Module 6: Skills — Extending What Your Agent Can Do

Out of the box, your agent can think, remember, and talk. Skills are how you give it new abilities — specific tools and integrations bundled up so your agent knows how to use them.

---

What a skill actually is

A skill is a package that adds a new capability to your agent. It comes with:

• The tool itself — code that does the thing (fetch weather, control a browser, run diagnostics, etc.)
• Instructions — a SKILL.md file that tells your agent how and when to use it
• Scripts and assets — whatever the tool needs to run

When a skill is installed, your agent reads its instructions and knows it has this new capability available. It uses it when relevant, the same way it naturally decides to search the web or write a file.

---

Finding skills

The community skill hub is at clawhub.com. Browse by category or search for what you need.

To see what skills you already have installed:

openclaw skills list

---

Installing a skill

openclaw skills install <skill-name>

Example:

openclaw skills install weather
openclaw skills install healthcheck

That's it. The skill is now available to your agent. You don't have to configure anything in most cases — just install it and tell your agent to use it:

"What's the weather in Miami right now?"

Your agent knows it has a weather skill and uses it.

---

Useful skills to install

weather — Current weather and forecasts via wttr.in or Open-Meteo. No API key needed.

openclaw skills install weather

healthcheck — Security hardening, system health checks, firewall audits. Great if your agent runs on a server you care about.

openclaw skills install healthcheck

tmux — Remote-control terminal sessions by sending keystrokes and reading output. Useful if you want your agent to interact with other CLI tools.

openclaw skills install tmux

browser (built-in) — OpenClaw has browser control built in. Your agent can take screenshots, click things, fill forms, and scrape pages. Ask it:

"Go to [URL] and tell me what's on the page."

---

How agents use skills

Your agent doesn't need you to tell it "use the weather skill." It reads the skill's instructions when it loads and understands what the skill does. When you ask a question that the skill is relevant to, it uses it automatically.

This is why the instructions in a skill's SKILL.md matter. Good skill instructions tell the agent exactly when to use the tool, what inputs it needs, and how to interpret the output.

---

Building your own skill

Skills are just folders with a SKILL.md and whatever scripts it needs. If you have a tool you use regularly that your agent should know about, you can package it as a skill.

The structure:

my-skill/
├── SKILL.md          ← Instructions for the agent
├── scripts/          ← Executable scripts
│   └── do-thing.sh
└── README.md         ← Human documentation (optional)

SKILL.md should answer:

• What does this skill do?
• When should the agent use it?
• How do you call it?
• What does the output mean?

We cover building custom skills in Course 03.

---

Skills vs. just asking your agent to write code

Your agent can write and execute code on its own — so why bother with skills?

Skills are consistent. They work the same way every time, with tested code and clear instructions. Code your agent writes on the fly can vary. Skills are shareable. You build it once, share it to ClaWHub, the whole community benefits. Skills are auditable. You can read exactly what a skill does before installing it. No surprises.

For one-off tasks, let your agent figure it out. For things you do regularly, package it as a skill.

---

Next up

Module 7: Memory That Actually Works →