cordy27/canvotion-skill

Canvotion Skill is a library of instructions built for AI coding agents that need to create visual and video content. It pairs an npm command line tool called Canvotion with a set of tuned system rules so that a large language model can write and render graphics without a human doing the drawing by hand. The library covers two rendering engines. LeaferJS handles two dimensional animation: strict spatial layout, timing curves, motion along SVG paths, and lighting effects meant to look like a polished product launch video. Three.js handles three dimensional scenes, starting from basic meshes and going up through materials, lighting systems, and shader based post processing. Instructions are organized into four layers of rules under a references and authoring folder: planning the video structure and script, visual layout and self review for aesthetics, the actual 2D code including shapes, styles, animation, media, and voiceover, and Three.js specific rules for scenes, materials, lighting, and post effects. A manifest file tracks the schema version, and a SKILL.md file defines how the Canvotion CLI is meant to be used. To use it, a person installs the Canvotion CLI through npm, logs in, and creates a project. The repository's rule files are then loaded into an AI agent's system prompt, and the agent writes JavaScript code that the CLI can verify for syntax and visual correctness before rendering. Finished scenes can be exported as video, including a transparent alpha channel format meant for use in other editing software. This project is aimed at people building or using AI agents that need to generate marketing style animations or 3D scenes automatically, rather than at someone who wants to write animation code by hand. It is released under the MIT license, so it can be used, modified, and redistributed freely.