Design Apply logo

Design Apply

Community
Muluk-m
design-apply

Generate style-consistent frontend code using a saved design system (tokens.json + DESIGN.md) from the global library or a local DESIGN.md. Loads the design system + semantic roles + essence, re-screenshots the source for visual calibration, generates code constrained to the palette, typography, and component language, then runs a closed visual-verification loop (render → measure → iterate) instead of an eyeball checklist.

Overview

PublisherMuluk-m
Repositorydesign-distill
Skill namedesign-apply
Stars
11
Forks
2
Bundled files
Instructions only
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.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by Muluk-m on GitHub. Read the source before you install it.

Installation

Install the Design Apply 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/Muluk-m/design-distill/tree/main/skills/design-apply
  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/Muluk-m/design-distill.git /tmp/design-distill
mkdir -p .claude/skills
cp -r /tmp/design-distill/skills/design-apply .claude/skills/design-apply
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Design Apply 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 Design Apply 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 Design Apply 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.

Design Apply

Step 1: Load Design System

Resolve the design system based on priority:

The library lives at ~/.config/design-distill/ (override with $DESIGN_DISTILL_HOME). Resolve with ordinary file ops — there is no CLI.

Priority 1: Explicit style name from user input

If the user mentions a style name (e.g., "用 linear 做个博客主页"), read its structured set + view:

bash
DIR="${DESIGN_DISTILL_HOME:-$HOME/.config/design-distill}/<name>"
cat "$DIR/tokens.json"   # canonical: semantic roles + essence (preferred)
cat "$DIR/DESIGN.md"     # rendered view

If the directory doesn't exist, list available styles and report:

bash
ls -1 "${DESIGN_DISTILL_HOME:-$HOME/.config/design-distill}"

Priority 2: Local ./DESIGN.md

If no explicit style name but ./DESIGN.md exists in the current project:

bash
cat ./DESIGN.md

This is the project's own design system (e.g., the user is building a product and has documented their own design language).

Priority 3: No style available

If neither an explicit name nor a local DESIGN.md:

bash
ls -1 "${DESIGN_DISTILL_HOME:-$HOME/.config/design-distill}"

Display available styles and ask the user to choose one. If none exist, suggest running design-distill first.


Step 2: Re-Screenshot Source for Visual Calibration

Extract the source_url field from the loaded DESIGN.md.

If source_url is present

Use the bundled screenshot primitive (self-contained — no /browse dependency) to capture the source:

bash
node scripts/screenshot.mjs <source_url> --out <tmpdir> --viewports desktop --schemes light
# For a specific page type (e.g., pricing), re-run with that subpage URL

Load the captured images with the Read tool — look at the reference images before designing.

If node scripts/setup.mjs --probe reports tier: "token-only" (no browser), skip screenshots and design from the DESIGN.md values, clearly noting that visual calibration was skipped.

Why re-screenshot? Text descriptions drift toward stereotypes ("developer tool = dark mode"), while screenshots are facts. Vercel's default is light mode.

If re-screenshot fails (404, timeout, site unreachable)

Fallback to archived screenshots:

Check if archived screenshots exist:

bash
ls ~/.config/design-distill/<name>/screenshots/ 2>/dev/null

If archived screenshots exist:

  • Load them for visual calibration
  • Warn: "Using archived screenshots (source site unavailable)"

If no archived screenshots:

  • Generate based on document values only
  • Warn: "Source site unavailable, no archived screenshots. Generating from document values only."

If no source_url

The design system was extracted from a local project. Generate based on document values only, without screenshots.


Step 3: Style Calibration

Prefer the structured tokens.json (next to DESIGN.md) when present — it carries the semantic roles (color-primary, color-surface, color-text, …) and the essence (the ≤5 defining traits). Generate against the semantic roles, not raw hex guesses.

Looking at the screenshots (if available), tokens.json, and DESIGN.md, identify:

  • Essence first: the ≤5 traits from tokens.json.essence — these define the look; honor them above incidental detail
  • Semantic roles: which color plays primary / surface / text / status — map UI elements to roles, not to arbitrary palette entries
  • Core visual facts: background light or dark? primary font? CTA color?
  • Anti-patterns — what this design would never do (from the document's Anti-patterns section)

Step 4: Generate

Produce frontend code, HTML, or design artifacts based on the user's request.

Constraints:

  • Map elements to semantic roles (primary CTA → color-primary, page bg → color-surface, body → color-text) — colors come from the palette, no new colors
  • Fonts must match the document's font families
  • Reference the screenshot's actual proportions and visual density
  • Component shapes follow the document's component vocabulary and radius-button/radius-card

Step 5: Closed Visual-Verification Loop

Do not rely on an eyeball checklist. Render the generated output, measure it against the saved design system, and iterate on the concrete deltas until it converges.

Render target: a static artifact via its file:// path; a framework project via a local dev server URL.

bash
node scripts/verify.mjs ~/.config/design-distill/<name>/tokens.json <output.html|url> --out <evidence-dir>
# prints { score, pass, totalDeltas, instructions[], evidence }; exit 0 = pass

Loop (bounded + convergent — defaults: threshold 85, cap 3 iterations, stop on no-improvement):

  1. Run verify.mjs. If pass (score ≥ threshold) → done; present the result + evidence.
  2. Else apply the returned instructions (e.g. "color-primary should be #5e6ad2 (got #ff0000)") to the code and re-run.
  3. Stop at the iteration cap or when the score stops improving; present the best round with its remaining deltas.

Degradation: if verify.mjs reports visualVerification: false (no browser), it ran a token-only check and skipped screenshots — tell the user visual verification was skipped rather than implying it passed visually.

Frequently asked questions

What does the Design Apply AI skill do?

Generate style-consistent frontend code using a saved design system (tokens.json + DESIGN.md) from the global library or a local DESIGN.md. Loads the design system + semantic roles + essence, re-screenshots the source for visual calibration, generates code constrained to the palette, typography, and component language, then runs a closed visual-verification loop (render → measure → iterate) instead of an eyeball checklist.

Why use Design Apply on TypingMind?

Because you install it once and use it with any model. Design Apply 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 Design Apply in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/Muluk-m/design-distill/tree/main/skills/design-apply. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Design Apply?

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 Design Apply?

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

Is the Design Apply AI skill free?

Yes. It is published on GitHub by Muluk-m 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 👇