fix(pr-review): per-agent skip + no self-triggered shadow runs#3
Merged
Merged
Conversation
Two field reports from the first live rounds: 1. A press re-ran ALL agents even when some had already cleared the exact head SHA — the same-SHA skip was all-or-nothing. Now each agent is skipped individually when its sticky marker covers the head SHA AND its positive label still stands (the verdict is read from the labels; no session spent). Agents that failed or errored at this SHA still re-run, so a press doubles as a re-roll of open verdicts; all-positive at the current SHA no-ops the whole press. muse:force re-runs everyone; hand-removing a positive label forces that agent. finalize counts skipped agents as positive for the muse:approved rollup and leaves their labels untouched. 2. Adding muse:reviewing used the app token, whose labeled event spawns a new run of the caller workflow — its jobs skip, but the newer (skipped) run replaced the live round in the PR checks view. All label writes now use github.token, whose events never trigger workflows: no more shadow runs mid-round.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two fixes from the first live rounds of the review loop:
1. Per-agent skip (stop burning sessions on already-approved agents)
The same-SHA skip was all-or-nothing, so a press re-ran all agents even when security/tests had already cleared the exact head SHA. Now each agent skips individually when its sticky marker covers the head SHA and its positive label still stands — the verdict is read from the labels, no Jules session spent.
muse:forcere-runs everyone; removing an agent's positive label by hand forces just that agentfinalizecounts skipped agents as positive for themuse:approvedrollup, labels untouched2. No self-triggered shadow runs
muse:reviewingwas added with the app token, whoselabeledevent spawns a new caller run — its jobs skip, but that newer (skipped) run replaced the live round in the PR checks view. All label writes now usegithub.token, whose events never trigger workflows.Verification
bash -nall run blocks · actionlint cleanmuse:review→ only the non-clear agents spin sessions; PR checks keep showing the live round