fix(reputation): never cadence-downgrade a trusted submitter, and never claim a review was skipped when it ran

[JSONbored]

What happened

The repo owner's own PRs started showing:

⏸️ Required AI review was skipped by a submitter-reputation downgrade — the submitter's recent-submission signal downgraded this PR to deterministic-only checks.

…and on #9764, that bullet rendered directly above a successful AI review summary, holding the gate for a review that had already run.

Root causes (both verified live)

1. The cadence leg has no notion of an earned record. shouldSkipAiForReputation has two legs: the quality/burst signal, and isMachinePacedCadence (≥5 submissions in 24h with a median gap under 10 minutes). The maintainer's remediation campaign measured 50 submissions / 7.2-minute median gap — machine-paced by definition — so every PR was downgraded, even though the same module's quality signal rates them trusted (145 recent merges classified success; verified by running the production query against the live ledger).

The cadence check is a spend control against unproven accounts. A trusted submitter — the module's own definition of proven: 5+ recent merged successes at a low fail rate, recency-windowed — now short-circuits it. The exemption can't be farmed into a loophole: reaching "trusted" requires real merged PRs on that repo inside the window, and a single prompt-injection row hard-locks the signal to low before success counting — pinned by an invariant test.

2. The hold finding described the heuristic, not reality. forceAiReview (the panel checkbox / @loopover review) and aiReviewAllAuthors both bypass the reputation skip in shouldStartAiReviewForAdvisory — but maybeAddReputationSkipHold pushed the "was skipped" finding on reputationSkipped alone. So a forced re-run produced a real review and a bullet claiming it was skipped, and held the gate anyway. The hold now returns early on either bypass.

Operational note

LOOPOVER_REVIEW_REPUTATION is commented out on the ORB (with a dated note) until this ships in an image — the owner shouldn't be rate-limited by their own anti-abuse system in the meantime. Re-enabling is a one-line uncomment.

Tests

4 new: the trusted-exemption regression (6 merges + 8 PRs at 4-minute pace → no downgrade), the prompt-injection invariant (banked merges + one injection → still downgraded), and one regression per bypass for the hold (forced re-run, aiReviewAllAuthors). 39 + 15 pass in the touched files; tsc clean.

[loopover-orb]

Important

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏳ LoopOver is waiting…

LoopOver has seen this pull request and is waiting on CI checks to finish before reviewing it. This comment will update once the review runs.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed · 🟨 Waiting}

[superagent-security]

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.52%. Comparing base (f5711a3) to head (366f2fe).
⚠️ Report is 3 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9800      +/-   ##
==========================================
- Coverage   90.33%   89.52%   -0.82%     
==========================================
  Files         918      918              
  Lines      113936   113938       +2     
  Branches    26975    26977       +2     
==========================================
- Hits       102926   101998     -928     
- Misses       9681    10850    +1169     
+ Partials     1329     1090     -239     

Flag	Coverage Δ
backend	94.09% <100.00%> (-1.48%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/queue/processors.ts	94.75% <100.00%> (+<0.01%)	⬆️
src/review/reputation-wire.ts	100.00% <100.00%> (ø)

... and 3 files with indirect coverage changes