elevenlabs-stt
Pass
Audited by Gen Agent Trust Hub on Mar 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill relies on the
infsh(inference.sh) CLI tool to execute various audio processing tasks. Theallowed-toolssection explicitly limits the agent to theinfshcommand set. - [EXTERNAL_DOWNLOADS]: The documentation references an external installation script for the
infshCLI located athttps://raw.githubusercontent.com/inference-sh/skills/refs/heads/main/cli-install.md. This URL belongs to the vendor's official repository. - [PROMPT_INJECTION]: As a transcription tool, the skill processes external audio content that may contain spoken instructions designed to influence the behavior of an AI agent (Indirect Prompt Injection).
- Ingestion points: Audio files from remote URLs are processed into text.
- Boundary markers: Commands use JSON formatted input to isolate audio URLs from instructions.
- Capability inventory: Uses
infshfor transcription andnpx skills addfor extension management. - Sanitization: The skill produces text output based on audio; downstream agents should treat this output as untrusted content.
Audit Metadata