Skills & ClawHub

Extend your agent with 565+ community skills — or create your own in 5 minutes.

Skills are markdown-based extensions that teach OpenClaw how to interact with tools, services, and workflows. They follow the AgentSkills standard and are compatible with Claude Code and Cursor.

How Skills Work

Each skill is a directory containing a SKILL.md file with YAML frontmatter and natural-language instructions:

~/.openclaw/skills/my-skill/
└── SKILL.md

A minimal SKILL.md looks like this:

---
name: my-custom-skill
description: What this skill does
---

Instructions for the agent on how to use this skill...
Use {baseDir} to reference the skill folder path.

The frontmatter provides metadata; everything after the --- fence is injected into the agent's system prompt when the skill is active.

Loading & Precedence

Skills load from three locations. When duplicates exist, the highest-priority source wins:

1. Workspace skills — <workspace>/skills (per-agent overrides, highest priority)
2. Managed / local — ~/.openclaw/skills (shared across all agents)
3. Bundled — shipped with the OpenClaw install (lowest priority)

This layering lets you override a community skill for a specific workspace without touching the global copy.

ClawHub — The Skills Registry

ClawHub (clawhub.com) is the community registry for OpenClaw skills. Browse, install, and publish — all from your terminal.

Install a skill:

clawhub install <skill-slug>

Update all installed skills:

clawhub update --all

Publish your own skill:

clawhub sync --all

Popular Skills

A selection of high-quality community skills to get you started:

Creating a Custom Skill

You can have a working skill in three steps:

Step 1 — Create the directory:

mkdir -p ~/.openclaw/skills/my-skill

Step 2 — Write SKILL.md:

---
name: my-skill
description: A short summary of what this skill does
---

## Instructions

Detailed markdown instructions for the agent.
Reference files relative to {baseDir}.

Step 3 — Reload:

OpenClaw auto-watches skill folders, so your new skill is available immediately. If it doesn't appear, restart the session.

Advanced: Gating & Config

Skills can declare requirements in their YAML metadata so they only activate when prerequisites are met:

---
name: gemini
description: Use Gemini CLI for coding and search
metadata: {"openclaw":{"requires":{"bins":["gemini"],"env":["GEMINI_API_KEY"]}}}
---

Available requirement keys:

• requires.bins — binaries that must exist on PATH
• requires.env — environment variables that must be set
• requires.config — openclaw.json config paths that must be truthy

Per-skill configuration overrides live in ~/.openclaw/openclaw.json under the skills.entries.<name> key.

Security

Skills inject instructions into the agent's system prompt and can invoke tools on your behalf. Treat third-party skills with the same caution as any executable code:

• Read before enabling — review the SKILL.md of any skill you install from ClawHub.
• Use sandboxed runs — for untrusted inputs, run OpenClaw in its sandboxed execution mode.
• API key isolation — keys injected via skills.entries.<name>.apiKey stay in the host process and are never passed into the sandbox.

Token Impact

Every active skill adds to the system prompt. The per-skill overhead is approximately ~97 characters + name length + description length. Keep descriptions concise to minimize token usage, especially if you run many skills simultaneously.