The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of an external Python package (whiteboard-video-engine) from the author's GitHub repository. While this is a vendor-owned resource, it involves downloading and executing remote code on the local system.
[COMMAND_EXECUTION]: The engine is designed to execute external shell commands for model-based line-art extraction. These commands can be customized via environment variables (WHITEBOARD_INFORMATIVE_DRAWINGS_CMD, WHITEBOARD_ANIME2SKETCH_CMD), which allows the execution of arbitrary scripts or binaries if misconfigured.
[PROMPT_INJECTION]: The skill processes untrusted external data, including user-provided images and scripts, which are used as inputs for the video generation pipeline. This creates an indirect prompt injection surface where malicious filenames or script content could attempt to influence agent behavior or exploit the underlying CLI engine.
Ingestion points: User-uploaded images, SVGs, and text scripts are ingested through the render-photo, render-image, and run commands.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands in input data are present in the provided instructions.
Capability inventory: The skill uses a Python wrapper (scripts/whiteboard_cli.py) to invoke a CLI engine that performs file system operations, executes sub-processes (likely FFmpeg and model scripts), and accesses local files.
Sanitization: The skill instructions do not specify any sanitization or validation logic for the input files or their contents, relying entirely on the external engine's implementation.

whiteboard-video