coding-agent

Delegate coding tasks to Codex, Claude Code, OpenCode, or Pi agents as background processes.

Updated Aug 17, 2026
One-click install
npx skills add https://github.com/prabaljainn/my-claude-code-setup --skill coding-agent-prabaljainn
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: coding-agent
Source: https://github.com/prabaljainn/my-claude-code-setup/tree/main/claude/skills/coding-agent
Command: npx skills add https://github.com/prabaljainn/my-claude-code-setup --skill coding-agent-prabaljainn

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @anthropic-ai/claude-code, @openai/codex, @mariozechner/pi-coding-agent.

What problem does it solve? Running coding agents in the foreground blocks your session and leaves you without completion notifications. This Skill orchestrates Codex, Claude Code, OpenCode, and Pi coding agents as background processes with proper PTY configuration, process monitoring, and direct completion messaging so long-running builds, refactors, and PR reviews never stall your workflow. ## Core Features & Use Cases - Background Agent Orchestration: Launch any supported coding CLI with background:true immediately, then monitor progress with process log and poll actions. - Per-Agent Execution Modes: Uses pty:true for Codex, Pi, and OpenCode, and --print --permission-mode bypassPermissions for Claude Code, matching each CLI's requirements. - Completion Notification Routing: Injects a notify route into every worker prompt so the agent sends a direct openclaw message send on completion or failure instead of relying on heartbeat. - Use Case: You need to review three open PRs and fix two issues in parallel. The Skill clones each PR into a temp directory or git worktree, spawns a background Codex session per task, and each worker messages you directly when done. ## Quick Start Ask the agent to build a feature or review a PR with Codex in the background, for example: "Use Codex to build a dark mode toggle in my project and message me when it finishes."

Frequently Asked Questions about coding-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run Codex CLI in the background for coding tasks?▼

Launch Codex with bash using pty:true and background:true, for example codex exec --full-auto with your prompt. Codex requires a PTY and a trusted git directory, so initialize a git repo first for scratch work.

How do I review a pull request with a coding agent?▼

Clone the repo to a temp directory or create a git worktree, check out the PR branch with gh pr checkout, then run codex review --base origin/main in the background. Never review PRs inside the live OpenClaw project folder.

What is the difference between running Claude Code and Codex as background agents?▼

Claude Code uses --print --permission-mode bypassPermissions without a PTY, while Codex, Pi, and OpenCode require pty:true. Using the wrong mode causes the agent to fail or hang.

How do I get notified when a background coding agent finishes?▼

Inject a notification route into the worker prompt and require it to call openclaw message send with the channel and target on completion or failure. Do not rely on heartbeat, system events, or notifyOnExit.

Can I run multiple coding agents in parallel?▼

Yes, you can spawn many background sessions at once, for example one git worktree per issue with a separate Codex process each. Monitor all of them with process action:list and process action:log.

When should I not use a background coding agent?▼

Skip it for simple one-line fixes you can edit directly, for reading code where the read tool suffices, and for any work inside the OpenClaw state directory or live OpenClaw project folder.