onnx-asr-transcription
Pass
Audited by Gen Agent Trust Hub on Mar 1, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it processes external audio data and converts it into text that an agent might interpret as instructions. \n
- Ingestion points: Audio file paths provided to scripts/extract_timestamps.py and scripts/transcribe_long.py. \n
- Boundary markers: The skill does not implement delimiters or provide instructions for the agent to treat generated transcripts as untrusted. \n
- Capability inventory: The skill performs local file writes and suggests the use of shell commands like ffmpeg for pre-processing. \n
- Sanitization: No sanitization or filtering is applied to the transcribed text before it is saved to output files. \n- [COMMAND_EXECUTION]: The documentation provides examples for executing shell commands, including ffmpeg and onnx-asr CLI tools. \n- [EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python dependencies, including onnx-asr and onnxruntime (a well-known library from Microsoft).
Audit Metadata