Thread
HOWTO: Protect yourself from OpenClaw skill vulnerabilities:
Skills are powerful — they extend what your AI can do. But with great power comes great "wait, what does this actually do?"
Here's your two-step safety check:
1️) Read the SKILL.md — Every skill has one. It's the blueprint. Open it and see exactly what the skill is instructing the AI to do. Browse to the website.com/SKILL.md file or view it directly on ClawHub.ai yourself.
2️) Not technical? Let AI help you. Copy the entire SKILL.md text, paste it into any AI chatbot, and ask:
— "Does this skill do anything dangerous?"
— "What files or data does it access?"
— "Is this safe to install?"
The AI will translate the technical stuff into plain language and flag anything sketchy.
This is the beauty of open source — the code is RIGHT THERE. You don't need to be a developer to verify it. You just need to know where to look.
Stay safe out there.
Replies (5)
clawbot #skills = > key to spyware/stealing/etc
follow above n stay safe
Yep, for sure monkey like me with AI assist can outsmart any bad actor who deliberately put malicious code into this honeypot called ClawdBot or whatever you name it.
Be serious and do not give bad ideas to people who do not understand the risk.
This stuff is full of vulnerabilities, in main app and in this skills bucket of malicious code and yet people give this thing access to their computers in their networks where their bitcoin nodes are running, give them credentials to bank accounts, bitcoin wallets, telegram, email.
What is with you guys? Are you all want to be in the first line when someone selects the targets? Do you know who now has access to your data, to your machines, to your future?
Where “don’t trust. Verify” has gone?
I really do not understand that yolo. This is sad to see.
Guys, If you want to have ai agent then vibe code it yourself. Do some work. Learn something. It might be clunky and vulnerable but not as this supermarket of malware.
Yes, true, but i think this doesn't work at scale, power users will do this, any other user (the majority) will just fall in to the trap
the same goes for installing any software on your phone or your computer. you can only do so much for users.
Yes, right is just that the way someone can be attacked exploded, in ways that we aren’t even been able to see yet. This is why I’m cautious, I’m seeing software quality being degraded, critical infra taken down, and solid foundations that led us to these advances being forgotten. I’m not against all of this, I’m just trying to calibrate my compass accordingly
