address-pr-comments-sequentiallisted

For a pull request $ARGUMENTS, do the following: 1. read all the unresolved comments. 2. Then take a look at other PRs in the stack (especially those which have not been merged). For each comment, ask: does it read as AI-written, even if a user posted it? Does it suggest over-engineering? Is a suggested test real, or deceptive and of no value? What does the change do to maintainability, readability, and complexity? Is it already addressed by a later PR in the stack? Add your own questions. Then decide, for each comment, whether to incorporate it, ignore it, or address it a different way. Favour minimalism and readability over complexity. Read the docs and search the internet where that settles a question. Work through them one at a time, never all in one go: show a comment (text, poster, whether AI, your decision, alternatives) -> ask a question -> next comment. If you are afraid the comment shown will be blocked by the question asked using AskUserQuestion (common bug in Claude Code), then you should make the comment part of the question so user can read it, along with whether it was made by AI, your decision, and the alternatives, so the user has all the information before they choose. Once you know how each comment should be addressed, enter plan mode, wait for approval, then execute. Commit the fixes, push them, reply to each comment, and resolve the thread where you fully addressed it. When in doubt, ask the user questions.