The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides a bundled Python CLI script (scripts/transcribe_diarize.py) and instructs the agent to execute it for processing audio files. This is a standard architectural pattern for such tasks.
[EXTERNAL_DOWNLOADS]: The skill documentation recommends installing the openai package from standard registries (PyPI) via pip or uv. This is a well-known and trusted dependency.
[DATA_EXPOSURE]: The script reads local audio files and environment variables (OPENAI_API_KEY) to facilitate API requests. It explicitly instructs the agent never to ask the user to paste keys into the chat, adhering to secure secret management practices.
[PROMPT_INJECTION]: As the skill processes untrusted audio data and converts it into text that enters the agent's context (the transcription), there is a surface for indirect prompt injection. However, this is inherent to the transcription use-case and the skill itself does not provide dangerous capabilities that would make this surface highly exploitable.

transcribe