
exploded
Take a real UI component apart. A Claude Code skill + runtime that explodes a live component into paint layers, states and tokens. Nothing is redrawn, no num…
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up exploded (claude-skill project) into my current project. Found on https://claudeers.com/exploded Repo: https://github.com/diip3sh/exploded Homepage/docs: https://diip3sh.github.io/exploded Detected install method: claude-skill → # copy this skill into .claude/skills/exploded/ Category: skills. Platforms: cli, api, web. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: active; community-verified: false. Confirm the source before running anything.
# copy the skill dir into your project: # .claude/skills/exploded/ (or ~/.claude/skills/exploded/ for all projects)
git clone https://github.com/diip3sh/exploded
// compatibility
| Platforms | cli, api, web |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | MIT |
| Pricing | open-source |
| Language | JavaScript |
Exploded
Take a real UI component apart. A skill for Claude Code, plus the runtime it uses.
Name a component and Claude renders it, then pulls it apart in isometric into its paint layers: elevation, surface, stroke, inner surfaces, content and focus ring. You can switch its states and read every measurement and design token right off the layers.
The rule
Nothing is redrawn, and no number is typed.
- Every layer is the live component, cloned with the browser's own computed styles. Any paint that belongs to another layer is made transparent.
- Every number comes from
getComputedStyle, layout boxes and text ranges. - Every token name comes from the CSS rule the author actually wrote.
So the spec sheet can't drift from what ships.
What you get
- Layers, found automatically. Elevation, Surface, Stroke, inner surfaces (chips, thumbs, avatars), Content and Focus ring. Light paint (like white text) gets a backing in the colour of the surface it really sits on.
- Every state. Hover, focus, pressed, disabled, invalid, on/off, or your own.
:hover/:focus/:activeare forced by rewriting CSS rules, so no pointer is needed. The stack combines every state, so the focus ring layer waits, dashed and empty, at rest. - Redlines. Padding, radius, gap, type and shadow are printed on the layer you hover. Labels stay a fixed size on screen and never cover the glyphs they measure.
- Tokens. Press
Tand every value becomes the token the author wrote:--radius-md,--spacing × 2.5. Hard-coded values stay raw, so drift is obvious. - The live component sits beside it. Hover or press the real component and the figure follows.
- An audit. It checks that every paint lands on exactly one layer, and that every layer's copy lays out within 0.5 px of the live component, in every state.
Install the skill
npx skills add diip3sh/exploded
It works with Claude Code, Cursor, Codex and any other agent that reads skills. Or clone it yourself:
git clone https://github.com/diip3sh/exploded ~/.claude/skills/exploded
Restart your agent, then:
/exploded document our Button, all states, for the Next.js docs
/exploded spec sheet for the email field and the switch, dark theme
You don't need the command. Claude uses the skill on its own for component docs, spec sheets, redlines and anatomy diagrams. SKILL.md has the whole workflow.
Use the runtime directly
<link rel="stylesheet" href="runtime/exploded.css">
<script src="runtime/exploded.js"></script>
<script>
document.fonts.ready.then(() =>
Exploded.mount(document.querySelector("#fig"), {
target: document.querySelector(".btn"),
title: "Button · primary",
states: "pressable", // or "field", or your own
parts: { ".btn-kbd": "Shortcut chip" },
theme: "dark",
}));
</script>
React / Next.js:
import { Exploded } from "@/components/exploded/react/Exploded";
<Exploded fig={1} title="Button · primary" states="pressable">
<Button>Deploy</Button>
</Exploded>
Keys: 1–9 states · E apart · ↑ ↓ layers · ← → turn · T tokens · Esc reset. Hover opens it and dragging turns it.
Audit
node scripts/audit.mjs examples/gallery/index.html --out shots
This runs headless Chrome with no dependencies. It prints the layers for each state, runs the audit, and saves PNGs of each figure apart and together. Set CHROME_PATH if Chrome isn't in a standard place.
Try it locally
python3 -m http.server 8731
open http://127.0.0.1:8731/
Limits
- CSS from another origin can't be read, so its hover/focus rules can't be forced and its tokens can't be traced. Serve CSS from the same origin.
- Shadow DOM, canvas and WebGL components aren't captured. A canvas clones without its pixels.
- Transitions are frozen at their end state. The figure shows states, not motion.
- JS-driven states (spring animations) need an
applystate for now.
Credit
Inspired by Anatomy by Ryan. Same rule: the thing itself is the explanation.
MIT © diip3sh
// faq
What is exploded?
Take a real UI component apart. A Claude Code skill + runtime that explodes a live component into paint layers, states and tokens. Nothing is redrawn, no number is typed.. It is open-source on GitHub.
Is exploded free to use?
exploded is open-source under the MIT license, so it is free to use.
What category does exploded belong to?
exploded is listed under skills in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/exploded)
// retro hit counter
[](https://claudeers.com/exploded)
// reviews
// guestbook
// related in Claude Skills
An agentic skills framework & software development methodology that works.
AI coding assistant skill (Claude Code, Codex, OpenCode, Cursor, Gemini CLI, and more). Turn any folder of code, SQL schemas, R scripts, shell scripts, docs,…
💫 Toolkit to help you get started with Spec-Driven Development
