A sophisticated Project Manager layer for Claude Code that orchestrates complex task lifecycles using tmux and asynchronous hooks.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install claude-agent
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 claude-agent 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 Claude Agent?
The Claude Agent is a specialized skill designed to transform Claude Code into a comprehensive Project Manager operating system. Rather than acting as a simple remote controller, this skill enables the agent to understand complex requirements, design implementation plans, supervise execution, and perform quality checks. It bridges the gap between high-level project goals and low-level CLI execution, making it an essential component of modern Openclaw Skills workflows.
By utilizing tmux for interactive CLI management and custom hooks for asynchronous wake-ups, the Claude Agent can handle long-running tasks autonomously. It features dual execution modes—auto-approval for high-velocity development and manual approval for sensitive operations—ensuring that the developer maintains the right level of control without being bogged down by repetitive monitoring tasks.
Claude Agent Use Cases
- Managing complex coding tasks that require multi-turn reasoning and iterative refinement.
- Automating local knowledge base updates for AI coding assistants to stay current with official documentation.
- Orchestrating parallel development workflows across multiple branches using git worktrees and isolated tmux sessions.
- Providing a centralized control layer for MCP tools and custom system prompts within a CLI environment.
- Setting up event-driven notifications for task completion or permission requests via Telegram.
How Claude Agent Works
- Requirement Synthesis: The agent analyzes the user's task description and proactively asks for clarification to define acceptance criteria.
- Strategic Planning: It evaluates the technical stack and selects the optimal Claude model (Opus, Sonnet, or Haiku) and necessary MCP tools.
- Prompt Engineering: Based on local prompting patterns, it constructs a precise execution instruction for the underlying Claude Code instance.
- Execution Management: The skill launches Claude Code inside a background tmux session, supporting both single-turn print commands and multi-turn interactive sessions.
- Asynchronous Monitoring: A pane monitor script or a Stop hook tracks the CLI output, detecting when the agent is finished or requires manual intervention.
- Reporting: Once the task meets the quality standards, the agent generates a summary report of changes and configuration updates for the user.
Claude Agent Setup
To get started with this skill within the Openclaw Skills framework, follow these steps:
- Ensure tmux and Claude Code are installed on your system.
- Configure your
~/.claude/settings.jsonto include the completion hooks:
{
"hooks": {
"Stop": [{
"matcher": "",
"hooks": [{
"type": "command",
"command": "python3 /path/to/skill/hooks/on_complete.py"
}]
}]
}
}
- Set the required environment variables for notifications:
export CLAUDE_AGENT_CHAT_ID='your_telegram_chat_id'
export CLAUDE_AGENT_NAME='main'
- Verify the installation by running
claude --versionand testing the Openclaw notification system.
Claude Agent Data Schema & Taxonomy
The Claude Agent organizes its operational data and technical knowledge into a structured directory hierarchy:
| Directory/File | Description | Use Case |
|---|---|---|
knowledge/ |
Technical documentation and capability maps | Referenced during prompt design and tool selection |
state/ |
Version tracking and update timestamps | Used for automated knowledge base maintenance |
hooks/ |
Python and Bash automation scripts | Manages async notifications and tmux pane monitoring |
settings.json |
Core configuration file | Defines permissions, MCP servers, and execution hooks |
workflows/ |
Detailed process guides | Standard operating procedures for specific tasks |
Claude Agent Advanced Features
- Multi-Model Orchestration: Dynamically switch between Claude 3.5 Opus for complex architecture and Haiku for fast, low-cost iterations.
- Event-Driven Wakeups: Integrated Stop hooks that trigger Python scripts to notify the user via Telegram upon task completion.
- Intelligent Permission Management: Granular control over tool execution with the ability to define allowed and denied bash commands.
- Git Worktree Integration: Support for concurrent task execution by isolating sessions in temporary worktrees.
- Automated Knowledge Sync: A specialized workflow to fetch, diff, and update local documentation from official GitHub releases and web sources.
SKILL.md
Loading
Related Openclaw Skills
Miaoxiang Portfolio Management (mx_self_select)
A sophisticated portfolio management skill that allows AI agents to interact with Eastmoney accounts via natural language to track and manage stock watchlists.
OpenClaw Security Hardening
A comprehensive security framework for OpenClaw agents to prevent data leaks and block prompt injection attacks.
Doubao Image Gen
A professional AI image generation skill leveraging the Doubao Seedream model for high-fidelity text-to-image conversion and batch processing.
Boss Agent
The Boss Agent acts as a central orchestrator that decomposes user commands and coordinates specialized sub-agents to execute complex workflows.
Openclaw AI Coding Assistant
A sophisticated AI coding agent skill designed to automate complex development workflows and codebase refactoring.
Sloth D2C: Figma Design to Frontend Code
A powerful Design-to-Code (D2C) skill that automates the conversion of Figma design nodes into clean, framework-specific frontend components.
Featured*
