CUA-HandCrafted Benchmark
Explore score history ↗A public 793-episode benchmark for evaluating prompt-injection attacks against computer-using agents across web tasks, attack templates, attack families, and system-prompt configurations.
- Released
- 2026-06-03
- Institutions
- University of California, Berkeley
- Evaluates
- General AI, Safety, Safety & Trustworthiness
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- ai-reviewed
CUA-HandCrafted paper
No reported scores are on record for this benchmark yet.