← Back to Skills Marketplace
danielsinewe

OpenClaw Coding Agent Playbook

by Daniel Sinewe · GitHub ↗ · v0.1.1 · MIT-0
cross-platform ✓ Security Clean
333
Downloads
0
Stars
0
Active Installs
2
Versions
Install in OpenClaw
/install openclaw-coding-agent-playbook
Description
Delegate coding tasks to Codex, Claude Code, Pi, or OpenCode from bash with safe launch modes, background monitoring, and repo-isolated review workflows.
README (SKILL.md)

Coding Agent

Use this skill when you need another coding agent to implement or review changes in a repository.

Use this for

  • Building features that need multi-file edits.
  • Large refactors where you want parallel workstreams.
  • PR review or issue triage in an isolated checkout.
  • Repeatable "run/fix/verify" loops that benefit from background execution.

Do not use this for

  • Tiny one-line edits you can apply directly.
  • Read-only code lookup tasks.
  • Work inside ~/.openclaw for external repo reviews.

Execution mode matrix

  • Codex: use pty:true.
  • Pi: use pty:true.
  • OpenCode: use pty:true.
  • Claude Code: use --print --permission-mode bypassPermissions (no PTY required).

Examples:

# Codex (PTY required)
bash pty:true workdir:~/project command:"codex exec --full-auto 'Add request timeout handling to API client and update tests.'"

# Claude Code (print mode, no PTY required)
bash workdir:~/project command:"claude --permission-mode bypassPermissions --print 'Refactor auth middleware and explain risk areas.'"

# OpenCode (PTY required)
bash pty:true workdir:~/project command:"opencode run 'Add retry/backoff for webhook delivery.'"

# Pi (PTY required)
bash pty:true workdir:~/project command:"pi 'Fix failing Vitest suite in src/api and keep behavior unchanged.'"

Background session pattern

# Start
bash pty:true workdir:~/project background:true command:"codex exec --full-auto 'Implement issue #142 and run tests.'"

# Monitor output
process action:log sessionId:XXX

# Check running state
process action:poll sessionId:XXX

# Reply when prompted
process action:submit sessionId:XXX data:"yes"

# Stop session
process action:kill sessionId:XXX

Safe review workflow

Never run PR review in your live OpenClaw repo. Use a temp clone or a worktree.

# Temp clone review
REVIEW_DIR=$(mktemp -d)
git clone https://github.com/org/repo.git "$REVIEW_DIR"
cd "$REVIEW_DIR" && gh pr checkout 130
bash pty:true workdir:"$REVIEW_DIR" command:"codex review --base origin/main"

# Worktree review
git worktree add /tmp/pr-130-review pr-130-branch
bash pty:true workdir:/tmp/pr-130-review command:"codex review --base main"

Prompt template for long tasks

Task:
- [clear scope]
- [constraints]
- [required checks]

When finished:
1) Summarize changed files and why.
2) Report test/lint commands and outcomes.
3) Run:
openclaw system event --text "Done: [brief result]" --mode now

Failure recovery checklist

  • If output stalls, poll once before killing.
  • If agent asks for missing context, answer directly and continue the same session.
  • If the session dies early, relaunch with a tighter prompt and explicit file scope.
  • If multiple retries fail, switch to a fresh worktree to remove local state noise.

Rules

  1. Match the user-requested agent when explicitly specified.
  2. Keep each session scoped to one repository and one objective.
  3. Prefer --full-auto for implementation and conservative flags for review.
  4. Send user updates at start, major milestones, blockers, and completion.
  5. Do not silently take over coding work when orchestration mode was requested.
Usage Guidance
This playbook is internally consistent with its goal of orchestrating other coding agents, but it instructs the agent to clone and execute code and to use aggressive agent flags (e.g., --full-auto, --permission-mode bypassPermissions). Before using it: (1) Only run against repositories in isolated sandboxes (temporary clones, ephemeral containers, or dedicated CI worktrees) and never inside your real OpenClaw home. (2) Inspect and vet the external CLIs (claude, codex, opencode, pi) — know where they come from and what their flags do. (3) Avoid --full-auto or permission-bypass unless you understand the exact consequences; prefer conservative/review modes for code changes. (4) Do not provide secrets or prod credentials to sessions started from this playbook. (5) If you need higher assurance, run the playbook in an environment with network and filesystem restrictions (container/VM) and log the commands the agents run so you can audit changes.
Capability Analysis
Type: OpenClaw Skill Name: openclaw-coding-agent-playbook Version: 0.1.1 The coding-agent skill bundle provides a legitimate framework for orchestrating external AI coding tools (Claude Code, Codex, etc.) within OpenClaw. It includes proactive security guidance, such as using temporary clones and git worktrees to isolate code reviews from the main environment. While it utilizes high-privilege flags like '--permission-mode bypassPermissions' for Claude Code, these are functional requirements for automated delegation and are used transparently within the documented scope of the skill (SKILL.md).
Capability Assessment
Purpose & Capability
The name/description match the SKILL.md instructions. The listed required binaries (claude, codex, opencode, pi) align with the skill's purpose of delegating work to those agents; no unrelated credentials, paths, or installs are requested.
Instruction Scope
The instructions explicitly tell the agent to clone repositories, run agent CLIs with --full-auto or permission-bypass flags, start background sessions, and execute changes in working directories. Those actions are coherent with an orchestration playbook but are inherently high-risk because they involve executing code from external repos and using agent flags that may grant broad permissions. The skill does advise isolation (temp clone/worktree) and not to operate inside ~/.openclaw, which is good practice.
Install Mechanism
There is no install spec and no code files — the skill is instruction-only. That minimizes on-disk installation risk. The playbook relies on external CLIs being present on PATH rather than downloading code itself.
Credentials
The skill requests no environment variables, no credentials, and no config paths. Requiring presence of agent CLIs is proportionate to the stated goals. The SKILL.md does not ask for unrelated secrets.
Persistence & Privilege
Flags show the skill is not forced-always and has no install-time persistence. It is user-invocable and allows normal autonomous invocation (the platform default), which is expected for an orchestration playbook.
How to Use
  1. Make sure OpenClaw is installed (local or Docker)
  2. Run the install command in chat: /install openclaw-coding-agent-playbook
  3. After installation, invoke the skill by name or use /openclaw-coding-agent-playbook
  4. Provide required inputs per the skill's parameter spec and get structured output
Version History
v0.1.1
Rewrite and tighten playbook: clean frontmatter, explicit execution-mode matrix, safer review isolation flow, long-task prompt template, and failure-recovery checklist.
v0.1.0
Initial release: practical orchestration patterns for Codex, Claude Code, Pi, and OpenCode coding workflows.
Metadata
Slug openclaw-coding-agent-playbook
Version 0.1.1
License MIT-0
All-time Installs 0
Active Installs 0
Total Versions 2
Frequently Asked Questions

What is OpenClaw Coding Agent Playbook?

Delegate coding tasks to Codex, Claude Code, Pi, or OpenCode from bash with safe launch modes, background monitoring, and repo-isolated review workflows. It is an AI Agent Skill for Claude Code / OpenClaw, with 333 downloads so far.

How do I install OpenClaw Coding Agent Playbook?

Run "/install openclaw-coding-agent-playbook" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.

Is OpenClaw Coding Agent Playbook free?

Yes, OpenClaw Coding Agent Playbook is completely free, licensed under MIT-0. You can download, install and use it at no cost.

Which platforms does OpenClaw Coding Agent Playbook support?

OpenClaw Coding Agent Playbook is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).

Who created OpenClaw Coding Agent Playbook?

It is built and maintained by Daniel Sinewe (@danielsinewe); the current version is v0.1.1.

💬 Comments