Design Loop

Design Loop — Autonomous Site Builder

Build complete multi-page websites through an autonomous loop. Each iteration reads a task, generates a page, integrates it, verifies it visually, then writes the next task to keep going.

Overview

The Design Loop uses a "baton" pattern — a file (.design/next-prompt.md) acts as a relay baton between iterations. Each cycle:

1. Reads the current task from the baton
2. Generates the page (via Claude or Google Stitch)
3. Integrates into the site structure (navigation, links)
4. Verifies visually via browser automation (if available)
5. Updates site documentation
6. Writes the NEXT task to the baton — keeping the loop alive

This is orchestration-agnostic. The loop can be driven by:

• Human-in-loop: User reviews each page, then says "next" or "keep going"
• Fully autonomous: Claude runs continuously until the site is complete
• CI/CD: Triggered on .design/next-prompt.md changes

Generation Backends

Detecting Stitch

At the start of each loop, check if Stitch is available:

1. Check if @google/stitch-sdk is installed: ls node_modules/@google/stitch-sdk 2>/dev/null
2. Check if STITCH_API_KEY is set in .dev.vars or environment
3. Check if .design/metadata.json exists (contains Stitch project ID)

If all three are present, use Stitch. Otherwise, fall back to Claude generation.

Stitch SDK Reference

Install: npm install @google/stitch-sdk. Set STITCH_API_KEY in environment or .dev.vars.

import { stitch } from "@google/stitch-sdk";

// Create a project
const result = await stitch.callTool("create_project", { title: "My Site" });

// Reference an existing project
const project = stitch.project("4044680601076201931");

// Generate a screen
const screen = await project.generate("A modern landing page with hero section", "DESKTOP");

// Get assets
const htmlUrl = await screen.getHtml();    // Download URL for HTML
const imageUrl = await screen.getImage();  // Download URL for screenshot

// Edit an existing screen (prefer this for refinements)
const edited = await screen.edit("Make the background dark and enlarge the CTA button");

// Generate variants
const variants = await screen.variants("Try different colour schemes", {
  variantCount: 3,
  creativeRange: "EXPLORE",     // "REFINE" | "EXPLORE" | "REIMAGINE"
  aspects: ["COLOR_SCHEME"],    // "LAYOUT" | "COLOR_SCHEME" | "IMAGES" | "TEXT_FONT" | "TEXT_CONTENT"
});

Device types: "MOBILE" | "DESKTOP" | "TABLET" | "AGNOSTIC". Model selection: pass "GEMINI_3_PRO" | "GEMINI_3_FLASH" as third arg to generate().

Other operations: stitch.projects() lists projects, project.screens() lists screens, project.getScreen("id") fetches one.

getHtml() and getImage() return download URLs. Append =w1280 to image URLs for full resolution. Auth: STITCH_API_KEY required (or STITCH_ACCESS_TOKEN + GOOGLE_CLOUD_PROJECT for OAuth). Errors throw StitchError with codes: AUTH_FAILED, NOT_FOUND, RATE_LIMITED.

Stitch Project Persistence

Save Stitch identifiers to .design/metadata.json so future iterations can reference them:

{
  "projectId": "4044680601076201931",
  "screens": {
    "index": { "screenId": "d7237c7d78f44befa4f60afb17c818c1" },
    "about": { "screenId": "bf6a3fe5c75348e58cf21fc7a9ddeafb" }
  }
}

Stitch integration tips:

1. Persist project ID in .design/metadata.json — don't create a new project each iteration
2. Use screen.edit() for refinements rather than full regeneration
3. Post-process Stitch HTML — replace headers/footers with your shared elements
4. Include DESIGN.md context in prompts — Stitch generates better results with explicit design system instructions

Getting Started

First Run: Bootstrap the Project

If .design/ doesn't exist yet, create the project scaffolding:

1. Ask the user for:

   • Site name and purpose
   • Target audience
   • Desired aesthetic (minimal, bold, warm, etc.)
   • List of pages they want
   • Brand colours (or extract from existing site with /design-system)

2. Create the project files:

project/
├── .design/
│   ├── SITE.md           # Vision, sitemap, roadmap — the project's long-term memory
│   ├── DESIGN.md         # Visual design system — the source of truth for consistency
│   └── next-prompt.md    # The baton — current task with page frontmatter
└── site/
    └── public/           # Production pages live here

3. Write SITE.md from the template in the "SITE.md Template" section below
4. Write DESIGN.md — either manually from user input, or use the design-system skill to extract from an existing site
5. Write the first baton (.design/next-prompt.md) for the homepage

Subsequent Runs: Read the Baton

If .design/next-prompt.md already exists, parse it and continue the loop.

The Baton File

.design/next-prompt.md has YAML frontmatter + a prompt body:

---
page: about
layout: standard
---
An about page for Acme Plumbing describing the company's 20-year history in Newcastle.

**DESIGN SYSTEM:**
[Copied from .design/DESIGN.md Section 6]

**Page Structure:**
1. Header with navigation (consistent with index.html)
2. Hero with company photo and tagline
3. Story timeline showing company milestones
4. Team section with photo grid
5. CTA section: "Get a Free Quote"
6. Footer (consistent with index.html)

Execution Protocol

Step 1: Read the Baton

Read .design/next-prompt.md
Extract: page name, layout, prompt body

Step 2: Consult Context Files

Before generating, read:

Critical: Read the most recent page's HTML to extract the exact header, navigation, and footer markup. New pages must use identical shared elements.

Step 3: Generate the Page

Option A: Claude Generation (Default)

Generate a complete HTML file using Tailwind CSS (via CDN). The page must:

• Match the design system from .design/DESIGN.md exactly
• Reuse the same header/nav/footer from existing pages (copy verbatim)
• Be self-contained — single HTML file with Tailwind CDN, no build step
• Be responsive — mobile-first, works at all breakpoints
• Include dark mode if the design system specifies it
• Use semantic HTML — proper heading hierarchy, landmarks, alt text
• Wire real navigation — all nav links point to actual pages (existing or planned)

Write the generated file to site/public/{page}.html.

Option B: Stitch Generation (If Available)

If Stitch SDK is available:

1. Build the prompt by combining the baton body with the DESIGN.md system block
2. Call project.generate(prompt, deviceType) to generate the screen
3. Download the HTML from screen.getHtml() to .design/designs/{page}.html
4. Download the screenshot from screen.getImage() to .design/screenshots/{page}.png
5. Post-process the Stitch HTML:
   • Replace the header/nav/footer with your project's shared elements
   • Ensure consistent Tailwind config
   • Wire internal navigation links
6. Save the processed file to site/public/{page}.html
7. Update .design/metadata.json with the new screen ID

For iterative edits on an existing Stitch screen, use screen.edit(prompt) instead of regenerating.

Step 4: Integrate into the Site

After generating the new page:

1. Update navigation across ALL existing pages — add the new page to nav menus
2. Fix placeholder links — replace any href="#" with real page URLs
3. Verify cross-page consistency — header, footer, nav must be identical everywhere
4. Check internal links — no broken links between pages

Step 5: Visual Verification (If Browser Available)

If Playwright CLI or Chrome MCP is available:

1. Start a local server: npx serve site/public -p 3456
2. Screenshot the new page at desktop (1280px) and mobile (375px) widths
3. Save screenshots to .design/screenshots/{page}-desktop.png and {page}-mobile.png
4. Compare visually against the design system
5. Fix any issues (broken layout, wrong colours, inconsistent nav)
6. Stop the server

If no browser automation is available, skip to Step 6.

Step 6: Update Site Documentation

Edit .design/SITE.md:

• Mark the page as complete in Section 4 (Sitemap): [x] {page}.html — {description}
• Remove any consumed item from Section 5 (Roadmap) or Section 6 (Ideas)
• Add any new ideas discovered during generation