An agentic exploitation-evaluation harness: grade whether a coding agent can turn a known bug into an exploit, and contain it while it tries.
By chatting or signing in you agree to the Terms and chat-message logging (revocable in History).