ClawHavoc: hundreds of credential-stealing skills flood the ClawHub marketplace: why it is not safe to use

What happened

Security firm Koi Security checked every skill on ClawHub, a marketplace of add-ons for a popular AI agent, and on February 1, 2026 reported 341 harmful ones. Most hid fake setup instructions that told users to run commands, which quietly installed a rented password-stealing program. That program grabs saved passwords, browser data, login keys, and crypto wallets, then sends them to servers the attackers control. Some skills also stole bot credentials, hid backdoors inside working code, and edited the agent's memory files so they survived being deleted. A later rescan on February 16 counted 824 bad skills, and another lab has since logged 1,184 over time.

What it means for you

A skill is just code someone else wrote. If you install one from an open marketplace, it can run on your computer with your access. Anyone can publish, and names or star counts can be faked, so a skill that looks helpful may be built to rob you.

What to do instead

• Never paste and run setup commands you do not understand.
• Only add skills from a source that reviews and tests them first.
• If something feels off, remove the skill and change important passwords.

AgentPod lists only skills we have reviewed and tested, so you are not judging safety alone. No list is perfect, but a checked catalog beats an open free-for-all.

Sources

• https://www.koi.ai/blog/clawhavoc-341-malicious-clawedbot-skills-found-by-the-bot-they-were-targeting
• https://thehackernews.com/2026/02/researchers-find-341-malicious-clawhub.html
• https://unit42.paloaltonetworks.com/openclaw-ai-supply-chain-risk/