gJO=S zLZkBt)Q66EKO^&lfj3oI4t&%p!tt0>3*#mP3^@uSjkWVGDSJhX{5V+zzi1+fGt}ld z)cbiNRSjA0@-^i$SdGqwHf4D(lm|TQ2hvf0inWMX?(MPONEg5EZo(D)VjO@j@BY1e z2rq6l6o@iYj&)@9VKUi}+78YzVfsfi(k+4uAut@*qq$)QBh>{j$BBm+xAy|Z$9K!8 zGry9_8P?n0yWL$GzEqZ* zx9eqwdFYWHrxN(nDmTf?{8%se$T64t7tsH9#Js*PaIt>?dGph2{ZByJ7}(oe{EsUZ zld2=X!GIEa{Z37DSx6x0)@y-Y#}qmw)m&(EQNj_CrQvchiXXyo**k+Dv9ILFMR#+a z{p_(Y#3#9309_;+!#c%iL`wxpj%1~jUnNSF)--q4&D&Pv&{z;%rnxXpc0Kz6M|*WO zxAbv@r^Az!Bz238PE!a!qp7(oWVP-VAu{_+b`c&!cYqC zMpVED5#=7JRV}G)$|d72nJw3O$3-{Y-BeN7>xj`pojPKuB}3m5@OC+1PtB2r?Z{rk zFZAWzlf<)ee*=yx5VU3+A3@*I{_?&9_ZuF;9l`$h3GYQLavLxxxevLP$EuGs4WlX$oJAld zA9Ba5`Feux)&3h;tVbAb_mGsbz7CciJZ6rRK9MG&qdy#2oH*yr4&z~ld!~_UjM=Du z!?Tzw-N_P$l)eVu&W1rx#`s>y znP?Ec0=>ZgxuA!Sow0uf@MLD}!1RBb43|Lv@&o>#B4=P|@XtC}n*<%{!GG{{PYrh| ziegEbGDhJ(Hx*H6JxHZnmfXa>G4LjeltA>hA-hmBX?sj`aKLkX9%G^j(xm!GHX3j=)%F&p-6kWtR6_zrA8D)UcI5rk8gEKp2TG&BkSB?*{(;Px zcHFxT`lgF2B(Tf@%%93l<01gO>XD5a{D?+eJT99(z6cqVo(*m?m?8xD+O*=;EI0oN z4Hc?HN)RH#DAmLPX5~gqm^RNsg5Q~BDm{6&Hz(tsham)NOQv*I%zhne#W%qBgU!}x zivbWO+VM{lpMKnk(HV>bdj-QG&H2KK|G63;ot4-2oBxL0Y0p<_n0C>oVtwpA9B0us zqvWNut43|g7Ye&!&ynZs$5h1hV$UV^cl^ITN}gcwOw>Qj1L9vR3;)kYu{N{&=RIqR z+J-GQJA%)OI?OV>Eucp=Mwvd%IzR-yxxxw~O{)(CP{ssqHH5<$?N|&E%4MsY`&tSu zrNaQYU{`hqGqai4c!M=a)4J?dz9UKIXQ#4z#n*R{0GL9B>{ zw1KU}Zm}yfG(|1H=IllMZD0yq7zqm1?19)bu`k3f=;}4=e}*s4dHMjL=Ciz9jJnlM zD2_6djz@bEvrv#5R}g={adn%)QY-1TLE=!&hxkb{(BC01vsSul#wyt_)(t9r5Cr3X zC&f)r8B>-(X1;tgh=8`7D*yft7zaU$b+!AsZq@aUP7WJuGldpMaB>cNp&hU%w(u@SxoTUcfy>0kYePMtfT zJ8vvkrTe1l&gE`P6a0uTb!L)avU-i5|rA#2Y@ZYJ1K{%?0RIM(gXg!m` znajy6nMlIhr%FE&DQTr-loptg9Ig}*Vf}Gf+gp{}%j;O8z}~)_tWiVaj$gOyT7m@~ zLz9lC?Wh@;Igm|@E`?48dy<98j+md(IFw__Vj`g0uV3+iUo#g_R5Isg;xBzYs1UpP znyg$Vd8v({2~be1kD9Cbq%Vt)#$1AFZqLrqi)?U#nk0~l># zM#D68Vq3Jv>*n@#=rsrkK{OGXd{v^h?(5}(%}eQ@Fpvh}jijW21bi4c+g>zCdoMBt z6M=QywP+0qq6x+{HyD=A=!cvb+}{pTfF`3{4B3wKCs|?HAQqdBCT*r)Ukc)=H7J+L zvTE2egk?E18FkHB1%Venm_1PMP8+!!L^rtMXJ7}U-mya=75x}&nRxTCpa>O~^V!{* zR*Eh@ZHc*l3jJPA?UqiN)q`W~FE1H*)jPZh@i`Y?9jQCEcDlTuVsbLUJpgdhH`+31_pID+6qpr5%;Is6!^4?hYYa?CR3LhX-jjQwjeU>U5`{@WLpO+DT0?3iIQUq_TlBs%*=5{@HJatQ}<6OnuT=aHx=g06E9>wr2<9M zcig`Q?Y1I~nOC(X&52hzd3a>r&ch#|pi5D8M9FG&6vw~8Vy(nD5#@oojb(1s65-mB zNHR-GVH(RmuZ5tat~FgQpi+k&(rp}IaE-Bvx|ymdRi0pi-x*lp8ZPV@_02eNY6j)9 zk=SxSNiMn)oF5Nof_(@xleFu;gL5b>Xf#)5*)+nevt6qj4_b8JouZJPa5PBOiA z?6nZ5yNBEJF0=FC?L&Uj(_X!R{~N^vRTcuWKiSTypAJ0Me;5D;j>aw)wst=tj!D~+ zI~PM3xxPoCJ|3t}{zVvu5bjJz012)F6+W9A)e^b&<65t2TRFmw zf>biIoL|yB@T~Qp^|M2fytQIS=C^OiKbt|4^2qkbP(r*&lm(uv$t8fM#U%jp@AdF& zuke-y?=z!C1VAU}{&=oM?`YGjwVtD{GsPOlryCPtxLcY8(VDflu!-CSIz%il#jpuB zkmuJT`-Xl(@83df{lyU*KkUk6Q6f;6#x6iTj*gqfZ&q(L!3*r@r1m$`|gY zf^7O#sYs;IsZ?-)KvIA}Jhn}&e7XsXhg5Zdm>>SfvLKLf6%}D=K>ZnSnrvb5zV~e7%rnh0wrS#C z%&s@AOwBU(1N}RudeA^0mV1a?ys=30G13ii;GWTy1AKoh$_D7)z z5oQ7I^`)(NJLHZN1xHi|ocrL~fio+fRVXyFZe{f3TOlV6zNS8^bztM`aoq{Ta%z|k zL)MOb&omgcSjZ6XQum&_HygK>rM2;Zv_etEDnwCj*Vbj}YH(%dGSU>FGI9^P!f;h> z%hu&UFunkQU+Bv{s}r*CCc38uhyEJm(s_w-L<+CUB>U=aC!%$y{>5*j_QW;ar187k zwEfG6HoNvB*FRv#N|!ALl<;FVk*qI;BozeZgy}Y8EKP&I>Mm0Ug@Vwh{CkW$Hxl5o zRQEzNz$z>BP3v&&tNN*_A~w%`x%CSNJ0T)%S$#o*!1PjCm^4r;4%1HS5i6>}cAQEY zRNOXg+gGh_VBrY4SyVE>v|13^tB#*?xb&tnor{m-tTG z>$Y)MX^*vldjhutFjGsxyk$vA>?L@Qhd{yEqsj2YB8|bk(K9~d2bs>xwpd2tn$6Nm zqm(b}ka~u}&?~%>Nj%@W$Po(w>6NU9)&a%Ax-MGv@%!sR>ENrV7#Vj{!s~oHxb+!! zsdo{qYPfe7{VPfKK)LgQQHK0&I@5u-F>ivTDZpc!Mewz9ti#*G01B1CI0+skGij3Z557#)u7p=lDK2% zY2CnZ!0k7Ul!t&L`DC!FTRb{=e+XoywZXdXk$(s%-Jr_CBw!lZ$Ed27qK~NGQ%GV{ z>HAL5_rES7ZDhNw)*o?}^g|{7*Gt&{qXk@?U2Og*Vy>!Vx4?knGgI4Ui4UX{-XK^I zH6*;^0u#_;L3;tHa9G@0%{mf`BaT4$a>*e#4?hwDUb9#hG1K02yp3$efZ?7`T+4>( zs2Utnfq_OyshZ)?_^^SQtC=N0l^%bd7(fUn;E|vlK`lY0pal}bRFuuIIzN?qnpfz( z;6F$vo*mhh_$URla7CXjK6j3OJoj2A?YyH_#+*pw4}?gThsN#(e_tm#Xdc@Kto$fy6PGpY*EO0p1G zN>nKrKqf^?>z|P<1a`Be0@`Yln%~nsp@^hP-z6NISOUmnTRuQUPU-p2iqqj#X?JM7 zj|ht%%AxO)6(zAqAh=vU=9T06HN)B18-6o~^hhg{4y0a(fSt*5hfv>-M#U1NS^_Zc zkDqS`--X+Asfw??Uryv1%s@iMGG!R@jO0W)PuwyiI^vZQtC!pDaC5|2l3cgiu2npE z(NffW?}FYwqIF}hc-T$1c%wE6w}VOOePQ>Pd%Ro+O*b|9)Zml!8}Xy#puBKknEi&L zVQ*nwFBUS@J0wvt);>@}4`;tM;uDA;SyFP~BJ_386J0zY5IyoC+6jtpKeD2Nd{iDp ztKWW;TqfDqElFvOE7Z&#v3iBgTab;;t?6|un_4C$qqFd@X$R?yHXKd}c3m(pzC zphwE-r#$v}Apb6yU)zUs!TzY<5dgQUEdCCj-5zXW3T@G)g3iK=g3w4t+Sqc*W$Dq8 zHlc9Hh+X0!2JT1qmxu)PwL3jm8|UP=O(!nTL~C~mn|O?LP2vDvQ-T*!V~pJ|C#KKt zuh7Hsr;X8IU|;o$u8V%Rr=~sV@}2kKV)cf^Fu zx+GlBt=u>{+KcXG^E^{+Afy5|r;7MjR5Ap#4`bcs;BiF`huEeVQI7NrQ)(d$JRiZn z2%BB!;Gb6EPnn!{$pJ#SUJ8T*TdjwK7v}@)CJijZg#!pz=)_6icN{qK=tdgQkdRcl zrVrD2E@xH0N98?3Sm;AkN)nL$k!Cb-0P-BZrALQ?C}-w~$2?%~!YLLE_xm-Z2zg%$ zRQsLN#3kYykx12D{vxAMEUX5V+Emn0c|^zVSE=z#s8PQ_>}WERR0n?+8)x5E0KLD& zeEDLZr9EKX>z)cO-u2NbD2NyPvqG4{FjWd1_sUA6(9a9k1+^!@9dU|VGv~9_4VOeF zneBr0>mz`Oa2mWgMf^s&O&EhRa5!1$*G$GJ*S|Gj8hU(#3v;y9;w*SJi#wDiI)$}7fgf;ql9 zRt&ZwY?Zp=Cy=|kU~~nJtF{L0Tp!iKmf5Z?4(=R`N#%VSBbqg2~Tw?(X#o~M51086po2j*i-ei z8>$fq-mW<0u$4(fSA>?AJD;Du+!F>EVEV#r8(8ggB3YpYWUu|S-*}?C%L^5$KMy9F z+Vh9wa${>0;)o%zC6&L8dMPe(c0trT$#x+v=NqMGf zHkNr>avZ>CdNw`-6|4IZHA(klLKHk`&=c*?iq@xaMa zemLC=T{I4+Qxv=s>v!teCcIbmGTwtdFB1v>6T&1LnQN#JLDk#boV0PayQ9ulBV&!1RInY*?%yMJH8`5j#Z0z&42>e(YR zCCy3wYm87SIU2f2>AMM_A1~l4pb?XLaUWoc(~L&|KLCi_!Jkj?e$pK&pPR^kpGV-i z&0iqQ?X06V3ZghclyV1Qe6yjXzO4Rd8#@!0oH<&4bL8&k{f|gE{t*fOJ@3q0Fu1kf zgZEt=BdQ^Axw)sk$H=(;c3N+}-D|x?ehL|352smQ8)1mC_{`ImnbwA>6Ijx_I&QO< zuH3^*YOYM`TRgK#B&S2TR)pjzY%1!z0H*;RpTvb!gk&821mdS{>{psbN8#hgy}D4g z+^h*Fb<7v=rM|Pv1K8R?Q$_2hQoO$Q2&Xw+n^klDsHl5mkxBXjNt4b4mz6kw%3 z9{3DzBAlpRqXX2Vf&zu<6j;$<#Nf>IZ&|vwhXfAiri<8m5ASg1ict;A ze`MEL*NE?4m~_^|4+CP5{GRrtBFu(rhbjEk6{x*Tf*#xh z89V`X7M#K25V;{3%-k%Ut;%pGxTK0!WRMk4YUWwMX(mSOL=|o>H^Q~^N8#%{BYVde zyIrkw9?6#6i|8Yo`7fUS6zu`tL8+0ZBN=*A44W!tYXs}GT-rVL5s3I}#PsRok+;XU zDuZF=v=`%tY1lzH$4fE?Z&q!TlNKgrb_wMgXHrXF%37p&O*K!gVHbYzvS7myZOXj@ z(=%Vwq_k74r&$2_sQzXrarj-ICeNRIG@rvhxCNUoL;r9uHk^ALm2 zi{BJu2@;+d_%ItEpYSs==sVB6#bUZSN~6YzC=G@RAj!;hv7`%0{IxLhGc^j}#I$%t zB-D2z0?~y~)gCZ2#ZXwTk{-Gd2hOR|tP8%P$kBst$UGTKKptQZi=IIw@y^ju08MWX zcLv|y5LnDK1@WkLBF|y6%-^>L-99|la|zHWK_rmWKr`xd9^(DV#C*xcoajinXZTEi7xuzAXMK>`(2GGKKO-KHQ9KgW4e*)nnz&97nRw?n(;Kap=I!u5Aqj z;5p_)8UVBWr4b{Rs>kw?UytPEwP10N^|EU6zEKgtybqYf-#gt5=uv6UK=otEp?kA+ zrro_MGA-q8#KrQi|G4783)OfwwAUk(ARCx4S#<0KXh( z|KNp_?fI(f{M`tDzhQ1F)xuA4kSbqo99E6Mhnx@}j+yZ+9%b(d3yaqvlvwpxG2MxJ z0*WhAHEZvH>NDf+d^)_m_cHD1KJ|DupL1WHG8jh!*1)XV;xRW%@a6BPYAEtP;53=*)w6TmQ4O8DiPY(1zr2 z@)C?PCB`naR4U5OmfZG(Fp0p?7KeTu7ohIcMlhoNbHjX)q3P9(qaWTWO*fF&HETlw zw}R#Gr1hg}_!i;C(r*h>2~i$QJnT2uj>JzgfUe6l^?NQ}a{E;D{g|n14Dut<8@(rf zjZ=?C{;Z9U-1QZii5<&#bV0Az3RDT}hpml%t}a-vv4ZI7&@&sOc2aD80t0*h%18sA zDm(GYB*-CzTDQOBvgc-@Yp|VLs6S3Ie-@|&iJ~2uda;tdk^GrN?JDaH8Krs*e|efJ zoldpzpk_2W+*nVthlNBmx=2_Lgv*{tu&>sWf(zAyS6@QAE?E7xe9@NWQdiv9qwl{^ zs)|jGBJqP#9H{@M=*A|d1}@glKOhZJo3s6?ct6&*#!bAMVfZY`tb`9lK-E&kPsMvQ z$RYx^4TQqLEJBstXq~|r?IX)7$|vC)>3M$HJg10c0Q7|3Ucbn+>#D;BE-As6~Z zAjKp@v8t*E0`SL ziqqw=#;GH6o+kW1dNkm%Bp)rN)yX^!(k9-rS$LM?xlo%LTDJkh)f@Mep#*S zqmreF4cOEB43{$7muVBgfPzXMyeM}O_92ORNGyf%G2|lliX)9U*{N0_Q(zFCrlf;( zr_PjFXur8x?mEuKt}S1za6VmHNxQDw(;NRE2J4_}Vuc5r*hqxubegV67J%)v3_GnP8Je><$|t1lPG? zEp4OdYj{is1Yhd5FSEzk)XYR~#P*3<-ty&KIq4`gxe=B^9J807U7lT7n{P0VOaoV- zX>x*{+#LH=!c#=-nIdu&w9h_Ur}?n^;L;#- z-^)lt&jn~JFG<&*mPM88bTSf6nuh^-6%(gTiW;1ggI#)8--z6!A7C{QQiX4 zYsmA=C6}I){Fw5M=6-+o9DRvVoZmGo+S1dntZ)C)CLan7XU0GH3;i*K{!1boI~o|d zSpO5^4^iFu=V_Slv7TTV{F;JlYR+mZ=tF@+L$gLaNMk3jb&#|oxfTQ|6w1;x#ri+V zUo{o2Y4g5@>z>Yc_uc~!yaM-%&lZ2B6~XFfpfdv1X!bzOt>-=As=IpD;yl5ncAE0Z zlTvK6&j?c~H0M$m)sdn(wv|pY$Tk|W>i5C>m~BD8fP42TO9@I(AwnD8xn{swnTl9S z*(c|AWe?d}#6c7N7{j30t|n`<%NWJ@bcFiN zEz~LmjX2BvGHWt{O>Y>}b(yT=JK}cfwty8TfQ&x1VCz%h>nc?+E z3Cm}TzGS!;g8^=WG|g95R=y^C1b0ShtXVVV)lny3W&n#q(ljg*YgN@I$qYtMA;(Gl z8K>p=)}Z=r!&97f94~R_X#giYKqtpCqY+{!5zWP& zqJS(y*hi$(&{^;W^KIDSA7Ps#RRe{kTVL+7mwn34hfM=|-?Nc7M&!2UHuzNw>}=8C znUGdTX@klgOzcyL2jRM{t^+ueaT3oqXr>5Uv(eG+)xjDxU;9+cIU|d*a2$Fg32x5K z!RLM#$J$>MC?$!fth|)n%$4WkAF+G7sh%F|~DPq9ZT%GDBG- zYk-neo+U6g>*ipQjYM&nA{1rudLt~@XGFywe=LuHww&=3Y&7rPT1@<^_C6qd2!;ZR z!KeV@-(B1~eK#{@(@Ee6sr3W8E*mm>dpmFOTY9R9f#m**91ImsjWHLM&r?J$WHZIK zJfeg4okAPq?n0~dQbfT#2q-a#j9I`kfh0sM!yYh;Wd!(+R?G@#fD!c6IOzwkAmur$nHMjAv-?Yc}EmLh8Xqr)oCATM!j@in+gCEYfchU>Sw)__6aw z-Y!o7Q3A^YbJoojX-$dEV&OUwsUgYIPB_q5$uP0QIEoDtVwyD-SnO`#45X9%;(#(o zf|QfDTv5}~h!Is@X}(;tj4%D2o&_`>p35Mb%N78&t=dkdg1ytgBj2*+^z#Zca$uny zCkO_2-5cw%z;)Kik~m8jlicvSLNx|E^+Yuzgr@`wC+be;+<1xPt;Dr%7z4;M#%Lw3 zKE>0=8ReTb*jq(jT610fn!kybq)?^f*G z(n9{!;pvynY!U5d;58E*O=+oC8|C)>k<$pK@&^&g8638PwU>-XFmPAwrXO8@ixhC? z#|k&CCoLU=%eP_7-CYd_u}SZOm%CR&daD4D2_#t7A57B|iVY?Xs;bn#*+n#jfNpxleD-2`17Tm@$RWBk^ggr3S1K10*VS5n2 zksN)m(&xw9=sa|(P#hvTuuftlnCp6C(1FWbnsGb@iZ=6+&>*>8gruhP{+(Gv@=96% zhk?@kFwp5PC3y(rc8S z#7(s`_@@;nh6#}AyS7Q6w=8Eu77IjxDKk6esdmXvV|yHT&bh7_DzwWH$vna!8a^C5 zCjyBiDk+62etstgX#WOP4Co#B7u=*E2G0Sg#h#N=u!W!@{KAICESd@1CrP(pgy&Xp zlTKMbJq;T_03ZqoyBmiHPBRji0mi8Fk3-j1q18jPo4T1+$NW2;YY>JRCYQ&TobHbN z0i&r->$jwDmbeJT-xFQsV_j7&eDGQ?F()KjV5}$_UE+oH_QTr0e-Hy`69`)ngxe>_ z#eNgqewd{o)PI5T363+->d`Im6`q4w%_8AC(2U1c6-4RmEJ;pB3XYr4l4S+6v8>>> zf6}dUg%AuluutvCS8d6W{Y;Qr{Dqcd`^EytEmui7O%=pbOfKW_OY#iBE>cBS3s!77 zrS3Y{eQC62u-Yic8uqfwJfVychtmX|gWUt=KD^BLlu-IXM-B*25tMFr@XH;9>aAt5 zhc0;?A4BggSBjFpqvuO9qZc&ZQZ6TI7qb!Kov{q@kaT7 zAumRqgL!^40~zmD=x{N4_P(oM;aI^(pk>~8!_V}87@!L2 z+7friW)l_JnVPFp#|k0t#)avcS=F?&UC_NgCC~N6DH4jyFkJG-Ak8#*y{blj8Q>06fcS!xUqLeCXC0GtUkWXS4BTm$FqiB9kw zDv|cdIp?t(^=!#5SNJGuvY2^AlLTOJ3@{D{{uihGe-jt zCnE#v|M5T7WbHOM{&^#TFi+U&U&LWyO%CK~Z`cXjA`!UW$?J>`MjAs)h(f8LqHQqu z^@OXGNUVI*n>zbJAA%Tgp8kbiS z&_5M^D-4t=ECNCv(|gu#CUA}+L6IJRkkSC;NYo;-*Jwl&aZ1evvN*<$A<7>C;NvxbRcl2fLz$)@X9&pl;DfBN9P6Uo$%R7&g&r?_Z%O@?g z%lUF(%Ub*O{b1kgu1!0NR*4u;2szOm5Hy8p{ad2?DV}61SjG6ap9TP_Zk-bAE<6Qe zzqtgS1(Yxh=Tbf)5^1WH1ey{QCy8HH*{D$hm_L6WQ@qM4KOX>iMzHW5sT|4yCaeJi zhBaGq{Zwcdx!z2wOD!ME76W6>XHZ@I91r=Y|ISbIpl5BhoRx#PJY3+eb$OlYtYWYd zYu^K8BZWM0fg*79c`U`v!{Z>w2$JETv8G{$3Lg+w&s7P9XI-VjTOP-ad;Ulz+>-6) zwO6&@z8iIzeA#ickH*YhUfR#%T#YAv(qrQc1%quls60v0r4jtBpXK7hj%LIWe z`EemI=bEmznebkq+GGLSlt5##QA_qV*NKA2VQ6TJr5wDAj=9sCHNB0j%!Io_ES1-Q z^J=JS84q^-eSU0cdarLz`R^7gq%vgd{#l9j>_lJ~l!@8pZjY8_l;QF2S7fU3w~HEQUJfwj7IdV*>Y@GFpR?oS5dd6L}T@SE?O1WSgAb z_JN{}qs8PbD;qYGk{b?aD3g@YoYF9a#(XkG@;e@u6D&^G>DlpV*HCk`N$8R5>^ z^rCf9jY}TSyl^F!>v2_MjVl*43>0=fE{oNY=7zpCJZYVbX=`ag@;+MqA-)wIv>UhJ z*a81HpB=I&lB4)($I|@t!TzhcVe05&;rvgXby1d$`{{*sAFD^t3kyIcw9@<*`CCWR zN%N~oYYqrpASFpg$qJdfA*T?c-)_X=uWzs`oa%+gJ4JS9VP40LsX{)qCfd->0j`Lt|Tjj>Nc(l3MlP)s%}q ztq^Mpys9E(vF-wn*dI-P6jy%s4J zoOE9lVuwwpsKO#L_qckciDt#h9eO87i#pDS_r=1|LNX;VXv6I3*~9wvv%lyMWIcDY zw;b!n--Pv=XJ4`l!#L`c5WQCv3sCAWhBZaKO|J4u9AIvACETY+U#{Y3z8ETVu=pd_#LOJVf$F z%!2rErgwa7S`@YnjMa9G0#ErkFIckRNwe2;Mx$2Txk$8I#HWXIwD%M#v84Hk!#eOf zgL^UHjN{{A4M+)J2T4c$%*^kh^lzywSsAPtxlW#IF8t^D>+^EV#?$n>mc4SeiB`K; zlHZM2a27{xIfKX18qJTkHi43vzmt17M>>Btju-KbSiN_nX&WYyXAJ7-$517@U9pc=qK`&vMA3c#=p$ZsgpUJ$k#gqIGUy0hz`- zV2^QFQX-tmviH}ZPyxe?(!_ksj{+c-FQFL3o4C_Y=?wD(aAXJ|%?OSuUerRWzas`0 zC@&7PAl+eNhpCf z@4zag=Wx8G4CK&$7$m(^q~YYIzyU-wClqpDP7%Jsw3p)awY2>_af2^^FHUWAQ#SuG zUEd7o2g=@1()|=DgrsUW`;=CId~*nEeJ4Vc@89<}W~{RTM@3nKhVuN<9=7NKc zr?xiV^j~&EzIF%P|wUgcKo zjm@eG4m0QPz6Obt{i8Nc3*O!7Gg3t>e}rwb>1w{Icrdw&DiJAV)CU&HgTCsDtlX5n z+IuvoXqLF@uWD8rQy7x?ZoRLj3feDSxNsp%2ZDZ{04ZZW)vdH+1^BbF4|v2rf!v99 z$&UzGd(J_B9jDTr{!Q_pG`LkOfZLY-<#^1s>a#xUoaNp!Q3|&BfARIs-+6Xz*Jzx^ zw$a#XY&J<_+s+l+wv)zIW81cU#daF2Np^aV?;U%;_xC*ehw~pe$2itpYaSC?=%@|> zVDMTgyX4b?P2ug}-h?!$pxZyUk0`Ko@o%Dyg&9F%#E5S)CK;L4K%z-@C-XFA54EcX z;$28Na^jzthOebZz_8pbmYTBbd*E_NT&TUxwC=pKdqCClcc16oy%CCgb-PW8?3|j* zV!Mg+QFVbzUfg~qltvJ-qZp7C1M-tMLX9N;9+hIZDHQS5v+f_#7T(nHt z(t##sR$OfK-L_QSF+f8$pltX?*%q^myLupfxr?aF%T_l|V_;fRcnhcoxiqGeNgy2* z&`Fu(_6yhnjd31eJrepp2Ap_xFYOsrDMi_U#RFovO4?aLrmxl0ixCB)< z+quxajm5&F@$d&zsxQLYdTfqpVZsWg088FaGw8}s4#yklBWmD5RJ~FNz8oF?NaoG$RtzserR-6tNJj zDg9`D%IweS&}P6AtoYL+Cp83Duss}vK9rrz80Hf{{jO3y*n`}GtUAC3&Ia~Ju&3_2 z27=LE^P+Gd4hnkemb9g9ZTOEoDfs*yP0_jFlb)K5{4Y4B$ymBgA~z6QV090R>`}6b zrj=g|%ajpb4qEDIE-^&M`~AsAKO>M^J8mG=7!sx@<3i`{&)zF=^34UDbWBi^Gb*u9 zolcAd6Q$s!6rK2o;~3|P>Nhgbl4)`)NV&ymMEN?bh(D{{~lI?=cpT)>v-axgS)q$UDCaSjcG0>|r4 z759<9jbyHIcKM33uGob(}>|zF{zjP%sxWS|#XJp6VetFR@zD?;vBc z9?m7~nct6Y1fmjXwo)LH8AxTu!$G-{lVk=(-+IsNgRa%{G`cbV9>YbmP?+9>goEil zNtH9U@Hrl;&0^9Jh$MUcXoLyOvW5ayU2UcsV*;7#@+_u}fRX1To!P@KM*!jseV`M; zaMrL3mNONek>?x+d(&ba%s(IHo1&4{g-w7>gG<)gsI-2-g~cPIetSko?E`+(G_FjI z&JpMu>odkm)zNjpATluV)W_EvN#op*(&13UTC0i6t+Pyn_hMcHGxqyx&C|c^d3%TK zCC;Dfu$|AF@&5{D{6CUh3p0Sj=X~45E64@@)rUQ!V?Fp$NRAw8lV2OupMGE{COW(s*8zunkV zI&ZY0lCuA52hx0j<2RB9(kEwE^IqG zr{yLf4KQlfpUm8UC14G1;gBH6 z3I&PjBSHOD+ea1Qg%JZ(a}wdK#I|6HfGQWMWDN${Z`_fRUN&7|@lq?OI|Fm}f^>j_XMJguF zY+~+SbCgTY7#SoK6u`I3Bj{+N;bY2vXaBxhS@}#t{~&{VJG*g>hQLJFiMQRmINLYN zu`9>|oeX0B$NA0v+T>u#bdEfz1A=IyisJs|?T_tCO?ufI={C8PbhBCbiYIeH8YbWO zx&W3K)tgwT?K_woGJS;Ubs@MNG{+grm5+QnhZ*qE9+Y>pgq<(6GLn0_L}Sg=M#_Ff zWpr_cVxUoWT(_VO`clWafc78}jkwSBOI7y~+CH@OM~}|QkirwwK--zeE_r*Y)|;9S zrHWu`YUzbk?}AN$ccT@9Hfn>9cbPYPi7qHiPE(YsdjpIG%|NSAiZ>`0Q%ln+>#tj( zi4RrZgE)iN5l%XusK8Xw6h~oN_X#`jm#;5~gLR+DW#i>%P%p%Fqir<=?w~_D4eiGA zOt>%fcgWpX)9!b4QgRF$y-QdS&%``22jWn>WqwQ~7M^%Wo}eRsa=%43Q;j%#*&EvF z@*>P5jStvzf5CKYVzp(a{0749ww#PI;~-S2KS)MXbi`+J7ueTOw$Md>Oyh>T3~Mj^ zkiX`*!hdP{t~eXaX7NHg!#UIf5O*~`p{km3t?PitaYwCG`)}Gj3`MJ#)t1b^GNQufLx@3_I#>($|Xk#BmG${JDm6MnItus{di@`5`}iqed~&Hx0j?dB#x105K%!xr#+1$iYK z@n>B4kock)YCOw`I zHa~?TQL5mYf*Bm{%B6@331w@Lc_ybr=9K8-IY4hXCJh8ej0TD7f$QjH?XjN&z94Xb zk-iTbH9HN5@ryA-g_caxq(A1k_q6uMCW*hDt+mI(YvKMq6~#Q`#(xg$nP z0lc>k4;zS5T8(!)W`*(U$ZXEY>Ep1khPPEvhwz$)iY+N3=Vcx|$mIK~mnMIetVxI{ zKRM3Iu`Qq4Q-5duQ=RU}SMY64(zI3PfhHav@x8?80;(U>d*(WbKfYx03+2OH?K>bw zOIJ`r!wpd>2 z-X`bigQ)6FD}^M=dmgmaKf8H>Q72n%Fi?aZz!3-W6Di;J8kgoBb$;>BM6v}N&pCa@ z4kb-$nVZ<)Y3}m%6D1>C&U%PNIq2JV;cW7Xeu7*|0L~houfuGsi1tx(j?dVKZH|6C zHYuupefCr-q92VFB~K6Yew1I-VZSo?pf+Zc3#gO}m-iKeG?&t1O8Rs^gMX4%%jqqSs7#0$vnYcTRK1ToGWb1qz*-olB(S^1vaO;Kp=qg zfXJ@sDuc9gRVBK8s{AtTrF9%r^(YCk3&mKEH+T{UYdQMoEv~}j66>W7AqwHwW!6a* zD0ji}eLSw1c5FOK-jeg=pH08i&Frmp)FI->##MgatBBDQMFvT9xM_*pKRT{OBAwVL zG_+i#e0xQyU9^#9{}?jZ*CSDnfb)N&X`&S$R4V8yQT4>bZ<08nHAQ(C8Pra=c#pPg zu|6Yz``xMynr`y`Ixbw@nZ@u_OsI@SE2c29YKzsGNpHcESB}H?9qb`A{@Sj?W)r<~VN%OvZfT8@C+T094jcIykrbCQ zKV-{sc}g6a`DnW76?f^GGAvjO+|IoFhS)BU#m*$;bWh1mcWd2n)j7gm&MM zmJ@<#c9VOr%`$2Q!xn0F2Mx-e*M~g}nrwybQz9ycM}%K@14c($NTgc5wMAdhiAT*z z6Umy&SpE0_6sXT|9miX-1n1(WD2(RT87rx0Z{r>NBC27rr+?{{a+eHUQeXBO2ArF&1Hv87u5MxM9=M9z$Q z9>%!igc&QgsMO6dl^2yNY4`~d?mSZa&SIcd45< zM2nMnDZ!4%32#?KbvCHzx0HHVJ_fv)diw`_+tnoUFuU(?D!C|^5Ut2DSR|Ck&Yo&g zpBd_uxvtKwLVJ5UvP1K3=e-M|@?3E4HmRSe8e2hFSYrT1A|`K_=B$0||Vh zz7?4igq(H~WxI!h<+hz8YQ5rp-l<{xF>9{?SN5cpHIL&%bD`dTwCQ75hX@&2Qm{k* z#%wwD)LB%oxW0plK=aO0fSE8CvALvbw`H0>;sU!(w;S+c&d);Tr9harXPq=u@L}2| zp7J4QDgWwO*>bfeH#$jf>l;zCf6iln#B^CBe5fkx>~dX=*Jcax`1_hIElSs~WC%uA zdWpgp(`*!u?z+(U=0mrmRao9=f(AYHQ~ke#n^d;>7mMFC+WV)JQWB3`Ifd+j1w19g<)J-SjLB&}iL^k!;40#;JuAZi3YL zoNT=hemG;S@gKcCj|TX!{7)Ev45-A4w1)J&zD0d1HJ7e(_K(aYbh#>Et^37j$!Hz2 zNi!czO9WW;MVNOL(LnTIVhZ`!xxYUwiCeOxa|cOF2nm-K#*qo z#SIkZwk|p_9IRbG!K>VhYrwt7R{9?CwHuxpTI;R$j!=C&OiViysL*z0c&xJabTU!% ze2o0ntuA|rbkN%KZW+7SI=H0AU!9`c3VC6LS7x-8bM(GpSxZm9`CDoEbit$kM(xtu zIXEU;T{<6!9H z#ADNtwi(d6SgRwY7z6$Ox^+{oE?di)t4rDfL=iAu(pE{wCFh(V24L73j5|kY#WD#% z@Xr3A-Bq$x9e^M9%H+Ixseed8H%SUA!V~{a10BVfPIs8guX?Wrv{Jz~#i}nRpTw!i zuLx{U8Rs0lLb)QM0ZuEpAvJNCMzV>EdE-?a!T4Ppw&!u#5fDSU3zQ(5O(U z6k>L%fZ3a;ph|U6JVUCN-NZ;O`BnSeUD!CGfD+H|OwNIED-qdVBK-Y_9K3m>Md&^k zNn836T^n`(`HFVsqc)Z|l0vb?>SZ0`A05f+lTDcr)?D^L&k48}FG)gp`HINGpB|Cu zL>`Pel;7;OXqxA<<5mE1hS&qn&o?j}ZHK-ayfjlDVNxJQMx)7b?rkB+W2Q>g0#U)tzOMgnW%1{+Dlkg0w;kMV)EXT@UL`Zg*Vyhr!XZhP>d{h$FVSIrnZW#M>@W`*Ij z6I|@BJ~>+8jGMOauKtI%j!!eQ0a)g_jrMU5B^B&VH=Pu^bR`obKCL?0xWaB`sV$w>o zjjnp(13I|0VpviR&*!AAZkMnvR}cQ*?byW;NT}XgJRBblaI|p5;rj4cMAXKu zJU=#MIl6VGX2m6yQG=+lP-w1o;- z6)eo*{ZTc!#T}`G@e8UcXq|wnT>d$(?D6U+c-1hJ37AHA?h+UlmfCT>`Zc;Du}fBW zu-@S^aI%E6-&$0d*4d8juEU={n-%cjd4PXPz=f$1orMVF?RAYAR(E!HUsWSgep=Q4 zG!@S~`NiGS!o4Y&tO6|3`$eQ8`vSe=Jxl+0JMR}~? z`C#Nv#{R6VZ+(>ovySy(T%m(GNER;;_pt1ZTzlmyEU7mJSEKu=s<{?5V&iaP!F%}% zDPKnw`Vmqrh@h05v`T(aAB8Ov^)B#bn3S~L4hJlRzVWCY>3`=o6flsop9+&>m!hco zp@`9;u<|;J=~MQAcoY=ewa~?tNUWxO!MFUCGwVUybVa$BIj(sO1Chsy=b;v3lTa4; zJ=?5WiUiCTLc?P!n@?hi-Dn}eKXY*R>@vPK)!XvNFFZs_ar(9rlceHKcc+yO>YC=} zoQ>ybfjD}1_>I&)?!lB-8mx(F+u9kAl9eoi}W9l!S%GdciG_4kSusph4&vnXAmud^xE}jPs zXkjR$efTEN_u@?s<8HNA5CITC_sX!$&-T~q-hVyO%k12=JwLA#7Rdiu6=!4MWa409 z@K0`-+J@t28=db_^-LjBgVHcb9LVqlpPxNL+W0blgSRQl=ea5QZ%cO|;(ZRu#a-Aj@a2(wvo-C`P$Z88aA3B!wM8TogAb_i- zaK{bf$)iXAtxvVD?2_k2pCbyiLR1K%6b{+^R!Xy+9%dCYnOjeOlI(~z^S84{*X)m$ z-=lDx?FJDJ@oibGBhp7-;|<7IgF1aT6R^MeQ@Ira#;<&aQQUmej zW!xH^!*Zb5LV~*+w0sgq21ukI5aWjFv{~IeBQ%6n7(NC_B_+4xj_6>O#W8-yi(GOQ z!hByisZdo8Ta54`_jMU3HAiaq?YAU0Pfs9wo_=}!8NT_aV!Ov-SjjLS{x-dMSWJgh ztg>7JnaXX3zI=I6@vS~@^bD4U^yWV2rC-RGiDMkV^F>%1aA#C#B^0Pee{nJ_iPl6N zw;ZTIk_Z#s6zo5{g6ARLVJ1A%40bUY!b-MD0Z)@8-RUntI33NlqC|%8n>vleU_5l* zB9N8RmkO&&Kk<_NjR+G?0~LVI8~-8CaC<*C1$zZp>=@=BA4|8k^#ye!S?i$ZKsN|KzX(lXSkN^tJ_rW zFd84`sawW{7kLpg0`9d-Yd#a9)XzVbt4CZWtKv<NDn4n?snwv+uZ>$OXgRC%;gcpAdJw#_w9j zJ_kNsc0<$I-V9y7%KE(Pn@_}}p7YPpBTDoQ$&}>8t0h*bruS;@&Dmb* z2Ld&7ziwz!RfP7UJh;M^>}e8bN}MVM0LghE-^usWm+IYNy-H5hrKhNCijpweS`t@Y zG>CeBpj}K+&kbU<1!S}*I;*lp;WhAhyYF049?FCB2ZBpnZ#Vw($9r@^H-U}GW(9qu zStP|0FINkwjnB{cofN1y4T@HKrrDL(wg?Tb}wJ1KHV$QkR*Ai`;t@IV&oF+cpe3CGrIelu0hn(K6?k7xU)w9mq zO7d7NAnV_bPZ5Jl9E4A8ks$hiUY6PbYynmVfA1L4ss9K-{I>qyF-Y1Kr5NELX(szY zvOqsY7wxw-?ZQ7C($$@JM_kOy7cT*Nl+<5q#ooH7HUl+bKo$D3neM)a+qHf7YzyoK<$53k8my0xHeY;z! zk&@)fowcks_IyVBEc|}QAyn3I|IQIfQu00ia8Ix7O0OgtYD3nigIp$lZMz|=WsTql z^CyncToGrkMn;$RMij5icz(gMJZ1`x{tXrsg%8y%R z^48bVEKM$U9j+icX^53HGRjEGMjvNfaj$h-n{M*6ikNxdq50!1eien;sDO@b0f*t8mOT`h6CDGdZtHz_g+pIC^GI>O8E9#x-Zf)RP5G zk{BmlZK<()HU83FS4)&~LcC3>#>|d>`AMo&48!pjecAp+=K*%2tOC~Sg&MEbdKFNC z&+g7Ake!*;tTLtjpdf@2{^p~*&+rNXyN@U8d6CTUS~qo((5JBxHtYFWjv0hY#|wS$ z%Pl1Sh#?(|k1~ZRh}k%HDiSOHtx2U9H{$0vYpTFtHP$H(L)H9uHJg~k+!f9&I(LQ8%+qM0(Cti8npiM93%7(4)f+Vx6F8bd>q7L5-v&B z6b~aTVpQu8{r3?MBk3mTCyj}OrGT`9oK>2j^-%_Gq?#3gi7j_{c)eq!cT!(ccqMz8 zV^`CP7A}audE7+HASr@1WyZxdkB|25Lp2?ruDO@F)#zQ5z|14!Ean4raun1OPE&tr zGoN_pxTP8*&6$l}ueFKy5$Q^HLQBPyyv8>b6kD*cal@(#B!flts3tKqE>65jVDdUP zhvmY@5M(1XfpVs9BeKO9hVO+`6Ic1X+Lwk1J%VGKwlsImAN9H}zmnAjH`2+2pl332JsEacxJ; z=HXtSiC$q^-I(vhdF#}0o9V~JpMO2mCY<#OEI*%V_n&r%|0L41adxmX|0~jrPWfla z`FA0$decKuJCZL5K1XQoU!|#5Hdl!RROUMLE0x0MidiyWI^YP`y z9(#6=r8yLSC{4|V8RJ;~dHun+HNAlPqf-cA+6F^02Wb7#z1G|HBF9&jm8{fFyLI4p zZy(aI)8daiwJo~b$|Nr)M}dY8OWHiD?Bny^c<(9)gCjIQRCTpDv@| zYCATdmO-OSUJ@b4a`byJaJL(UFoAe(j-h}r!zpjmIRPeNL6?hxr2R(&Mr{S{TC}bh z9p_@+-sPY8c>ZEc%#VyqJXFDj4J(AqK%}E3c{oBx69O ziZyo)$k@a#KNKh7H>*a=q^5Z)$k#MD?B7*s7vHya2*d1MjKkhxrIN!L*DeZ&unO`B zx_?0YPHF^(J3#jo{>1tE5?)=9>^-%AWq`!^((8rr;Fit4SR4k9c>>v0(8U+j*g<&@ zpL}rK%pKeo$6cof>GyN;LZ##TG+_l$tC1^OPb}&ft5@qm0O!Qr2Fqw`e|e$9)Y7N$ zJO|?5n^uijtMXFUllmd!MBqDzyWNa(d?*FP6b)zzxN!j23#1xHq?7F7 zyr*4|-rG0nFkig0+2wgP*{pgJTE(Hp<)GZr)o$JI{qIU1|23e|Bbs?HpIVXE&mxKc zG@!NyP8Kfz(7eeyv5A78H19J;qJuw$Bvs=ELpau}f=+=2p4J8&`y@zc3LTR%wklph z_r@~_w|rS>5c(k(BkNHID?#+AQ}7h+h~vg8?eu*y6(=lyC7uQpXT1W9pDjOt zNHvEeCP69@`3BX?sl8QgkLYu8l9B-#0V9UOlv!A06!ml(m6}k|%%5ZonFnt(AcCa7 z4|4-ktHyV~%{`zUnh zy_~so=^=bn6fsG6B{qYE%6Wb-<6R3TKPcLd8|CR|6eDo!SDaQg*fXKOzg-lI5Wy7%K^cj~Uv9~s*9JqkefEvjK9z#pHZ+i4W^_n$km)ttdKZ?I;FzZsgF2ops4N#J2 zern$53liln?45d5&rzEK-!8olU9|Ss)xdL6f((UP{^`rSL#U{qOJ|t{sD$#IMsm}1 zT$jE*dw%i&NtCby7Die|R&X}NC;^?!>7x=%wbVP`ifGw0%O=HI*+0yv)s-$^(0PYd zF=)#$;Y{H5wZ?2HewcUH&!x=M+TS1uc6()$e~E5aK8E#F4Xi%+6V6)Bt+ScN9(etGDHCvsVw6nxYF zt$z`im^Zn-wUzpYEqj>&()m$1KlphF4WjF=S-hloh{lu_|wA;I>XkQE~Q#C zV8eHcMY9VUmxAFE9?5sQAkJYP8(1cV5&Ywmw8>}O!^F7N?T@Z5iht*WWW7Z3>$D7wDPv}xbTa+ch!%cPjm*LN&!8dFo9g(*91o?f?m9Noo_+gC z^Y{>YH%>y`XmRdLGeJ`P59m4bu4}Qf3B;#@Yd86{jWFw|@hEF=qK&A~qy9Bv74h9S ztY>vyhZBksH;uO$IM6D(lIA^M;{b*_t4$XdH(Cw0%}a3qMN8}EiDev1r;!*=%;3Nm zEHkWaohI>LDwk10LsY9uz&FR=Q%n}Iia&$hll+EAPT@vzxU8ZTtkyLU#Uvk#NNaua zHLY3c_1i2dt-jCf8~hB69`df1lU+O)J?NULeJMVmvwtiOGbFS)v-qsJ|1-@3d3TT1 zA_27{I0zWG=}pqwq#+i{-94E6wAZ$L$PmwW*1qGY1NBzj}G0EBmr zMQ`U)nR)_uIihvwN)eH+pjL50XX!0dzj9yzk&JW?lKUNT>34;h9TxuFEce*;uI@d|^QJNde$hP}m#uWX&|ZJ**-Yf+kEN zsag{e<#C6H|mQyT{Wwa?MEIVOqw&c$-xEC@sRwN3mEQNuZSBtiLeYJDXY^Q!{<^u9#<;n-4#dgdQ^()z}hU0L0btT z|6KJW0}^yFsp7;{wDM~CQc;Xz*$F9yR9?lsBf8XwIY2#q`cfjPqNX3SK3IoWnruL? zM>*^R6tg|KwvdxNag33VoKS?CsvzEdj*L`L(LkRcb20;Bo57qQ0uY{Hm6#rTq8z%C zSIa>0zvB4gODS&eMcmR|KPIlM7{ z>HaIST7I4^wS`#Xi?+bj)$n|p+B`CTKAVKM9cz1K4iXX6yZK?@R}*4ke8U9K6nazx zdK}q8WspqcBzq>>Zb`7rDuow{-v~CkRG&2v0Wc21qf*rSu*Cfq=U|_JsLdo%z5Dxd ztVtn0tffsqlLX5B`3|utY-a{ij&?#JyH{RbHiB&*azC|LSe26=B6OEIaGSvm|qDp)MA$4-gsgUXRuC^ehon7o_MX(5f=`UJoCe@BU zwpx29_s3OBwHn0pMaXUM&)03-oJf83g6b}; zUKJKUJcW?L14s9qSSn%}c#c4M(YP9f!EW(K@}AV+_)gXQ<-{T&##UWc?N?2H+!wx} z_1F=xY8K9t7K$p5n$;i~eyv3vlpG6KhdIwzRRe4c?f#iQwR(OA5l0V7Vvbi!7tf~> zw6W9(HKAstCb<*Kd~{8%(|i$pU8NkLZ>6OnIpyT}-4%ik3v_wZbJ0ICL8=;4sn*%}$rV3VQP)2C=L1TXPYYlX2-|Cow)+CWkiQ*hUZygU5i%h|Rie z*@zW`2knIH!tJb<0D(h=LmVL6?VR0dI!ws0IU2x#?VL2#9wZ zD_dRg#lriw;voP1K%=RnTwe);XH)@Yw%=aYjK(FxVIW(fMMvcT}I`~W_b zD8SHf?10L#A$*B4;wXY%5tUYKT*nPwGK@s5d8QuzwDDO+u%ztncZX0ZolfrKB+TfI zd8kAzDlBfiA6T*br?A zZ;s8!m8(azH%7$yZT_q%n%44FwGgg>V+x^DtJD>))oWY@Kv7lrQxu`|o+q258^I1O zn+J8J9fgqqKh&!KWf=?_)}toiyEl&Q0k85*^My_G3meA;CbuPgSk0!kN}DhaGKE@O z1hWsJu=}W`adogb>+# zEF`kFg8Xk%f}^o{82}>S7OhkAR=`=9p+qrCIh;^OV`?)aCcW%_3v7IjF)de8sz`vy zY8e!YBEnn;B+*0UFW$#5;&VYYVm)N~fjKhXxYf#p_GsbB#Ta=@ezR+VD%tkE^#K_f zqT?begtbOr(+28q$C-q7We!kGB#S^buS+@D-5g}QL^r=ytr32Gp6sv+GZm{zaA`{Ns&nLmnrcs;1|=YsKZCqN`Sfl z>>|4T3_y{PXtDu5$+d_Bj8!269R)4LANI`*t; zUJhfIWDMa@DCu8-aip;+z!%M|hvAPkw~sCX0U0(pobbP30qg$*3vTFlT(Vz#L~&9S z&%PhC#gS7Ai8-9B6hDZh+o+#>CxYL6hdeSf+ep0q{K1M?dU9qFyG+ENARzWyNsrK|$-{m1Vbqele zw^9CoP{Hup)dE;5DXOi*G=&`4;pG{sIAC`iaP^;2f&KqQ1;)RqfDuCnVo`G&8rRZF z$h#>!ELWgKV(crNGpmBWvb1f%eJ^o4_}gHA9IIDl24Twje?TFj{P!@e)7OFSm=d4p z1Fuz~BgvYmZf`U?hf1vbw$AA$_WS1al#Qu_BAe)v<%0I_!#Afi$6-Z!Z##B>IyHrN z#7fMxT$Q$}m38j|Y~{Y_eEe~u$6^TlO{%`OH2M2iKGzwP4ml=;-27;b=BYCSxEo1= zc`(Ew*NmY(`3K>n$+1#AIjCck)`h4oyF^!lFBo=}k$5DVW>8X?@H5zhK@uvhe_Ux+ z4}+}=s9J_b#Rs)-$}}ZYK5=1G_`)0;sxv5!(YPXAxw{lh9;0DrcYhM(JoJ!R4r=p$ zin8#6z00;Yjs<$t#ij3g&(t^}9qm$urtu257dAPS!KwcfO6D+{iSQ$C~Q?D2EFnp$*4;~$0m|R z^&d=_Kdqc(EzIpsxi7=8((0tcxC>IKbyo&z(M@x}dVB#F<@c~H78SzN#51psy%R*7 z2({3{*Tv8}Wb*00W}ER^7TDH;wwQYZ!wKgn12* z6mVrXwvKP)qE*%u4{vGp!O3Ci8^^RIuj-H3HF-s$&gvjxXJQsYkWt0{&4T4e*wzpo zioG0rI^x-!4Y(Gqf9H0X89{jhx?>NS2VR* z>HPG~({!^50D^YtQ4H9F?LblBp4|q^JA03R2w_F8svdob1EJ+H@UVe4XRUBUP@)kE zMK!VfhnhJ;{$#S_B(?xj9=Gvg-LfDVf_E%-1iI^*Q#IQ^pb!mfIUmj5@Jp`w#Ak6v zsa9hldM(5hcg^nFT+G$!jsk}28N1Erb;YNSNyP`;mNbGOSv-mX;(ADblUF^8Eq>2Z z)F+{GLEaD%22rKlQdIKWh#egNIo7=UB}!BhqCupeL=Tie!^ON?(QR0wWE5NU7k1pJ zS;C6jD7b8l4(~*4_tGgc(+YAWHy58qr#Q}ecTpV0BLDA%8^ySCwj)1B+WS;i3%+(| zyMoVFd{O;A^$^oI@xwmvb|MWMZ2U@U=n9!W#->~B&E)N0hP4O4yN=W+^R@9`3~P1< zj!qVTsp9D5f3j6S3mMT>Tl5oIzDr4K7V0#}E~%EMmyRw)+I&aIBro-gXV#o;;Cqci z2uElDwuU7>oUyiwCE)k zW7AHA59AW_p;pyeNF#&d#L2@4(ecvQm>^0?$^1;al634^J8+;GzOt=0eH_PGLmpEy z=8SN(V%q-!&yhq<&nTZ{9K_PPh}kp3`pJ0VjO7G{kbN><*w;yku(^S{m<4k2UxHWdCXz@i7p}OX_v}a~(ETWQo`+zTGUCwN{FO@d`0>*H-dyEpyzP5pON(oDL(C=C034%=+0UN>9*xD8t1yX z261fen5KMX4Wyc2tZKDJp%9y#tovVw9moykb+iO5_`{C`CFPnaFk6A7tO!v$%>aL3 z66hrQ3fmBGgTulx9g-c?T1At$18PW_#H2ya-A{ z;Dn4+(J_fyQA&-6--GpH#FEM)MWrKk{cmrdjFz$WfSgWd!qoDvFw4GK0_T6xloBDFLD`jV>h%G>{tVH= zeE%UP+rOA;-a=M}m|VqM9Q$>#@>1h-VZI+L-=4QLp&?cQ{;1pa62JV{sHm;MHyXz~b+XFY^`D&B2ring2 zcU~a;IiXw(8@a2?at|7JMi$^rpfy~vi5q>ETz&W9O~=Ks1`0?0N1Mv}^= zW|l-krX|D$@=To%S9iz9gDSm=QNz6X2~O^$d%lF3c2%xn%43NK5adP|M^`a(Ma8 z%*2|IL^t^{j;5hXVG-p$S?;+7typ#t9tCL>ts-9!N{DJ}a|K7{hY}RXaAW~LbLb-6 zUCj%-rda)`#e#EZ_!(irgQ74^Bun`$sM}^JTG;~)O;EnxFE~AQA{mQJ)sm`P&q3^D9d;`3e-lZ^V3dbkWeb;ze zCQ2`aS!n~4`3sRzPZ;lHQ~tx=%F=c3`wX%6%eBTBlvo4H5VTq2hSbQsEC|#u z;BU5RqXI)9APhU!Df#O|=wiWJsS)KBQ}bjDVABN6_bo)aO1WfSS%z2GU{-T^+R zvMdzkFeS82`&}xU^OqffxaJcGXGWuzRJUpUFol9^GA8S~E9qv<^2GytFYof5zbXKCw5 z*U7mXFyb@#T)Ta}GExsovrcY+V7=|0%UW9it-Y;`m|9_|Nn`@rz(6;hAhV= zBF7Tv>@Rhe^hY$AC+caoCH3;xT`wI*r?)#c;LLN8qvrDBubs|XeZuE4rW|D`DOrVu zIm~WiW*nSrW6J8J6Mc>H#_wylKffa$uA7&yx1ue~cV&*8MRu6d)KMf%Qk-cWL0hBLSq3R8$@{1n?Pc>bHy{8-N4H~d6I%e^n+>vA$5*kkY!L-`WXx3Zy{ zgQoL4l`!Y)j5G-!WrPcr-XTEdisV#l^#P79o(t+wgG;@nyx5DqLI`M2$jAo0Z-0Iz z1}x^MPai0#lqgNR5q4e_fwGSKCCBz`o!#&dC@aBGu2Kak19wtphWcfqC51V+w_-!a zvKVIcgfIm}^y6G9M>Ea%y6(}3GHd~bRLCk$Pn)Vaz$vt}+t{-7PJo_*+|whMpD9Ui zA?~+^4W3_olQTRq+ystLx)lvT16$|V&q~Ma9|9*6tHQe59m9`ruY(>T^k+RE^+&Cw z?xmeR8dp9-3fphZuv-RkJN!yl%%YZq1}{U($ar?_J4;`RFmCQnUvFO8kI8@2`TQ4& zL1_3CWa`KC;_7Ez|3DT1Hm=5&hE@*twEv&!g`M3$J_?EQa^VX&5hOtC#2x?|=!pHiwYt^f`Y~HoL z4yZg&dJsN1e*8+%?Nj#bVZ$J%T3`tY3a{d|p0@ zfl-)QQmDx~x#AjuQMr~Xv13`>5E#ib0ie8@n*pHVEvi*sNn2zi)VUnJvQ4Q~eARVi zcRUx|#Gl#c7@VeFm`_F_K_RHtNBP2M)d($qP)te37wJp>$oRoS=~wBZIciMfDgp8% zr~J;i6GY@LIA3Do#R5X7M>cfksD!DbNky3mbR3j^e!;s9QW;{&JMhR-3S5bW*R?vpDPfme%R7XUj$tOG<{S zb=y#Ex@;fSt!bHnm!S?CW--3+Y%PSY*`EE^RGzY1Adu;2PM7n8TK%UAu-E;^+IlhJ zznlR7;jrXC!Nu?=NG)zsAOvs9>wui)P5xsAm@bJKz7l2OjL(0wp2}*P&g{N3MwY!l zyuZv4A0bQ#J5uIGDXWf%p^+@gVwCYcJ;TWo!;EEPN>TME*b(~F@eF9pZ!`icQCVXW zSZAto4xcJf$c3!s7kfb+O$aXuwJL}?FlZ`*QfL4hH2#MAHe6n9MRdWApRrY?^7AOY zj+yDI4B1N)*=d8ORll-zTxr=-B(3l37yKSsn+LqeZNsV$H5|8tD?qp;unyer+kXFg z`#YmWe*&j~DV=ZngA&Krms6K&(eIU+H#YSl2}C!-vGV3w@eD(hmXd@f%eN_j@~-?1 zpzE5iTnC(wb@j;f{+QC$`qhrJaqJ1w5^$Xy$Bv)5W9cW5l5 z0KB+%SmgFBL z%YQPPu(!6*wKH`y{D%XwnXIL-{{Jdt86W!S($fa;@oy)zXLxZ~%4rPoM(&M-yP6q(mR&bmez*(ewxrRlZ^Ja^21L21x)MS4pwVgC zZPLED(AB>>)-QI#>}EEIYu`tRK)iz!$ipCt{wU$A%&$=}$cjX?HUDNWDCm4y$)c6M zMA7f4z=%ulQ?QntcS$ZqAC;J#_O#rB+y(mgKwTQ?8BY^QJxQ?&tGv=P+C>98s&#otEzoZjoJ}sd6@G;yR>k9 z(ZZkBB-g^N8zL|!e;U>z*OJNVC#+}l8X#rkl$TP}`c-abE zW`TsvRt)t5|M1F!K#K9Y>?N@ciIWiVblOv)$~m6%dwPG_4(o|#CkN;8^r zDE!Q}U%CzVhcv^vlsRl>AIdH<-`3plndmI-$a0j!OIrbr|0dE)M?}^^f71V@Mo-c` zW}eSWff_s(!4ofATNMxy#^i*L3gN4gqLP`<0#QAn6o|B($W@WQqSvcUWGApL@T_r? zPsDbo-e7ofMB%u`_>Pu-xC8S>RSG1ma(pUXTt35BuCe2(-Mk}`)|sFGZ&$#2>&9K1 zpYDn0r+dQw58ac!gRb3AP@w65sr@VE<;?qjnyBk))-wT?6LENws+!=^LFf#a^Q<8< zxoV6gk~jlcLyK-tG4=8i_jTks?XLH$tM>BicO0~JrF4W=<=8o7fp-FVBoM7!fA25G z>D^RA-#lVNK|houzme2^|8X=xU2%Xzx^;YOxV~k&T#3*|+{Mu~bqfl!eIk#RL}MFos)sLV_E9h{^Q zhin`D8ZnLYj;sF8{?%x5kkT=f&f28XvD+7C*+HHeap}rZ#=W|ie4S9v3*P~~-5KVS zrQ$mie$o~dxL~=q>~uQWD|TsonYDh>*Hhfk#)9LuL`o9&`kks}*5;QITG(dRoWp@Bgx8aG<&H5cz3yPJd3I|I1(Hp!?rs4MzpJ zpL_whA75J5KrrUeyrcp~zB2)Vd8@MoMTurE3q*PwBJGIChD?rFEAl&aRV_nUb6gb|D9u5d znMfISDfamdu)huACW4lElNE>T5=IJs<>LwI#FW+{h|-Op^;THoAYw1-8h zD99fwv#7yYG^bg{nP^mA^T$=*&{HWAEjJ8IOEvml55laVOp?2#7GL1Nv9w|l?nT{=ugqc%+ zhBR)A!ooV_!}00ws}HXamzFz!52sjW;~B3zZSc29zlF5(U`QXZ1#`7ffjxH`9kKm3 zB4y}T8U|a2P#X!$i`JYYJKuW!lLCx*fd`wvsB%&yez2n%5Py@Z1=fN~Ddw3>;#PV` zmr0A(s{61GR~*^3QY~qp6eisrg-~U*f4-_=^Q~Nx5PAa7ogTbjbAGP7^XdLJUPh`^ zrLe+^_YL#)uY{*QO%@&Ep9+Ndhcxt`H;bPN{O`eVrowpDkM;KEF{SV%0bZe8+E#-S z5sK`XfM#O|iYh?ag~1vI7Dt_y76tHj)P4bRfo1@jkgLlS7VdN^6etj^g<`R1Xy1uw zI^ysfAJnQ{6AxPQx2?AejE#nXu9gZAVw)c1?;O&0y-+~tK&-2jN!{KEMR=-0K02_X zUTuEn_5{&@UxdB{a3h9lwheT}T!BrFngW=&YWx zR-Rlrh<}7NzKMy7mOO3%gxxu;K-#;f(q#kz#N?Oo$8{Baxzvi17Xqqe#7f|}nPrEy z1t#6f$P{{KkI!I6vMNN)jhG*1oYU)85XbA(oPMyr<_Y#g5D?s}9#f%9Gvk~#&%vqd zMl(}_`ljUJOTONBj`!Y4I81%m48cJxZ6`LgeNRJKq&5O<4g%;c0qsnP=J(qh#@LRO z^etgMxUJxQzwcCl>bGMTguqTST1cQzxbJN>iG%>TQzMGT$ES;J@ioP-si4V@{=OML zWz8I1zA>Q*Y?R8lJSnpX{tSyn*YoDN*>__uCh< zXHLQQ{%z8#;5+e%Ik$v-Lc-Phs9=pJh=2`Qin^s0V?CQl;^(>yDGZt-;*k8Bo0=7V;Z zn6%1UHb~n3F-qlKH=g}8zIiqd*sN&JK(raqd%`tnq-$b5Lo7>W;& z=JMvmktSflsjvT|A2nUVUmueVH{Bi+VU(Ymj4Pv>m7E1D@#(?Gjk;kq2Xe1(CGCMC zBfd(2F&G=){dO*0tiX=#jy|+M`Z6&NYkX{sI;F!?K*A^$Dp}z@Bg*X#HU^=@8-}&l zjSzahaAk{?|ImJ!#b+rthAJV20@-BWu(|L-c)-fX5<#Xxoj@h%A@4>%9qQ=WSArM2 z?k@|B2z@@Gy8IDs?qHdOa^kJbPq-yM5^00f3dr}s?B5{+8GDBu(>TGzb_4-`B};l( zs`+E@fd+O|%a8*rjBu|>iG!Rbu`95vYy}{*Ov+ZZ5sb@11ljB}p~MV#p&(suHU7Pg z2Bb8wXs3kre>BE6PSo3sm2=UQuE2pFOuLTA+a8-_lxsFckqNI@So$*GP>h?}$9d(b z7k6Mi1fp)`vPBhXQAraoO|tF->Dst9MzHOt44RaN){hR#u;ONITyDorSTQguQye+! z&2r2c@c>@$nupihvtfaP#n^yi^)vo*YqmdccWZ5axU|%kh4U$7<(YA$)JI$BaZRiR zR5O%%Kvqf3EHQloR|j*y`jVa^TKjO@o?XRb7DhmRo%WrI)J8@0C#IdnSNnVosAYfD z9$?%>&=+jv6Q22DeD}1HFuxygbSX1(v9r6m@o#&(5&WbhxiRZ%;JbhKsZa@B#(nnd zzdsLv*bgrke~XloKi324Mtr0`SWKR-iabqoJ94&qa>qx>(0CQXUQ8cMYq6i{i3`*w z@EeN3r>EEyGdnTa5z|W@f@WVUs6NAMJERW5dS^=G#wPh7x^-@*b#nOya^Sxgih6lk z*Z;MK{vONf0*nFY0xK`2MN$%T%UQ@W)Y@TMn)wn}d*YcU=uF<4jXLfkG=jZ7&?5JbRek zE}0T7vR#!>J-67=lm1~rw^bHI#?`a-0ihOm4vkjpWt2nx^1cxWMvM1AX zHHl^Z`9Uv!7|8H5cgI}*0YYXNrt$^?hrtp~fF!7X-|LUgN&0IXha7ZNSQO2MaO=bt3^mhq(aXpPPS~Xh<(>>a zSe&e=NbOU`Wvr2NbvY^7vE^)t%msa&7!*-vGEYM7Mm0(7AB%;typgt=JS97+$6crE z__V*H&}*$Sq!$+)4HnD=7xYT@X-TC*$jb4nJc!?Y%vLbL%tfWF@eeUYC6(wS{#*}V zu`Au{!*~03z^9#C!eK0`R!WSTx3#AsA*8U4YN?xJS0f`W{GbA6UFRGPd%2FsTAjQ@3CSdMImtOP*L4zQQ&CX zGw#)AG;5pHG?X)9i22}NgRe!vOUkg}Tv zY!Z_~s)P{AONzL%xxX@ih#L(x3cV?2ZUb+raO9G(>)pC*SE zr5-w7#BjM}(KRzXESwZ^De^4>o|oRj;(j-BE}&pPaTB&J20_!gXb~>2qIK;T=7NZT z^K3a{O|=9c?vI-*E(ZpLZAgon@pv1%MD+HkJRdI*!NKH!OWlmb#c=s zE2?X~oH<%BxU{{_8t9w=Jhge#ZJ#72v_e3LNT6lq>l2#CL5YhQNG|pskI47Gno$PW z#e{#<8J$17oISwkQCw&Cj^#=>lW zS%B^N!0ksRgaQ!z18w3WydsVq^7B2XYEb%x93rbw{?cS>rn~Q$XJN2z@~N4dQXW?2 z#S_COgWKI3KRUYdx?HkCz|d9J0Z2YScl*2_E$B`x^Y!H9LHSG4zkd1gt_-YGwm7{`6^6* zN^ODSGKTn%K?Nlz4=#L_eFT}9TOVl`GGpWkSTTrJ&DAj;SZ4M0^l>Jh*+S6=2?Mak#gljX za9~5Nrpii{W1|}tGfj6y`Fl$$&~A?iIu;g}+3vQ7qM0n3hR50J00jh+&A1YZ@+_!4 zBQ^YxKXuc&nDl6nfo_B+fieE@Nqr%X8|USy)-eN{r&y=j{#D~Q zGliGrv&??lKQ_MmQz8wpMeSNt&<VJxR|EIYB(@4X~)bgjcJ(d2;-|Q!?vqc;i zQa&UPMp(lXc73fDqCu&_0@jCOxz)Ua&Gvv56WQqD&H}hDaiH0L}?)=Jg|v zwU~fw#E&I-{Ebx7Z|7l?D8+-UhoX-JnvAFx44ojudtGzeQ|sLjV~UIfeQH;)lkB;B zLzBjOSt6f6PzY_>V|WpSJz9RC6Tx?|GzM~MFdK(AlvL8J$C$GN9No=ELhqsBpasA+ zjm&WuhpG^1&L3A8dE@97=5f0niCAuMh*VOR{;XGE0Lh3vmwmj5wbwS|GRdZEvk%<^ zce{Qi1pRvfGa;VJ-|aN=)r*70xJ7AvLLD?UJkzcl^xI9l7b#cO?@RkABvF5r4sUe^{veYGfuTBmOn9m)m zh2LR;WYnz~Z}RBg4IGG2>AvgGzIP?jMzL|B-Uy5}xB2}6!K%<7?obCCHX{(Uo<0}A4Ios zIv5X4?h?IZ3dWdOMKZg`mC~Nc@=qPie60$+F2c=QKTp7}Q7_SJSeNAq1h}LdV}tI8 zU;o7_bOq+gnEXK>&3+)QA4%>1n%z46Gr03ikhA{55r((#$)&FZ_#k676d4I_0SfiO zo0JTCfO1B}sv?<4f97@y0Xy~6wd;S_sGoHPr>4K(-)Ug*i2yORx-Dy%*k_ASOX4{$ zlfI(}?9S1x*?o6(T8@i?#-X}wJM{tS>&sgL(3`7r@Uuf76Ymp)nG4e7f1&eiBP`MC=eZVJNsHcRpY9cxzTw)gfCH_(4MN_Z<;E zDb5OylCYwxuFIpF&q@RKjA&TE_K;p&7nbR@=s^!H*)j0eZkD*TN9q};9XPN3l<$xIVM+iabu^?eDwwt|ZvXd44jXq+IA5bRK=yh5&4^h>H#>M?pt z0e5vfXu}sc?0CRj;Y$fw0Tk z%Fed}@+p=->@4)|*QApW_Du1N(){A?uU6fsaGoOb1gMN}8I~SwNYcUOCvAtLk{7~r znx5G)=!!Dg!ub39-1M*SJ5o1*|M|TyZ$v6@^0UqD`~XV->3iQr-`e26NTmsX|4BjU z`KJVa;!?%);XoT=p8&`=%kX*gH%TlJnMRR2*nb3;KcrI6CpMUd($Rufd^?YBj_tqB z?ATxr6i@3=_Qam0Pf&&5DK|)w2Quq+5hhBtc&5VW@Hwm{*nOMH;*T)$9i+qxi)uk$ zm2kuzE5F+|OMmc5ls-JPtUx|ipUJHhRh*1pfkgMb1br`0N1uE9W7U&%X*x82{6^2} zlBTQnPd%6d-Dw^s!(>T3i}U^MuRx|NZhVYDP|KdS0g=3WKEOSCK4oa+VULIeWZE;7 zVv3PLCYy0219$rTt%oR6l2<_=uk(2M6p2&h3Y#ZvGal3EnzZ{hUe3~hQ`rgkZV@87V?YJ?Y(So4(k;dr^_=XY+Si|qXi!`W2hiU zQaEn%$C#3ARW3c2(DwNBqWfp5vAK=N4MEz_Uqm6Q-RKNnTb4REK1SlCZs5D zDm};1SocAcc>2~o(s+~CO1dg;Go=Ldhu2;%krI%t(tSjOTaAdhfBfzo2+%OJ=4-!6 z4F@A|Ni9Ay#iMKW1mi$j!w!lr5fQ_irIjio4WUmXG$?W6T+DBp3`G&gLqJ=_tCy!M zpSmE(YTDR^Fu^QhJxNveqBq})Zu-T>H`zZwm#GZn#qnRPrOp`*Q~K-6)et%l`a3S?NqY?%R0wC(f=oV#%wcQB z>*>x&rRh)-5K8PD7j`;w_|vO9ctYti3?ZOuU;UbO;kFyEP8g#2^Q6m7x+8fTE|#BC z)BL`tke_LlAL68*1>W5Xq=r|cvfS+t7I%G_8B$asMUZYWBr2UlLHmMso z%4|s-1R7}n)|KCgS+Q#E;%|peZ^3+`)DLEHVf}>5m@Pmz$Xk5qmIzsBpRLcJbKPwM&v@@l&^>qae1!^nGrMc~?i%UnVvrc%L z2b4cW@v{aF(hk-vP0wtaD!gb%)OKOUqg1VuT_#(VcfrFR-`6vvRa?J-BaSv0Zl_Q1x#E zEUmMdMvRR?tu9?5_?!^0F^8Dmz45%ta&zzO8m=4ID%A3K>q-xKfXg~Lh$in`y_Udt zOEmPX=nctgqUAjT+Wz@$go29aZyv&K+S2Vk^9Md#6Ue~O()WC8Sn&_6j-#> z=oRv3>EQ#iWKpSgNz^VR%5>OqKEjm+@k*PRWVEg3`6oMTNB!0R(E=dnbn7Mc#9` z>$x6c7bqs0K%fP?8U3mKyAd?AwQ+1o9(?+(ap(_o&a=qnfBe&n%H-HFJ5@PeVxICg zmYEnP>%nY(Ls<}FyZf({+!byaMddj@THUlYlB ziyQ$>?04jN1M@?9gBjPEG6>{Xsy;}8WV&3KU}wzi75YB=po)IWDqw(EiU^0Ailqc9Z00G@b z23mZn_MA3=a8AuY71=hNPLAh;7A zZ1WQz>|Ma^UEpcaT-Dq+{R@T4f1AA+*VQU5-6-6kf|0gw3@uwThbm?%Bv#S^fAcDH zeXNJjRM%c9*`*V8i5$GD;`2ct90Z zD82S+`|R1`t&$A|dJ{nf-%)Sb+c){h4c z-Ibh$6mMSrCXb8nx78GuXu;ABl8OJ+tr}1hwgt5oq)qMCqVR+cijXyV9f%kV$po$} zL(Y-jj%V|4JtY<-&d0{XAih#K$;l);)#^ri7zu(pjNenhHF$g)iU&F9*8AkA5`Moi z70d$pmXc;wrb{F-QkanjB5cQsm9x_3*O{y-Lk0$=sm=k&Ky(aEN`2)REt%(*6~Y(| z_A;N+ZT53Q3r7TjQ1{2>BvJJvzK3i$zOV$o4mm+dfN#>8LUrmz%FiZk8emM(`M$%{ zYYX+DJU8aLI=Bq2*Yyo{J4~&X57NNXateG9sB5$kJj_mhlZSzdvyfncJ+y817c5_P z!kWb_OykPHsA11*gZ~HvsH90w-RP5cN^{du`1f@#ThbEZ#maZYKN#QsWjoAYR*k3p zvvsKaT>mH6`+sjCDxyL{lE0@#r6y;gC8#N;CZ>Pdg$0HkJDEudYH_Mj+B*3WQ7W;i zk4(5-=la*i+VqFWO5@<-5H3A6O*u6!7D+2jGe$*5 z9(71jA~`KJNjX{}u2iGsV;A-qXRlyKe=lEYhsbc>@~<1W%aUfAAp!u1{@nHd|Hc+S zOb;u2Lwj02B}Gvgd4)vv@wk0fr_E)Rp8de%+_BSUj*F+X2Dq4!H^MS4|%zd?m+d?AgWG)D{BYch=*L-;?K3_~Pa% zj2CJW=0c+`sdawib9>pMt7S%nxer0mjLLRTn?YXhUMlO7i7GF*GAh;DQm^iXY7|cd zDjd1c`|34?R4Xa#MEcumY97(v=$}rBC3`Yg)Sg-+any>V@scMh+qQ3?;w;_X(}crD=#tUm#h^C#<|3XkAW*s|5Yk$*LYky`9cIbFUk=89& zuU4)>CWWJRoIPJi-^4wtlo4P~ntBr4yH(Xx$49b%1^dN{w)pz%*rCcK6(4%hsCiq` zSDA{Tco43rpZfJjHgruW_ClXq+gU|WvGlEK*(>Y=*eFGB3dQ4Sgd`{9X!Ny%l6qlx zr?MV%!t4o-XL=5P{D#hAJ*3T(1U@%zE!!b<+tNgDpY`K*`m7s=ZU0Z8mvCDTyy1vCX<&s>-B$TK~0{xg>7tUh5O zv&)VFLhP*_g7fgAUNkIBK#<61eb~B?gK~B>LZhNOJw@yNn;;tE;HE;JB+%c5s0|oO zO1ph$B@)MEfMHlltBL{u@s`Qqr^)>|3J98{o_f63mLL#qihHaS9lHZoHp!&A@)j@1 z4GbMmN{M_Y1G*z4%Dapm0x3ldR+^?^Sl({=9YkJQ@0C{NQT3yq6c>m6@<z_XX&g ztr{+R=bW&<UisX0CR{I4M*xzkwIb zylpo#J;4l5L=FJi2($imeo5o`Cv6jA`LGL_wTUEs3OBqI9-Qy1zKt6~tf52ga0IUn8= zRMbi2Yr&`I`wWFQqBTV8q~ub|b9f03EMdULv!R_;dm1rRgN0V~<6GA4l zq+TFo2s>xn-;f5T$#_u6`DHwnRi#r9mT5Vs1-ihEs$5et*VZ+%>|13EXczW-=~2gN zez&>`?kZQ6aJ$v%6~VaoDcdC~La%@S?<2!=z=PABlF#KzkF1IiAD~|o_0L-kP9|8a zqqk74l%3O@ZNeuk$cCiaGEw;Lj&{Gi! z{3LuMZCJ{ym-|d_ir}#I7ue8g`1i*ryXO(Ru zt+dD>U6QA-D2a9Y4qxd7?FXUz7Vq7tTVyCnfkvWgnz+mdXtH}@Kd*1@I4<~^e06=! z#<6sd$ZRVStt0zPN`Ju7c~HA5`&WX-@}Z+)QD5*oIM1~F9felAP%n-PZT5Y4d~ofC z@?7o(yP^wa;0h4zBPel0uD5Lyd@I4*U>*lC=m0`LEL);DXE4DPH59!{n#5;Cu4;~x z;WxjgkP38&u+DXYIiC`87vR&y*a?-$D>Z&Z&9?PU;W+yZ-Jb9#Iv)ZVzq`79vqQ$P zc#%(+w^tS}EO#I-NXt*UOh$QldCv0;(F|!V=suE*+(lu+%7w1`g--4T;V|lmEf$X$ zXFpTlj}ob=XVoc%WOHcmkD{WPitZJZ8TP`gH?Yku}$f`p!$iN9E-K;uH_@u{Y1 z`l2Awu)j5uX)fHs+X4OXitp$bt-xBH?$QABIhylImg!qpmNEq9NF3-CaR?3x=y0di z#2jn(1JE}ScXc7>znni+6EH7Q?-q_?40Z>jA#3tVEpFp|g>_Q9< zee6i+raEAv#kE+6SyC~l?@v$V%`w9Qprh9e(zk44c{|yn8**RDu$2vhq6DmgHlBEN z0;qv%dIvB0!T=GXqK1Jka~|>aJAq*(-IOE_!^#?gsh;zrgJHsk9=gL{WUP2YNv+%a z4CysC@w(lf>wRT$;%gM}^OZ#&wL={;mqQ9d=xLMLlwa@U9=Bb12;TGI3VRP;RyT(I zZq(B!whO+s&$WM&YPMABm5UM3wn#BRoS|zJ(+3e`>XLvIOKK~qo$CsEzvs;eS)fL% zZp1Aj1WSsC35YlZSpBn=!E~JU55 zQWd%eup{FZsrYPJ3X@kOe32g2R~BWz%Qx5@w+YyUNgQ{4AUOO>eWnCd&cU5g8esaC z=5i7WFe)qmIjq){(<4O2TgSg-I4j7m)5$G(udW5~n0UR2Dk-m@iFSD8sLEI9`C)~J zQdHzDr?VokqZP3|ldjJ_2B6e|Twk+}HDK(3Wc+rzX%2f`lge5<_>dJC7_a9!!<>=G zi= zsq*SNtI43W{JA!BX6xz}NSL5SbH?WJaCSqrYC%&SQONIGsM-0&?wRR_Buz$5lb5iY zJp-G_G~?$&P1lNy9L#fRfPgi8lxK?Op=eVAjP-``fQAO4MuB?KPHa4)YIMkEW{Ag) zDq}MaiIiN6N#!?NSlIG$$E6xxyL6&^JDy6rueb0V-d{I#uj}70qsx+Az?kspXbKiw zG1~dqbl^d&o@2QyN5ah3ts+j>|22pZSSn?@nKB&clKBQ4DSYTj zWZwmOrUo9ujkGBUUM=6^h>R|K+N9+HMJ&zG!N;YraFmGa$0@FGo3Wr-@~U~wJxqs1%n}t94xI zobCNIna;>HZ&gAsC5q_+kG%}~IY%f;`aq$?WB9_@3Rc0I4MiFIeQO2h zyl)bHLia1m75gok?_pM-cTx(`#0^kqn@OS90X;mg=ITE06qxN3_AHLH(%?F9|Hv%a zjuMZii0L!zr=vOz&oA-DCPQT%Rny#!=WwO309N3KkT8vROI+Agdfu zdv7H_J@=)HN0bXg%Si$eZ!d5Qd%ne0JcGvLz4OAXF40Bd6lm6a3pA^;=Ud_IS#VWS z=ez@(z1c&uPMDXF5LDcrR#tCV?L>VVEglLZwYDUJi^I21Z*fDUWODVxMV$In@iLOJ z5iD_+J|!aL8R?5cAgI8DP>U4lw9kQX&r5v3=cs1MYwMNL8O-xwM)Dw%UNweaM1V;{ z8ilKgRV)?}z_pSMQCUlGEnYS9$T}CN#e9bZYfyrP*jSAP5jl*G3+S>cnR?|eY?4`t zI_*`Vv`I6I^%ABNl9NG!U_KkI`~&oiS|AkgT-?xWU#}CHZo1O`UMujcgM+26#fiZh z3x1E_-YbsypzDt=Qfvb(FeJ0+8e(#LD5U#KG596hwgI%KqUxUWXrmx0lsoC zK^nscr9{J3gDNCq<{=zA&}ljcZ;cWJcUu{NVdRQBe|M$j${F*5GxfZ7;!SWM9ChJ-JFxz3A#=QRV=nB@`aK#beP5mMUV zszmHlObsOgftI^&8d51?-7>A@rOC$CV$mH$G7Ev4z~6bn zjRp#k?Cnuznj;6k0k8HwP-q~W`V!^Hfj2D<@N`VOc0{TxYHN1b?s#%0Tx$wki-XTq zQBbsy2(_3K`<;r7lip3|!V8N%&>P;qwvz24{PUR`#B6Q6sI}>P?F*mfauq9hooU7YqkDX~73M*)D{p;)PeA zXO6lHlxodQ*WB;+xN04jHx0+`g+KE1vzNJ??rT(O-s*AcAx``>pROEcH>ZjM`a~je z83n_RHjdk}TyWYls?*cccI0E-^r=RBN+nYB_*~;OULT3o2e5ZKeh0-eCiTb%X1b)+ zkh+>h%u(~awlpX*t4{}yd`Z8WNqs- z)I<#nBEfyOu=VXCJih*1$vgdJwIIQ|;5g5Sa=l?_lApcZSsJ1uLyBsHU~yqiJEDuv z?qk^Z1*Rmo#`9cI8+MqUF_7lEnl1aEsE+u1&`WSWn~D zXr%%Kc%0@2WpzkNI^rVG48#x~9-|K;?%kfjL`mu!_+}@&P+`3*~{Ogu8|zgfw%b`Ox%-Y zh9}k6s^A?uop<75n2Xz&`P*WIZ&%pQzKW+NfYkhTLHFO5a#m}UR0zXbL5ypAp>U0w zw_~N9w(j=Q6SJ=6je%y2jI{+PXiAOEK~uN#s&7lPhz0udGoFmasiPY)zK`d)-n*cG zs#X;`)f56(Q1ktle#lXyRQ_BW@5ifvtjg;u~1dSN?M$^X{QLQ^18ZEX5zC#xf9 z^_PfqfeGeK`IDEnN4)Z(a0f6Pl)`{|GeQj$e637`UHM;!YhJVDCB%KkKwylP3xJ*_ zbpboOh8%p2Q}N{Na^B1n#qs8xdw*}Uuu9<%GQBSSU6To3n375p@%nl^lMsozTDNBD zcDRMa7by*7uI@Al5AtQU(q;%qDqp*1%|xGJiru)pNt3M9f@fb8?$m31Ur4@^)yrZ4 zLNNzrW?{6mP~|y`_UIBoq)X^{%mkkd=3V-K7<%hDo5qYoK-(&&K%=PElG-H2F~aH zglkxXS<4S5Ht9dyv3}pzm_4_NNAMp6mPvQAYm|Y73BcQ_*Sb=?ZX#Q^(@UJT-l(hX z0g+DAkKM&juzzoWz=05s_kVT9S-v{s|37!^WF&<}<&;HhlqO@pdgVQ@)MC8fRi)JN zJT2rYn1+STUelr#cdlUS6z8)CIpuMUt6h(s!*osEOyC zUGIW6b-P@6S$zO$v;OCX&S?c~*Cyt#X6^yqxN^o|(@TVX2-m0dx!F4Ak1e)f0$`H{ zJ2s(7_gl{P1RIO48c)LeMQh#M1e=ZLdHF6r=7Td>p(CSDb-~`jkC9^bujn7mnXJSZ_y*t4VfvL%qa(&EkF|=)o^*}D0 z?r1?vQmLp|Q{h;f>Lv6w;(R(lNq?>XmO$aLD}XS$2oXil0T-Eo=_+BMYRJ zv>sIxWFF4{+dikWoY}7wPIpwDu5+7Q0aSoSAFM=|mJnl$8G;UwG{imisBxc&PRo?k zRL_iZ3_RU5peaw-v%nZcH?5BZ2E()?RC33m6(z^z?>$W93r- z2?|Wc0~s_k8fMTzO$X7lb}HLoG2M%Q9g>>ke-A@1pg|e$1Y`|~g6V@yF7b@?=ngN^ z`g&0?HL&*n;wAsjPT&V?{gN3s2#7rJ-{~m-?>%!RQDJ!{5$EpTb460E-`9GlNj8Hx zV%GbD1SAHa=ZUX~x=@c!7__`T&81I`-DoNcNyNx9=A)U8CsaC1RdwP9Kd-yGro^XT z{nqbUP56S1KJx%sPyyINv~n2zzeNvFk`l$UX#=g)LZp1Pi8a+OC&IIdP{=>?9{l=f z`0M=ov;sS-C&xBwW3RU8m@2vZibkd$^Kw4H^0hK|$UB?z!?F#6p9rh#BgeX8d-LPf zh!0yrFDNLtvxv^a#4~fZC%`J|3LdpRybQ)}X08DCfd**9J$;#;7GprhF?r3ZEMr0~ zOIlFS4C3_!xw`ZF=8^9|l~`3x&F>w&Fes4$y%?znhggsL3bD*BV&_Mw{>z9PBPKZ( zSmHm3ffyd36g=Z*)liQ) z{K5|quE+@Rct7%xv0h(BnDu<)Z+0fU{OiMi(KUfSv|NDT(yEgWaZ&oP*_&NAL`t31 zfXgw$g6RV^*%aUEHq&M`7+8g7`U$$xTWrHyuV9|kk+jzWqraylQJs2s>gK=Gf`Ysg z*0*HOb$tZlT*A%hV+lKscFih@x;cd)N?@(7I*^gh2Q_weKg|U`@c&v#9n`yzSe=8)@W3os0%qkgt zAqN;A_Qjp+su4zZ5uMEl@g#(kMe?!bNgGZ_>_l_93J==X5G9&g=qXP=C`V`PuCf;% zLR$g3dmk(~ljb7p!WZqMSEg^_>+#_Eqqa%X{rG-HpUF=JFk4QKWevp=>^y@oN5NP; zhzcdV4VN|C^(JK1qzH7X6x3IDq&fZR&L6bEUX(T~+Ji{MR!$*YGg6`*Er;zvM$~FON%)AARLInG2zUv=$R3j6NQ}S_2_k*6Z2Uc{K z>`HY~#z|Y87?Zwx!sVH@QQ64IzY!gRh&DDK;aosWz68l{$U-!W%J9g^G(N#zP;N$> zNzk#ncvN-abUqvciEpsj&a=p#5znNkA{Y)_X1kS&a=p9t?5Zc^@^x#;PZYD!Hiz?r zHNiOxKtS;W&5ZADA-x&&7!x_hv$Zb~>#-vULEzK3P<4I+xMG@_v!Z@n7H_60qJ$s` z-v0e9)|?6qkF2E=4gjoqTSlu6{n!Y@x_tOCO{BwA|21%UlFMcQ!ZUm}$RLBz2JgSC zQKcs}GnNn(5<`PR{NND)H&;_?F`_@vtwwVj91raYt@&cy<&9f)6diz~ojVx#E8CQf zVR2P7aoO?jDIX;k0o^5t*ce7@**9o%M8zzoR~So?Qw?OkbOO8M0wY%ryI?ixUQlc* zha+_MVvgt$4z-LdpNgUD)mGme_LOZ*oTx zPy(u&iim5Yq`|^N{fUD&&~5HRrA-}m+)66~bC@WDp*KR@NMa}K5zRgeo`r^QQ!7Uh z17s0+ARLBHZMY3csbl2B3=wQ%nn4m>BB=4QwRfFJC<%r%a-6L{ZO!9&QwoTD>j`&f zWL%tsw-%AaW67bNesu*3+Wb{xFn0QGj`K5s_4?O9kg)QY6%x0bP+w6B(%`p->D(fj z-E(sYCF65QyWldNyYg#m3}gRCwDndvQvc?kb9RK#H^8HI(lXhk)ax1ii*}W});${m zp#3J8Aq$t?^f*Bh5$b+O%ATw;dt;3@&b%BE6LRRsUT-Ng|ArHK1Bs7Vt4d&E^u0KP zc=pq(Y_}QXR*r@};}(d3xyr-_ekc)QOV?dk0oMS<4@Lx6JEw^El?jZ|1eVVh#)I*q zX_`&f+{s=d4lx1T6@5hM9{5TY4hpD8s;? zF)pn;Qt!h5ah8efET9x%Jb5wAd2mJ`_Cn+o(EG%!(hCS|*q~$s!9~MFe&svv~ zs$BeL*q5a@_x76YiQuy4^e|hQ#a_El6%)5}kKmDDeq<;4;dno-uvdy0+I_`?Kv@qX zAQ&{ENWxwz@py-j1h}$YA|Tc$33gpyPB~2WinX5>(>J-2nd$mwRBtgQrxegV2ghJA z%>*JR1o&8CA54F^Hsow~!)^e@sex<`xt}m@ny_e0A1|!c_3Gyp=tAEoX^{R&35mrw z2CJKaDLymqtT`H=t+V87jQz6!qG|lh_ayR{eYY0noR{R5#!Ur^SMi$1p8g-=iM_w} zzYApW8Cs=X!usa$ZVJz9JBv$be)%`IKbX*8FngKlSt^~GVE;nwa)7>_LyGOgU2dPx z_lYC33(}SZ>rXW|wpotJY|#P>X+`v34L(D;g$Y4M<^chg@->df75@>S{BL?L!Q4ADrJsM#Nh%`#^^N zW3ama*+Be=H^)4Ph>j>H9btpPrkyuIeOL@P76OB`x~ zk#4~0)--yhR7#QbC35j|3Rur@sVO%#GwI$RB6dbiAdIB*fe&w(*Si1hrMXi8t$T_N z;-!OKE4B&-Yk)C;F&bWjOqd=vZEQ2un~xaTxqkh!>jxMjRWVA$x9W>W@qt4MQl}hO zaG*e)Ks7bU5!f8*?qAft90j~5<#wEpIKo&D<!oZb|eS{iB^?mS`k5KG9bIk$>bnB zt~O0^0Pl||>iy{w#ynwhF^P9!q)dNQ#eLj`2~r#hLT0wwMYJ`96tX2y-f=VEX+|FX z`~)-aDO!~>{-9?c7d`cewb|;pgfZ&X?wj1i1XN_tIWrtQ%z(V{y7flRxxP%viab^l z-&UHi>i5f?hj^V81_wh+J6~_KW$(?Qblrwz%D?`^J2PJNg`-JZLRqS{hnHZR0nE$c zoC;=V;(A9+%;f81Iv?QA;ax8DMiPGZTNn8)3v3eb_RcL2AiH_YOVJ}r_u8L#VjXcZ z%-Ln~)_Y7*f|to%$<=;%R(v)NcV4%(92+pryk;PHUQC+vL5DYeLp+1T_k>LC3!u~Y zlKXLP@(gsmdgZ;1wtltcKi}8f666yAqA4yv8P~dRKLTG~H`(W|+^3)?_%i^vv5zV+ zg#f&P%U81xIU_eN*lI*Q0$8Lm1Y8ZEkEMQN^}V)kBiACXWChHH7zgDORoL@8=lx{s zbN;fT@~?4?dA7YRqpF7bDjhyH&}CbhDfr2{6}zg@>!OJ`=P-WnG=(Xfzo`eb@Rn*_ zX*Typp}D6RJ8ymO7~7)ZG;vXdT3~IsNMZz7(+WPyfv}D_OzCkLT$a5;>#gxd4lJo_p;pmY~~^ney=LWqV`QC%t3??VqsRl?p0Y!oAd zoB;@j0awu%%%*2K=-C=Tzgt3Hk@4K&xeTW0tbm()tW%n2#1UIvHfr=Pz$l<=x;25$Wlj5Mlqj`5YGBn>Qw zU5snl)0}50_=^7YwI(he%q8FD=;s{-3dqZ2^`Nc?=Jv+`t`OYbtpBRQbO(onRe{Wq zMQy0ErkGU;$vT@-d-a;+m>n+w7yHF_^I`G{AMgBk$e4gMKe+NLq?jzuY^fo2nwAvj ztgb&i!&3>O5Bpgj;+~(e@t?g3U*i{hb^$wkAlQNo;wt z^xR2TQ`#7hF;B9CM5*cmrN|n_vt$78%OBoAKX4rQYP(%X@;z*yNq3HM0S7LQ5J6&h z?zTEC4Yv-cq{7f_17!aC-=$V=7f<*^_so`kyQVI?n|YI5pd^xx=$VQ$Ep3trroul8 z=)ggA2T{WX#G|T{c-|j|R@W*R^I!kS$*yHaB8W;L3r*Y{8uGlerm7`K3v9PDzs`aj z2zKuSd*9v)A&|ieD}4BOh5R|70D*e#T`s2VPZvKR?~FY?1{&EPA*A1ASq-&6#71~! zs`17~t&59c#H@Ko%;WUMQS}A$E8Y|a;+(@fc>CVhSloN2RMZoPg)46Sc{!WS zo+*WD4B7lmu?t@Grl*^=t>)i3#KpG#i+Zxo{>2*{m%b94N3i160!z8f`MZ6*E+t;^ z=Lo~L$Mty#Q=U-TYwj+C$uAht(UK@Fm2VFZb#-u`?slas$dz=@LzbA$fbsQ`C$~)y z4Uq1_S*xq`P7elsyx+g>&r3<>1}kBrzngui>fjxvUU-9X)?bc%$}J`5?%S*1KgA3T zmRaA%8*mOl`;Wo);0H=-Uh1`?D`qF4g5u$cqxIf061+aGcQkV@5Xun5KvX&G1Z}6` zOf*87w5wI__Z@4JB`ex2**geH9aq9WAI3@TDW%#&TPo zllDe{aF>#H+a)ct$KVz1JC22&se4tGbB7;2YS|#QMmNYl>n~ND5KQKz^(111iTUXK z&cb+2*k)F}Pv|+|DBw`FmOcb==(=U5dfoarWFOCCK3K}mXQ-Q*-I^gTN1h1KMTawX zNeMv?vfeJ-vpX<)QpNAG+n?UN|ogTUz>P75KJ!o(Q_CZc$$DS{` z;AON>u5)C!Mb7rT{JM9eJt%{NDoSykIQ#fDYylWl6Y4qdXgDt^?k)K_-@l66#jr4C z(+@sNm}dJC;zp{UOFsX2UvtACmUUtHVe_4gAkXtVb&B~8Sq_uiaz!Sp7~TTd{@5q> zoV_lj_BKmV3eh01s~P_}Ycc#=zUc2|7aSL!fc@aK5Z-syH4Kz}M@Z zEU}C#b44JH87AnE9-I#RUPsiz+-&Q_$jYg?Gc1?J{!FyPC#fO5q;z^|*VxnywXeB# zEhyYeN_0!$@0stJN5vR|1>cx&=l!{7-mX*My*VN7j}PyC=ciM(qhk}M_HdMLRT@E8 z3M+!5l?K$*cP6K}aR)zhW%k01Z*|ksJu!E!Ow75Wjl^{fY1O10V@Kik1{pQX)Xp8w zr(D+1Uc-v6w%I6ryWu7-r+fUbG;@{%d5=-bmgk3uVD^WQ9>fX@+`AQmcaeA!BRhov z`QWzg+HN2C*3Y-NSdJ#0aauaAxUcm#Fg$U-)Ar^s=hNE0#+Z{shS{76D}rg1qe++f z^V{j|A=?(HW93ebTcma(S8PA>BsWO+SZX=$1BH_0R2IbHKAQ82Z9Y^0K1`(fU?HyH zCS^5b=zeFyIZ6!ewv4CJjE3)_1O0dqeHX!i0e;IDrq`C6Q_k=NoP`d6M8ZWfj4DN_ z@3H637=o6Q>jt0S*A^9z9Osv?=M%^Se*t`#lZ6Q#%m78Y(J(uq+TXU}7B_JD9o?k|ejv{yP+GHYqal(YkJ?Y>&dpNRHwBu*g8IN)IRe^I9q-Hn+S zw{zX+^LyI%EQKg;*5edt4+szC9A)YdrZPu$MxS=$#0E(hV1|+Ku7e(_3*PFxhL~JE z+RN0b7GNV1N%HG#a|tjQ$&rD#Bc$S9qT?u9nVLfv&+qJPGE+!~ zpD&6fJ0&M|j#5NKZX}2zhB)s`uGW1FHw|Fw0$RFA6Lk>}JsYpy`$sfpDj~LeRYR}K zEXKL^clkrP7lo&$bjzlihh!ao@vy_Kj!X~k81%SnVr=lzcGZn^TKz)mxU#%dPucczVfpBE zF70ZS@d`gVv%_BvzE)}LNSkb?yL6-*)@217SnaXod?WT1H(z)?Z_U&8Uf!T)yknj{ z$rHG^e&S*`Pq##tuhs!z(3=~Zdcgx+MV)58bYB^fdNpS>u8S*`q~3d$KZL3q2m5Lq zX?-AE&VG{1!Xi&orQycql%GGJ)v14gzZ4k(w=f=r!ebv}d z96`)D$02XG+R32a5Kcb@dT!6*8rMi(U2TtJ?sAWbpp1h$YBh;joFg*eNLQ7^m}b`E z2D{1#pZeIt^4Rz;?F11)gL76Z=3r@ey*IG5miJ0C0-NHB(Nr)y3*Lq4E zG<{_JmVg7_*bE4+Lb>_D96)3+19&c35qBZ8$+}CNMjMWYtp# zj*3Rr23#4CyrS;m5V5tvwA;&Z3!c`xN;?E25-+?mD}8gN?|@>@F$w^ru@MS$i5ZZ}y=2%eybYsfy-5rl57jbT zL<5bGF#mv#k7fD&bcT#lNmUfd*MJ?J03}L5#uOoFX5THwuC&SP0i_bz^w6jAnEB!D zljie-ZB3b|?w}S@-D_%22~1|)O5+<7!utm4B6l@}T4`oyI49$1+ULV3nV4`qVdNHz z?evIbDHYhF2Z&qL*5)h8x`|iR0TiLzw*`Jz2LOI?m^wOA&F3Hln{$c6A=qN*L8H^8 zbWJU6a)M}|0GBt`)Q3z0+-9t)XP`g$(!-kk4tL`3DNiakRQ< zO>c8o!&K(?NoeW4bj0q^Qn6m~d3rl!6t5q<8xjvCe-Z{VCQSv%wO(8qj-mSv3Ckx( zSK1k1@xIqr_`!kqgTbsoVl(8-=;!YExOk1O(aj>d$(x+9C`pYoL$M|MHXyvE8FqyN zVVK~AN&m2dIWLs>Q9M*{iuVo!3*a}Pb-WM@H9*i)*)6XNk^#SOu%BOIYTZUJu=B-&0>QIwe+#IX%7 zg_b|X^kdyqp-YZ

5*Lv%2aB=XJx)HnwdX|ph^;53E zQoHb0Jzp%73EZWb5coGk2J|D3wj5cxXP|D)>_SfI-AT!SJF4!bG?s}NWb$9_bku0P zp$+Dn-XyTtUJsKy0Y~CIDu|G%VVzlMjPF7*ySxgBZ}#2%_+cL#6oRw8!F&24UOG)p z&bwiv0M)tFPHcC|n==OTV5IUS{+!tuwniqA9y=lvX*8$^Etii#(m&H7`2$%(H zzeUdX4sOt@LB_&zRNlI4?gigZEKGx2{Txof(Q^CFnBIMcBz-^lBF_G6JuisKOvLaevMdLc3QWoV~&mL!qe zFHAoFriMgN5D;=xLIaT0u#xyZGJ77t*A1GWQ?mM}pF(m7i>E5pivrJ89oGdV;TDr$Tk(w*JKA9zGB`6X9g`5W9xYyYR1Qf zP@#aUpn03lVcX~SWH--&Z6|y4na$O8>Hr2w#ysS+eCsD+ad9JqE$D7>pN_$Do+MMJ zCGm?=DN7R;_+s7PxcO`~I6Cr-u!MhNKYRW>O}f@{#-IcUiw`=#!9U1vT+5Vdw(&^i}rWS)&LvX!rGm0723w2ac(h@t650*&WBYe9% zZh)?HfUk%Jkd$&Sf3Mno4}_K|x`0=7KQkguc0APe4tfA1+erJ|m3k zut9gGSOMPFF}e?hGWlI#h2v!&I>}&IT566z`g1C#Sx-Q5tpnA&q|+a#M?1VY+=?!9;n-)!0Y)}l%< zU=?^B496+tvBA~e1kY`4Ovf*P%6`cFK`QvP<;JECfK)t~FB-GCJ6BstTJ)m2R^+Qw zH&B2{OKH2c1|2k}|GbRyOufPBPY(67*BDWcj{QjPjJ0jmE&E8#Lp6h5fZ{PAnJH-pZ)NT@}uX>w$bWD8qR0a z(EWuhV#Qu$_^Os`Ck>6^E}(`n>TKJ~Od{GlM*>05fM=FjT?FIXpiI){`kThb)Rkl^ zP^iz#B_b_gz3pNi1-05X9Uv-XmKG9}SLJ#w*1?yO0#g@tA%w2#98~0QF$zE}uOaiN z9+X^6Ix0VD{phd5GI#X_L6+GqkMEZtvFTmjwct>VuxV5DmQQ~;3%*F`U>HI|lv8#$ z>Em`xZS!ywSNd(l);SDsxbYD4F_Nog3$Y$q2<;Q>8V$NGC!T#TbFBu9GfjI55+G#< z&D$PfNfoQh+oqS!4dS=g{xm|R|F9r8)E%M~^RF&;`Y}H*??;xP7 z@nf-R2rFA38t&J$3yYyKni`=tyPuzkL9Zu}!-9Sbd7V;bH7ntj{)=dVGl42Y(4^NP z*tMCogiTcw`W!|U+U2zV3q2Ii-GR_Ye$4r(g{oP;2NVx}Xre@KR^V8K>=m>5ikTTw zcd{6d!ZH&z925|l9~X4A{tyqNmSB|Sh=%^&VrlCdc=*jdWPPARg3-1yRA1c0+k0V~ z;Nx>Rgzw|T$`X#INmC)3ceP6?8KJ{bcj#%>g*2`#*u~C0ziup4HHVjl)Gwzf?(=kR zX<7rjwM30+rhKF;Ge*F+*?r?4x^7ko)R`mi;@2UfnezLJ@>qbt*9JW$0qk&Os$|fm z;qycSSDF8)4)XBslLvFh(T|ZjnO0<6y?InvJhY_OH*=fQ7cj!V4gNu*?TC$GhZz>P z<{D4WGIM>+UeWEzja|@9c@CdN_wTHM${VJJSnc|djoQ2?A;cwhRXeXJkBuw5QJk{1 z(MV3m_sW8N@%%ohta*aJFAK{1GE9ub3bbcaD!_Kz{@Zct`Uhx71S_Ut-I z0I~26zwwf`>h{{7EdgQSFmne%$?%y>nbnM$ERp0jO^ zxp*6j%NQPZAp7q&QCLMMU#3YHh{0lS!)Uwg@?DM%e(_7kL&gd5$IDLqdkY<@7@)OX zLcxPQ6YaXo+Q;+(p}y)l#{FCN!UigdxKB$7FGAnYb{EINbzDPa)=o9@NTxs?P1M4o z!z42!!z4z?H;R=43qui2MJ7X^Qu(-^w!5&)wH#-}=C+zQa%BNtmo2*V2=?qsR)%kX zeO5?Nr>*J*8OU8okoQ39Z^j+(#(|$%Mt}5Ns^Jk>t_BooarSt&m6`F|cWm_l-#VJT z+`+t{xprszvsV=kjUUea z8@jnO&yLD9NK>+8ebqF%auXo@tits4Uwz}8XatLRM3rO;V!S*#Jrt-aNa^@qU$Z{3 zxWFruDNhqjQZcRk1JJp0HgL#2DKPa%0xDrz#Tb;;u5Eq+I+vDu_3)Y6i?wMd&OEW) z=K_KdK^Vyv?}*G$&L=TYXy7XdE@yX1PXVP(5kcX@O&Rs3nk#eXdg zGDJ{iE-%iS?SAk5set$tR-?oJbyF2oSrRiuzOXo=np4ciOO{2 z^0|_j`13V%yGk2&_$y024>V>>6Yx5hU@QvbQMl+K*>?wRhZeA^c&m@(hzIhqAvH7V zb)p@&{w}JM#~Vpry4L?SH#U;FNF~;;#W06X8`yE7b4ipP!yhuEL~*?wyx^f2ef-tR zSL(XiNFiH7jL^{~p{R8XMVzVE?a1HJr)L}HAIIIbI5tt??}Y&#ruIFgX6tX|&*~<> z;(D0x+*9`+Xo~Y8I$gmJVX2=VGgSe_1o?~7P&s_5GEjRXLqU=l4tj=YP>b+Z(n}vu z`t_JSEs=Jqd}l^*_P0K}V(%u$Z#;0{7lc;%LgomD2(4$$(-|-CTz?sxL=Xs@C6(9n z$&vZm3tQ;#_j3C0m)cl4_6%?}&^zglBbp)3-)A83U<4F9PTm&)aI>B!U{L{(6W5Qk zy)VN&7lTA8j(1HmPxK6Sa@$cuNzb%*BG4!7@GonoU5zZdJNc7O{ZU1E12>#FBfQ+^ zO362o{;EAKo&ySK+ICmW1|ifU)f9{draIa7Z~o^#j7e5+eEBq#rXi>CmD?-QeIdah zms>g~q8HfKnl`em#7{IvxYr(yC*evlnj3fgw?$MdEv&ypS4%Qw>r5n&*K1FstWADh zpagm=WtZLc>6#Ak!hZua$1<(QnTPjec|@g!TNNn09_ue@;MDWo+Z2icBX5tm_45Sy zk_>uN6pn*I0rG}1rdQ+N=W2x%wAOTsx&p6auwW}SMq^;F)Y>1Sq?`J~?bdvEpgrr; z_?HQ~P7L|Vs)o0i?`^WG@(Zd}_%gsuWH4olvua6l@9i&c_`ziW_WW*zTG)znVm%d4 z5cE8ux_CW^6`ch25T`Fc4>UuiBk_NyeQtPg8tfTgQc4nhV8{P7%%qwV6;I&I&J3Io zmZ4i~^r%sM@TtPWws7cMAf{M`>NS9|J*3T6 zP@>E(zzyZEw4Rct1mkm`vlM^-#-8O|&_2;5UL^{AmQn@g3el2)wTIc;4Yccl2=(kR{*`KI;+-8CM-zs-k@-s`VZ#B-LPe?hHciSgA~


a_5P40CkQ9Kd9;qg^iCZ&}eBkO3j^E@_%J+mJ{UP1G z@9!yT_@`uv=M-_Q^KmNNkhNpRQmDUfLAQen#`prtccv{T^l}&i%}I!b%@mdE2bh2x z1O7>mn3SotvOOt>P7K_(!pB~)9y!9(out6k-ATI1&XRcxW@ zk-J={_x4hXQ^ir9OeNPO%AS!gjI?G)dzYj(qN)A~&+%{o$clBqWG{3b_kkEzh)I9X zOyam`Uw=`XAR3f2w|j`+-~!g|zF50~^dp@+lONszJ(=I_jGrU3RQCFqG5nbzi}(Ej z14nrf+U@MGC$wF`FWzwt>IYE*orZNoKL9WA?5Wy}%*PIen6KJ&Z(uKWbh(|Rnww0Z zres}1?q$Td z_#6yqtX_Ni{B9XTQrS0b+h-Dq3Y_WpJBli=N%cW4GE08ijd(*(x8>`+Y-kifn$9sD z!~SIsT^(;M7R^pj zH!wP(vuGH{V|x!ek!4zSM+csDjMvoB-x5d264tIm;LnVY{AXOY5S^v94QE01g;db> zP*7)I`-A#@)J2mQL2l(jrQOd?^LQ_~#F)rW4<{_DCEt)7%+c$H*SlRnv$0dR+%JM% z&aq79nm8!-{#;Q9;!POR%YM$EVg1gHHJA>t%2Eom>$#s$ zo!6pt9fkC>j{h9K3(89nfoD3-O~3et88Yl~Fok&M{W~-#yk5_lWL0!|K~O^6+nX8w z8M9W#Z;Tp-1wGgV%i#f${w4XKr>icM@F1Uu;jO^}qRHeCD|hfW%uB}oC)C~F%j4K! z4omXA!}FmE8k0LCMf8Hge zpp#pM2lPE>09nPmb@?w{1O145^(j2fAq_bg;3bp2L=xr(hiGoOIG5Xk74AxMJeGz| z(#T@xrh1X8nQRgdjxSo)sQ0;G91cs>_$1&gae}Xvokk)K?fVKTmoH3Avfl081hJkT zSxyOu8cvc<)_rLx7lJr(MR^ECFR#fKFXl+p&V$&M9-Lq%f6XryoJ1?-waR>~$)izz z+_AjXU{`?HHkf9NZyI3xjz)XBxQSiP&B=i&crj|s(6nHQ$p{2dW%b7qoa#*;`|I=nPEnL-^HDp4_&@`nHJ(A08*(orW>Rcl4_rnZen3 zL@E0n*pVbQ(_IbV&^d-o_lAyCL>N3g|KFqkPO zEl_WTE@dD?KWaH9yZ$_S8uC}L;}QFS3BEg040-y(1>mf#6*XhCp~)(!aNESqSiN{G z;XT2%D<|YXpM*bplR8c|x_+PNNmOoHb8!b&YG`b#7y%YLBJBI`#*Zc4(w}g-K3sql z4%uF=kR&l4Vbu;k4_+8TxO%lYe`PqQ9n*82QSG{ze^BD_`Bz1?JhMNlXa4;uTPcqm z1B6}Bc{4vWy8IhKo7%FP7G~zk9d-s`yMnsk|1-@aGp7HJbPL}Itn-g<z(3t8Yq%ZzxA;_`Kfxn1Bx9ae(bT1~5P zKyGk;j_2!MHO*(2_nL3YiTXfyZ5$`Zn*14&8y&f#hPKqpFcJgu;tA1?g2aMLLwv!H zqujsB(y!l>X>C&2(b0?dLtOZaaG;ND=dyob!=|hJb};^1ZpQ2Xz8@^mx80`4=A!(x zJx=%ftE~z0{865nEjiHlt}2SH@*@)NZ^o}Q>Qn|d)?>WC!GS~D!gZxR@0J_c)T@&z zg_AW;E_TCd?)Z42+|uUqYw1~z4pX@T;tfYux4l2Uf5vFMe&W#6{bbsLcsHC8IXyX@ z>;Z9;9ypANXt%(-SY#2&#<@;k-I#tgpgk@9Y#|K5H;lUB-o|mHxX_B|VfbcfUm620 zLPFLkzx#s`e6oX9B>k;3Mi&2*Ru2sVxkG228P5+R#Y~eox5!`o-eB|laJZJYo~j{R z9-oM}Z`Kh5{!292k?`#$r!(~nT7|5pXC;pz$*8W6r{|&vb;XUKz;OF&Z}vk5--dIV z3!wTBf7|EN(PL9zVIo!ogGbbE1|W1dyQ~LEk~8evh-&n-9pG+BH~9Dn{B3zl+YjsZ zQrU5PK_K&c%?}JnuU-bFRm9Q;_TLj_kZHwU-BH)!{Kog*4!gQ_*!fO+gzQdcpChBVB%=fP2^GrB1 z``J3#Yb%uZtLzv74CWe|^wEO|WavKe&vYzVp>v<|#pmUzR*OnVz|Mznl&7SB6dftE z|5vcMvHsAJ(M0mSz6KZs7k@IoN6D4XqtiH-1O)~X6jrF;tyjb<`k8qf@8fmi-LOlt z+N|uig1V2;bKo{eI||;3!~6&rpjo$q?es0N;MJ{Q`V<7~l%P=R`RT82j23SxR_tn( zV+y5==wmzQl5PoR9G6|zm&IVuJ2v>-^Gwp(_XyAwiB5WCoIPu-H3I9(A=xy3P?a&o zJI?!avmSjFX_ocAtP>17ac$ZbbpmB)e1k}VW};q+(x9@5O#?#YoKp^R%FY(ww071lg|j!nGwE#5Td!YvskM`I5F>tFyA^Mm@cQ6G>?gTd8Jw zMQ~a2K*QCU30Zyft6?{hpD-C=hW>{5MuO>0q#e(A=<<06>fr96Fgd-5k7S!x9zeK| zPXyf{FShf-d+Qeeg*j#4?C)zD6Ip=aBh;%1^@H}-)}nQ2+pbjb`DwQi$BE{E5C!5o z*$#)?wTD%dFEhvVQpPf|bE@LqxzY78oPsnc7#cX*7l-8+H#Ufn0;U1j&EGG+-Z(fA zkbhnMA1e;{*RTKKxn$7Sx3sf#*4O`{%>GXV=-Ei_t*?7^zT^P@9Rcra1pXfgwuY8= z{}qAuu12Z+*lCIDii79|*>_CN`FK|6-N@&k%kadsC!*&D^(q-TvypuO{dL5LbaZt?#@K>I@Y_cAVN{sY0u*wD`I zzb4@g67aY0YhAB>RsCP86#rvi+Dm%>1hBWTv~xFg|8Fr=il}&8eN_?~`#;OzY5q?L zXOI87kW@)WLjt}MsPrX@@b6X3u=|I*Gtk=fzvU1K{eHDi1_b2K!+-t(N!Ne4yV!lP zXggZkng5q0u17Sr0uew!7$`wNzSPeD*NOvp2LB1+Y-#gfyme#`GONBwchhtJk5Kym zxEBYk&iups|GWrJ|6iYlqb7y;p|3<9eyzrTvH1Rv6$kWO{Xd1BcU)HG`^OI=j>JXk z*V4>^dn1$!jc_C8!mapBc;Hz+gXsuop=m}AP+Z`)+?u4Bd(X_h&_qRZ=E~IH=k)zb z-RFFGeuo$TzF*gU?fbg#bDn>b-iCxn*woZlhpQu-^u&rp*!E|Iz$_ZOoomQiDP&qa z-0UjcO#V6Ew@-FT)mI#*)$0EHa~4XUA0K#NU?nW_Jls|OUHP|9cFIvNuYVGc@t;LT z$v?_W6n{5zR-N2Bk~Ct9B;hwS3!dzhY9%-++GSxDY;<#4&GVHc$p^{w{Q``f%33z% zAd`jpkvV^{_dYDG7nUY+x^*gq%{Db?044~3@uP|L({F-+hb)Fg|^ zs$mY*XY{*x;lG*&9>#$?a1fS`|vvd-8ay$Xi1YBg1QdWqTp;{9FdJSK$`O=g^G z3o91KN~;T29KD!;G>){P2ovdTYBr1aAp*Hf4N2QfS=MoD|Vl+m1to@i%U2j}~u z0Ic{(X@T(Fgr+llf_+juyWExGlJskam%?3V{>orBoz3G?zb&5i{0ZvYz0R(iko&-E|t-t%mJBX)dO5NS*Nr6sJw8|H!CG|$4FWO#`>ImYM8A51L zHd~b0;2hymg|UAg@%_52B*h{^#fZ-~6K0ItZ1SkWq4#eMx~GvOYw02io(nxxLC5)4 zg=US@Y4fz^)_dQ-)yN=8?Ur~+XaWnK?36{98P8-+wA(xz#X05f-Q`GdtB?Vz7Xal# zP`-mV5z*~&!9la{8FJBI{)qmvhCqAXozvW&)~oH>zp}L*?E9{lqTE&fjNl3{<`sH` zy_ke%>&xNFRc`-^fjqlO<$Wa^4@L65r&qek_$7>%`2V=w_&W{vpN2i3{*1<6&{~b1 zglDrGtOk9Q!IS2UN4F2|A1+DXAjee~kgabKq&?d1iFUQ}#NNB)rYcY=;IW{hYA3PV zoUSN`!DIMt$(WWLUIrgPYp(Q^Upo=x2(u~HlY~om@2WZZybi(kHl)1*f$^VN#Q3}|yZbOvn1JTdc{j=vs^ zPQZpXK(zAM9gOMOtzY^_4JnQiJ^@x1gZI@zM#HbQdlK%ejG}Y1kyQqxREo(o^fbr0 zJm!{uW%G~LLvq%)Rf<>lDdD;PpP>6~c1SUKBb%h^3#R(`sD0Y)^k|8d7n{e7^pT_# zM2Q%}T}lvIoYtvxnCvc3`Wuz=)81_;MERGM#w}ikAx#!fw9ZeycHKcgbqxI(j8gDq zrxXcR(=B2^X@e!x5JbJ%LyK2<@PEDtNL}zSl91KFaPUn5k9r+jR#*-pR^8e*E^yPC0L6 z7&~jpk0w1iaT}dm&je))J=j7Zc70(}qqcb?CiRAMcZdIo<{mYjftWCt%luVi29k>( z{117$;K@$$o<$fABxqNnolW6-O#I_+Q*6xdLK@L=PA+03hb^9H_|B|TzkLbEnb}Bb zsE?O2ip%Y=5lm3?ob%rBn@ku>RL!&X1mk3T-#h6g^JX@br2AP)Mw88qA^$Y6a$j;W zG5q+)kYS0+VV+-O3~O@&DVcK9RS8}jAFdpE=x-B<+l9TZoF@n>uh;Gy{?N-|m|S$= zjYW#7va43(V3wZU^3JkHD47u!rKQUhV-Tl_sdxX?f#>`nib-Xnk7^Q%UQd?uQI&wx zDEpG^F{00ntW6jOha-_t&aU00^FtgR!iHN+4*m@Y#iZpEj_|0cg{85WMvz1dB}-F+ zF}n?ROaWQ88nmT`zx>;*y--#~@7)kYNH&%^wF_ni!aJn?Xy!!5{x*a$f_Y1CpXuYL z)smzwR%NET5keSlCkFgZwmpAwqRH*O(C*|;N;`dLP=j=lIy)QGe0XY4mc<{l)lo`2 zEl*$+^rp;x*ERjT-A>ek(b!&zb_z$jQ11S88oSl*u(?>4*_bh`YcAYxGTcveoURKP z$)Tk|dGh0&66fJKr!X25gHp4YVDx&H$e+{=8t@Kg*Z!#eqAd1R1QM;W8yF@#>vH}O z$l?cv7*T826O7rQVU80i^&R5}#gs$Y7KI7TAry>fM=)WkYmct22p^7Bh4+-77{)?J zV9UaTOFTwL^9@MIS_Myb%9&l1Vn%`eX=r;u~6$JLzPi{QxU8WdWd^NMinI67k4=FGGo^uIuq)L^AJE~&UW z9vq2I7uECJ14EyWMaD9IsuX8TdBXX7E=96&t@PjLXTfQz>y(MBWetX58eXC~_Mjnr z@8>$o)2y;>)C^|plEq5Y?20&ejWM9;!RC&HpfSgpV##8zfBVf4PzJ= zOXaQZOlZ*xt678%f|$xnX&8i!-~aIau4W+oWIYTd9-Ab@5{4lTKS=0u7+C)7zp_@# zIH;x=4(Y^nb<<8zCVFDBzRdvdWylaMR3#3-O{H==JstY4@h@7d7Y@qGXDDrC9Z48A zm({@hKYh>52{nO9ou*vR(=38ReuzL4%`|VCx9ASXVQ_FA zsaLdg%vm+Xpkqn%}=+p=bh&Kg^sFdePu0~<6A=Iwg#sSM>Dkh9U|ptU|S^ZSCzKoRR}04ny%{m zh@FEJoj}aCLT4(Q z-dw1ZA%_w$qWjr;5t7Z}iYBVD^j!F%+=)*`dJs!~`yrg-pa~-X+@8=3bmW_jN=qG$ zVVnfCu0))BCcULI;;9D8k|@snsRV;MU^Fm`FZgLfQYUz^3riDiRc#^RxUFm>m^5W_ zLLc;J8)0;@G^H=$5P5S;TaaP?ot^|hp81eRTvy#?4CRW(p)k?UI-@S<3`B~-{!|oc z`gdwhlzhLSup!Nw;J61`#MV@B#O^a|H9_eMb;O2`w*Ryq@(ch>bi{UB2nd}yv)DRA zl`(P9&^~BLykt{%8=)}gxlyu>dk`Wb7VQXJvEa#0Ilhx192yIb-XpCxS7Aem%&Mz{ zFv*sD(S(XCi#^=9tJhoGWd=tp#-(ys~u z(cMe#FZEi=1CFWKm5G)w=Uq}&E2114dvu&BiFo6l+NMDtA{)*`uOvp<;9>-X0Xu;_ z|M9iI4oV&f%ZZTHzG?{Kg0+6T8t?e23M$0C6s2l9muC=XtijBL88N)emOs!IzfMpJ zlhJ^25|}VE_V@RD0~?(jpzKg~wI>j|YyM-yHv47RsG}-jJdGd>LxSFIVWSD_T%T2U zV4iak$_4ojA{>7CCwtZAdIl_LU|Z#a8pWs~Cc2_70jrw zUX&!<4=i}HQ?j-)lq=F``unhigu9(MKRK#dhb<+k|B+q!-kd%Q%>buzZ3Q z!j}u4>=g6wYD^3fWM#z`M*{*Ssnj#2e*W_ZA!y<>g;s~5hZnf+p@7!JP)kfUS05to_PgcJqk|8s4j_RU?}*h|L;A)U*4` zHCv~^4U1y`A-Z9eKu+NsyfZ$Z$8Lo*6Oj7Et;`i{tJO5=X)pO;F#5F|;TyjEt-`>Bjl$ZgAV-mc}if+}rBvJzM1wS!@3lB)I0 zFL%!{Ny;i-M3KQq?xYZ00wLPHwC-iwmdJH;pkc8PB-HRFc1kc=SX94L_lqzbxJnOH z{p-!z1cRvno0+dT{P4(QWbb*{D2R&rz9FHQEUYIwKWS-=!{}u`hoQuM@997Wu^My^ z4c$SP=`qP?52{-;R5wwai=m8UXI1+|pl{p7aNTyP%H784ZUn+w#T?%?jzj3IoLG@q z*uFOr#DCwAV56Jz*P?ny)W<;Un6jFzOQaygBn!!d_ndPNV=BBH{wFG?eHP=yu+C_} z{d1|+VC?CLM=>Nn*~lPP7qj-EJvA#Xz)5Z|G<+iE+ii^Eq?5iCMV+l0w3MVbFO=mm z=75@Fr9;P#b&4Ge*;5DLhz7HZK8=Kes zoDz0hK1z@~L@^>qwI&SQio`oV%-VN8YY;OUW#BqAB)V^Ydjcxtb%0-TOmQwOX+C^L z6ldP2YKk2%1v>eNcu4&*&40%*^|vUc(Zf1Zkdtp!q~853*OBx}Cn}9@AIvz=4oiHU zlFB?>g%lB7U72T6lNctR6_r`3@zvs>obOckJz~cZ2tQhG)bL8lgH2$gbfqMfz90~T zgDq$Ed$;Q!M`=4TNC~O9nLwiHZ20i=wx_;9Ecvuhy6^kHG6-wV!s8PED393QgmfjA z$g4L9!-hjsTu^7#?M&;<3ywm$$FNNnD?!8CjAUh(PZCd#3eHC)jmN>V7z9o35r{*h zVM(vuj5X(*B9z8K95D!bykZoa4ljYR1W_rl&d*2CNPiljbly#GD9CJM#Wp-MWMDPq zh_UD%#Kf0Vx{RtGU`TDDLo4sNw3llU0}X2`HC*;xhM|xVW8>EiLyDQ9DhN-i5QvKv zgcA|{hZ-B=9^y&mdM4E*j6_#7P2?$0Yt;+IdMd+4;>f_ihZ-}2w5%;MtDo8oBU|7y zqR*xGBLrN5A&`iKeJ373j58pJRbawk0*PZDw{z9_My0TA>jv|ROIfa0GgxsvHn;wU zH^@srv{Oovht(;!csB zQyC?SwRAD{297%mhz&~;L$JtH2EqLY+$1N~SreXn{yrpGiCQE|^X@kUWYROp;Ii|# z)kJA~1UOWgbm4Cb>mb3w>QuLFwIRrvacQRi4^~!`G@5A3jwH%*E7&a~5 zf3h-DKMWW!$%qgx*F|~He91x%av5>!%9&}-1lfyT_Ux!V<=85+IuPLYxZ46k?-5{SA8p1&l+75?zcnb zem%RDRV4^#6!V%1&zBzPit6#fRAqu#5>6P={Jdi8plNe_pqY~@&D6FNipfBm@4deD z?YQnJ`D4;0mipl?Lcwci?40Yw#LyFOvubcjC+#tFtn2X4s=jC!k*) zJXkE!XK{~9nSh-ZlPj79!TK>%R(%5HJW*}hu1;i_IFj??7bhfa1LM3(l)jS*!-_P) z68VJ=3qW_(kF?cL5_DDEe3Qd%`oR;cm6F_4ZGmxtZ|yk>&U2cRnvOStVW zgRrQ&nj2ry2upoYQOWRqNEmcLbY*_cxcSh^PK1Kkk1Tmc7zXOV2Ok@{_rb>Gm%7T* zlJtr}EOz;81gu{ES=}VfLPzoxjw~9fc(ZbKj z3a6fp{A-o|aGw#GX9avwjEBkN3CPSCSB@+#bqEAXaPmhzBs*YovPI#kavBhib^e4Q{3i2;>brnytU^V67h~M z$_gC0jxekyY(5AkqCv&=@)_k$xStrE2|EeKp=Is$$x2(V3`2O|LUSUz&GthC;$$P! zkfPqD`$4nKVIeWy^g6C)#AzJ()(~qV)7HFGNxpJ547w3tJH2w2aBztgu2b0ktr;uV zN93ScG9dMc!I^id5SX3j_w9Sb-)u!teFpW28akItKqfZcyO6i~We}?7ePkWc_YT}5 z47|(8W)r7eAAB|q>3BMpB}x*AQ$A%Q8E;@|!t&jvEeY%?_+yjhWg(%9scPgv0x?I?J#W_* z_Dp;%MY;ousGQ)Tg-{}`?B({It4yEBp8z}ppAl<|w}yZmdR#4Ho3bJA+UCf&JJKMK z7}sAI2*>SmTUgCaztm^^XTXHPjYV65C;~xTQ9X+qp~E-g=!;Wjx}=r008>siA$Hms#%yyK-7IrPRB2eD z3qVy-a7ER;=Td|81|#klu{Qs&p6l}U;7o+2L{54<;W#up+=(D5tHiPQ# z6)P+qyb>#HgWa@fxRzfLj>8?NwXtKZ>v=&-<&S;js?u1M>1vLN-xl_Ay>2C6me8xB zW6w7PVvV)im^1bZyg2%Tj6rS|=bB|^sxf9$RJ2R$ru(zgb)uzWAo-~>Vyz5D`tR4c zh+l8*v13g%%oU?57urk$vc(z98agv8-}2t>1V}Vl6(HV=)F6w-YGoUU750Q7$q>hc zow=Cl*Dhuh`Q03ME5bQARwF77OP`ILdWZ^+pfE@1=5s7mKSUC$V~)ebFgahy;Hkv z8D!b4syZ<_YKRRV{NfSnEE7`u98_ElJ&(A->b6bIk#D7#@Qt}bU>RxSkIN%Pf555> z32peU!eM`h2=VHtw%0?uLQj`e@fvr4(J-=NDKU2Ft3hRe*`z8SPjNl2U@!5%rVQIG zGe0l7-Qj)%xy$SaaYRkEJ+9_JPgboh9-5;2+K&5f`y53=y9c1yZI(MlU=Egu8g37r z5Cj)k0F8?bztd_4KabloW6vLH&({NT9T3s< zH}0t+aV9nqPAuAf?hzO)23i*T&BG4~gLOi^jdgoRLk#&FFrr2dKUO0W3i-V8^lfpQ zL*X@ZQ~?_ER82`TSha2kYsAB1D`k0OF~!X-$eR!#p>$)!0C^FA>h60;cE+T6Nj4Dze~rC-IcKR)6XU~2Pg>t z4eE=$b-$t~ eZtsqvpY`t0Gt39|4yQz3_#2L`TQA((lKu}gc-Vgc literal 0 HcmV?d00001 From 04f432a0a36306392d8246baf029728a80466676 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 23 Aug 2026 11:38:09 +0100 Subject: [PATCH 78/79] refactor(skills): clarify contracts and context --- CONTEXT.md | 28 ++++++- README.md | 20 ++--- skills/code-review/SKILL.md | 6 +- skills/commit/SKILL.md | 4 +- skills/create-tickets/SKILL.md | 6 +- skills/debug/SKILL.md | 2 + skills/define-product/SKILL.md | 4 +- skills/elicit-with-context/SKILL.md | 6 +- skills/elicit/SKILL.md | 4 +- skills/implement/SKILL.md | 10 ++- skills/maintain-agents/SKILL.md | 2 + skills/maintain-decision-records/SKILL.md | 4 +- skills/maintain-ubiquitous-language/SKILL.md | 4 +- skills/measure-code-complexity/SKILL.md | 10 ++- skills/modular-design/SKILL.md | 6 +- skills/pull-request/SKILL.md | 2 + skills/research/SKILL.md | 4 +- skills/review-architecture/SKILL.md | 4 +- skills/review-pull-request/SKILL.md | 24 ++---- skills/specify-feature/SKILL.md | 4 +- skills/tdd/SKILL.md | 28 +++---- skills/tdd/references/TEST-DESIGN.md | 79 ------------------- skills/test-design/SKILL.md | 59 ++++++++++++++ skills/test-design/agents/openai.yaml | 5 ++ .../references/GENERATIVE-TESTING.md | 0 .../references/TEST-DOUBLES.md | 0 skills/write-skill/SKILL.md | 18 ++++- skills/write-skill/assets/skill-template.md | 2 + skills/write-skill/scripts/validate-skill.js | 17 ++++ 29 files changed, 210 insertions(+), 152 deletions(-) delete mode 100644 skills/tdd/references/TEST-DESIGN.md create mode 100644 skills/test-design/SKILL.md create mode 100644 skills/test-design/agents/openai.yaml rename skills/{tdd => test-design}/references/GENERATIVE-TESTING.md (100%) rename skills/{tdd => test-design}/references/TEST-DOUBLES.md (100%) diff --git a/CONTEXT.md b/CONTEXT.md index 384e58a..f62495c 100644 --- a/CONTEXT.md +++ b/CONTEXT.md @@ -6,12 +6,34 @@ **Predictability**: The degree to which a skill makes the agent follow the same process on every run, without requiring the same output. -**User-invoked skill**: A skill the user selects explicitly. This is the default skill type in Propulsion. +**User-invoked skill**: A skill the user selects explicitly. This is Propulsion's default invocation mode. **Model-invoked skill**: A skill the agent may select autonomously or invoke from another skill. Use this exception only when autonomous discovery would naturally help during ordinary coding work often enough to earn its permanent context cost. **Invocation policy**: Client-specific metadata within a skill bundle that controls whether an agent may select that skill implicitly. The same intent may require different policy fields in different clients. -### Composition +### Skill types -**Router skill**: A lightweight skill that invokes other skills to produce a combined outcome. It contains only coordination unique to that outcome; called skills remain authoritative and the router neither repeats nor overrides their context. +**Skill type**: The authority category recorded in a skill's frontmatter as `metadata.type`. Propulsion uses exactly `performer`, `router`, `teaching`, or `utility`. + +**Performer skill**: Owns one bounded requested outcome end to end, including its actions, artefacts, verification, handoff, and stopping boundary. + +**Router skill**: Resolves the inputs needed to coordinate other skills and returns their results. It contains only routing unique to that outcome; called skills remain authoritative and the router neither repeats nor overrides their context. + +**Teaching skill**: Supplies knowledge and decision rules inside a caller's workflow. It does not own an independent operation, mutation, artefact, verification, handoff, or stopping decision. + +**Utility skill**: Owns one bounded reusable operation and returns its result, whether the operation runs in the calling agent, a fresh agent, or a deterministic tool. + +### Software quality + +**Deterministic measurement**: A reproducible tool-produced observation of one named property of a fixed software candidate, with its scope, components, configuration, limitations, and comparison basis preserved. An attention threshold requires appraisal but is not by itself a defect, an overall quality score, or permission to weaken behaviour or design. + +### Skill benchmarking + +**Authentic workflow arm**: One isolated benchmark run that gives a skill set the same implementation task, starting repository, model, tools, permissions, and environment as the other arms while allowing it to load and follow its own required skills and workflow. + +**Adaptive single-run comparison**: One fresh run per benchmark arm, followed by another run only for an arm whose result is tied, failed, or surprising enough that run-to-run variation could materially change the conclusion. + +**Skill-context tokens**: The exact tokenizer count of a benchmark arm's skill discovery metadata and every skill or reference file loaded during the run, reported separately from whole-run model usage. + +**Runtime token usage**: The execution harness's complete input, cached-input, output, and reasoning-output token counts for one benchmark arm, excluding the independent quality evaluator. diff --git a/README.md b/README.md index 0684ab2..dce137c 100644 --- a/README.md +++ b/README.md @@ -97,7 +97,7 @@ $implement The session produces a minimal local change with retained tests, applicable quality evidence, and completed independent review. Small, understood work can begin here without a product definition, feature specification, or ticket. Implementation does not commit or publish the change. -`$implement` applies `$modular-design`, uses `$tdd` for behaviour-changing work when a usable suite can exercise it, and requires an independent `$code-review`. It uses existing project infrastructure and reports unavailable evidence instead of installing unrelated tooling or manufacturing confidence. +`$implement` applies `$modular-design`, uses `$tdd` for behaviour-changing work when a usable suite can exercise it, and requires an independent `$code-review`. `$tdd` applies the reusable `$test-design` teaching for durable behavioural tests. Implementation uses existing project infrastructure and reports unavailable evidence instead of installing unrelated tooling or manufacturing confidence. #### `$commit` @@ -151,17 +151,17 @@ The session establishes an evidence-backed root cause and stops there when diagn $debug → $commit → $pull-request ``` -## Utility skills +## Supporting skills -These skills are independently invokable outside the main lifecycle path and may also be composed by lifecycle skills when their trigger applies. +These utilities and routers are independently invokable outside the main lifecycle path and may also be composed by lifecycle skills when their trigger applies. -| Skill | Invoke it to… | -| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- | -| `$review-pull-request` | Independently assess someone else's pinned pull request without changing their branch or publishing a review unless requested | -| `$elicit` | Resolve material user-held information and decisions one question at a time | -| `$elicit-with-context` | Resolve software-project questions through `$elicit`, applying `$maintain-ubiquitous-language` and `$maintain-decision-records` when their triggers apply | -| `$research` | Investigate a material subject with high-trust evidence and persist a trusted cited report | -| `$maintain-agents` | Keep confirmed repository-wide agent guidance lean and current | +| Skill | Type | Invoke it to… | +| ---------------------- | ------- | ----------------------------------------------------------------------------------------------------------------------------- | +| `$review-pull-request` | Router | Independently assess someone else's pinned pull request without changing their branch or publishing a review unless requested | +| `$elicit` | Utility | Resolve material user-held information and decisions one question at a time | +| `$elicit-with-context` | Router | Resolve software-project questions through `$elicit`, always maintaining shared language and preserving qualifying decisions | +| `$research` | Utility | Investigate a material subject with high-trust evidence and persist a trusted cited report | +| `$maintain-agents` | Utility | Keep confirmed repository-wide agent guidance lean and current | Use `$elicit-with-context` to work through questions or decisions within a software project. For non-software work, use `$elicit`. diff --git a/skills/code-review/SKILL.md b/skills/code-review/SKILL.md index 37438f6..b0a8565 100644 --- a/skills/code-review/SKILL.md +++ b/skills/code-review/SKILL.md @@ -1,6 +1,8 @@ --- name: code-review -description: Reviews a fixed code change through isolated specification and engineering-standards agents. Use for direct diff review or when another delivery skill invokes independent review. +description: Reviews a fixed code change through isolated specification and engineering-standards agents. Use when a fixed candidate needs independent diagnostic review. +metadata: + type: utility disable-model-invocation: true --- @@ -34,7 +36,7 @@ Direct the Standards reviewer to understand every changed line in its necessary - repository-required and changed-risk harness selection, measurement-path integrity, and honest evidence limits; and - recognised maintainability shapes and specialist risks exposed by the change. -Require the Standards reviewer to load and apply `$modular-design`, plus `$tdd` when its prerequisites hold. These skills supply teaching knowledge within the Standards review; they do not become delegated workflows or own findings, commands, remediation, or a verdict. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. +Require the Standards reviewer to load and apply `$modular-design` and `$test-design`. When a concrete maintainability shape needs recognised vocabulary, load [Code Smells](references/CODE-SMELLS.md); use it to investigate a mechanism and consequence, never as finding authority or a removal checklist. Require the Standards reviewer to invoke `$measure-code-complexity` against the fixed candidate and comparison base with read-only authority. Treat its current output as diagnostic evidence: inspect each trigger in source and test context, and retain a review suggestion only when a concrete maintainability consequence survives falsification. Report unavailable or incomplete measurement in the review limitations. diff --git a/skills/commit/SKILL.md b/skills/commit/SKILL.md index 394f29e..907fd6a 100644 --- a/skills/commit/SKILL.md +++ b/skills/commit/SKILL.md @@ -1,6 +1,8 @@ --- name: commit -description: Creates atomic Conventional Commits for eligible reviewed work while preserving unrelated changes. Use when the user asks to commit completed work. +description: Creates atomic Conventional Commits for eligible reviewed work while preserving unrelated changes. Use when completed local work is ready to record. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/create-tickets/SKILL.md b/skills/create-tickets/SKILL.md index 764c929..f2906ab 100644 --- a/skills/create-tickets/SKILL.md +++ b/skills/create-tickets/SKILL.md @@ -1,6 +1,8 @@ --- name: create-tickets description: Creates implementation-ready tickets from approved work. Use when a complete work definition needs vertical decomposition in the project's configured ticket system. +metadata: + type: performer disable-model-invocation: true --- @@ -10,13 +12,13 @@ Turns one **approved work definition** into a confirmed, verified set of small v ## Prerequisites -Read the root `AGENTS.md` for an explicit ticket destination such as `The project uses Linear for tickets.` or `The project uses local Markdown for tickets.` When absent, ask the user which destination the project uses, invoke `$maintain-agents` with the answer, verify the instruction, and resume. For an external destination, require its installed integration and writable account; otherwise stop with the exact access needed. Never guess or silently fall back. +Read the root `AGENTS.md` for an explicit ticket destination such as `The project uses Linear for tickets.` or `The project uses local Markdown for tickets.` When absent, ask the user which destination the project uses, invoke `$maintain-agents` with the answer, and resume from its result. For an external destination, require its installed integration and writable account; otherwise stop with the exact access needed. Never guess or silently fall back. ## Process ### 1. Establish the work and destination -Inspect the request, approved work definition, repository guidance, linked product, context, decision, research, and current-system evidence needed to understand the work. Treat the definition as authority for intended outcomes and settled constraints, and the repository as evidence of the current system. +Use the already-loaded root guidance. Inspect the request, approved work definition, and only the additional linked product, context, decision, research, and current-system evidence needed to understand the work. Treat the definition as authority for intended outcomes and settled constraints, and the repository as evidence of the current system. For an external destination, inspect its native item, body, status, estimate, parent, dependency, and read-back capabilities before planning writes. For local Markdown, use one file per ticket under `docs/features//tickets/`. diff --git a/skills/debug/SKILL.md b/skills/debug/SKILL.md index 5be7e8d..6de1def 100644 --- a/skills/debug/SKILL.md +++ b/skills/debug/SKILL.md @@ -1,6 +1,8 @@ --- name: debug description: Diagnoses software defects and conditionally makes causal, verified repairs. Use for failures, regressions, runtime errors, flaky behaviour, or performance faults. +metadata: + type: performer disable-model-invocation: true --- diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index 9e4651e..bf8bbbc 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -1,6 +1,8 @@ --- name: define-product description: Defines a breadth-first product and system requirements foundation. Use when externalising a new product idea or deliberately revising PRODUCT.md. +metadata: + type: performer disable-model-invocation: true --- @@ -46,7 +48,7 @@ Record assumptions, dependencies, constraints, risks, and open questions distinc Walk every intended user and lifecycle stakeholder across the journeys, capabilities, and applicable technical requirements. Resolve material omissions, duplicates, broken transitions, contradictory ownership, untraceable requirements, hidden feature depth, and requirements that prescribe unconfirmed design. Check that every normative requirement is necessary, feasible enough for this stage, unambiguous at its level, source-traceable, and capable of later verification. -Have `$elicit-with-context` present one complete synthesis covering the entire proposed foundation, non-blocking unresolved items, and exact document effect. Obtain explicit confirmation of the whole synthesis. If the user rejects or corrects it, leave `PRODUCT.md` unchanged, reopen every affected branch, repeat the relevant coverage walk, and confirm the complete revision. The confirmed synthesis authorises its agreed `PRODUCT.md` write without another permission prompt. +Invoke `$elicit-with-context` with the proposed foundation, non-blocking unresolved items, and exact document effect. Use its confirmed synthesis as authority for the agreed `PRODUCT.md` write. If coverage changes afterward, resolve the affected foundation again before writing. ### 5. Write the foundation diff --git a/skills/elicit-with-context/SKILL.md b/skills/elicit-with-context/SKILL.md index 841f651..995ca0b 100644 --- a/skills/elicit-with-context/SKILL.md +++ b/skills/elicit-with-context/SKILL.md @@ -1,6 +1,8 @@ --- name: elicit-with-context -description: Elicits shared understanding while maintaining software-project context. Use when a software-project skill needs user-held information or decisions. +description: Elicits shared understanding while maintaining software-project context. Use when software-project work depends on material user-held information or decisions. +metadata: + type: router disable-model-invocation: true --- @@ -10,4 +12,4 @@ Elicits shared understanding while maintaining project context. ## Process -Invoke `$elicit` for questioning, applying `$maintain-ubiquitous-language` and `$maintain-decision-records` throughout. +Invoke `$elicit` for questioning. Apply `$maintain-ubiquitous-language` throughout so stable project terms stay current. Invoke `$maintain-decision-records` only when its description applies. Return `$elicit`'s confirmed synthesis. diff --git a/skills/elicit/SKILL.md b/skills/elicit/SKILL.md index e5c2813..b5a855c 100644 --- a/skills/elicit/SKILL.md +++ b/skills/elicit/SKILL.md @@ -1,6 +1,8 @@ --- name: elicit -description: Elicits material information and decisions from the user one question at a time until shared understanding is confirmed. Use when a request or consuming skill depends on user-held context, clarification, or choices that cannot be safely discovered. +description: Elicits material information and decisions one question at a time until shared understanding is confirmed. Use when user-held context, clarification, or choices block a bounded outcome. +metadata: + type: utility disable-model-invocation: false --- diff --git a/skills/implement/SKILL.md b/skills/implement/SKILL.md index 4e42e60..37847f2 100644 --- a/skills/implement/SKILL.md +++ b/skills/implement/SKILL.md @@ -1,6 +1,8 @@ --- name: implement -description: Implements one ticket or small confirmed software change as a minimal, locally checked, independently reviewed candidate. Use when the user asks to build agreed work locally. +description: Implements one ticket or small confirmed software change as a minimal, locally checked, independently reviewed candidate. Use when agreed software work is ready to build locally. +metadata: + type: performer disable-model-invocation: true --- @@ -18,7 +20,7 @@ Keep the outcome to one implementation-ready ticket, diagnosed repair, or small ### 2. Load the teaching knowledge -Invoke `$modular-design` against the fixed work. Invoke `$tdd` when the change affects observable behaviour and an existing usable suite can exercise it through a credible seam. Retain their applicable guidance throughout implementation; they teach within this workflow while this skill retains ownership of edits, commands, evidence, and completion. +Invoke `$modular-design` against the fixed work. Invoke `$tdd` when the change affects observable behaviour and an existing usable suite can exercise it through a credible seam. Apply their guidance throughout implementation. ### 3. Build and check the candidate @@ -26,7 +28,7 @@ Implement the smallest coherent change that completely satisfies the confirmed o Inspect the complete diff and every in-scope untracked file. Account for each production and test addition, dependency, generated artefact, and measurement-path change. Run applicable repository-required checks. Preserve exact results, unavailable evidence, and residual risks. -After applicable checks are current, invoke `$measure-code-complexity` with the fixed candidate, comparison base, and implementation authority. Resolve its returned dispositions by applying authorised coherent improvements through the applicable teaching skills, retaining supported justifications, or invoking `$elicit-with-context` when resolution would change agreed behaviour, contracts, architecture, or scope. Rerun every affected harness and the measurement after a change. Freeze one complete candidate only when its checks and complexity measurement are current and every trigger is resolved or explicitly retained for handoff. +After applicable checks are current, invoke `$measure-code-complexity` with the fixed candidate, comparison base, and implementation authority. Resolve each supported concern by applying an authorised coherent improvement, retaining a supported justification, or invoking `$elicit-with-context` when resolution would change agreed behaviour, contracts, architecture, or scope. Rerun every affected harness and invoke the measurement again after a change. Freeze one complete candidate only when its checks and complexity measurement are current and every trigger is resolved or explicitly retained for handoff. ### 4. Review and adjudicate @@ -37,7 +39,7 @@ Invoke `$code-review` on the frozen candidate and its authorities. Validate ever - **Rejected finding:** falsified, unsupported, superseded, or outside the confirmed scope. - **User decision:** resolution would change agreed behaviour, contracts, architecture, or scope, or needs evidence that cannot be obtained safely. -Apply required corrections and accepted improvements through the applicable teaching skills. Invoke `$elicit-with-context` for a user decision. Rerun every affected harness after a candidate or measurement-path change. Freeze and review the complete new candidate after any material change. Continue until every finding has a supported disposition and no applicable correction or accepted improvement remains. +Apply required corrections and accepted improvements. Invoke `$elicit-with-context` for a user decision. Rerun every affected harness after a candidate or measurement-path change. Freeze and review the complete new candidate after any material change. Continue until every finding has a supported disposition and no applicable correction or accepted improvement remains. ### 5. Hand off the local result diff --git a/skills/maintain-agents/SKILL.md b/skills/maintain-agents/SKILL.md index db21b8c..423386d 100644 --- a/skills/maintain-agents/SKILL.md +++ b/skills/maintain-agents/SKILL.md @@ -1,6 +1,8 @@ --- name: maintain-agents description: Maintains concise root AGENTS.md instructions. Use when creating the file or adding, refining, or assessing repository-wide agent guidance. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/maintain-decision-records/SKILL.md b/skills/maintain-decision-records/SKILL.md index f1c916a..101cf9d 100644 --- a/skills/maintain-decision-records/SKILL.md +++ b/skills/maintain-decision-records/SKILL.md @@ -1,6 +1,8 @@ --- name: maintain-decision-records -description: Maintains rare architecture decision records. Use when explicitly invoked to test and preserve an accepted technical or architectural decision's durable rationale. +description: Maintains rare architecture decision records. Use when an accepted technical or architectural decision may warrant durable rationale. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/maintain-ubiquitous-language/SKILL.md b/skills/maintain-ubiquitous-language/SKILL.md index affcd2f..cd27424 100644 --- a/skills/maintain-ubiquitous-language/SKILL.md +++ b/skills/maintain-ubiquitous-language/SKILL.md @@ -1,6 +1,8 @@ --- name: maintain-ubiquitous-language -description: Maintains project-specific language in CONTEXT.md. Use when explicitly invoked to add, change, challenge, rename, deprecate, or reconcile a project term. +description: Maintains project-specific language in CONTEXT.md. Use when project terminology must be added, changed, challenged, renamed, deprecated, or reconciled. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/measure-code-complexity/SKILL.md b/skills/measure-code-complexity/SKILL.md index 448de92..add0e04 100644 --- a/skills/measure-code-complexity/SKILL.md +++ b/skills/measure-code-complexity/SKILL.md @@ -1,6 +1,8 @@ --- name: measure-code-complexity description: Measures changed-code complexity with pinned cross-language metrics and advisory review triggers. Use when implementing, reviewing, or directly assessing code complexity. +metadata: + type: utility disable-model-invocation: false --- @@ -34,11 +36,11 @@ For each triggered metric, load only its named reference: [Cyclomatic Complexity Disposition every trigger as one of: -- **Improved:** apply an authorised, coherent simplification; rerun affected tests and this measurement. -- **Justified retention:** retain the shape with a concrete code- and test-based reason. +- **Supported concern:** state the concrete maintainability consequence, evidence, and narrow improvement direction. +- **Justified shape:** state the concrete code- and test-based reason the shape is appropriate. - **Unresolved:** state the missing authority or evidence and leave the trigger open. -Respect the caller's boundary. An implementation caller may edit and rerun; a review caller remains read-only and returns an evidenced suggestion when the consequence warrants one. +Measure and interpret only. Leave every source edit, test run, and decision to apply an improvement to the caller. ### 3. Return the evidence @@ -49,4 +51,4 @@ Return the fixed scope, compact summary, triggered locations and values, baselin - Keep the bundled metric definitions and bands intact unless explicit repository policy overrides a threshold. - Preserve all metric components; do not combine them into Maintainability Index, CRAP, or another scalar grade. - Reject score-only changes that merely move branching, add indirection, hide parameters, or replace one clone with a shallow abstraction. -- Stop after every trigger has a supported disposition or is explicitly unresolved. Do not create repository policy, durable reports, commits, pull requests, or external state unless the caller separately authorises them. +- Stop after every trigger has a supported disposition or is explicitly unresolved. Do not modify source or tests, run implementation checks, create repository policy or durable reports, commit, publish, or change external state. diff --git a/skills/modular-design/SKILL.md b/skills/modular-design/SKILL.md index e005c07..935991c 100644 --- a/skills/modular-design/SKILL.md +++ b/skills/modular-design/SKILL.md @@ -1,6 +1,8 @@ --- name: modular-design description: Teaches structural design through information hiding, deep modules, cohesive ownership, stable interfaces, and change locality. Use when writing or assessing code, boundaries, contracts, dependencies, seams, or change propagation. +metadata: + type: teaching disable-model-invocation: true --- @@ -12,7 +14,7 @@ Teaches a caller to make the smallest coherent change through deep modules whose ### 1. Fix the structural basis -Read the confirmed behaviour, caller outcome, repository guidance, domain language, architecture decisions, relevant source, contracts, and tests. Follow explicit project authorities and established language and framework idioms when they conflict with this baseline. +Use the caller's confirmed behaviour, outcome, repository guidance, domain language, architecture decisions, relevant source, contracts, and tests. Inspect only evidence that is missing or stale. Follow explicit project authorities and established language and framework idioms when they conflict with this baseline. Treat a **module** as any cohesive capability with an interface and implementation: a function, object, component, package, process, service, or tier-spanning slice. Treat its **interface** as everything callers must know, including operations, data, invariants, ordering, errors, configuration, effects, and material performance characteristics. Do not equate modules with files, classes, layers, or deployment units. @@ -75,7 +77,7 @@ Trace one plausible current-project change through the proposed owner, interface Reconsider the design when the same decision remains duplicated, elements that always change together are split, an owner changes mostly for unrelated reasons, callers still coordinate hidden sequencing, or one conceptual change requires scattered edits. Do not substitute file size, class count, method count, churn, coupling scores, complexity thresholds, or speculative agent-context savings for this trace. -Finish with the owner, caller-visible contract, hidden decisions, locality trace, triggered technique or material comparison, and unresolved structural uncertainty. The caller retains its outcome, edits, commands, tests, verification, review findings, artefacts, and stopping decision. +Keep the owner, caller-visible contract, hidden decisions, locality trace, triggered technique or material comparison, and unresolved structural uncertainty explicit in the caller's workflow. ## Rules diff --git a/skills/pull-request/SKILL.md b/skills/pull-request/SKILL.md index 4e29f4a..3981cc5 100644 --- a/skills/pull-request/SKILL.md +++ b/skills/pull-request/SKILL.md @@ -1,6 +1,8 @@ --- name: pull-request description: Publishes the current branch as one concise verified pull request. Use when the branch is ready for external review. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/research/SKILL.md b/skills/research/SKILL.md index e912bb6..9134b21 100644 --- a/skills/research/SKILL.md +++ b/skills/research/SKILL.md @@ -1,6 +1,8 @@ --- name: research -description: Investigates a material subject against high-trust sources and writes a trusted cited report. Use when a user or calling skill intentionally requests durable research. +description: Investigates a material subject against high-trust sources and writes a trusted cited report. Use when a material question needs durable external evidence. +metadata: + type: utility disable-model-invocation: true --- diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index a49cdda..e1da368 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -1,6 +1,8 @@ --- name: review-architecture description: Reviews a whole codebase for high-impact evidence-backed modular-design improvements and writes an interactive HTML report. Use for architecture or structural maintainability reviews. +metadata: + type: performer disable-model-invocation: true --- @@ -16,7 +18,7 @@ Inspect the request, repository guidance, root `PRODUCT.md` and `CONTEXT.md`, ap Inspect relevant Git history after mapping the repository. Use repeated cross-file changes and hotspots as supporting evidence and ranking signals without excluding quieter areas. Distinguish project authority, current observation, inference, and unknowns. -Invoke `$modular-design` with the authorities and current evidence. Apply its structural teaching while retaining ownership of investigation, candidate qualification, ranking, the report, and stopping. Invoke `$research` only when a material candidate depends on external framework, platform, or technical evidence that the repository cannot establish; read its report before deciding whether the candidate qualifies. +Apply `$modular-design` to the authorities and current evidence. Invoke `$research` only when a material candidate depends on external framework, platform, or technical evidence that the repository cannot establish; read its report before deciding whether the candidate qualifies. ### 2. Retain only top-level candidates diff --git a/skills/review-pull-request/SKILL.md b/skills/review-pull-request/SKILL.md index 7349343..7f563e8 100644 --- a/skills/review-pull-request/SKILL.md +++ b/skills/review-pull-request/SKILL.md @@ -1,29 +1,17 @@ --- name: review-pull-request -description: Reviews a supplied pull request through independent specification and engineering-standards assessment. Use when the user wants a read-only pull-request report. +description: Reviews a supplied pull request through independent specification and engineering-standards assessment. Use when a pull request needs a read-only independent report. +metadata: + type: router disable-model-invocation: true --- # Review Pull Request -Resolves one pull request and routes its complete change through `$code-review`, then returns the independent findings without changing the pull request or repository. +Routes one fixed pull request through `$code-review` and returns its independent findings unchanged. ## Process -### 1. Resolve the pull request +Resolve the supplied pull-request identifier against the current repository unless another repository is named. Use native host metadata to fix its repository, number or URL, and exact base and head revisions. Return the exact blocker when the pull request is ambiguous or inaccessible. -Resolve the supplied pull-request identifier against the current repository unless the user names another repository. Use available native host tooling to inspect its metadata, commits, complete patch and changed files, relevant surrounding code, discussion, existing check evidence, repository guidance, linked specification or work item, and applicable product and technical authorities. - -Fix one exact, non-empty pull-request candidate with its repository, number or URL, base and head revisions, complete changed paths, and authorities. Return the exact blocker when the pull request is ambiguous, inaccessible, or incomplete. Do not infer intended behaviour from the implementation, commit messages, or implementation rationale. - -### 2. Invoke the independent review - -Invoke `$code-review` with the fixed candidate, complete pull-request context, intended-behaviour authorities when available, applicable Standards authorities, existing check evidence, and a read-only boundary. When no implementation-independent behavioural authority exists, let `$code-review` omit Spec and continue Standards. - -Treat existing checks as supplied evidence. Leave focused review probes to `$code-review`; do not run a separate review, verification workflow, or full project check suite. - -### 3. Return the report - -State the pull-request identity and exact reviewed scope, then return `$code-review`'s separate Standards and Spec results, priorities, findings, evidence, limitations, omitted axes, and counts without substantive rewriting or adjudication. - -Stop after the report. Do not modify files or branches, publish comments, change pull-request or tracker state, approve, request changes, merge, commit, push, release, or deploy. +Invoke `$code-review` with that fixed pull-request candidate and a read-only boundary. Return its result unchanged. diff --git a/skills/specify-feature/SKILL.md b/skills/specify-feature/SKILL.md index 8a012e9..fed97ba 100644 --- a/skills/specify-feature/SKILL.md +++ b/skills/specify-feature/SKILL.md @@ -1,6 +1,8 @@ --- name: specify-feature description: Creates a decision-complete specification for one feature. Use when a high-level feature request or capability needs collaborative definition before ticket creation. +metadata: + type: performer disable-model-invocation: true --- @@ -14,7 +16,7 @@ Turns one high-level feature description into one user-confirmed specification c Inspect the request and the project evidence relevant to the feature. Resolve discoverable facts before asking the user, and distinguish evidence of the current system from intended future behaviour. -Invoke `$elicit-with-context` whenever material user-held information or a decision is needed. Bring it the evidence, gaps, viable options and consequences. The user decides the feature's direction; make informed recommendations, challenge ambiguity and expose overlooked paths without turning a suggestion into a decision. +Invoke `$elicit-with-context` with the relevant evidence and unresolved decision whenever material user-held information or a decision is needed. Use its confirmed synthesis as authority for the feature direction. Invoke `$research` when durable external evidence is needed to resolve the feature. Use its report as evidence, not as authority for a product or solution decision. diff --git a/skills/tdd/SKILL.md b/skills/tdd/SKILL.md index 7712cee..512e0d6 100644 --- a/skills/tdd/SKILL.md +++ b/skills/tdd/SKILL.md @@ -1,6 +1,8 @@ --- name: tdd description: Teaches test-driven development through red-green-refactor with durable behavioural tests. Use when a usable suite can exercise changed behaviour through a credible seam. +metadata: + type: teaching disable-model-invocation: true --- @@ -10,27 +12,25 @@ Teaches a caller to implement one observable behaviour at a time and retain a sm ## Prerequisites -Apply TDD only when the work adds, changes, or repairs observable behaviour and an existing usable suite can exercise it through a credible seam. Otherwise return the missing condition to the caller; proceed through the strongest project-native feedback without claiming TDD. Create a suite or framework only when that infrastructure is part of the agreed work. +Apply TDD only when the work adds, changes, or repairs observable behaviour and an existing usable suite can exercise it through a credible seam. Otherwise surface the missing condition so the caller can use the strongest project-native feedback without claiming TDD. Create a suite or framework only when that infrastructure is part of the agreed work. ## Process ### 1. Fix the slice, seam, and oracle -Read the confirmed behaviour, repository instructions, applicable modular constraints, relevant production code, and current tests. Identify the focused command and run its baseline. Keep pre-existing failures visible; continue only when they cannot hide the next Red or Green signal. +Use the caller's confirmed behaviour, applicable modular constraints, and current production and test evidence. Inspect only evidence that is missing or stale. Identify the focused command and run its baseline. Keep pre-existing failures visible; continue only when they cannot hide the next Red or Green signal. -Select a **tracer bullet**: the smallest complete vertical slice that produces one requested observable outcome. Choose the narrowest seam that exposes the promise, hides changeable structure, and retains the production risk that matters. Derive the expected result before Green from a requirement, accepted example, contract, invariant, trusted reference, separate model, or accepted prior behaviour. Current output is characterization, not proof of correctness. Load [Test Design](references/TEST-DESIGN.md) and apply its boundary, oracle, determinism, and durability rules. +Select a **tracer bullet**: the smallest complete vertical slice that produces one requested observable outcome. Apply `$test-design` to choose its seam, oracle, test shape, and any conditional technique. Current output is characterization, not proof of correctness. -If the caller already wrote this production behaviour, discard only caller-owned uncommitted implementation and restart test-first. Preserve pre-existing and user-owned work. For an existing partial implementation, demonstrate sensitivity against the pre-change revision, a safely disabled behaviour, or a controlled known-bad variant. When none is safe, retain an independently meaningful test as regression or characterization evidence without calling the work TDD. +For an existing partial implementation, demonstrate sensitivity against the pre-change revision, a safely disabled behaviour, or a controlled known-bad variant. When none is safe, retain an independently meaningful test as regression or characterization evidence without calling the work TDD. ### 2. Red -Write one focused behavioural test before its production change. Arrange only necessary state, act once through the supported seam, assert the complete promised outcome, and clean up acquired resources. Assert unchanged state only when its preservation is promised and the action can actually affect the observed value. Name one realistic production break the test must detect and one hidden structural change it must survive. - -Keep internal collaborators real. When an uncontrollable boundary needs substitution or observation, load [Test Doubles](references/TEST-DOUBLES.md), choose the least powerful double, and assert only contractually meaningful effects. +Write one focused behavioural test before its production change, following the selected test design. For a bug, reproduce the incorrect result and assert the independently established desired result. Structure the test so failure is reported against the promise: an unexpected exception that aborts before any oracle assertion is not Red unless the test explicitly establishes that successful completion is the promise. When count or order at a system boundary is promised, assert it directly rather than inferring it from final state. -Run the focused test. Red exists only when an oracle fails for the intended missing or incorrect behaviour; a pass, syntax error, fixture failure, or environment error is invalid. Correct an in-scope test or environment defect and rerun, or return the blocker. Preserve the exact Red command and failure. +Run the focused test. Red exists only when an oracle fails for the intended missing or incorrect behaviour; a pass, syntax error, fixture failure, or environment error is invalid. Correct an in-scope test or environment defect and rerun, or surface the blocker in the caller's workflow. Preserve the exact Red command and failure. ### 3. Green @@ -40,16 +40,14 @@ Add only enough production code for the current behaviour. Treat confirmed preco Improve production naming, cohesion, duplication, ownership, and interfaces without adding behaviour. Keep behavioural tests unchanged through production refactors. When structural movement alone breaks a test, move its observation back to the promised outcome. Clarify test names or fixtures only without weakening the oracle or counterfactuals. Run the focused tests after each material step. -If refactoring exposes a new material structural decision not resolved by the caller's modular constraints, return it to the caller for resolution before continuing. +If refactoring exposes a new material structural decision not resolved by the caller's modular constraints, pause the cycle until the caller resolves it. ### 5. Rationalise the retained suite -After production refactoring, identify the unique behaviour, boundary, invariant, or risk protected by each affected test. Remove obsolete implementation-detail assertions, broad snapshots used only as change alarms, and examples that add no distinct failure detection. Consolidate equivalent cases with a table or property only when the resulting oracle and failure diagnosis remain clear. - -Mentally mutate wrong constants, branches, validation, effects, and empty or default results. Strengthen the smallest test or input that would let a realistic defect survive. +After production refactoring, reapply `$test-design` to the affected suite while it remains Green. -### 6. Complete and return +### 6. Complete the cycles -Repeat Red, Green, Refactor, and rationalisation for each remaining confirmed behaviour. When example cases cannot economically cover a large input or state space, or a direct oracle is unavailable, load [Generative Testing](references/GENERATIVE-TESTING.md) and use only the triggered technique. Run the complete relevant suite. +Repeat Red, Green, Refactor, and rationalisation for each remaining confirmed behaviour. Run the complete relevant suite. -Return the behaviours delivered; retained tests and the unique protection each adds; exact Red and Green evidence; refactors; commands and results; unresolved baseline failures; fidelity limits; and other missing evidence. The caller retains ownership of broader checks, independent review, the complete work outcome, commits, and publication. +Keep the delivered behaviours, retained tests and their unique protection, exact Red and Green evidence, refactors, commands and results, unresolved baseline failures, fidelity limits, and other missing evidence current in the caller's workflow. diff --git a/skills/tdd/references/TEST-DESIGN.md b/skills/tdd/references/TEST-DESIGN.md deleted file mode 100644 index 8713180..0000000 --- a/skills/tdd/references/TEST-DESIGN.md +++ /dev/null @@ -1,79 +0,0 @@ -# Test Design - -Apply this guide to every retained test. - -## Select a durable seam - -Use the narrowest seam that is: - -- **Observable:** exposes a supported result, state, error, effect, persistence, navigation, or user-perceivable outcome. -- **Stable:** hides algorithms, collaborator graphs, storage layouts, rendering wrappers, generated selectors, and other replaceable decisions. -- **Predictive:** exercises the behaviour and technology whose production failure matters. - -Add a broader companion test only when the narrower seam cannot retain a material integration risk. Test a pure rule through its public function, a database guarantee against the real engine when its semantics matter, and an end-user interaction through the rendered application when wiring or accessibility is part of the promise. - -| Concern | Brittle or circular test | Durable behavioural test | -| -------------- | -------------------------------------------------------------- | --------------------------------------------------------------------------------------- | -| Expected value | Repeats the production reduction to calculate `expected` | Uses the accepted example `10 + 5 = 15` or an independent invariant | -| Domain result | Calls a private pricing method or asserts helper delegation | Calls the supported quote operation and asserts its public result | -| UI | Asserts CSS classes, wrapper tags, child indexes, or DOM depth | Interacts by role or label and asserts visible text, value, focus, state, or navigation | -| Persistence | Inspects private tables, columns, or ORM calls | Writes and reads through supported interfaces; reloads when durability matters | -| Error | Asserts a private branch or stack shape | Asserts caller-visible type, code, promised message, recovery, and unchanged state | -| Snapshot | Accepts a broad incidental object or markup snapshot | Asserts selected semantic fields, unless exact bytes or markup are the contract | -| Bug | Treats the existing crash or output as the expected result | Reproduces the wrong result, then asserts a requirement-derived correction | - -Representation is valid only when that representation is itself a published contract. - -## Build an independent oracle - -Prefer, in order: - -1. a confirmed requirement, accepted worked example, published protocol, or domain decision; -2. a law, invariant, or contract independent of the production algorithm; -3. a trusted external reference or separately implemented model; or -4. accepted prior behaviour when preservation is the requirement. - -Encode the expectation as an independently reasoned literal or predicate. Never call the production helper, copy its algorithm, approve newly captured output, or assert only that execution completed unless completion is the entire promise. When no credible oracle exists, return the missing behavioural decision. - -## Make the test deterministic and complete - -Arrange local explicit state, perform one meaningful action, assert the complete promised outcome, then release acquired resources. Add a negative or unchanged-state assertion only when the requirement makes it material and the action has a real causal path to the observed value. Do not clone, replace, or otherwise isolate that value before the action and then claim its preservation as evidence. One behaviour may need several cohesive assertions; split multiple independent actions or rules. - -Control time and randomness through an injected clock or deterministic source at the system boundary. Use fixed instants and seeds and preserve a minimized failing counterexample. Give each test isolated data and restore process state, files, transactions, services, timers, and environment values it changes. Avoid sleeps, execution order dependencies, live external services, and shared mutable fixtures. - -```typescript -// Brittle: wall-clock timing and an assertion too weak to catch the wrong expiry. -expect(issueToken().expiresAt).toBeGreaterThan(Date.now()); - -// Durable: controlled time and the exact promised interval. -expect(issueToken({ now: fixedClock('2030-01-01T00:00:00Z') }).expiresAt).toBe( - '2030-01-01T01:00:00Z', -); -``` - -## Challenge and rationalise - -Ask both counterfactuals before retaining a test: - -- If promised behaviour broke, would this test fail? -- If only hidden structure changed, would this test remain unchanged and pass? - -After refactoring, map every affected test to unique protection. A focused example remains useful when it specifies a distinct rule or gives a clearer failure than a consolidated form. Otherwise remove duplicates or combine them: - -```typescript -// Repetition with no distinct structure. -expect(priceFor(1)).toBe(10); -expect(priceFor(2)).toBe(20); -expect(priceFor(3)).toBe(30); - -// One rule with the same protection and clearer extension. -for (const [quantity, expected] of [ - [1, 10], - [2, 20], - [3, 30], -]) { - expect(priceFor(quantity)).toBe(expected); -} -``` - -Do not consolidate cases with materially different setup, outcome, boundary, or diagnostic value. Fewer tests are better only when no unique defect detection is lost. diff --git a/skills/test-design/SKILL.md b/skills/test-design/SKILL.md new file mode 100644 index 0000000..e4ad25c --- /dev/null +++ b/skills/test-design/SKILL.md @@ -0,0 +1,59 @@ +--- +name: test-design +description: Teaches durable behavioural test design through stable seams, independent oracles, determinism, and counterfactual challenge. Use when writing, reviewing, or rationalising tests. +metadata: + type: teaching +disable-model-invocation: true +--- + +# Test Design + +Teaches a caller to retain tests that detect promised-behaviour defects without resisting hidden structural change. + +## Process + +### 1. Fix the behavioural basis + +Use the caller's confirmed behaviour, applicable contracts, and current production and test evidence. Inspect only evidence that is missing or stale. Treat representation as a valid oracle only when that representation is itself a published contract. + +### 2. Select a durable seam + +Use the narrowest seam that is: + +- **Observable:** exposes a supported result, state, error, effect, persistence, navigation, or user-perceivable outcome. +- **Stable:** hides algorithms, collaborator graphs, storage layouts, rendering wrappers, generated selectors, and other replaceable decisions. +- **Predictive:** exercises the behaviour and technology whose production failure matters. + +Add a broader companion test only when the narrower seam cannot retain a material integration risk. Test a pure rule through its public function, a database guarantee against the real engine when its semantics matter, and an end-user interaction through the rendered application when wiring or accessibility is part of the promise. + +Prefer supported operations and observable outcomes over private methods, helper delegation, ORM calls, CSS classes, wrapper tags, DOM depth, stack shape, or broad incidental snapshots. For a defect, reproduce the wrong result and assert a requirement-derived correction rather than treating the existing failure as the expectation. + +When a slow, unavailable, destructive, non-deterministic, externally mutating, or otherwise uncontrollable boundary needs substitution or observation, load [Test Doubles](references/TEST-DOUBLES.md). Keep internal collaborators real and do not introduce a production abstraction solely to enable mocking. + +### 3. Build an independent oracle + +Prefer, in order: + +1. a confirmed requirement, accepted worked example, published protocol, or domain decision; +2. a law, invariant, or contract independent of the production algorithm; +3. a trusted external reference or separately implemented model; or +4. accepted prior behaviour when preservation is the requirement. + +Encode the expectation as an independently reasoned literal or predicate. Do not call the production helper, copy its algorithm, approve newly captured output, or assert only that execution completed unless completion is the entire promise. When no credible oracle exists, surface the missing behavioural decision in the caller's workflow. + +When examples cannot economically cover a large input or state space, or a direct expected output is unavailable but an independent relation, model, or comparator exists, load [Generative Testing](references/GENERATIVE-TESTING.md). Use only the technique whose trigger and oracle hold. + +### 4. Make the test deterministic and complete + +Arrange local explicit state, perform one meaningful action, assert the complete promised outcome, then release acquired resources. Add a negative or unchanged-state assertion only when the requirement makes it material and the action has a real causal path to the observed value. One behaviour may need several cohesive assertions; split multiple independent actions or rules. + +Control time and randomness through a deterministic source at the system boundary. Use fixed instants and seeds and preserve a minimised failing counterexample. Give each test isolated data and restore process state, files, transactions, services, timers, and environment values it changes. Avoid sleeps, execution-order dependencies, live external services, and shared mutable fixtures. + +### 5. Challenge and rationalise + +Apply both counterfactuals before retaining a test: + +- If promised behaviour broke, would this test fail? +- If only hidden structure changed, would this test remain unchanged and pass? + +Map every affected test to unique protection after refactoring. Keep a focused example when it specifies a distinct rule or gives a clearer failure. Otherwise remove duplicates or consolidate cases only when the resulting oracle and failure diagnosis remain clear. Fewer tests are better only when no unique defect detection is lost. diff --git a/skills/test-design/agents/openai.yaml b/skills/test-design/agents/openai.yaml new file mode 100644 index 0000000..bd620c2 --- /dev/null +++ b/skills/test-design/agents/openai.yaml @@ -0,0 +1,5 @@ +interface: + display_name: 'Test Design' + short_description: 'Design durable behavioural tests' +policy: + allow_implicit_invocation: false diff --git a/skills/tdd/references/GENERATIVE-TESTING.md b/skills/test-design/references/GENERATIVE-TESTING.md similarity index 100% rename from skills/tdd/references/GENERATIVE-TESTING.md rename to skills/test-design/references/GENERATIVE-TESTING.md diff --git a/skills/tdd/references/TEST-DOUBLES.md b/skills/test-design/references/TEST-DOUBLES.md similarity index 100% rename from skills/tdd/references/TEST-DOUBLES.md rename to skills/test-design/references/TEST-DOUBLES.md diff --git a/skills/write-skill/SKILL.md b/skills/write-skill/SKILL.md index ceaf67f..b5d9f09 100644 --- a/skills/write-skill/SKILL.md +++ b/skills/write-skill/SKILL.md @@ -1,6 +1,8 @@ --- name: write-skill description: Creates and revises effective agent skills through confirmed requirements, minimum-sufficient design, and baseline-controlled evaluation. Use when authoring or maintaining a reusable agent skill. +metadata: + type: performer disable-model-invocation: true --- @@ -12,7 +14,7 @@ Creates minimum-sufficient agent skills whose observed behaviour justifies their ### 1. Confirm the contract -Inspect the request, an existing skill and its direct composition when revising, and task-scoped evidence without changing the target. Invoke `$elicit` to confirm the recurring job or capability gap; user and model triggers plus non-triggers; inputs and prerequisites; required actions, material branches, authority, and stopping boundary; expected output or handoff; direct composition; necessary resources; and observable success evidence. Keep the target unchanged until `$elicit` returns the confirmed synthesis. The authoring contract is explicit and closed. +Inspect the request, an existing skill and its direct composition when revising, and task-scoped evidence without changing the target. Invoke `$elicit-with-context` in a software project and `$elicit` elsewhere to confirm the recurring job or capability gap; task triggers and non-triggers; skill type; inputs and prerequisites; required actions, material branches, authority, and stopping boundary; expected output or handoff; direct composition; necessary resources; and observable success evidence. Keep the target unchanged until elicitation returns the confirmed synthesis. The authoring contract is explicit and closed. ### 2. Establish the matched baseline @@ -27,11 +29,20 @@ Record each result, loaded path, output, tool calls, and iterations. Design only Apply the **KISS principle**: retain the common path and material branches; remove speculative edge cases, unnecessary artefacts, and ceremony that cannot change the evaluated outcome. Keep coupled actions and artefacts together when they serve one outcome. Split independently useful capabilities when their trigger families, outcomes, authority, or success boundaries differ. The candidate has one coherent behavioural purpose. +Classify the skill by the authority it owns, not by whether it calls tools or fresh agents: + +- A **performer** owns one bounded requested outcome end to end, including actions, artefacts, verification, handoff, and stopping. +- A **router** resolves inputs and coordinates other skills' public contracts, adding only routing unique to the combined outcome. +- A **teaching** skill supplies knowledge and decision rules inside a caller's workflow without owning an independent operation, mutation, artefact, verification, handoff, or stopping decision. +- A **utility** owns one bounded reusable operation and returns its result. + +Set `metadata.type` to exactly `performer`, `router`, `teaching`, or `utility`. Remove responsibilities that belong to another type or split them when each capability is independently useful. + Choose invocation from actual use. Make regularly useful skills model-invoked. Make occasional, deliberate, or composition-only skills user-invoked. Set `disable-model-invocation` and `agents/openai.yaml` consistently. The skill enters context only through its intended route. ### 4. Design the loading path -Use a 1-64 character lowercase name with single hyphen separators and match the directory name. Write the description as one line that starts with a third-person action verb, says what the skill does, includes `Use when`, `Use for`, or `Use to`, and front-loads natural trigger words. Add a human-readable display name and short description to `agents/openai.yaml`. +Use a 1-64 character lowercase name with single hyphen separators and match the directory name. Write the description as one line that starts with a third-person action verb, says what the skill does, includes `Use when`, `Use for`, or `Use to`, and front-loads natural task triggers derived from its content. Describe the skill in isolation as something the agent may select; base its trigger on the work, not whether the user, caller, or another skill invoked it or the current invocation policy. Add a human-readable display name and short description to `agents/openai.yaml`. Use **progressive disclosure**: @@ -41,7 +52,7 @@ Use **progressive disclosure**: - Put reusable output material in `assets/`. - Put deterministic transformations and checks in `scripts/`. -Keep required coupled guidance together. Prefer direct one-level references. Remove unreachable and unlinked resources. Trace direct composition across trigger, inputs, authority, result, owner, and stopping boundary. The smallest sufficient context is reachable when needed. +Keep required coupled guidance together. Prefer direct one-level references. Remove unreachable and unlinked resources. Trace direct composition across trigger, inputs, authority, result, owner, and stopping boundary. A composed skill consumes current caller evidence and loads only missing or stale evidence; an independently invokable skill remains self-sufficient; a fresh agent receives a complete task-scoped packet. A router relies on each invoked skill's public contract and does not repeat its loading, process, or result handling. The smallest sufficient context is reachable when needed. ### 5. Draft the isolated candidate @@ -50,6 +61,7 @@ Create the candidate outside the requested destination. For revisions, start fro Use [the skill template](assets/skill-template.md). Load [the section guide](references/SECTIONS.md) only when optional sections or subsection layout are needed. Write commands, not essays: - Use direct verbs such as _Load_, _Check_, _Run_, _Fix_, _Stop_, and _Return_. +- Match commands to the declared type: performers direct the complete outcome; routers direct only input resolution, invocation, and result routing; teaching skills state principles and decisions for the caller to apply; utilities direct one operation and its result. - Give each step or subsection one coherent behavioural idea and one observable outcome. - Use short concrete words and clear fragments; expand conditions, risks, and handoffs that compression could hide. - State each normative meaning once. Remove filler, hedging, generic knowledge, and repeated enforcement. diff --git a/skills/write-skill/assets/skill-template.md b/skills/write-skill/assets/skill-template.md index 196bd21..1db761a 100644 --- a/skills/write-skill/assets/skill-template.md +++ b/skills/write-skill/assets/skill-template.md @@ -1,6 +1,8 @@ --- name: skill-name description: Creates the intended outcome. Use when the skill's confirmed trigger applies. +metadata: + type: performer disable-model-invocation: true --- diff --git a/skills/write-skill/scripts/validate-skill.js b/skills/write-skill/scripts/validate-skill.js index 418b06b..1e8a74a 100644 --- a/skills/write-skill/scripts/validate-skill.js +++ b/skills/write-skill/scripts/validate-skill.js @@ -310,6 +310,23 @@ if (frontmatter) { if (typeof frontmatter['disable-model-invocation'] !== 'boolean') { fail('disable-model-invocation must be a boolean.'); } + + const metadata = frontmatter.metadata; + if (!metadata || typeof metadata !== 'object' || Array.isArray(metadata)) { + fail('metadata must be a YAML mapping.'); + } else { + const allowedTypes = new Set([ + 'performer', + 'router', + 'teaching', + 'utility', + ]); + if (!allowedTypes.has(metadata.type)) { + fail( + 'metadata.type must be performer, router, teaching, or utility.', + ); + } + } } const openai = parseYaml(openaiSource, 'agents/openai.yaml'); From 1a7b61715f8c693f76c93d0b1199ef674c56fc73 Mon Sep 17 00:00:00 2001 From: Adam Hainsworth-Potter Date: Sun, 23 Aug 2026 12:36:14 +0100 Subject: [PATCH 79/79] refactor(maintain-ubiquitous-language): rename context to glossary --- CONTEXT.md => GLOSSARY.md | 2 +- README.md | 2 +- skills/define-product/SKILL.md | 6 +++--- skills/define-product/assets/product-template.md | 2 +- skills/maintain-ubiquitous-language/SKILL.md | 12 ++++++------ .../{context-template.md => glossary-template.md} | 2 +- skills/review-architecture/SKILL.md | 2 +- 7 files changed, 14 insertions(+), 14 deletions(-) rename CONTEXT.md => GLOSSARY.md (99%) rename skills/maintain-ubiquitous-language/assets/{context-template.md => glossary-template.md} (85%) diff --git a/CONTEXT.md b/GLOSSARY.md similarity index 99% rename from CONTEXT.md rename to GLOSSARY.md index f62495c..984a19d 100644 --- a/CONTEXT.md +++ b/GLOSSARY.md @@ -1,4 +1,4 @@ -# Propulsion Context +# Propulsion Glossary ## Language diff --git a/README.md b/README.md index dce137c..9a95d7d 100644 --- a/README.md +++ b/README.md @@ -45,7 +45,7 @@ Enter here to establish or deliberately revise the product and system requiremen $define-product ``` -The session produces a user-confirmed `PRODUCT.md` and the corresponding project language in `CONTEXT.md`. It stops before defining individual features, selecting feature-specific architecture, or creating tickets. +The session produces a user-confirmed `PRODUCT.md` and the corresponding project language in `GLOSSARY.md`. It stops before defining individual features, selecting feature-specific architecture, or creating tickets. ### Define diff --git a/skills/define-product/SKILL.md b/skills/define-product/SKILL.md index bf8bbbc..fe09898 100644 --- a/skills/define-product/SKILL.md +++ b/skills/define-product/SKILL.md @@ -14,7 +14,7 @@ Turns a product idea or existing system into one confirmed `PRODUCT.md` that sup ### 1. Establish the evidence boundary -Locate the project root; read its repository guidance and check root `PRODUCT.md`, `CONTEXT.md`, applicable ADRs and research. For an existing system, also inspect the smallest representative public contracts, entry points, data, tests, configuration, and operational evidence needed to understand the product. Record absent foundation files as absent evidence rather than skipping the check. +Locate the project root; read its repository guidance and check root `PRODUCT.md`, `GLOSSARY.md`, applicable ADRs and research. For an existing system, also inspect the smallest representative public contracts, entry points, data, tests, configuration, and operational evidence needed to understand the product. Record absent foundation files as absent evidence rather than skipping the check. Treat the user as authority for intended direction, repository and runtime evidence as evidence of current behaviour, and cited external sources as authority only for the claims they establish. @@ -58,6 +58,6 @@ State each fact and requirement once, then link its identifier. Keep journeys, r ### 6. Verify and stop -Verify intended direction against the confirmed synthesis, current claims against inspected evidence, terminology against `CONTEXT.md`, and material external claims against cited research. Check authority, scope, journeys, capability and technical-requirement coverage, identifier uniqueness, source traceability, internal consistency, explicit non-blocking uncertainty, product-level acceptance, and foundation readiness. Confirm the absence of delivery status, priorities, roadmap, releases, tickets, detailed feature acceptance, selected feature implementation, internal architecture inventory, and unsupported precision. +Verify intended direction against the confirmed synthesis, current claims against inspected evidence, terminology against `GLOSSARY.md`, and material external claims against cited research. Check authority, scope, journeys, capability and technical-requirement coverage, identifier uniqueness, source traceability, internal consistency, explicit non-blocking uncertainty, product-level acceptance, and foundation readiness. Confirm the absence of delivery status, priorities, roadmap, releases, tickets, detailed feature acceptance, selected feature implementation, internal architecture inventory, and unsupported precision. -Report changed files, supporting evidence, verification, non-blocking unresolved items, and limitations. Stop after `PRODUCT.md` and the context changes owned by invoked skills are verified. Do not create feature specifications, ADR decisions, work items, scaffolding, implementation, releases, or deployments. +Report changed files, supporting evidence, verification, non-blocking unresolved items, and limitations. Stop after `PRODUCT.md` and the glossary changes owned by invoked skills are verified. Do not create feature specifications, ADR decisions, work items, scaffolding, implementation, releases, or deployments. diff --git a/skills/define-product/assets/product-template.md b/skills/define-product/assets/product-template.md index 6ab3829..7d81795 100644 --- a/skills/define-product/assets/product-template.md +++ b/skills/define-product/assets/product-template.md @@ -116,4 +116,4 @@ ## Evidence and traceability -{Link supplied or inspected research, current-system evidence, applicable ADRs, and `CONTEXT.md`. Do not invent dates, roles, or provenance. Do not add a traceability matrix when the identifiers and inline source traces already establish coverage.} +{Link supplied or inspected research, current-system evidence, applicable ADRs, and `GLOSSARY.md`. Do not invent dates, roles, or provenance. Do not add a traceability matrix when the identifiers and inline source traces already establish coverage.} diff --git a/skills/maintain-ubiquitous-language/SKILL.md b/skills/maintain-ubiquitous-language/SKILL.md index cd27424..f5f2b7d 100644 --- a/skills/maintain-ubiquitous-language/SKILL.md +++ b/skills/maintain-ubiquitous-language/SKILL.md @@ -1,6 +1,6 @@ --- name: maintain-ubiquitous-language -description: Maintains project-specific language in CONTEXT.md. Use when project terminology must be added, changed, challenged, renamed, deprecated, or reconciled. +description: Maintains project-specific language in GLOSSARY.md. Use when project terminology must be added, changed, challenged, renamed, deprecated, or reconciled. metadata: type: utility disable-model-invocation: true @@ -8,20 +8,20 @@ disable-model-invocation: true # Maintain Ubiquitous Language -Applies **Ubiquitous Language** by keeping project-specific terminology consistent in one root `CONTEXT.md` glossary. +Applies **Ubiquitous Language** by keeping project-specific terminology consistent in one root `GLOSSARY.md` glossary. ## Process ### 1. Inspect the language -Inspect the relevant conversation or caller handoff, root `CONTEXT.md`, applicable repository guidance, and only the project evidence needed to understand the language. Treat a stable project-specific meaning as resolved when that context makes it clear. Challenge uses that conflict with an established term. +Inspect the relevant conversation or caller handoff, root `GLOSSARY.md`, applicable repository guidance, and only the project evidence needed to understand the language. Treat a stable project-specific meaning as resolved when that context makes it clear. Challenge uses that conflict with an established term. ### 2. Update inline -Update `CONTEXT.md` as each meaning resolves, before the surrounding work continues. Do not request approval or announce the skill. Add no separate verification or handoff for a routine glossary update; follow the surrounding task's normal checks and reporting. +Update `GLOSSARY.md` as each meaning resolves, before the surrounding work continues. Do not request approval or announce the skill. Add no separate verification or handoff for a routine glossary update; follow the surrounding task's normal checks and reporting. When ambiguity, conflict, rename, or deprecation could materially change the meaning, ask one direct question that distinguishes the alternatives. Preserve the existing entry until the answer resolves the meaning. -Create the root file lazily from the [context template](assets/context-template.md). Preserve unrelated terms and local format. Edit only affected entries, prefer updating an entry to adding a duplicate, and reconcile resolved changes across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence. +Create the root file lazily from the [glossary template](assets/glossary-template.md). Preserve unrelated terms and local format. Edit only affected entries, prefer updating an entry to adding a duplicate, and reconcile resolved changes across affected entries. Retain an alias or deprecation only while it helps interpret current project evidence. -Keep `CONTEXT.md` a glossary, not a specification or implementation record. Give each canonical term one short project-native paragraph; group terms by domain only when useful. +Keep `GLOSSARY.md` a glossary, not a specification or implementation record. Give each canonical term one short project-native paragraph; group terms by domain only when useful. diff --git a/skills/maintain-ubiquitous-language/assets/context-template.md b/skills/maintain-ubiquitous-language/assets/glossary-template.md similarity index 85% rename from skills/maintain-ubiquitous-language/assets/context-template.md rename to skills/maintain-ubiquitous-language/assets/glossary-template.md index 3f5bcd3..5d11007 100644 --- a/skills/maintain-ubiquitous-language/assets/context-template.md +++ b/skills/maintain-ubiquitous-language/assets/glossary-template.md @@ -1,4 +1,4 @@ -# {Project Name} Context +# {Project Name} Glossary ## Language diff --git a/skills/review-architecture/SKILL.md b/skills/review-architecture/SKILL.md index e1da368..f7723d6 100644 --- a/skills/review-architecture/SKILL.md +++ b/skills/review-architecture/SKILL.md @@ -14,7 +14,7 @@ Finds the most consequential current opportunities to improve a codebase's modul ### 1. Establish whole-codebase evidence -Inspect the request, repository guidance, root `PRODUCT.md` and `CONTEXT.md`, applicable ADRs, current revision, and starting worktree state. Treat the whole repository as the review scope while honouring explicit exclusions. Map its capabilities and inspect the source, tests, contracts, schemas, configuration, and operational evidence needed to understand current structure. Exclude dependencies, generated output, vendored code, and obsolete paths from recommendation coverage. +Inspect the request, repository guidance, root `PRODUCT.md` and `GLOSSARY.md`, applicable ADRs, current revision, and starting worktree state. Treat the whole repository as the review scope while honouring explicit exclusions. Map its capabilities and inspect the source, tests, contracts, schemas, configuration, and operational evidence needed to understand current structure. Exclude dependencies, generated output, vendored code, and obsolete paths from recommendation coverage. Inspect relevant Git history after mapping the repository. Use repeated cross-file changes and hotspots as supporting evidence and ranking signals without excluding quieter areas. Distinguish project authority, current observation, inference, and unknowns.