The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The mixkit.py script is designed to fetch audio assets (music and sound effects) from the well-known stock media service mixkit.co. This behavior is central to the skill's purpose of creating scored product films.
[COMMAND_EXECUTION]: The skill relies heavily on ffmpeg and ffprobe for media processing. These tools are invoked across several scripts (beats.py, mix.py, peaks.py, loud.py, scan.py, and render.mjs) to handle tasks like beat detection, audio mixing, and video encoding. Analysis shows these commands are executed using argument lists rather than raw shell strings, which reduces risk.
[DYNAMIC_EXECUTION]: The core rendering logic in render.mjs uses Playwright to execute dynamically generated JavaScript within a headless browser context. This allows the skill to render frames based on the seek(t) logic defined in the index.html templates.
[INDIRECT_PROMPT_INJECTION]: As the skill processes user-supplied briefs and asset descriptions into rendered HTML, it possesses an attack surface for indirect prompt injection. However, the skill implements a mandatory human-review step (the Step-4 gate) where a user must approve the beat map and stills before the full video is built, significantly mitigating the impact of potential adversarial data.
[CREDENTIALS_SAFE]: No hardcoded credentials or sensitive environment variable exposures were detected. The skill uses standard package registries and official media sources.

launch-video