Skill Evolution

This tracks how you actually use and modify skills, then suggests improvements based on patterns it detects in your edits. It monitors success rates, builds confidence scores for suggestions, and keeps versioned snapshots so you can roll back if something breaks. The collaborative persuasion type means it asks before applying changes rather than auto-updating everything. Useful if you're maintaining a skill library and want data-driven refinement instead of guessing what needs work. The safety rails are solid: it alerts on performance drops over 20% and won't re-suggest things you've rejected. Think of it as a feedback loop that learns from how you correct AI-generated code or documentation.