A self-evolution engine that consumes research logs and performs system self-checks to optimize AI agent architectures automatically.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install harness-evolve
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 harness-evolve 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 Harness Evolve?
Harness Evolve is a sophisticated self-improvement framework designed for AI agents to close the loop between research and action. It functions as a bridge that takes research discoveries and transforms them into tangible system improvements. As a key component in the ecosystem of Openclaw Skills, it manages the entire lifecycle of an agent's growth, from identifying redundant rules to proposing major architectural shifts.
The skill operates on a tiered execution model, ensuring that safe, non-breaking changes are applied immediately while complex logic updates are submitted for human review. By integrating Harness Evolve into your workflow, your AI agents gain the ability to conduct self-diagnostics, manage technical debt, and ensure their internal configurations remain aligned with their evolving operational goals.
Harness Evolve Use Cases
- Automating the transition of research findings into live system configurations.
- Continuous monitoring of AI agent log files for performance degradation or format regression.
- Managing complex rule sets by identifying and eliminating logical redundancies.
- Generating structured architectural proposals for human-in-the-loop review cycles.
- Ensuring AI agent persona settings and prompt templates stay synchronized with project changes.
How Harness Evolve Works
- Scans the project root to identify core configuration files, research logs, and memory directories.
- Extracts and prioritizes pending research items (P0/P1) from research logs for potential conversion.
- Performs a multi-point system self-check covering rule redundancy, complexity debt, and log anomalies.
- Classifies optimizations into three tiers: Direct Execution (Tier A), Review Proposals (Tier B), and Pattern Observation (Tier C).
- Executes safe updates and commits them to version control with structured commit messages.
- Generates a comprehensive daily evolution summary for downstream reporting and monitoring.
Harness Evolve Setup
To integrate this engine with your Openclaw Skills setup, ensure your project contains a CLAUDE.md or README.md for configuration discovery.
# Configure your research and log directories
mkdir -p research logs
# Initialize the research log if it doesn't exist
touch research/harness-log.md
# Run the evolution cycle
# The agent will automatically detect and process pending items
Harness Evolve Data Schema & Taxonomy
Harness Evolve uses a structured approach to manage system evolution data:
| Component | Path | Description |
|---|---|---|
| Research Log | research/harness-log.md |
Source of truth for new insights and mapping targets |
| Review File | pending-review.md |
Queue for Tier B proposals requiring human approval |
| Evolution Summary | research/evolve-today.md |
Daily report of all evolution activities and system health |
| Configuration | CLAUDE.md |
Primary source for identifying agent root and rule files |
Harness Evolve Advanced Features
- Multi-tier optimization logic ensuring zero-risk direct execution for documentation and safe files.
- Automated Git integration for version-controlled evolution history.
- Pattern recognition for identifying recurring manual tasks that should be abstracted into new Openclaw Skills.
- Custom notification hooks to alert developers when high-priority architectural proposals are ready for review.
- Complexity debt tracking to prevent prompt and rule bloating over long-term agent lifecycles.
SKILL.md
Loading
Related Openclaw Skills
Harness Research
Harness Research is an automated intelligence engine that tracks, analyzes, and maps the latest AI agent research papers directly to your system's improvement goals.
Creem Agent
An autonomous store operations manager that monitors SaaS health, manages failed payments, and handles churn winbacks for Creem.io merchants.
Axios Malware Scanner
A specialized security diagnostic tool to identify and mitigate the 2026 axios supply-chain attack on local development environments.
Memory System
A comprehensive long-term memory framework for AI agents featuring Markdown-based storage, semantic search, and structured feedback loops.
QMD Plus
QMD Plus enhances local markdown search by using LLMs to expand simple queries into sophisticated lexical and semantic search variations.
XiaoHongShu Content Migration
An automated solution for scraping XiaoHongShu notes and cross-posting them to Facebook Meta Business Suite and WordPress via REST API.
Featured*
