正在加载视频...

视频加载失败

Broooooo this is so freaking good 🤯 This open-source skill makes AI coding agents generate better-looking UIs. Taste Skill works with Claude Code, Codex, Cursor, Gemini CLI, Lovable, v0 and more. Stops the generic frontend slop. Perfect for developers, vibe-coders & designers who actually want the AI output to...

93,540 次观看 • 1 个月前 •via X (Twitter)

33 条评论

Manish Kumar 的头像
Manish Kumar1 个月前

Follow @Manixh02 for more.

Jony · AI与科技 的头像
Jony · AI与科技1 个月前

就烦生成出来全是同一套紫渐变卡片。能塞进 Claude/Codex/Cursor 里用,有点东西。

zzxvt 的头像
zzxvt1 个月前

full circle, we’re back to terrible flash sites without the flash

mjrode 的头像
mjrode1 个月前

Nothing inspires confidence in an anti-AI-slop tool like an AI-slop landing page.

Manish Kumar 的头像
Manish Kumar1 个月前

check out this new Ai App too

AI Apps API 的头像
AI Apps API1 个月前

The useful part is that this is a standard the agent actually reads. Style guides have always existed and coding agents ignore them, because a guide written for humans is context the model never knows to apply. Packaging it as a skill flips which side has to remember. Same reason agent written code needs its own quality pass, nothing tells the model your repo has a rule until it is already writing.

Saman Pandey 的头像
Saman Pandey1 个月前

So goated!!

Luís Rodrigues 的头像
Luís Rodrigues1 个月前

The underrated part of AI coding is that generating code is easy. Generating taste is the hard part.

Moez Zhioua 的头像
Moez Zhioua1 个月前

Got it. The anti slop approach makes AI‑generated UIs feel intentional. Rule: Keep UI scaffolding minimal, let design drive the code

Dharmvir 的头像
Dharmvir1 个月前

So amazing

Alfin 的头像
Alfin1 个月前

Man is this a promo?

Manish Kumar 的头像
Manish Kumar1 个月前

no bro, i use labels when I do promos 😭😭

Saul Flores Jr. 的头像
Saul Flores Jr.1 个月前

I like that carousel animation looks really smooth. Honestly, I'm willing to try anything to make websites look more designed by humans than manufactured by AI.

abhinav ramesh 的头像
abhinav ramesh1 个月前

Cool! Will check it out

Taz 的头像
Taz1 个月前

Deva in July 2026: "production costs are now 0. The real moat is taste" Devs in August 2026:

Abhishek Swami 的头像
Abhishek Swami1 个月前

Does Gemini CLI exist?

Mohammed Tawakkal Ahmed 的头像
Mohammed Tawakkal Ahmed1 个月前

So damn cool man

dhinna ship .ico 的头像
dhinna ship .ico1 个月前

Still need to check how it respects existing component boundaries in Cursor

Mohit 的头像
Mohit1 个月前

Share more like this

Heinrich 的头像
Heinrich1 个月前

Looks really cool Thanks for sharing

MDC 的头像
MDC1 个月前

Thanks this was much needed honestly

Noel 的头像
Noel1 个月前

Anything that pushes agents away from generic purple-gradient UI is welcome. Curious how opinionated Taste Skill is — does it enforce a design system, or more like anti-slop heuristics?

AnotherGameDev 的头像
AnotherGameDev1 个月前

Thought you a shit poster, thought you spreading shit Honestly did, there are so many “shit ai stuff posters” here But fuck. You are not. Wanted to say that. Follow

fudes 的头像
fudes1 个月前

The people shipping AI products don't need your permission. They don't need your analysis. They don't need your tweet.

uslab|AI×受託開発の検証ログ 的头像
uslab|AI×受託開発の検証ログ1 个月前

"Generic frontend slop" nails it — the agent has no criteria to check its own output against. I hand the design criteria over first, then have it generate a few variants and judge them against that 👀 Does Taste Skill score the output too, or only guide the generation?

Samay 的头像
Samay1 个月前

damm so cool

AI Mastery Guide 的头像
AI Mastery Guide1 个月前

Working across that many coding agents at once is genuinely useful.

Odeh Judge Anigboro | Building LifeSync 的头像
Odeh Judge Anigboro | Building LifeSync1 个月前

I believe more innovations are coming as we allow knowledge takes it place

Dark Hair Cult 的头像
Dark Hair Cult1 个月前

This one or impeccable?

Satyam 的头像
Satyam1 个月前

crzy

Santiago Vega 的头像
Santiago Vega1 个月前

vaya, esto lo necesito

Neeraj 的头像
Neeraj1 个月前

Great share!

jendy 的头像
jendy1 个月前

A design skill should be judged like a linter, not a mood board. Give it the same brief and compare hierarchy, accessibility, design-system fit, implementation complexity, and regressions. “Less generic” matters only when the result is repeatable and reviewable.

相关视频