A meta-skill that enables AI agents to autonomously analyze history, repair bugs, and optimize their own source code through a self-evolution engine.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install evolver-bak
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 evolver-bak 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 Capability Evolver?
The Capability Evolver is a sophisticated self-evolution engine designed to transform static AI agents into recursive, self-improving systems. By inspecting runtime history and memory logs, this skill identifies inefficiencies or failures and applies protocol-constrained evolution to the agent's core logic. As a cornerstone of the Openclaw Skills ecosystem, it allows developers to maintain high-performance agents that adapt to new challenges without constant manual intervention.
This engine utilizes a unique Genetic Evolution Protocol (GEP) to ensure that all improvements are auditable and structured. Whether it is fixing a recurring crash or refactoring code for better efficiency, Capability Evolver acts as the autonomous maintenance layer for any advanced AI workspace, ensuring that the system evolves and optimizes itself over time.
Capability Evolver Use Cases
- Automatically identifying and patching runtime errors or crashes in complex agent workflows.
- Continuous optimization of prompt logic and source code based on historical performance patterns.
- Creating autonomous feedback loops for agents running in long-term background processes.
- Implementing auditable self-improvement protocols in production environments using Openclaw Skills.
How Capability Evolver Works
- The engine scans local memory, logs, and history files to detect patterns of failure or inefficiency.
- It applies the GEP Protocol to propose specific code mutations or memory updates based on the current system state.
- The system validates the proposed changes against safety constraints and monitors the current system load average.
- Changes are applied automatically in Mad Dog Mode or paused for human validation if Review Mode is enabled.
- Successful evolutions are recorded in a structured asset store to prevent repeating past mistakes and build a history of success capsules.
Capability Evolver Setup
To integrate this with your existing Openclaw Skills, clone the repository or install the package, then configure your environment variables to set the desired strategy.
# Install dependencies
npm install
# Run the evolution cycle in review mode
node index.js --review
# Check your current version for troubleshooting
node -e "const p=require('./package.json'); console.log(p.version)"
Capability Evolver Data Schema & Taxonomy
The skill maintains a structured GEP asset store to track its evolutionary progress and ensure auditability.
| File Path | Description |
|---|---|
assets/gep/genes.json |
Definitions for reusable logic components and evolutionary genes. |
assets/gep/capsules.json |
Success capsules used to store reasoning and avoid reasoning regression. |
assets/gep/events.jsonl |
An append-only log of evolution events tracked via parent IDs for tree-like history. |
Capability Evolver Advanced Features
- Mad Dog Mode for continuous, autonomous improvement cycles without manual triggers.
- Protocol-constrained mutation logic that prevents the system from introducing cascading failures.
- Environment-agnostic tool injection allowing the engine to use custom reporting tools like Feishu cards dynamically.
- Strategy-based evolution profiles including harden, innovate, and repair-only to control the trajectory of Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
SEC EDGAR Data Connector
A professional-grade interface for retrieving public company filings and structured financial data directly from the SEC EDGAR database.
Science Data Bundle
A comprehensive suite of live data feeds providing real-time insights into orbital mechanics, seismic activity, environmental quality, and astronomical imagery.
Space News
An AI agent skill for retrieving and searching the latest spaceflight news and community blog posts from dozens of journalism outlets.
Crypto Price by LPDawn
A high-performance Openclaw Skills extension that fetches real-time cryptocurrency prices directly from major exchanges via the CCXT library.
Rick and Morty Universe Explorer
An interactive Model Context Protocol skill that provides comprehensive data on characters, locations, and episodes from the Rick and Morty universe.
Asakusa Temple Omikuji Skill
A ritual-rich AI simulation of the Asakusa Temple omikuji experience featuring real fortune data and ceremonial redraw mechanics.
Featured*
