diagram-design
Also known as: cathrynlavery diagram-design, editorial diagram system, AI coding diagram pack
The diagram-design repository by Cathryn Lavery launched in 2026 and quickly became one of the most-starred diagram resources on GitHub, reaching more than 48,000 stars by October. The premise is that AI coding agents are good at generating diagrams from Mermaid (a markdown-like diagram language) but the output often looks generic or cluttered. diagram-design provides an alternative: a library of carefully art-directed diagram templates that agents can populate with real content.
Each template is a single, self-contained HTML file that embeds its own SVG. There are no shadows, no gradient fills, and no Mermaid dependencies. The design choices are intentional: clean editorial style that reads well in docs, articles, README files, and landing pages. The 42 types cover architecture diagrams, sequence diagrams, flow charts, data models, timelines, and more.
For builders, the practical use is to drop the relevant template file into a Claude Code or Codex session and ask the agent to populate it with the specifics of the system being documented. Because the template is plain HTML and SVG, the agent can edit it directly without needing a design tool. The result is a diagram that looks considered rather than auto-generated, which matters when the output ends up in a client-facing doc or a public repo.