Taste Web Performance logo

Taste Web Performance

Community
madebymustafa
taste-web-performance

Core Web Vitals discipline for agent-built pages: LCP, CLS, INP budgets, font loading that doesn't shift layout, images that don't bloat, CSS/JS that doesn't block. A designed page that loads like a template is half a page. Use when generating or auditing any web page or component. Stack on the design-taste core skill.

Overview

Publishermadebymustafa
Repositorydesign-taste
Skill nametaste-web-performance
Stars
198
Forks
10
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 madebymustafa on GitHub. Read the source before you install it.

Installation

Install the Taste Web Performance 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/madebymustafa/design-taste/tree/main/skills/taste-web-performance
  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/madebymustafa/design-taste.git /tmp/design-taste
mkdir -p .claude/skills
cp -r /tmp/design-taste/skills/taste-web-performance .claude/skills/taste-web-performance
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Taste Web Performance 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 Taste Web Performance 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 Taste Web Performance 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.

taste-web-performance

Speed is part of the design. A hero that paints in 900ms with a shifted headline is a broken hero.

1. THE BUDGETS (Write These Down First)

Target p75 on mobile hardware (the median phone, not your Mac):

  • LCP < 2.5s — the hero text or hero image must be the largest contentful element and must paint fast.
  • CLS < 0.1 — no element moves after it's first painted. Every font, image, and ad-sized box is a CLS risk.
  • INP < 200ms — no click waits on a long task; no JS does heavy work on the main thread at interaction time.
  • Total page weight ≤ 1.5MB for a marketing page; a "simple" page should ship under 500KB.
  • State the budget in the build notes — a page without a stated budget is a page without a plan.

2. FONT LOADING (The #1 Agent Failure)

  • Never more than two font families, never more than ~4 styles (the design-taste skills already cap this — enforce it in code).
  • font-display: swap on all webfonts, and the text must not shift when the font lands: reserve the space or accept ≤ 3px movement.
  • preload the hero font face (woff2 only); load the rest lazily or from the correct unicode-range subsets.
  • No @import in CSS (render-blocking), no font loading scripts that block first paint, no loading fonts the page doesn't use on the current route.
  • If the pack's display font is decorative, consider system-font fallback for everything below the hero — the headline is where the font earns its weight.

3. IMAGES (Every One Gets a Decision)

  • Every image: explicit width/height (or aspect-ratio) — the CLS fix that costs nothing.
  • Format: AVIF/WebP at runtime, PNG only for real transparency, never a 2MB JPEG on a marketing page.
  • srcset/sizes for responsive delivery; loading="lazy" for everything below the fold, fetchpriority="high" for the LCP image.
  • Decorative imagery is CSS (gradients, noise, patterns) — an agent that "solves" a background with a 400KB image has failed the design, not just the budget.
  • Hero images are cropped at the correct aspect ratio in the markup, not squeezed by CSS.

4. CSS & JS DELIVERY

  • CSS: one file, or per-route critical CSS; no framework stylesheet for a 2-section page, no purged-out-of-the-box utility soup.
  • JS: defer everything that isn't interactive-critical; no framework for a static page (a landing page has no business shipping React).
  • No synchronous third-party scripts (chat widgets, analytics) in the critical path; analytics loads after interaction if it loads at all.
  • Animations run on transform/opacity only (see taste-motion); anything animating top/width/box-shadow on scroll is a jank generator, not a feature.

5. THE VERIFY RITUAL (Always Run It)

  1. Build, then run Lighthouse on a throttled profile (mobile, Slow 4G) — not on a localhost Mac.
  2. Report the three vitals + weight. Any budget miss gets fixed or gets a written reason.
  3. Re-check after any image/font change — those are the two things that regress silently.
  4. prefers-reduced-motion and offline-safe fonts are not optional extras; they're part of the contract.

6. CHECKLIST (Fail the Page If…)

  • Any font loaded without swap + dimensioned text, or more than two families.
  • Any image without dimensions, without format decision, or lazy-loading the hero.
  • Any layout shift visible on reload (test it; shifts lie about themselves).
  • Any render-blocking script/stylesheet, @import, or framework shipped to a static page.
  • Page weight or vitals over budget with no stated reason.
  • No Lighthouse run logged for the page.
  • The page looks designed but behaves like 2015.

Frequently asked questions

What does the Taste Web Performance AI skill do?

Core Web Vitals discipline for agent-built pages: LCP, CLS, INP budgets, font loading that doesn't shift layout, images that don't bloat, CSS/JS that doesn't block. A designed page that loads like a template is half a page. Use when generating or auditing any web page or component. Stack on the design-taste core skill.

Why use Taste Web Performance on TypingMind?

Because you install it once and use it with any model. Taste Web Performance 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 Taste Web Performance in TypingMind?

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

Which AI models can use Taste Web Performance?

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 Taste Web Performance?

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

Is the Taste Web Performance AI skill free?

Yes. It is published on GitHub by madebymustafa 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 👇