Skip to content
Sign inJoin waitlist

Getting started — Kendi AI Prompts

The pack builds Kendi one prompt at a time. Each prompt describes one part of the site exactly — sizes, spacing, text styles, motion, content — and names the screenshots that show how it should look. This guide takes you from the download to a finished site.

  • An AI tool that writes and runs a website: a chat builder (Lovable, Bolt, v0, Replit) or a coding agent (Cursor, Claude Code, Codex, Windsurf). Notes for each are in Notes for each tool.
  • A plan with room for about 35 to 45 messages: the 25 prompts, plus a few follow-ups to fix details. Free plans usually run out before the end.
  • A web browser to open Prompts.html.
Folder or file What it is for
Prompts.html Open this first: every prompt with a Copy button, the screenshots to attach, and your progress
prompts/ The same prompts as Markdown files (01-foundation.md … 25-finish.md)
references/ Screenshots of every page layout: home-desktop-1.png (1200 px wide, top to bottom), home-tablet-1.png (810 px) and home-phone-1.png (390 px wide, strips side by side)
content/ The demo text as JSON: site.json, pages/*.json, collections/*.json
assets/ images/ (placeholders at the right sizes, the logo and the sponsored ad’s logo), icons/, icon.svg (the browser tab icon) and apple-icon.png (the home-screen icon)
checklist.md Section heights per page, for agents that can measure what they built
DESIGN.md The design tokens and style rules in one file, for coding agents
# Prompt What it builds
1 Foundation The project, the design tokens (colours, text styles, spacing, motion), the rules for breakpoints, images and the two kinds of form, the two logo files and the icons
2 Components Buttons and their rolling arrows, text links, category pills, the logo and icons, the menu links, the Categories menu and the hamburger, the breadcrumb, the post cards (stacked, row, small, over a photo), the Spotlight items, the category and Instagram tiles, the sponsored ad, the team card, the slideshow, the photo stack, the newsletter form, the contact form’s fields and button, rich text and the list that loads on scroll — everything the pages are made of
3 Layout Header (the fixed white bar, the Categories menu and the phone menu) and footer with its Instagram photos, and the site details (site.json)
4–13 Content The collections: the 64 posts (in ten parts, with their full text), then the categories, authors, team, Instagram photos and legal pages
14–24 Pages One prompt per page layout: Home, About, Contact, Categories, Category, Authors, Author, Blog, Post, Legal, 404
25 Finish Browser titles and descriptions, share image, tab and home-screen icons, sitemap and robots, a last check of every page

Each prompt builds on the ones before it, so keep the order. Some page prompts reuse a section written in an earlier one (“The Sidebar section from prompt 14 — reuse it”): the Breadcrumb comes from About (prompt 15), the Sidebar, the Sponsored ad and the Newsletter band from Home (prompt 14).

  1. Open Prompts.html and choose Lovable, Bolt, v0, Replit at the top.
  2. Start a new, empty project in your tool.
  3. Copy prompt 1. In the tool’s message box, attach the screenshots listed on the card (drag them from the references folder, or use the attach button), paste the prompt and send it.
  4. Wait until the tool has finished and the preview works. Compare it with the screenshots; if something is off, ask for that one fix in a short message (“On desktop, the Contact us button should sit at the right end of the header, as in the screenshot”).
  5. Tick the prompt in Prompts.html and go on with the next one.

Before the page prompts, add the pack’s placeholder photos to the project: upload the assets/images folder into public/images/. The prompts expect them there, and the pages only take the right shapes once they are.

The Content prompts (4–13) are long on purpose: they carry the full text of every post, and the categories, authors, team, Instagram photos and legal pages. Send them as they are. The tool saves them as files and shows nothing new yet — that is expected. So is one thing after the Layout prompt: the header’s Categories menu and the footer’s Categories column and Instagram photos stay empty until the Content prompts have saved the collections.

Some page prompts list more screenshots than a tool takes in one message (the Blog has seven desktop slices and six phone images). Attach the desktop ones first; if your tool has a limit, send the rest in a short follow-up (“Here are the phone screenshots for the same page”).

  1. Copy the whole pack into your project folder and name it kendi/.
  2. Open Prompts.html, choose Cursor, Claude Code, Codex, Windsurf at the top, and copy the kickoff prompt.
  3. Send it to the agent. It copies the content, the images and the icons into the project, then works through the prompts in order, opening the screenshots for each page. The Content prompts are skipped — the files are already there.
  4. Check each page as it lands, in the browser, at phone and desktop width. Agents that can run a browser can also compare their result with kendi/checklist.md.
  • Put the site next to the page’s screenshots at 1200, 810 and 390 px wide: same sections, in the same order, with the same alignment and spacing.
  • Resize the window slowly from phone to desktop: the layout changes at 810 px and 1200 px and never scrolls sideways.
  • On a computer, hover the buttons (the label slides, the arrow rolls), the menu links, the cards (the photos grow), the footer links and the Instagram photos; open the Categories menu in the header; let the hero slider and the Inspiration carousel step on their own and use their arrows; scroll a post and see the sidebar stay in view on desktops; scroll Articles and see more posts load; send each form in demo mode; follow a link — the page changes at once, without a transition, and the header stays fixed at the top.

Photos are grey placeholders until you add your own — see Making it yours.