Post-task retrospective for agent self-improvement

Summary

This skill runs after a non-trivial task to review performance, categorising outcomes and identifying failures. It writes structured lessons to auto-memory, promoting reusable rules and updating the agent's knowledge base.