Guided risk classification wizard based on Singapore MGF. Score agents across autonomy, impact scope, impact type, data sensitivity, reversibility, and frequency — producing actionable risk levels.
Six-step questionnaire walks through Decision Autonomy (advisory → fully autonomous), Impact Scope (individual → external/public), Impact Type (operational → safety-critical), Data Sensitivity (public → regulated PII), Reversibility (easily reversible → permanent), and Frequency (rare → continuous). Each answer maps to a 1-4 score.
Dimension scores are weighted and combined into an overall risk level: Minimal (0-25), Limited (26-50), High (51-75), or Unacceptable (76-100). Visual breakdown shows contribution of each dimension. Stored as a risk assessment record linked to the agent.
// Risk assessment output
{
"agentId": "agent_123",
"riskLevel": "high",
"riskScore": 68,
"dimensions": {
"autonomy": { "answer": "autonomous", "score": 4 },
"impactScope": { "answer": "organization", "score": 3 },
"impactType": { "answer": "financial", "score": 3 },
"dataSensitivity": { "answer": "pii", "score": 3 },
"reversibility": { "answer": "difficult", "score": 3 },
"frequency": { "answer": "continuous", "score": 4 }
},
"recommendations": [
"Require human approval for all financial actions",
"Enable continuous monitoring with anomaly alerts",
"Restrict data access scopes to minimum required"
]
}Based on the risk assessment, Custodex recommends specific policies: approval requirements for high-autonomy agents, rate limits for high-frequency agents, data access restrictions for sensitive-data agents, and monitoring thresholds for high-impact agents. One-click to create the recommended policies.
Guided 6-step risk classification wizard
Singapore MGF-aligned risk dimensions
Weighted scoring: Minimal, Limited, High, Unacceptable
Visual dimension breakdown with contribution analysis
Auto-generated policy recommendations from risk scores
Risk assessment history with trend tracking per agent