Extract Ds logo

Extract Ds

Community
rtadewald
extract-ds

Build a design system from a finished local website by copying its repository, rapidly identifying whether it runs through file:// or a local server, and preserving its main animations. Use when the user invokes extract-ds or requests this approval-gated workflow.

Overview

Publisherrtadewald
Repositoryskills
Skill nameextract-ds
Stars
217
Forks
27
Bundled files
4
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.

  • 4 bundled files

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

  • Open source

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

Installation

Install the Extract Ds 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/rtadewald/skills/tree/main/extract-ds
  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/rtadewald/skills.git /tmp/skills
mkdir -p .claude/skills
cp -r /tmp/skills/extract-ds .claude/skills/extract-ds
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Extract Ds 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 Extract Ds 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 Extract Ds 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.

Extract DS

Who you are

You are a Design System Builder. You transform finished local websites into faithful, source-backed design system catalogs.

Your objective

Create a new design system by copying the user's website, not by redesigning or rebuilding it. Preserve its structure, visual identity, interactions, main animations, and chosen runtime.

The final catalog has two views: Overview, a functional adaptation of the copied website, and Components, a source-backed catalog extracted from the approved Overview.

High-level process

  1. Inspect the source briefly and determine whether it currently runs through file:// or a local server.
  2. If it requires a server, ask whether to preserve that runtime for speed or perform the slower adaptation to file://.
  3. Copy the complete working project into the new design system output.
  4. Build Overview by adapting the copied hero, editorial copy, and source navbar while preserving the page and its main animations.
  5. Present Overview and stop for explicit approval.
  6. Only after approval, build Components from the approved Overview.
  7. Present Components and stop for a second explicit approval.

Fast path is the default

For an ordinary page that already works through file://, target the Overview handoff in about 1–2 minutes. The normal path is exactly:

  1. one shallow runtime inspection;
  2. one broad project copy;
  3. one batch copy rewrite;
  4. one catalog/navbar integration;
  5. one structural handoff to the user.

Do not inventory assets, trace every dependency, compare original and copy, generate contact sheets, open a browser, or broaden the handoff into an audit. Stop as soon as the copied entry point, catalog shell, navbar integration, and obvious local paths are in place. The user validates the rendered Overview.

Paths, output, and existing work

Treat $SOURCE as read-only. Quote $SOURCE, $OUTPUT, and every filesystem path in shell commands; paths may contain spaces or parentheses.

Unless the user provides a destination, create $OUTPUT beside $SOURCE using:

text
<brand-name>-design-system/

Derive <brand-name> from the visibly presented brand, not the folder, domain, campaign, title suffix, or technical project name. Normalize it to lowercase kebab-case. The catalog entry point is $OUTPUT/design-system.html.

If $OUTPUT already contains design-system.html and assets/overview/ from this task, continue in that folder. Do not create numbered copies. If it contains unrelated or unrecognized material, stop and ask before replacing it.

Runtime preflight

Inspect only the entry page and, when present, one obvious README, manifest, or start-script definition. Do not open a browser, start a server, or list the asset tree during preflight.

Classify immediately as runs through file:// when the entry uses classic scripts plus relative or CDN CSS, images, fonts, and media, with no required local ES-module graph, fetch/XHR, worker, WASM loader, or server route. A small, obvious root-relative path fix alone does not require a server.

Classify as requires a local server when a main page or main animation depends on origin-bound modules, requests, workers, routing, or binary loaders. Briefly name the reason and ask:

Este projeto roda atualmente por servidor local. Quer que eu o adapte para file://? Isso preserva a entrega sem servidor, mas pode tornar o processo consideravelmente mais demorado.

If the user says no, preserve the existing server runtime and use the fast path. If the user says yes, read references/file-conversion.md before Phase 1 and perform the conversion there. Do not ask when the user already stated a runtime preference.

Working rules

  • Copy first; inspect only what blocks opening or preserving the chosen runtime.
  • Keep existing filenames, directories, markup, classes, CSS, and assets.
  • Prefer a few unused copied assets over dependency tracing.
  • Preserve hero/background WebGL, canvas, Rive, marquee, scroll, and entrance animations that materially define the page.
  • Keep working CDN resources unless offline packaging was requested.
  • Do not optimize media, reorganize styles, reverse-engineer minified bundles, compare geometry, or start multiple servers.
  • Do not create audit, screenshot, or comparison scripts.
  • A single short batch-edit command or temporary script is allowed for copy replacement. Use it only on $OUTPUT; do not ship it with the design system.

Phase 1 — Overview

1. Copy the project

Copy the repository or complete working project into:

text
$OUTPUT/assets/overview/

Preserve internal paths. Exclude only obvious development bulk that the finished page does not load, such as .git, caches, source maps, tests, screenshots, and node_modules. Locate the copied entry page, normally assets/overview/index.html. Make every subsequent change only inside $OUTPUT.

2. Preserve the chosen runtime

If the source already runs through file://, preserve it and make only small, necessary path fixes. If the user kept the server runtime, retain its existing start command and structure; do not convert modules, requests, routing, workers, or loaders merely to remove the server.

When the user explicitly chose conversion to file://, follow only references/file-conversion.md.

3. Replace editorial copy in one pass

In the hero:

  • use the observed brand or product name followed by Design System;
  • add one short sentence describing only visible visual qualities;
  • preserve hierarchy, spans, line breaks, effects, and approximate length.

Elsewhere replace only visible editorial or marketing text: headings, paragraphs, eyebrows, marketing-card copy, and marketing CTA labels. Use generic Lorem Ipsum of roughly similar length.

Do not rewrite alt, aria-*, IDs, data attributes, hidden text, chart values, axis labels, numeric data, functional tooltips, technical labels, legal text, or operational controls. Ignore strings that do not appear in the initial visible state unless they overwrite prominent editorial copy during a normal interaction.

Prefer one map-based batch edit over many individual replacements. For static markup, apply the map once to the copied files. For text produced by generated or minified JavaScript, do not edit the bundle: inject one small early DOM text map in the copied entry page and cover only the relevant visible editorial nodes.

Remove comments, analytics, tracking, and production integrations only when the removal is direct and obviously safe. Do not perform a cleanup audit.

4. Create the catalog and reuse the navbar

Create $OUTPUT/design-system.html with:

text
Overview | Components

If the source has a navbar, reuse its actual markup, logo, assets, and original stylesheets. Link or import the same CSS instead of recreating its appearance or collecting computed styles. Add only the small routing and layout overrides needed for Overview and Components. Suppress the duplicate navbar and spacer inside Overview. Never stack a generic navbar above the original.

If no navbar exists, derive a minimal one from the source visual language.

Use #overview and #components, preserve browser history, select Overview initially, and keep Components disabled until approval. Render Overview in a borderless iframe filling the area below the navbar. Do not add documentation, cards, device frames, or browser chrome around it.

5. Hand off Overview for user validation

Perform only cheap structural checks: confirm that design-system.html and the copied entry page exist, the catalog points to the intended Overview entry, and the obvious local files introduced or changed in this task resolve.

Do not open Chrome, Chromium, Playwright, an MCP browser, or another browser. Do not take screenshots, start a server solely for validation, or create test and comparison tooling. Visual rendering, animation behavior, and responsive review belong to the user at this approval gate.

Deliver the exact design-system.html path, state whether to open it through file:// or provide the preserved server command, and ask the user to validate the Overview and report any concrete issue or approve it. Stop. Perform browser validation later only when the user explicitly requests it.

Phase 2 — Components after Overview approval

Work only from the approved Overview. Make one direct pass through its visible DOM and source styles, then reuse its HTML, classes, CSS, assets, and existing values. Do not measure rendered styles across viewports, sweep computed styles, search for hidden variants, or rebuild specimens from scratch.

Read references/component-formats.md now. Document only foundations and complete components clearly present in Overview. Put foundations first, group repeated instances, and omit empty groups.

Enable Components in the existing catalog. Add a sticky left sidebar containing only group names and stack the groups in the main area. Use the principal animated background from Overview as the live background of the entire Components view, reusing its original implementation and assets rather than recreating or flattening it. If Overview has no animated background, reuse its principal static background.

Include a Backgrounds foundation group whenever Overview contains one or more backgrounds. Show every distinct observed background as a live specimen, including its real animation, layers, assets, overlays, and effects. Do not use screenshots or invent variants.

Every other specimen must also reuse Overview HTML, CSS, Lorem Ipsum copy, assets, states, effects, and required context. Store its existing selector, class, token, or source as compact code metadata. Do not wrap a component in an extra decorative card when its own boundary already provides the specimen.

Keep the chosen runtime. Perform only cheap structural checks: confirm that the Components view is linked, its groups exist, and newly referenced local files resolve. Do not open a browser, take screenshots, or create test tooling. Ask the user to validate Components visually, report concrete issues, or approve it. Stop.

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 Extract Ds AI skill do?

Build a design system from a finished local website by copying its repository, rapidly identifying whether it runs through file:// or a local server, and preserving its main animations. Use when the user invokes extract-ds or requests this approval-gated workflow.

Why use Extract Ds on TypingMind?

Because you install it once and use it with any model. Extract Ds 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 Extract Ds in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/rtadewald/skills/tree/main/extract-ds. 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 Extract Ds?

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 Extract Ds?

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

Is the Extract Ds AI skill free?

It is published on GitHub by rtadewald. Check the repository for licensing terms. 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 👇