The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple system tools, including ffmpeg, ffprobe, playwright, node, python3, and awk, via shell and Node.js scripts to handle video rendering, audio synthesis, and quality assurance.
[DYNAMIC_EXECUTION]: The rendering pipeline generates and modifies HTML and JavaScript (e.g., through wire-audio.mjs and wire-grade.mjs) which is then executed in a headless browser environment to create the video frames.
[DATA_EXFILTRATION]: The skill uses the serve command to expose rendered files over a Tailnet URL. While this is a common workflow for sharing assets with users in cloud-based agent environments, it involves the network exposure of local data.
[INDIRECT_PROMPT_INJECTION]: The skill's reliance on external inputs such as a user-provided 'brief' and music metadata creates an attack surface for indirect prompt injection, which is mitigated by the skill's specific task-oriented instructions.
[PROMPT_INJECTION]: The creative direction instructions include phrases like 'no constraints on creativity' and 'do not ask permission,' which are characteristic of behavioral overrides but are contextually limited to the creative aspects of video design.

motion-launch-video