Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Core Web Vitals Fixer for LCP INP and CLS

    1

    Finds what hurts LCP, INP and CLS, maps each issue to the file that causes it, applies framework-aware fixes, and rates each metric GOOD / NEEDS IMPROVEMENT / POOR plus SHIP / FIX FIRST / BLOCK

    $7

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Core Web Vitals Fixer for LCP INP and CLS

    Core Web Vitals Fixer for LCP INP and CLS

    Example session with this skill installed

    Our Next.js 16 product page fails Core Web Vitals on mobile. PSI field data says LCP 3.8 s, INP 340 ms, CLS 0.21. Find the causes in the repo and fix them.

    • Read your context and instructions
    • Compiled the core web vitals
    • Generated the UI component

    Core Web Vitals report: /products/[slug] (PHONE)

    Stack: Next.js 16.2, App Router, SSR · Data: FIELD (your PSI paste, CrUX 28d) + LAB (Lighthouse 13, median of 3)

    Verdict: FIX FIRST

    Metricp75SourceStatus
    LCP3.8 sFIELDNEEDS IMPROVEMENT
    INP340 msFIELDNEEDS IMPROVEMENT
    CLS0.21FIELDNEEDS IMPROVEMENT

    No metric is POOR, so not BLOCK. All three miss GOOD.

    Diagnosis

    LCP element: hero <Image> in app/products/[slug]/Gallery.tsx:42. Lab LCP 4.1 s = TTFB 600 ms + load delay 1,900 ms + load 1,100 ms + render 500 ms. Load delay is the biggest part: the image has loading="lazy" and only renders after the client carousel hydrates.

    INP: no attribution data yet. A DevTools trace of the size filter click shows a 280 ms handler in components/SizeFilter.tsx:18 re-rendering 120 product cards.

    CLS: reviews widget injected above the fold (app/products/[slug]/page.tsx:77, no reserved height); fallback font swap on the H1.

    Fixes (ordered by expected impact)

    1. Render the hero server-side with high priority (LCP)

    // Gallery.tsx:42
    <Image src={hero} alt={name} fetchPriority="high" loading="eager"
      sizes="(max-width: 768px) 100vw, 50vw" />
    

    Move the first slide out of the client carousel so it is in the initial HTML.
    Expected impact (estimate): removes most of the 1,900 ms load delay seen in lab.

    2. Split the filter update (INP)

    const [isPending, startTransition] = useTransition();
    onChange={(v) => { setChecked(v); startTransition(() => setFilter(v)); }}
    

    Expected impact (estimate): handler work leaves the click's critical path; confirm with RUM.

    3. Reserve space for reviews and fix the font swap (CLS)

    .reviews-slot { min-height: 320px; } and switch to next/font/local for the H1 font.

    Measure

    npx lighthouse <url> --form-factor=mobile --output=json x3 before/after; add web-vitals/attribution RUM for INP. Field p75 settles about 28 days after deploy. Re-check CrUX on or after Nov 16, 2026.

    Gallery.tsx

    TSX · React component

    Generated

    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

    Search Console says your pages fail Core Web Vitals, and PageSpeed Insights gives you a list of generic audits. The hard part is working out which element, script or component in your repo causes the problem, and which change will actually move the 75th percentile.

    This skill gives your coding agent a fixed method for that job:

    • Pulls field data from the CrUX API or PageSpeed Insights API, and lab data from the Lighthouse CLI, or works from a report you paste. Every number is labeled FIELD or LAB.
    • Names the LCP element and splits it into TTFB, load delay, load duration and render delay, so the fix targets the biggest part.
    • Finds slow interactions for INP (input delay, processing, presentation) and the sources of layout shift for CLS.
    • Searches the repo and ties each problem to a file and line.
    • Applies fixes that fit your stack: Next.js, React, Vue/Nuxt, Astro, SvelteKit, WordPress themes or plain HTML. Covers image priority and sizing, fonts, render-blocking CSS and JS, third-party scripts, hydration, code splitting, yielding with scheduler.yield and a fallback, CSS containment, reserved ad and embed space, and bfcache.
    • Gives each fix an expected impact labeled as an estimate, plus a before and after measurement plan.
    • Rates each metric GOOD, NEEDS IMPROVEMENT or POOR against Google's p75 thresholds, then gives an overall SHIP, FIX FIRST or BLOCK.

    The reference file holds the current thresholds, the API request formats, the Lighthouse 13 audit renames, and framework patterns as of October 2026, with source links.

    Who it's for: frontend developers, agencies and site owners who need to pass Core Web Vitals and want code changes rather than a score.

    Honesty note: the skill does not invent metric values or promised savings. Lighthouse cannot measure INP, and the skill says so. Field data takes about 28 days to reflect a fix.

    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.

    Needs field data from CrUX, PSI or your own RUM to judge INP and give a pass or fail. Lab runs alone give only a provisional verdict. Impact numbers are estimates until you measure them. Fixes to third-party tags, CDN or server setup may need someone else to make them, and field data takes about 28 days to show a change.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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 1 month ago

    Nothing to install. It runs in your AI agent.

    Optional
    Node.js 22.19+Lighthouse 13Web access

    Node.js and Lighthouse are for lab runs. Web access fetches CrUX or PageSpeed Insights field data; you can paste your report instead.

    Permissions

    Terminal / Shell
    Network Access
    Read Files
    Write Files

    Allowed Hosts

    chromeuxreport.googleapis.com
    www.googleapis.com
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Two files: SKILL.md, which holds the method, the verdict rules, the output template and the guardrails, and references/cwv-facts.md, which holds the thresholds, the CrUX, PSI and Lighthouse formats, the Lighthouse 13 audit mapping, the web-vitals RUM snippet, and the framework fix patterns, with source URLs.

    Any agent that reads the SKILL.md format, such as Claude Code, Codex CLI, Cursor, VS Code Copilot and Gemini CLI. How you install and trigger it depends on the agent.

    Nothing is required. Paste a PSI or Lighthouse report and it works from that. For automatic data, use Node 22.19+ to run the Lighthouse 13 CLI, plus an optional Google API key in an environment variable for the CrUX or PSI API. Live lookups need web access.

    It cannot see real user interactions without field or RUM data, so it cannot measure INP in the lab. Impact figures are estimates. CrUX needs enough traffic on a URL, and field results lag a deploy by about 28 days. Scripts injected by a tag manager or CMS may not be in your repo.

    It follows a fixed order: data first, then LCP subparts, then mapping to files. It uses verified thresholds and current API and audit names, labels FIELD versus LAB data, refuses to invent numbers, and returns the same verdict and template every time.

    New in Frontend & Web Apps