Autonomous LLM pentesting powered by autoresearch. Multi-agent framework where Claude + Codex collaborate to discover novel AI safety issues. - View it on GitHub