speech-to-text
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted audio data from external URLs for transcription. The resulting text is returned to the agent context, which could allow instructions embedded in the audio to influence the agent's behavior.
- Ingestion points: Audio files fetched from user-provided URLs in
SKILL.mdexamples. - Boundary markers: Absent; no specific instructions provided to the LLM to ignore potentially adversarial content in the transcript.
- Capability inventory: The skill has access to the
Bashtool, though it is restricted tobeltCLI commands. - Sanitization: Not implemented in the skill instructions.
- [EXTERNAL_DOWNLOADS]: The skill references installation scripts and CLI tools from the
inference-shorganization on GitHub. These downloads are necessary for the skill's primary purpose and originate from the service provider's official repository.
Audit Metadata