The Agent Skills Directory

[DYNAMIC_EXECUTION]: The module scripts/browser.mjs dynamically resolves and loads the playwright package by searching multiple locations, including the current working directory and a directory specified by the FILM_DIR environment variable. This pattern can be exploited if a malicious package is placed in those locations.
[DYNAMIC_EXECUTION]: The skill's core engine in assets/engine/main.js dynamically imports and executes a user-provided JavaScript file (spec.js), which serves as the film's definition.
[EXTERNAL_DOWNLOADS]: The shell script scripts/get_fonts.sh fetches font assets from the official Google Fonts GitHub repository. The process is secured with hardcoded SHA-256 checksum verification for every downloaded asset.
[COMMAND_EXECUTION]: Multiple Python scripts (render.py, post.py, mix.py, etc.) and Node.js components utilize subprocess and child_process to run external utilities such as ffmpeg, ffprobe, and npm. Command construction consistently uses list/array formats to prevent shell injection.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted media and script files. While it includes a comprehensive sanitizer in assets/engine/sanitize.js to strip dangerous elements from SVG logos, the overall architecture allows for the execution of complex logic from user-controlled files.

scenario-kinetic-film