gladia-pre-recorded-transcription
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted audio and video content from local paths or remote URLs, creating a surface for indirect prompt injection.
- Ingestion points: The
transcribe()method in both Python and TypeScript SDK examples accepts arbitraryaudioInput, which can be an HTTP(S) URL or a local file path, as documented inSKILL.mdandreferences/transcription-options.md. - Boundary markers: The instructions and examples do not include delimiters or specific guidance for the agent to disregard instructions contained within the transcribed output.
- Capability inventory: The skill enables reading local files (via upload), making network requests (via the
callback_urlparameter), and deleting remote job data usingdelete(). - Sanitization: The documentation does not specify any sanitization, filtering, or validation of the transcribed text before the agent interacts with it, such as when using the
audio_to_llmfeature.
Audit Metadata