deepgram-python-audio-intelligence
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill documentation includes a command to fetch additional product knowledge and MCP setup components via
npx skills add deepgram/skills. This is an official extension mechanism for the vendor's ecosystem. - [INDIRECT_PROMPT_INJECTION]: The skill processes external audio files and streams, which could potentially contain spoken instructions designed to influence the agent's behavior after transcription.
- Ingestion points: Data enters the context through
client.listen.v1.media.transcribe_urlandtranscribe_filecalls inSKILL.md. - Boundary markers: No explicit delimiters or instructions are used in the examples to isolate the transcribed text from the agent's control logic.
- Capability inventory: The skill interacts with the Deepgram API for transcription and performs local file reads.
- Sanitization: The examples demonstrate direct printing and processing of the results without sanitization or filtering steps.
Audit Metadata