agentleFS
plugin87plugin87/full-stack-design-skills/.claude/skills/web-performance-optimization/SKILL.md

web-performance-optimization

Skill10 starsChanged 4 months agoMITView on GitHub

Optimize web applications for speed, responsiveness, and efficiency. Use this skill for Core Web Vitals analysis, Lighthouse interpretation, asset optimization strategies, network performance tuning, runtime performance improvements, monitoring setup, and performance budgets. Triggers include: page speed optimization, Lighthouse audit review, Core Web Vitals improvement, image/CSS/JS optimization, bundle size reduction, first-paint delays, layout shifts, interaction delays, performance monitoring questions, and performance budgets. Essential for anyone responsible for user experience, loading speed, or application responsiveness.

9 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

19 rules. 0 are common, 16 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 9 edited versions we read keep these.

  • 16

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use a CDN for fast global delivery

    line 48
  2. Distinctive:

    Avoid render-blocking CSS/JavaScript

    line 49
  3. Distinctive:

    Use server-side rendering for critical content

    line 50
  4. Distinctive:

    Use Web Workers for heavy computation

    line 76
  5. Distinctive:

    Avoid layout thrashing (read DOM, then write.

    line 77
  6. Very specific:

    Use requestAnimationFrame for animations

    line 78
  7. Distinctive:

    Avoid inserting DOM elements above the fold after load

    line 108
  8. Very specific:

    Use CSS transform: translate() instead of changing top/left (doesn't cause reflow)

    line 109
  9. Distinctive:

    Test with slow networks to catch layout shifts that appear only on slow connections

    line 110
  10. Distinctive:

    Not scored, but useful for investigation

    line 149
  11. Distinctive:

    No download; instant rendering

    line 214
  12. Distinctive:

    Remove unused CSS (PurgeCSS, Tailwind)

    line 245
  13. Distinctive:

    Use CSS-in-JS with automatic code splitting (styled-components, emotion)

    line 248
  14. Distinctive:

    Consider moving to Edge runtime (Cloudflare Workers, Vercel Edge) for faster execution

    line 254
  15. Distinctive:

    Don't block main thread (use Web Workers for heavy computation)

    line 332
  16. Distinctive:

    Use CSS transforms instead of top/left (GPU-accelerated)

    line 335
  17. Very specific:

    Use font-display: swap to avoid font swap causing layout shift

    line 469
  18. Distinctive:

    Run Lighthouse audit (lab test)

    line 476
  19. Distinctive:

    Test in isolation (fix one thing at a time)

    line 487

“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Web Performance Optimization

A practical guide to measuring, diagnosing, and improving web application performance.

Performance is a user experience feature. A fast site converts better, keeps users engaged, and ranks higher in search. This skill covers the metrics that matter, tools that measure them, and concrete optimization strategies.

Core Web Vitals: The Three Metrics That Matter

Google's Core Web Vitals are the three quantitative measures of user experience. They directly impact search rankings and user satisfaction.

1. Largest Contentful Paint (LCP)

What it measures: How fast the largest visible element (text, image, video) appears on screen.

Target: < 2.5 seconds

Why it matters: Users perceive "fast" as "appears quickly." If the main content takes 4+ seconds, users bounce.

What counts as LCP:

  • <img> elements
  • <image> inside <svg>
  • Backgrounds with background-image
  • Text blocks (paragraphs, headings)
  • Videos (poster frame)
  • Canvas elements

What doesn't count:

  • Off-screen images
  • Transparent images
  • Text in canvas

Example LCP Timeline:

0s    — Page begins to load
0.8s  — Hero image starts loading
1.2s  — Hero image finishes (LCP = 1.2s)  ✓ Good (< 2.5s)
2.1s  — Page fully interactive

Optimization strategies:

  • Preload critical images (<link rel="preload">)
  • Compress images aggressively (WebP format, lazy loading for below-the-fold)
  • Use a CDN for fast global delivery
  • Avoid render-blocking CSS/JavaScript
  • Use server-side rendering for critical content

2. First Input Delay (FID) → Interaction to Next Paint (INP)

What it measures: Delay between user input (click, tap, keystroke) and browser response.

Target: < 100ms (FID) or < 200ms (INP)

Why it matters: Slow response feels like the app is frozen. Users think something broke and abandon the interaction.

FID vs. INP:

  • FID (First Input Delay): Only measures the first user interaction. Being fast once doesn't mean consistently fast.
  • INP (Interaction to Next Paint): Measures all interactions throughout the page lifetime. More realistic; now recommended over FID.

Example Input Timeline:

User clicks button at: 0ms
Browser processes JavaScript: 0-80ms
Browser repaints: 80-85ms
Paint completes: 85ms
→ INP = 85ms ✓ Good (< 200ms)

Optimization strategies:

  • Break long JavaScript into shorter tasks (< 50ms each)
  • Defer non-critical code (code splitting, lazy loading)
  • Use Web Workers for heavy computation
  • Avoid layout thrashing (read DOM, then write; don't alternate)
  • Use requestAnimationFrame for animations

3. Cumulative Layout Shift (CLS)

What it measures: Unexpected layout shifts that happen after the page loads. Every time an element moves without user interaction, CLS increases.

Target: < 0.1

Why it matters: Layout shifts are annoying. Ads suddenly pushing content down, buttons moving mid-click—users hate it.

Common culprits:

  • Images without declared dimensions
  • Ads inserted dynamically
  • Embeds (video players, maps)
  • Web fonts causing text reflow

Example Layout Shift:

User is reading article at position 200px.
Ad loads below: shifts layout down 80px.
User's eyes jump; reading is disrupted.
CLS += 0.15 (poor)

Worse: User clicks where they thought a button was,
but button has shifted. Click lands on wrong element.

Optimization strategies:

  • Reserve space for images/ads before they load (set explicit width/height)
  • Load critical fonts synchronously; non-critical asynchronously
  • Avoid inserting DOM elements above the fold after load
  • Use CSS transform: translate() instead of changing top/left (doesn't cause reflow)
  • Test with slow networks to catch layout shifts that appear only on slow connections

Lighthouse: The Diagnostic Tool

Lighthouse is an automated audit tool that measures performance, accessibility, best practices, and SEO. It runs in Chrome DevTools, as a CLI, or via web (PageSpeed Insights).

Understanding Lighthouse Scores

Lighthouse scores are weighted averages of metrics:

Performance score: 0-100

  • LCP: 25% weight
  • INP: 30% weight
  • CLS: 25% weight
  • First Contentful Paint (FCP): 10%
  • Time to First Byte (TTFB): 10%

Score ranges:

  • 90-100: Excellent ✅
  • 50-89: Needs work ⚠️
  • 0-49: Poor ❌

Interpreting Lighthouse Reports

A Lighthouse report shows:

Metrics (top section)

  • Quantitative measurements (LCP 2.1s, INP 85ms, CLS 0.05)
  • Visual representation of how you're doing vs. targets

Opportunities (middle section)

  • Ranked by potential impact (high impact fixes first)
  • Examples: "Eliminate render-blocking resources," "Defer offscreen images"
  • Estimated savings (e.g., "Saves 2.5s")

Diagnostics (bottom section)

  • Additional info: unused CSS, accessibility issues, security warnings
  • Not scored, but useful for investigation

Common Lighthouse Opportunities

Eliminate render-blocking resources (high impact)

  • Problem: CSS/JS files loaded in <head> block page rendering
  • Fix: Inline critical CSS, defer non-critical JS, use async/defer attributes

Defer offscreen images (high impact)

  • Problem: All images load immediately, even those below the fold
  • Fix: Use lazy loading (loading="lazy" attribute or Intersection Observer)

Unsized images cause layout shift (high impact)

  • Problem: Images load without width/height, causing reflow
  • Fix: Always declare width and height attributes or CSS aspect-ratio

Reduce CSS (medium impact)

  • Problem: Large unused CSS files
  • Fix: Remove dead code, split by route, use CSS-in-JS with automatic splitting

Reduce JavaScript (medium impact)

  • Problem: Large bundle size
  • Fix: Code splitting, tree-shaking, lazy loading routes

Asset Optimization

Images: Biggest Bang for Buck

Images typically account for 50-80% of page weight. Optimizing images has the highest ROI.

Image optimization checklist:

  • ✅ Use modern formats (WebP with JPEG fallback)
  • ✅ Compress aggressively (quality 75-85 is often imperceptible)
  • ✅ Serve responsive sizes (srcset, sizes attributes)
  • ✅ Lazy load below-the-fold images (loading="lazy")
  • ✅ Declare dimensions (prevent layout shift)
  • ✅ Use a CDN with image optimization (Cloudflare, Akamai, Fastly)

Example: Optimized Image Tag

<!-- Before: 2MB, unoptimized -->
<img src="photo.jpg" alt="..." />

<!-- After: 150KB, optimized -->
<picture>
  <source type="image/webp" srcset="photo-400w.webp 400w, photo-800w.webp 800w" sizes="(max-width: 600px) 100vw, 50vw" />
  <source type="image/jpeg" srcset="photo-400w.jpg 400w, photo-800w.jpg 800w" sizes="(max-width: 600px) 100vw, 50vw" />
  <img src="photo-800w.jpg" alt="..." width="800" height="600" loading="lazy" />
</picture>

Result: 95% size reduction, faster LCP, no layout shift.

Fonts: Load Strategically

Web fonts are often invisible bottleneck. Default behavior blocks rendering.

Font loading strategies:

Strategy 1: System Fonts Only (Fastest)

font-family: system-ui, -apple-system, sans-serif;
  • No download; instant rendering
  • Sacrifice brand consistency for speed

Strategy 2: Critical Font (Synchronous)

<link rel="preload" href="font.woff2" as="font" type="font/woff2" crossorigin />
  • Loads early; unblocks rendering after download
  • Good for display fonts (headings)

Strategy 3: Non-Critical Font (Asynchronous)

<link rel="preload" href="font.woff2" as="font" type="font/woff2" crossorigin media="(prefers-reduced-data: no-preference)" />
  • Loads in background; doesn't block
  • Good for body text

Strategy 4: Variable Font (Modern)

@font-face {
  font-family: 'Recursive';
  src: url('recursive.var.woff2') format('woff2-variations');
  font-weight: 100 900;
}
  • Single file for all weights/styles (400 normal + 700 bold = 1 file instead of 2)
  • Huge savings for large font families

CSS & JavaScript: Load Only What You Need

CSS optimization:

  • Remove unused CSS (PurgeCSS, Tailwind)
  • Split by route (critical styles inline, route styles async)
  • Compress and minify
  • Use CSS-in-JS with automatic code splitting (styled-components, emotion)

JavaScript optimization:

  • Code splitting by route (lazy load pages, modals, etc.)
  • Tree-shaking (remove dead code)
  • Minification and compression
  • Consider moving to Edge runtime (Cloudflare Workers, Vercel Edge) for faster execution

Network Optimization

Caching Strategy

Browser cache (HTML Cache-Control header)

  • Set long expiry for static assets (1 year for versioned files)
  • Set short expiry for HTML (1-5 minutes or no-cache with ETag)
# Versioned assets (never change)
Cache-Control: max-age=31536000, immutable

# HTML (check for updates frequently)
Cache-Control: max-age=300, must-revalidate

CDN cache

  • Edge caching for global delivery
  • Geographically distributed servers reduce latency
  • Example services: Cloudflare, Fastly, Akamai

Service Worker cache

  • Offline support
  • Predictable performance (don't wait for network)
  • Good for PWAs

Compression

Gzip/Brotli compression (most important)

  • Compress text (HTML, CSS, JS) at 70-90% reduction
  • Enable on server (nearly zero CPU cost for modern servers)
  • Most browsers support it automatically
# Enable Brotli (better than gzip)
Content-Encoding: br

Lazy Loading

Intersection Observer API (modern, recommended)

const observer = new IntersectionObserver((entries) => {
  entries.forEach(entry => {
    if (entry.isIntersecting) {
      const img = entry.target;
      img.src = img.dataset.src;
      observer.unobserve(img);
    }
  });
});

document.querySelectorAll('img[data-src]').forEach(img => {
  observer.observe(img);
});

Native lazy loading (simplest, browser support ≈95%)

<img src="..." loading="lazy" />

Runtime Performance

Rendering Performance

Frame budget: 60fps = 16ms per frame

  • Rendering (layout, paint): ≤ 10ms
  • JavaScript: ≤ 3-5ms (rest is browser work)

Avoid jank (stuttering):

  • Don't block main thread (use Web Workers for heavy computation)
  • Batch DOM reads/writes (all reads first, then all writes)
  • Use requestAnimationFrame for animations
  • Use CSS transforms instead of top/left (GPU-accelerated)

Example: Layout Thrashing (BAD)

// Bad: read-write-read-write alternates
for (let i = 0; i < 10; i++) {
  element.style.width = element.offsetWidth + 10 + 'px'; // read then write
}

// Good: batch reads, then batch writes
const widths = [];
for (let i = 0; i < 10; i++) {
  widths.push(element.offsetWidth); // all reads first
}
widths.forEach((w, i) => {
  elements[i].style.width = (w + 10) + 'px'; // then all writes
});

Memory Leaks

Common causes:

  • Detached DOM nodes still referenced in JavaScript
  • Event listeners not removed
  • Timers (setInterval) not cleared
  • Circular references in objects

Detection:

  • Chrome DevTools Memory profiler
  • Heap snapshots to find retained objects
  • Performance monitor for memory growth over time

Monitoring & Observability

Real User Monitoring (RUM)

Synthetic (lab) testing like Lighthouse is good, but doesn't capture real-world conditions. Real User Monitoring measures actual user experiences.

Key RUM metrics:

  • Core Web Vitals (LCP, INP, CLS)
  • Time to First Byte (TTFB)
  • Page load time
  • Error rates
  • User interactions (clicks, scrolls, form submissions)

Popular RUM services:

  • Google Analytics (free, built-in Web Vitals)
  • Datadog (premium)
  • New Relic (premium)
  • Custom implementation using PerformanceObserver

Custom RUM implementation:

// Measure Core Web Vitals
new PerformanceObserver((list) => {
  for (const entry of list.getEntries()) {
    console.log('LCP:', entry.renderTime || entry.loadTime);
  }
}).observe({ entryTypes: ['largest-contentful-paint'] });

new PerformanceObserver((list) => {
  for (const entry of list.getEntries()) {
    console.log('CLS:', entry.value);
  }
}).observe({ type: 'layout-shift', buffered: true });

Performance Budgets

A performance budget is a maximum limit on page metrics. It prevents performance regression.

Example budget:

  • JavaScript bundle: ≤ 150KB (gzipped)
  • CSS: ≤ 50KB (gzipped)
  • Images: ≤ 500KB
  • LCP: ≤ 2.5s
  • INP: ≤ 200ms
  • CLS: ≤ 0.1

How to enforce:

  • CI/CD checks (fail build if budget exceeded)
  • Lighthouse CI (automated Lighthouse runs on every PR)
  • Bundle size monitoring (bundlesize, size-limit)

Example CI check:

# GitHub Actions
- name: Lighthouse CI
  uses: treosh/lighthouse-ci-action@v10
  with:
    uploadArtifacts: true
    temporaryPublicStorage: true
    budgetPath: ./lighthouse-budget.json

Common Performance Issues & Fixes

Slow LCP

Problem: Page takes 4+ seconds to show main content.

Diagnosis:

  • Largest element is a hero image? Problem is image loading
  • Largest element is text? Problem is font loading or render-blocking CSS

Fixes:

  • Hero images: Preload, compress, use CDN
  • Fonts: Use system fonts initially, load custom font asynchronously
  • Render-blocking CSS: Move to <head>, inline critical styles, defer rest

High INP

Problem: Page feels unresponsive (clicks take 500ms+ to register).

Diagnosis: DevTools Performance profiler will show long JavaScript execution times.

Fixes:

  • Break JavaScript into shorter tasks (use scheduler.yield() in modern browsers)
  • Defer non-critical work (debounce, lazy load)
  • Use Web Workers for heavy computation

High CLS

Problem: Page layout shifts unexpectedly.

Diagnosis: Lighthouse report identifies specific culprits (images, ads, fonts).

Fixes:

  • Declare image dimensions
  • Reserve space for ads/embeds before load
  • Use font-display: swap to avoid font swap causing layout shift

Performance Testing Workflow

Step 1: Establish baseline

  • Run Lighthouse audit (lab test)
  • Set up RUM (real-world monitoring)
  • Define performance budget

Step 2: Identify bottlenecks

  • Lighthouse Opportunities section (ranked by impact)
  • Performance profiler for runtime issues
  • Network tab for slow assets

Step 3: Prioritize & fix

  • High-impact fixes first (e.g., image optimization)
  • Test in isolation (fix one thing at a time)
  • Re-measure to confirm improvement

Step 4: Monitor & maintain

  • Set performance budget in CI/CD
  • Monitor RUM dashboard weekly
  • Alert on regressions

When to Use This Skill

✅ Do use for:

  • Core Web Vitals optimization and interpretation
  • Lighthouse audit review and prioritization
  • Asset optimization strategy (images, fonts, CSS, JS)
  • Network and caching optimization
  • Runtime performance improvements
  • Performance monitoring setup
  • Performance budget definition and enforcement
  • Performance testing methodology

❌ Don't use for:

  • Specific tool setup (Google Analytics, Datadog config — see tool-specific skills)
  • Framework-specific optimization (React.memo, Vue computed properties — see framework skills)
  • Server optimization (Node.js, database tuning — see backend skills)
  • SEO optimization (though performance helps SEO)

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.