What This Skill Does

The Coding Agent skill provides programmatic control over various AI coding assistants, including Codex CLI, Claude Code, Kiro CLI, OpenCode, and Pi Coding Agent. It allows you to launch and manage these agents as background processes directly from OpenClaw using bash commands. This is essential for integrating AI-powered coding into automated workflows and scripts. The skill leverages a pseudo-terminal (PTY) to ensure proper interaction with these command-line coding tools, preventing common issues like output corruption or process hangs. You can manage sessions, monitor progress, send input, and terminate processes, offering a robust solution for programmatic code generation and assistance.

Installation

To install the Coding Agent skill, use the following command in your OpenClaw environment:

clawhub install openclaw/skills/skills/dandysuper/openclaw-kirocli-coding-agent

Use Cases

This skill is invaluable for a range of development and automation tasks:

• Automated Code Generation: Generate boilerplate code, write unit tests, or create small scripts based on predefined prompts.
• Code Refactoring: Automate the process of refactoring code within existing projects.
• Bug Fixing Assistance: Integrate AI to suggest or implement fixes for detected bugs.
• Prototyping: Quickly generate code for new features or components during the prototyping phase.
• Learning and Exploration: Experiment with different AI coding models and their capabilities programmatically.
• CI/CD Integration: Incorporate AI code reviews or generation steps into your continuous integration and delivery pipelines.

Example Prompts

1. bash pty:true workdir:~/my_project background:true command:"codex exec 'Generate a Python Flask API endpoint for user registration'"
2. process action:log sessionId:XXX (Replace XXX with the actual session ID)
3. bash pty:true workdir:~/my_project command:"kirocli --prompt 'Refactor this Javascript function to be more efficient'"

Tips & Limitations

• PTY is Crucial: Always use pty:true when running coding agents, as they are interactive terminal applications. Failing to do so will likely result in unpredictable behavior.
• Background Processes: Utilize the background:true parameter for longer tasks to avoid blocking your main terminal. Remember to use the sessionId returned to monitor and control the process.
• Working Directory: The workdir parameter is highly recommended, especially when interacting with existing projects. It sets the context for the AI agent.
• Timeout: Be mindful of the timeout parameter to prevent processes from running indefinitely if they get stuck or take too long.
• Agent Specifics: Each coding agent (Codex, Claude Code, Kiro CLI, etc.) may have its own specific command-line arguments and best practices. Consult their individual documentation for optimal usage.
• Output Interpretation: While the log action provides output, you may need to parse and interpret it programmatically, especially for complex code generation tasks.
• Resource Intensive: Running AI models, especially large ones, can be resource-intensive. Ensure your environment has sufficient CPU, memory, and potentially GPU resources.