- Home
- Skills
- Frontend & Web Apps
- Website Design Auditor: Make AI-Built Sites Look Pro
More screenshots
Works with the AI Agents you already use
Website Design Auditor: Make AI-Built Sites Look Pro
Turn a rough AI-built page into a polished one: renders it at phone, tablet and desktop, scores the design 0-100 and hands your agent a ranked fix list with annotated screenshots.
$19
About this skill
Your agent built the landing page, said it looks great, and on a phone the pricing table scrolls sideways, a badge sits on the headline and the grey copy is unreadable. It never saw the page.
Measured, not guessed: on a typical AI-generated landing page it found 25 problems (7 HIGH, score 0/100); after applying its fix list in order, the same page scored 100/100 with 0 findings.
Website Design Auditor gives your agent eyes. One Python script renders the page in headless Chromium at 375, 768 and 1440 px, measures every visible element and returns:
- A ranked fix list with the CSS selector, the viewports, the measured fact and a concrete fix (
color: #9ca3af -> #70757e,repeat(auto-fit,minmax(min(100%,16rem),1fr))). - Annotated screenshots per viewport: numbered boxes match the fix list.
- A 0-100 score with a published formula, so the agent loops: audit, fix, re-audit until PASS.
- 14 checks: sideways overflow (with the culprit element), text overlapping text, clipped text, WCAG contrast sampled where the text sits on gradients, tap targets under 44 px, font and colour sprawl, near-duplicate colours, spacing off the 4/8 px grid, images without alt or size, layout shift, lines over 80ch, heading structure, viewport meta.
- Design rules in the skill (type scale, spacing scale, palette tokens) that the agent applies while fixing.
Benchmark (real runs, pinned commits):
| Page | Score | Findings |
|---|---|---|
| Seeded AI landing page, before / after | 0 / 100 | 25 / 0 |
| StartBootstrap agency | 69 | 12 |
| tailwindtoolbox Landing-Page | 75 | 10 |
| StartBootstrap landing-page | 74 | 9 |
| news.ycombinator.com (live) | 64 | 18 |
It caught white on #ffc800 buttons at 1.55:1, white text on a gradient's gold end at 2.22:1, and a page 804 px wide on a 768 px tablet. Six false-positive patterns these sites exposed were fixed and covered by tests (36 tests). Benchmark runs took 5-24 s for all three viewports (up to 35 s on a heavily loaded machine).
What it does not do: judge taste (a page can pass and still be bland), score text over photos (reported as unknown), or test hover and open-menu states. Without Playwright it runs a partial static check and says so.
Works in Claude Code, Cursor, Codex CLI, Copilot and any SKILL.md agent.
Private by design: no telemetry, nothing uploaded; it only opens the page you point it at and what that page loads (--offline blocks all network) and never writes into your project.
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.
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 27 days ago
- Software
Python
Git
Permissions
Allowed Hosts