
antislop-ui-ux-skill
A skill (Claude Code, Codex, OpenCode) that de-slops an existing screen, or builds a new one without the AI-generated look.
Install with your AI
Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.
Install and set up antislop-ui-ux-skill (claude-skill project) into my current project. Found on https://claudeers.com/antislop-ui-ux-skill Repo: https://github.com/gustavo-henriq/antislop-ui-ux-skill Homepage/docs: — Detected install method: claude-skill → # copy this skill into .claude/skills/antislop-ui-ux-skill/ Category: other. Platforms: web. Read the repo's README for exact setup and env vars, then install it and wire it into my project. Claudeers Health Verdict: unknown; community-verified: false. Confirm the source before running anything.
# copy the skill dir into your project: # .claude/skills/antislop-ui-ux-skill/ (or ~/.claude/skills/antislop-ui-ux-skill/ for all projects)
git clone https://github.com/gustavo-henriq/antislop-ui-ux-skill
// compatibility
| Platforms | web |
|---|---|
| Operating systems | — |
| AI compatibility | claude |
| License | — |
| Pricing | open-source |
| Language | HTML |
take the "AI slop" out of an interface
A skill (Claude Code, Codex, OpenCode) that de-slops an existing screen, or builds a new one without the AI-generated look.
"AI slop" is a screen that looks assembled by someone who doesn't use the product: every block with the same visual weight, text explaining what a drawing would show, repeated actions, gradients/glass/glow used as decoration. The goal is for every element to exist for a reason the user can feel.
Before and after
The full gallery — every pattern the system flags, side by side with the fix (slop on the left in red, correction on the right in green):

One image per rule is also in docs/pares/, and the whole gallery is rendered in anti-slop/galeria.html (open it in a browser).
What it does
- De-slop a screen that already exists (app, Figma, or a prototype it just made).
- Build from scratch a new screen without inheriting the AI look.
- Ask before deciding — intent, screen role, visuals, priorities. That's the main differentiator.
- Search references with a cheap subagent — it browses curated sources, takes screenshots, and returns only the outline; the main agent decides and applies (anti-token protocol).
- Propose thematic effects tied to the product's universe (a stars app gets constellations, a music app gets a waveform).
- Run layered checks: signal audit, aesthetic checklist, redundancy test, "outside look", second pass.
How it works (the loop)
- Read the project's taste (
gosto.md) and rules (receitas.md). - Spot the slop signals, pointing each one in place.
- Ask, propose a grayscale wireframe, and only then implement.
- Self-contained HTML prototype (base64 images, real states).
- Double check + outside look + second pass, and report what it found.
Structure
deslop/
├── SKILL.md # the core: the loop and the rules
├── estetica-ia.md # quick checklist (12 signals, with CSS and "when it's legitimate")
├── tipografia.md # fonts without the AI look (indie, no serif, neutrals that pass)
├── vidro.md # real glassmorphism × AI glass (with CSS recipes)
├── componentes.md # anatomy of table, metric, chart, chat, form
├── fontes-de-referencia.md # curated sources (Mobbin, 60fps, Land-book, Godly, …)
├── gosto.md # template: the project's choices and rejections
├── receitas.md # template: palette, surfaces and components
├── anti-slop/
│ ├── catalogo.md # 80+ patterns in categories (with sources)
│ ├── galeria.html # 20 slop × correction pairs, rendered
│ └── referencias/ # real slop examples ("AI SaaS" templates)
├── inspiracao/ # curated good references, by theme (landing, motion, dark…)
└── docs/ # images: full gallery and one image per rule
Anti-slop collection
catalogo.md— 80+ "AI slop" patterns and quality issues, in 10 categories (typography, color, layout, motion, copy…), each with what gives it away and the fix.estetica-ia.md— the 12-signal checklist to run on every screen.galeria.html— the visual version: each signal as a slop × correction pair, side by side.
Install
Claude Code
Copy the skill folder into the project:
cp -r skills/deslop my-app/.claude/skills/deslop
Then fill in gosto.md and receitas.md in the project's folder (they are templates). Everything else lives in the skill.
Codex / OpenCode
Same idea: place the folder in skills/deslop (or the agent's skills directory) and point gosto.md/receitas.md to the project.
Why only gosto.md and receitas.md per project
The skill is canonical and generic. Each project keeps only what is its own — the taste (choices and rejections) and the rules (palette, surfaces, components). The anti-slop collection, the sources and the rules live here, to avoid duplication.
Golden rule
Every reference becomes a concept (hierarchy, rhythm, metaphor) — never a copy (palette, font, literal layout). Good references come contaminated too: take the idea, drop the aesthetic.
Author
Made by Guto · @henrik404
// faq
What is antislop-ui-ux-skill?
A skill (Claude Code, Codex, OpenCode) that de-slops an existing screen, or builds a new one without the AI-generated look.. It is open-source on GitHub.
Is antislop-ui-ux-skill free to use?
antislop-ui-ux-skill is open-source, so it is free to use.
What category does antislop-ui-ux-skill belong to?
antislop-ui-ux-skill is listed under other in the Claudeers registry of Claude-compatible tools.
// embed badge
[](https://claudeers.com/antislop-ui-ux-skill)
// retro hit counter
[](https://claudeers.com/antislop-ui-ux-skill)
// reviews
// guestbook
// related in Other
Anti-AI-slop design skill for Claude Code, Cursor, and Codex.
Huashu Design · HTML-native design skill for Claude Code · Claude Code 里 HTML 原生的设计 skill · 高保真原型 / 幻灯片 / 动画 + 20 设计哲学 + 5 维评审 + MP4 导出 · Agent-agnostic
Open source Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents. Built for multitasking, organization, and programmability.