Agents working on PRs in this repo can flood review threads by posting a separate reply to every individual review comment — often 7–10 in a row, each with a "Fixed in <sha>" status header that mirrors the reviewer's items 1:1. A recent example is #2704, where the latest Copilot review cycle generated 9 inline comments and the author posted 9 near-identical reply comments back.
A human maintainer wouldn't do this. A human would push the fix commit and leave one summary comment per review round.
AGENTS.md currently has no guidance on how to respond to PR reviews, so agents fall back to their default per-comment-reply behavior.
Proposal
Add a short "Responding to PR Review Comments" section to AGENTS.md with directive bullets covering:
- Agent self-disclosure (name the agent, model, and the human being acted for)
- One top-level summary comment per review round — do not reply on every comment
- Inline replies only when context is needed (disagreement, deferral, non-obvious fix)
- Never click "Resolve conversation" — that belongs to the reviewer or PR author
- No emoji, no celebratory framing, no checklist mirroring the reviewer's items, no restating what the reviewer wrote
- Re-request review once per round (when all feedback is addressed), not after every intermediate push
Agents working on PRs in this repo can flood review threads by posting a separate reply to every individual review comment — often 7–10 in a row, each with a "Fixed in
<sha>" status header that mirrors the reviewer's items 1:1. A recent example is #2704, where the latest Copilot review cycle generated 9 inline comments and the author posted 9 near-identical reply comments back.A human maintainer wouldn't do this. A human would push the fix commit and leave one summary comment per review round.
AGENTS.mdcurrently has no guidance on how to respond to PR reviews, so agents fall back to their default per-comment-reply behavior.Proposal
Add a short "Responding to PR Review Comments" section to
AGENTS.mdwith directive bullets covering: