garden-skills: Skillset for AI Agents to Design and Produce Videos

This is a collection of skills that connects to AI coding agents like Claude Code and Cursor to automate design, frontend development, and video production. The core idea is to empower agents with visual polish and editorial judgment, going beyond simple code generation. It is released under the MIT license and has recorded over 10,000 stars.

The web-video-presentation skill converts scripts or text into web presentations at 1920×1080 resolution. Scenes transition based on clicks and keyboard inputs, and 23 built-in themes can be applied for immediate use as screen recording videos. It also supports narration audio synthesis via TTS integration options.

web-design-engineer refines AI-generated web artifacts from mere functional implementation into visually memorable results. It offers 25 anchor style recipes, including Linear, Aesop, and Bloomberg, and allows users to choose between expansion, preservation, or complete overhaul modes when redesigning existing products. It is designed to avoid generic AI UI patterns through an anti-cliché blocklist.

The gpt-image-2 skill generates posters, UI mockups, infographics, and more using GPT Image 2 and OpenAI-compatible APIs. It automatically detects three runtime modes—local generation, host-native delegation, and prompt-only support—to determine the execution path. It provides 79 structured prompt templates and a library of over 160 public examples.