video-understand

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The script scripts/understand_video.py executes shell commands to invoke ffmpeg, ffprobe, and whisper CLI tools. The implementation uses the subprocess.run method with a list of arguments rather than a single string with shell=True. This is a security best practice that prevents shell injection vulnerabilities.
  • [EXTERNAL_DOWNLOADS]: The skill includes logic to install the openai-whisper package from the Python Package Index (PyPI). The script checks if the package is available and, in interactive sessions, prompts the user before executing pip install. The package belongs to a well-known organization and is relevant to the skill's stated purpose of transcription.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted video files. 1. Ingestion points: Input video file path argument in scripts/understand_video.py. 2. Boundary markers: Absent. 3. Capability inventory: subprocess.run calls to media processing tools and file system writes for extracted frames. 4. Sanitization: The script uses argument lists for command execution to prevent injection and verifies the existence of the input file before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 06:34 PM