Search papers, labs, and topics across Lattice.
Chongqing University
3
0
4
A single poisoned skill can stealthily compromise LLM agents, activating only under specific conditions, raising alarms about the security of AI skill supply chains.
Malicious instructions hidden in images can bypass existing skill scanners, exposing a critical vulnerability in LLM-based systems.
Coding agents are vulnerable to a new class of stealthy, automated prompt injection attacks via poisoned skills, achieving high success rates even in realistic software engineering tasks.
