What problem does it solve? AI coding assistants repeat the same mistakes because generic skill instructions never learn from a project's actual bug history. This Skill closes that loop by analyzing accumulated fix patches, recurring error patterns, and codebase conventions, then writing project-specific rules that make every other AI Factory skill smarter over time. ## Core Features & Use Cases - Incremental patch analysis: Reads patches from .ai-factory/patches/ using a cursor file with an overlap window, extracting every independent prevention point into a registry mapped to target skills. - Stale rule detection: Compares skill-context rules against updated base SKILL.md files, classifying them as fully covered, conflicting, or partially overlapping, and collects user decisions before applying changes. - Safe rule application: Writes improvements only to .ai-factory/skill-context/<skill-name>/SKILL.md for built-in aif-* skills (never editing package-owned files), while allowing direct edits for custom skills, and logs every evolution with a timestamped report. - Use Case: After ten bug-fix sessions, run the evolve command to discover that six patches involved null-reference errors, then automatically add a prioritized null-check guard rule to the fix skill's project context. ## Quick Start Ask the AI to run /aif-evolve all to analyze accumulated patches and propose project-specific rule improvements for every installed skill.
