What it is
Anatomy is an Agent Skill, a folder with a SKILL.md file, that teaches a coding agent to draw explanatory figures. Ask it to explain a concept and it invents a physical object whose working is the concept, for example a pin-tumbler lock for a cache key that only matches when every input matches, then builds it from dozens of small shaded parts in isometric view, inside a card with a title, a hint and a live readout.
Figures come out as standalone HTML, a single SVG file or React/Next.js components, in light and dark themes, with optional WebGL shaders for fire, water and light and a 3D mode in beta. The skill checks its own output by zooming in on joints. Its kit uses only Node's built-in modules, so there are no API keys or packages to install.
It installs with `npx skills add wheresryan22/anatomy -g` and runs in Claude Code, Codex, OpenCode, Antigravity and MiniMax Code. It is MIT-licensed; version 1.1.0 came out on 12 October 2026.
Our review
Anatomy shows what a skill can carry that a prompt cannot. Its SKILL.md fixes a workflow (collect the real numbers, invent the object, write a parts list, build the geometry once, paint back to front, then audit), and a zero-dependency kit does the drawing: an isometric camera, helpers for round parts and pipes, a WebGL layer and React components. What comes back is source you own, a Node build script and the page it writes, not a picture from a model you cannot edit.
Its checking is stricter than most skills attempt. The kit's audit proves that nothing passes through anything and every end sits on a mount; headless Chrome scripts take close-ups and fail any line that ends in mid-air, and the agent is told to read every PNG it makes. That rigour is also the bill. The README calls it slow on purpose: the Raptor engine example takes a few minutes just to build, and a full 3D orbit audit runs 15 to 20 minutes for 30 parts. No token counts per figure are published, so a team has to measure its own.
The range is narrow by design. There are no axes or data series, it will not draw people or exact maps, and an idea with no mechanism to build comes out weaker. The 3D mode is a beta proved on its own example and one test figure; a 380-part stress test ran at about 20 frames a second at 1,440 pixels wide. In Codex the sandbox may stop the headless Chrome checks, which leaves a figure built but unverified until someone approves the command.
The project is also very new. Version 1.0.0 shipped on 10 October 2026 and 1.1.0 two days later, so its 19,000 stars measure attention more than track record. (Disclosure: the explainer figures on our AI stack pages were drawn with its kit.) Against a Mermaid diagram, which takes minutes and explains with labelled boxes, Anatomy asks for a long session and returns one figure whose moving parts do the explaining: a fair trade for a launch-page hero, a poor one for internal docs that change every week.
Where it earns its place
A developer-relations engineer at an API company
Explain rate limiting on the docs landing page with something readers can interact with
Prompt /anatomy with the real bucket size and refill rate; the skill invents the object, builds it with the kit and writes a standalone HTML page to commit. Expect one long agent session per figure on an existing Claude Code or Codex plan; the licence costs nothing.
A front-end engineer maintaining a Next.js documentation site
Add an interactive figure without pulling in a charting or animation library
Ask for React output: geometry modules, a server component for the static art and one client component for motion. Only React 18+ is required, and under reduced motion the figure goes still, so accessibility settings are respected.
A content lead at a 30-person infrastructure start-up
Get a distinctive hero figure for a launch post without commissioning an illustrator
One engineer runs the skill and iterates in conversation, so the cost is agent usage and that engineer's session rather than an illustrator's fee. Figures can also be printed as A4 posters. The risk is a concept with no physical mechanism, which the README says comes out weaker.
Before you commit
- Setup
- Install with `npx skills add wheresryan22/anatomy -g` (version 1.7.1 of the skills CLI wants Node 22.20+) or a git clone into the agent's skills folder; it copies about 10 MB, examples included. The kit needs Node 20.10+, and Chrome or Chromium for the self-checks.
- Integrations
- Claude Code, Codex (invoked as $anatomy), OpenCode, Antigravity and MiniMax Code through the Agent Skills format. Output is standalone HTML, a single SVG, React/Next.js components, or A4 PDFs via scripts/print.mjs.
- Data and privacy
- No API keys; the kit and checks run on your machine, and prompts go only to the model your agent already uses. Publishing to the community gallery is opt-in and opens a GitHub pull request only after you confirm what will be sent.
- Cost at scale
- Free under MIT. The real cost is agent usage: every figure is a long session of building, auditing and screenshotting in both themes, and the project publishes no token figures, so budget from a trial figure on your own plan.
- Limits
- Not a chart library; no people or exact maps. Shaders need WebGL (the SVG still reads without it), the 3D mode is beta, and a full 3D orbit audit takes 15 to 20 minutes for 30 parts.
What's good
- Audits its own geometry and screenshots
- Ships editable source: HTML, SVG or React
- No npm dependencies; Node 20.10+ only
- Runs in five coding agents via Agent Skills
What's not
- Long sessions; no per-figure token cost published
- No charts, people or maps
- 3D mode beta, about 20 fps on a 380-part figure
- Codex sandbox can block the Chrome checks
Best for
Technical writers and developers who want interactive explanatory figures for docs, posts and landing pages.
Alternatives
Anatomy: common questions
- Is Anatomy worth it?
- A skill that turns a concept into a machine readers will stop and play with, and charges for it in long, token-hungry sessions. We rate it 4.0 out of 5.
- How much does Anatomy cost?
- Free and open source under MIT. Runs on the agent subscription or API you already use.
- What are the downsides of Anatomy?
- Long sessions; no per-figure token cost published; no charts, people or maps; 3D mode beta, about 20 fps on a 380-part figure; codex sandbox can block the Chrome checks.
- Who is Anatomy best for?
- Technical writers and developers who want interactive explanatory figures for docs, posts and landing pages.
- How much does Anatomy cost for a team or in production?
- Free under MIT. The real cost is agent usage: every figure is a long session of building, auditing and screenshotting in both themes, and the project publishes no token figures, so budget from a trial figure on your own plan.
- Is Anatomy safe for company data?
- No API keys; the kit and checks run on your machine, and prompts go only to the model your agent already uses. Publishing to the community gallery is opt-in and opens a GitHub pull request only after you confirm what will be sent.
- What are the best Anatomy alternatives?
- The closest alternatives we have reviewed are Anthropic Skills.
Answers checked Oct 22, 2026.
Make Anatomy? Show this rating on your site
