Company news
Zenity Labs Launches AI Total: Free Sandbox for Vetting AI Agent Skills
Published 2 October 2026
WHAT HAPPENED
Zenity Labs launched AI Total, a free sandbox tool that tests AI agent skills for malicious behavior by running them in a controlled environment with planted credentials and sensitive files, then records all domains contacted, files touched, and commands executed.
WHO IS INVOLVED
Zenity Labs developed the tool. Researchers tested it against thousands of public skills and found malicious campaigns affecting 1.7 million installs, skills with 250,000 installs hiding in top rankings, and skills instructing agents to download attacker files or replace Claude's own skill creator.
WHY IT MATTERS
AI agent skills from public registries are the fastest-growing and least-inspected dependency layer in the AI stack, with attackers embedding malicious payloads that only activate at runtime rather than in visible code. Teams need a way to verify actual behavior before deploying skills, not rely on code inspection or trust scores.
- Type
- Product launch
- Category
- AI Security
- Source
- linkedin.com ↗
See Zenity’s full company record.
Look up its funding history, investors, hiring and competitors in Mandos. Choose a platform plan, or request access to one company for free.
- Funding history and investors
- Hiring trends and departments
- Competitor comparisons
- Scores and supporting data