- Home
- Skills
- Frontend & Web Apps
- Screenshot to UI Code: Clean HTML/CSS/React from an Image
More screenshots
Works with the AI Agents you already use
Screenshot to UI Code: Clean HTML/CSS/React from an Image
- Pulls design tokens (colors, type scale, spacing, radii) into :root variables, so one edit restyles the page.
$5.99
Screenshot to UI Code: Clean HTML/CSS/React from an Image
Example session with this skill installed
Rebuild this pricing page from the screenshot. Plain HTML/CSS, no libraries.
- Read your context and instructions
- Compiled the screenshot to ui
- Generated the design tokens stylesheet
Summary
Rebuilt the pricing page from the screenshot (1280×800, fictional "Lumen" page) as one self-contained index.html: header with nav, centered headline, three plan cards (Starter $9, Pro $29 featured, Team $79) and a footer note. Plain HTML/CSS, no libraries. All colors, spacing and type sizes live in :root variables.
Inventory (written before coding)
- Header: logo mark + "Lumen", nav "Product / Pricing / Docs / Blog", button "Start free"
- Hero: h1 "Simple pricing for growing teams", subtitle "Pick a plan today. Change it any time."
Cards: STARTER $9 / month (3 features, outline button "Choose Starter"); PRO $29 / month with a "MOST POPULAR" badge (4 features, filled button "Choose Pro"); TEAM $79 / month (3 features, outline button "Choose Team")
- Footer note: "Cancel anytime. No hidden fees."
- Icons: one check-in-circle icon, repeated in every feature list (decorative)
Report
Built: index.html (new file, nothing overwritten)
Exact (read straight from text/structure): all visible strings, 3 plans in left-to-right order,
Pro is the featured one, 3/4/3 feature counts, nav order.
Estimated (from pixels): colors (#f5f6fb, #e6e8f2, #12132b, #6366f1 ...), font sizes, spacing on a 4px grid,
radii. I could only look, not sample pixels. Send me the real design tokens and I will replace them in one
place (:root).
Decisions you didn't see in the image:
- Responsive, mobile-first, with four breakpoints: under 640px smaller headline; under 720px the header
wraps (logo + button on row 1, nav links on row 2); under 960px the cards stack in one column;
960px and up, three columns with the Pro card raised.
- Hover states, :focus-visible outlines, button transitions (switched off for prefers-reduced-motion).
- "Pricing" link marked aria-current="page" (assumed). Links use href="#" as placeholders.
Not done: React version (not requested), dark mode, real links, mobile menu, monthly/annual toggle (not in the image).
Not rendered: I did not open the page in a browser. The three places most likely to differ:
1. Card heights and vertical rhythm
2. Type widths (system font vs the original)
3. Small colors and the check-icon shape
Checklist: one h1, landmarks, lists, buttons and links = pass; decorative icons aria-hidden = pass;
focus outlines = pass (read from code, not tested in a browser);
contrast: footnote #8a8ca5 on the page background is about 3.0:1 = FAIL, reproduced as asked;
white on #6366f1 is about 4.47:1 = borderline; 320px layout = not checked in a browser.
The full index.html is the deliverable of this run; this listing shows the written part of the output.
index.html
CSS · design tokens
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop getting div soup from "rebuild this screenshot". Give your agent a design image and this skill makes it work like a careful front-end developer:
- Writes down every region, component and string first, so nothing is invented or dropped.
- Pulls design tokens (colors, type scale, spacing, radii) into :root variables, so one edit restyles the page.
- Builds a real responsive layout with grid and flexbox, mobile-first, and writes down every layout decision the screenshot could not show.
- Runs an accessibility pass: landmarks, one h1, real buttons and links, visible keyboard focus, contrast pairs checked.
- Ends with a report: what was exact, what was estimated from pixels, and what was not rendered.
Tested here, one run each: the same agent rebuilt the same fictional pricing page from the same screenshot, once without this skill and once with it. On desktop both matched the screenshot. On a 390 px phone the plain version broke (menu over the logo, plan cards cut off); the skill version stacked cleanly. Its CSS used variables for colors, spacing and type (97 uses versus 0), had 4 documented breakpoints and visible keyboard focus. Results vary between runs.
What you get: SKILL.md (7-step procedure), a token-extraction guide, a delivery checklist, React output rules and a complete worked example.
Honest limits: colors and sizes read from a picture are estimates (it says so). It does not render the page unless your agent can open a browser; it tells you which parts are most likely to differ. It will not invent copy, logos or photos. You attach the image; it does not capture your screen. Tested as instructions given to a Claude agent; other agents are untested.
Privacy: instructions only. Nothing is installed and the skill itself makes no network calls.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Frontend & Web Apps.
Measurements are estimates. No network access or automated screen capture. Does not generate animations or complex state logic unless visible. Focuses on structure over brand asset recreation.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 12 days ago
Permissions
Read Files: your agent reads the screenshot or file you attach. Write Files: only to create new files (index.html and optional React files); it never overwrites an existing file without showing you first. Browser is optional: only if your agent can open the page to compare. The skill itself runs no scripts and makes no network calls.