diagram-design

diagram-design is a diagramming Skill open-sourced by Cathryn Lavery, with the plugin listings published by her company, LittleMight. Describe what you need in plain words, such as an app architecture diagram, a quadrant of projects by impact and effort, or the steps of a process, and the agent picks a diagram type, states its plan, and writes one HTML file you can open with a double click, drawn in inline SVG. It covers 45 diagram types, including flowcharts, timelines, org charts, Venn diagrams, funnels, Gantt charts and several data charts, each in light, dark and full-editorial styles. Point it at a website and it pulls that site's colors and fonts so every later diagram matches; it can also redraw existing draw.io, Mermaid and Excalidraw diagrams in this style and export PNG or SVG for slides and social cards. What sets it apart is that it writes taste down as concrete rules: the accent color goes only to the one or two things that matter, lines are 1px hairlines, there are no shadows, and every coordinate sits on a 4px grid. The author is open about the limits: the same prompt can lay out differently twice, the result isn't a canvas where you can drag boxes around, and for diagrams that live in Git and change often, Mermaid is still the better fit.

Who it's for:people who want a good-looking diagram for an article, report or course without dragging boxes around in design software. You'll need an AI tool that supports Skills, such as Claude Code, Codex or Cursor.

Before you start: install with npx skills add cathrynlavery/diagram-design, or through the plugin marketplaces for Claude Code, Codex, GitHub Copilot and others. The author's privacy policy says the Skill has no server, no accounts and collects no data; what you provide is processed by your AI provider. Generated diagrams request fonts from Google Fonts when opened in a browser or exported to PNG, and removing the font link makes them fall back to system fonts. Official builds come only from this repository, and listings under other names are unofficial copies. Output quality depends on the model you use. The license is MIT.

Based on the project's public documentation. Not tested by AI Inspiration World.