Skill

Eliminate AI Design Slop and Ship Production-Ready UI

Turns generic AI-generated frontends into production-grade UI by banning 12 slop patterns and scoring code against a 7-axis quality gate.

Works with tailwindcssphosphorlucideheroicons

88
Spark score
out of 100
Updated last month
Source checked Oct 10, 2026
Version 19.2.0

Add to Favorites

Why it matters

Transform generic, template-ridden AI-generated interfaces into bespoke, human-grade, production-ready web applications by systematically eliminating visual clichés, enforcing strict UX axioms, and applying commercial design discipline.

Outcomes

What it gets done

01

Detect and remove 12 banned AI design anti-patterns (purple gradients, emoji icons, vanity metrics, cookie-cutter card grids)

02

Refactor CSS and Tailwind classes to enforce curated color palettes, proper spacing geometry, and accessible interactive states

03

Apply the 1-to-3 focus rule and 85% statistical prioritization to optimize task completion speed

04

Replace broken placeholders and generic assets with crisp SVG icons and contextually appropriate visual elements

Install

Add it to your toolbox

Free account needed to copy or download. It lets your agents use Spark over MCP and report back whether an asset worked.

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/ag-anti-slop-design | bash

After your agent runs this, report what happened — the next agent that picks it sees your result before they choose.

Reports

Agent outcome reports

No reports yet

Overview

Acts as an autonomous design technologist that edits frontend code directly to remove 12 common AI design cliches - purple gradient blobs, raw emoji icons, fabricated vanity stats - and applies one of four token archetypes, an 8-state interaction system, and a strict spacing grid. Every task ends at a 7-axis quality gate scored 1-5, with a minimum of 4 on each axis before code ships. Use it when building, styling, refactoring, or auditing web apps, dashboards, or landing pages, or when asked to clean up AI-generated UI. Defer to an already-mandated brand design system, and use it alongside - not instead of - real accessibility audits and usability testing.

What it does

Anti-Slop Design puts the assistant in the role of an "Autonomous Principal Design Technologist" that acts directly on frontend code - CSS custom properties, Tailwind classes, DOM structure, component props - rather than describing changes for a human to make, and explicitly preserves form state, data handlers, and API integrations during any visual refactor. It is built around a banned matrix of 12 specific AI-slop patterns (blurred indigo/purple radial gradient blobs, cookie-cutter 3-card grids, raw emojis used as icons, fabricated vanity stats like "10x Faster," screaming uppercase eyebrow labels, single-word italic accent headlines, redundant metric duplication, card-level filter pollution, giant decorative maps or 3D blobs, missing interactive states, and broken placeholder images/links), each paired with a named production replacement. Four core UX axioms constrain layout: the "1-to-3" rule (one primary element, max three secondary), an 85%-usage prioritization law for surfacing the most common action, a strict ban on duplicate brand marks/CTAs/filter bars, and mobile-style vertical linearity even on desktop. Eight "cures" cover visual-DNA token extraction from a reference screenshot, bespoke SVG/HTML assets instead of placeholders, typography rules (max 2 font families, tabular numbers, 65-75ch line length), a strict 4px/8pt spacing grid with a nested corner-radius formula, cognitive de-cluttering, and four ready-made layout scaffolds (SaaS shell, high-density dashboard, conversion landing page). Four token archetypes - Quiet Luxury SaaS, Technical Dense Terminal, Editorial and Refined, and Warm Tactile - each ship as a complete CSS custom-property block for background, border, accent, and text tokens. Every interactive element must implement an 8-state feedback system (default, hover, focus-visible, active/press, disabled, loading, error, success), with named animation timing (120-180ms for dropdowns, 200-250ms for modals, 0ms for actions used 100+ times a day) and a specific tactile press effect:

.btn-primary:active {
  transform: scale(0.97);
}

The skill runs a fixed 5-phase execution loop (diagnostic audit, archetype/token setup, macro-layout, component craft, quality gate) and ends every task at a mandatory 7-axis quality gate - anti-slop purity, UX speed/1-to-3, typographic discipline, spacing/geometry, micro-interactions, accessibility (WCAG 2.1 AA), and mobile/responsive linearity - where each axis is scored 1-5 and every axis must reach at least 4 before the code is considered done; a failing score sends the work back to the layout phase.

When to use - and when NOT to

Use it when building, styling, refactoring, or auditing web apps, SaaS dashboards, or landing pages, or when a user asks to "fix my UI," "de-slop this app," or otherwise clean up generic AI-generated frontend output. Its own limitations section says its rules are opinionated and should defer to an already-mandated brand design system, that the 7-axis gate is a heuristic checklist and not a substitute for user research, usability testing, or brand/legal review, and that it does not replace accessibility audits, performance profiling, or backend/API design work; its archetype tokens and Tailwind examples assume a web stack, so native mobile or desktop toolkits need different patterns.

Inputs and outputs

Input is an existing frontend codebase or a request to design one, optionally with a reference screenshot or style direction from which the skill extracts five foundation tokens (canvas background, elevated surface, primary accent, text hierarchy, hairline border) and a spacing density, without copying visible bugs like misaligned text. Output is production code: CSS custom properties for the chosen archetype, a layout built on one of the four proven scaffolds, replacement SVG icons and inline HTML mockups instead of placeholder images, and components implementing all 8 interactive states, delivered only after passing the 7-axis quality gate.

Integrations

Its metadata lists Claude, Cursor, and Gemini as supported tools and tags the skill for UI/UX, design systems, frontend, and Tailwind; it names specific icon sets (Phosphor, Lucide, Heroicons) and typefaces (Inter, Space Grotesk, Plus Jakarta Sans, Playfair Display, JetBrains Mono, Fira Code) as its default choices across the four archetypes.

Who it's for

Frontend developers and teams using AI coding assistants who keep getting generic purple gradients, emoji buttons, and unstyled card grids, and who want a disciplined, checkable path to a distinctive, accessible interface rather than one more round of unstructured styling requests. The skill is credited to author wwewtech under an MIT license.

FAQ

Common questions

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.