Memory Management Skill
This skill provides guidance on effectively using Claude's memory system to persist knowledge, decisions, and learnings across sessions.
When to Use This Skill
- •After completing significant work that yielded learnings
- •When making architectural or design decisions
- •When encountering and solving bugs that might recur
- •When establishing patterns or preferences
- •When needing to recall past context
Core Concepts
Memory Types - Choose the right type for durability:
- •
episodic: What happened (90-day TTL) - •
procedural: How to do things (permanent) - •
decision: Choices and rationale (permanent) - •
preference: User/project preferences (permanent)
Scope - Where memories live:
- •
core: Global, cross-project - •
project: Specific to current project
Importance - Affects decay rate:
- •
high: 2x TTL multiplier - •
medium: Standard TTL - •
low: 0.5x TTL multiplier
Quick Reference
| Task | Command |
|---|---|
| Store a decision | /wicked-mem:store "..." --type decision |
| Find past context | /wicked-mem:recall "query" |
| Check memory health | /wicked-mem:stats |
| Archive old memory | /wicked-mem:forget mem_id |
On-Demand Recall
Memories are pulled on-demand, not preloaded. When context is needed:
- •Auto-triggered: Prompts like "why did we decide..." auto-search memories
- •Manual fallback: If no context is injected and you need history, use:
code
/wicked-mem:recall "query terms"
When to proactively recall:
- •User asks about past decisions but no memory was injected
- •Task relates to previous work in the project
- •Making decisions that should reference past rationale
- •Debugging an issue that might have been solved before
Workflow Guides
| Task | Guide |
|---|---|
| What to store and how to structure it | Storing Decisions |
| Effective search and filtering | Effective Recall |
| Managing memory lifecycle | Memory Lifecycle |