A sophisticated integration for delegating software engineering tasks to autonomous AI agents through managed background processes and interactive terminals.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install coding-agent-backup
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 coding-agent-backup 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 allows developers to harness the power of autonomous AI engineers like Codex, Claude Code, and Pi directly within their development environment. By leveraging Openclaw Skills, this tool provides a structured way to spawn agents that can explore files, write code, and execute shell commands to solve complex tickets. It is designed to handle everything from feature creation to deep codebase refactors, ensuring that agents operate in isolated working directories to maintain project integrity.
Central to this skill is the support for pseudo-terminals (PTY), which is critical for interactive CLI agents that require a real TTY for colors and input handling. Whether you are running a quick one-shot command or a long-running background session, the Coding Agent manages the lifecycle of these external processes, providing tools for monitoring logs and interacting with the agents as they work.
Coding Agent Use Cases
- Delegating the creation of new features or entire applications to autonomous agents.
- Performing automated code reviews on PRs by spawning agents in temporary directories.
- Refactoring large-scale codebases through non-blocking background processes.
- Parallelizing bug fixes across multiple issues using git worktrees and simultaneous agent sessions.
- Conducting iterative coding tasks that require autonomous file exploration and testing.
How Coding Agent Works
- The user triggers the coding agent using the bash tool, ensuring the pty:true parameter is set for interactive compatibility.
- The skill identifies the target directory (workdir) to focus the agent's context and prevent unauthorized file access.
- For complex tasks, the agent is launched in background mode, returning a unique sessionId for tracking.
- Developers monitor progress using process:log and process:poll actions to see real-time output from the agent.
- If the agent requires clarification, the user can send input via the process:write or process:submit actions.
- Upon completion, the agent can optionally trigger a system event to notify the user that the task is finished.
Coding Agent Setup
To use the Coding Agent, ensure you have one of the supported binaries installed (claude, codex, opencode, or pi). You must use a bash tool that supports pseudo-terminals. To start an interactive session, use the following command structure:
bash pty:true workdir:~/your-project command:"codex exec 'Implement a dark mode toggle'"
For long-running tasks, enable background mode:
bash pty:true workdir:~/your-project background:true command:"claude 'Refactor the database schema'"
Coding Agent Data Schema & Taxonomy
The skill manages agent interactions through process sessions and directory scoping. Below is the primary data structure used for session management:
| Attribute | Description |
|---|---|
| command | The specific agent CLI command being executed (e.g., codex exec). |
| pty | Boolean; must be true for agents to render terminal UI correctly. |
| workdir | The filesystem path where the agent is restricted to work. |
| sessionId | A unique identifier for background processes used to fetch logs or kill tasks. |
| background | Boolean; determines if the agent blocks the terminal or runs asynchronously. |
Coding Agent Advanced Features
- Support for multiple AI backends including gpt-5.2-codex and Anthropic-powered Claude Code.
- Background execution mode for long-running engineering tasks with session management.
- Git worktree integration for massive parallelization of issue resolution.
- Pseudo-terminal (PTY) allocation to support high-fidelity interactive CLI experiences.
- Automated completion notifications via system events to reduce idle time between tasks.
- Specialized flags for sandboxed execution including --full-auto and --yolo modes.
SKILL.md
Loading
Related Openclaw Skills
Clawdbot Documentation Expert (clawddocs)
A specialized AI agent skill designed to help users navigate, search, and configure the entire Clawdbot ecosystem through structured documentation access.
Expense Tracker
A privacy-focused AI financial assistant that logs, categorizes, and tracks your spending using natural language processing.
Market Snapshot
A high-performance utility for fetching stable JSON market snapshots including price, liquidity, and volume for any token on the Solana network.
claw-relay
A networking skill that routes AI agent traffic through residential IPs using Tailscale exit nodes without custom proxy software.
Get Focus Mode
A lightweight macOS utility to programmatically retrieve the current Focus mode status for context-aware automation.
Paperless-NGX CLI for AI Agents
A comprehensive CLI integration for Paperless-NGX that allows AI agents to search, retrieve, and organize digital documents using the ppls tool.
Featured*
