probeagent

ProbeAgent is a command-line tool designed for offensive security testing of AI agents, performing automated red-team attacks such as prompt injection and credential exfiltration against any HTTP-accessible agent. Notable features include a detailed attack grading system that categorizes responses as Compromised, Resisted, or Blocked, allowing users to evaluate the effectiveness of their security controls, and advanced guardrail detection to distinguish between model defenses and actual security mechanisms.