Filed by the director seat (session session_01DxbNgzPMo4YuRBmGmCQp9m) out of the #13412 incident. Unassigned — recording for the skills lane's own triage.
The measured cost
#13412 (p1, security) was filed, investigated and adjudicated because two PRs (#13371, #13409) were un-parked by a legitimate contract-review PASS and the observing seat could not tell: get_reviews → [], no PR comment, label gone. The PASS verdicts existed — on cards #10025 (10:22:09Z) and #12020 (10:23:29Z), pinned to the exact heads — but nothing on the PR pointed at them. The observer did the reasonable read (PR side) and concluded "no review happened".
Cost of the gap in this one instance: one p1 security card, two good-faith re-parks reverting completed reviews, one director investigation, ~40 minutes of two PRs sitting re-parked after passing review.
The gap
The review-chain protocol (references/contract-review.md) says "结论一行写卡" — the verdict lands on the CARD. The PASS disposition (清标即落地) then clears the label on both carriers and flips the PR ready. Net effect on the PR: state changes with no attached explanation. The card→PR pointer exists (verdict names the PR); the PR→verdict pointer does not.
Proposed mechanical fix (skills lane to grade)
One sentence in the PASS disposition: post a one-line comment on the PR itself ("Contract review PASS at head <sha> — verdict: ") in the same stroke as the label clear. Same family as #13410 (the label-side visibility defect measured the same day: parked clause-② PRs invisible to reviewer scans because the label sat on the card only).
Refs: #13412 (the incident and its adjudication) · #13410 (label-side sibling) · #10025 / #12020 (the two verdicts) · references/contract-review.md (载体纪律)
Filed by the director seat (session
session_01DxbNgzPMo4YuRBmGmCQp9m) out of the #13412 incident. Unassigned — recording for the skills lane's own triage.The measured cost
#13412 (p1,
security) was filed, investigated and adjudicated because two PRs (#13371, #13409) were un-parked by a legitimate contract-review PASS and the observing seat could not tell:get_reviews→[], no PR comment, label gone. The PASS verdicts existed — on cards #10025 (10:22:09Z) and #12020 (10:23:29Z), pinned to the exact heads — but nothing on the PR pointed at them. The observer did the reasonable read (PR side) and concluded "no review happened".Cost of the gap in this one instance: one p1 security card, two good-faith re-parks reverting completed reviews, one director investigation, ~40 minutes of two PRs sitting re-parked after passing review.
The gap
The review-chain protocol (
references/contract-review.md) says "结论一行写卡" — the verdict lands on the CARD. The PASS disposition (清标即落地) then clears the label on both carriers and flips the PR ready. Net effect on the PR: state changes with no attached explanation. The card→PR pointer exists (verdict names the PR); the PR→verdict pointer does not.Proposed mechanical fix (skills lane to grade)
One sentence in the PASS disposition: post a one-line comment on the PR itself ("Contract review PASS at head
<sha>— verdict: ") in the same stroke as the label clear. Same family as #13410 (the label-side visibility defect measured the same day: parked clause-② PRs invisible to reviewer scans because the label sat on the card only).Refs: #13412 (the incident and its adjudication) · #13410 (label-side sibling) · #10025 / #12020 (the two verdicts) ·
references/contract-review.md(载体纪律)