The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill dynamically loads and executes JavaScript files from the films/ and pieces/ directories within a headless Chromium instance. This is done via document.write in assets/film.html and by bundling files into a single bundle script in scripts/film-render.mjs and scripts/film-score.mjs which is subsequently executed in the browser context.
[COMMAND_EXECUTION]: The scripts film-render.mjs and film-score.mjs execute external commands using spawn, execSync, and execFileSync. Specifically, they invoke ffmpeg for video encoding/audio muxing and python3 to locate the imageio_ffmpeg executable using a dynamic command string.
[EXTERNAL_DOWNLOADS]: The rendering scripts include logic to automatically import and potentially trigger the installation of the playwright package and its Chromium browser dependencies via npm and npx commands if they are missing from the environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted brief files (FILM-BRIEF.md, SCORE-BRIEF.md) which are used to instruct sub-agents. There is a lack of explicit sanitization or boundary markers when these files are interpolated into the sub-agent prompts defined in references/agent-prompts.md, creating a surface for indirect instructions to influence the agent.

motion-film