Route coding-agent requests into the right ACP or acpx session flow.
This skill appears to be prompt/routing guidance rather than executable software, with no required secrets and no declared remote endpoints, so the overall risk is low. Its main function is to route requests into ACP/runtime or acpx flows; the open-source repository and very high community adoption materially reduce supply-chain risk.
The materials explicitly state that no keys or environment variables are required. There is no indication of collecting, storing, or forwarding API keys, tokens, or other sensitive credentials, so credential exposure risk is low.
Both the metadata and objective checks indicate no remote endpoint hosts. While the README discusses routing requests to external coding harness/ACP flows, it does not declare that this skill itself connects to any specific remote service or sends user data to unknown hosts.
This object is classified as prompt-only, and the provided content is README/routing logic rather than executable code. Although it references driving other runtimes via `sessions_spawn` or `exec`/`acpx`, that is instructional guidance and does not mean the skill itself directly executes code on the host.
The materials do not declare any permissions to read or write local files, databases, system resources, or the user workspace, nor do they request extra authorization. As a prompt-style skill, any data access would depend on downstream host tooling rather than direct access by the skill itself.
The source is an open-source GitHub repository with extremely high community adoption (about 377k stars), which are strong risk-reducing signals. The unspecified license and unknown maintenance status are minor information gaps, but without other red flags they do not justify a higher-risk rating.
Copy the install command and let the AI configure it · recommended for beginners
Please install the "acp-router" skill from askskill: 1. Download https://raw.githubusercontent.com/openclaw/openclaw/main/extensions/acpx/skills/acp-router/SKILL.md 2. Save it as ~/.claude/skills/acp-router/SKILL.md 3. Reload skills and tell me it's ready
Route this request to Claude Code: refactor my Node.js API and add unit tests for the new routes.
The request is recognized as a coding-agent task and routed into the appropriate Claude Code ACP or acpx execution flow.
Use ACP to spawn a background coding thread, analyze why this repository's CI is failing, and suggest fix commits.
A background session thread is created per the explicit ACP request and routed into the appropriate ACP runtime workflow.
Send the following development request to the most suitable agent: inspect dependency conflicts in a Python project and generate repair steps.
The skill selects the most suitable coding-agent entry point and routes the task into the corresponding session mode.
When user intent is "run this in Claude Code/Cursor/Copilot/OpenClaw/OpenCode/Gemini/Qwen/Kiro/Kimi/iFlow/Droid/Kilocode (ACP harness)", do not use subagent runtime or PTY scraping. Route through ACP-aware flows.
Codex is special: plain chat/conversation binding and control should use the native Codex app-server plugin (/codex bind, /codex threads, /codex resume) instead of the default ACP path. Use ACP for Codex only when the user explicitly names ACP//acp/acpx, or when spawning background child sessions through sessions_spawn where a native Codex runtime spawn is not available yet.
Trigger this skill when the user asks OpenClaw to:
/acp, or acpxMandatory preflight for coding-agent thread requests:
OpenClaw ACP runtime path below; do not use message(action="thread-create") for ACP harness thread spawn.Choose one of these paths:
sessions_spawn / ACP runtime tools.acpx path (telephone game): use acpx CLI through exec to drive the harness session directly.Use direct acpx when one of these is true:
acpx drivingDo not use:
subagents runtime for harness control/acp command delegation as a requirement for the useracpx is availableUse these defaults when user names a harness directly:
agentId: "openclaw"agentId: "claude"agentId: "codex" only for explicit ACP/acpx requests or background ACP runtime spawnagentId: "copilot"agentId: "cursor"agentId: "droid"agentId: "opencode"agentId: "gemini"agentId: "iflow"agentId: "kilocode"agentId: "kimi"agentId: "kiro"agentId: "qwen"These defaults match current acpx built-in aliases.
If policy rejects the chosen id, report the policy error clearly and ask for the allowed ACP agent id.
Required behavior:
sessions_spawn with:
runtime: "acp"thread: truemode: "session" (unless user explicitly wants one-shot)message with action=thread-create; sessions_spawn is the only thread-create path.task so the ACP session gets it immediately.agentId explicitly unless ACP default agent is known.Example:
User: "spawn a test codex ACP session in thread and tell it to say hi"
Call:
{
"task": "Say hi.",
"runtime": "acp",
"agentId": "codex",
"thread": true,
"mode": "session"
}
When the user asks to start a coding harness in a thread, treat that as an ACP runtime request and try to satisfy it end-to-end.
Required behavior when ACP backend is unavailable:
…
List chats, review message history, and send iMessage or SMS from CLI.
Summarize or transcribe URLs, videos, PDFs, and local files quickly.
Create, search, and manage Bear notes on macOS via grizzly CLI.
Run or recover OpenClaw macOS signing, notarization, and release promotion.
Automatically add redacted agent transcripts to GitHub PRs or issues.
Audit and harden OpenClaw hosts for security and operational health.
Run headless agent communication, sessions, and multi-agent workflow orchestration.
Configure skills, tools, and AI coding agents from one capabilities.yaml file.
Let AI coding tools safely call Codex for development and automation tasks.
Connect multiple AI coding agents to collaborate on development tasks efficiently.
Control multiple logged-in AI web sessions via MCP with file and image handling.
Manage skills across multiple coding agents on macOS in one place.