autopilot

Get a high-level overview of everything autopilot has memorized and learned

Search and retrieve memories from autopilot's memory system

Review what autopilot learned during this session

Analyze memories and apply structural changes — update CLAUDE.md, create skills

Use this agent when analyzing session transcripts for learnable patterns, when the user runs /autopilot:evolve, or when batch-processing memories for CLAUDE.md integration. This agent specializes in identifying preferences, conventions, corrections, and workflow patterns from conversation history, then proposing structured updates to CLAUDE.md files. Examples: <example> Context: User has completed a coding session and wants to capture learnings. user: "/autopilot:review" assistant: "Let me use the memory-analyst agent to analyze this session for learnings." <commentary> The session review command triggers the memory-analyst to scan conversation history for patterns, preferences, and corrections worth remembering. </commentary> </example> <example> Context: User wants to apply accumulated learnings to their CLAUDE.md. user: "/autopilot:evolve" assistant: "I'll use the memory-analyst agent to analyze your memories and propose CLAUDE.md updates." <commentary> The evolve command triggers the memory-analyst to analyze all stored memories, identify what's missing from CLAUDE.md, and propose integration changes. </commentary> </example> <example> Context: The PreCompact hook has fired and context is about to be compressed. system: "PreCompact event triggered" assistant: "Memory-analyst will flush important learnings before compaction." <commentary> Before context compaction, the memory-analyst quickly identifies and stores any learnings from the current context that haven't been persisted yet. </commentary> </example>

This skill should be used when the user says "remember this", "learn this", "save this preference", "store this convention", "don't forget", "always do X", "never do Y", "update my preferences", or "add to memory". It also activates when a clear user preference, correction, or workflow pattern is detected that should be persisted across sessions.

This skill should be used when the user asks "what do you remember", "recall", "search memories", "what have you learned", "what are my preferences", "do you remember", "show me what you know about", or "what conventions does this project use". Also activates when the user asks to "forget this", "remove this memory", "delete this preference", or when stored learnings need to be retrieved to inform a coding task, style decision, or project setup.

4 hooks across 4 events