elevenlabs-stt
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes audio data from external URLs to generate text transcripts. This introduces an attack surface for indirect prompt injection, where malicious instructions embedded in the audio could be followed by the agent in a multi-step workflow.
- Ingestion points: 'audio' field in the JSON input passed to the 'belt' CLI tool (SKILL.md).
- Boundary markers: The skill does not define specific delimiters or warnings to ignore instructions within the generated transcripts.
- Capability inventory: The skill is configured to use the Bash tool, allowing for subsequent command execution based on processed data.
- Sanitization: No sanitization or filtering of the transcribed output is mentioned before further processing.
- [EXTERNAL_DOWNLOADS]: The documentation references external installation scripts and resources hosted on GitHub at
raw.githubusercontent.com/inference-sh/skills/. This is part of the standard setup for the underlying service infrastructure.
Audit Metadata