Optional ReadonlyblockedOptional ReadonlyblockedCustom message when model output is blocked
ReadonlycontentContent filter configuration across safety categories
Optional ReadonlycontextualContextual grounding filters for response accuracy
Optional ReadonlydescriptionGuardrail description
Optional ReadonlyinterventionIntervention alarms on this guardrail, covering every guardrail version. Requires the
module-level guardrailAlarms block, which sets the notification target and timing.
Use cases: alerting on spikes in blocked content, intervention-rate monitoring
AWS: CloudWatch alarms on AWS/Bedrock/Guardrails InvocationsIntervened
Validation: Optional; GuardrailInterventionAlarmProps; at least one threshold
Optional ReadonlysensitiveSensitive information filters for PII and custom patterns
Custom message when user input is blocked