feat(review): per-repo opt-in to let a confident AI-judgment blocker gate the merge

[JSONbored]

Summary

• Same repro as fix(review): deterministic linked-issue gate never checks issue-PR scope match, only existence/openness #3906 (metagraphed PR fix(review): preserve gate verdict on auto-review skip #3910): the AI reviewer correctly flagged a real content-correctness defect under "Blockers" (a registry provider slug semantically wrong for its domain — "gittensory" instead of "gittensor"), yet the gate's top-line disposition stayed approve/merge recommended. Root cause: applySurfaceGate's guard feat(scoring): add situational score projections #3 unconditionally lets a decisive deterministic surface merge override even an AI-judgment-only failure — correct and intentional for gittensory's own code repo (fix(review): decide whether AI-judgment blockers should gate the verdict, not just annotate it #2592: other deterministic nets like codecov/patch/tests/lint/secret-scan make AI judgment supplementary there), but content/registry repos like metagraphed have no equivalent net for "this value is semantically wrong" defects.
• Adds gate.aiJudgmentBlockers: "gate" | "advisory" (default "advisory", byte-identical everywhere that doesn't opt in) as a YML-only config knob — deliberately mirroring contentLane's own YML-only shape (no DB column, no dashboard toggle) rather than the heavier DB-backed gate-setting-wiring template, since this only ever has an effect for a repo already running the registry content lane.
• When "gate": applySurfaceGate's AI-judgment-only override is skipped, so the finding falls through to the same union+failure path any other blocker takes — decision is no longer merge. evaluateWithSurfaceLane's companion advisory.findings cleanup (which strips the AI finding from the public-comment feed when the override fires) is also skipped in "gate" mode, since the finding is now a real blocker and the public comment needs to keep showing it.
• Preserves the existing "advisory" (default) behavior byte-identically — confirmed by running all 46 pre-existing content-lane-wire.test.ts tests unchanged and green, plus a new explicit test proving aiJudgmentBlockersMode: "advisory" produces output toEqual the unset-default case.

Scope

• Conventional Commit title.
• Focused — one config field + its wiring, no unrelated changes.
• Follows CONTRIBUTING.md.
• Owner PR — linked issue eligibility gate does not apply. Advances feat(review): per-repo opt-in to let a confident AI-judgment blocker gate the merge decision (not just annotate it) #3907 (does not close it — see Notes on the metagraphed-side follow-up).

Validation

• git diff --check
• npm run typecheck
• npm run docs:drift-check — unaffected (this field lives only in FocusManifestGateConfig, not RepositorySettings, so it's outside GATE_MODE_MANIFEST's scope, which only scans src/types.ts).
• npx vitest run test/unit/content-lane-wire.test.ts test/unit/focus-manifest.test.ts test/unit/config-templates.test.ts — 612/612 passing, including all 46 pre-existing content-lane-wire.test.ts cases unchanged.
• New tests: applySurfaceGate with aiJudgmentBlockersMode: "gate" reproducing PR fix(review): preserve gate verdict on auto-review skip #3910's exact shape (a confident ai_consensus_defect provider-misattribution finding survives a decisive surface merge, decision becomes failure); the explicit-"advisory"-equals-default proof above; and two full evaluateWithSurfaceLane end-to-end tests — one opted-in (gate.aiJudgmentBlockers: "gate" via a real parsed manifest, confirms decision !== "success" AND the finding stays in advisory.findings) and one opted-out (zero-config manifest, confirms decision === "success" and the finding IS stripped, byte-identical to the pre-existing REGRESSION test).
• Scoped coverage: src/review/content-lane-wire.ts 100% lines / 98.75% branches (the one remaining branch is in runRegistrySurfaceGate, a function this PR does not touch — confirmed via git diff, zero changes in that function). packages/gittensory-engine/src/focus-manifest.ts's uncovered spots are likewise pre-existing and unrelated (verified each one against git diff — none fall within the aiJudgmentBlockersMode field I added).
• npm run test:workers, build:mcp, test:mcp-pack, ui:openapi:check, ui:lint, ui:typecheck, ui:build, npm audit — not re-run locally (no worker/MCP/OpenAPI/UI-component code touched); CI runs them authoritatively.

Safety

• No secrets, wallet/hotkey/trust-score/reward data anywhere.
• No behavior change for any repo that doesn't opt in — confirmed byte-identical via the full existing test suite passing unchanged.
• No UI changes — no UI Evidence section needed.
• No docs/changelog changes needed (.gittensory.yml.example + config/examples/gittensory.full.yml documentation is the deliverable itself, not a changelog).
• Real design tension, documented in the code (not hidden): the guard this issue makes opt-outable is a deliberate, incident-motivated safety net (its own existing regression test cites a prior hallucinated-RAG-claim false positive as the reason AI-judgment-only findings must never veto a decisive deterministic verdict by default). Opting a repo into "gate" mode reintroduces exactly that risk as an explicit, per-repo, documented trade-off — the doc comments on both the manifest field and applySurfaceGate itself say so plainly, and the default stays "advisory" everywhere.

Notes

• Deliberately out of scope, per the issue's own 4th deliverable: actually setting gate.aiJudgmentBlockers: "gate" in JSONbored/metagraphed's own .gittensory.yml is a separate follow-up PR in that other repo, gated on this shipping first — not bundled here.
• Confidence bar, as specified: only a finding the AI reviewer itself already placed under "Blockers" (not "Nits") is eligible — this PR doesn't add a new confidence-bar mechanism, it reuses the existing isAiJudgmentOnlyFailure/AI_JUDGMENT_BLOCKER_CODES classification that already gates which findings are "AI-judgment-only" in the first place.

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	gittensory-ui	a80e7ae	Commit Preview URL Branch Preview URL	Jul 08 2026, 09:19 AM

[codecov]

Codecov Report

❌ Patch coverage is 83.33333% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 93.72%. Comparing base (11fc55b) to head (a80e7ae).
⚠️ Report is 3 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
packages/gittensory-engine/src/focus-manifest.ts	0.00%	0 Missing and 1 partial ⚠️

❌ Your patch status has failed because the patch coverage (83.33%) is below the target coverage (99.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4171      +/-   ##
==========================================
- Coverage   93.72%   93.72%   -0.01%     
==========================================
  Files         387      387              
  Lines       36309    36312       +3     
  Branches    13300    13303       +3     
==========================================
+ Hits        34032    34034       +2     
  Misses       1621     1621              
- Partials      656      657       +1     

Files with missing lines	Coverage Δ
src/review/content-lane-wire.ts	98.55% <100.00%> (+0.04%)	⬆️
packages/gittensory-engine/src/focus-manifest.ts	99.06% <0.00%> (-0.10%)	⬇️

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.