Affected schemas and content slots: New strict
CoverageSeedRegistryV1Schema, 28 registered V1 surface facts, four
registered but incomplete V1 scheduled-category candidates, and zero complete
canonical COV-001 slots. No content artifact is released or placed by this
task.
Rollout strategy: Commit the deterministic JSON seed and concise report,
enforce exact regeneration and adversarial extraction tests in CI, keep every
unresolved/unregistered row assigned to the portfolio migration owner, and
require later COV/product work to supply missing canonical policy/lifecycle
facts rather than defaults.
Rollback strategy: Revert schema, generator, generated outputs, CI,
capability commentary, audit reconciliation, and ledger closure together.
Never retain a generated registry whose source digests or reviewed counts no
longer match the repository.
The generator parses GENERATION_SURFACE_VALUES and
V1_AUTONOMY_LAUNCH_CATEGORIES through the TypeScript compiler AST. It
accepts only static string/number/array/object literals and rejects spreads,
computed members, runtime evaluation, or missing symbols.
The nine authoritative version TODO ledgers are parsed with exact checkbox
markers. All P0/bootstrap and P1/launch rows are retained, including checked
rows; every unchecked [ ] and blocked [~] row is retained regardless of
phase. Checked post-launch rows outside that scope are intentionally omitted.
Every row carries a stable content-derived ID plus path, line interval, file
digest, symbol or heading provenance. The output is deterministic and its
--check mode fails on source/output drift.
V8 and V9 each contribute seven canonical product gates and seven conflicting
implementation-alternate gates. Both contracts remain in the registry. A
reviewed migration rule maps by semantic name/component rather than the
conflicting ordinal and keeps unmatched canonical gates explicit.
Sixteen required V8/V9 artifact rows and six placement rows are anchored to
architecture/TODO sources. Exact-semantic duplicate groups retain every
underlying requirement row and surface status/priority contradictions.
The four live autonomy categories are current scheduled candidates, not full
V1 coverage and not complete COV-001 records. Their missing definition and
lifecycle fields are enumerated instead of invented.
The registration numerator is intentionally a count of source facts admitted to
this seed, not a coverage percentage. Requirements and conflicting alternate
gate rows remain unregistered. The four scheduled candidates are registered
source facts but unresolved canonicalizations, which is why unresolved exceeds
unregistered by four.
Every unresolved row is assigned to team:oshun-content-coverage-migration,
with escalation to oncall:oshun-platform. This is explicit migration
accountability, not an invented content-business owner.
migration:coverage.ledger-exact-semantic-v1 links exact normalized duplicate
requirements while retaining each row's independent source locator, status,
priority, and text.
migration:coverage.gate-name-mapping-v1 retains canonical and alternate
V8/V9 gate contracts, records the non-bijective semantic mapping, and leaves
alternate suites unregistered until product code is migrated.
pnpm verify:content-coverage:seed-registry passes focused ESLint, the
generator TypeScript project, 10/10 Node test cases, deterministic output
drift checking, and the contracts source typecheck. Tests freeze the exact 28
surfaces, four categories, per-product extraction counts, independent
unchecked/blocked coverage, all four seven-gate suites, artifacts, placements,
reconciliation retention, owner completeness, deterministic regeneration,
strict unknown-field rejection, missing-owner rejection, and stale-summary
rejection.
pnpm verify:contracts:content-slot passes focused ESLint, all 17 COV-001
contract tests, and the contracts source typecheck with the new registry
export included.
NX_DAEMON=false pnpm exec nx run @oshun/contracts:build --parallel=1 passes,
and a runtime probe validates the committed registry through both
@oshun/contracts and @oshun/contracts/content-coverage exports.
Prettier checks pass for all changed sources. The 11 MiB generated JSON and
its generated report are byte-stable drift-gate outputs and are excluded from
lint-staged reformatting to avoid both generator drift and unnecessary memory
pressure.
Source and authenticated-record pre-commit linting, formatting, asset locking,
and V5 document validation passed. The hook's inferred broad typecheck was
deliberately bypassed with SKIP_TYPECHECK=1 on this 15 GiB host; the focused
generator and contracts typechecks, package build, and post-merge gates above
all passed independently.
The schema rejects unknown weakening fields, invalid row IDs and source
digests, undeclared source references, duplicate IDs/source files, stale
counts, unknown reconciliation rules/rows, missing owner assignments, checked
non-launch imports, and inconsistent inclusion reasons.
This task does not claim four categories cover V1, that a source tool surface
is a required content slot, that TODO completion proves a live artifact, or
that any slot is localized, accessible, rights-cleared, fresh, placed,
releasable, or released. COV-003 through COV-005 and the product slices own
those closures.
No customer/admin UI, mobile flow, API behavior, persistence migration,
provider request, generated media, or Unreal asset changed. Playwright/mobile,
accessibility-browser, database, provider, media, and Unreal execution are
therefore inapplicable to COV-002.