The Agent Skills Directory

[SAFE]: The skill utilizes established and reputable command-line tools, specifically FFmpeg and ffprobe, for video metadata extraction and frame processing. These tools operate within the user's local environment without unauthorized network access.
[SAFE]: The provided shell scripts (video-info.sh, extract-frames.sh, and extract-audio.sh) implement rigorous validation for user-supplied parameters, such as frame rates and timestamps, to prevent potential command injection or unexpected behavior.
[SAFE]: The workflow incorporates a dedicated cleanup mechanism that removes the temporary working directory (located in /tmp) upon completion of the analysis, following best practices for local file system usage.
[SAFE]: The documentation identifies standard external dependencies, including the openai-whisper Python package and the ffmpeg system utility, and provides neutral instructions for their installation through official package managers.
[SAFE]: The skill is designed to process external media data as its core functionality and includes specific logic to safely handle atypical media streams, such as attached MJPEG thumbnails in social media downloads.
Ingestion points: The skill ingests video files provided via arguments and optional user-supplied transcript files.
Boundary markers: Logic is present to identify and skip non-video streams (attached pictures/thumbnails) during extraction.
Capability inventory: The skill is granted access to the Bash, Read, and Agent tools to execute its internal scripts and coordinate sub-analysis.
Sanitization: All numeric and time-based inputs are validated against strict regex and range checks before being passed to shell commands.

analyzing-video