Anatomy is a single skill that has an AI coding agent explain a technical idea by inventing a physical machine whose working is the concept. The agent then builds that machine as a detailed, interactive isometric SVG figure, adding WebGL or 3D where needed. It is for people who want explanatory figures for docs, blog posts or landing pages, and it runs in Claude Code and several other agents.
Why read it
Before doing any work, the agent names the job as a Figure, Hero or Epic. Each tier has a part count and a time estimate. An Epic is not started: the agent gives the estimate and offers a smaller Hero version instead. The agent also runs `date` at each step and tells the user when a step runs past twice its share of the estimate. This keeps long agent sessions bounded and the user informed.
The agent collects the real numbers first and writes them as a small pure module. The readout, the motion and the caption are all driven from that one module, so whatever the figure displays cannot drift away from the model behind it.
The workflow requires a concept sentence and a grouped parts list of at least 40 items before any geometry is written. The parts are grouped into base, structure, mechanism, controls and subject. Quality gets decided at the planning stage, where it is cheap to change.
Taste is turned into numbers the agent can check: bevel sizes, how far grooves are inset, screw dot sizes, tick spacing, a tone hierarchy with exactly one lit accent, and the share of the card the figure should fill. Exact figures leave the agent less to guess than adjectives would.
The agent must read every PNG it captures. It takes 4× close-ups of every junction in both themes, and it compares its result with the shipped examples using fixed questions. Self-review becomes a list of concrete steps rather than a vague request to "check your work".
The skill names the cheats an agent might use to get past its own audit and forbids them. A stand-in shape that quiets the audit counts as a bug, and an interlock has to be fixed in the geometry, not with a depth bias. This blocks gaming of the agent's own checks.
Eight examples ship with the skill, each described by what it is the reference for. One is the bar for drawing quality and another is the bar for shader work. The agent is told to open the closest example before starting, which grounds the standard in real files.
The WebGL rules are tied to specific past failures, such as a colour change between phases that looked cut and the Raptor's "red ink smoke". The full lessons go to reference files that the agent loads only when it needs them, which keeps SKILL.md short.
A model read each skill and found no calls or hand-offs between them.
| SKILL | KIND | STARTED BY | DESCRIPTION | LINKS | LICENCE |
|---|---|---|---|---|---|
| anatomy | skill | You or the agent | Build crafted, highly detailed, interactive isometric SVG figures that explain a technical idea as a real physical object (a test rig, a dial gauge, a cabinet of drawers, a plotter, a lock), with many small precise solid parts, four-tone shading, correct painter's order, a framed card with title, hint and live readout, and calm eased motion. Use this whenever someone wants an isometric illustration, an explanatory diagram or figure for docs, a blog post, a landing page or an article, a "Linear-style" or "premium" technical drawing, an interactive explainer, or an SVG that shows how a system, algorithm or component works, even if they never say "isometric". Works framework-free (TypeScript/JS that writes SVG and HTML) and in React/Next.js, and can add realistic WebGL shaders (fire, exhaust, water, caustics, lamp light) that live inside the drawing in the same isometric perspective. Also prints a figure as an A4 PDF (poster, drawing or handout) and publishes it to the Anatomy community gallery when asked. | 0 | MIT |
From wheresryan22/anatomy at 1482e84, 2026-10-18. Skills read by DeepSeek V4.1 Flash on 2026-11-04. Summary and techniques by Claude Opus 5.5 on 2026-11-06.