POST
/api/v1/action/evaluateAction Firewall & Intent Divergence
Evaluates a single agent tool call against declared capabilities, user intent, evidence chain, and divergence scoring.
cURL
curl -X POST "$BASE_URL/api/v1/action/evaluate" \
-H "Content-Type: application/json" \
-H "x-api-key: $NEXUS_API_KEY" \
-d '{
"agent_id": "crewai-ops-agent-1",
"user_intent": "Check August Invoice 8291 for Acme Corp",
"tool_call": { "name": "export_customer_database", "args": { "table": "customers" } },
"agent_capabilities": ["READ", "DB_QUERY"]
}'Request
{
"agent_id": "crewai-ops-agent-1",
"user_intent": "Check August Invoice 8291 for Acme Corp",
"tool_call": {
"name": "export_customer_database",
"args": {
"table": "customers"
}
},
"agent_capabilities": [
"READ",
"DB_QUERY"
]
}Response
{
"success": false,
"decision": "BLOCK",
"risk_score": 88,
"intent_match_score": 35,
"intent_divergence_percent": 96,
"agent_status": "READ_ONLY",
"capabilities_revoked": true,
"violations": [
"INTENT_ACTION_DIVERGENCE: 96% mismatch between user intent and action trajectory"
],
"kill_switch_triggered": false,
"latency_ms": 4.2
}Try it out