Schematic Figure logo

Schematic Figure

Organization
lgwrightlab
schematic-figure

Draw a publication-quality scientific schematic, slide or poster figure (setup, concept, mechanism or overview figure, Nature style) as vector SVG/PDF/PNG from code, optionally with Blender-rendered 3D elements under vector labels. Use when asked to draw, redraw, replicate, revise or design a schematic or concept figure for a paper, talk or proposal, from a reference figure, a written description, or a research note. Options (all optional) - mode, style, route (flat, vector3d, blender), depth (draft, standard, full), web (on, off).

Overview

Publisherlgwrightlab
Repositoryschematic-figure
Skill nameschematic-figure
Stars
116
Forks
15
Bundled files
74
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.

  • 74 bundled files

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

  • Open source

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

Installation

Install the Schematic Figure 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/lgwrightlab/schematic-figure/tree/main/.claude/skills/schematic-figure
  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/lgwrightlab/schematic-figure.git /tmp/schematic-figure
mkdir -p .claude/skills
cp -r /tmp/schematic-figure/.claude/skills/schematic-figure .claude/skills/schematic-figure
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Schematic Figure 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 Schematic Figure 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 Schematic Figure 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.

Schematic figure

A schematic here is a Nature-style concept or setup figure: flat vector art, sometimes with 3D hero elements, 5-7 pt Arial labels, thin leader lines, one idea per panel. Code draws every mark (figure/figure.js, run in headless Chrome). The same source gives a vector PDF with live text, an SVG and a PNG, and the tools can measure it. For a patent drawing sheet, use the patent-drawing skill.

What our development benchmark measured: the skill drew the main figure of 28 published papers. A blind judge model rated how much each figure looks published, from 0 to 100. The skill's lead over a plain model appears when the agent must decide what the figure shows. The full skill scored 36.6. Removing the design process (brief, concepts, spec) cost the most, 6.8 points. Two drafts, the critic, the reader test and the gate did not move the judge's score. The critic removed most of the errors it named. So spend the effort on the plan. Use the checks where an error is costly.

Options

The user may give any of these. Infer the rest (step 0).

OptionValuesDefault
modedesign, replicate, describe, revisefrom the request: a reference image is replicate, a description with no image is describe, an existing figure with changes is revise, else design
stylea style the user names, or a reference figure the user givesyour own choice for the purpose and the field
routeflat (2D, illustrated finish), vector3d (library 3D parts, L.lightScene3), blender (illustration-mode or photographic render; "photoreal" or "photographic" means the photographic one, step 0)for a physical object, the route that suits the hero; say why in the Assumptions line (step 0)
depthdraft, standard, full (table in step 3)full (steps 1-11 of reference/process.md: two drafts and a choice, your own critique pass, your own cold read, the gate), also for a one-line request. Standard or draft only when the user names it
reviewown, independentown: you do every check yourself, in one agent. independent: an independent critic (step 9) and blind readers (step 10) run as separate agents. Only when the user asks for it, for example "with an independent physics check"
webon, offon: search the web freely for anything that helps the figure (other figures, layouts, styles, reference images, background on the subject), in any field. Never look for the figures of the paper you draw for. off: no web access, as in the benchmark

0. Intake (every run)

  1. A new design starts from the brief alone: do not copy the look of the project's other figures unless the user asks. In revise mode, or when the user asks to make a set of figures consistent, find the project's existing figures first and inherit their colour key, notation and symbols (step 1 of reference/process.md).
  2. Infer every option the user did not give, and ask no question. When the route is not set for a figure with a physical object, pick the route that suits the hero. When the user named no style, choose the look yourself for the purpose and the field.
  3. Photographic words. When the user asks for "photoreal", "photographic", "realistic glass" or a similar look, use the Blender route with the photographic 3D look (the kit style render, reference/process.md step 3). Write 3d=render in the Assumptions line. Without such a word, choose as before: an illustration render matches flat panels best.
  4. Assumptions line first. Your first message to the user states what you picked, before any tool call, file write or drawing work. It is one line with the reason for each choice, for example Assumptions: mode=design, style=own (clean flat, physics readers), route=blender, 3d=illustration, depth=full, web=on, width=183 mm. Repeat the same line, updated if a choice changed, at the top of the final report (step 11 of reference/process.md).

1. Plan (every run, most of the effort)

  1. Brief → docs/brief.md: the one sentence a reader must take away, the key ideas (the user's own criteria first, verbatim), what the figure must not imply, its scope, the format and the width.
  2. Structure → the top of docs/spec.md: name the figure's skeleton before any concept. When the content is one principle with several cases, lead with one unifying schematic of the principle. Then give one sub-schematic per case. Each sub-schematic reuses the principle's glyphs and changes only what the case changes. Link them visibly: the same glyph, an inset, a shared colour or an arrow from the principle. Other skeletons are a pipeline, a controlled comparison, a setup with its signal, or a zoom from system to part. Write which one, and why.
  3. Facts → docs/facts.md: before you draw any physical device, write its geometry record. Give the element order along each path, each element's orientation and angle, the beam or flow paths, what sits in front of what, and the source of each item. Mark an item that no source gives as unknown. Ask about it, or keep that part of the drawing vague. Do not draw a device before its record exists.
  4. Concepts and spec: write two or three concepts (one or two at draft depth), score them as step 2 of reference/process.md says, and keep one (two at full depth, for the two drafts). Write the spec card (templates/spec-card.md) and the source of every element (step 3), with the Hero: line and its route.

2. Build

Follow steps 3-8 of reference/process.md: sources, look-up of real objects (standard and full depth), text, the layout at thumbnail size, the panel-by-panel build with render, look and fix, and the gate. Read reference/lib-api.md before you write code, reference/style.md before you choose sizes or colours, and reference/principles.md (P1-P31) before the plan.

3. Check (by depth)

DepthChecks
draftThe gate (node tools/check.mjs) with no FAIL.
standardThe gate, then the faithfulness pass of step 9 (every count, polarity, direction, order and connection against the facts), then your own cold read of the thumbnail.
fullThe standard checks, then two drafts and a choice (step 6), your own critique pass for physics and claim errors (step 9), and your own reader test (step 10).

With review=independent, the critic of step 9 and the readers of step 10 are independent agents. When you cannot start one, say so in your report, so that the caller can run it.

4. Deliver

Step 11 of reference/process.md: the final gate, the full render, the caption and the report. The report starts with the Assumptions: line.

Modes

ModeYou getYou deliver
designa research note, abstract or paper text, and a purposea figure idea (what to show and how) and the figure
replicatea reference figure (image)your own drawing of the same figure: same content, layout, style, size
describea written description (the reference is sealed from you)the figure the description specifies
revisean existing figure (yours or a previous version) and requested changesthe same figure with exactly those changes

The rules of each mode (what you may look at, the revise-scope diff) are in reference/modes.md. Read the file for your mode before step 1.

Files

FileWhat it holds
reference/process.mdSteps 1-11 in full, the project layout, and every tool with its options.
reference/modes.mdThe rules of the replicate, describe, design and revise modes.
reference/principles.mdP1-P31, each with its reason and its exceptions.
reference/style.mdFormats, sizes, calibrations and the rubric.
reference/lib-api.md, reference/lib-<domain>.mdThe drawing library and the domain parts.
templates/spec-card.mdThe spec card.
foundation/The engine, the tools, the Blender kits and the demos.

Bundled files

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

and 14 more files.

Frequently asked questions

What does the Schematic Figure AI skill do?

Draw a publication-quality scientific schematic, slide or poster figure (setup, concept, mechanism or overview figure, Nature style) as vector SVG/PDF/PNG from code, optionally with Blender-rendered 3D elements under vector labels. Use when asked to draw, redraw, replicate, revise or design a schematic or concept figure for a paper, talk or proposal, from a reference figure, a written description, or a research note. Options (all optional) - mode, style, route (flat, vector3d, blender), depth (draft, standard, full), web (on, off).

Why use Schematic Figure on TypingMind?

Because you install it once and use it with any model. Schematic Figure 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 Schematic Figure in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/lgwrightlab/schematic-figure/tree/main/.claude/skills/schematic-figure. 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 Schematic Figure?

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 Schematic Figure?

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

Is the Schematic Figure AI skill free?

Yes. It is published on GitHub by lgwrightlab 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 👇