sf-service-voice-toolkit
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of live voice transcripts, which are untrusted inputs from external callers. This creates a surface for indirect prompt injection if the data is passed to an LLM without proper handling.
- Ingestion points: The skill uses
lightning-service-cloud-voice-toolkit-apiand queries theConversationEntryobject to retrieve live call text (SKILL.md, reference.md). - Boundary markers: Documentation includes a 'Guardrails' section that explicitly instructs developers to 'Treat transcript text as untrusted input — never echo into Apex, prompts, or rendered HTML without sanitization.'
- Capability inventory: The skill is designed to feed transcript data into UI components, AI prompts, and server-side automation (SKILL.md).
- Sanitization: The skill highlights the requirement for sanitization but does not provide specific implementation logic in the provided code snippets.
- [SAFE]: The skill references official technical documentation from Salesforce at
developer.salesforce.com.
Audit Metadata