What does this skill do?

The Coding Agent skill lets an agent delegate substantial development tasks to AI coding assistants running as background processes — not trivial edits or read-only lookups.

It is built for feature development, pull-request reviews, large refactors, and issue-to-PR automation loops, with strict rules for background execution, notifications, and Git preparation.

It classifies source references as trusted or untrusted, spawns isolated worktrees for repository work, and requires workers to report completion through a defined channel.

  • Delegates coding work to background workers

  • Handles feature builds, PR reviews, and large refactors

  • Prepares Git worktrees and trust-classifies sources

  • Monitors workers and reports status on milestones

How it works

1

Choose the worker

The agent selects Codex, Claude Code, or OpenCode and respects any user-specified preference, initializing in the background.

2

Prepare Git

For repository work it establishes the remote and default branch, classifies the source ref's trust, and spawns an isolated worktree.

3

Launch and notify

It launches the worker from a prompt file with a notification block, so the worker reports completion through the defined channel.

4

Monitor and report

It monitors the process, respawns or asks for guidance on failure, and reports status on milestones and completion.

Best use cases for this skill

  • Feature builds: Delegate a full feature to a background coding worker.

  • PR reviews: Hand off a pull-request review to a coding assistant.

  • Large refactors: Run a comprehensive refactor as a background task.

  • Issue-to-PR: Automate an issue-to-PR loop end to end.

Frequently asked questions

What does the Coding Agent skill do?

It delegates substantial coding work — feature development, PR reviews, large refactors, and issue-to-PR workflows — to Codex, Claude Code, or OpenCode running as background workers.

When should an agent use it?

Use it for background feature builds, PR reviews, comprehensive refactors, and issue-to-PR automation — not for simple edits or read-only code inspection.

How does it handle repository safety?

It classifies source references as trusted or untrusted, spawns isolated worktrees from canonical defaults, and avoids auto-rebasing existing branches.

How do workers report back?

Each worker prompt includes a notification block, and workers must report completion through the defined message channel rather than relying on heartbeats.

Use from source

Ask your Buda agent to install this community skill.

Copy this prompt into your own Buda agent. It includes the source repository and install command so the agent can complete the setup in natural language.

Prompt to send to your Buda agent
Please install the "Coding Agent Skill" skill into this Buda agent.

Source repository: https://github.com/openclaw/openclaw
Install command: npx skills add openclaw/openclaw --skill coding-agent -y -a universal

Please run the install command, verify the skill is available, and tell me any setup notes or follow-up steps from the source repository.
Explore marketplace

Install workflow

  1. 1Copy the prompt
  2. 2Send it to your Buda agent
  3. 3Done