Search papers, labs, and topics across Lattice.
2
0
4
Public AI agent skill marketplaces are riddled with security holes, with over 13% of skills containing critical vulnerabilities and confirmed malicious payloads still available.
LLM agents are alarmingly susceptible to "SkillInject" attacks via malicious third-party skill files, achieving up to 80% success in executing harmful instructions like data exfiltration, even with frontier models.
