skill-evolution

Automate skill library maintenance by continuously evolving skills through a prioritized queue. This process analyzes the entire corpus, triages priorities based on staleness and risk, and applies bounded improvements to prevent silent decay.

The evolve loop works by analyzing audit-state, triaging items into a priority queue, executing one bounded action, verifying results, and checkpointing progress. This cycle ensures resumable and verifiable corpus maintenance across the skill library.

Yes, you can run corpus maintenance on a standalone repository. The system supports standalone workspace mode with pilot lanes, allowing you to analyze and evolve skills outside of a full Skill Graph deployment.

Audit-state.json is used as provenance for skill evolution. It records the audit status, staleness, and risk data that the prioritized queue analyzes to triage and execute bounded improvement cycles across the skill library.

The best way to prioritize skill reforms is by analyzing the entire skill library and ranking items based on audit status, staleness, and risk. This triage process feeds a bounded, verifiable execution queue for safe improvements.

You can resume a bounded improvement cycle by using the checkpoint system. The loop executes one bounded action per item, verifies results, and checkpoints progress to ensure the maintenance process remains resumable.