Can your agent survive this?

Enter any prompt - safe or malicious. See a plausible tool call it could produce, then watch the real Verdicter policy engine allow or block it in real time.