Collusion Detection
Spot coordinated agent misbehavior.
ML-powered analysis of agent pair interactions. Scores every agent pair on a 0-1 risk scale with confidence ratings. Detects patterns like synchronized denials, shared scope abuse, lineage evasion, sibling coordination, and generation-depth abuse. High-risk pairs flagged automatically with trend indicators.
// Collusion scan result
{
"agentIds": ["agent_a", "agent_b"],
"score": 0.87,
"confidence": 0.92,
"patterns": [
"lineage_evasion",
"sibling_coordination",
"synchronized_denials"
],
"trend": "rising"
}