INTERACTIVE THREAT LAB

Real-Time Autonomous Attack Simulator

Test AgentShield's cryptographic and policy defenses against simulated adversarial attacks. Each test sends real payloads to the running Express backend and inspects the live firewall response.

ATTACK VECTOR 01RFC 8032 Guard

Tampered Intent & MITM Attack

Simulates a man-in-the-middle proxy that alters the payment recipient or inflates the amount without having the agent's Ed25519 private key.

ATTACK VECTOR 02Replay Defense

Cryptographic Nonce Replay

Captures a previously executed valid payment signature and re-submits it in an attempt to trigger a duplicate payment without authorization.

ATTACK VECTOR 03Policy Engine

Unauthorized Overspend Breach

Simulates a rogue or prompt-injected agent attempting to drain funds by submitting a $600 payment that violates the strict $500 hard cap.

GOVERNANCE TEST 04Human-in-the-Loop

High-Value Intent Escalation

Submits a $150 intent exceeding the $100 automated threshold. Verifies that payment execution is frozen until a human officer approves it.