voice-agent
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides documentation and implementation examples for the Zavu voice agent framework. It exclusively uses official vendor-owned CLI tools (zavudev) and libraries (@zavudev/functions) for agent management and execution.
- [PROMPT_INJECTION]: The skill architecture involves processing caller speech transcripts, which is a surface for indirect prompt injection. This is a fundamental characteristic of voice agents and is addressed through platform-level controls and standard prompting practices.
- Ingestion points: Call transcripts derived from caller audio (SKILL.md).
- Boundary markers: Example prompts use markdown section headers (# Personality, # Environment) to structure instructions for the agent.
- Capability inventory: Defined tools can perform network operations (fetch) and telephony control actions (handoff, outbound dialing) (SKILL.md).
- Sanitization: The examples rely on the platform's speech-to-text processing and the LLM's instruction following rather than custom input sanitization.
Audit Metadata