Mission F0: Factory current-state closure v0.1 — repin to afi-config@9497afc, canonical v1.3.0 official composition, residue scrub

[Gio2050]

Mission F0 — Factory current-state closure v0.1

Owner-authorized cleanup: afi-factory now reflects ONLY the current canonical AFI architecture (Evidence V3 era, five-lane provider runtime), ahead of Mission Atlas-0.

Pin movement

• Old pin: faa6a8a3d387a573c24f0421c02d806044a22d9e (MANIFEST/CI/drift test) with stale f91ce4465b9c… references still active in README, .afi-codex.json, src/schemas.ts, operator onboarding, and the official hashes.json — two parallel stale pins.
• New pin: 9497afc24bf380b21701bee453c13ebdf8881a26 (afi-config main, EV3-GOV) — ONE pin everywhere: closure MANIFEST, drift test, CI checkout, official hashes.json, README, codex; guarded by tests/current-state-residue.test.ts (full-length and shorthand pin scans).

Vendored closure delta

• Re-vendored the two files EV3-GOV touched: pipeline.schema.json and composition-ref.schema.json — their governance citations now name afi.scored-signal-evidence.v3 as the current evidence contract. Other eight files byte-unchanged; MANIFEST sha256 pins refreshed; src/generated/ unchanged (codegen embeds descriptions, not $comment citations).
• Byte-verification: all 10 vendored files byte-compared against afi-config@9497afc (local cmp + the drift test's byte-equality pass with AFI_CONFIG_DIR; CI checks out the pinned commit and re-verifies on every run).

Official composition refresh (stale pipeline snapshot deleted)

• Deleted templates/official/froggy-trend-pullback/ — the retired v1.0.0 composition (post-merge aiMl augmentation, degrade lanes, 1.0.0 plugins, retired provider/secret capability claims, old hashes b8d9b734…/269ae355…/6d54c8b7…). FLPR-GOV D-FLPR-6 removed pipeline v1.0.0 from the canonical registry; Factory still taught it as current.
• Added official/froggy-trend-pullback/: byte-identical copies of the canonical afi-config@9497afc registry records — registered manifest froggy-trend-pullback v1.3.0 (five provider-instance-backed fail-fast lanes, EV3-GOV D-EV3-5(1)), canonical analyst-strategy config, the seven bound plugin records (2.0.0 lanes / merge 1.1.0 / scorer 1.0.0) — plus hashes.json re-pinned to the canonical composition hashes:
  • manifestHash df3372da… (= canonical config pin), analystConfigHash e34471de… (= canonical registration), pluginSetHash 5384e1c0… (= runtime composition provenance).
• Because provider selection is authored at the manifest layer and afi.pipeline-template.v1 parameterizes values only, the official provider-backed composition is not template-produced; no template claims official identity (guarded). Template authoring remains a fully tested engine capability (self-contained proof template + init scaffold flow).
• Operations layer follows the assets: factory.templates.list → factory.official.list (canonical identity + hash pins in the listing); factory.plugins.list reads the official plugin records. 14 operations, registry-projection invariant unchanged.

Deleted residue (verified, no archive copies)

Stale V2 citation wording (via re-vendor), the v1.0.0 template family (11 files), both stale pins from all current-facing text, stale conformance-fixture plugin identities, stale operator-onboarding secret-profile example, stale README official-template section, stale src/schemas.ts pin comment, stale test fixture templates/official path. No compatibility aliases, no archive directories, no migration notes.

Guards added

tests/current-state-residue.test.ts: one canonical pin everywhere (40-hex + afi-config@<sha> shorthand scans); no scored-signal-evidence.v1/v2 / EvidenceV1/2 / evidence.v1/2 identifiers in Factory-authored text; no Pipehead / generic-DAG / social-category returns; dependency set exactly ajv/ajv-formats/commander and no recordHash/replayHash/providerInvocations/store tokens in src/; templates/ may not return; official dirs must carry manifest+config+hashes+plugins and never a template. tests/official-artifacts.test.ts now byte-compares every official artifact against AFI_CONFIG_DIR (CI always provides it) and pins the canonical hash literals.

Test proof

• npm run typecheck — clean. npm run build — clean (pretest).
• AFI_CONFIG_DIR=<afi-config@9497afc> npm test — 207 passed, 0 failed, 0 skipped (18 files), including the governed-schema drift guard's byte-equality pass, hashing KATs, official-artifact byte-identity + canonical hash pins, and the new residue guards.
• Adversarial review: 6 independent verification lenses + refutation pass over the diff; all 3 confirmed in-repo findings fixed in 6ae891b (stale f91ce446 comment, stale fixture path, shorthand-pin guard gap).

Boundary + scope confirmations

• Factory builds no evidence records, computes no recordHash/replayHash, persists nothing, executes no providers/graph/scoring/UWR — boundary proven by source scan + guard test.
• No runtime behavior outside Factory changed; no other repository modified; nothing deployed; Atlas-0 not started.
• Known follow-up (afi-docs, out of this mission's write scope by design): AFI_Full_Architecture.md:144 enumerates factory.templates.list in the Factory operation roster; the current name is factory.official.list.

🤖 Generated with Claude Code

[cursor]

Bugbot is not enabled for your account, so this pull request was not reviewed.

Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs.