V3 Launch Readiness

Gate: v3-launch-readiness.v1

Verifier: pnpm verify:v3 launch-readiness

This gate closes V3/V3_TODOS.md §75. It aggregates every final launch readiness item into 39 required release gates (the named gate IDs in libs/oshun/analytics/src/v3-launch-readiness.ts) with concrete evidence paths and verifier commands. Since the 2026-05-31 de-circularization (V3/VERIFICATION_AUDIT_2026-05-31.md), the verifier scripts/v3/verify-v3-launch-readiness.mjs actually executes every verifier command those 39 gates declare — 41 distinct constituent commands: 40 verify:v3 * sub-gate verifiers plus gate 25's avatar-pipeline vitest run (the prior doc cited only the 39 manifest count, which caused a 39-vs-40 discrepancy; all counts are real and are indexed below). Each named gate's measured status is the AND of its own commands' results, and the verifier refuses to pass while the committed evidence artifact disagrees with the run it just performed. The aggregate dispatches each verify:v3 * command through the exact route in scripts/verify-manifest.json, including composite routes, and fails closed when a route is absent. --write-evidence regenerates the artifact from a real run; hand-editing it green is impossible without the verifier failing on drift. The aggregate gate is green only when:

• §1 through §74 have no remaining unchecked or waived local tasks.
• GA inventory is published for Tara, Saraswati, and Lilith Commons.
• All 41 constituent verifier commands exit green.

Current status (2026-07-24): NOT GA-ready. V3/V3_TODOS.md carries [~] not-locally-actionable tasks and a subset of the constituent sub-gates fail for exactly those reasons — the measured state is 30/41 constituent commands and 24/39 named gates passing (see V3/launch/launch-readiness-release-gate.json, machine-generated). A passing sub-gate is structural verification, not operational proof — see the honest-scope note in V3/VERIFICATION_AUDIT_2026-05-31.md:308-318.

The machine-readable evidence is V3/launch/launch-readiness-release-gate.json (regenerate only via node scripts/v3/verify-v3-launch-readiness.mjs --write-evidence). The gate fails closed when any required item is missing, not passing, lacks a verifier, lacks traceable evidence, when the GA inventory aggregate is not published, or when the recorded evidence disagrees with the measured run.

The artifact uses oshun.generated-evidence-metadata.v1 and the release-readiness evidence class, whose freshness SLO is 24 hours. Its metadata binds the measured payload to the generator version, source commit, exact write command, host, and manifest-contract digest. The consumer also fails closed when that metadata is stale, future-dated, incomplete, tampered, references an unavailable commit, or describes another contract. Execution, evidence write, and release publication are separate lifecycle states: the current red result records a completed verifier and written evidence while remaining honestly unpublished for GA release.

Index: the 39 manifest gates#

Each row maps a manifest gate ID to the verify:v3 * sub-gate(s) it declares and to its human-readable topic file, or — where no topic file exists — its primary evidence path.

The union of the verify:v3 * commands above is exactly the 40 distinct constituent sub-gates the launch verifier executes; together with gate 25's avatar-pipeline vitest command (also executed since 2026-06-12) the verifier runs 41 distinct constituent commands per sweep.

Index: root V3 gate topic files not aggregated by §75#

These per-topic gates exist with their own verifiers and evidence but are not constituents of the launch-readiness aggregate. They are listed here so the topic-file inventory is complete (previously ~15 of these were absent from any prose index).