The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates video processing by instructing the agent to execute a local Node.js script located at ~/.claude/skills/video-effects/run.cjs with user-supplied file arguments.
[EXTERNAL_DOWNLOADS]: The skill depends on a local 'engine' directory named video-effects and requires the installation of standard dependencies such as ffmpeg, Node, Puppeteer, and torch to perform video rendering.
[INDIRECT_PROMPT_INJECTION]: The skill processes external video files (.mp4) provided by the user. While these files are ingested as command-line arguments for the transformation script, there is no evidence that the content of the video files is parsed or interpreted by the agent in a way that could lead to instruction injection. Evidence Chain: 1. Ingestion: my-clip.mp4 (user file). 2. Boundary: Not explicitly defined. 3. Capability: Local script execution via Node.js. 4. Sanitization: Handled by the underlying shell/script environment.

magnifying-lens