exepal
Stay safe › Skill check

OpenClaw skill scanner: is this skill safe to install?

Paste a ClawHub link, a GitHub skill folder or a SKILL.md. Skill Check reads every file the way your agent will and tells you, in plain English, whether to install it. Free, no sign-up.

Looks for the tricks used by the 2026 ClawHub malware skills: "paste this in Terminal" set-up steps, download-and-run, password-protected zips, macOS quarantine bypass, fake password pop-ups, Keychain, browser, wallet and API-key theft, reverse shells, hidden instructions and leaked keys. Rules only, no AI model. Nothing is stored. 3 free checks a day.
Watch · 34 s

A bad skill, caught in 34 seconds

A skill that claims to track a Solana wallet, and what the check found.

  • Verdict: don't install, 5 findings, 4 high
  • A command to paste into Terminal and one that turns off the Mac's quarantine warning
  • Free in the browser; $0.005 a skill over the API

Check a skillMore videos on YouTube

Plays from YouTube when you press play. Nothing loads from YouTube before that.

Why check a skill before you install it

A skill is a folder of instructions and scripts your agent follows with your permissions: your files, your browser, your API keys. In February 2026 researchers found hundreds of malicious skills on ClawHub (the "ClawHavoc" campaign). They looked like crypto trackers, YouTube helpers and updaters, and their set-up step asked you to paste one command into Terminal. That command installed a password and wallet stealer.

What it checks

  • Malware set-up steps: "open Terminal and paste this", curl … | bash, base64-encoded commands, downloads of .dmg, .pkg or .exe files, password-protected archives, fake "prerequisite" helpers.
  • macOS stealer tricks: removing the quarantine flag (xattr -d com.apple.quarantine), turning off Gatekeeper, fake password dialogs through osascript.
  • Theft: Keychain and saved browser passwords, crypto wallets and seed phrases, ~/.ssh, .env files, your agent's own keys (~/.openclaw), environment variables sent to a server, webhook collectors and tunnels.
  • Hidden instructions: invisible Unicode, HTML comments and instructions to hide what the agent does from you.
  • Other red flags: reverse shells, long encoded blobs, code built and run at runtime, bundled binaries, leaked API keys.

Ordinary skills (call an API, run a script in the skill folder, read a file you point at) are not flagged.

Let your agent check skills itself

Skill Check is also an API and MCP tool, skill_check, at $0.005 a skill, part of the 20 free calls a day on a free key. Add it to your agent and ask it to check every skill before installing.

curl -X POST https://api.exepal.com/v1/skill-check \
  -H "Authorization: Bearer YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"url":"https://clawhub.ai/owner/skill-name"}'

MCP server: https://api.exepal.com/mcp. See how to connect your agent.

Questions

Is OpenClaw's ClawHub not already scanned?

ClawHub runs VirusTotal and its own scanner on skills it hosts. That catches known malware files well. Skill Check reads the skill's instructions and scripts the way your agent will, so it also catches set-up steps that ask you to paste a command, hidden text and data-stealing instructions, and it works on skills from GitHub, skills.sh or a pasted SKILL.md too.

What happened with ClawHavoc?

In February 2026 security researchers found hundreds of malicious skills on ClawHub. Most looked useful (crypto trackers, YouTube tools, updaters) and had a "prerequisites" step that told you to paste a command into Terminal or download a password-protected zip. That installed a password and wallet stealer on macOS and Windows.

Does 'safe' mean I can trust it?

It means none of the known bad patterns were found in the files we could read. A skill can still do something you don't want with permissions you give it, so read what it asks for, and never paste a command a skill tells you to run without knowing what it does.

Does it work for Claude Code and Codex skills?

Yes. They use the same SKILL.md folder format. Paste a GitHub link to the skill folder or the SKILL.md itself.

Do you keep the skill?

No. The files are fetched, scanned in memory and thrown away. Only a count of checks is logged.

Next checks

Keep going

palnext step

Check skills before your agent installs them.

The same check is an API and MCP tool (skill_check) at $0.005 a skill, with 20 free calls a day on a free key.