azure-speech-to-text-rest-py
Pass
Audited by Gen Agent Trust Hub on Apr 14, 2026
Risk Level: SAFEDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill reads local audio files to perform transcription services via external APIs.
- Evidence: Uses
open(audio_file_path, "rb")and transmits the data to remote endpoints. - Context: Transmission is directed to official Microsoft Azure subdomains (
*.stt.speech.microsoft.com), which is an expected behavior for the skill's purpose. - [EXTERNAL_DOWNLOADS]: Recommends and utilizes standard Python libraries for network operations.
- Evidence: Instructions include
pip install requestsand the implementation usesaiohttpfor asynchronous requests. - [PROMPT_INJECTION]: The skill ingests and processes audio data which is then transcribed into text, creating a surface for indirect prompt injection.
- Ingestion points: Input audio files via
audio_file_pathparameter inSKILL.md. - Boundary markers: No explicit delimiters or instructions to ignore embedded content are present in the code snippets.
- Capability inventory: Network communication (
requests.post,aiohttp.ClientSession.post) and file system access (open) withinSKILL.md. - Sanitization: No sanitization, validation, or filtering of the transcribed text is performed before it is returned to the agent context.
Audit Metadata