What problem does it solve? Codebases accumulate noisy comments that narrate obvious code, restate types, or duplicate documentation, making reviews harder and comments less trustworthy. This Skill enforces the dd-trace-rb comment conventions so every comment earns its place by explaining why, warning of hazards, or citing external sources. ## Core Features & Use Cases - Comment quality rules: Defines when to write a comment (non-obvious tradeoffs, thread-safety hazards, external citations, field invariants) and what to never write (narration, section banners, type restatements, copy-pasted comments). - Comment removal guidance: Provides rules for reviewing and deleting existing comments, including fixing the underlying defect (renaming, moving types to sig/) rather than just deleting. - Style and ratio checks: Enforces terse one-line comments and a ratio check of roughly one comment per 15 lines of new code, with exemptions for declaration-only blocks and @public_api YARD docstrings. - Use Case: While reviewing a pull request in lib/ or ext/, ask the AI to evaluate whether each new comment follows the repo conventions and to suggest deletions or renames where comments merely narrate the code. ## Quick Start Review the comments in my current diff and tell me which ones violate the repo comment conventions and how to fix them.
