retellai-policy-guardrails
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a webhook handler in
references/policy-code.mdthat processes external user transcripts to determine call flow (consent and escalation triggers). - Ingestion points:
references/policy-code.mdprocessesreq.body.transcript. - Boundary markers: Not applicable for this code-based implementation.
- Capability inventory: Modifies call state via
retell.call.updateand influences the final response generated by the agent. - Sanitization: Employs regex and string inclusion checks to identify specific user intent for safety and compliance.
Audit Metadata