The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/understand_video.py uses subprocess.run to execute ffmpeg and ffprobe for frame extraction and video metadata retrieval. The commands are passed as lists of arguments, which is a secure practice to prevent shell command injection.
[EXTERNAL_DOWNLOADS]: The skill offers to install the openai-whisper Python package from the official Python Package Index (PyPI) if it is not already present on the system. This installation only occurs after an interactive user confirmation.
[INDIRECT_PROMPT_INJECTION]: The skill extracts text transcripts from video files. This creates an indirect prompt injection surface if the video contains adversarial spoken content intended to influence the behavior of the agent that processes the resulting transcript.
Ingestion points: Video file input processed in scripts/understand_video.py.
Boundary markers: None identified in the prompt interpolation logic.
Capability inventory: Local file system writes, execution of system binaries (ffmpeg, ffprobe), and package installation via pip.
Sanitization: No sanitization or filtering is performed on the extracted transcription text before it is returned in the JSON result.

video-understand