Skip to content

docs(review-pr): add false-positive rules from warp-server feedback#31

Draft
captainsafia wants to merge 1 commit into
mainfrom
update-review-pr-feedback-20260619
Draft

docs(review-pr): add false-positive rules from warp-server feedback#31
captainsafia wants to merge 1 commit into
mainfrom
update-review-pr-feedback-20260619

Conversation

@captainsafia

Copy link
Copy Markdown
Contributor

What

Adds two evidence-backed false-positive rules to the review-pr skill's Review Scope, derived from human feedback on agent-authored review comments.

Why

Ran the warp-server update-pr-review workflow over the last 7 days of oz-for-oss review comments on warpdotdev/warp-server PRs (15 agent comments that received human replies; all on code PRs). Most comments were validated by maintainers, but one recurring failure mode stood out:

  • warp-server PR #11910 (3 review threads): the bot asserted incorrect PostgreSQL semantics around chr(0) / NUL bytes and proposed a replace(..., chr(0), '') fix that would have guaranteed a query failure (chr(0) itself raises null character not permitted). Worse, those suggestions contradicted the bot's own earlier, correct comment in the same review asking to remove chr(0). The maintainer had to rebut the same point twice.

Changes

Two new bullets in .agents/skills/review-pr/SKILL.md:

  1. Verify language/runtime/database semantics before claiming code is broken; a confident CRITICAL/IMPORTANT comment with a wrong fix is worse than no comment. Lower severity and describe the risk when unsure rather than prescribing a concrete fix.
  2. Keep comments mutually consistent within a single review (don't propose removing a construct in one comment and adding it back in another).

Spec PRs in the window had no agent review comments (only human-to-human design discussion), so review-spec is intentionally left unchanged.

cc @captainsafia

Conversation: https://staging.warp.dev/conversation/a5946d89-3ced-4943-ba85-f2ef0ccd30be
Run: https://oz.staging.warp.dev/runs/019ee065-8132-7316-8773-a4abce136cd1

This PR was generated with Oz.

Add two evidence-backed rules to the review-pr skill, derived from human
replies to oz-for-oss review comments on warp-server PRs over the last week:

- Verify language/runtime/database semantics before flagging code as broken;
  a confident CRITICAL/IMPORTANT comment with a wrong fix is worse than none.
- Keep comments mutually consistent within a single review.

Both are backed by warp-server PR #11910, where the bot asserted incorrect
PostgreSQL chr(0)/NUL-byte semantics, proposed a fix that would guarantee a
query failure, and contradicted its own earlier (correct) comment; the
maintainer rebutted the same point across three threads.

Co-Authored-By: Oz <oz-agent@warp.dev>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant