evolution-loop

You can automate skill updates by scanning session logs, failure autopsies, and decision logs for recurring patterns, then auto-generating targeted skill modifications or new skills to resolve them. This process drives continuous self-improvement and tracks improvement metrics over time.

Failure autopsy analysis for self-improvement is the process of examining past failures and decision logs to identify recurring patterns. The evolution-loop automates this by scanning session outputs to detect issues and auto-create skills that prevent similar failures.

You should run pattern detection on session logs after a major session, lengthy operations, or when recurring issues are suspected. Operating at these points allows the system to identify repeating failure patterns and queue necessary skill updates efficiently.

Yes, you need retrospective notes, failure autopsy outputs, and decision logs. The evolution-loop requires access to these inputs to accurately detect recurring failure patterns across sessions and generate valid skill updates.

You measure the effectiveness of auto-created skills by tracking improvement metrics over time. The evolution-loop validates effectiveness by monitoring these metrics and applying guardrails to ensure the new skills actually resolve the detected failure patterns.

The best way to auto-generate new skills from discovered patterns is to use an automated evolution loop that scans session logs and failure autopsies, detects recurring issues, and directly authors skill updates while tracking their adoption metrics.