An autonomous self-evolution engine that enables AI agents to learn from execution history, extract insights, and optimize their own strategies.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-evolver
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 agent-evolver 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 Agent Evolver?
Agent Evolver is a sophisticated framework designed to provide AI agents with a continuous learning loop and self-improvement capabilities. By analyzing execution results—particularly failures—the skill extracts actionable insights and stores them in a persistent experience library. This allows agents to evolve beyond their initial programming, adapting their behavior based on real-world outcomes to increase success rates over time.
As a core component of the Openclaw Skills ecosystem, Agent Evolver bridges the gap between static prompting and dynamic intelligence. It utilizes Large Language Models (LLMs) to diagnose technical errors and vector databases to perform semantic lookups of historical solutions, ensuring that your agents become more efficient and resilient with every task they perform.
Agent Evolver Use Cases
- Automated failure analysis and root cause detection for complex agentic workflows.
- Continuous strategy optimization to improve success rates in repetitive tasks.
- Accumulating a specialized knowledge base for specific domains like code generation or data analysis.
- Implementing semantic search to retrieve proven solutions for similar historical errors.
How Agent Evolver Works
- The system monitors agent execution and triggers an analysis when failures occur or when optimized performance is requested.
- The engine uses an LLM to analyze the error logs and task input to determine the underlying issue.
- A structured ExperienceCapsule is generated, containing the error classification and a proposed solution or strategy adjustment.
- The insight is persisted in a SQLite database and indexed via ChromaDB for high-speed semantic retrieval.
- During future executions, the agent queries the evolution engine to apply relevant historical lessons to the current context.
Agent Evolver Setup
To integrate this skill into your workflow, ensure you have the necessary environment variables and dependencies configured.
# Install required dependencies
pip install chromadb openai pyyaml
# Set your API credentials
export OPENAI_API_KEY="your_api_key_here"
# Optional: Customize the database storage path
export EVOLVER_DB_PATH="~/.evolver/evolution.db"
# Run the evolution CLI to analyze a result
python3 scripts/evolution_cli.py analyze --result "ValueError: negative input"
Agent Evolver Data Schema & Taxonomy
The skill organizes learning data using a structured ExperienceCapsule model, ensuring that every lesson is searchable and actionable.
| Attribute | Type | Description |
|---|---|---|
id |
String | Unique identifier for the experience record. |
task_type |
String | Categorization (e.g., code_generation, calculation). |
status |
String | The outcome status of the original task. |
error_type |
String | The technical category of the analyzed failure. |
solution |
String | The refined strategy or fix derived from the experience. |
keywords |
Array | Tagged metadata used for classification and filtering. |
Agent Evolver Advanced Features
- Vector-based semantic search using ChromaDB for finding contextually similar historical experiences.
- Multi-task support tailored for code generation, data analysis, and document processing.
- Dynamic strategy versioning that allows agents to manage and rollback optimization attempts.
- Automated performance statistics providing deep insights into agent improvement and error distribution trends.
- Seamless integration with existing Openclaw Skills to enhance overall agent autonomy.
SKILL.md
Loading
Related Openclaw Skills
Sync Trending
An intelligent tech-monitoring agent that finds trending repositories and autonomously verifies them within your project's context.
WorkIQ for Microsoft 365
A natural language CLI tool for querying emails, meetings, documents, and Teams messages within the Microsoft 365 ecosystem.
McKinsey PPT Design Framework
Automate the creation of professional, consultant-grade PowerPoint presentations using a McKinsey-style design framework and python-pptx.
SVG Draw
A specialized skill for generating high-quality vector graphics and converting them to PNG format without external Python graphics libraries.
macOS Suite Read-only
A read-only utility to fetch Mail, Calendar, Notes, and Stocks data from macOS in structured JSON format for AI agents.
macOS Suite
A powerful automation toolkit for controlling native macOS applications including Mail, Calendar, Reminders, and Notes via AI-driven commands.
Featured*
