The Agent Skills Directory

[COMMAND_EXECUTION]: The skill orchestrates video editing by executing ffmpeg and ffprobe commands via the subprocess.run module. These commands are invoked using argument lists rather than shell strings, which effectively prevents common shell injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input data from clip_plan.json and sources_manifest files. It employs a mandatory validation process including schema checks, type verification (e.g., duration parsing), and geometric constraints to ensure the input data conforms to expected formats before processing.
Ingestion points: clip_plan.json, sources_manifest file, and video path argument.
Boundary markers: The skill validates the structure and content of the JSON input files against specific schemas.
Capability inventory: File system read/write access, and execution of video processing tools (ffmpeg, ffprobe).
Sanitization: Implements strict type checking, range validation for numeric inputs, and uses path objects for file handling.
[DYNAMIC_EXECUTION]: Complex ffmpeg filtergraphs are generated at runtime based on the segments defined in the clip plan. This dynamic generation follows strict templates and is used solely for the intended purpose of concatenating and processing video segments.

video-cut