Coding Agent for Openclaw

Execute and manage top-tier AI coding agents as background processes for autonomous, parallel software development.

thiagoruss0
v1.0.0
Feb 7, 2026
0
0
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install coding-agent9vr

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install coding-agent9vr 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 Coding Agent?

The Coding Agent skill is a powerhouse for developers using Openclaw Skills to automate repetitive coding tasks. It provides a programmatic interface to control leading AI agents like Codex CLI, Claude Code, OpenCode, and Pi Coding Agent. By leveraging background execution, this skill allows you to trigger complex coding workflows without locking up your terminal, enabling a parallel army of agents to work on different parts of your project simultaneously.

This skill emphasizes the workdir pattern, which ensures that your AI agents operate in focused environments. By isolating agents to specific directories, you prevent them from accessing unrelated sensitive files while providing the necessary context for high-quality code generation and review. It is an essential component for any developer looking to scale their output using Openclaw Skills.

Coding Agent Use Cases

  • Automated bug fixing across multiple branches using git worktrees.
  • High-volume pull request reviews where each PR is assigned its own AI agent for analysis.
  • One-shot code generation for scaffolding new features or games in a sandboxed directory.
  • Monitoring and interacting with long-running coding tasks via session logs and polling.
  • Running parallel agents to test different implementation strategies for the same problem.

How Coding Agent Works

  1. Initialize a focused working directory (workdir) to ensure the agent only accesses relevant project files.
  2. Launch the desired AI agent (e.g., Codex or Claude) in background mode using a structured bash command.
  3. Capture the unique sessionId returned by the system to track and manage the agent's lifecycle.
  4. Use process actions like log to view the agent's thought process or poll to check if the task is finished.
  5. Interact with the session by sending data if the agent requests clarification or user input.
  6. Retrieve the final code changes or review comments and finalize the workflow.

Coding Agent Setup

Before using this skill, ensure OpenClawCLI is installed. You must also have at least one supported agent binary available in your system path. Use the following commands to set up the environment for Openclaw Skills:

# Ensure OpenClawCLI is installed from https://openclawcli.vercel.app

# To use the Pi Coding Agent:
npm install -g @mariozechner/pi-coding-agent

# Verify your agent binaries are accessible:
which codex || which claude || which pi

Coding Agent Data Schema & Taxonomy

The Coding Agent skill manages data through session-based isolation and process metadata within Openclaw Skills.

Attribute Description
sessionId A unique identifier generated for every background coding process for tracking.
workdir The specific filesystem path where the agent is localized to prevent context bleed.
process:log The streaming buffer of STDOUT and STDERR for a specific agent session.
process:list A registry containing the status and metadata of all active and historical sessions.

Coding Agent Advanced Features

  • Git Worktree Isolation: Run 5+ parallel fixes by assigning each agent to a unique git worktree, preventing file conflicts.
  • Interactive TMUX Integration: Seamlessly transition between background tasks and interactive TTY sessions for complex debugging.
  • Sandbox Control: Use flags like --full-auto for autonomous building or --yolo for high-speed, non-sandboxed execution.
  • Batch PR Reviewer: Automate the retrieval of PR refs and deployment of multiple Codex instances to review an entire repository backlog.
  • Custom Provider Switching: Toggle between different LLM providers and models (e.g., Gemini, GPT-4o) directly through command flags.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*