chore(skills): retry the plan-revise agent once before losing its findings - #615
Merged
Merged
Conversation
…dings The revise-after-pass step's agent() call never checked its return value: if that agent died mid-edit (observed live during /ship 590, a transient 529 Overloaded error), the loop silently continued to the next review pass on a plan file that could be partially/inconsistently edited, with that pass's accepted findings never folded in. Add the same death-check the review-runner step already has, retry once with an explicit instruction to reconcile any partial prior edit, and return status:'error' if it dies twice. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01LwFPT465eDJqYcRa8HGNLz
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What & why
plan-review-loop.workflow.mjs's revise-after-pass step never checked itsagent()call's return value — unlike the review-runner step, which already returns
status:'error'if the agent dies. Found live during/ship 590(2026-08-05): thatagent died twice on a transient
API Error: 529 Overloaded, and the loop silentlyadvanced to the next review pass on a plan file that could be left
partially/inconsistently edited, with that pass's accepted findings never confirmed
folded in. In the observed case a later pass happened to independently re-surface the
same gap, so nothing was actually lost — but that's luck, not a guarantee.
Adds the same death-check pattern the review-runner step already has: on death, retry
once with an explicit instruction to re-read the whole plan file and reconcile any
partial prior edit before folding in the findings; if it dies twice, return
status:'error'instead of silently continuing on an unknown plan state.Rebased onto current
main(this file was untouched by #614's ChatGPT-planner work,which lives in separate new files).
Checklist
src//tests/change, no coverage impact🤖 Generated with Claude Code
https://claude.ai/code/session_01LwFPT465eDJqYcRa8HGNLz