Anatomy Reveal logo

Anatomy Reveal

Community
annnguyentt
anatomy-reveal

Build any physical object (a product, a tool, furniture, an instrument, a shoe, a vehicle, a piece of jewellery, kitchenware, a toy, even a fruit or a flower) as a crafted isometric "anatomy" figure made of many small convex parts, then turn it into (1) an interactive dissection page you can drag, turn and take apart, and (2) a product-reveal film (MP4, 9:16, 4:5, 1:1 or 16:9, 30 fps, motion blur, white paper look, 10 s by default, any length from 3 to 60 s). The film tumbles in, whips to macro shots of the object's details, shows one feature working, explodes into its parts and snaps back together for a hero shot. Works from a photo of the object or from just its name. Use it when someone asks for an exploded view, a teardown or dissection drawing, an anatomy figure, or a product reveal or product motion video of an object.

Overview

Publisherannnguyentt
Repositoryanatomy-reveal-skill
Skill nameanatomy-reveal
Stars
0
Forks
0
Bundled files
34
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 34 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by annnguyentt on GitHub. Read the source before you install it.

Installation

Install the Anatomy Reveal AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/annnguyentt/anatomy-reveal-skill/tree/main
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/annnguyentt/anatomy-reveal-skill.git \
  .claude/skills/anatomy-reveal
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Anatomy Reveal in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Anatomy Reveal on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Anatomy Reveal is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Anatomy Reveal

The object is never a stock 3D model or a traced picture. It is built part by part from its real dimensions, with the Anatomy kit's 3D mode (template/kit):

  • every panel, screw, dial, glass element and board is its own convex solid;
  • each part is shaded in four tones by how it faces the light;
  • parts are painted in an order solved from separating planes.

Because the object is real geometry, the same model turns through any angle, comes apart, and drives both outputs.

template/ is a complete working project. Its worked example is a dual-boiler espresso machine with a built-in grinder (82 parts), but nothing in the pipeline is specific to espresso machines: any object the user names or shows works the same way. Start every job from the template and replace the object.

Read before building:

  • references/inputs.md: getting from a photo or a name to dimensions, features and a parts list.
  • references/objects.md: how any kind of object comes apart, and how to build curved, organic, soft and large shapes from convex parts.
  • references/object.md: how to build the object so it turns through every angle and comes apart cleanly.
  • references/shots.md: the film's beats, the timeline format, eases and framing.
  • references/look.md: the paper look, materials and line weights.

Inputs: two cases

A. The user uploads an image of the object.

  • Study it for silhouette, proportions, materials, colours, surface decoration, visible controls, ports and markings, and anything that moves.
  • If you can tell what product it is, also look up its real dimensions and internals as in case B. Otherwise estimate the size from a feature of known size and say what you assumed.
  • The image is a reference only. Rebuild every shape, colour and ornament as geometry: never trace it, embed it, or map its pixels onto anything.

B. The user only names the object.

  • Research the real product:
    • overall dimensions in millimetres, and the main sub-dimensions;
    • materials and colourways;
    • every visible control, port and marking;
    • how it opens, folds, slides or turns;
    • what is inside (teardowns are the best source).
  • Use web search when you have it. Without it, use what you know and list your assumptions.
  • If the name is ambiguous (a model family, several generations), pick the most common variant and say which.

Either way, finish with a short spec before any geometry: the dimensions, the feature list, the parts list grouped by assembly, and the one or two features that will move in the film. Show it to the user if anything is uncertain.

Workflow

  1. Size it. An object is a "Hero": 100–150 parts, about 2–4 hours for the object and about 1 hour for the film. Tell the user, then build. Have the object rendering on a test page early, and deepen it from there.

  2. Gather the facts for case A or B above (references/inputs.md).

  3. Scaffold: node <skill>/scripts/new-project.mjs <dir>. It copies template/, installs puppeteer-core and builds the example, which proves the toolchain works on this machine.

  4. Build the object in object/geometry.mjs, following the rules in references/object.md:

    • a hollow shell of convex panels;
    • explode assemblies whose outer parts lead;
    • feature groups (hinges, turns, slides);
    • anchors for every shot;
    • markings as face planes.

    Iterate with these, and look at every image you make:

    • node object/test-page.mjs --light;
    • node object/shots.mjs object/test-light.html out/t/a '{"yaw":-35,"pitch":12}' '{"yaw":-35,"pitch":20,"apart":1}';
    • node object/scan-apart.mjs <yaw> <pitch> from five views (front, back, top, below, side), until each says no pairs pass through each other and no forced picks.
  5. Describe the object in object/film.mjs. The figure page and the film read everything object-specific from this one file:

    • FILM: the title, the scene radius (about 0.8 × the largest dimension in mm), an accessible label, the bounding box;
    • FIGURE: the page title, starting view, legend keys and caption (with {parts}, {assemblies} filled in);
    • FEATURE_DEFAULTS and FEATURE_KEYS: each moving feature, its key on the page, its values or step;
    • featureValues and featureTravel: how features map to groups and how far they move a part.
  6. The anatomy page. Run node object/build.mjs and --light to write object/figure.html and object/figure-light.html. Serve the folder over http to look at it in a browser.

  7. Choose the length and the ratio. Ask the user if they have not said (default 10 s, 9:16), then build the film page with node motion/build.mjs --duration <seconds> --ratio <9:16|4:5|1:1|16:9> (3–60 s). Whips keep their length; drifts and holds stretch or compress. For films over about 12 s, add beats (more detail shots) rather than only stretching.

  8. Choreograph motion/timeline.mjs beat by beat (references/shots.md):

    • frame every shot with node motion/render.mjs poses '[…]' out/p and pick by eye;
    • check timing with node motion/render.mjs stills 0,1.2,2.3,3.5,4.6,5.6,7.9,10 out/s;
    • node motion/scan.mjs must report every frame clean.
  9. Render: node motion/render.mjs blur out/blur --step 4 --most 40, then node motion/encode.mjs out/blur out/<name>.mp4.

    • The render sizes each frame's motion blur from how far the picture moves: whips get up to 40 sub-frames, holds 1.
    • It takes about 2 minutes per second of film. Run it in the background and keep other browser work off the machine.
  10. Check the MP4 itself. Extract frames (ffmpeg -i out/<name>.mp4 -vsync 0 out/check/f_%03d.png), contact-sheet a dozen of them, and look.

  11. Deliver: the MP4, the two figure pages and the project folder. Say what you did not verify. The film is silent; offer to cut it to a track the user supplies.

Rules

  • Geometry only. No images from the user, the web or a generator end up in the figure or the film. Decoration such as engraving, filigree or print is drawn: SVG on a face plane, or raised convex pieces.
  • True to the object. Use real millimetres, the real count of buttons and screws, and the real way it opens. Detail comes from real parts, never noise.
  • Convex parts only, never sunk in. The build refuses a pair it cannot separate: move or split the part.
  • One continuous take. Change subject with a whip (about 0.24 s); holds keep drifting; the last half-second is the only stillness.
  • Paper look. The film is drawn on warm white paper in the kit's light theme: ink outlines, four-tone fills. Keep it unless the user asks otherwise.
  • Neutral markings by default. Label parts generically (SoC, crown, battery). Draw a maker's logo or trademark only when the user asks for a faithful replica.

Files

PathWhat
scripts/new-project.mjsCopies the template into a new folder, installs, builds the example.
template/kit/The Anatomy kit (MIT), with extra materials (glass, pcb, sensor, signal, porcelain, lavender, knit, coffee, bean, milk, smoke) and two debug hooks in turn.mjs.
template/object/geometry.mjsThe object: parts, groups, explode assemblies, feature poses, anchors.
template/object/scene.mjsProjection, build, orient() (yaw, pitch, roll in view space), explodeValues().
template/object/film.mjsEverything object-specific the page and the film read: titles, scene radius, labels, bounds, feature channels and keys.
template/object/build.mjs, live.jsThe interactive anatomy page.
template/object/test-page.mjs, shots.mjs, scan-apart.mjs, capture.mjsThe bare test page, posed screenshots, the explode pass-through sweep, page captures.
template/motion/timeline.mjsKeys, eases, labels, the retiming behind --duration and the framing behind --ratio.
template/motion/build.mjs, look.cssThe film page: frame(t), pose(state), speed(t).
template/motion/render.mjs, encode.mjs, scan.mjsStills, poses, the adaptive-blur render; averaging and H.264 encode; the per-frame order check.
template/scripts/browser.mjsFinds Chrome, Chromium or Edge on macOS, Linux or Windows (CHROME_PATH overrides).

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Anatomy Reveal AI skill do?

Build any physical object (a product, a tool, furniture, an instrument, a shoe, a vehicle, a piece of jewellery, kitchenware, a toy, even a fruit or a flower) as a crafted isometric "anatomy" figure made of many small convex parts, then turn it into (1) an interactive dissection page you can drag, turn and take apart, and (2) a product-reveal film (MP4, 9:16, 4:5, 1:1 or 16:9, 30 fps, motion blur, white paper look, 10 s by default, any length from 3 to 60 s). The film tumbles in, whips to macro shots of the object's details, shows one feature working, explodes into its parts and snaps back...

Why use Anatomy Reveal on TypingMind?

Because you install it once and use it with any model. Anatomy Reveal is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Anatomy Reveal in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/annnguyentt/anatomy-reveal-skill/tree/main. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Anatomy Reveal?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Anatomy Reveal?

As many as you like. As long as a model supports skills, you can use Anatomy Reveal with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Anatomy Reveal AI skill free?

Yes. It is published on GitHub by annnguyentt under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇