explainx.ai350K+ learn from us
TrendingAI News TodayExplainx for Business
Pricing
explainx.ai

Upskill in AI — 16 free pathways, live workshops & bootcamps, and 50+ courses from practitioners. Plus the skills, tools, and MCP servers to practice on.

follow us

follow on google

Add explainx.ai as a preferred source

corporate training

support@explainx.ai

get started

Find your pathTake Free Evaluation

community

Join the community

learn

mind: share how you thinkpathways — start freeworkshopsbootcampscoursescompare Explainxcertificationsmock testsexplainx universitycorporate traininglearn skills & mcp

discover

skillsmcp serversexplainx mcpestacktoolsmdx readeragentsllmsdesignsdictionarypeopleagi trackerfelony benchranks

company

aboutvisionmissionteaminstructorsteach on explainxpartnershipscommunityhackathonscareers

content

daily AI newsstate of AI — live resultsblogreleasespromptsgeneratorsresource libraryfor LLMsexplainx.ai kids

solutions

all solutionsdeveloper upskillingmarketing upskillingproduct manager upskillingleadership upskilling

newsletter · weekly

Get AI news, tools, and insights in your inbox.

supportcontactprivacytermsdata rightshow we create contentsubmission guidelines

© 2026 AISOLO Technologies Pvt Ltd

explainx.ai

On this page

  • What diagram-design does
  • Install in one command
  • Match your brand once, reuse it everywhere
  • Import Mermaid, draw.io, Excalidraw and PlantUML
  • Export to SVG and PNG
  • Semantic patterns and optional motion
  • How it compares with other ways to draw
  • Limits and what to check first
  • Who should try it
  • Related reading
← Back to blog

explainx / blog

diagram-design: The Agent Skill That Draws Brand-Matched Diagrams in Claude Code and Codex

Open Source, Claude Code, AI Agents, Agent Skills, GitHub Trending

Part of AI Tools and Apps

diagram-design is a trending open skill that makes Claude Code, Codex and Cursor draw one-file HTML and SVG diagrams in your brand. Install, import Mermaid, export, limits.

Nov 19, 2026·9 min read·Yash Thakker
add explainx.ai
go deep
diagram-design: The Agent Skill That Draws Brand-Matched Diagrams in Claude Code and Codex

diagram-design is an open agent skill that makes your coding agent draw diagrams as one self-contained HTML and SVG file, in your own brand colours and fonts. It gained roughly 330 GitHub stars in a single day this week. You install it with one command, ask your agent for an architecture diagram or a sequence in plain words, and get a file you can open by double-clicking. The repo's tagline is blunt: no shadows, no Mermaid slop.

This post covers what the skill does, how to install it in Claude Code, Codex, Cursor and other hosts, how its import and export commands work, and what to verify before you trust it. All details come from the project README. We did not run the skill ourselves for this article, so behaviour described here is the README's account of it.

What diagram-design does

The README describes an agent skill that draws diagrams as one self-contained HTML plus SVG file, in your brand. You can ask for things like:

  • "Make me an architecture diagram of my app: frontend, backend, database, Redis cache."
  • "I need a quadrant showing Q2 projects by impact vs effort."
  • "Give me a sequence of a bearer call with token refresh on 401."
  • "Onboard diagram-design to https://yoursite.com", which pulls your palette and fonts so later diagrams match your site.

The agent picks the type, states its plan, and writes one .html file. The README headline says 51 diagram types, while the repo description says 45; we could not reconcile the two, so check the gallery for the current list. The reference folder lists type files for architecture, data flow, database schema, ER, flowchart, Gantt, heatmap, kanban, Sankey, sequence, state, swimlane, timeline, UML class and many more.

If you have read our explainer on Mermaid slop, the pitch will sound familiar. Generic auto-laid-out AI diagrams look the same everywhere. diagram-design answers that with a design system: a 4px grid, a single style guide as the source of truth for colours and fonts, per-type layout budgets, and primitives for nodes, arrows and labels.

Install in one command

The cross-agent route works with any host that reads Agent Skills, which the README lists as Cursor, Claude Code, Codex, Qoder, GitHub Copilot, Gemini CLI, Cline, Windsurf and Amp:

bash
npx skills add cathrynlavery/diagram-design

The skills CLI finds skills/diagram-design/SKILL.md and installs the whole skill, including references, assets and scripts, into each host you select. If your selection resolves to more than one skills folder, it asks for an install method and recommends symlinks so one update reaches every tool. Pass --copy for independent copies. This standalone install does not follow marketplace updates; run npx skills update diagram-design to pull them.

Native installs give you the slash commands too:

table · 2 cols
HostInstall command (per README)
Claude Code/plugin marketplace add cathrynlavery/diagram-design, then /plugin install diagram-design@diagram-design
Codexcodex plugin marketplace add cathrynlavery/diagram-design, then codex plugin add diagram-design@diagram-design
GitHub Copilotcopilot plugin marketplace add cathrynlavery/diagram-design, then copilot plugin install diagram-design@diagram-design
Factory Droiddroid plugin marketplace add https://github.com/cathrynlavery/diagram-design, then droid plugin install diagram-design@diagram-design --scope user
Pipi install https://github.com/cathrynlavery/diagram-design, then /reload
KiroImport the skill from the repo's skills/diagram-design subdirectory URL

Two practical notes. In Claude Code, third-party marketplaces have auto-update off by default; the README says to run /plugin, open Marketplaces, select diagram-design and enable auto-update. And the README states that official builds come only from this repository: listings under any other name are unofficial copies. That matters because agent skills run with your agent's permissions. For another recent example of an agent-tool repo with a multi-host installer, see our report on the app-store-compliance guard.

Match your brand once, reuse it everywhere

Onboarding is the feature that separates this from a one-off prompt. You give the skill a URL and it extracts the palette and fonts into a style guide. If the project has no style guide yet, the skill asks whether to run onboarding, paste tokens manually or proceed with the default, per the README.

For agencies and freelancers, there are named profiles. You onboard a brand once, save it as a profile, then add a .diagram-design marker file containing profile: <slug> to each client project. Marker projects read the profile from ~/.diagram-design/profiles/<slug>.md, so parallel workspaces can use different brands without overwriting a shared style guide. The profile library is shared across Claude Code, Codex, Factory Droid and Pi.

Import Mermaid, draw.io, Excalidraw and PlantUML

If you already have diagrams, the skill redraws them rather than converting them. The commands in the README include:

text
/diagram-design:import-drawio platform.drawio
/diagram-design:import-mermaid README.md --diagram=all
/diagram-design:import-excalidraw whiteboard.excalidraw --size=slide-16x9 --detail=simplified
/diagram-design:import-plantuml sequence.puml

Supported inputs are broad. For draw.io it reads .drawio, .drawio.xml, .drawio.png with an embedded diagram, and .drawio.svg, including compressed payloads. For Mermaid it accepts .mmd, .mermaid and fenced blocks in Markdown. For Excalidraw it accepts scene files but not PNG or SVG exports. PlantUML import covers sequence and class diagrams only. The README says import parses text only: no rendering, no JavaScript, no browser, no network, no PlantUML runtime, and no followed click targets.

The four dials

The point of import, in the README's words, is not conversion but fitting the output to where it goes. Four dials control that:

table · 3 cols
DialOptionsWhat it changes
Formathtml, svg, png, html+pngThe deliverable
Sizedoc-inline, doc-wide, slide-16x9, slide-4x3, social-og, social-square, print sizes, fitThe viewBox and type ramp; a slide gets larger text
Detailfaithful (up to 24 nodes), balanced (up to 12), simplified (up to 7)How much of the source survives
Audienceengineer, mixed, executiveThe wording, not the count

The audience dial has a nice example in the README: "Auth Service / JWT · RS256 · :8443" becomes "Auth Service / token check → Sign-in" for an executive audience. Every import ends with a fidelity ledger that lists what was merged, collapsed or dropped, such as a decision node folded into an edge label or an unconnected sticky note removed. What never carries over: source coordinates, palette, fonts, draw.io's diagonal connectors, Mermaid's automatic layout and PlantUML skinparams. What always carries over: components, relationships, grouping and direction.

Export to SVG and PNG

Diagrams ship as HTML, but you can export the diagram itself for Figma, slides or social cards with /diagram-design:export-diagram in Claude Code, or /export-diagram in Pi and Factory Droid. Flags include --svg-only, --png-only --scale=3 and --registry. SVG export injects Google Fonts so it renders standalone in browsers, Figma and Illustrator; --system-fonts skips that for offline use. PNG uses Playwright at 2x by default and needs pip install playwright && playwright install chromium once.

Note that exports are diagram-only: editorial cards and headers from the "full" template variants are not included. For the full editorial layout, use your browser's print-to-PDF or a full-page screenshot. The --registry flag writes a JSON sidecar of block metadata for diagrams that use the traceable block decomposition pattern.

Semantic patterns and optional motion

When behaviour matters, the skill picks a semantic pattern first and a visual type second. The README names nine routed patterns: fan-in queues and bottlenecks, repeated stage slots, unstructured-input transformation, paired policy traces, secure paved roads, governance catalogs, compensating security layers, traceable block decomposition and lifecycle phase maps. Each defines its triggers, a budget and a static fallback.

Motion is optional and defaults to none. Modes are none, reveal, step and loop, each with a complete static first frame, deterministic timing and controls. Reduced-motion output shows the full static frame. The README says motion HTML uses one reviewed controller from template-motion.html, and that arbitrary or modified inline scripts, remote assets, CSS imports and executable HTML attributes are rejected. Ordinary output is static and script-free. That is a sensible security stance for a tool that generates HTML files you will share.

How it compares with other ways to draw

table · 3 cols
ApproachStrengthWeakness
Mermaid in MarkdownText source, renders in many toolsAutomatic layout, generic look
draw.io or ExcalidrawManual controlHand work, hard to keep a brand consistent
An agent drawing raw SVGFastInconsistent spacing and style without rules
diagram-designDesign system, brand profiles, imports, exportsDepends on the agent following a long skill; young project

The gap it targets is consistency. Agents can already draw an SVG, but the result changes with every prompt. A skill with a style guide, grid and per-type budgets pushes the agent to the same output each time. Our write-up of a Claude Code skill for isometric interactive figures shows the same idea in another style, and Karpathy's push for diagrams, HTML and video to understand LLM outputs explains why visual outputs are gaining ground.

Limits and what to check first

  • Unverified claims. The README says 51 types, the description says 45. We did not count them, and we did not generate diagrams. Open the gallery at skills/diagram-design/assets/index.html in a clone to judge quality yourself.
  • Agent dependence. Quality depends on the agent following the skill. Weaker or smaller models may drift from the style guide. The /doctor command checks your environment.
  • PNG export needs Playwright and Chromium, which is a heavy one-time install.
  • Imports are lossy by design. The detail and audience dials drop nodes. Read the fidelity ledger before you ship a diagram to a client.
  • Third-party skill risk. Read the repo's PRIVACY.md and SECURITY.md, install only from the official repository, and review commands before running them in a work tree.
  • Maintenance. It is a single-author project with a company behind the plugin listings and a large commit history (473 commits per the repo page), which suggests active work but no support guarantee.

If you build with several agent plugins, the broader ecosystem trend is covered in our look at agent plugin standards across OpenAI, AWS, Cursor and GitHub.

Who should try it

  • Developers writing docs and READMEs who want architecture and sequence diagrams without opening a drawing tool.
  • Founders and marketers who need slide-sized or social-sized diagrams that match the brand site.
  • Teams migrating old Mermaid or draw.io files into one consistent look.
  • Agencies that juggle several brands and can use the profile marker per client.

Start small. Install with npx skills add cathrynlavery/diagram-design, ask for one architecture diagram of a project you know well, and compare it with what you would draw by hand. If the result passes your own review, onboard your brand and reuse it. Follow @explainx_ai for more trending agent tools.

Daily briefFrom Yash · 350K+ students taught

Catch up on AI

One AI story explained in plain English, plus one thing to try. Every day, free. Unsubscribe anytime.

Related reading

  • What is Mermaid slop? The generic AI diagram problem
  • Anatomy of a Claude Code skill for isometric interactive figures
  • Karpathy on understanding LLM outputs with diagrams, HTML and video
  • Agent plugin standards across OpenAI, AWS, Cursor and GitHub
  • app-store-compliance: a pre-submission guard for agent-built apps
  • Snap Easy Lens CLI with Claude Code and Cursor
Spotted something out of date? Let us know.
Daily briefFrom Yash · 350K+ students taught

Catch up on AI

One AI story explained in plain English, plus one thing to try. Every day, free. Unsubscribe anytime.

  • ✓ Free
  • ✓ One email a day, no spam
  • ✓ Unsubscribe anytime

People in this article

  • Andrej Karpathy →AI researcher and educator
Explore people in AI →
Yash Thakker

Written by

Yash Thakker

Yash is an AI expert with over 300K learners. Join his workshops →

View Yash Thakker in People in AI →

Related posts

Nov 19, 2026

app-store-compliance: An Open Playbook and Guard That Stops Your AI Agent Shipping a Rejected App

A new open repo, mjmirza/app-store-compliance, gained about 540 GitHub stars in a day. It bundles a playbook of 117 documented App Store and Google Play rejection patterns, a guard script that blocks a risky submit, and a skill for AI coding agents. Here is what it does, how to run it in two commands, and what a static scan cannot see.

Nov 13, 2026

Chrome DevTools MCP: Let Your Coding Agent Debug a Live Chrome Browser

Chrome DevTools for agents, shipped as the chrome-devtools-mcp repo, lets a coding agent drive and inspect a live Chrome browser through MCP or a CLI. It trended on GitHub again in week 46 with 55,000 stars. Here is what it does, how to install it, and what to switch off.

Nov 13, 2026

LobeHub: Super-Agents in Your Team Chat, With 85K Stars and a Community License

LobeHub, the open-source chat and agent workspace, now pitches itself as "Super-Agents on Your Team". It trended on GitHub in week 46 with about 1,200 new stars in seven days and 85,000 in total. Here is what it does, how to self-host it, and what its license means.