Hacker News
Show HN: Nyx — Multi-Turn Adaptive Offensive Testing Harness for AI Agents
Nyx is a new Show HN project (20pts, 8 comments) that autonomously probes AI agents for failure modes including logic bugs, instruction-following failures, edge cases, jailbreaks, prompt injection, and tool hijacking through multi-turn adaptive conversations. As more agents ship to production, automated red-teaming tools like Nyx fill a critical gap — the tool represents the growing 'AI security testing' category alongside Anthropic's own internal red-teaming and HackerOne's AI Red Teaming platform.
Source
↳ Follow the thread