- Home
- Skills
- Frontend & Web Apps
- Turn Your AI Agent into a Senior UI Reviewer
More screenshots
Works with the AI Agents you already use
Turn Your AI Agent into a Senior UI Reviewer
Most UI problems that cost you sign-ups are measurable: grey text at 2.3:1 contrast, a 23 px button nobody can tap, an email field with no label.
$9.99
Turn Your AI Agent into a Senior UI Reviewer
Example session with this skill installed
Our landing page is sample/bad.html. Before I send it to the developer, check it against the UI rules and tell me what to fix. It has no dark theme and no dashboard.
- Read your context and instructions
- Compiled the turn your ai
- Generated the app files
Example output (real run on sample/bad.html)
The script output is saved in sample/bad-audit-output.txt; this report is built from it plus a look at sample/bad-phone.png.
UI audit: sample/bad.html (8 Nov 2026)
Verdict: FIX FIRST Score: 52/100 over 27 rules (11 PASS, 6 WARN, 10 FAIL)
Out of the score: 4 N/A, 4 NOT CHECKED. Groups judged: typography, spacing, colour, interface states, mobile. Dark mode, dashboards and process rules were left out on purpose (no dark theme, no dashboard, one page).
Score working: (11 + 0.5 x 6) / 27 = 14 / 27 = 51.9, rounded to 52.
| Group | PASS | WARN | FAIL | N/A | Not checked |
|---|---|---|---|---|---|
| Typography (8) | 3 | 3 | 2 | 0 | 0 |
| Spacing (6) | 3 | 2 | 0 | 0 | 1 |
| Colour and contrast (8) | 2 | 0 | 3 | 2 | 1 |
| Interface states (7) | 1 | 0 | 2 | 2 | 2 |
| Mobile (6) | 2 | 1 | 3 | 0 | 0 |
| Total (35) | 11 | 6 | 10 | 4 | 4 |
Findings (FAIL first)
| ID | Rule | Viewport | Measured | Element | Fix |
|---|---|---|---|---|---|
| CC1 | Text contrast 4.5:1 | both | 2.32:1 | p.lead (#aaa on white) | Use #222 (15.91:1) |
| CC1 | Text contrast 4.5:1 | both | 3.32:1 | both inline links (#2a9d8f on white) | Use #1d6f64 (5.98:1) |
| CC2 | Button text contrast | both | 3.32:1 | form>button, white on #2a9d8f | Button fill #1d6f64, white text (5.98:1) |
| CC5 | Links marked beyond colour | both | no underline | links "story", "origins" | text-decoration: underline |
| TY1 | Running text 16 px on phone | phone | 14 px and 12 px | p.lead, span.tag | 16 px for both |
| TY5 | Uppercase for 4 words or fewer | both | 12 words in capitals | span.tag | Sentence case, shorter: "Free shipping on orders over 30 pounds" |
| ST1 | Visible label on every input | both | placeholder "Email" only | form>input | Add <label for="em">Email address</label>, give the input id="em" |
| ST2 | Button names its outcome | both | "OK" | form>button | "Start my subscription" |
| MO1 | Touch target 44 px, never under 24 | both / phone | button 35 x 23; input 212 x 27 | form>button, form>input | min-height: 48px; button min-width: 120px |
| MO2 | Form text 16 px | both | 13 px | form>input | font-size: 16px |
| MO3 | Right keyboard | both | type="text" on an email field | form>input | type="email" autocomplete="email" |
| TY2 | Line height 1.4 to 1.6 | both | 1.10 | p.lead (body rule) | line-height: 1.5 on body |
| TY3 | 45 to 75 characters per line | phone | 116 per line | p.lead | Shorter text and max-width: 36rem; the real fix is cutting the paragraph |
| TY7 | Tighten large headings | both | 40 px, spacing normal | h1 | letter-spacing: -0.02em |
| SP1 | One spacing scale | both | 14 (6 uses), 6 (4), 13 (4), 10 (3) | page-wide | Move to 8, 16, 24 |
| SP2 | Group gap 1.5 x the gap inside a group | eye | promo line touches the form; blocks 10 to 14 px apart | span.tag, form | 16 to 24 px between blocks, 8 inside the form |
| MO5 | Main a |
bad.html
ZIP · project files
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
About this skill
"Looks fine to me" is not a test. Most UI problems that cost you sign-ups are measurable: grey text at 2.3:1 contrast, a 23 px button nobody can tap, an email field with no label. This skill turns your agent into a reviewer that measures them instead of giving opinions.
It checks any landing page, app screen or dashboard against 50 concrete rules in 8 groups: typography, spacing, colour and contrast, interface states, mobile, dark mode, dashboards and process.
The proof
A real run on a sample coffee-subscription landing page (the files ship inside the skill, so you can check them):
- Before: 52/100, 10 FAIL and 6 WARN: text contrast 2.32:1, a 35 x 23 px button, a placeholder instead of a label, 13 px form text, a button that just says "OK".
- After the fixes: 0 FAIL, 0 WARN at desktop and phone width, re-checked by the same script.
What you get
- SKILL.md for Claude Code and OpenAI Codex
- RULES.md: 50 rules, each with a number or a yes/no test, thresholds and sources
- ui_audit.py: a Playwright script that measures the page at desktop and phone width
- A report format: PASS / WARN / FAIL per rule, the element, the measured value and the exact fix
- Sample pages: a faulty page, its fixed version, the real audit outputs and phone screenshots
- Testing notes and a changelog
How to use
- Install once:
pip install playwrightthenplaywright install chromium(about two minutes). - Ask your agent: "Audit this page against the UI rules", with a URL or a local HTML file.
- Read the scored report, let your agent apply the fixes, then ask it to re-check.
Known limitations
- A screening tool, not an accessibility certificate.
- Some rules need a human eye (visual hierarchy, image quality). The report marks them as eye checks and never guesses.
- Web pages only: it does not inspect native iOS or Android apps.
- The script needs Python 3.9+ and Playwright.
FAQ
My agent can already review UI. Why this? It can give opinions. This gives measurements: the contrast ratio, the tap-target size in pixels, the exact element, so fixes are right first time and can be re-checked.
Does it work on dashboards and dark mode? Yes. Six rules cover dashboards (units, scales, chart clarity) and six cover dark mode.
Will it change my code? Only if you ask your agent to apply the fixes. The audit itself only reads the page.
Mac or Linux? Yes. Tested on Windows 11; Playwright runs the same on macOS and Linux.
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.
Requires Python 3.9+ and Playwright installation. Cannot measure elements behind logins without manual HTML export. Text over images or gradients results in unmeasured contrast.
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 7 days ago
- Software
PythonPlaywrightChromiumpip
Network access is only used to open the page you ask it to audit; local files need none. Nothing is sent anywhere.