diff --git a/packages/extension/packs/ee-test-measurement/FIRST-TASK.md b/packages/extension/packs/ee-test-measurement/FIRST-TASK.md
new file mode 100644
index 000000000..72f0c7195
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/FIRST-TASK.md
@@ -0,0 +1,24 @@
+# First task — the shakedown exercise (declared; NOT yet run)
+
+**The task:** one non-QICK instrument class through the strumento device
+schema against **mock instruments** — a signal-generator + spectrum-analyzer
+fit cycle via the task-record seam: author a device instance for the bench
+class (schema extension in scope this pass), run a fit cycle
+(`instrumento fit lorentzian`-class) against a mock measurement, record the
+task (manifest, progress stream, terminal result), and fold a **T1b,
+tier-labeled verdict** (residuals against declared tolerance) end-to-end
+through the loop.
+
+**Exercise level: shakedown** — the loop plumbing end-to-end on synthetic
+substrate. It certifies the pack's mechanics (schema → task record →
+verdict → fold), not the domain's real capability. Substrate-touching
+exercise (a real bench run) is the tracked next milestone, gated on
+hardware access that is not wired in this build — stated, not assumed.
+
+**The polyglot seam rides the same exercise:** the task record the mock run
+produces is written by the Python task-record writer and read by a Julia
+consumer — one round-trip exercising the contract's shape across languages.
+
+**Exit:** the run exists with a run ID, a tier-labeled verdict, and the task
+directory intact — then this pack is EXERCISED (shakedown), and this
+declaration updates to point at the run.
diff --git a/packages/extension/packs/ee-test-measurement/PACK.toml b/packages/extension/packs/ee-test-measurement/PACK.toml
new file mode 100644
index 000000000..b9658c25e
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/PACK.toml
@@ -0,0 +1,53 @@
+# The EE / test-&-measurement domain pack (plan-20260920 step 10; amicode #1328).
+# The strumento substrate — device model, task records, calibration store —
+# is this domain in quantum clothing; this pack generalizes it to instruments.
+# SEEDED, not exercised: the manifest + references resolve; the first task
+# (a mock-instrument fit cycle) is the shakedown exercise, the next milestone.
+#
+# SCOPE BOUND (honest): instruments the device schema actually models. The
+# schema today models QICK-class hardware; the first task extends it to the
+# bench signal-generator + spectrum-analyzer class (mock). Everything the
+# schema does NOT model is out of scope until modeled — stated, never assumed.
+schema_version = "1"
+id = "ee-test-measurement"
+name = "Electrical Engineering / Test & Measurement"
+version = 1
+scores = []
+curricula = ["curricula/instrument-bringup/CURRICULUM.md"]
+skills = [] # the strumento skill generalizes at exercise time; none shipped yet
+payload_schemas = ["PAYLOADS.md"]
+benchmarks = ["benchmarks/README.md"]
+
+[[instruments]]
+kind = "strumento device schema (QICK-class, existing)"
+coverage = "modeled"
+
+[[instruments]]
+kind = "bench signal generator (mock, schema extension this pass)"
+coverage = "modeled-at-first-task"
+
+[[instruments]]
+kind = "spectrum analyzer (mock, schema extension this pass)"
+coverage = "modeled-at-first-task"
+
+[verification]
+contract = "VERIFICATION.md"
+bindings = [
+  "measurement residuals: T1b (deterministic given the recorded artifact; residuals against declared tolerance, honestly noise-bounded)",
+  "closed-loop recalibration: T1b over the recorded calibration-store chain",
+  "hardware claims: un-tiered until measured on real instruments (mock runs certify nothing about hardware)",
+]
+
+[first_task]
+declaration = "FIRST-TASK.md"
+level = "shakedown"
+
+[onboarding]
+primary = "instrument-bringup"
+head = "instrument-bringup"
+
+[corrector]
+name = "task-record contract (the strumento seam: the directory is the truth; verdicts re-derived from the record)"
+paths = ["FIRST-TASK.md", "VERIFICATION.md"]
+integrity = "corrector/integrity.toml"
+tier = "open"
diff --git a/packages/extension/packs/ee-test-measurement/PAYLOADS.md b/packages/extension/packs/ee-test-measurement/PAYLOADS.md
new file mode 100644
index 000000000..41f494c78
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/PAYLOADS.md
@@ -0,0 +1,10 @@
+# EE / test-&-measurement — typed payloads
+
+The pack's payload vocabulary (additive; the schema package is the arbiter):
+
+- `measurement-result` — {quantity, value, tolerance, artifact-ref, provenance} — T1b-verified residuals against the recorded artifact
+- `calibration-ref-advance` — {ref, parent, config-content-id, per-key-diffs} — the calibration store's content-addressed chain
+- `task-record-terminal` — {state, error_kind (script | transport — who fixes them), gates-forwarded, authored_by} — the strumento seam
+
+Every payload carries `tier` (T1a | T1b | T2 | T3 | un-tiered | pre-tier) —
+the enum travels with the claim through every render path.
diff --git a/packages/extension/packs/ee-test-measurement/VERIFICATION.md b/packages/extension/packs/ee-test-measurement/VERIFICATION.md
new file mode 100644
index 000000000..f78994cbd
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/VERIFICATION.md
@@ -0,0 +1,21 @@
+# EE / test-&-measurement — the verification contract
+
+Tiered and claim-class-bound (spec-20260920-171500 D4): a tier tag asserts
+what the instrument establishes, never more.
+
+| Claim class | Tier | The gate |
+|---|---|---|
+| Measurement residuals (recorded artifact vs model, declared tolerance) | **T1b** | deterministic given the recorded artifact — the measurement is stochastic, the residual check over the recorded artifact is not; tolerance declared per contract, residuals honestly noise-bounded |
+| Closed-loop recalibration (the measure → believe → measure-again chain) | **T1b** | the calibration store's content-addressed chain: each ref advance re-derived from the record, never self-graded |
+| Hardware claims (real instruments) | **un-tiered** | mock runs certify NOTHING about hardware — the tier appears only when real instruments are measured; the scope bound holds |
+| Bringup state (device model vs manifest) | **T1b** | the strumento device-schema validation (typed errors, `ok: false` carries the error class) |
+
+**The re-tiering rule:** upgrade on new mechanical evidence (provenance
+stamped); downgrade/quarantine when an instrument is invalidated, cascading
+to the claims it certified. Historical records render `pre-tier` with
+provenance until retro-labeled via the deterministic class→tier table.
+
+**The task-record contract is the trust boundary:** the directory is the
+truth; verdicts forwarded from gates are re-derived from the record before
+routing; only declared gates forward at all; forwarding requires a
+non-empty `authored_by`.
diff --git a/packages/extension/packs/ee-test-measurement/benchmarks/README.md b/packages/extension/packs/ee-test-measurement/benchmarks/README.md
new file mode 100644
index 000000000..9cf6c22a5
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/benchmarks/README.md
@@ -0,0 +1,6 @@
+# EE / test-&-measurement — benchmarks
+
+Declared, not yet populated (honest v1): the benchmark surface lands with
+the first-task exercise — the fit-cycle run's measured table (residuals,
+wall-clock, retries per fit kind) is the first bench row, cited with its
+run ID. Performance and quality claims cite bench rows, never adjectives.
diff --git a/packages/extension/packs/ee-test-measurement/corrector/integrity.toml b/packages/extension/packs/ee-test-measurement/corrector/integrity.toml
new file mode 100644
index 000000000..db243a1ce
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/corrector/integrity.toml
@@ -0,0 +1,6 @@
+# sha256 integrity manifest for the corrector paths (relative to the pack dir).
+# Verified at load time (WS1 #369 pattern): a mismatch makes the pack BROKEN.
+# Regenerate: shasum -a 256 <path> from the pack dir.
+[files]
+"FIRST-TASK.md" = "971f100a28f673ccabd73a0662abd58f4dd14c6bcf2d5e4b123febc526891547"
+"VERIFICATION.md" = "3d53456c5a60b266cf79e3d0b09e6f634b6bb3439871f415c0730f706d895eb2"
diff --git a/packages/extension/packs/ee-test-measurement/curricula/instrument-bringup/CURRICULUM.md b/packages/extension/packs/ee-test-measurement/curricula/instrument-bringup/CURRICULUM.md
new file mode 100644
index 000000000..460675001
--- /dev/null
+++ b/packages/extension/packs/ee-test-measurement/curricula/instrument-bringup/CURRICULUM.md
@@ -0,0 +1,19 @@
+---
+type: curriculum
+schema_version: "1"
+id: instrument-bringup
+version: 1
+description: >-
+  The skeleton seed for the EE domain's guided path: device-model definition,
+  mock bringup, the fit cycle, the task-record discipline. Stages are
+  placeholders until the first-task exercise informs them — a curriculum is
+  authored from measured practice, not speculation.
+stages: [] # deliberately empty: the skeleton exists so the manifest resolves; content lands with exercise
+memory_hooks: [] # none yet — hook-card genesis proposes them from recurring questions
+---
+
+# Instrument bringup (skeleton)
+
+The guided path from a device description to a verified fit cycle. Authored
+from the first-task exercise when it runs — the curriculum grows from what
+the loop actually needed, not from a speculative plan.
diff --git a/packages/extension/packs/sensing-metrology/FIRST-TASK.md b/packages/extension/packs/sensing-metrology/FIRST-TASK.md
new file mode 100644
index 000000000..58145fd6b
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/FIRST-TASK.md
@@ -0,0 +1,17 @@
+# First task — the shakedown exercise (declared; NOT yet run)
+
+**The task:** filter-function validation against synthetic noise spectra —
+compute the filter response for a representative filter-function family
+against declared synthetic spectra, fold the verdict with confidence
+intervals (the estimation payload) end-to-end through the loop, tier-labeled
+T1b. The synthetic ground truth is declared up front: the harness compares
+deterministically over it.
+
+**Exercise level: shakedown** — loop plumbing on synthetic substrate; it
+certifies the pack's mechanics, not the domain's real capability.
+Substrate-touching exercise (real NV data) is the tracked next milestone —
+the NV testbed is not wired in this build, stated plainly.
+
+**Exit:** the run exists with a run ID, a tier-labeled verdict with CIs,
+and the artifacts intact — then this pack is EXERCISED (shakedown), and this
+declaration updates to point at the run.
diff --git a/packages/extension/packs/sensing-metrology/PACK.toml b/packages/extension/packs/sensing-metrology/PACK.toml
new file mode 100644
index 000000000..933d460db
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/PACK.toml
@@ -0,0 +1,45 @@
+# The sensing / metrology domain pack (plan-20260920 step 10; amicode #1328).
+# The NV-center substrate is the existing exploratory bet (INTENT's
+# exploratory tier); this pack seeds the domain: filter functions, noise
+# spectroscopy, metrology-grade estimation with confidence intervals.
+# SEEDED, not exercised: the first task (filter-function validation against
+# synthetic noise spectra) is the shakedown exercise, the next milestone.
+schema_version = "1"
+id = "sensing-metrology"
+name = "Sensing & Metrology"
+version = 1
+scores = []
+curricula = ["curricula/noise-spectroscopy/CURRICULUM.md"]
+skills = [] # the sensing physics rides the flagship's Lindblad/filter-function material at exercise time
+payload_schemas = ["PAYLOADS.md"]
+benchmarks = ["benchmarks/README.md"]
+
+[[instruments]]
+kind = "NV-center testbed (data substrate, not wired in this build)"
+coverage = "unmodeled-until-hardware"
+
+[[instruments]]
+kind = "synthetic noise spectra (validation substrate)"
+coverage = "modeled"
+
+[verification]
+contract = "VERIFICATION.md"
+bindings = [
+  "filter-function validation vs synthetic spectra: T1b (exit-code + artifact contract — the validation harness over the synthetic ground truth)",
+  "estimation results (noise spectral density, filter response): T1b with CI folding — the estimation payload carries lo/hi bounds and the method tag",
+  "real-testbed claims: un-tiered until real data is measured",
+]
+
+[first_task]
+declaration = "FIRST-TASK.md"
+level = "shakedown"
+
+[onboarding]
+primary = "noise-spectroscopy"
+head = "noise-spectroscopy"
+
+[corrector]
+name = "synthetic-spectra ground truth (the validation harness: declared spectra, deterministic comparison)"
+paths = ["FIRST-TASK.md", "VERIFICATION.md"]
+integrity = "corrector/integrity.toml"
+tier = "open"
diff --git a/packages/extension/packs/sensing-metrology/PAYLOADS.md b/packages/extension/packs/sensing-metrology/PAYLOADS.md
new file mode 100644
index 000000000..6925f95f1
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/PAYLOADS.md
@@ -0,0 +1,8 @@
+# Sensing / metrology — typed payloads
+
+- `estimation-result` — {quantity, value, lo, hi, method, artifact-ref, provenance} — the CI-folding payload (the EstimationResult class)
+- `validation-verdict` — {claim, tier (T1b), ground-truth-ref, delta, tolerance} — the synthetic-spectra comparison
+- `filter-function-artifact` — {family, parameters, response-array-ref, sha} — the computed response, content-addressed
+
+Every payload carries `tier` — the enum travels with the claim (un-tiered
+for real-testbed claims until measured; pre-tier for historical records).
diff --git a/packages/extension/packs/sensing-metrology/VERIFICATION.md b/packages/extension/packs/sensing-metrology/VERIFICATION.md
new file mode 100644
index 000000000..8b6d5cf52
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/VERIFICATION.md
@@ -0,0 +1,13 @@
+# Sensing / metrology — the verification contract
+
+Tiered and claim-class-bound (spec-20260920-171500 D4):
+
+| Claim class | Tier | The gate |
+|---|---|---|
+| Filter-function validation vs synthetic spectra | **T1b** | exit-code + artifact contract: the validation harness compares computed filter response against the declared synthetic ground truth — deterministic over the declared spectra |
+| Estimation results (noise spectral density, filter response, with CI) | **T1b** | the estimation payload carries value + lo/hi bounds + the method tag; the CI folds into the ledger as an EstimationResult-class payload |
+| Real-testbed claims (NV hardware) | **un-tiered** | synthetic validation certifies NOTHING about the real testbed; the tier appears only when real data is measured |
+| Metrology-grade comparisons (cross-method) | **T1b** | reproducible comparison over recorded artifacts; the better method wins on declared metrics, never vibes |
+
+**The re-tiering rule** and the `pre-tier`/retro-labeling discipline: as in
+the EE contract (spec-20260920 D4, one rule for every domain).
diff --git a/packages/extension/packs/sensing-metrology/benchmarks/README.md b/packages/extension/packs/sensing-metrology/benchmarks/README.md
new file mode 100644
index 000000000..ddfd7e3a9
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/benchmarks/README.md
@@ -0,0 +1,5 @@
+# Sensing / metrology — benchmarks
+
+Declared, not yet populated (honest v1): the first bench row is the
+first-task validation's measured table (filter families × spectra shapes ×
+deltas vs tolerances), cited with its run ID.
diff --git a/packages/extension/packs/sensing-metrology/corrector/integrity.toml b/packages/extension/packs/sensing-metrology/corrector/integrity.toml
new file mode 100644
index 000000000..8bdf1679f
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/corrector/integrity.toml
@@ -0,0 +1,6 @@
+# sha256 integrity manifest for the corrector paths (relative to the pack dir).
+# Verified at load time (WS1 #369 pattern): a mismatch makes the pack BROKEN.
+# Regenerate: shasum -a 256 <path> from the pack dir.
+[files]
+"FIRST-TASK.md" = "e75ac866f0a60eaf4852422d0083ab5806b8f66cf1549472b33748f845c94cfa"
+"VERIFICATION.md" = "fcfc12df1318989ee034a38d0483a9ca3e36fa3563a989982f550d673f0eba06"
diff --git a/packages/extension/packs/sensing-metrology/curricula/noise-spectroscopy/CURRICULUM.md b/packages/extension/packs/sensing-metrology/curricula/noise-spectroscopy/CURRICULUM.md
new file mode 100644
index 000000000..40b69584e
--- /dev/null
+++ b/packages/extension/packs/sensing-metrology/curricula/noise-spectroscopy/CURRICULUM.md
@@ -0,0 +1,18 @@
+---
+type: curriculum
+schema_version: "1"
+id: noise-spectroscopy
+version: 1
+description: >-
+  The skeleton seed for the sensing domain's guided path: filter-function
+  selection, noise-spectroscopy estimation, metrology-grade comparison.
+  Stages are placeholders until the first-task exercise informs them —
+  authored from measured practice, not speculation.
+stages: []
+memory_hooks: []
+---
+
+# Noise spectroscopy (skeleton)
+
+The guided path from a sensing question to a CI-bounded estimation. Authored
+from the first-task exercise when it runs.
diff --git a/packages/extension/src/scores/pack_lint.ts b/packages/extension/src/scores/pack_lint.ts
new file mode 100644
index 000000000..e24305b37
--- /dev/null
+++ b/packages/extension/src/scores/pack_lint.ts
@@ -0,0 +1,133 @@
+// The pack lint (plan-20260920 step 10; amicode #1328) — "seeded" means
+// something. The loader (packs.ts) loads and reports; the LINT is the bar the
+// acceptance reads: schema validation PLUS reference resolution PLUS the
+// domain key set. A pack counts as SEEDED when its manifest validates, every
+// path-bearing reference resolves, and it carries the full domain keys
+// (curricula, payload schemas, verification contract with claim-class
+// bindings, instruments, benchmarks, first task). The flagship quantum pack
+// carries no first task by design — it reads FORMALIZED, not seeded (the
+// grandfathered status, spec-20260920 D2/D7).
+//
+// A manifest is cheap; a domain is not (the r3 advisory, verbatim): the
+// exercised bar — the first task run end-to-end through the campaign loop
+// with a tier-labeled verdict — is a run state, never a manifest state; the
+// lint records the declared level (shakedown | substrate) and stops there,
+// honestly.
+import * as fs from "node:fs";
+import * as path from "node:path";
+import { validateFile } from "@amicode/schema";
+import { parse as parseToml } from "smol-toml";
+
+export interface PackLintFinding {
+  kind: "schema" | "dangling-ref" | "missing-domain-key";
+  detail: string;
+}
+
+export interface PackLintResult {
+  dir: string;
+  id: string;
+  seeded: boolean;
+  exerciseLevel: "shakedown" | "substrate" | undefined;
+  findings: PackLintFinding[];
+}
+
+/** The domain key set — required for SEEDED, optional at the schema layer so
+ *  the flagship's WS1 manifest stays valid without them. */
+export const DOMAIN_KEYS = [
+  "curricula",
+  "payload_schemas",
+  "verification",
+  "instruments",
+  "benchmarks",
+  "first_task",
+] as const;
+
+interface PathBearing {
+  field: string
+  p: string
+}
+
+function collectReferences(manifest: Record<string, unknown>): PathBearing[] {
+  const refs: PathBearing[] = []
+  const rel = (field: string, v: unknown) => {
+    if (typeof v === "string") refs.push({ field, p: v })
+  }
+  for (const s of (manifest["scores"] as string[]) ?? []) rel("scores", s)
+  for (const sk of (manifest["skills"] as { path: string }[]) ?? []) rel("skills", sk.path)
+  const templates = manifest["templates"] as Record<string, { path: string }> | undefined
+  if (templates) for (const t of Object.values(templates)) rel("templates", t.path)
+  const corrector = manifest["corrector"] as { paths: string[]; integrity: string } | undefined
+  if (corrector) {
+    for (const c of corrector.paths ?? []) rel("corrector.paths", c)
+    rel("corrector.integrity", corrector.integrity)
+  }
+  for (const c of (manifest["curricula"] as string[]) ?? []) rel("curricula", c)
+  for (const c of (manifest["payload_schemas"] as string[]) ?? []) rel("payload_schemas", c)
+  for (const b of (manifest["benchmarks"] as string[]) ?? []) rel("benchmarks", b)
+  const verification = manifest["verification"] as { contract: string } | undefined
+  if (verification) rel("verification.contract", verification.contract)
+  const firstTask = manifest["first_task"] as { declaration: string } | undefined
+  if (firstTask) rel("first_task.declaration", firstTask.declaration)
+  return refs
+}
+
+/** Lint one pack dir. Never throws: findings are the interface. */
+export function lintPackDir(dir: string): PackLintResult {
+  const manifestPath = path.join(dir, "PACK.toml")
+  const findings: PackLintFinding[] = []
+
+  if (!fs.existsSync(manifestPath)) {
+    return { dir, id: "<no manifest>", seeded: false, exerciseLevel: undefined, findings: [{ kind: "schema", detail: "no PACK.toml" }] }
+  }
+
+  const validation = validateFile(manifestPath, "pack")
+  if (!validation.ok) {
+    findings.push({ kind: "schema", detail: validation.errors.join("; ") })
+  }
+
+  let manifest: Record<string, unknown> = {}
+  try {
+    manifest = parseToml(fs.readFileSync(manifestPath, "utf8")) as Record<string, unknown>
+  } catch (e) {
+    findings.push({ kind: "schema", detail: `unparsable TOML: ${e instanceof Error ? e.message : String(e)}` })
+    return { dir, id: "<unparsable>", seeded: false, exerciseLevel: undefined, findings }
+  }
+
+  // Reference resolution: every path-bearing field resolves, dangling = named.
+  for (const { field, p } of collectReferences(manifest)) {
+    const resolved = path.resolve(dir, p)
+    if (!fs.existsSync(resolved)) {
+      findings.push({ kind: "dangling-ref", detail: `${field}: ${p} does not resolve (${resolved})` })
+    }
+  }
+
+  // The seeded check: full domain key set AND a clean manifest.
+  for (const key of DOMAIN_KEYS) {
+    if (!(key in manifest)) {
+      findings.push({ kind: "missing-domain-key", detail: `${key} absent — not a seeded domain pack (formalized/legacy packs omit it by design)` })
+    }
+  }
+  const exerciseLevel = (manifest["first_task"] as { level?: "shakedown" | "substrate" } | undefined)?.level
+
+  return {
+    dir,
+    id: String(manifest["id"] ?? "<no id>"),
+    seeded: findings.length === 0,
+    exerciseLevel,
+    findings,
+  }
+}
+
+/** Lint every pack under the ordered roots — the acceptance's validator. */
+export function lintPacks(roots: string[]): PackLintResult[] {
+  const results: PackLintResult[] = []
+  for (const root of roots) {
+    if (!fs.existsSync(root)) continue
+    for (const entry of fs.readdirSync(root, { withFileTypes: true })) {
+      if (!entry.isDirectory()) continue
+      const dir = path.join(root, entry.name)
+      if (fs.existsSync(path.join(dir, "PACK.toml"))) results.push(lintPackDir(dir))
+    }
+  }
+  return results
+}
diff --git a/packages/extension/test/pack_lint.test.ts b/packages/extension/test/pack_lint.test.ts
new file mode 100644
index 000000000..ed088a3f6
--- /dev/null
+++ b/packages/extension/test/pack_lint.test.ts
@@ -0,0 +1,104 @@
+// The pack-lint tests (plan-20260920 step 10; amicode #1328) — "seeded"
+// means something, and these tests are the acceptance's own validator:
+// they lint the REAL bundled packs. The EE and sensing packs must be SEEDED
+// (clean manifest, every reference resolving, the full domain key set); the
+// flagship quantum pack must read FORMALIZED, not seeded (no first task, by
+// the grandfathered design) — and its references must still all resolve.
+import { describe, it, expect } from "vitest"
+import * as fs from "node:fs"
+import * as os from "node:os"
+import * as path from "node:path"
+import { lintPacks, lintPackDir, DOMAIN_KEYS } from "../src/scores/pack_lint.js"
+
+const bundledPacks = path.resolve(import.meta.dirname, "..", "packs")
+
+describe("the pack lint", () => {
+  it("lints the bundled packs: EE and sensing are SEEDED; the flagship is formalized, not seeded", () => {
+    const results = lintPacks([bundledPacks])
+    const byId = new Map(results.map((r) => [r.id, r]))
+    expect(byId.get("ee-test-measurement")?.seeded).toBe(true)
+    expect(byId.get("sensing-metrology")?.seeded).toBe(true)
+
+    const flagship = byId.get("quantum-control")
+    expect(flagship).toBeDefined()
+    expect(flagship?.seeded).toBe(false) // formalized, not seeded — the grandfathered design
+    // its missing-domain-key findings are the EXPECTED marker of formalized status…
+    const missing = flagship?.findings.filter((f) => f.kind === "missing-domain-key")
+    expect(missing?.length).toBe(DOMAIN_KEYS.length)
+    // schema-clean against the extended schema (the additive extension does not break the WS1 manifest)
+    expect(flagship?.findings.filter((f) => f.kind === "schema")).toEqual([])
+    // KNOWN REAL FINDING, recorded: the flagship's skill paths are SNAPSHOT-LAYOUT
+    // paths (../../skills/transmon resolves in the VSIX/extension-snapshot layout
+    // where all skills coexist — the repo tree carries only the meta skills), so
+    // in-repo the lint flags them dangling. The step-14 formalization reconciles
+    // the path layout; until then this is the honest finding, baked into the test
+    // so the reconciliation cannot land silently.
+    const dangling = flagship?.findings.filter((f) => f.kind === "dangling-ref")
+    expect(dangling?.every((f) => f.detail.startsWith("skills:"))).toBe(true)
+  })
+
+  it("the seeded packs declare their exercise level honestly (shakedown, not substrate)", () => {
+    for (const id of ["ee-test-measurement", "sensing-metrology"]) {
+      const r = lintPacks([bundledPacks]).find((x) => x.id === id)
+      expect(r?.exerciseLevel).toBe("shakedown") // declared; NOT exercised — a run state, never a manifest state
+    }
+  })
+
+  it("a dangling reference fails the lint, named", () => {
+    const dir = fs.mkdtempSync(path.join(os.tmpdir(), "pack-lint-"))
+    fs.writeFileSync(
+      path.join(dir, "PACK.toml"),
+      [
+        'schema_version = "1"',
+        'id = "broken"',
+        'name = "Broken"',
+        "scores = []",
+        "[onboarding]",
+        'primary = "none"',
+        "[corrector]",
+        'name = "gate"',
+        'paths = ["nope.md"]',
+        'integrity = "nope.toml"',
+        "curricula = []",
+        "payload_schemas = []",
+        "instruments = []",
+        "benchmarks = []",
+        "[first_task]",
+        'declaration = "also-nope.md"',
+        'level = "shakedown"',
+        "",
+      ].join("\n"),
+    )
+    const r = lintPackDir(dir)
+    expect(r.seeded).toBe(false)
+    const dangling = r.findings.filter((f) => f.kind === "dangling-ref")
+    expect(dangling.some((f) => f.detail.includes("corrector.paths"))).toBe(true)
+    expect(dangling.some((f) => f.detail.includes("first_task.declaration"))).toBe(true)
+  })
+
+  it("a manifest missing the domain keys reads formalized (the seeded-marker findings, not errors)", () => {
+    const dir = fs.mkdtempSync(path.join(os.tmpdir(), "pack-lint-2-"))
+    fs.writeFileSync(
+      path.join(dir, "PACK.toml"),
+      [
+        'schema_version = "1"',
+        'id = "minimal"',
+        'name = "Minimal"',
+        "scores = []",
+        "[onboarding]",
+        'primary = "none"',
+        "[corrector]",
+        'name = "gate"',
+        'paths = ["gate.md"]',
+        'integrity = "integrity.toml"',
+        "",
+      ].join("\n"),
+    )
+    fs.writeFileSync(path.join(dir, "gate.md"), "# the gate contract\n")
+    fs.writeFileSync(path.join(dir, "integrity.toml"), "schema_version = \"1\"\n")
+    const r = lintPackDir(dir)
+    expect(r.seeded).toBe(false)
+    // schema-clean and reference-clean — the ONLY findings are the domain-key markers
+    expect(r.findings.every((f) => f.kind === "missing-domain-key")).toBe(true)
+  })
+})
diff --git a/packages/schema/schemas/pack.schema.json b/packages/schema/schemas/pack.schema.json
index 296a60c0e..f35abf091 100644
--- a/packages/schema/schemas/pack.schema.json
+++ b/packages/schema/schemas/pack.schema.json
@@ -2,33 +2,51 @@
   "$schema": "http://json-schema.org/draft-07/schema#",
   "$id": "https://amico.harmoniqs.co/schema/pack/v1",
   "title": "amico pack manifest",
-  "description": "The unit of generality for a domain pack (autoresearch studio WS1, #369): one manifest per domain bundling skills, a template registry, the corrector with its integrity manifest, the catalog schema, the onboarding score, and an eval corpus pointer. Owned by the schema package; one validator consumed by the extension, the CLI, and CI. A score is one FIELD of a pack, not a peer of it. The per-entry tier (open vs an entitlement bundle id) encodes the open-core boundary: flowing a feature back to open-source is a manifest edit plus a content move, not a migration. Pack ids are deliberately NOT an enum — the taxonomy is manifest data, revisable without a schema version bump.",
+  "description": "The unit of generality for a domain pack (autoresearch studio WS1, #369): one manifest per domain bundling skills, a template registry, the corrector with its integrity manifest, the catalog schema, the onboarding score, and an eval corpus pointer. Owned by the schema package; one validator consumed by the extension, the CLI, and CI. A score is one FIELD of a pack, not a peer of it. The per-entry tier (open vs an entitlement bundle id) encodes the open-core boundary: flowing a feature back to open-source is a manifest edit plus a content move, not a migration. Pack ids are deliberately NOT an enum \u2014 the taxonomy is manifest data, revisable without a schema version bump.",
   "type": "object",
   "additionalProperties": false,
-  "required": ["schema_version", "id", "name", "scores", "onboarding", "corrector"],
+  "required": [
+    "schema_version",
+    "id",
+    "name",
+    "scores",
+    "onboarding",
+    "corrector"
+  ],
   "properties": {
-    "schema_version": { "enum": ["1"] },
+    "schema_version": {
+      "enum": [
+        "1"
+      ]
+    },
     "id": {
       "type": "string",
       "pattern": "^[a-z0-9-]+$",
       "description": "pack id, kebab-case, free-form (quantum-control, qec, calibration, ...)"
     },
-    "name": { "type": "string", "minLength": 1 },
+    "name": {
+      "type": "string",
+      "minLength": 1
+    },
     "version": {
       "type": "integer",
       "minimum": 1,
-      "description": "pack content version — bump when content moves between tiers"
+      "description": "pack content version \u2014 bump when content moves between tiers"
     },
     "scores": {
       "type": "array",
-      "minItems": 1,
-      "items": { "type": "string", "minLength": 1 },
-      "description": "score dirs this pack contributes, relative to the pack dir"
+      "items": {
+        "type": "string",
+        "minLength": 1
+      },
+      "description": "Score/curriculum dirs (paths relative to the pack dir). A seeded domain pack may carry none yet \u2014 the curriculum skeleton rides `curricula` and composed interviews come at maturity."
     },
     "onboarding": {
       "type": "object",
       "additionalProperties": false,
-      "required": ["primary"],
+      "required": [
+        "primary"
+      ],
       "properties": {
         "primary": {
           "type": "string",
@@ -44,32 +62,48 @@
     },
     "skills": {
       "type": "array",
-      "items": { "$ref": "#/$defs/tieredPath" },
+      "items": {
+        "$ref": "#/$defs/tieredPath"
+      },
       "description": "skill dirs this pack contributes, relative to the pack dir"
     },
     "templates": {
       "type": "object",
-      "additionalProperties": { "$ref": "#/$defs/tieredPath" },
-      "description": "template registry: name → tiered path (e.g. solve → the vetted template)"
+      "additionalProperties": {
+        "$ref": "#/$defs/tieredPath"
+      },
+      "description": "template registry: name \u2192 tiered path (e.g. solve \u2192 the vetted template)"
     },
     "corrector": {
       "type": "object",
       "additionalProperties": false,
-      "required": ["name", "paths", "integrity"],
+      "required": [
+        "name",
+        "paths",
+        "integrity"
+      ],
       "properties": {
-        "name": { "type": "string", "minLength": 1 },
+        "name": {
+          "type": "string",
+          "minLength": 1
+        },
         "paths": {
           "type": "array",
           "minItems": 1,
-          "items": { "type": "string", "minLength": 1 },
-          "description": "the gate artifacts — must resolve OUTSIDE agent-editable trees (a load-time property, not a convention)"
+          "items": {
+            "type": "string",
+            "minLength": 1
+          },
+          "description": "the gate artifacts \u2014 must resolve OUTSIDE agent-editable trees (a load-time property, not a convention)"
         },
         "integrity": {
           "type": "string",
           "minLength": 1,
           "description": "path to the sha256 integrity manifest covering every named path, relative to the pack dir"
         },
-        "tier": { "$ref": "#/$defs/tier" }
+        "tier": {
+          "$ref": "#/$defs/tier"
+        }
       }
     },
     "catalog_schema": {
@@ -81,20 +115,118 @@
       "type": "string",
       "minLength": 1,
       "description": "pointer to the pack's eval corpus (e.g. a Prova corpus ref)"
+    },
+    "curricula": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "Curriculum paths (relative to the pack dir). The curricula substrate is core; the content is pack-declared."
+    },
+    "payload_schemas": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "Typed-payload schema references (paths, relative to the pack dir)."
+    },
+    "verification": {
+      "type": "object",
+      "required": [
+        "contract",
+        "bindings"
+      ],
+      "properties": {
+        "contract": {
+          "type": "string",
+          "description": "Path to the per-domain verification contract (claim-class bindings, tier assignments)."
+        },
+        "bindings": {
+          "type": "array",
+          "items": {
+            "type": "string"
+          },
+          "description": "Claim-class bindings in short form (e.g. 'simulation: model-trajectory claims at T1a; hardware claims un-tiered') \u2014 the full contract doc carries the details."
+        }
+      },
+      "description": "The verification-contract reference with claim-class bindings (the tiering is honest: tiers assert what the instrument establishes)."
+    },
+    "instruments": {
+      "type": "array",
+      "items": {
+        "type": "object",
+        "required": [
+          "kind"
+        ],
+        "properties": {
+          "kind": {
+            "type": "string"
+          },
+          "coverage": {
+            "type": "string",
+            "description": "modeled | modeled-at-first-task | unmodeled \u2014 the scope bound, stated not assumed"
+          }
+        }
+      },
+      "description": "The instrument list with the scope bound: instruments the schema actually models \u2014 each entry states its coverage."
+    },
+    "benchmarks": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "description": "Benchmark paths (relative to the pack dir), each resolving at lint time."
+    },
+    "first_task": {
+      "type": "object",
+      "required": [
+        "declaration",
+        "level"
+      ],
+      "properties": {
+        "declaration": {
+          "type": "string",
+          "description": "Path to the first-task declaration doc."
+        },
+        "level": {
+          "type": "string",
+          "enum": [
+            "shakedown",
+            "substrate"
+          ],
+          "description": "shakedown = the loop plumbing end-to-end (mock/synthetic); substrate = touching real domain substrate. The exercised-to-maturity ladder is tracked per manifest."
+        }
+      },
+      "description": "The first-task declaration \u2014 a pack is exercised when it has run end-to-end with a tier-labeled verdict."
     }
   },
   "$defs": {
     "tier": {
-      "anyOf": [{ "const": "open" }, { "type": "string", "pattern": "^[a-z0-9-]+$" }],
+      "anyOf": [
+        {
+          "const": "open"
+        },
+        {
+          "type": "string",
+          "pattern": "^[a-z0-9-]+$"
+        }
+      ],
       "description": "open (default when absent), or the entitlement bundle id gating this entry"
     },
     "tieredPath": {
       "type": "object",
       "additionalProperties": false,
-      "required": ["path"],
+      "required": [
+        "path"
+      ],
       "properties": {
-        "path": { "type": "string", "minLength": 1 },
-        "tier": { "$ref": "#/$defs/tier" }
+        "path": {
+          "type": "string",
+          "minLength": 1
+        },
+        "tier": {
+          "$ref": "#/$defs/tier"
+        }
       }
     }
   }
diff --git a/packages/schema/test/pack.test.ts b/packages/schema/test/pack.test.ts
index a214d8727..d5734047a 100644
--- a/packages/schema/test/pack.test.ts
+++ b/packages/schema/test/pack.test.ts
@@ -51,8 +51,12 @@ describe("the pack kind", () => {
   it("requires the onboarding primary score — a score is a FIELD of a pack", () => {
     expect(validate(pack({ onboarding: { head: "overture" } }), "pack").ok).toBe(false);
   });
-  it("requires a non-empty scores list", () => {
-    expect(validate(pack({ scores: [] }), "pack").ok).toBe(false);
+  it("allows an empty scores list — the seeded-domain-pack contract (plan-20260920 step 10)", () => {
+    // A seeded domain pack legitimately carries no composed scores yet: the
+    // curriculum skeleton rides `curricula` and composed interviews come at
+    // maturity. The flagship keeps its non-empty list; minItems was the
+    // quantum-era single-pack contract, relaxed deliberately for domains.
+    expect(validate(pack({ scores: [] }), "pack").ok).toBe(true);
   });
   it("requires corrector.integrity — the threshold condition is a load-time property, not a convention", () => {
     const c = { ...(pack().corrector as object) } as Record<string, unknown>;