# COV-005 evidence pack — fail-closed content coverage release reconciliation

## Work-package charter

- **Task ID:** `COV-005`
- **Owner:** Oshun portfolio content-coverage contract maintainer;
  implementation owner: Codex
- **Accountable product:** Shared V1–V9 autonomous-content platform
- **Priority:** P0
- **Target milestone:** 2026-07-15 SOTA gap-closure program
- **Risk tier:** High release-integrity, unknown-content, stale-evidence,
  placement, rights, accessibility, and lineage risk
- **Dependencies:** COV-001 canonical content slots; COV-003 dependency,
  invalidation, and graph-integrity contracts; COV-004 inventory semantics;
  AUD-003 reproducible verification; AUD-006 generated capability truth
- **Affected schemas and release paths:** New strict release-candidate,
  launch-manifest, artifact-evidence, complete shipped-observation, migration,
  decision, blocker, and metric schemas; deterministic reconciliation and a
  typed fail-closed assertion API. Product release-path callers are not added by
  this task.
- **Rollout strategy:** Publish through `@oshun/contracts/content-coverage` and
  the root contracts namespace, enforce the focused contract suite in CI, then
  require each product release path to construct a closed-world candidate from
  authoritative slot, graph, evidence, invalidation, and shipped-observation
  sources and call `assertContentCoverageReleaseAllowed` before release.
- **Rollback strategy:** Revert the release-closure contract, graph diagnostic
  correction, tests, exports, CI step, audit qualification, and ledger closure
  together. Never retain a release integration that converts missing candidate
  fields, evidence, or shipped observations into a passing default.
- **Evidence-pack location:** This file

## Implemented release-closure contract

1. A release candidate is a strict, versioned closed-world snapshot containing
   its product, evaluation time, required canonical slots, authoritative launch
   manifest and registrations, canonical slot records, artifact catalog,
   complete shipped observation, COV-003 dependency graph, and pending
   invalidations. Unknown fields, duplicate identities, cross-product slots,
   future observations/evidence, omitted manifest requirements, and invalidation
   plans bound to another graph revision are rejected.
2. Every required slot must exist, be released, carry fresh lifecycle evidence,
   and link at least its required quantity of active catalog artifacts.
   Bidirectional slot/artifact linkage is checked. Active or shipped artifacts
   with null or unknown slots fail closed.
3. Artifact source and autonomy policy must match the canonical slot. Every
   active artifact must have current, non-revoked release, placement, gate,
   rights, accessibility, and lineage evidence. Missing placement remains an
   explicit blocker even when it is also included in the complete missing- or
   stale-evidence diagnosis.
4. Launch registrations must exactly cover the authoritative manifest. A
   requirement counts only when it has required canonical slot IDs, registration
   evidence, and a non-future registration time. Partial or unknown-slot
   registrations do not count toward the 100% target.
5. The independently supplied shipped observation declares `complete-snapshot`.
   Observed artifacts absent from the artifact/slot registry block release and
   contribute to the target-zero metric. Catalog and observation shipped state
   is reconciled in both directions.
6. Every COV-003 graph diagnostic blocks release. Required slots and active
   artifacts must be represented and correctly bound in the graph, and every
   pending invalidation effect blocks until reconciled.
7. The decision exposes required-slot, launch-registration, and active-artifact
   evidence coverage in basis points, each with a 10,000-basis-point target. It
   also exposes observed shipped artifacts absent from the registry with a
   zero-basis-point target. `assertContentCoverageReleaseAllowed` throws a typed
   error containing the full decision when any blocker remains.

## Import, migration, and deterministic identity

V1 import accepts serialized JSON or in-memory input through the same strict
schema. V0 intentionally lacks the shipped observation, dependency graph, and
invalidation state; migration therefore requires an explicit context carrying
all three and never fabricates them. Import rejects malformed JSON, unsupported
versions, and weakening fields. V1 migration is idempotent.

For the same parsed candidate, reconciliation sorts blocker records and hashes
the candidate plus stable blockers with SHA-256. The resulting decision carries
the full candidate digest and a `coverage_release_` identifier derived from the
first 40 hexadecimal digest characters. Callers supply all timestamps and stable
references; the reconciler has no clock or random input.

## Artifact identity

| Artifact                                                       | SHA-256                                                            |
| -------------------------------------------------------------- | ------------------------------------------------------------------ |
| `libs/contracts/src/content-coverage/release-closure.ts`       | `e84ed16eeea0fa006fcaa5a92c7ee89510d910140506f852c063350bd9306020` |
| `libs/contracts/src/content-coverage/release-closure.spec.ts`  | `2559ea29ce7be7a14e1c273f39c0146eb8aaa9c430e5a309f748b6ee474f0c52` |
| `libs/contracts/src/content-coverage/dependency-graph.ts`      | `e49c8760350572ac6f6474b8d6c8d1be389a22b62a6d5a9c9bc8560c99b405ea` |
| `libs/contracts/src/content-coverage/dependency-graph.spec.ts` | `34c92c01b99790895c5b1fb27017894fb8fe92044a72c8ede7c40b148c2ad5d5` |
| `libs/contracts/src/content-coverage/index.ts`                 | `bbc437a74d6626f716f7e38f9d3bad7f064c1580c03b5bbded59c9785dab9f26` |
| `.github/workflows/content-slot-contract.yml`                  | `3cd72920751d39e153af5f85c395fdddb7fac7f960693e25dee5a3f2ed589bd4` |
| `package.json`                                                 | `9f729f451e3b2686407d6d19485f3bdb4760400c7dfa1a24796980597b030826` |

## Automated and adversarial verification

- `pnpm verify:contracts:content-release-closure` passes targeted ESLint,
  **21/21** Vitest cases with one worker, and the contracts source typecheck.
  The suite covers the full allowed decision and every COV-005 blocker family.
- Adversarial cases cover absent required slots; wrong source/autonomy policy;
  insufficient quantity; missing, stale, revoked, and expired evidence; missing
  placement; stale slot lifecycle; incomplete and omitted launch registration;
  unknown registration slots; broken bidirectional linkage; null/unknown
  artifact slots; unknown shipped artifacts; both shipped-state mismatch
  directions; dependency integrity and missing bindings; pending invalidation;
  typed blocking errors; explicit v0 migration context; strict JSON import;
  duplicate identities; future evidence; and graph-revision drift.
- A COV-003 analyzer defect exposed by the adversarial duplicate-node case was
  corrected: duplicate content-slot diagnostics now deduplicate their node ID
  list before validating their own output. The regression is included in the
  existing dependency suite, which passes **28/28** cases.
- Compatibility gates remain green: content slots **17/17**, COV-004 inventory
  **10/10** with no generated drift, and COV-002 seed registry **10/10** with 68
  registered / 7,185 unregistered rows.
- `NX_DAEMON=false pnpm exec nx run @oshun/contracts:build --parallel=1` passes
  before and after remote integration. A runtime `tsx` probe verifies the
  release schema, evidence-kind constants, and reconciler through the root flat
  export, root `ContentCoverageContracts` namespace, and
  `@oshun/contracts/content-coverage` subpath.
- The gates ran sequentially. Resource checks observed about 10–11 GiB available
  RAM, 30 GiB free swap, and 38 GiB free disk; the build was polled at least
  every 30 seconds and stayed near 10–11 GiB available.
- Source pre-commit ESLint, Prettier, asset locking, and V5 documentation
  validation passed. When the hook expanded into a prohibited 93-project
  workspace typecheck, it was stopped and rerun with `SKIP_TYPECHECK=1` under
  the host resource policy. The focused contracts typecheck and resource-gated
  package build passed before and after remote integration.

## Negative proof and claim boundary

- This task proves a strict contract, deterministic in-process reconciler,
  metrics, and fail-closed assertion API. It does not prove that a database,
  API, worker, deployment gate, or any V1–V9 release path calls that assertion.
- The COV-002 seed and COV-004 inventory still report zero complete canonical
  slots, and no populated product dependency graph or complete product release
  candidate is committed. Product slices still own canonicalization, graph
  population, resolved quantities, current artifact evidence, and integration.
- `complete-snapshot` is a required caller attestation in this schema. This task
  does not implement a signed live inventory scan, production alert, or
  independent deployment probe, and the green fixture is not release evidence.
- Evidence references are strict typed links, not proof that their underlying
  claims are true. Rights, accessibility, lineage, placement, and release
  evidence must still be produced and verified by authoritative product systems.
- No user-facing web/admin/mobile UI, persistence migration, media pipeline, or
  Unreal asset changed. Playwright/mobile/browser accessibility, database,
  provider, media, and Unreal execution are therefore inapplicable to COV-005.

## Publication identity

- **Implementation source commit:** `34aa7cc35c224fa369774f503f7a99e324895013`
- **First remote-integrated tested source:**
  `93404a57474a99c8f8f40993c43d16c67251b268`
- **Publication metadata commit:** `bd93d4fdea6ad0e743f08de1570bdf3604c473ca`
