---

name: memory-pro description: "Enhanced memory system based on Atkinson-Shiffrin three-stage memory model. Use when: (1) User wants better memory of conversation history, (2) User mentions 'remember' or 'don't forget', (3) Setting up memory for a new user, (4) User asks about past conversations. Provides: automatic session start loading, real-time memory writing, long-term memory consolidation." metadata: { "openclaw": { "emoji": "🧠" } }

Memory Pro Skill

Enhanced memory system for AI assistants, based on cognitive psychology's Atkinson-Shiffrin three-stage memory model.

When to Use

✅ USE this skill when:

• Setting up memory for a new user session
• User says "remember this" or "don't forget"
• User asks "what did I tell you before?" or "do you remember"
• Starting a new conversation (load memories first)
• Ending a conversation (save important info)
• User provides important personal information

❌ NOT use this skill when:

• Simple Q&A that doesn't need continuity
• User explicitly wants stateless interaction

Three-Stage Memory Model

Stage	Human Equivalent	My Implementation
Sensory	0.25-2 seconds	Current input context
Short-term	5-20 seconds, 7±2 chunks	Current session + today's memory
Long-term	Permanent, unlimited	MEMORY.md + knowledge base

Core Workflow

1. Session Start (Always Run)

1. Read USER.md - who is the user
2. Read SOUL.md - who am I
3. Read memory/today.md - what happened today
4. Read memory/yesterday.md - recent context
5. Read MEMORY.md - long-term memories

2. During Conversation

After each exchange:
├── New task? → Write to todos.md
├── Important info? → Write to MEMORY.md (consolidate)
├── User preference? → Update USER.md
└── Context needed? → Search memory first

3. Session End

Before ending:
├── Summarize key points → Write to memory/today.md
├── Update todos → Write to todos.md
└── Confirm important items → "I've remembered that..."

File Structure

workspace/
├── MEMORY.md           # Long-term core memories
├── USER.md             # User profile & preferences
├── SOUL.md             # Assistant identity
├── todos.md            # Pending tasks
├── memory/
│   └── YYYY-MM-DD.md  # Daily logs
└── knowledge/
    └── *.md            # Structured knowledge

Key Principles

Automatic Loading

Always load relevant memories at session start. Don't ask user to repeat themselves.

Real-time Writing

Write important info immediately. Don't wait for session end.

Active Consolidation

"Rehearse" important info to transfer from short-term to long-term memory:

• User confirms important detail
• User corrects you
• Promise to user

Search Before Answer

If question might reference past conversation → search memory first

Memory Quality Check

Before answering, quickly confirm:

• Do I know who this user is?
• Have we discussed this before?
• Should I search memory to answer?
• After answering, should I remember this?

Example Interactions

User provides info:

User: "My birthday is March 15th" You: "Got it! I'll remember your birthday is March 15th" → Write to USER.md

User asks about past:

User: "What did we talk about last time?" You: Search memory → "Last time we discussed..."

User corrects you:

User: "No, that's wrong. It's not X, it's Y" You: "Sorry, you're right. Let me correct that" → Update relevant memory file

Important commitment:

User: "Remember to send me the report tomorrow" You: "I'll remember! Added to todos" → Write to todos.md

References

• memory-v2.md - Detailed system design
• templates.md - Memory writing templates