comment-policy

To stop redundant code comments from bloating your AI coding agent context, enforce a write-time comment policy that strips WHAT-restating prose and changelog notes while retaining only non-obvious constraint tripwires and external rationale retrieval pointers.

For AI coding assistants, you should keep only tripwire comments that flag non-obvious constraints editors could break, and retrieval pointers that link to external rationale, removing all duplicate rationale and WHAT-restating prose.

You can enforce a comment policy across AI code-writing sub-agents by using a pre-written push block that delegates the FR-22 rules, ensuring all sub-agents apply the retention and removal logic even if they do not auto-load the skill.

The FR-22 write-time comment policy works with development workflows using AI-powered coding assistants like Claude Code or Cursor, automatically applying comment enforcement as you write or edit code files.

The best way to review code comments before an AI agent reads source code is to apply an automated pre-commit comment review that eliminates redundant comments and enforces retention of only useful tripwires and retrieval pointers.

Over-commented code files reduce AI agent performance because redundant WHAT-restating prose, changelog notes, and duplicate rationale waste valuable context budget, leaving less room for the actual source code logic the agent needs to process.