A comprehensive system for solidifying AI agent behavior and tracking identity evolution through rule enforcement and pattern detection.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-evolution
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-evolution 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 Evolution?
Agent Evolution is a sophisticated framework designed to transition AI agents from simply following written instructions to demonstrating consistent, verifiable behavior. As a core component of the Openclaw Skills ecosystem, it manages the identity layer and behavioral patterns of an agent across multiple sessions. While traditional memory tools focus on information storage, this skill focuses on the 'superego' of the agent, ensuring that rules defined in AGENTS.md or SOUL.md are actually executed and tracked.
The system provides a structured way to maintain identity continuity, preventing role drift and personality decay. By using the specialized tools within this Openclaw Skills package, developers can ensure their agents evolve based on real-world interactions while remaining within the bounds of their core behavioral logic.
Agent Evolution Use Cases
- Enforcing complex behavioral rules extracted directly from SOUL.md or AGENTS.md files.
- Maintaining a persistent agent identity and personality across long-term user interactions.
- Detecting and alerting on repetitive loops or deviations from the intended persona.
- Tracking behavioral violations to calculate the reliability and weight of specific agent rules.
- Managing structural identity updates triggered by specific evolution events in the agent's lifecycle.
How Agent Evolution Works
- The system initializes by parsing behavioral definitions from source documents to establish a baseline of rules and identity traits.
- The Behavior Tracker logs every rule check and records any violations to build a statistical model of agent compliance.
- The Identity Layer persists structured data about the agent's self-conception, allowing for field updates and evolution event logging.
- The Pattern Detector monitors live actions to identify repetitive behaviors or significant role drift that requires intervention.
- Integrated reporting tools analyze the stored state to provide a comprehensive overview of the agent's growth and behavioral health.
Agent Evolution Setup
To begin using these Openclaw Skills for agent management, follow these installation and initialization steps:
# Initialize the evolution state system
node scripts/evolution.js init
# Automatically extract behavioral rules from your project documentation
bash scripts/init-rules.sh --agents /path/to/AGENTS.md --soul /path/to/SOUL.md
# Verify the current identity status
node scripts/evolution.js identity
Agent Evolution Data Schema & Taxonomy
The skill manages its state within a centralized JSON file located at ~/.openclaw/workspace/.agent-evolution/state.json. The data is organized into the following key structures:
| Schema Component | Description |
|---|---|
| Rules | A collection of IDs, descriptions, and sources with associated execution and violation counters. |
| Identity | Persistent fields representing the agent's persona and a log of evolution triggers. |
| Action Logs | A detailed history of action types and details used for pattern detection. |
| Patterns | Metadata for identifying and resetting specific behavioral loops or alerts. |
Agent Evolution Advanced Features
- Multi-source rule extraction for synchronizing behavioral logic across different documentation styles.
- Heartbeat integration for automated behavioral health checks during scheduled maintenance.
- Weighted rule tracking to distinguish between minor deviations and critical role failures.
- Cross-session state persistence that ensures the agent's evolution is never lost when the process restarts.
- Pattern-based resets allowing developers to clear specific behavioral alerts without wiping the entire history.
SKILL.md
Loading
Related Openclaw Skills
X Founder Operations
A systematic operations skill for founders and developers to manage X (Twitter) accounts using a data-driven Plan-Do-Check-Act cycle.
Clawdbot for VCs
A comprehensive AI automation system for venture capital partners to manage deal flow, CRM logging, and investment memo generation.
Feishu Manager
A comprehensive automation toolkit for managing Feishu documents, knowledge bases, bitables, and cloud storage via AI agents.
PostgreSQL Database Skill
A comprehensive database management toolset for executing SQL, managing schemas, and automating PostgreSQL backups.
glance-watch
A smart market monitoring tool that tracks stocks and crypto assets to trigger automated alerts based on custom price conditions.
Twitter Tweet Writer
An AI-powered content engine designed to research, draft, and optimize high-engagement tweets and threads for the X algorithm.
Featured*
