audio-transcribe
Pass
Audited by Gen Agent Trust Hub on Apr 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONNO_CODE
Full Analysis
- [NO_CODE]: The core logic script 'transcribe.py' referenced in the workflow instructions is not included in the provided skill files.
- [EXTERNAL_DOWNLOADS]: The skill requires the installation of standard machine learning libraries (mlx-whisper, pyannote-audio, torch, soundfile) from the Python Package Index and downloads pre-trained models from HuggingFace.
- [COMMAND_EXECUTION]: The skill uses the Bash tool to execute the transcription pipeline, manage Python virtual environments, and perform HuggingFace authentication.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it processes untrusted audio files and external transcripts (Markdown, VTT, JSON) for analysis. Ingestion points: audio and transcript files provided by the user. Boundary markers: absent; no specific delimiters or instructions to ignore embedded content are defined. Capability inventory: Bash (subprocess execution), Read, Write, and AskUserQuestion. Sanitization: absent; the skill does not specify procedures for validating the content of transcripts before generating summaries or action items.
Audit Metadata