How OpenClaw Skills Work

Skills are not decorative

Skills are one of OpenClaw’s core extension mechanisms. They change what the model knows it can do, which means they affect both capability and risk.

The three-layer loading order

The Orange Paper describes a clear precedence:

1. workspace-level skills
2. user-level skills
3. bundled skills

If the same skill exists in more than one place, the higher-priority layer wins.

What a skill really is

At minimum, a skill is a directory with a SKILL.md file. That file tells the agent:

• what the skill is for
• when it should trigger
• what instructions it should follow
• what tools or environment variables it needs

This is why the skills system feels powerful with relatively little ceremony.

ClawHub is useful, not automatically safe

The Orange Paper documents a large ecosystem, but it also documents quality and safety problems:

• 13,729 registered skills
• thousands filtered out as low quality, duplicate, or malicious
• real supply-chain incidents

The correct takeaway is that discovery is easy, but trust is expensive.

How to install and think clearly

The right order is:

1. identify one repeated task
2. search for the smallest skill that solves it
3. read the source
4. verify what files, commands, or secrets it touches
5. install only after that review

This is slower than app-store thinking, but much safer.

Why AIClawGuide curates the marketplace

Because the Orange Paper makes the risks explicit. Phase one of this site does not try to mirror the whole skill ecosystem. It deliberately stays:

• manual
• editorial
• free or open source only
• explicit about official-source versus community entries

Where to go next

• Open the Skills Hub
• Read OpenClaw Security & Cost Control
• Read OpenClaw Architecture