Fake AI Agent Skill Collects Emails, Demonstrates Marketplace Vulnerability
Monitor skill marketplaces for fake AI agent skills that collect email addresses.
Monitor skill marketplaces for fake AI agent skills that collect email addresses.
Summary
Security firm AIR constructed a fake AI agent skill and pushed it through a popular skill marketplace and an Instagram advertisement.
The malicious skill reportedly reached roughly 26,000 agents, including some on corporate accounts.
Despite being flagged as harmless, the payload simply collects the user’s email address and performs no other action.
All skill security scanners tested by AIR marked the skill as safe, highlighting a gap in marketplace vetting.
The demonstration shows how attackers can exploit trust in skill ecosystems to harvest contact information.
Security teams should scrutinize new skills and monitor for unexpected data collection.
The incident underscores the need for stricter validation processes in AI agent marketplaces.
Key changes
- Fake AI agent skill built by AIR
- Pushed through skill marketplace and Instagram ad
- Reached ~26,000 agents
- Some on corporate accounts
- Payload harmless, collects email address
- Security scanners marked it safe