agentsclimarketplace

React performance

Skill jgamaraalv/delivery-loop/.claude/skills/react-performance

React performance: unnecessary re-renders, memoization (React.memo/useMemo/useCallback), React Compiler, Context, concurrent features, code splitting, and Core Web Vitals.From its SKILL.md

Install
npx -y skills add jgamaraalv/delivery-loop --skill react-performance

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

9.0 KB, ~1.9k tokens by cl100k_base, as published. Nobody here has run it

Performance Debugging Workflow

Always follow this order — measure before optimizing:

  1. Reproduce — Document exact steps, browser, device, network conditions
  2. Measure — Profile with React DevTools Profiler + Chrome Performance tab. Never measure development builds.
  3. Identify — Form a specific hypothesis about the cause
  4. Fix — Apply the minimal fix needed
  5. Verify — Measure improvement under same conditions

Rule #1: Do NOT measure the development build. For production profiling, alias:

  • react-dom$react-dom/profiling
  • scheduler/tracingscheduler/tracing-profiling

Rule #2: Simulate real user conditions (slow CPU, throttled network).

Why Components Re-render

Components re-render for exactly three reasons:

  1. Its state changed
  2. Its parent rendered (unless wrapped in React.memo)
  3. A consumed Context value changed

Renders are either necessary or unnecessary. Any single unnecessary render is rarely a problem — it's the accumulation over time.

State Architecture (Fix Before Memoizing)

Before reaching for memoization, fix your state architecture. These are free performance wins:

Colocate State

Move state to the closest component that needs it. State at the root re-renders the entire tree.

// BAD: searchQuery in App re-renders Header, Cart, Footer
function App() {
  const [searchQuery, setSearchQuery] = useState("");
  return (
    <>
      <Header />
      <SearchBar query={searchQuery} onChange={setSearchQuery} />
      <Cart />
    </>
  );
}

// GOOD: searchQuery lives in SearchSection
function App() {
  return (
    <>
      <Header />
      <SearchSection />
      <Cart />
    </>
  );
}
function SearchSection() {
  const [searchQuery, setSearchQuery] = useState("");
  return <SearchBar query={searchQuery} onChange={setSearchQuery} />;
}

Derive, Don't Store

If a value can be computed from existing state/props, compute it — don't store it.

// BAD: synchronized state
const [items, setItems] = useState<Item[]>([]);
const [filteredItems, setFilteredItems] = useState<Item[]>([]);
// Must keep in sync — bugs + extra renders

// GOOD: derived value
const [items, setItems] = useState<Item[]>([]);
const filteredItems = items.filter((i) => i.active); // or useMemo if expensive

Lift State Intelligently

Only lift state to the lowest common ancestor that needs it — no higher.

Memoization Decision Guide

Three memoization tools, each with specific use cases:

ToolWhat it cachesUse when
React.memoComponent render outputProfiler shows component re-renders due to parent, not its own props/state
useMemoComputation resultExpensive calculation runs on every render, OR referential stability for memoized children
useCallbackFunction identityFunction passed as prop to React.memo-wrapped child, OR used in dependency arrays

Key rule: useMemo/useCallback are pointless without React.memo on the receiving child (or a dependency array consumer).

When NOT to Memoize

  • Simple/cheap components that render fast
  • Props that change on almost every render anyway
  • Components without memoized children receiving the value
  • Simple calculations (the comparison overhead exceeds computation cost)

React Compiler

The React Compiler is a Babel plugin that automatically applies memoization at build time. It analyzes your code and inserts useMemo/useCallback where safe.

Prerequisites: Code must follow the Rules of React — components must be pure, props/state immutable, no side effects in render.

What the Compiler Handles

  • Stabilizing callback identities (replaces manual useCallback)
  • Memoizing derived values (replaces manual useMemo)
  • Memoizing JSX output (replaces many React.memo wrappers)

What You Still Need Manually

  • React.memo: Impure components, 3rd-party library components, explicit render boundaries
  • useMemo: Truly expensive computations the compiler can't prove safe, custom equality logic
  • useCallback: Complex closure semantics, library APIs requiring stable function identities

Context API Performance

Context is a broadcast mechanism — every consumer re-renders when the value changes. The fix is splitting.

The Two-Context Pattern

Separate state from dispatch/actions into two contexts so components that only dispatch never re-render when state changes (full Provider example in references/context-optimization.md).

Split by Domain

Never create a "mega context" with unrelated state. Split into ThemeContext, AuthContext, UIContext, etc.

Concurrent Features

useTransition vs useDeferredValue

useTransitionuseDeferredValue
WrapsThe action (setState call)The value (result)
Use whenYou control the state updateYou don't control the update
ProvidesisPending booleanCompare current vs deferred value
EffectMarks update as low-priorityValue lags behind during urgent updates

Neither makes anything faster — they make the UI feel faster by prioritizing urgent updates (typing) over expensive work (filtering). Code examples for both, plus Suspense and data-fetching patterns, live in references/concurrent-features.md.

Bundle Performance

Code Splitting Checklist

  1. Route-based splittinglazy() + Suspense for each route
  2. Heavy component splitting — Lazy-load modals, charts, editors
  3. Conditional feature splitting — Admin panels, premium features
  4. Bundle analysis — Use webpack-bundle-analyzer or source-map-explorer
  5. Tree shaking — Use named exports, check sideEffects in package.json

Core Web Vitals Quick Reference

MetricTargetReact Optimization
LCP (Largest Contentful Paint)< 2.5sSSR/SSG, preload critical assets, optimize images
INP (Interaction to Next Paint)< 200msBreak long tasks, memoize, debounce/throttle, Web Workers
CLS (Cumulative Layout Shift)< 0.1Set image dimensions, reserve space for async content, skeleton loaders

React Fiber — How Rendering Works

React Fiber is a cooperatively-scheduled rendering engine using a linked-list tree structure.

  • Two trees: Current (what DOM reflects) and Work-in-Progress (draft being prepared)
  • Yielding: React yields to browser every ~5ms, allowing paint/input handling
  • Lanes: Priority system using bitmasks — SyncLane > InputContinuousLane > DefaultLane > TransitionLane > IdleLane
  • Commits are never interrupted — once render phase completes, DOM updates happen synchronously

Understanding this helps explain why useTransition works: it assigns updates to lower-priority lanes.

References

Each file is loaded on demand — read one only when the task needs that depth (progressive disclosure).

  • references/profiling-and-debugging.md — the systematic measure-before-optimize workflow, the React DevTools Profiler, the Chrome Performance tab, and production-profiling build setup · read when measuring, reproducing, or diagnosing a slowdown before touching code.
  • references/memoization-patterns.mdReact.memo, useMemo, and useCallback in depth with custom comparators and the dependency-array rules · read when applying memoization or deciding which of the three to reach for.
  • references/react-compiler.md — React Compiler setup, how it auto-memoizes, migration steps, and what still needs manual memoization · read when adopting/migrating to the React Compiler or auditing what it can't cover.
  • references/context-optimization.md — the two-context (state/dispatch) Provider pattern, the mega-context anti-pattern, and domain splitting · read when a Context value re-renders too many consumers.
  • references/concurrent-features.mduseTransition, useDeferredValue, and Suspense with full code and data-fetching patterns · read when making expensive updates non-blocking or wiring Suspense.
  • references/bundle-and-loading.md — code splitting, lazy()/Suspense route and component splitting, lazy-loading patterns, tree shaking, and bundle analysis · read when reducing bundle size or implementing lazy loading.

What ships with it: 6 files

33.9 KB alongside SKILL.md

Gives 0 of the 12 instructions most performance cost skills give in ~1.9k tokens

Counted across 803 of the 1,058 authors here whose files we hold, read 2026-08-07

  • Keep skill files under 500 lines or tokensin 82 of 803, across 16 files
  • Use imperative form in instructionsin 80 of 803, across 9 files
  • Draft assertions while test runs are in progressin 75 of 803, across 9 files
  • Create two to three realistic test promptsin 74 of 803, across 9 files
  • Write skill descriptions to be pushyin 72 of 803, across 7 files
  • Save test cases to evals JSONin 72 of 803, across 6 files
  • Ask questions about edge cases and input formatsin 72 of 803, across 7 files
  • Save timing data immediately when runs completein 70 of 803, across 5 files
  • Include all trigger conditions in the skill descriptionin 69 of 803, across 3 files
  • Launch all test runs in a single turn or simultaneouslyin 69 of 803, across 3 files
  • Capture intent before writing a skillin 67 of 803, across 1 file
  • Import directly instead of barrel filesin 52 of 803, across 15 files

Said here and by no other author read

  • profile production builds only
  • simulate real user conditions during profiling
  • derive values instead of storing them
  • split contexts by domain
  • read referenced files on demand for depth

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 326,852. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.