KairaxisKairaxis
Freetaste skillmedium risk

Capability · rank 03

design-taste-frontend

Ship landing pages, portfolios, and redesigns with intentional art direction instead of generic AI patterns.

Taste Skill reads the brief, infers an appropriate design language, calibrates layout, motion, and density, and applies audit and pre-flight rules that reduce templated frontend output.

Listed owner

Leonxlnx/taste-skill

Package source

tasteskill.dev

Category

TASTE

TASTE
$npx skills add https://github.com/Leonxlnx/taste-skill --skill "design-taste-frontend"
OverviewSandboxUsageExamplesLimitationsVersionsRelated Skills

Overview

Taste Skill reads the brief, infers an appropriate design language, calibrates layout, motion, and density, and applies audit and pre-flight rules that reduce templated frontend output. Review the sandbox comparison, limitations, permissions, and pinned version before install.

Outcome

Ship landing pages, portfolios, and redesigns with intentional art direction instead of generic AI patterns.

Who it is for

Builders using AI coding agents for marketing sites, portfolios, and frontend redesigns.

What it is not for

Dashboards, data tables, multi-step product UI, or final accessibility and brand approval.

Runtime export

codex_skill · markdown_export

TASTE domain: ui_ux_taste

Sandbox comparison
Generic baseline vs design-taste-frontend on the same sample input · pinned v2.0.0-experimental

Sample input

Create a premium landing page from a short brief without producing a templated AI-style interface.

Model: gpt-4.1-mini · temp 0.2

Generic baseline

45/100
Use a centered hero, purple gradient, rounded CTA buttons, three equal feature cards, and subtle animations.

Skill-enhanced

93/100
Design read: premium technical product for design-conscious buyers, using a restrained editorial system.
Dials: variance 7, motion 5, density 3.
Direction: asymmetric hero with one real product visual, a locked neutral-and-cobalt palette, and varied section layouts.
Pre-flight: keep the hero within one viewport, use one CTA label per intent, verify contrast and reduced motion, and remove decorative UI that carries no meaning.

Rubric scores

+48 lift
Brief-aware direction2 → 5
Layout distinctiveness2 → 5
Visual-system coherence2 → 5
Pre-flight completeness1 → 4

Assumptions

  • Sample input is representative of a real production task
  • Same model and temperature for baseline and skill-enhanced runs

Comparison runs through the shared sandbox engine on registry sample cases (deterministic v0 path).

Usage

  1. 1. Inspect trust badges, limitations, and permissions on this page.
  2. 2. Run the sandbox comparison to see baseline vs skill-enhanced output.
  3. 3. Install the pinned version with the command below for your agent runtime.
  4. 4. Agents can also fetch /api/v1/skills/design-taste-frontend/manifest.
$npx skills add https://github.com/Leonxlnx/taste-skill --skill "design-taste-frontend"

Example prompts & use cases

1 / Prompt

Build a premium product landing page without defaulting to generic AI design patterns

2 / Prompt

Audit this portfolio before redesigning it, preserving the brand elements that already work

3 / Prompt

Infer a design direction and calibrate layout variance, motion, and visual density from this brief

Limitations

  • The default v2 ruleset is experimental and may change before the stable v2.0.0 release.
  • Its visual rules are contextual guidance and still require review against the product, brand, and audience.

Audience

Builders using AI coding agents for marketing sites, portfolios, and frontend redesigns.

Not for

Dashboards, data tables, multi-step product UI, or final accessibility and brand approval.

Version history

v2.0.0-experimental

May 2026current

Experimental v2 rewrite with brief inference, design-system mapping, redesign audits, and a hard pre-flight check

v1.0.0

Mar 2026

Original dial-driven anti-slop frontend skill, retained upstream as design-taste-frontend-v1