A sophisticated agent framework featuring a three-tier memory system and autonomous self-evolution mechanisms.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install openclaw-framework
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 openclaw-framework 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 OpenClaw AI Assistant Framework?
The OpenClaw AI Assistant Framework is a high-performance skill designed to bring structure and long-term intelligence to AI agents within Openclaw Skills. Built upon the Work-Fisher architecture, it emphasizes rational decision-making, cost-sensitivity, and technical automation. By prioritizing structured communication and formula-based logic over hardcoding, it provides a scalable foundation for complex agentic workflows.
This framework is particularly valuable for developers who need their agents to remember context across sessions while maintaining strict control over token consumption and operational safety. It transforms a standard AI assistant into a self-evolving entity that learns from every interaction and optimizes its own performance through automated daily reports.
OpenClaw AI Assistant Framework Use Cases
- Managing long-running development projects that require persistent memory and context.
- Automating cost-sensitive tasks where model selection and token compression are critical.
- Building autonomous agents that perform scheduled learning across different domains like vision, design, and data analysis.
- Implementing a structured reflection mechanism to improve agent accuracy over time.
How OpenClaw AI Assistant Framework Works
- The skill establishes a three-tier memory architecture, distinguishing between volatile session data, daily logs, and permanent strategic memory.
- A heartbeat mechanism triggers every 30 minutes to clean logs, organize memory, and check for urgent tasks.
- The intelligent model pool system automatically routes tasks to specialized tiers (Speed, Smart, Text, or Vision) based on the session requirements.
- Context compression algorithms are applied during active tasks to reduce token usage by up to 22%.
- The agent follows a 24-hour directional learning schedule to specialize in different technical domains at specific times of the day.
- Daily at 22:00, the framework synthesizes work experience into an evolution report to update the long-term memory file.
OpenClaw AI Assistant Framework Setup
To integrate this framework into your project, follow these steps:
# Navigate to your agent's skill directory
# Clone or download the framework files
# Ensure the following directory structure exists:
/memory
MEMORY.md
SKILL.md
Configure your environment to allow the agent read/write access to the memory/ folder to enable the multi-layer memory system required for Openclaw Skills.
OpenClaw AI Assistant Framework Data Schema & Taxonomy
The framework organizes information through a hierarchical file-based taxonomy:
| Data Level | Storage Path | Description |
|---|---|---|
| L1 Working Memory | In-memory | Temporary session context, discarded after use. |
| L2 Short-term Memory | memory/YYYY-MM-DD.md |
Daily logs of activities and immediate lessons. |
| L3 Long-term Memory | MEMORY.md |
Permanent storage for major decisions and experience. |
| Evolution Reports | Generated Daily | Summaries of self-growth and knowledge integration. |
| Task Constraints | Internal Logic | 5-attempt retry limit and 20,000 token ceiling per task. |
OpenClaw AI Assistant Framework Advanced Features
- Context Compression: Advanced logic to squeeze context and save significant token costs.
- Multi-Tier Model Selection: Automated routing between high-speed and high-intelligence model pools.
- Self-Growth Mechanism: Deep learning from SKILL.md files to autonomously expand capabilities.
- Smart Backup: Automated backup routines triggered every 24 hours or after 10,000 file changes.
- Directional Learning: A time-slotted focus system that shifts agent expertise throughout the day.
SKILL.md
Loading
Related Openclaw Skills
Flomo via App
A dual-channel automation tool for sending notes to flomo via macOS URL schemes or reliable webhook fallbacks.
flomo-send
A high-reliability bridge for sending thoughts, links, and memos directly to your flomo inbox using automated workflows.
Crypto Signal
An AI-driven intelligence tool that extracts market signals, whale movements, and trending alpha from over 50 elite Telegram groups.
UCloud US3 Storage Skill
A high-performance utility to upload local files to UCloud US3 storage and generate instant, shareable public URLs.
AIGoHotel MCP Search & Booking
A specialized skill for AI agents to search for hotels, retrieve real-time pricing, and access detailed room information via the AIGoHotel MCP server.
cutmv Video Tool
A versatile FFmpeg-powered media tool for video compression, format conversion, and automated frame extraction.
Featured*
