self-improvement

To capture agent workflow errors and user corrections for continuous learning, the skill appends structured entries into local learning logs using consistent schemas, preserving failed operations and corrected outputs for future review.

The best way to log unexpected tool failures during multi-agent operations is to use hook-based shell scripts that detect errors and append structured learning entries to the local .learnings directory.

Yes, you can promote validated learnings into project workspace guidance by elevating broadly applicable items from local learning logs into structured guidance documents for agents and tools.

You track feature requests and outdated knowledge discovered during agent-assisted work by instructing the agent to log these items using the skill's feature-request and error schemas into local learning logs.

No specific dependencies are required; the skill operates independently using included shell scripts to detect errors during agent workflows.