The Agent Skills Directory

[SAFE]: The skill contains no malicious code, obfuscation, or unauthorized data access patterns. All operations are aligned with the stated purpose of video analysis.
[COMMAND_EXECUTION]: Uses standard shell commands for ffmpeg and ffprobe to inspect metadata and extract frames. These commands are template-based and execute within the expected operational scope of a media analysis tool.
[EXTERNAL_DOWNLOADS]: Recommends installing the well-known ffmpeg utility through official system package managers like brew or apt-get during setup.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user video files, which is its primary function. The potential attack surface is managed through structured prompting.
Ingestion points: External video files (input.mp4) provided by the user.
Boundary markers: Employs structured output templates (JSON) and predefined analysis categories (Colors, Typography, Layout) to constrain AI vision interpretations.
Capability inventory: Utilizes subprocess execution for FFmpeg utilities and file system access for frame management.
Sanitization: Relies on FFmpeg's media decoding as a processing layer and provides explicit semantic analysis instructions to the vision model.

video-analyzer