State of OpenClaw Security
We scanned 6,493 skills from ClawHub and rated their security across 6 dimensions. Updated February 2026.
Why does this matter?
ClawHub — the official OpenClaw skill directory — has zero security ratings. Anyone can publish a skill with no review. ClawStack is the first platform to independently scan and rate every skill.
Security Grade Distribution
Based on permissions, author trust, network requests, and more
Category Safety Rankings
Average security score by category (higher is safer)
How We Score Security
Each skill is rated across 6 dimensions, with a total possible score of 100 points.
Security Scan
30 ptsCisco mcp-scanner YARA analysis. Detects prompt injection, data exfiltration, credential harvesting, and more.
Permissions
20 ptsSensitive permissions requested: file write, shell exec, network access.
Author Trust
15 ptsGitHub account age, followers, and contribution history.
Network Safety
15 ptsExternal URLs and domains contacted. Suspicious endpoints flagged.
Community Trust
10 ptsCommunity safe/suspicious votes from ClawStack users.
Auditability
10 ptsOpen source repo available with complete, readable source code.
Share This Report
Help the OpenClaw community stay safe. Share the State of OpenClaw Security report with your network.