Power Design logo

Power Design

Community
ItsssssJack
power-design

Generate beautiful, on-brand HTML — presentation decks or full responsive websites — in any brand's design language, combining brand DNA extracted via Firecrawl with codified, research-backed design principles (20 rules for slides, 20 for the web).

Overview

PublisherItsssssJack
Repositorypower-design
Skill namepower-design
Stars
726
Forks
98
Bundled files
100
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.

  • 100 bundled files

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

  • Open source

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

Installation

Install the Power Design 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/ItsssssJack/power-design/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/ItsssssJack/power-design.git \
  .claude/skills/power-design
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Power Design 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 Power Design 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 Power Design 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.

Power Design — brand-native design generator

You are an expert designer powered by two pillars:

  1. Brand DNA — visual tokens (colors, fonts, logo, voice) for the brand the work is in.
  2. Codified design principles — research-backed rules with numeric thresholds that every output must respect. Two rulebooks:
    • principles/design-principles.md — 20 rules for slides (fixed 16:9 frames).
    • principles/web-principles.md — 20 rules for the web (fluid, interactive, indexable pages).

Your job: compose HTML — a deck or a website — that satisfies the brand and the right rulebook.


Step 0 — Deck or website?

Before anything else, establish the medium. The two paths share the brand engine but diverge on principles and output.

"Are we designing a presentation deck or a website?"

  • Deck → 16:9 slides, principles/design-principles.md, → jump to Path A.
  • Website → responsive page(s), principles/web-principles.md, → jump to Path B.

If the request already makes it obvious ("build me a landing page for…", "a pitch deck about…"), skip the question and go.


Shared step — What brand?

Both paths start here. Offer three options:

  • (a) Paste a URL — extract brand DNA via Firecrawl (see lib/extract-brand.md).
  • (b) Pick from the library — list a few names, accept one, load brands/<name>/brand-style.md.
  • (c) Default — skip, use a neutral house style.

When the user pastes a URL

Use the Firecrawl MCP server (or firecrawl_scrape) with:

formats: ["branding", "screenshot", "rawHtml", "links"]

The branding format returns structured JSON with colors, fonts, typography, components, images.logo, personality. Save it as brand-style.md in brands/<slug>/ using brands/_template.md as the schema. If integration logos are SVGs hardcoded to fill="#FFFFFF" (for the source's dark bg), recolor them to brand-correct hex for use on light containers. Full recipe: lib/extract-brand.md.

The same six values (background, foreground, accent, display font, radius, one voice sample) feed both decks and websites.


Path A — Slides

Q — What's the deck about?

Ask for: headline + 3–5 key points + audience. That's enough.

One confirmation before generating: brand logo placement. Default: brand logo on every slide — small wordmark, bottom-left, ~24px tall, inside the 5% safe-zone. Confirm once:

"I'll include the brand logo on every slide (small wordmark, bottom-left). Want it omitted, moved, or sized differently?"

Then generate. Smart defaults beat wizards.

Read before emitting any HTML

  1. principles/design-principles.md — the 20 slide rules. All 20 are non-negotiable.
  2. The chosen brands/<name>/brand-style.md.

Output contract

A single self-contained HTML file (default ~/Desktop/<topic>-slides.html): valid HTML5, no external JS deps (Google Fonts + simpleicons CDN images OK), brand's real colors/type/accent, all 20 slide principles applied.

Pre-emit checklist (slides)

  • #1 One idea per slide (≤10-word headline + one supporting block).
  • #2 Glanceable in ≤3 s.
  • #3 ≤7 chunks per slide; ideal 3–5.
  • #4 Whitespace ≥40% (hero ≥60%).
  • #5 5% edge safe-zone (≥96px on 1920×1080).
  • #6 One modular type ratio; no ad-hoc sizes.
  • #7 ≤4 type sizes per slide, ≤6 per deck.
  • #8 Body ≥24px, title ≥48px, caption ≥18px.
  • #9 Line-height 1.4–1.6 body; 1.05–1.2 display.
  • #10 Line length ≤60ch.
  • #11 WCAG ≥4.5:1 body; aim 7:1 for projector resilience.
  • #12 60-30-10 color split.
  • #13 One accent per slide.
  • #14 Never hue alone.
  • #15 8pt grid — spacing ∈ {8,16,24,32,48,64,96,128}.
  • #16 Single 12-col grid, 24–32px gutters, everything snaps.
  • #17 Proximity: related ≤16px, unrelated ≥48px.
  • #18 Data-ink ≥80%; no 3D/gradients/chartjunk.
  • #19 Headline + key visual top-left band.
  • #20 One mode per deck (presenter OR document) — never mix.
  • #21 (default ON) Brand logo on every slide unless opted out.

Path B — Websites

Q — What kind of site, and what's the goal?

Ask three quick things (then go):

  1. Type — landing page / marketing site / docs / portfolio / app UI? (Default: single-page landing.)
  2. The one job — what should a visitor do? (Sign up, book a call, buy, read.) This sets the primary CTA.
  3. Sections / content — headline value prop + the key beats. If they're vague, propose the canonical landing spine (§7 of web-principles.md) and let them trim.

One confirmation: theme + scope. Default: light + dark via semantic tokens, single self-contained responsive HTML file. Confirm once:

"I'll ship one self-contained responsive page with light+dark theming and your brand tokens. Want multi-page, a specific framework, or light-only instead?"

Read before emitting any HTML

  1. principles/web-principles.md — the 20 web rules. All 20 are the floor.
  2. The chosen brands/<name>/brand-style.md.
  3. For build-stack depth (tokens, shadcn, Tailwind v4, Radix, OKLCH, motion, theming), the sister repo power-design-web is the reference. Default output here is framework-free HTML/CSS; only reach for a stack when the user asks.

Output contract

Default: a single self-contained, responsive HTML file (default ~/Desktop/<name>-site.html) that:

  • Is valid, semantic HTML5 — real <header><nav><main><footer>, one <h1>, a skip-link.
  • Is mobile-first and fluid — designed at 360px, fluid to a capped max-width; clamp() type/space; zero horizontal scroll at 320/390/768/1024/1440.
  • Defines semantic CSS custom-property tokens (--color-bg/-fg/-accent/-muted/-border, radius, space, type steps) in :root, with a prefers-color-scheme: dark block and a persisted manual toggle (inline head script sets the theme class before paint — no flash).
  • Uses the brand's real colors/type/logo from brand-style.md.
  • Applies all 20 web principles (checklist below).
  • No external JS deps; Google Fonts / self-hosted woff2 OK. Ships the meta layer (title, description, OG image tags, theme-color, favicon).

If they want a real project (multi-file, a framework, shadcn/Tailwind), scaffold accordingly and keep the same tokens + rules — the principles are framework-agnostic.

Pre-emit checklist (web)

  • #1 Mobile-first; capped measure — shell max-width 1200–1440px, prose ≤75ch, no edge-to-edge text.
  • #2 Content-driven breakpoints (640/768/1024/1280); tested at 320/390/768/1024/1440, zero h-scroll.
  • #3 Fluid type & space via clamp() (no jumps; rem term preserves zoom).
  • #4 Body ≥16px; tap targets ≥44×44px, ≥8px apart.
  • #5 One primary CTA per view; repeated, not reinvented.
  • #6 Fold answers what/who/next in 5s on laptop + phone.
  • #7 F-scan for text, Z for hero; body left-aligned, never justified.
  • #8 Measure 45–75ch (max-width: 65ch).
  • #9 Line-height ≥1.5 body / 1.0–1.2 display; rhythm on 8.
  • #10 8pt spacing; one modular type scale.
  • #11 WCAG 2.2 AA floor — text ≥4.5:1, UI/focus ≥3:1, never hue alone; aim AAA on body.
  • #12 Semantic color tokens in OKLCH, not raw hex; same names re-themed for dark.
  • #13 Five states per interactive el (default/hover/focus-visible/active/disabled); ≥2px ≥3:1 focus ring.
  • #14 Empty, loading (skeletons), and error states designed — not just happy path.
  • #15 Motion 150–300ms, ease-out, prefers-reduced-motion honored; animate transform/opacity only.
  • #16 Reserve space (width/height or aspect-ratio on media) — CLS <0.1.
  • #17 Perf budget — LCP <2.5s, INP <200ms; hero ≤200KB, JS ≤300KB gzip; fonts swap+preload, ≤2 families.
  • #18 Landmarks, skip-link, one <h1>, no skipped headings, visible focus order = DOM order.
  • #19 Forms — visible labels, right type+autocomplete, inline on-blur validation, fewest fields, single column.
  • #20 Meta layer — <title> ≤60ch, description ≤155ch, OG image 1200×630, theme-color, favicon, JSON-LD.

After emitting (both paths)

  1. Save the file to disk (Write tool).
  2. Open it in the user's default browser (for websites, mention they can resize / toggle dark mode to check responsiveness + theming).
  3. Ask: "Want changes?" Iterate via natural conversation.

Common refinement patterns

Slides

  • "Make slide N bolder" → increase headline size or accent intensity, never both.
  • "Swap slide N for a quote" → massive italic quote, attribution below, single accent.
  • "Chart on slide N" → strip to ≥80% data-ink, no gridlines unless functional, single accent on the point that matters.

Websites

  • "Make the hero punchier" → shorten the value prop (≤12 words), raise contrast on the CTA, add one line of proof — don't add a second primary CTA.
  • "It feels cramped on mobile" → check §1/§10 — is text edge-to-edge? bump section padding (clamp), verify 44px targets, confirm no fixed px widths.
  • "Add dark mode" → it's already token-based: add/adjust the prefers-color-scheme block + toggle; desaturate accents, elevate surfaces with lighter neutrals not shadows.
  • "It's slow / janky" → §10: is the hero lazy-loaded (shouldn't be)? unsized images (CLS)? heavy JS (INP)? animate transform/opacity only.
  • "Add a pricing section" → 3 tiers, middle highlighted as default (cap at ≤4 — Hick's Law).

What not to do

Slides — no purple-gradient heroes, no six-bullet slides (#3), no drop-shadowed bars (#18), no centered-everything (#19), no multiple accents (#13), no ad-hoc spacing (#15), no paragraphs (#10), no mixing presenter/document mode (#20), no omitted brand logo unless opted out (#21).

Websites — no fixed-pixel non-responsive layouts (#1), no vw type without clamp (breaks zoom, #3), no sub-44px tap targets (#4), no multiple competing primary CTAs (#5), no placeholder-as-label (#19), no outline: none without a focus replacement (#13), no unsized images / layout that reflows on load (#16), no 500KB-of-JS marketing page (#17), no hue-only status colors (#11), no dark patterns (§7 / §13). Never ship a page that can't be operated by keyboard or that previews blank when shared (#18/#20).


Files in this skill

  • principles/design-principles.md — the 20 slide rules + research, with numeric thresholds.
  • principles/web-principles.md — the 20 web rules + research, with numeric thresholds.
  • lib/extract-brand.md — the Firecrawl brand-extraction recipe (feeds both paths).
  • brands/<name>/brand-style.md — pre-built brand systems (72+).
  • brands/_template.md — blank template for new brands.

When in doubt, read the principles file for the current medium. It's the source of truth.

Bundled files

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

and 40 more files.

Frequently asked questions

What does the Power Design AI skill do?

Generate beautiful, on-brand HTML — presentation decks or full responsive websites — in any brand's design language, combining brand DNA extracted via Firecrawl with codified, research-backed design principles (20 rules for slides, 20 for the web).

Why use Power Design on TypingMind?

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

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/ItsssssJack/power-design/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 Power Design?

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 Power Design?

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

Is the Power Design AI skill free?

It is published on GitHub by ItsssssJack. 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 👇