The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts/extract.py script executes media tools using subprocess.run with list-based arguments, protecting against command injection from user-provided file paths.
[EXTERNAL_DOWNLOADS]: The skill correctly identifies ffmpeg as an external dependency and directs users to the official project site, ffmpeg.org, a well-known service.
[PROMPT_INJECTION]: The risk of indirect prompt injection from video content is handled by the skill's structured reasoning workflow.
Ingestion points: Video file paths provided by the user.
Boundary markers: Defined workflow phases in SKILL.md that isolate visual analysis.
Capability inventory: Subprocess calls limited to ffmpeg/ffprobe for frame extraction.
Sanitization: Use of pathlib and secure argument handling in the Python script.

video-analyzer