iart-ai/motion-skills

What it solves

It addresses the difficulty of creating high-quality motion graphics via AI prompts, specifically solving for timing, easing, safe areas, and frame-accurate rendering. It prevents the "hope-and-pray" approach by providing agents with professional workflows and a verification loop to inspect their own visual output.

How it works

The project provides "skills"—self-contained folders consisting of documentation (SKILL.md) and references that teach AI coding agents how to execute specific motion-graphics workflows. These skills are not model fine-tunes or plugins, but rather knowledge bases that agents can auto-discover. For visual artifacts, the system uses a deliver-and-verify loop (rendering a frame, taking a screenshot, and checking it) supported by a toolkit of scripts to probe MP4s or render HTML.

Who it’s for

It is designed for AI coding agents (such as Claude Code, Cursor, and Codex) and the users who employ them to create content for TikTok, Reels, YouTube, podcasters, e-commerce sellers, brand marketers, educators, and frontend developers.

Highlights

• 54 open-source skills organized into 17 installable packs covering everything from kinetic typography to WebGL and Manim math animations.
• Deliver-and-verify loop allows agents to render frames and screenshots to self-correct their work.
• Agent-native format compatible with 40+ agents without requiring model fine-tuning.
• Broad modality support including Remotion, Manim, GSAP, Three.js, and transparent lower thirds for professional editing software.