skill-evolution-manager

To extract user feedback for skill documentation, trigger a session review that analyzes dialogue and translates actionable changes into a structured evolution JSON. This workflow captures preferences and fixes to continuously improve skills without losing historical context.

The best way to merge incremental skill updates is using the merge_evolution.py script to persist evolution data, followed by smart_stitch.py to align and document learnings directly into SKILL.md files under a dedicated best practices section.

Yes, you can apply feedback-driven evolution across multiple skills. The workflow operates across various skills during review sessions, generating incremental JSON changes that can be merged and stitched into their respective SKILL.md files to maintain consistency.

Stitching evolution data into SKILL.md preserves improvements by automatically writing structured learnings into a dedicated 'User-Learned Best Practices & Constraints' section, ensuring that feedback-driven changes survive subsequent skill updates and historical context is never lost.

The merge_evolution.py script requires a JSON string input containing arrays for fixes and preferences, such as '{"preferences": ["default to quiet prompts"], "fixes": ["path normalization"]}', which it then persists as structured evolution data for the specified skill.

No external dependencies are required to run the skill evolution and stitching scripts. The workflow operates independently using the included scripts to parse JSON feedback, merge evolution data, and stitch updates directly into local markdown documentation files.