nadav-design

Orchestrates end-to-end UI/UX setup, planning, screen building, auditing, and release gating for Hebrew RTL web projects.

Updated Sep 15, 2026
One-click install
npx skills add https://github.com/Nadav011/nadavai-claude --skill nadav-design-nadav011
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: nadav-design
Source: https://github.com/Nadav011/nadavai-claude/tree/main/plugins/nadavai/skills/nadav-design
Command: npx skills add https://github.com/Nadav011/nadavai-claude --skill nadav-design-nadav011

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Setting up and maintaining a consistent, high-quality UI/UX system on a Next.js or Vite project involves dozens of error-prone steps: RTL infrastructure, fonts, design tokens, accessibility gates, Lighthouse budgets, and release checks. This Skill runs that entire loop one verified step at a time, with evidence at every step, so nothing is skipped, assumed, or silently broken. ## Core Features & Use Cases - Seven workflow modes: setup (one-time project foundation: RTL, fonts, tokens, DESIGN.md, PRODUCT.md, DoD gate, CI wiring), plan (feature PRD and mockups), screen (build or redesign one screen with critique and hardening), audit (map every screen), ship (pre-release measurement), retro (process review), and next (gap scan showing where the project stands). - Hard quality gates: DoD score of 100 across every route via Playwright and axe, Lighthouse 100 thresholds, zero lint findings, and a mandatory visual change gate requiring before/after screenshots at 375px and 1440px for any user-visible change. - Hebrew RTL-first enforcement: logical CSS properties only, dir="ltr" islands for numbers and Latin names, Hebrew formatting via Intl he-IL, and multi-brand support through per-vertical token scopes. - Use Case: On an existing React app, run setup to archive legacy design artifacts, migrate to shadcn RTL mode, install a Hebrew font, wire tokens and a styleguide route, and establish a CI gate that fails any pull request below the quality bar. ## Quick Start Ask the assistant to run /ux-setup on this project to configure the complete UI/UX environment step by step.

Frequently Asked Questions about nadav-design

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I set up RTL support for a Hebrew Next.js or Vite app?

Run the setup mode, which enables shadcn RTL via components.json, sets dir="rtl" and lang="he" on the shell, installs a Hebrew font such as Rubik or IBM Plex Sans Hebrew, and enforces logical CSS properties through lint rules and a pre-commit hook.

How do I enforce a UI quality gate in CI for a React app?

The setup mode installs a Definition of Done suite using Playwright and axe that scores every route of the app, plus a Lighthouse CI configuration asserting 100 on performance, accessibility, best-practices, and SEO. The GitHub Actions workflow fails any pull request below the thresholds.

Does this workflow support Vite projects or only Next.js?

Both are supported. Step 1 of setup detects the framework family from the lockfile and config, then records the token file, shell entry, UI directories, package manager, and dev command so every later step uses the correct paths for that project shape.

Can I use this with a multi-brand or multi-tenant design system?

Yes. Setup detects brand scopes such as data-brand attributes or per-vertical build modes and asks before configuring them. Shared foundations live in a root DESIGN.md while each vertical gets its own token scope and design file, with DoD runs executed per vertical.

What happens when a visual change is made automatically by a tool?

A visual change gate applies to any user-visible change, including ones made by eslint --fix, shadcn migrate, or token replacements. The workflow stops, shows before and after screenshots at 375px and 1440px, and waits for an explicit keep, change, or revert decision.

Why does the Lighthouse step refuse to report a score sometimes?

Scores are only measured on an idle machine because load can swing performance results several-fold. If the 1-minute load average is too high or the PageSpeed Insights API quota is exhausted, the run waits or falls back to local lhci rather than reporting an unreliable number.