Skills & Plugins
Extend OpenClaw with Skills and create your own
Extend OpenClaw with Skills and create your own
OpenClaw skills are plugins that add tools, knowledge, and workflows. To pick which skills to try, browse the full list of available skills from ClawHub (calendar, Slack, GitHub, smart home, and hundreds more). Install the ones you want with one command; use openclaw skills list to see what's already installed on your machine. For Model Context Protocol tool servers (a different wiring path from packaged skills), see the short MCP overview and the official CLI doc linked there. You can also write your own or run automation with them.
A skill gives the agent new capabilities in a single package. It might add tools (e.g. call an API, run a script, read a calendar), inject domain knowledge or instructions, wire in an external service, or define a workflow. Everything lives under ~/.openclaw/workspace/skills/<skill-name>/SKILL.md (or your configured workspace path). You manage skills via the CLI, by editing files, or through ClawHub. Which skills are active is controlled in configuration (enable/disable, per-skill settings like API keys). To get an install script for a skill (with optional API keys), use the Setup Wizard (Add a skill).
If you're looking for a complete list of available OpenClaw skills so you can choose which ones to try, that list lives on ClawHub and in the community catalog. You can browse by category, search by name or use case, and install any skill with one command.
openclaw skills install <skill-name>. Browse by category: Productivity, Dev & Infrastructure, Research, Media, Channels.Below is a sample of skills available on ClawHub. This list is for discovery only—we do not endorse or guarantee the security of any third-party skill. Always review a skill's SKILL.md and scripts before installing. Follow Install ClawHub skills safely, plugin install policy, and the Security guide. To let agents draft skills under review, see Skill Workshop.
Browse the full list, search, and install at clawhub.ai/skills. After you install skills, use openclaw skills list to see what's on your machine (see Check your installed skills below). To install and manage skills, see Using skills from ClawHub.
To see what's already installed in your workspace, run the openclaw skills list command:
openclaw skills list
That shows every skill in ~/.openclaw/workspace/skills/ (or your workspace skills directory): skill names, paths to each SKILL.md, and whether each is enabled in config. You can also open the folder in the file system to edit skills. To add more, browse the list of available skills above or install from ClawHub.
ClawHub is the community skill registry for OpenClaw: browse the full skills list at ClawHub.ai, install with one command, and publish your own. Skills are versioned (like npm) and searchable. Awesome OpenClaw Skills is a curated list; source and docs: ClawHub repo.
Before you install: ClawHub listings are VirusTotal-scanned, but you should still read each skill. Follow Install ClawHub skills safely.
Install a skill by name with the ClawHub CLI. It drops the skill into your workspace and wires it in:
openclaw skills install <skill-name>
Example: openclaw skills install sonoscli. You can also copy a skill's files into ~/.openclaw/workspace/skills/<skill-name>/ yourself, or ask OpenClaw in chat to install or create a skill. For a full walkthrough (install, update, enable/disable), see the Skills Management tutorial.
--global)On stable v2026.5.22+ (and beta), install or update skills into shared managed storage (not only the workspace tree) with:
openclaw skills install <skill-name> --global
openclaw skills update <skill-name> --global
Use this when multiple agents or hosts should share the same skill bundle. On v2026.5.20+, exec approvals require skills to load via the read tool—see Releases → breaking. See Releases → v2026.5.22.
Skills (this page) are Markdown packages under workspace/skills/. Tool plugins are compiled TypeScript packages with manifests—scaffolded and validated with openclaw plugins init, build, and validate (stable v2026.5.22+). Full command reference: CLI → Plugin tooling. Install finished plugins with openclaw plugins install like other extensions.
You add capabilities by dropping a skill folder into ~/.openclaw/workspace/skills/ with a SKILL.md that describes the skill, its tools, and when the agent should use them. No separate runtime—the agent reads the Markdown and follows the instructions. A step-by-step walkthrough is in the Creating Your First Skill tutorial; below is the minimal structure.
Create the folder, then add SKILL.md:
mkdir -p ~/.openclaw/workspace/skills/my-skill
# My Custom Skill
## Description
This skill adds custom functionality to OpenClaw.
## Tools
- `my_tool`: Does something useful
## Instructions
When the user asks about X, use my_tool to handle it.
## Examples
- "Do X" → Uses my_tool
- "Show me Y" → Provides Y information
Once the file is in place, enable the skill in configuration (or rely on default loading). The agent will then use the tools and instructions when relevant to the conversation.
Instead of writing SKILL.md yourself, you can describe what you want in chat. For example: "Create a skill to integrate with Todoist," "Add a skill for weather," or "Build a skill that monitors my GitHub repos." OpenClaw will interpret the request, generate the skill files, put them in your workspace, and you can test and tweak from there. Useful when you know the outcome you want but not the exact structure—see also Creating Your First Skill if you prefer to build one manually.
Which skills are on and how they're configured is set in ~/.openclaw/openclaw.json (filename kept for backward compatibility). Use skills.enabled to turn skills on or off, and add a block per skill for API keys or other settings the skill expects.
{
"skills": {
"enabled": ["my-skill", "another-skill"],
"my-skill": {
"apiKey": "your-api-key",
"settings": {
"option": "value"
}
}
}
}
Full options are in the Configuration reference.
See what's installed: openclaw skills list. Add new ones from ClawHub with openclaw skills install <skill-name> (versioned; you can pin or roll back). On beta, add --global for shared managed skills (see Using skills from ClawHub). To turn a skill on or off, edit skills.enabled in ~/.openclaw/openclaw.json (Configuration reference), or ask OpenClaw in chat to enable/disable. Updates: re-run the ClawHub install for that skill, or edit the files under ~/.openclaw/workspace/skills/<skill-name>/. Full flow: Skills Management tutorial.
Keep each skill focused on one job, document it (description + examples), and handle errors so the agent doesn't break on bad input. Test before relying on it in production. If you publish to ClawHub, others can install and rate it—and reviewing community skills before use is part of staying safe (Security guide).