high-perf-browser

Diagnose and optimize web performance across network protocols, resource loading, caching, and Core Web Vitals.

Updated Jul 8, 2026
One-click install
npx skills add https://github.com/HafidJoss/Lummy --skill high-perf-browser-hafidjoss
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: high-perf-browser
Source: https://github.com/HafidJoss/Lummy/tree/main/agent/skills/high-perf-browser
Command: npx skills add https://github.com/HafidJoss/Lummy --skill high-perf-browser-hafidjoss

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Slow page loads usually stem from latency and round trips rather than bandwidth, and fixing them requires understanding how DNS, TCP, TLS, HTTP versions, caching layers, and the browser rendering pipeline actually interact. This Skill provides a systematic framework for diagnosing slow sites and applying the correct protocol, loading, caching, and rendering optimizations. ## Core Features & Use Cases - Six-domain performance framework: Covers network fundamentals (TCP/TLS handshakes, slow start, DNS), HTTP/1.1-to-HTTP/3 protocol evolution, critical rendering path, caching strategies, Core Web Vitals (LCP, INP, CLS), and real-time transports (WebSocket vs SSE vs long polling). - Quick Diagnostic scorecard: Eight pass/fail checks (TTFB, LCP, INP, CLS, content hashing, HTTP/2+, render-blocking resources, compression) produce a 0-10 score with a specific fix for each failure. - Deep reference guides: Five reference documents provide copy-paste configurations such as Nginx OCSP stapling, service worker cache-first/network-first recipes, WebSocket reconnection with exponential backoff, and resource hint ordering. - Use Case: A user reports their e-commerce site feels slow on mobile. Run the Quick Diagnostic, find LCP at 4.2s and no content hashing, then apply the LCP preload pattern with fetchpriority="high" and immutable cache headers for hashed static assets. ## Quick Start Ask the assistant to run the web performance diagnostic on your site and explain which of the eight checks fail and how to fix each one.

Frequently Asked Questions about high-perf-browser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I improve Core Web Vitals scores like LCP, INP, and CLS?

Improve LCP by preloading the largest element with fetchpriority="high" and keeping TTFB under 800ms. Improve INP by breaking long tasks with scheduler.yield() and deferring non-critical JavaScript. Improve CLS by setting explicit image dimensions and reserving space for dynamic content.

What is the difference between WebSocket and Server-Sent Events?

WebSocket provides full-duplex bidirectional messaging for chat and collaboration, while SSE is server-to-client only over plain HTTP with automatic reconnection. SSE is simpler and proxy-friendly; choose WebSocket only when the client must also push data frequently.

Should I use HTTP/2 or HTTP/3 for my website?

Enable HTTP/2 on all servers as the baseline since it multiplexes streams over one connection, then advertise HTTP/3 via the Alt-Svc header through a CDN. HTTP/3 over QUIC eliminates TCP head-of-line blocking and benefits mobile users on lossy networks.

Why is my site slow even with fast bandwidth?

Latency, not bandwidth, is usually the bottleneck because each request pays DNS, TCP handshake, and TLS negotiation round trips. Reduce round trips with preconnect hints, connection reuse, TLS 1.3, CDN edge caching, and fewer third-party origins.

How do I configure Cache-Control headers for static assets?

Serve content-hashed static assets with Cache-Control: max-age=31536000, immutable so browsers never revalidate them. Use no-cache with ETag for HTML documents so they revalidate with cheap 304 responses, and no-store only for sensitive data.

When should I not use domain sharding and file concatenation?

Avoid these HTTP/1.1 workarounds on HTTP/2 or HTTP/3, where they become counterproductive. Domain sharding defeats multiplexing by opening extra connections, and concatenation prevents granular caching of individual modules.