coding-agent

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

Updated Jun 19, 2026
One-click install
npx skills add https://github.com/AmirulAndalib/Vilvona-AI --skill coding-agent-amirulandalib
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: coding-agent
Source: https://github.com/AmirulAndalib/Vilvona-AI/tree/main/skills/coding-agent
Command: npx skills add https://github.com/AmirulAndalib/Vilvona-AI --skill coding-agent-amirulandalib

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 lacks reliable completion tracking. This Skill offloads that work to background coding agents with a defined notification route so results come back without manual polling. ## 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: Every worker prompt includes a notification block so the worker reports success or failure via openclaw message send instead of relying on heartbeats or exit events. - Issue-to-PR Automation: Drive full workflows from a GitHub issue through branching, implementation, testing, review, and PR creation, returning the issue URL and sessionId immediately. - Use Case: You receive a bug report, create a GitHub issue describing the fix, and hand it to a Codex worker that implements, tests, and opens a PR while you continue other work, then get a message when it finishes. ## Quick Start Ask the agent to delegate a specific coding task to a background worker, for example: run a background Codex worker in this repo to implement the feature described in issue 42 and notify me when the PR is open.

Frequently Asked Questions about coding-agent

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

FAQPage Schema
How do I run a coding agent in the background?

Write the worker prompt to a temp file, then launch with background:true and the agent-specific command, such as codex exec with pty:true or claude --permission-mode bypassPermissions --print. Include a notification block so the worker reports completion via openclaw message send.

What is the difference between Codex, Claude Code, and OpenCode workers?

Codex and OpenCode require pty:true when launched, while Claude Code runs without a PTY using --permission-mode bypassPermissions --print. Codex also needs a trusted git repository, so scratch work requires initializing a temp git repo first.

How do I automate an issue-to-PR workflow with a coding agent?

Create or reuse a GitHub issue as the durable spec, then give the worker the issue URL, repo, base branch, expected PR, and notification route. The worker branches, implements, tests, reviews until no actionable findings remain, and opens the PR.

When should I not use a background coding agent?

Avoid it for simple edits, read-only code lookup, ACP thread-bound work, or anything inside OpenClaw state directories. Never run background agents in the main OpenClaw checkout; use an isolated checkout instead.

Why is my coding worker not sending a completion notification?

The worker needs a real notification route captured before spawning, with channel, target, and optional account or thread values in its prompt. Heartbeats, system events, and notify-on-exit are not reliable, so verify the openclaw message send command in the prompt.