Top downloaded skill in ClawHub contains malware
1password.com
February 5, 2026
7 min read
🔥🔥🔥🔥🔥
63/100
Summary
OpenClaw's agent skills provide extensive access to files, tools, browsers, and terminals, creating potential vulnerabilities. The system's long-term memory feature can capture user behavior, increasing the risk of exploitation.
Key Takeaways
- OpenClaw's agent skills have real access to files, tools, and long-term memory, making them a target for modern infostealers.
- Running OpenClaw on company devices poses significant security risks, and users should treat any prior usage on work devices as a potential security incident.
- The markdown format used for skills in OpenClaw can include executable commands and scripts, posing a risk of malware delivery.
- The Model Context Protocol (MCP) does not guarantee safety, as skills can bypass it through social engineering and direct execution commands.
Community Sentiment
Positives
- OpenClaw's potential to automate customer support by integrating backend actions could significantly enhance user experience and efficiency in handling inquiries.
Concerns
- Security concerns surrounding OpenClaw are alarming, with users granting full permissions without adequate safeguards, raising serious questions about trust and safety.
- The current state of oversight in the AI skill ecosystem is negligent, with no effective security model in place, making installations highly risky.
- The article's lack of detail and reliance on AI-generated content undermines its credibility, leaving readers frustrated and skeptical about the claims made.
