ci(pr-review): launch review agents in the foreground - #564
Open
RyanCodrai wants to merge 1 commit into
Open
RyanCodrai wants to merge 1 commit into
RyanCodrai wants to merge 1 commit into
Conversation
#506 told the orchestrator to keep waiting inside its turn on the agents it launches. It now launches the parallel bug scans with run_in_background: true, gets 'Async agent launched' back, schedules a wakeup and ends its turn — and a one-turn SDK run never wakes, so no REVIEW_VERDICT line is written. Three runs in a row on #563 ended 'stated no verdict' this way (37392243274, 37393324699, and one before), each after ~50 s and 11-14 turns. The prompt now says how: launch every agent with run_in_background: false, several in one message to run them in parallel, all of which return within the same turn. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01C1gifr9K4ThHHxjoNctZTL
Contributor
|
✅ Passed review — No defects found in the workflow prompt change; YAML and quoting are intact.. (run) |
This branch has not been deployed
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Unblocks the review gate, which has stopped producing verdicts.
What happens. The review orchestrator launches its parallel bug-scan agents with
run_in_background: true, gets "Async agent launched" back, schedules a 20-minute wakeup ("fallback while waiting on review agents") and ends its turn. The action runs one SDK turn, so the wakeup never fires and the agents never report: noREVIEW_VERDICTline, and the gate reads "the review stated no verdict". Three runs in a row on #563 ended this way (runs 37392243274, 37393324699), each after ~50 s and 11–14 turns; the execution logs show therun_in_background: trueinputs and theScheduleWakeupcall.Why #506 doesn't cover it. #506 told the orchestrator to keep waiting inside its turn — but an agent already launched in the background has nothing to block on.
The change. One sentence in
--append-system-prompt: launch every agent withrun_in_background: false; several calls in one message still run in parallel and all return within the same turn. Plus a comment recording why.The workflow runs from
main, so this takes effect for/reviewon any PR once merged — #563 needs it to get a verdict.🤖 Generated with Claude Code
https://claude.ai/code/session_01C1gifr9K4ThHHxjoNctZTL