Research

Published security research on the AI agent ecosystem.

The OpenClaw Security Assessment

Ongoing — Last updated February 2026

The most comprehensive security analysis of the AI agent skill ecosystem. 908+ ClawHub skills audited across 87 rounds. 5,411+ findings documented (806 critical, 1,377 high). 115 attack patterns across 25 attack classes, mapped to the OWASP Agentic AI Top 10. 5 active responsible disclosures with framework maintainers. 12 historical attack studies (Stuxnet to Cambridge Five) yielding 68 detection patterns.