A sophisticated dual-track memory system that prevents AI context loss through automated daily journaling and long-term knowledge distillation.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install xiaolongxia-memory
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 xiaolongxia-memory 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 robust architecture designed to solve the 'amnesia' problem inherent in large language models when context windows expire or sessions reset. It functions as an external cognitive layer for AI agents, ensuring that critical decisions, user preferences, and project progress are never lost. By utilizing Openclaw Skills, this system transforms volatile chat histories into a structured knowledge base.
The system operates on a dual-track principle: a daily diary acts as an inbox for raw interactions, while a centralized memory file serves as a curated repository of distilled wisdom. This approach ensures that the agent remains lightweight and efficient while retaining access to high-value historical data.
Memory Manager Use Cases
- Configuring memory management for new AI agents to ensure continuity across sessions.
- Implementing heartbeat-based periodic memory grooming and consolidation.
- Fixing issues where AI agents fail to remember previous instructions or user preferences.
- Adapting structured memory workflows for external platforms like Coze, Dify, or WorkBuddy.
- Managing long-term project context and technical research history.
How Memory Manager Works
- Initialization: The agent loads long-term context from MEMORY.md and recent context from the last two days of diary files at the start of every session.
- Active Recording: After high-value interactions, the agent automatically writes key points into a daily markdown file (memory/YYYY-MM-DD.md).
- Heartbeat Monitoring: The system checks a JSON state file to see if the last maintenance occurred more than 48 hours ago.
- Distillation: During a heartbeat trigger, the agent reviews recent diaries to extract significant insights and updates the main MEMORY.md file.
- Optimization: The agent proactively suggests starting new sessions when context grows too large, knowing all essential data is safely persisted.
Memory Manager Setup
To deploy the Memory Manager within your Openclaw Skills environment, establish the following directory structure and configuration files:
# Create the memory directory
mkdir -p workspace/memory
# Initialize the state tracker
echo '{"lastMemoryMaintenance": null}' > workspace/memory/heartbeat-state.json
# Create the primary memory file
touch workspace/MEMORY.md
Update your AGENTS.md to include instructions for reading MEMORY.md and the current day's diary during startup.
Memory Manager Data Schema & Taxonomy
The system organizes data using a hierarchical markdown-based taxonomy to ensure human readability and machine parsability:
| File | Format | Description |
|---|---|---|
MEMORY.md |
Markdown | Curated long-term insights, preferences, and project milestones. |
memory/YYYY-MM-DD.md |
Markdown | Raw logs of daily conversations, decisions, and events. |
heartbeat-state.json |
JSON | Tracking metadata for the last maintenance timestamp (lastMemoryMaintenance). |
SOUL.md |
Markdown | Definitions for the agent's recording habits and memory triggers. |
Memory Manager Advanced Features
- Periodic automated grooming: Automatically cleans up obsolete data every 48 hours via heartbeat tasks.
- Multi-platform compatibility: Portable prompt templates allow the memory logic to function on Coze, Dify, and other LLM orchestrators.
- Context Window Compression: Works alongside native platform compression to prioritize high-value persisted data over raw history.
- Active Reminder System: The agent monitors its own context length and triggers a 'memory save' before a session reset.
SKILL.md
Loading
Related Openclaw Skills
XLSX Spreadsheet Master
A professional-grade skill for automating complex Excel workflows, including formula recalculation, financial modeling, and data visualization.
Stock Monitor Pro
A professional-grade intelligent stock monitoring and alert system designed to track market anomalies and technical indicators with multi-condition resonance.
Xiaohongshu Search
A browser-based automation skill designed to search Xiaohongshu notes, extract engagement data, and analyze consumer trends for market research.
Downloader
A streamlined automation skill designed to download and store remote resources from any HTTP/HTTPS URL directly to a local directory.
NotebookLM PDF Cleaner
A precision tool designed to mask the NotebookLM footer badge on exported slide-decks for a professional, presentation-ready finish.
EmailOctopus Skill
A powerful integration for managing EmailOctopus lists, campaigns, and templates through AI-driven automation and secure credential management.
Featured*
