What problem does it solve? Over time, installed skills accumulate overlapping principles that belong in shared rule files, but manually cross-reading dozens of skills to find them is impractical. This Skill automates the discovery of recurring principles across skills and proposes concrete additions or revisions to your rules, with user approval required before any change. ## Core Features & Use Cases - Deterministic Inventory: Shell scripts exhaustively scan skill frontmatter and rule file headings, outputting structured JSON for analysis. - LLM Cross-Reading with Verdicts: Batched subagent analysis matches candidate principles against full rules text and assigns verdicts — Append, Revise, New Section, New File, Already Covered, or Too Specific. - User-Gated Execution: Every candidate is presented in a summary table; rules are never modified automatically, and results are saved to results.json. - Use Case: After installing several new skills, run a monthly rules maintenance pass to surface principles like "treat LLM output as untrusted" that appear in multiple skills but are missing from your security rules. ## Quick Start Ask the AI to run rules-distill to scan all installed skills and propose updates to your rule files.
