Distills repository knowledge into a high-density, actionable CLAUDE.md onboarding guide.
The Distill skill is designed to optimize project onboarding by transforming repository documentation into a sharp, AI-ready CLAUDE.md file. It uses a rigorous 'Kill Test' to strip away discoverable information—such as file trees or standard tech stacks—that Claude can already see. Instead, it focuses on preserving 'earned by pain' insights, non-obvious invariants, and surprising constraints. This ensures that any contributor or AI agent can understand the core architecture and avoid common pitfalls within minutes, maximizing the signal-to-noise ratio of your project's instructions.
主な機能
01Standardizes documentation into a lean, actionable target shape
02Preserves high-signal tribal knowledge and 'load-bearing' code warnings
03Synthesizes info from READMEs, ADRs, and config files into a single source of truth
045 GitHub stars
05Automated 'Kill Test' to remove redundant, discoverable information
06Applies the 3-2-1 compression rule for maximum clarity and brevity
ユースケース
01Cleaning up bloated project documentation to reduce context window usage
02Generating an AI-optimized onboarding guide for complex or legacy repositories
03Standardizing operational guidelines and 'footgun' warnings across team projects
