skill-optimizer-lawvable

To turn conversational feedback into skill improvements, identify actionable signals from your session and evaluate them against Complete, Precise, Atomic, and Stable criteria to generate structured update proposals.

Updating SKILL.md based on session signals requires evaluating each signal against four quality criteria. Fully qualified signals generate direct update proposals, while unqualified signals trigger requests for clarification.

Feedback analysis decides between direct updates and observations by evaluating each conversational signal against Complete, Precise, Atomic, and Stable criteria to determine if it can be added directly or requires further review.

You can generate changelog entries from session corrections. When a conversational signal meets all quality criteria, the system proposes direct updates for both your SKILL.md and CHANGELOG.md files.

Automated workflow improvement cannot process incomplete or imprecise signals directly. When signals fail to meet the Complete, Precise, Atomic, and Stable criteria, the system must request clarification or record them as observations for future review.

You need to run the self-improve command after finishing a skill session to start analyzing improvements. This maps the conversational signals from your recent interactions into actionable updates for your files.