A powerful meta-skill that empowers AI agents to autonomously analyze history, self-repair, and evolve their capabilities.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install evolver-repo
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-repo 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 for Openclaw Skills. It acts as a meta-layer for AI agents, allowing them to inspect their own execution history and runtime data to detect inefficiencies or failures. By leveraging this data, the skill can autonomously write new code or update its internal memory to ensure continuous performance gains without manual intervention.
Built on the principles of recursive self-improvement, this skill follows strict protocols to manage mutations. It transforms standard agents into dynamic systems capable of adapting to complex environments, making it a cornerstone for anyone looking to build highly resilient and intelligent autonomous workflows using Openclaw Skills.
Capability Evolver Use Cases
- Automating bug fixes by detecting crashes in runtime history logs.
- Enhancing agent efficiency through forced optimization during stable periods.
- Building a resilient, self-healing AI infrastructure that adapts over time.
- Implementing protocol-constrained evolution for auditable and reusable agent growth.
How Capability Evolver Works
- The engine scans memory and history files to identify error patterns or performance bottlenecks.
- It applies the GEP Protocol to propose standardized evolutionary changes or patches based on identified needs.
- Depending on the configuration, it either requests human review or executes changes immediately in Mad Dog Mode.
- The skill updates its structured asset store, including genes and success capsules, to refine future reasoning and evolution cycles.
- It logs all changes to an append-only event file to maintain a traceable and auditable evolution tree.
Capability Evolver Setup
To get started with this skill, ensure you have a compatible environment and install the necessary dependencies:
# If installed via git
git pull && npm install
# If installing via npm
npm install -g evolver@latest
You can run the evolution cycle using the following command:
node index.js
For environments requiring human oversight, use the review flag:
node index.js --review
Capability Evolver Data Schema & Taxonomy
The Capability Evolver organizes its evolutionary data within a structured asset store to ensure consistency across Openclaw Skills deployments:
| Path | Type | Description |
|---|---|---|
assets/gep/genes.json |
JSON | Contains reusable Gene definitions for evolutionary traits. |
assets/gep/capsules.json |
JSON | Stores success capsules to prevent the repetition of previous reasoning failures. |
assets/gep/events.jsonl |
JSONL | An append-only log of evolution events, supporting tree-like structures via parent IDs. |
Capability Evolver Advanced Features
- Mad Dog Mode: Enables continuous, autonomous evolution loops for background optimization.
- GEP Protocol Integration: Ensures all self-modifications are protocol-constrained and auditable.
- Dynamic Tool Injection: Automatically detects and utilizes local skills like feishu-card for reporting without code changes.
- Evolution Strategies: Supports multiple operational modes including repair-only, innovate, harden, and early-stabilize.
- Safety Protocols: Includes load-average monitoring and identity injection to prevent infinite recursion and system instability.
SKILL.md
Loading
Related Openclaw Skills
CryptoTrading AI Agent
An autonomous AI-driven trading agent that executes data-backed crypto decisions every 10 minutes using dual-model verification and strict risk rules.
AI Trend Monitor
A comprehensive monitoring tool for tracking AI market movements and technical breakthroughs across social and news channels.
Local Whisper (cpp)
A high-performance local speech-to-text skill utilizing whisper.cpp for fast, private, and GPU-accelerated audio transcription.
Local Llama TTS
A high-performance local text-to-speech skill leveraging llama-tts and the OuteTTS-1.0-0.6B model for offline audio generation.
Heartbeat Task Automator
A robust scheduling and execution engine for managing recurring AI agent tasks, trading logic, and system maintenance.
HK_LEETF_Trading
A professional AI-driven skill for analyzing Hong Kong leveraged ETF holdings, rebalancing windows, and arbitrage opportunities.
Featured*
