A framework for building AI agents that evolve through accumulated memory, modular skills, and refined behavioral protocols.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install evolutionary-model
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 evolutionary-model 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 Evolutionary Model?
The Evolutionary Model represents a shift from static chatbots to dynamic, self-improving assistants. By utilizing Openclaw Skills, this framework ensures that an agent grows smarter with every session, accumulating context and capabilities that are specific to its owner's needs. The model operates on the philosophy that an agent's intelligence should not just come from the underlying LLM, but from a structured, evolving codebase of files that the user owns and controls.
At its core, the model organizes an agent into four distinct layers: Identity, Memory, Skills, and Protocols. This architecture allows the agent to persist across sessions, learn from its mistakes, and gain new abilities without manual reconfiguration. By adopting this approach through Openclaw Skills, developers create a compounding value loop where the agent becomes increasingly irreplaceable as it documents more project context and specialized workflows.
Evolutionary Model Use Cases
- Setting up a highly personalized AI agent from scratch that understands your specific business context.
- Onboarding development teams to standardized AI-native workflows using shared Openclaw Skills.
- Auditing existing AI agent configurations to identify and fix gaps in long-term memory or operational reliability.
- Creating a portable knowledge base for an AI assistant that can be moved between different LLM providers without losing context.
How Evolutionary Model Works
- The agent initializes Layer 0 (Identity) by reading SOUL.md and USER.md to understand its persona and the owner's preferences.
- During each session, the agent references Layer 1 (Memory) by reading SESSION-STATE.md and previous daily logs to maintain context.
- The agent monitors for specific triggers defined in the when_to_use metadata of available Openclaw Skills to activate specialized capabilities.
- As the agent works, it records new insights and decisions into a daily log, which is later distilled into curated long-term memory files.
- If a mistake occurs, the agent performs a root cause analysis and updates Layer 3 (Protocols) in AGENTS.md to prevent the error from recurring in future sessions.
Evolutionary Model Setup
To begin building an evolving agent with Openclaw Skills, initialize the minimum viable directory structure.
mkdir -p clawd/memory clawd/skills
touch clawd/SOUL.md clawd/IDENTITY.md clawd/USER.md clawd/AGENTS.md clawd/MEMORY.md
Next, define your agent's core personality in SOUL.md and your personal preferences in USER.md. Once the structure is ready, the agent will begin reading these files at the start of every session to establish its operational context.
Evolutionary Model Data Schema & Taxonomy
The Evolutionary Model organizes data into a hierarchical file structure to ensure the persistence of Openclaw Skills and session data.
| Component | File Path | Description |
|---|---|---|
| Identity | SOUL.md |
Defines personality, values, and core operating principles. |
| User Context | USER.md |
Stores owner name, timezone, and communication preferences. |
| Memory | MEMORY.md |
A curated file for long-term project knowledge and decisions. |
| Session Log | memory/YYYY-MM-DD.md |
Raw logs of daily interactions and task outcomes. |
| Skills | skills/{name}/SKILL.md |
Modular instructions with when_to_use metadata. |
| Protocols | AGENTS.md |
Fixed operating rules and safety constraints for the agent. |
Evolutionary Model Advanced Features
- Trigger-based automation using when_to_use frontmatter in Openclaw Skills to ensure the right tool is used at the right time.
- Proactive heartbeat systems that allow the agent to run scheduled check-ins and maintenance tasks independently.
- Decoupled script execution where Python or Bash helpers in the scripts folder provide extensible power beyond text generation.
- Mistake-to-Protocol evolution where the agent self-corrects by updating its own AGENTS.md file after analyzing failures.
- Provider-agnostic architecture allowing you to move your entire Openclaw Skills library between different AI models and platforms.
SKILL.md
Loading
Related Openclaw Skills
Personal OS Onboarding
A conversational setup tool that creates the foundational memory and identity files for your autonomous AI agent.
SeaArt Video Generator
A powerful AI video generation skill for creating cinematic content from text prompts or static images using the SeaArt platform.
OpenClaw LanceDB Memory System
An intelligent memory system powered by LanceDB providing AI agents with persistent, semantic, and cross-session memory capabilities.
Step3-VL-10B Fine-tuning Skill
A technical workflow for fine-tuning the Step3-VL-10B vision-language model using LoRA or full-parameter methods on GPU clusters.
CPA Manager
A professional CLIProxyAPI maintenance tool for inventory scanning, 401/quota cleaning, and auth file management.
Community Onboarding
An adaptive AI agent skill designed to onboard new users by providing clear, hype-free explanations and low-risk entry points into a community.
Featured*
