Last released May 10, 2026
Benchmark prompt-injection resilience and tool safety for dual-LLM agent architectures.
Supported by