A multi-agent orchestration framework that launches three distinct AI personas to collaborate, refactor, and review code until a unanimous consensus is reached.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install three-minds
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 three-minds 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 Three Minds Multi-Agent Collaboration System?
The Three Minds skill is a powerful multi-agent collaboration system built for the Claude Code CLI. By utilizing Openclaw Skills, it enables developers to move beyond single-prompt interactions and instead deploy a team of three specialized AI agents that share a single working directory. These agents are not just conversational; they possess the capability to read project files, modify source code, and execute tests to verify their work.
This skill is designed to handle complex engineering tasks by simulating a professional team environment. Whether using the default configuration of an Architect, Engineer, and Reviewer, or a custom security-focused squad, the system ensures that every change is vetted from multiple perspectives. The collaboration continues in iterative rounds until all agents provide a YES vote, ensuring that the final output meets a high standard of technical excellence.
Three Minds Multi-Agent Collaboration System Use Cases
- High-stakes code reviews where security, performance, and readability must be audited simultaneously.
- Complex refactoring tasks that require structural changes followed by detailed implementation and verification.
- End-to-end feature development where an Architect defines the pattern and an Engineer implements the logic.
- Automated bug hunting and fixing across large directories using specialized debugging personas.
- Improving project documentation and standardization through collaborative editing and consensus.
How Three Minds Multi-Agent Collaboration System Works
- The user initiates the skill with a specific task and target directory using the Openclaw Skills environment.
- Three distinct AI agents are initialized based on a selected or custom configuration (e.g., Architect, Engineer, and Reviewer).
- In each round, the agents take turns analyzing the project, applying code changes, and running necessary terminal commands.
- After performing their tasks, each agent casts a vote: YES if the task is complete and correct, or NO if further improvements are needed.
- The workflow proceeds through successive rounds of collaboration as long as there is a dissenting vote or until the maximum round limit is reached.
- Upon completion, the system generates a comprehensive audit trail of the discussion and all file modifications.
Three Minds Multi-Agent Collaboration System Setup
To use this skill, ensure you have Node.js 18+ and the Claude Code CLI (claude) installed. You can then trigger the multi-agent workflow directly from your terminal.
# Install the skill and run a basic collaboration task
three-minds "Review and improve the error handling in this project" --dir ./src
# Execute a task using a specialized code-review team
three-minds "Audit for security vulnerabilities" --config code-review --dir ./
# Run a deep refactoring session with a higher round limit
three-minds "Refactor to functional components" --max-rounds 10
Three Minds Multi-Agent Collaboration System Data Schema & Taxonomy
The skill maintains a transparent data structure to track the multi-agent session history and consensus logic.
| Data Point | Format | Description |
|---|---|---|
| Session Logs | Markdown | A file named three-minds-{timestamp}.md containing the full transcript and voting history. |
| State Export | JSON | A complete data object of the session when using the --output flag. |
| Agent Personas | JSON | Configuration defining agent names, emojis, and specific system prompts. |
| Project Context | Filesystem | Direct modifications to the source code within the specified --dir. |
Three Minds Multi-Agent Collaboration System Advanced Features
- Custom Agent Personas: Define your own team of experts by creating a JSON configuration file with custom system prompts and roles.
- Consensus Mechanism: A strict voting protocol that prevents tasks from being marked complete until all three agents agree on the quality of the work.
- Shared Workspace Execution: Agents can see each other's file changes in real-time, allowing for true collaborative coding and testing.
- Flexible Round Management: Control the depth of the AI's 'thinking' process by adjusting the maximum number of collaboration rounds.
- Multi-Output Logging: Simultaneously output real-time terminal progress while saving detailed Markdown and JSON reports for later audit.
SKILL.md
Loading
Related Openclaw Skills
Claude Code Skill
A powerful interface to control Claude Code via MCP, enabling persistent sessions and multi-agent collaboration for coding tasks.
OpenClaw Async Task
A powerful background execution tool for Openclaw Skills that prevents HTTP timeouts during long-running operations.
Encrypted File Reader
A Python-powered utility for reading content from protected or encrypted local files within secure enterprise environments.
nadfunagent
An autonomous AI trading agent for the Nad.fun ecosystem that automates market scanning, token analysis, trade execution, and profit distribution.
Xian Node Management
A professional toolkit for deploying, configuring, and managing Xian blockchain nodes using Docker and CometBFT.
Xian SDK for Python
A comprehensive Python SDK for building decentralized applications, managing cryptographic wallets, and interacting with smart contracts on the Xian blockchain.
Featured*
