Applies Test-Driven Development principles to AI process documentation to ensure reliable agent behavior and eliminate documentation loopholes.
Writing Skills adapts the Red-Green-Refactor cycle specifically for creating AI agent capabilities. Instead of writing instructions and hoping they work, this skill guides you through creating pressure scenarios with subagents to witness baseline failures, writing documentation to fix those specific failures, and refactoring to close rationalization loopholes. It ensures every skill added to Claude Code is battle-tested, token-efficient, and functionally verified before deployment.
主な機能
01Loophole and rationalization detection
02Claude Search Optimization (CSO) patterns
03Subagent-based pressure testing
040 GitHub stars
05Red-Green-Refactor cycle for documentation
06Token-efficient skill formatting
ユースケース
01Creating new reusable skills for Claude Code or Codex
02Refining complex process guides to be more reliable
03Verifying existing documentation against agent rationalization