A skill for Claude Code and Codex · Design and frontend
Structure first, a written direction, tokens and local fonts, a different composition per section, and an anti-generic review at the end. It's the skill that keeps my pages from looking like 2010 or like an AI template, and it's the one that runs the most around here.
Solo skill: unlocks right away, no guarantee after download. Purchase terms.
Or get all 25 skills plus 9 bonuses in the Supreme Pack for R$ 497.
The problem
It's not that the model gets it wrong, it's the order: it opens the CSS before knowing what the page needs to do. Three symptoms that give away a page built on autopilot, and what changes once the skill takes the lead.
| Without the skill | With frontend-expert |
|---|---|
| You ask for the landing page and the AI just starts writing: headline, paragraph, two buttons, and three identical icon cards. Nice for ten seconds, identical to every other one out there. | It asks for the goal, who's arriving and from where, builds the section list with a persuasive job for each one, and only then decides what the page is going to look like. |
| A single font, almost always the same one, pulled from Google in neighboring weights. Nothing has hierarchy, and the page has to ask an outside server for permission before it loads. | Two local font families with a real size jump between them, one accent color, and zero external requests. The direction gets written into the project, and whoever touches it next reads it first. |
| The same composition five times in a row, no image after the hero, a flat background from top to bottom. The page turns into a document, and the eye stops scrolling. | Every section shifts composition, has an image where it needs one, and depth where it makes sense. At the end, a review measures the page and fails whatever still looks like a template. |
What you get out of it
Three pages from my own site, built with this exact process. No designer involved: the skill asks for the structure, writes the direction, builds it with the kit, and fails whatever still looks generic.
How it works
One action, who's landing there and through which traffic, and whether it's a landing page, corporate site, software, or dashboard. It reads the reference for that type before anything else.
Calls in copy-thief (the copywriting skill, free on my GitHub) and returns the section list, each one with a different persuasive job. No CSS gets written before that list exists.
Picks one of 12 directions with color, typography, and motion already locked in, adapts it to your brand, and logs it in the project: color, font pairing, signature device, and image plan. Whoever touches the project next reads this first.
Starts from a complete template, pulls in ready-made components (hero, pricing, steps, FAQ, closing band, and more), only the effects the page actually uses, and the local fonts it installs itself. Zero Google Fonts.
A checker measures fonts, typography, imagery, composition, depth, CTAs, motion, and accessibility, and fails the page if it repeats a composition or has no image after the hero. Then it screenshots at three widths against an 11-point rubric, with a minimum score on each.
Before you buy
No API key required. Photos and illustrations are your own, or whatever another skill generates.
It detects when a project already has a direction and switches to maintenance mode, without redesigning what you've already polished by hand.
Its limits
The skill picks the direction, builds the page, and fails whatever looks generic, but a good photo is still on you: it generates a neutral placeholder at the right size and marks where the real image goes. When you don't have the photo, generate one with whatever image tool you already use and bring it back for the skill to place.
In React, it covers tokens for Tailwind v4 and shadcn and explains how to port the kit over, but the kit itself is plain HTML, CSS, and JS. If you live in Next with your own design system, you get value from the process and the rubric, not the templates.
And it doesn't replace copy-thief: it calls in the copywriting skill in step 2 because structure comes before visuals. copy-thief is free, so that doesn't cost you anything extra.
What you receive
The price
Just frontend-expert
R$97one time payment
Bought it solo and now want it all? Message me on WhatsApp and I will deduct what you already paid.
Supreme Pack
R$ 1.665R$497or 10x on a card
Buying all 25 solo adds up to R$ 1.665. The pack pays for itself by the ninth skill. See everything the Supreme Pack includes
02 · Design and frontend
Before you ask
Anthropic's is a well-meaning paragraph: it asks for a distinct aesthetic and stops there. This one has an ordered process (structure, written direction, composition, build, review), a kit to copy from instead of improvising, local fonts, and a review that actually measures the page and fails it when needed. It was also calibrated on real Brazilian pages, tuned to the industries, tone, and traffic patterns found there.
The process and the rubric apply to any stack. The kit itself is plain HTML, CSS, and JS; for React, it brings a token reference for Tailwind v4 with shadcn and a roadmap for porting components. If your project is Next with its own design system, you use the direction and the review, not the templates.
No. It generates a neutral placeholder at the right size and writes the image plan: how many, what type, and where. The real photo is yours, or comes from whichever image generation tool you already use.
It calls in copy-thief during step 2, since the section list comes before the visuals. copy-thief is free, on my GitHub, under the MIT license. Without it installed, frontend-expert builds the structure on its own, with less rigor on the copy.
That's exactly why it exists here: the pages on my own site and the ones I deliver to clients all come out of it. The license covers work you do for clients; what's not allowed is reselling the skill.
One-time payment, 12 months of updates. No refunds after download; the Supreme Pack comes with a 7-day guarantee.
