---
title: Taste Skill Anti Cookie Cutter
category: product
entity_type: skill
price: $15
canonical: https://forgehouse.ai/skills/taste-skill-anti-cookie-cutter/
lang: en
hreflang_alt: https://forgehouse.ai/tr/skiller/taste-skill-anti-cookie-cutter/
last_updated: 2026-06-20
---

# Taste Skill Anti Cookie Cutter

> AI-uretilmis "slop" tasarimi (generic hero + 3 kolonlu feature + gradient blob + stock…

Detect the 'AI slop' smell in a design: generic hero gradient blobs, three-column feature grids, stock-photo teams, lorem-ipsum verbs like 'Empower your business': and replace it with craft-driven, reference-based work. It runs a twelve-pattern audit (seven or more matches equals an AI-slop verdict), then rebuilds with real-product benchmarks instead of imagination, variable-font typography tension and a tight constraint set.

## Use cases
- Auditing a landing page that 'looks amateur' or 'AI-generated'
- Replacing a cookie-cutter hero with a craft-driven alternative
- Pulling real-product benchmarks before designing from scratch
- Running a three-iteration craft loop with surgical fixes
- Setting a constraint config (5 colors, 2 fonts, 3 radii, 2 shadows)
- Catching slop automatically in CI with a pattern-detection gate

## Benefits
- Turn 'it looks AI-made' rejection into a measurable, fixable verdict
- Replace imagination with real-product references so output stops reading generic
- Build typography tension that signals craft instead of a single dead font weight
- Cut iteration waste by catching slop before the design reaches review

## What’s included
- A twelve-pattern detection rubric with a slop-verdict threshold
- A reference-first benchmark query snippet to pull real product screens
- A three-iteration craft loop: craft, four-dimension audit, surgical fix, then pivot
- A variable-font axis CSS layer for display/body/caption typography tension
- A Tailwind constraint config that caps colors, fonts, radii and shadows
- A fifteen-item anti-pattern list and a CI taste-audit workflow

## Who it’s for
Designers and teams who keep landing on the same generic template and need to make output read like a professional product, not AI output.

## How it runs
Gradient blob hero, three icon columns, Empower and Transform in the copy: recognize it? A 12-pattern rubric scores the slop, real product screens replace imagination as the source, and one surgical fix lands per iteration until the verdict flips.
1. Scores the design against a 12-pattern detection rubric (gradient blob hero, 3-column icon grid, single font weight everywhere, default Tailwind palette, placeholder verbs like Empower and Transform): 7+ matches is an AI slop verdict, 4-6 means mandatory refactor
2. Pulls a real benchmark before producing anything: a Refero query for 20 actual product screens in the same sector and tone, then extracts the shared composition, typography and palette patterns from their design decisions. Imagination is not a source, synthesis is
3. Locks a constraint set into the config: maximum 5 color tokens, 2 font families, 3 border radii, 2 shadows, one timing function, so decision fatigue cannot leak generic choices back in
4. Produces iteration 1 with variable font axis tension applied: display around 720 weight, body around 480, caption around 380, line height 1.6 and a 38-45ch reading measure instead of one flat weight everywhere
5. Runs the 4-dimension visual audit on the output: composition (is there a visual anchor), typography (axis tension and rhythm), color (max 5 tokens, max 3 saturation levels), spacing (8px baseline grid), then rescores the 12-pattern rubric
6. Applies one surgical single-dimension fix per iteration; if the same area still fails after three iterations, it forces a concept pivot instead of more tweaks on a dead idea

## FAQ
### My landing page technically works but clients say it 'feels amateur.' Is that what this addresses?
Exactly that. The vague 'looks AI-made' reaction gets converted into a twelve-pattern audit, gradient blobs, three-column feature grids, stock-photo teams, dead font weight, where seven or more matches equals a slop verdict. Once the problem is measurable, the rebuild path is concrete instead of another blind redesign.

### Taste sounds subjective. How do you turn it into something measurable?
By auditing patterns, not feelings. The rubric checks twelve concrete slop signatures, then rebuilds with real-product benchmarks pulled before designing, a constraint config capping the palette at 5 colors, 2 fonts, 3 radii and 2 shadows, and variable-font typography tension. A three-iteration loop with surgical fixes replaces endless full redesigns.

### Will it generate finished designs for me?
No. It diagnoses slop, supplies references, constraints, and the iteration discipline, but the design work still happens in your hands and tools. It also includes a CI taste-audit gate, which catches regressions automatically, yet it cannot invent a brand identity where none exists.

## Price
$15, one-time, no subscription. VAT included.

Related guide: [Building a design system with Claude Design](https://forgehouse.ai/guides/claude-design-design-system/)