memory-manager

What This Skill Does

The Memory Manager for OpenClaw is a professional-grade architecture designed to solve the common 'context window' problem. By implementing a three-tier memory system—Episodic (what happened), Semantic (what I know), and Procedural (how-to guides)—this skill enables agents to store, categorize, and retrieve information with extreme efficiency. It handles auto-snapshots, compression monitoring, and complex semantic searches, ensuring your agent never loses critical context even as its data footprint grows.

Installation

To install this skill, use the ClawHub registry command in your terminal: clawhub install openclaw/skills/skills/marmikcfc/memory-manager After installation, execute ~/.openclaw/skills/memory-manager/init.sh to scaffold your local memory directory structure.

Use Cases

• Long-term Project Continuity: Perfect for agents managing long-running tasks where session context would otherwise be lost.
• Knowledge Management: Use this to distill raw chat history into structured knowledge base files (Semantic Memory).
• Automated Workflows: Build a library of procedures (Procedural Memory) that your agent can reference to repeat complex tasks accurately.
• System Health: Prevent data loss by proactively monitoring memory capacity and triggering automated snapshots before the agent reaches capacity limits.

Example Prompts

1. "Check my current memory usage and run a snapshot if we are reaching the 85% capacity threshold."
2. "Search my procedural memory for how I handled the last Moltbook API integration and output the steps."
3. "Organize all notes from today into the episodic logs and extract any new facts about our project into the semantic folder."

Tips & Limitations

• Proactive Management: Integrate the suggested commands into your agent's periodic heartbeat; running organize.sh at the end of each day ensures your knowledge remains searchable and deduplicated.
• Storage Limits: While the skill manages compression, always be mindful of your local machine's disk space if dealing with massive logs.
• Naming Conventions: Stick to the generated file structure. Manual creation of files outside the episodic, semantic, or procedural directories may cause the search utility to miss your data. Regular maintenance of the semantic files helps keep retrieval accuracy high.