The Agent Skills Directory

[COMMAND_EXECUTION]: Uses ffprobe and ffmpeg to identify media formats, extract audio from video tracks, and re-encode audio to MP3 format for compatibility with the Whisper API.
[CREDENTIALS_UNSAFE]: Reads an authentication token from the user's home directory at ~/.openkbs/projectJWT to authorize transcription requests to the OpenKBS proxy.
[EXTERNAL_DOWNLOADS]: Communicates with the vendor's transcription endpoint at https://proxy.openkbs.com/v1/audio/transcriptions to perform speech-to-text processing.
[DATA_EXFILTRATION]: Temporarily stores processed audio chunks in a project-relative directory (site/_tmp/) to allow the remote transcription service to fetch the media files via a public URL during the transcription process.

file-transcribe