configuring-guardrails

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Celigo GuardrailImport of type ai_agent evaluates per-record input by sending the parent-provided mapped data (including guardrail.aiAgent instructions) to an LLM for custom judgment, so any outsider-authored free text that flows into the guardrail’s input mapping can be ingested at runtime.