What problem does it solve? Codebases accumulate stale, redundant, or misplaced comments—PR references, discussion notes, and duplicated documentation—that clutter files and mislead readers. Manually auditing every comment against a consistent discipline is slow and error-prone. ## Core Features & Use Cases - Comment audit and in-place editing: Each comment in the target files is kept, deleted, or rewritten according to comment-discipline tests, with edits applied directly via the Edit tool. - Flexible targeting: Audit files changed on the current branch relative to the default branch by default, or pass explicit file and directory paths (processed recursively) to clean up legacy code. - Dry-run mode: The --dry-run flag produces the full decision summary without applying any edits, supporting triage before committing to changes. - Diff audit and structured summary: After editing, the diff is audited for side effects on sibling docs or tests, and a per-file Markdown table reports each decision with line, action, test, and reason. - Use Case: After implementing a feature on a branch, run the skill to strip PR references and redundant header comments, tighten eslint-disable rationales, and receive a summary table of every decision before opening a pull request. ## Quick Start Ask the AI to run the revise-comments skill on the current branch's changed files, or pass explicit paths such as "revise-comments src/lib" to audit a specific directory.
