菜单
Use when the user wants to install OpenClaw, set up OpenClaw, onboard OpenClaw, get started with OpenClaw, run the OpenClaw installer, configure the OpenClaw daemon, or deploy OpenClaw on macOS, Linux, or WSL2.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Audit one or more OpenClaw agents, analyse what they require (complexity, tool use, context window, volume, autonomy, sensitivity), and recommend the best-fit OpenRouter models by pricing tier with ready-to-paste openclaw.json config snippets. Use when the user says "what model should I use", "recommend a model for my agent", "audit my agent models", "find a cheaper model", "openrouter model suggestions", "which model fits my agent", or invokes /openclaw-model-advisor.
Create, optimize, and maintain OpenClaw agent workspace files — SOUL.md, IDENTITY.md, AGENTS.md, MEMORY.md, and GOALS.md. Use this skill whenever the user mentions OpenClaw agent setup, agent personality, agent identity, agent workspace files, SOUL.md, IDENTITY.md, AGENTS.md, MEMORY.md, GOALS.md, agent persona, "set up my agent", "configure my agent", "make my agent", or wants to scan/optimize/audit an existing OpenClaw workspace. Also trigger when the user wants to create a new agent from scratch, refine an agent's personality, add goals or memory structure, or review workspace files for consistency. This skill conducts targeted interviews, generates production-ready markdown files, and can scan existing workspaces to optimize them.
Diagnose and fix OpenClaw agent and gateway problems — heartbeats not firing, cron jobs failing, agents not responding, wrong routing, model errors, workspace issues, channel connection failures, and openclaw.json config errors. Use this skill when the user reports: agent not responding, heartbeat stopped, cron job not running, wrong agent getting messages, model errors, gateway won't start, channel disconnected, agent sending to wrong place, memory not writing, session errors, openclaw doctor errors, or any "something isn't working" situation with OpenClaw. This skill reads the openclaw.json config, interprets error patterns, walks through diagnostic steps, identifies the root cause, and produces the exact commands or file changes needed to fix it.
Brainstorm, design, and document OpenClaw agents and multi-agent teams. Use this skill whenever the user wants to plan what agents to build, figure out what agents they need, design agent capabilities and permissions, plan a multi-agent architecture, decide how agents should work together, or create documentation for their agent team. Also trigger when the user says things like "what agents do I need", "help me plan my agents", "I want to set up multiple agents", "document my agent", "agent permissions", "agent roles", "agent team", or asks about agent capabilities, responsibilities, tool access, or inter-agent communication. This skill produces a comprehensive Agent Team Documentation package (markdown + optional .docx) covering every agent's role, persona, permissions, tools, boundaries, and team interaction rules.
Surgically update existing OpenClaw agent workspace files based on conversational change requests — without rewriting the soul, persona, or hard limits. Use this skill when the user says things like "update the agent", "tell NEXUS about the new client", "HUNTER's ICP has changed", "add this to APEX's memory", "RPS paid their invoice", "change SENTINEL's heartbeat", "update the goals", "the client list changed", "add a new cron job", "remove the stale goal", "SCRIBE needs to know about ISO 14001 scope change", or any request to change, add, remove, or refresh content in an existing agent's workspace files. Also trigger when business state changes (new client, lost client, invoice paid, tender outcome, staff change, new service, pricing update) that should be reflected in agent memory. This skill identifies which files and sections need changing, shows exactly what will change before touching anything, makes surgical edits that preserve the persona soul, and outputs ready-to-apply file patches.
Run the complete IBS OpenClaw client onboarding process — from scoping a new client's agent requirements, through designing their agent architecture, building their workspace files and openclaw.json, to producing a professional handover package. Use this skill when IBS is onboarding a new client onto OpenClaw, productising the agent deployment service, creating a client's first agent setup, or when the user says things like "onboard this client", "set up OpenClaw for a new client", "create an agent for our client", "deploy agents for [company]", or "build the OpenClaw setup for [client name]". This skill runs the full IBS OpenClaw product workflow: client interview, needs assessment, architecture design, workspace generation, openclaw.json configuration, deploy script, and professional handover documentation.
| name | openclaw-setup |
| description | Use when the user wants to install OpenClaw, set up OpenClaw, onboard OpenClaw, get started with OpenClaw, run the OpenClaw installer, configure the OpenClaw daemon, or deploy OpenClaw on macOS, Linux, or WSL2. |
Guide the user through installing and configuring OpenClaw, a self-hosted gateway connecting messaging apps to AI coding agents.
node --version.Detects Node.js, installs if missing, installs the CLI globally via npm, and launches the onboarding wizard.
macOS / Linux / WSL2:
curl -fsSL https://openclaw.ai/install.sh | bash
Windows (PowerShell):
iwr -useb https://openclaw.ai/install.ps1 | iex
Skip onboarding (install CLI only):
curl -fsSL https://openclaw.ai/install.sh | bash -s -- --no-onboard
npm install -g openclaw@latest
openclaw onboard --install-daemon
If sharp build errors occur:
SHARP_IGNORE_GLOBAL_LIBVIPS=1 npm install -g openclaw@latest
pnpm add -g openclaw@latest
pnpm approve-builds -g
openclaw onboard --install-daemon
git clone https://github.com/openclaw/openclaw.git
cd openclaw
pnpm install && pnpm ui:build && pnpm build
pnpm link --global
openclaw onboard --install-daemon
Run openclaw onboard --install-daemon to start the interactive wizard. It configures:
~/.openclaw/workspace).Use openclaw configure to reconfigure later.
Installed automatically by openclaw onboard --install-daemon. Auto-starts on login.
Installed as systemd user service. Enable lingering to survive logout:
sudo loginctl enable-linger $USER
After install, verify with:
openclaw status # Quick health check
openclaw gateway status # Service status
openclaw doctor # Config validation
openclaw dashboard # Open browser UI
openclaw agents add <name>
Creates a separate agent with its own workspace, sessions, and auth profiles.
export PATH="$(npm prefix -g)/bin:$PATH"SHARP_IGNORE_GLOBAL_LIBVIPS=1 prefixpnpm approve-builds -g after installsudo loginctl enable-linger $USER| Path | Purpose |
|---|---|
~/.openclaw/openclaw.json | Main config file (JSON5) |
~/.openclaw/workspace/ | Agent workspace (skills, prompts) |
~/.openclaw/credentials/ | Channel auth data |
~/.openclaw/agents/ | Per-agent state and sessions |
See references/installation-guide.md for detailed per-platform steps.