coding-agent

Delegates coding tasks to Codex, Claude Code, or OpenCode as background workers.

1|Updated Jun 9, 2026
One-click install
npx skills add https://github.com/JEROME-PRAKASH-L/openclaw --skill coding-agent-jerome-prakash-l
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: coding-agent
Source: https://github.com/JEROME-PRAKASH-L/openclaw/tree/main/skills/coding-agent
Command: npx skills add https://github.com/JEROME-PRAKASH-L/openclaw --skill coding-agent-jerome-prakash-l

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @anthropic-ai/claude-code, @openai/codex.

What problem does it solve? Long-running coding work like feature builds, large refactors, and issue-to-PR loops blocks interactive sessions and requires manual babysitting. This Skill offloads that work to background coding agents (Codex, Claude Code, or OpenCode) with proper notification routing so results come back to your chat channel automatically. ## Core Features & Use Cases - Background Worker Delegation: Launch Codex, Claude Code, or OpenCode as background processes with PTY and permission settings tuned per agent. - Completion Notifications: Workers report results back through openclaw message send to a captured channel, target, and thread route. - Issue-to-PR Automation: Drive full workflows where a worker branches from a GitHub issue, implements, tests, reviews, and opens a PR. - Use Case: You receive a bug report, create a GitHub issue, and hand it to a Codex worker that implements the fix in an isolated checkout and messages you on Telegram when the PR is ready. ## Quick Start Ask the assistant to delegate a specific coding task, such as implementing a feature or fixing an issue in a given repository, to a background coding agent and notify you 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 Claude Code or Codex as a background coding agent?

Write the task prompt to a temp file, then launch with background:true. Codex and OpenCode need pty:true, while Claude Code runs without PTY using claude --permission-mode bypassPermissions --print, reading the prompt file via stdin redirect.

Codex vs Claude Code vs OpenCode for background coding tasks?

All three are supported as workers; if the user names a specific agent, that one is used. Codex requires a trusted git repo (even a scratch one), Claude Code runs with bypassPermissions print mode, and OpenCode uses opencode run with PTY.

When should I not use a background coding agent?

Do not use it for simple edits, read-only code lookup, ACP thread-bound work, or anything inside ~/.openclaw, $OPENCLAW_STATE_DIR, or active OpenClaw state directories. Never run background agents in the main ~/Projects/openclaw checkout; use an isolated checkout instead.

How do background coding workers notify me when finished?

Each worker prompt includes a notification block with channel, target, and optional account, reply_to, and thread_id. The worker sends exactly one completion or failure message using openclaw message send; heartbeat, system events, and notify-on-exit are not used.

What do I do if a background coding worker hangs or fails?

Monitor the session with process actions like poll, log, and list. If the worker fails or hangs, respawn it or ask the user; do not silently hand-code the task yourself, and do not kill slow workers without cause.