The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes ffmpeg and ffprobe using the subprocess module to process audio and video files. These calls use argument lists rather than shell strings, following security best practices to prevent command injection.
[EXTERNAL_DOWNLOADS]: During its first run, the skill downloads transcription models from Hugging Face using the faster-whisper library. Hugging Face is a recognized and well-known service for hosting AI models.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted media files to generate text transcripts, creating a surface for indirect prompt injection if the resulting text is used in subsequent agent prompts without validation. • Ingestion points: Media files provided via the --input argument in scripts/transcribe.py and scripts/detect-language.py. • Boundary markers: None present; the transcribed text is output directly into SRT, WebVTT, or JSON files. • Capability inventory: File system writes for subtitles and transcripts; subprocess.run execution for media processing. • Sanitization: No specific sanitization or filtering is performed on the transcribed text content.

speech-captions