Repository navigation
Review: the reviewer can read and cite recorded decisions, and an open finding is not raised twice (#634) - #636
Merged
Merged
Conversation
…n finding is not raised twice (#634) The reviewer's prompt points at $PENDULUM_IN/decisions.md for the decisions it cuts for length, but the review stage first wrote that folder just before the answer pass. Each reviewer call now starts with the issue and every trusted decision written there, uncut, each under its key, and the cut note tells the reviewer to read them before it judges. An entry a maintainer's decision assigns outside the PR can be judged {"met": false, "decision": "<key>"}. reviewPost keeps it only when the key is exactly the key of one of the stage's trusted decisions (a map lookup): it adds no comment, and the review body lists the entry as not met, assigned by that decision, with its author and link. Any other key is a plain not met. reviewPost adds no "not met" comment for an entry whose finding from an earlier review of Pendulum's is still open on the PR. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017VqP6hg8ahWzteM9zqi8pU
…rst) Review of #636: reviewsCover and coveredHead differ when a marked review has no commit_id; gating both on scope.first keeps them in step. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_017VqP6hg8ahWzteM9zqi8pU
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Closes #634
What changes
reviewerCall, which writes$PENDULUM_INfirst:issue.mdanddecisions.md, the latter with every trusted decision, uncut, each heading carrying its key. The write uses only what the stage already holds, so it reads nothing from GitHub and does not touch the checkout while the required checks run beside the reviewer. The prompt's decisions section shows the same keys, and its cut note now says to read the file "before you judge". The worker's section and note are unchanged.deferred:{"met": false, "decision": "<key>"}.reviewPosttakes the stage's trusteddecisions(authors with write access) and looks the key up in aMapof their keys. On an exact hit it adds no comment and the body line readsnot met; assigned by <author>'s decision <link>. Any other key, and every caller that passes no decisions (a session'spost-review, the bench replay), is a plain not met with its comment.metstays false. Decisions now carryurl(GitHub'shtml_url). Documented inreviewerPrompt, the review skill'sreference.mdanddocs/PLAN.md;bench/skills.jsonre-recorded.unaddressedReviewComments, never a third party's or a writer's comment) andreviewPostadds no "not met" comment for an entry one of them already carries. The body line says its finding is already open. A failed read posts the comment as before; an outage ends the stage as on any read.Trust
The key is only ever matched against the stage's own trusted decisions. Text in the reviewer's answer, the PR body, a reply or an untrusted comment cannot supply one: a key that is not in that list changes nothing. Whether the decision really assigns the entry elsewhere stays the reviewer's judgement, as the issue says.
Tests
Three regression tests, each failing on main:
handoff.test.mjs: a reviewer call starts withdecisions.mdholding every trusted decision under its key, those the prompt cuts included (realopenHandoff,reviewerCalland the sandboxed write).exit.test.mjs: a trusted key gets no comment and the body names the decision; an unknown key is a plain not met.exit.test.mjs: an entry with an open finding gets no second comment.One existing assert changed:
foundation.test.mjspins the review job'sreviewPost(...)call as source text, so its pinned string gains the two new arguments.🤖 Generated with Claude Code
https://claude.ai/code/session_017VqP6hg8ahWzteM9zqi8pU