After your AI agent fumbles through a task or takes three detours to find information it should have known, run this to figure out what went wrong in your context network. It analyzes interaction transcripts across six dimensions like knowledge gaps, navigation issues, and whether the agent violated boundaries between planning docs and implementation files. The output tells you exactly what context nodes to add, what relationships to document, and which guidance rules need sharpening. Think of it as a post-mortem tool that turns every confused agent interaction into actionable improvements for your context architecture.
npx skills add https://github.com/jwynia/agent-skills --skill context-retrospective