A professional memory architecture for AI agents implementing episodic, semantic, and procedural storage to prevent context loss.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install memory-manager
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install memory-manager using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Memory Manager?
Memory Manager is a professional-grade architecture designed for AI agents to handle context with the same complexity as human cognitive systems. By implementing a three-tier memory pattern—episodic, semantic, and procedural—this tool ensures that agents don't just store data, but organize it for optimal retrieval. Using Openclaw Skills like this one allows developers to move beyond flat text files and implement a structured knowledge graph approach.
This system is built for agents that need to maintain long-term context without the overhead or privacy concerns of external vector databases. It provides local-first, human-readable storage that allows for 18.5% better retrieval efficiency. By monitoring memory usage and providing automated snapshots, it acts as a safeguard against the context loss typically associated with LLM token limits.
Memory Manager Use Cases
- Preventing critical context loss during high-token agent sessions.
- Categorizing raw agent logs into actionable procedural workflows.
- Building a local, private knowledge base of facts extracted from agent interactions.
- Auditing agent history through chronological episodic event logs.
- Detecting and managing memory compression risks before data is purged.
How Memory Manager Works
- The system initializes a specialized directory structure to house different types of agent memories.
- Episodic logs capture chronological events, providing a daily history of agent actions.
- The detection engine monitors memory volume and issues warnings at 70% and 85% capacity thresholds.
- An organization script parses raw memories to identify and migrate facts into semantic storage and workflows into procedural storage.
- Search utilities allow the agent to query specific memory tiers (episodic, semantic, or procedural) for highly relevant context retrieval.
Memory Manager Setup
To begin managing agent context with Openclaw Skills, run the initialization script to set up the directory structure:
~/.openclaw/skills/memory-manager/init.sh
You can then check for existing memory risks or organize your current flat-file memory logs:
# Check compression risk
~/.openclaw/skills/memory-manager/detect.sh
# Organize existing memories
~/.openclaw/skills/memory-manager/organize.sh
Memory Manager Data Schema & Taxonomy
The skill organizes data into a tiered hierarchy to maximize retrieval accuracy and human readability:
| Directory | Type | Description |
|---|---|---|
memory/episodic/ |
Episodic | Chronological event logs (e.g., YYYY-MM-DD.md). |
memory/semantic/ |
Semantic | Distilled facts, concepts, and permanent knowledge nodes. |
memory/procedural/ |
Procedural | Reusable step-by-step guides and workflow patterns. |
memory/snapshots/ |
Backups | Compressed snapshots taken before memory purging. |
memory/legacy/ |
Archive | Original files preserved after migration to the tiered structure. |
Memory Manager Advanced Features
- Multi-tier memory architecture including episodic, semantic, and procedural patterns.
- Automated compression risk detection with tiered warning thresholds (Safe, Warning, Critical).
- Intelligent organization engine that extracts facts and workflows from flat logs.
- Type-specific search capabilities to narrow down context retrieval by memory category.
- Usage statistics and growth rate monitoring to track agent learning over time.
- Local-first privacy, ensuring that agent identity and memory stay entirely on your hardware.
SKILL.md
Loading
Related Openclaw Skills
YouTube-Hook
An AI-driven YouTube content strategist that transforms any topic into high-performing video scripts, SEO-optimized metadata, and visual concepts.
moa-debate
A Mixture of Agents workflow that simulates a rigorous Oxford Union-style formal debate to stress-test arguments and explore complex topics.
Agent Passport
Agent Passport is a comprehensive security framework providing OAuth-style consent-gating, behavioral contracts, and threat-shielding for autonomous AI agents.
Zongjie: Structured Memory & Experience Management
A structured memory management system that distills complex troubleshooting and development experiences into a permanent knowledge base for AI agents.
PCAP Analyzer
A high-performance network forensics skill that transforms raw packet captures into actionable security reports using tshark.
Error-Driven Evolution
A structured system for AI agents to transform mistakes into imperative, executable rules that prevent recurring errors.
Featured*
