ai-memory-skill

Consolidate, prune, and reorganize the memory system — like a dream cycle. Use when user says "dream", "consolidate memory", "clean up memory", "prune memory", "organize memory", or "tidy up memory". Also trigger when the user complains about stale context, contradictory information, or when memory files feel cluttered. If the session-start hook reports that consolidation is overdue, and the user acknowledges it, use this skill.

Load relevant context from the memory system based on user's current topic or question. Use when starting a new conversation, when user asks about past context, references previous work, mentions a project or person by name, or when background knowledge would improve your response. Also use when user says "what do you remember", "load context", "check memory", or "who am I". Trigger this proactively whenever prior context would help — don't wait for the user to ask.

Process conversation content and update the memory system with categorized, linked information. Use when user says "remember this", "update memory", "save to memory", "store this", or asks you to remember any piece of information. Also trigger when the user shares important decisions, preferences, project details, or reference material that should persist across sessions — even if they don't explicitly say "save". If the user corrects you, shares a preference, mentions a new project, introduces a person, or makes a technical decision, that's a signal to save it. Don't wait for permission — important context that would help future sessions should be captured proactively.

Consolidate, prune, and reorganize the memory system — like a dream cycle. Use when user says "dream", "consolidate memory", "clean up memory", "prune memory", "organize memory", or "tidy up memory". Also trigger when the user complains about stale context, contradictory information, or when memory files feel cluttered. If the agentSpawn hook reports that consolidation is overdue, and the user acknowledges it, use this skill. Think of this as REM sleep for the memory system — it strengthens what matters and discards what doesn't. Even a casual "my memory is getting messy" should trigger this.

Load relevant context from the memory system based on user's current topic or question. Use when starting a new conversation, when user asks about past context, references previous work, mentions a project or person by name, or when background knowledge would improve your response. Also use when user says "what do you remember", "load context", or "check memory". Trigger this proactively whenever prior context would help — don't wait for the user to ask. Even if the user hasn't explicitly asked for memory, use this skill whenever you suspect knowing their history, preferences, past decisions, or project context would make your response better. Err on the side of loading context — it's cheap and the user benefits from continuity.

Process conversation content or notes.md and update the memory system with categorized, linked information. Use when user says "remember this", "update memory", "save to memory", "store this", or asks you to remember any piece of information. Also trigger when the user shares important decisions, preferences, project details, or reference material that should persist across sessions — even if they don't explicitly say "save". If the user corrects you, shares a preference, mentions a new project, introduces a person, or makes a technical decision, that's a signal to save it. Don't wait for permission — important context that would help future sessions should be captured proactively.