Repository navigation
pr-pipeline: re-run the review lanes on lane-source-changed instead of labeling needs-human #10076
Description
Problem
When review evidence unbinds as lane-source-changed (plugins/source-control/skills/babysit-prs/scripts/babysit_review_evidence.py:202), pr-merge holds the pull request on its risk score and labels it needs-human. A person can do nothing about it except ask for the reviews again.
Examples (escalation text: evidence-unbound:claude-review: not bound: lane-source-changed; evidence-unbound:claude-security-review: not bound: lane-source-changed): #9890, #9831, #9821, #9806, #9800.
Owner decision (2026-10-25)
A stop is worth a person only for security, intent, or a real charge. Re-running a review is Claude usage, which the cost rule (#9904) allows.
Fix
On an evidence-unbound reason whose cause is lane-source-changed, pr-merge dispatches both review lanes on the current head (once per head) and judges again when they report, instead of labeling. The unbinding rule stays as is: it is the security control that stops a head from forging its own review.
Done when
- An unbound-by-lane-source PR gets fresh reviews dispatched and no
needs-humanlabel. - A second unbind on the same head escalates as today (no loop).
Activity
github-actions commentedon Oct 25, 2026
Proposed work class: work-class: scoped
Basis: automated intake triage; the lanes App applied this class.
Size: Medium
Complexity: Medium
Uncertainty: Known
Flags: none
Blocked by: none
#10076 (comment)
This decision settles the open questions: dispatch mechanism and once-per-head marker, where the change lives, mixed-cause handling, and the design itself. Nothing is left to ask.
I understood the request as follows. When both review lanes unbind as
lane-source-changed, pr-merge dispatches fresh reviews once per head instead of labelingneeds-human. A second unbind on the same head escalates as today. The unbinding rule stays unchanged.Automated triage, applied by the lanes App.
melodic-automation-lanes commentedon Oct 25, 2026
Not admitted to the issue lane: no trusted-human acceptance criteria: a listed human must put the confirmed
- [ ]criteria in the body or post them as their own comment, or, on an issue a listed human filed, intake's newest refinement comment must hold them unedited. Admission takes the issue once each of those is met, when intake triages it again (.github/actions/resolve-config/admission-rules.json).melodic-automation-lanes commentedon Oct 25, 2026
The latest triage verdict holds nothing: the issue is at
status: ready.melodic-automation-lanes commentedon Oct 25, 2026
Refinement. A read-only run explored this repository, searched its issues and pull requests, and read the allowed docs to draft what follows. On an issue a listed human filed, it counts as the issue's spec once posted, while unedited; it is still model text, so check it.
Intended outcome: When a PR's review evidence is unbound as lane-source-changed, pr-merge dispatches both review lanes once on the current head, re-judges when they report, and escalates with needs-human only if the same head unbinds again.
Related work
Open questions
RECOMMENDED: Dispatch via workflow_dispatch from the pr-merge scripts and record a head-keyed HTML marker comment, like the needs-human-cleared marker, so a second unbind on the same head finds it and escalates. Basis: judgment; .github/scripts/pr-merge/clear-escalations.mjs:90
RECOMMENDED: Keep babysit_risk.py and babysit_review_evidence.py unchanged, so the unbinding rule stays as is. Detect the lane-source-changed cause in the pr-merge label path and skip the label there. Basis: plugins/source-control/skills/babysit-prs/scripts/babysit_risk.py:956
RECOMMENDED: Dispatch both lanes only when every unbound reason is lane-source-changed. Any other cause escalates as today. Basis: judgment
RECOMMENDED: Treat the 2026-10-25 owner decision in the body as settling intent only; the owner should confirm the dispatch mechanism and the one-retry rule. Basis: judgment
Sources read
Automated refinement by the lanes App. Answer with a comment whose first line is
**Decision recorded.**.kyle-sexton commentedon Oct 26, 2026
Decision recorded. The owner answered in the Refinement session (2026-10-25), deciding the open items below on the facts:
Question asked: Accept the body's Fix as the design, with the dispatch, placement and mixed-cause answers below?
The body's owner decision (2026-10-25) settles the intent:
Open items come from intake's refinement: #10076 (comment)
Decided:
workflow_dispatchfrom pr-merge, which already holdsactions: write, and a head-keyed HTML marker comment like theneeds-human-clearedmarker, so a second unbind on the same head finds it and escalates. Basis: origin/main.github/workflows/pr-merge.yml:134;.github/scripts/pr-merge/clear-escalations.mjs:90. Intake's answer kept.Decided: Leave
babysit_review_evidence.pyand its unbind rule unchanged. Skip the label in pr-merge's risk escalation,babysit_merge.py, which writes the<!-- babysit-merge-risk -->comment and the human-gated label, when every reason isevidence-unboundwith causelane-source-changed. Basis:plugins/source-control/skills/babysit-prs/scripts/babysit_merge.py:87;babysit_risk.py:956,971;babysit_review_evidence.py:213. Intake's answer corrected: it put the label path in the pr-merge.mjsscripts, butbabysit_merge.pyapplies it.lane-source-changedand the other for another cause, does pr-merge still dispatch?Decided: Dispatch only when every unbound reason is
lane-source-changed; any other cause escalates as today. Basis: judgment. Intake's answer kept.Decided: The body's Fix is the owner-filed design (dispatch both lanes on the current head once per head, judge again, keep the unbind rule). Items 1 to 3 are implementation within it. Basis: the issue body's Fix section. Intake's answer corrected: it asked the owner to confirm what the body already states.
Acceptance criteria:
needs-humanlabel. (Converted from the body's Done when, wording unchanged.)babysit_review_evidence.py's unbinding rule is unchanged.lane-source-changedescalates as today, with a test case.