The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts in assets/starter/scripts/ utilize ffmpeg and ffprobe for media processing tasks, such as padding audio tracks and verifying the final video's dimensions, framerate, and duration. These commands are executed with internally constructed file paths and pose no command injection risk.
[EXTERNAL_DOWNLOADS]: The prepare_audio.py script uses the edge-tts library to fetch narration audio from Microsoft's official speech synthesis services. This is a core feature of the skill for generating video voiceovers from text.
[DYNAMIC_EXECUTION]: The skill performs dynamic execution in two controlled contexts: scaffold.py uses runpy.run_path to load helper functions from internal scripts during project initialization, and render.mjs utilizes Playwright's page.evaluate() to execute the Motion Canvas animation engine within a browser for rendering frames.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied narration and mathematical definitions to create video content. It uses structured JSON schemas and text normalization to process this data, preventing the content from being interpreted as instructions by the underlying AI agent.

technical-explainer-video