Find agent skills by outcome

RIGHT: if data fetch takes >300ms

show skeleton. If fetch fails

If Tailwind: use Tailwind classes as primary

px as annotation

If CSS: use rem as primary

px as annotation

Before writing any specs

follow these formatting rules exactly:

**If the reader has to guess

infer

must have a rule. No as appropriate

no roughly

Your output will be read primarily by AI coding agents (Claude Code

Cursor)

Relationship to CLAUDE.md: This document assumes a CLAUDE.md design system file already exists in the project root. Handoff Notes reference tokens from CLAUDE.md but don't redefine them. If no CLAUDE.md exists

run the Design System Extraction prompts first.

- Tertiary: You (the designer)

when checking if implementation matches intent

- Primary: AI coding agents (Claude Code

Cursor

Description: Generates a design handoff document that serves as direct implementation instructions for AI coding agents. Unlike traditional handoff notes that describe how a design should feel

this document provides machine-parseable specifications with zero ambiguity. Every value is explicit

The core philosophy: If an AI reads this document and has to guess anything

the document has failed.

When to use: After design is finalized

before implementation begins. This replaces Figma handoff

### A structured handoff document optimized for AI implementation agents (Claude Code

Cursor

page.on('console'

msg => console.log('Browser log:'

Design Handoff Notes - AI First, Human Readable

# Design Handoff Notes — AI-First, Human-Readable

await page.waitForSelector('#element-id'

{ state: 'visible' });

await page.screenshot({ path: 'debug.png'

fullPage: true });

await page.fill('#password'

'password123');

await page.fill('#username'

'testuser');

Web Application Testing Skill (Imported)

---

console.log('Page title:'

title);

1. Inspect the existing codebase

components

Improve visual hierarchy

clarity