Configuration Effectiveness (agents.md#213): receipt, reach matrix, rule ledger, arms #342

Tracks the dokimi side of agentsmd/agents.md#213 (Configuration Effectiveness): what an agent was expected to load, what it loaded, and whether a rule earns its place.

Open PRs

• feat(repo): effective-instructions receipt for repo reach #335 effective-instructions receipt (repo reach --format receipt), shape from caioribeiroclw-pixel
• reach matrix, client × version (branch ce-reach-matrix; PR opens after feat(repo): effective-instructions receipt for repo reach #335 merges; nightly needs model secrets and REACH_MODEL_* vars), offered as the AGENTS.md conformance suite (feat(catalog): read trigger sets back and list skill package formats #211/ci(security): hash-pin every lock install; one hashed tools lock #135)
• feat(repo): dokimi:typecheck grader and repo scan --typecheck for Python fences #337 dokimi:typecheck grader, repo scan --typecheck
• feat(repo): rule ledger: repo compare --ledger with per-rule keep/review/insufficient-evidence verdicts #336 rule ledger: repo compare --ledger, per-rule keep/review/insufficient-evidence
• feat(repo): guidance:overlay arm and overlay reach check #338 guidance:overlay arm (agents.md#185)
• feat(repo): guidance:format=xml and guidance:as-skill arms #333 guidance:format=xml, guidance:as-skill
• feat(context): prefilter AGENTS.md rules against session logs (candidates for ablation) #334 context prefilter from session logs (ablation candidates)
• feat(repo): time to first useful result, with a rate in repo compare #332 time to first useful result

Merge order: #335 → reach matrix → #337, #336, #338, #333, #334, #332.

Open decisions

• POML arm in feat(repo): guidance:format=xml and guidance:as-skill arms #333: recommended drop
• max_docs raised by feat(context): prefilter AGENTS.md rules against session logs (candidates for ablation) #334, feat(repo): rule ledger: repo compare --ledger with per-rule keep/review/insufficient-evidence verdicts #336 and the reach matrix: the last one to merge sets 40
• batch_ledger cell id lacks the overlay value (feat(repo): guidance:overlay arm and overlay reach check #338)

Follow-ups (small)

• review_by per guidance glob in the ledger (feat(repo): rule ledger: repo compare --ledger with per-rule keep/review/insufficient-evidence verdicts #336)
• tokens per selection and catalog-listing cost in routing (feat(routing): estimate-only, all-vs-all repeats and cost, description queries #329, research I016)
• import CE trigger telemetry (trigger_telemetry.v1) as a second feat(context): prefilter AGENTS.md rules against session logs (candidates for ablation) #334 source; the JSON contract is with dakotafabro, rp-why: Configuration Effectiveness (CE) - Phase 1 block/agent-skills#52. A trigger rate picks candidates, and A/B gives the verdict.

Elsewhere

• failure → task replay: repo tasks: authoring replay tasks from merged PRs #294
• drift-stale cells: GitHub Action: pr-check / record / revalidate, .github/dokimi config, per-subject baselines #318
• scorecard export: Skill catalog integration (intel/skills): per-skill evals in catalog CI, results, promotion to validated #138, after model-targets #495 settles
• following (adherence) evidence: J4: adherence and task-impact evidence for instructions (AC-J04-005, AC-J04-006) #96