The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run() across several scripts (doctor.py, common.py, make_film.py) to interface with standard media processing tools like FFmpeg and ffprobe. These calls are scoped to local media processing tasks.
[EXTERNAL_DOWNLOADS]: The skill downloads Whisper models from official Hugging Face repositories. These downloads occur during the initialization phase via doctor.py or upon first use, and the skill is designed to inform the user before starting these downloads. Hugging Face is a recognized and trusted platform for machine learning models.
[REMOTE_CODE_EXECUTION]: The doctor.py script includes an optional --install flag that executes pip install faster-whisper. This is a documented administrative function used to set up the local environment and is explicitly triggered only after user confirmation, as specified in the skill instructions.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted audio and video content to generate text. It mitigates potential indirect injection risks by instructing the agent to never treat the transcript as authoritative and providing a mandatory review.md file for human verification of the output before it is used in further workflows.

transcribe