The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted media files to generate transcripts. Malicious instructions contained within the transcribed speech are written to the output markdown file, creating a surface for downstream agents to be influenced by the transcript content.
Ingestion points: Local audio/video files processed by whisperx in scripts/diarize.py.
Boundary markers: Output is formatted as a Markdown file with speaker labels and timestamps.
Capability inventory: Local file system read/write, network access for model retrieval, and use of ffmpeg and uv.
Sanitization: No filtering or escaping is applied to the transcribed text before it is saved.
[EXTERNAL_DOWNLOADS]: Fetches transcription and diarization models from Hugging Face's official repositories during initialization.
[COMMAND_EXECUTION]: Relies on ffmpeg being available in the system PATH for audio processing functionality.

speech-diarizer