audio-transcribe-note
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided audio data and transcripts which could contain malicious instructions. Mandatory evidence: 1. Ingestion points: Audio files are read in scripts/transcribe.py, and transcript text is processed in scripts/correct_transcript.py and scripts/build_note.py. 2. Boundary markers: Absent; system prompts do not use delimiters or instructions to ignore embedded commands. 3. Capability inventory: The skill has file system write access and performs network requests to OpenRouter. 4. Sanitization: Absent; the transcribed text is passed directly to the LLM.
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for the user to install the uv package manager from astral.sh, which is a well-known and reputable service in the developer community.
- [COMMAND_EXECUTION]: The check_dependencies.py script executes the uv sync command via subprocess.run to verify or install project dependencies. This is a standard and expected behavior for a Python project using this package manager.
Audit Metadata