Anatomy: A Claude AI Skill That Draws Any Concept as a Working 3D Machine

Anatomy is a new skill for Claude AI that turns abstract concepts into interactive 3D machines. Developed by Ryan, this skill allows users to describe an idea in a single line, and Claude invents a machine that visually represents the concept with moving parts. The resulting models are live isometric figures that users can turn, press, and read to understand how the idea works. Every figure is made of dozens to hundreds of small solid parts, and every number it shows comes from the same model that moves them.

The Core Rule: The Object Is the Explanation

Anatomy operates on a simple rule: the object is the explanation. Instead of using boxes and arrows, Claude creates a real, well-made thing whose working is the concept. Users describe an idea, and Claude invents a machine with dozens to hundreds of small solid parts that behave according to the described concept. For example, describing how a mechanical watch keeps time results in an interactive model where users can drag the crown to wind it, press S to slow time to a twentieth or a hundredth, and watch the pallet fork let the fifteen-tooth escape wheel through half a tooth at a time. Pressing E lifts the bridges apart to see the train underneath, with wheels cut to true ratios: the fourth wheel turns once a minute and the centre wheel once an hour.

Examples of Anatomy in Action

Anatomy has been used to create a variety of interactive models. A queue is represented as a marble run: two hoppers feed one sloping lane, and the marbles wait in order behind an escapement until a timer drum takes them. A cache that forgets its oldest entry is depicted as a bookcase with four places on top, most recently used at the left. More complex ideas, like a bloom filter, are represented as a fruit machine whose three reels are the three hash functions, and it only pays out when all three of a word's flags are up. Each model is created from a single line naming the idea, and Claude handles the rest. There are 33 models so far, all live in the collection.

How the Skill Works: A Fixed Order of Steps

Between the user's line and the finished figure, Claude works through a fixed order. Most of the quality is decided in the first four steps, before any geometry exists. The steps are: size it (Figure, Hero, or Epic), truth first (collect real numbers and behavior), invent the object (a machine whose mechanism maps one-to-one onto the idea), list the parts (at least forty, grouped into base, structure, mechanism, controls, and the one lit subject), plan the world, build the geometry once, paint back to front, frame it, make it live, and verify like a critic. Step two ensures that the parts, the readout, the screen-reader text, and the caption all read from the same small model, so none of them can disagree.

Three Sizes of Figures

The first step sets the budget. Every request lands in one of three sizes: a Figure is built without asking, a Hero is announced first then built, and an Epic is never started without the user — Claude gives the estimate, proposes a Hero that keeps the core of the idea, and lists what the full version would add.

How to Try Anatomy Today

To try Anatomy, open Claude AI and use the skill by describing an idea in a single line. For example, you can ask Claude to explain how an LRU cache evicts entries. Claude will then create an interactive 3D model that demonstrates the concept. The skill is designed to be intuitive — you almost never need to say what to draw. Claude handles the invention and visualization process. The skill writes framework-free TypeScript that outputs SVG and HTML, or React components for a Next.js site, in a dark or a light theme.