Loading video...
Video Failed to Load
Bruh... this AI skill actually makes AI-generated UIs look better 🤯 Taste Skill is an open-source skill that helps AI coding agents generate cleaner, more polished frontends, and it's compatible with popular agents like Claude Code, Codex, Gemini, Cursor, Lovable, and more :) Follow Ram Maheshwari for more 💎... show more
64,208 views • 3 months ago •via X (Twitter)
21 Comments

Well, you forgot to link the repo:

At least credit the original creator @LexnLin

taste is the one thing you can't npm install

Great. @grok give me the repo link of that taste skill

An open-source skill that works across Claude Code, Cursor and Lovable like that is a smart way to standardize UI quality 🎨

"taste" and it have that stupid AI tag on every title

Literally slop all over this

Design tokens help but the real failure mode is agents mixing utility classes with custom CSS mid-component, and you get a maintenance nightmare six months later. How does it handle existing codebases with legacy styling?

@LexnLin 🫡

Skills like this are useful because they encode judgment into the agent’s working context. The bigger pattern is that teams need reusable standards, not one-off prompts: design taste, accessibility, responsiveness, and review criteria all belong in the loop.

taste skill helps first screen but loading and error states still look messy

Frontend taste is hard to encode because what looks clean in one design system looks wrong in another. Does the skill adapt to the existing codebase's visual conventions or does it apply a fixed aesthetic regardless of context?

少点默认味仪表盘就很值了

The real question is whether taste can actually be encoded in a markdown file or if it's always going to lag behind whatever design trend was current when the skill was written.

@LexnLin

Skills that target output taste rather than just correctness are underrated. Most agent evals stop at "does it run" and ignore whether the generated frontend actually feels intentional.

encoding taste as a reusable skill is the right move — AI can generate UI but it can't judge it without something to calibrate against

What software do you use to make/edit these videos? 👀

any examples you built with?

the honest answer is most startups still struggle with frontend polish even with ai help, it's a skill that's still relatively hard to replicate

Design taste is becoming a real layer in AI coding workflows. The model can generate UI, but a reusable taste/spec layer is what keeps output consistent across iterations.

