coding-agent
Delegate coding tasks to Codex, Claude Code, or Pi agents via acpx (ACP protocol). Use when: (1) building/creating new features or apps, (2) reviewing PRs (spawn in temp dir), (3) refactoring large codebases, (4) iterative coding that needs file exploration. NOT for: simple one-liner fixes (just edit), reading code (use read tool), thread-bound ACP harness requests in chat (for example spawn/run Codex or Claude Code in a Discord thread; use sessionsspawn with runtime:"acp"), or any work in /clawd workspace (never spawn agents here). Prefer acpx over raw PTY/exec for all coding agent delegation.
- Installs
- 103
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://clawhub.ai --skill coding-agent- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads coding-agent automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
Coding Agent (acpx-first)
Use acpx for all coding agent work. It communicates over the Agent Client Protocol (ACP) — no PTY scraping, structured output, persistent sessions.
# Install (already global on this system)
npm i -g acpx
✅ Primary: acpx (preferred for all agents)
# One-shot (exec) — no session state saved
acpx codex exec "summarize this repo"
acpx claude exec "refactor the auth module"
# Persistent session (prompt) — reuses session per cwd
acpx codex "fix the failing tests"
acpx claude "add error handling to API calls"
# Named parallel sessions
acpx codex -s backend "refactor the API layer"
acpx codex -s frontend "update the React components"
# Auto-approve permissions
acpx --approve-all codex exec "run and fix failing tests"
# Fire-and-forget (background)
acpx --no-wait codex "long refactor task"
# Read prompt from file or stdin
acpx codex -f prompt.txt
cat prompt.md | acpx codex
# With working directory
acpx --cwd ~/Projects/myproject codex "fix the auth bug"
Built-in agent registry
| Alias | Resolves to |
|---|---|
codex | npx @zed-industries/codex-acp |
claude | npx -y @zed-industries/claude-agent-acp |
gemini | gemini --acp |
opencode | npx -y opencode-ai acp |
pi | npx pi-acp |
kimi | kimi acp |
Default agent is codex.
Session management
acpx sessions # list sessions
acpx sessions new # create new session for current cwd
acpx sessions new --name backend # named session
acpx sessions close backend # close named session
acpx sessions history --limit 20 # turn history
acpx codex status # check if agent process is alive
acpx codex cancel # cancel in-flight turn
⚠️ Fallback: raw exec (only when acpx unavailable or unsupported)
For Claude Code (claude CLI), if acpx isn't working:
# ✅ Fallback for Claude Code (no PTY needed)
cd /path/to/project && claude --permission-mode bypassPermissions --print 'Your task'
# ❌ Wrong for Claude Code
bash pty:true command:"claude --dangerously-skip-permissions 'task'"
For Codex/Pi/OpenCode raw fallback (PTY required):
bash pty:true command:"codex exec 'Your prompt'"
MCP Servers — Available in All Claude Code Sessions
Both MCP servers are pre-configured system-wide and connect automatically when running
claude from any workspace directory.
jcodemunch-mcp — Code Exploration
When to use:
- Exploring an unfamiliar codebase before making changes
- Looking up function/class/symbol definitions
- Getting a file outline without reading the whole file
- Tracing call chains or finding symbol usage
- Any code task where you'd otherwise open and read multiple files
Workflow:
# 1. Index the project once (idempotent — safe to re-run)
# In Claude Code prompt: "Use jcodemunch to index this project first"
# Tools available inside Claude Code:
jcodemunch: index_codebase { "path": "/path/to/project" }
jcodemunch: list_projects {}
jcodemunch: get_file_outline { "path": "src/server.js" }
jcodemunch: search_symbols { "query": "handleAuth" }
jcodemunch: get_symbol { "symbol_id": "..." }
Prompt pattern for Claude Code tasks:
cd ~/project && claude --permission-mode bypassPermissions --print \
'First index this project with jcodemunch, then use symbol search and outlines
to explore the codebase. Do NOT read full files — use targeted retrieval.
Task: <your task here>'
Rules:
- Always index before exploring a new project
- Use
get_file_outline+search_symbolsbefore reaching forread - Re-index after major refactors; skip for routine work
list_projectsto avoid re-indexing already-indexed repos
jdocmunch-mcp — Documentation Exploration
When to use:
- Exploring library/framework/API docs before implementing
- Finding config options, installation steps, API reference sections
- Onboarding to an unfamiliar tool
- Any task where you'd otherwise open and skim large README or doc files
Workflow:
# Tools available inside Claude Code:
jdocmunch: list_repos {} # check indexed
jdocmunch: index_local { "path": "/path/to/docs" } # index local docs
jdocmunch: index_repo { "url": "owner/repo" } # index GitHub repo
jdocmunch: get_toc_tree { "repo": "owner/repo" } # browse structure
jdocmunch: search_sections { "repo": "...", "query": "auth" } # find sections
jdocmunch: get_section { "repo": "...", "section_id": "..." } # read one section
jdocmunch: get_section_context { "repo": "...", "section_id": "..." } # section + context
Pre-indexed docs (already available — no need to re-index):
- OpenClaw docs: index with
index_local { "path": "/home/sean/.npm-global/lib/node_modules/openclaw/docs" }
Prompt pattern:
claude --permission-mode bypassPermissions --print \
'Use jdocmunch to index and explore the docs at <path/repo>.
Use search_sections and get_section — do NOT read full doc files.
Task: <your task here>'
Rules:
- Always
list_reposbefore indexing — avoid duplicate indexes - Pattern:
search_sections→get_section(never open full files) - Use
get_section_contextwhen you need surrounding structure GITHUB_TOKEN,GOOGLE_API_KEY,OPENAI_API_KEYare pre-configured for summaries + private repos
Installed CLIs
| CLI | Version | Auth | Notes |
|---|---|---|---|
claude | 2.1.71 | OAuth | --print --permission-mode bypassPermissions |
codex | 0.116.0 | OpenAI key (via ~/.bashrc) | pty:true required |
gemini | 0.29.5 | Gemini API key | pty:true recommended |
Bash Tool Parameters
| Parameter | Type | Description |
|---|---|---|
command | string | The shell command to run |
pty | boolean | Use for coding agents! Allocates a pseudo-terminal for interactive CLIs |
workdir | string | Working directory (agent sees only this folder's context) |
background | boolean | Run in background, returns sessionId for monitoring |
timeout | number | Timeout in seconds (kills process on expiry) |
elevated | boolean | Run on host instead of sandbox (if allowed) |
Process Tool Actions (for background sessions)
| Action | Description |
|---|---|
list | List all running/recent sessions |
poll | Check if session is still running |
log | Get session output (with optional offset/limit) |
write | Send raw data to stdin |
submit | Send data + newline (like typing and pressing Enter) |
send-keys | Send key tokens or hex bytes |
paste | Paste text (with optional bracketed mode) |
kill | Terminate the session |
Quick Start: One-Shot Tasks
# One-shot with acpx (preferred)
acpx --cwd ~/project codex exec "Your task"
acpx --cwd ~/project claude exec "Your task"
# With auto-approve
acpx --approve-all --cwd ~/project codex exec "Add error handling to the API calls"
The Pattern: persistent sessions + background
For longer tasks, use persistent sessions or --no-wait:
# Create a named session and start work
cd ~/project
acpx codex sessions new --name myfeature
acpx codex -s myfeature "Build a snake game"
# Check status
acpx codex status -s myfeature
acpx codex sessions history myfeature --limit 10
# Fire-and-forget (returns immediately)
acpx --no-wait --cwd ~/project codex "Long refactor task"
Reviewing PRs
⚠️ CRITICAL: Never review PRs in OpenClaw's own project folder!
# Clone to temp for safe review
REVIEW_DIR=$(mktemp -d)
git clone https://github.com/user/repo.git $REVIEW_DIR
acpx --approve-reads --cwd $REVIEW_DIR codex exec "Review this PR vs main branch and summarize issues"
Parallel Sessions with git worktrees
git worktree add -b fix/issue-78 /tmp/issue-78 main
git worktree add -b fix/issue-99 /tmp/issue-99 main
acpx --approve-all --no-wait --cwd /tmp/issue-78 codex "Fix issue #78"
acpx --approve-all --no-wait --cwd /tmp/issue-99 codex "Fix issue #99"
acpx codex sessions list
Auto-Notify on Completion
Append to your prompt so OpenClaw gets notified when the agent finishes:
... your task here.
When completely finished, run:
openclaw system event --text "Done: [brief summary]" --mode now
⚠️ Rules
- Use acpx first: Always prefer
acpxover raw PTY/exec for coding agent delegation. Fall back to raw exec only if acpx is unavailable or the agent doesn't support ACP. - Use MCP servers: Always prompt Claude Code to use jcodemunch for code exploration and jdocmunch for doc exploration — never brute-force read files
- Respect tool choice — if user asks for Codex, use Codex
- Be patient — don't kill sessions because they're "slow"
- Monitor with process:log — check progress without interfering
- Parallel is OK — run many agents at once for batch work
- NEVER start Codex in ~/.openclaw/ — it'll read soul docs and get weird ideas about the org chart!
- NEVER checkout branches in ~/Projects/openclaw/ — that's the LIVE OpenClaw instance!
Progress Updates
- Send 1 short message when you start (what's running + where)
- Update again only when: milestone completes, agent asks a question, error occurs, or agent finishes
- If you kill a session, immediately say why
FAQ
What does the coding-agent skill do?
Delegate coding tasks to Codex, Claude Code, or Pi agents via acpx (ACP protocol). Use when: (1) building/creating new features or apps, (2) reviewing PRs (spawn in temp dir), (3) refactoring large codebases, (4) iterative coding that needs file exploration. NOT for: simple one-liner fixes (just edit), reading code (use read tool), thread-bound ACP harness requests in chat (for example spawn/run Codex or Claude Code in a Discord thread; use sessionsspawn with runtime:"acp"), or any work in /clawd workspace (never spawn agents here). Prefer acpx over raw PTY/exec for all coding agent delegation.
How do I install coding-agent?
Run npx skills add https://clawhub.ai --skill coding-agent in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/coding-agent/ (personal) or .claude/skills/coding-agent/ (project).
Which AI agents can use coding-agent?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is coding-agent safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains coding-agent?
It is published by clawhub.ai in clawhub.ai. It has 103 recorded installs.
More from clawhub.ai
deep-skill-finder
clawhub.ai
最好的find skill的方式,能够基于你的任务,从实测效果和用户评价等真实使用反馈出发,寻找最匹配的高质量Skill,而不是只看下载量等热门数据。以下两种情况下都应使用本技能:① 当用户需要查找、搜索、推荐或安装 Skill,或者需要借助他人经验时——当用户说"找个 xxx 技能""股票分析别人怎么做的""找一找有没有现成的技能"等表达寻找意图,或直接表达想 find a skill / find skill / skill search的需求时;② Agent 自主判断需要外部 Skill 辅助——遇到不熟悉的任务,或对当前任务已经做过一些尝试仍无法解决、缺少合适工具时,可主动调用本技能查询实战经验并检索匹配的 Skill,无需等用户开口。
self-improving-agent
clawhub.ai
Captures and maintains learnings, errors, and corrections. Use when a command fails, the user corrects an assumption, a capability is missing, knowledge is outdated, or a better approach is found. Review relevant saved guidance before major tasks, reuse, and promotion; revalidate it according to what could change and how quickly.
humanizer
clawhub.ai
Remove signs of AI-generated writing from text. Use when editing or reviewing text to make it sound more natural and human-written. Based on Wikipedia's comprehensive "Signs of AI writing" guide. Detects and fixes patterns including: inflated symbolism, promotional language, superficial -ing analyses, vague attributions, em dash overuse, rule of three, AI vocabulary words, negative parallelisms, and excessive conjunctive phrases.
github
clawhub.ai
Interact with GitHub using the gh CLI. Use gh issue, gh pr, gh run, and gh api for issues, PRs, CI runs, and advanced queries.
summarize
clawhub.ai
Summarize URLs or files with the summarize CLI (web, PDFs, images, audio, YouTube).
Popular code quality & workflow skills
grill-me
mattpocock/skills
A relentless interview to sharpen a plan or design.
improve-codebase-architecture
mattpocock/skills
Scan a codebase for deepening opportunities, present them as a visual HTML report, then grill through whichever one you pick.
handoff
mattpocock/skills
Compact the current conversation into a handoff document for another agent to pick up.
domain-modeling
mattpocock/skills
Build and sharpen a project's domain model. Use when discussing codebase terminology, writing or editing a GLOSSARY.md, or recording or editing an ADR.
codebase-design
mattpocock/skills
Shared vocabulary for designing deep modules. Use when the user wants to design or improve a module's interface, find deepening opportunities, decide where a seam goes, make code more testable or AI-navigable, or when another skill needs the deep-module vocabulary.
diagnosing-bugs
mattpocock/skills
Diagnosis loop for hard bugs and performance regressions. Use when the user says "diagnose"/"debug this", or reports something broken/throwing/failing/slow.