| name | memory-core |
| description | Core memory management skill for OAMC (OpenClaw Adaptive Memory Core). |
| metadata | {"openclaw":{"emoji":"🧠","requires":{"plugins":["memory-x"]}}} |
Memory Core Skill
This skill provides the fundamental tools for the OpenClaw Adaptive Memory Core (OAMC).
It allows the agent to retain experiences, recall context, reflect on patterns, and self-evolve.
🛠️ Tools
1. memory_remember
Purpose: Store a new memory or experience.
- When to use: When the user tells you a fact, preference, or goal.
- Effect: Writes to the daily log (
memory/YYYY-MM-DD.md) and updates the search index.
2. memory_recall
Purpose: Retrieve relevant context.
- When to use: Before answering a question that relies on past interactions or user preferences.
- Effect: Searches themes, semantics, and episodes to provide dense context.
3. memory_reflect
Purpose: Mine patterns and suggest evolution.
- When to use: Periodically, or when you notice you are repeating mistakes.
- Effect: Scans recent history for high-frequency patterns and suggests new rules.
4. memory_evolve
Purpose: Update your own operating rules.
- When to use: When
memory_reflect provides a valid suggestion to update META.md.
- Effect: Writes a new Rule or SOP to
memory/META.md, permanently changing your behavior.
🔄 Workflow
- Ingest: Use
memory_remember to log important turns.
- Retrieve: Use
memory_recall to ground your responses.
- Reflect: Use
memory_reflect to identify patterns.
- Evolve: Use
memory_evolve to write new rules based on reflection.
🤖 Automatic Evolution
This skill includes a background process that:
- Runs
memory_reflect periodically (default: every 60 mins).
- If high-confidence patterns are found (e.g., repeated corrections), it automatically suggests or applies updates to
META.md.
- This ensures the agent evolves even without explicit user prompts to "reflect".