I built this skill after seeing one too many AI-generated pages with purple gradients, soft shadows, fake testimonials, and vague buttons.

The Editorial Minimalist UI skill fixes that by forcing the AI to start with the bones of the interface: structure first, typography second, then spacing, hierarchy, content, and surfaces before any decorative nonsense.

Hey, I’m Karo Zieminski 🤗

AI Product Manager and builder.

I write Product with Attitude, an AI newsletter for thousands of subscribers developing critical AI literacy the only way it sticks: through practice.

I maintain two tested Claude Skills libraries: one with skills I’ve built myself and one with free community-built skills shared in Attitude Vault by our community members.

SUBSCRIBE

What is Editorial Minimalist UI?

Editorial Minimalist UI is a product interface style that combines the restraint of minimalist UI with the hierarchy and rhythm of editorial design. It removes the obvious decorative junk and makes the remaining choices carry more weight.

Minimalism is not the absence of design. It’s the discipline of clarity: composition, functional elements, whitespace, typography, and removing every decorative detail that does not earn its place.

It treats typography as the main visual system, whitespace as structure, borders as depth, and copy as interface, not as an afterthought or filler.

That’s why this skill matters for AI coding agents. The agent needs constraints. And arguably, that’s what taste is: constraints with consequences.

Minimalist web design isn’t so much a specific design aesthetic as a set of principles or guidelines for design.

In practice, following these principles results in a simple website design that uses only the most essential elements to achieve the desired outcome from visitors.

— Toptal, Simplicity Is Key: Exploring Minimal Web Design.

Why AI-generated UIs Look Generic

AI can generate great design.

The problem is that many AI users can’t, at least not well.

That’s okay. Not everyone has to be a designer.

But if you want your page to look as if it were designed by a professional, vague prompts like “make a modern landing page” will not get you there.

Give the AI proper constraints, through a prompt or a skill, and you get a page with stronger structure, cleaner hierarchy, better spacing, sharper typography, and fewer decorative crimes against the internet.

That’s because when you ask an AI agent to “make a modern landing page,” it reaches for the statistical middle of modern landing pages.

That middle is crowded with every AI interface cliché: gradients, oversized rounded cards, generic icon sets, fake company names, vague CTAs, and placeholder copy.

This is why the skill bans certain things.

The Design Rule

Start with structure. Then typography. Then content clarity. Then surfaces. Then accent color. Then motion. Then imagery.

That sequence prevents the model from using visual noise to cover weak information architecture. It also makes the output easier to inspect. If the layout, hierarchy, and copy already work in warm monochrome, the design has a spine.

Accents can then clarify meaning instead of pretending to be meaning.

Editorial Minimalist UI Skill

Read more

This skill works in Claude Code and Claude Cowork, so you can use it both when coding and when managing longer agentic work sessions. It helps you keep an AI agent’s working space clean, focused, and useful.

Use it when:

• The agent starts repeating itself.

• The agent forgets an earlier decision.

• The agent invents a tool, file, or function.

• The AI suddenly asks what you’re working on.

• Output quality drops mid-session.

• Or you see each response getting more expensive.

What comes out:

• A handoff note that preserves the decisions.

• A persistent CLAUDE.md or memory/ entry for anything that should outlive the session.

• A scoped subagent prompt instead of one mega-agent juggling everything.

• A live artifact for views you’ll re-open later.

• Sometimes a scheduled task for work that can wait.

Hey, I’m Karo Zieminski 🤗

AI Product Manager and builder.

I write Product with Attitude, an AI newsletter for thousands of subscribers developing critical AI literacy the only way it sticks: through practice.

I maintain two tested Claude Skills libraries: one with skills I’ve built myself and one with free community-built skills shared in Attitude Vault by our community members.

SUBSCRIBE

The Skill

Read more

LinkedIn Teaser Generator turns a full Substack post into a LinkedIn teaser designed to earn attention and drive clicks.

It extracts the core argument, picks the strongest angle, writes a short post in your voice.

Key Design Decisions I Took

• Link goes in the first comment, not the post body. I just recently learned that LinkedIn’s algorithm suppresses posts containing outbound links. So teasing in the post and dropping the URL in a comment significantly increases reach.

• Hook patterns are constrained to three types (bold claim, specific result, pattern interrupt). These map to the formats that generate the highest dwell time on LinkedIn in 2026, where the algorithm rewards posts people actually stop to read.

• “No cringe” list bans overused LinkedIn phrases (”game-changer”, “let that sink in”, etc.) to keep your brand voice clean. If you have a separate Voice DNA skill active, it automatically defers to that instead.​

• Structured output gives you the post, a pre-written first comment with your Substack link, and post metadata (angle used, hook type, suggested timing), so you can copy-paste in under 30 seconds.​

Hey, I’m Karo Zieminski 🤗

AI Product Manager and builder.

I write Product with Attitude, an AI newsletter for thousands of subscribers developing critical AI literacy the only way it sticks: through practice.

I maintain two tested Claude Skills libraries: one with skills I’ve built myself and one with free community-built skills shared in Attitude Vault by our community members.

SUBSCRIBE

Using The Skill

Copy the skill script below and upload it to Claude Cowork. Once it’s saved, paste your Substack article link and type “linkedin”. The skill will extract the article, choose the strongest teaser angle, write the LinkedIn post, and generate a first comment with the link so you can publish faster.

Read more

That’s not how it works anymore.

AI Overviews and multimodal search engines like Gemini don’t just read our alt text, they parse it as a semantic unit. They extract meaning, map entities, and decide whether your image deserves to show up as a standalone answer snippet.

Which means our alt text isn’t a label anymore. It’s a micro-piece of content.

So we need to stop writing it like it’s 2019.

What Changed

Traditional Google Image Search matched keywords. You wrote “Substack growth chart,” Google matched it to a query, done.

AI-powered search does something different. It reads your alt text alongside the image, the surrounding copy, and the page structure -- then decides what the image means. An AI recognizes “Substack subscriber growth chart showing 340% increase after viral post” as a specific, meaningful entity. Not just a nice phrase.

Here’s the practical consequence: AI Overviews frequently surface images with their alt text as standalone answer snippets. If your alt text can’t stand on its own and still deliver value, you’re invisible in the new search landscape.

Hey, I’m Karo Zieminski 🤗

AI Product Manager and builder.

I write Product with Attitude, an AI newsletter for thousands of subscribers developing critical AI literacy the only way it sticks: through practice.

I maintain two tested Claude Skills libraries: one with skills I’ve built myself and one with free community-built skills shared in Attitude Vault by our community members.

SUBSCRIBE

The Principles That Matter

After digging through Google’s documentation, accessibility guidelines, and the emerging AIO optimization research, five principles hold up:

• Accessibility first, SEO second. Start by describing what a visually impaired reader would miss. That description is your alt text spine, you optimize from there, not the other way around.

• Entity-rich, not keyword-stuffed. Name specific tools, brands, concepts, and actions. One clean, natural keyword beats three awkwardly jammed in. Always.

• Context over appearance. A chart showing “revenue growth” matters more than “a bar chart with blue bars.” Describe the image’s purpose, not its surface-level visuals.

• 125-200 characters. Google indexes roughly 16 words as a ranking signal. Screen readers chunk at 125 characters. This range serves both without getting cut off.

• Write for AI parsability. Treat each alt text as a self-contained unit of meaning that an AI system can lift out and reuse as a standalone snippet.

I Built a Skill for This

I got tired of manually applying these rules every time I dropped an image into a Substack post. So I built a Claude Skill that does it automatically.

Here’s what it does: when you attach any image, it analyzes the visual content and outputs two things:

• An alt text (125-200 chars, entity-rich, naturally keyworded, accessibility-compliant)

• An SEO filename (3-7 hyphenated words, .webp)

It infers the keyword from the image itself so you don’t have to specify one each time -- though you can override it.

The alt text is built for both traditional Google Image Search and the newer AI Overview / AI Mode citations. It’s written as a standalone semantic unit that Gemini and other multimodal models can extract and reuse. Not just a keyword string for crawlers.

Getting Started

Copy the skill code below and save it as SKILL.md in your Claude skills folder. Once it’s saved, attach any image and ask for alt text. The skill handles the rest.

Read more