Agent Security Bench (ASB)

Evals

Benchmark / dataset

Attacks and defenses across 10 agent scenarios.

Category
Prompt injection & agent security
Insurance exposure
Prompt injection
Publisher
Rutgers (AGI Research)
Publisher type
Academic
License
MIT
Status
Active (<90d)
Last commit
2026-09-30
Stars
315
Notes
ICLR 2025.

View source on GitHub →

Wanich Risk Lab, 2026.