1 in 36 OpenClaw skills contains security threats — is yours one of them?
The AI Skills you use could be
your nightmare
Community skills can contain malware, data exfiltration, and backdoors. AgentGuard scans them all so you don't have to.
0
Executable skills analyzed
1/36
Contain security threats
Check any skill — free
Search our database of 6,400+ scanned OpenClaw skills
Real threats we found
These skills are live in the OpenClaw marketplace right now
Run Code
by etienneperot · 7.6K downloads
F — Shell execution + filesystem access
OpenRouter 380+ Models
by rbb-dev · 3.8K downloads
F — Code injection + SSH access
GitHub Copilot SDK Pipe
by Fu-Jie · 1.1K downloads
F — DB access + file deletion
Fileshed 1.0.0
by did100 · 802 downloads
F — Privilege escalation
Diagram Generator
by newnol · 747 downloads
F — Dynamic code execution
user_storage
by did100 · 532 downloads
F — Shell exec + file deletion
Protected in 60 seconds
No config, no Docker, no cloud account required
1
Install
One command. Works on Mac, Linux, and Windows. Detects your OpenClaw automatically.
npx @foxifi.io/agentguard init
2
Scan
Instantly grades every installed skill from A (safe) to F (dangerous). Shows you what's risky.
agentguard scan
3
Protect
Auto-fixes what it can, warns about the rest, and monitors for new threats 24/7.
agentguard protect
Free vs Pro
Free tells you what's wrong. Pro tells you where and how to fix it.
Finding descriptions (what's wrong)
✓
✓
Community threat definitions
✓
✓
File locations (file:line)
—
✓
Remediation advice + fix commands
—
✓
MCP runtime proxy + blocking
—
✓
Multi-instance (up to 5)
—
✓
Don't install blind
Every community skill is a piece of code running on your machine. Know what it does before it does it.
Get AgentGuard Free
npx @foxifi.io/agentguard init
Part of the Cyberforce Security Suite
AgentGuard is one product in a complete cybersecurity platform. Protect your APIs, endpoints, browsers, and AI agents — all from one team.
ByeBot Agents
API security — 8 AI agents, 229 skills protecting your endpoints
Watchdog
Endpoint scanner — 25 modules detecting hidden threats on dev machines
Striker
Offensive testing — 34 security checks across recon, assess & attack modes
Hammer SDK
Browser-side protection — 24 packages for client-side security
Explore Cyberforce Suite →
Support AgentGuard
AgentGuard is free and open source. If it saved your AI agents from a bad skill, consider supporting ongoing development.