A multidisciplinary web-excellence consultant that runs inside Claude Code — reviewing performance, accessibility, SEO, AI-discoverability, UX, design, Webflow and conversion in a single, repeatable pass.
Press → to begin · O for overview · keyboard-navigable & reduced-motion aware.
Performance, SEO and accessibility each live in a different tool — so nobody sees how a fix in one quietly breaks another.
Every reviewer checks different things in a different order — so you can't compare two pages, or track the same page over time.
Manual reviews don't scale and never run the same way twice — quietly missing the issues that cost rankings and conversions.
Growth needs pages that are fast, findable — by Google and AI — and that convert. This skill checks every discipline together, every time — so nothing slips through.
It runs inside Claude Code. Point it at a page and it runs 10 expert checks on it, then combines them into one review — sorting out any places where the checks disagree.
You get one to-do list, sorted by what matters most — not nine separate opinions, and never one thing fixed at the quiet expense of another.
It doesn't do the checks itself. It picks the right expert checks, gives them all the same evidence, weighs the trade-offs, and turns everything into one prioritized report.
Core Web Vitals sit inside Performance — 10 specialist skills feed these 9 disciplines.
Gathering the evidence once (step 3) is what lets all nine checks run cheaply — and keeps every finding tied to the same facts.
It sees which of the 10 expert checks are installed and matches each to a topic. If one is missing, it flags where the review will be lighter.
It works out what you gave it — a link, a whole site, a design file, a screenshot, a report, or code — and what you want: a review, a plan, or actual fixes.
It loads the page one time and records everything: the finished page, the raw HTML before scripts, all network requests, the console, desktop + mobile screenshots, real-user speed, and the findable/blocked check.
All nine topics review that same evidence at once. Each writes its findings the same way: an ID, how serious it is, the effort to fix, the proof, and the fix.
It merges everything, sorts out cases where two checks disagree, scores each finding by impact, effort and risk, and groups them into four priority tiers.
You get a short summary first. The full report, the checklist, and a PDF or HTML version are there whenever you ask for them.
Fixing one thing often hurts another. Instead of quietly picking a side, it shows the trade-off and tells you what to do.
Everything is captured once, up front — so every finding is backed by proof, not a guess.
Loads the page and records the network requests, console errors, a performance trace, and desktop + mobile screenshots.
Scores the page on Performance, Accessibility, Best Practices and SEO.
Real-user speed from Google (CrUX): LCP, INP and CLS from people who actually visited.
Grabs the raw HTML before scripts run, plus robots.txt and sitemap.xml — the “can it be found / is it blocked” check.
Pull out headings, image alt text, colour contrast and structured data straight from the page.
One shared set of facts — reused by all nine checks, so nothing is reviewed twice or guessed at.
Same skill — you decide whether it just reviews the page, or actually makes changes to it.
Point it at a page and it reviews everything, then hands you the report and a to-do list sorted by priority. Nothing on the page changes.
Use it to
Find out how good a page is, and exactly what to fix first.
Ask it to actually make the changes. On a live page it works carefully — one change at a time, checking nothing breaks:
Then it moves to the next change — so you always know each fix helped and didn't break anything.
| Metric | Field p75 | Target | Status |
|---|---|---|---|
| LCP | 2,592 ms | < 2,500 | Needs work |
| INP | 76 ms | < 200 | Good |
| CLS | 0.04 | < 0.1 | Good |
| robots.txt — URL allowed? | allowed · AI crawlers not blocked |
| In sitemap.xml? | listed |
| Structured data | Organization only · stale logo |
Real items from the Landbot homepage audit. Click an item to check it off.
Within each tier: order by impact first, then risk, then effort.
Desktop and mobile ship two different hero sections — different headline, value prop and CTA — with the desktop <h1> hidden via display:none on mobile. That single decision hurts four disciplines at once:
A perfect 1.00 Cumulative Layout Shift means the page visibly jumps under the reader on every visit — and it's in the field data, not just the lab.
Set explicit width & height on those 1,158 images.
One quick win moves the worst metric on the entire site — reserved space stops the layout collapsing as images load.
Because every report has the same shape, you can put two runs side by side and see exactly what moved.
| What we measure | Jun 18 · first run | Aug 1 · after quick wins | Change |
|---|---|---|---|
| Overall grade | B+ | A− | Up |
| LCP · load speed (real users) | 2.59 s | 2.31 s | −0.28 s |
| CLS · layout jump (real users) | 0.04 | 0.03 | still good |
| Muted-text contrast | 3.36 : 1 · fails | 4.6 : 1 · passes | fixed |
| Console errors | 1 | 0 | cleared |
| Structured data | Organization only | + SoftwareApplication | added |
| Open “High” issues | 4 | 1 | −3 |
The second column is an example re-run after shipping the Quick Wins — same audit, same page, a later date. Each new run just adds a column.
Google uses page speed as a ranking signal, and slow pages lose sales. On phones — where most people visit and pages are slowest — speed decides whether someone waits or leaves.
ChatGPT, Perplexity, Claude and Google's AI answers now reply to buyers directly, quoting pages they can easily read and trust. These checks make sure your pages can be read, your info is machine-readable, and AI isn't blocked from your site.
The same checks in the same order every time means results don't depend on who ran them. You can compare pages and track one over time, the easy wins are obvious, and the team stops re-hashing the same debates.
Every problem gets an ID, a severity and an effort tag — so marketing, engineering and design all describe it the same way. Less back-and-forth, fewer “is this really a bug?” threads, faster hand-offs.
Every quarter a fixable CLS 1.00, a blocked AI crawler, or a missing review score stays live is discovery and sales you don't get back — on your most important pages. Running the check is cheap and repeatable; leaving it isn't.
Even a small question runs the full check — so nothing gets missed.
In Claude Code: “audit <url>” → review the exec summary → “generate the report” → ship the Quick Wins first.
A senior web consultant that reviews every page across every discipline — fast, consistent, and evidence-backed.
Thanks — questions welcome. Press O to jump to any slide.