AgentPoison

Evals

Benchmark / dataset

Memory and knowledge-base backdoor poisoning of agents.

Category
Prompt injection & agent security
Insurance exposure
Data poisoning
Publisher
UIUC / Virtue AI
Publisher type
Academic
License
MIT
Status
Active (<90d)
Last commit
2026-10-14
Stars
248

View source on GitHub →

Wanich Risk Lab, 2026.