Skip to content
Productivity
Skill

/taste-skill

High-agency frontend design system for distinctive premium UI with anti-slop layout, motion, typography, and engineering rules.

From plugin
cowork-os
544157 skills

Claude Code

$ npx -y skills add CoWork-OS/CoWork-OS --skill taste-skill --agent claude-code

SKILL.md

taste-skill.SKILL.md
name: taste-skill
description: High-agency frontend design system for distinctive premium UI with anti-slop layout, motion, typography, and engineering rules.
version: "1.0.0"
metadata:
  author: CoWork OS Contributors <info@coworkosapp.com>
  source-author: "Leonxlnx"

Taste Skill

Use this skill when you need a strong frontend point of view, not a safe default.

The goal is not "pretty enough." The goal is a frontend that feels intentional, premium, and hard to confuse with generic LLM output.

1. Active Baseline Configuration

  • `DESIGN_VARIANCE: 8`

`1 = symmetric and quiet`, `10 = asymmetric and experimental`

  • `MOTION_INTENSITY: 6`

`1 = nearly static`, `10 = cinematic and highly reactive`

  • `VISUAL_DENSITY: 4`

`1 = airy gallery`, `10 = dense cockpit`

Treat these as working dials, not fixed constants.

  • Do not ask the user to edit this file.
  • Adapt the dials to the brief. If the user wants calmer output, lower them. If the user wants bolder output, raise them.
  • Use the dials to drive layout, animation, spacing, and card usage.

2. Required Working Model

Before coding, establish three things:

1. Visual thesis One sentence for mood, material, and energy. 2. Layout thesis What makes the page memorable: asymmetry, crop, white space, stacked motion, split hero, bento rhythm, etc. 3. Interaction thesis Two or three motions that materially change the feel of the interface.

Then implement real code. Do not stop at design commentary.

  • Do not leave "TODO: add animation here" style placeholders.
  • Do not output fake code snippets for components that are supposed to be shipped.
  • Do not ask for permission to make obvious design decisions once the brief is clear.

3. Architecture and Dependency Rules

Unless the user explicitly asks for something else, prefer React or Next.js.

Dependency verification

Before importing any third-party package:

  • check `package.json`
  • if the package is missing, state the exact install command before using it
  • never assume `framer-motion`, `lucide-react`, `zustand`, or any font package is already installed

React and Next.js rules

  • Default to server-safe structure in Next.js.
  • Keep global state out of server components.
  • If motion or highly interactive behavior is required, isolate it in a focused client component.
  • Do not spread animation logic across an entire tree when one leaf component can own it.

Styling rules

  • Use Tailwind for most styling unless the project clearly uses something else.
  • Check whether the project is Tailwind v3 or v4 before writing config-sensitive code.
  • Do not use v4 patterns in a v3 project.

Icon rules

  • Prefer `@phosphor-icons/react` or `@radix-ui/react-icons` when the project supports them.
  • Keep icon stroke weight consistent across the page.

Hard bans

  • No emojis in code, markup, UI copy, or alt text.
  • No `h-screen` heroes on mobile-sensitive layouts. Use `min-h-[100dvh]`.
  • No flexbox percentage math when grid solves the layout more cleanly.

4. Design Engineering Directives

LLMs drift toward familiar, overused UI patterns. Correct for that proactively.

Typography

  • Default headline energy: large, tight, decisive.
  • Avoid generic font stacks like Inter, Arial, Roboto, and raw system defaults when the brief calls for premium or creative work.
  • Favor distinctive but disciplined display fonts such as Geist, Outfit, Cabinet Grotesk, or Satoshi.
  • For dashboards and software UI, serif display choices are banned.
  • Pair a strong display face with a quieter supporting face or mono.

Color

  • Use a restrained neutral base plus one main accent.
  • Keep accent saturation controlled.
  • Avoid the default AI purple/blue glow aesthetic.
  • Do not mix warm and cool gray systems randomly.

Layout

If `DESIGN_VARIANCE > 4`:

  • do not center the entire hero by default
  • avoid "headline + two buttons + card grid" as the first screen
  • prefer split-screen, offset content, asymmetric negative space, anchored media, or broken-grid composition

Surfaces and cards

  • Cards are not the default.
  • Use cards only when elevation communicates hierarchy or interaction.
  • For denser dashboards, prefer dividers, border rhythms, and spacing over endless card mosaics.
  • If removing a card keeps the meaning intact, remove it.

Forms and states

Every serious surface needs full interaction coverage:

  • loading state
  • empty state
  • error state
  • success state
  • tactile active state

For forms:

  • label above input
  • helper text optional but useful
  • error text below input
  • consistent vertical spacing

5. Motion System

Motion should create presence and hierarchy, not noise.

If `MOTION_INTENSITY <= 3`:

  • limit motion to entrance, hover, and small affordance shifts

If `MOTION_INTENSITY > 5`:

  • add continuous micro-interactions
  • stagger reveals
  • use spring-based motion
  • make at least one interaction feel physically responsive

Preferred motion patterns

  • staggered entrances
  • layout transitions
  • hover-aware buttons
  • scroll-linked depth shifts
  • type or shimmer loops for active interfaces
  • floating or breathing status details

Magnetic interaction rule

If you build cursor-reactive magnetic behavior:

  • do not use React `useState` for per-frame cursor updates
  • prefer motion values or equivalent render-loop-safe primitives

Motion quality bar

  • no linear-by-default motion
  • no gratuitous infinite animation without purpose
  • no animation that only decorates and does not improve rhythm, focus, or affordance

6. Materiality and Visual Texture

When the design calls for glass or translucent layers:

  • go beyond raw `backdrop-blur`
  • add a subtle inner border
  • add a restrained inner highlight or inset shadow

Use atmosphere intentionally:

  • gradient meshes
  • noise overlays
  • soft vignettes
  • layered transparency
  • controlled shadows

But:

  • keep grain and noise on fixed overlays, not scrolling containers
  • do not turn every section into a s
Read more
@cowork-os
Ships withcowork-os

Rent the model. Own everything else. Free, open-source AI app for real work.

Stars on GitHub

Momentum

stars in total
544
Star history of CoWork-OS/CoWork-OS on GitHub, a chart by star-history.com

Other skills on cowork-os.

attack-path-analysis

attack-path-analysis

Use when Codex is already in the attack-path-analysis phase of a security scan or the user explicitly asks to trace a security…

@cowork-os@cowork-osInstall Skill
deep-security-scan

deep-security-scan

Use when the user asks for a deep, exhaustive, multi-pass, or variance-reducing repository-wide Codex Security scan. Run repeated…

@cowork-os@cowork-osInstall Skill
finding-discovery

finding-discovery

Use when Codex is already in the finding-discovery phase of a security scan or the user explicitly asks to discover candidate…

@cowork-os@cowork-osInstall Skill
fix-finding

fix-finding

Use when the user explicitly asks to fix and verify a validated or plausible security finding. Do not use as the primary trigger…

@cowork-os@cowork-osInstall Skill
security-diff-scan

security-diff-scan

Use when the user asks for a security review of a pull request, commit, branch diff, working-tree patch, or other Git-backed…

@cowork-os@cowork-osInstall Skill

More skills from other plugins.

i-have-adhd

i-have-adhd

Shape output for a reader with ADHD: lead with the next action, number multi-step work, restate state across turns, suppress…

@ayghri@ayghriInstall Skill
ab-test-analysis

ab-test-analysis

Analyze A/B test results with statistical significance, sample size validation, confidence intervals, and ship/extend/stop…

@phuryn@phurynInstall Skill
distilly

distilly

Set up the owner profile so a personal agent knows who it works for (first-run setup, \"get to know me\"), or distill colleague,…

Questions about taste-skill.

Which coding agents can use taste-skill?
taste-skill installs in Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and Windsurf with the skills CLI, using each agent's flag shown on this page. Flowy's auto-invocation, which fires a skill when your prompt matches, runs in Claude Code only.