agentsclimarketplace
Sign in

Perf checklist

Skill atuljha23/holocron/skills/perf-checklist

Performance quick reference — Core Web Vitals, bundle size, rendering, DB queries, caching. Use during implementation or review when perf matters.From its SKILL.md

Install
npx -y skills add atuljha23/holocron --skill perf-checklist

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

  • 2 stars2 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

3.0 KB, 752 tokens by cl100k_base, as published. Nobody here has run it

Performance checklist

Measure first. Opinion second.

Frontend — Core Web Vitals

LCP (Largest Contentful Paint) — ≤ 2.5s

  • Hero image preloaded, sized, fetchpriority="high".
  • Critical CSS inlined or loaded before render-blocking resources.
  • Fonts: font-display: swap, preloaded if above the fold.
  • No render-blocking JS on the critical path. defer or async.

CLS (Cumulative Layout Shift) — ≤ 0.1

  • width/height on images and iframes.
  • Reserved space for async content (skeletons, aspect-ratio boxes).
  • No ads / embeds inserted above existing content without reservation.
  • No web-font swap without matching metrics.

INP (Interaction to Next Paint) — ≤ 200ms

  • No long tasks (>50ms) on interaction paths.
  • Break up work: scheduler.yield(), idle callbacks, debounced expensive calcs.
  • Virtualize long lists.
  • Event handlers don't do layout thrash — batch reads then writes.

Frontend — bundle

  • Tree-shake-friendly imports. import { x } from 'lib' — not import lib from 'lib' unless required.
  • No moment — use date-fns or Temporal polyfill.
  • No full lodash — use lodash-es named imports or dropping to native.
  • Code-split routes. Lazy-load heavy components that aren't above the fold.
  • Images: modern formats (WebP/AVIF), responsive srcset, served from a CDN.

Backend — query layer

  • No query inside a loop over a result set (N+1). Batch with IN/JOIN/dataloader.
  • Every WHERE / ORDER BY / JOIN column used on large tables is indexed.
  • SELECT * replaced with specific columns on hot paths.
  • Pagination uses cursors, not offset, on tables that grow.
  • Queries have statement timeouts.

Backend — request

  • Timeouts on every outbound call.
  • Retries capped with backoff. Never infinite.
  • Circuit breakers for dependencies you can't trust.
  • Response compression enabled (gzip/brotli).
  • Static assets behind a CDN with long Cache-Control.

Cache

  • Layers have clear ownership: browser, CDN, app memory, database.
  • TTLs tied to staleness tolerance, not wish.
  • Invalidation path is obvious. If you can't invalidate it, you can't cache it.
  • Cache-Control: stale-while-revalidate for content that tolerates it.

Instrumentation

  • Real User Monitoring (RUM) for Web Vitals on production.
  • Server: p50/p95/p99 latency per endpoint.
  • Flame graphs available in production (sampled) for hot paths.
  • Alerts on regression — not just absolute thresholds.

Common mistakes

  • "Add caching" to mask a slow query — fix the query.
  • Memoizing in React before measuring — often makes things worse.
  • Index everything — indexes have a write cost; pick with intention.
  • Serve Gzip but forget to compress API JSON — biggest gzip wins are often in the API.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most caching build skills give in 752 tokens

Counted across 103 of the 134 authors here whose files we hold, read 2026-09-06

  • Prefer App Router and server componentsin 12 of 103, across 7 files
  • Stay on a recent Next.js 16.x releasein 12 of 103, across 7 files
  • Use the Bundle Analyzer to trim large dependenciesin 12 of 103, across 7 files
  • Use Turbopack for day-to-day developmentin 11 of 103, across 6 files
  • Fall back to webpack only for Turbopack bugs or webpack-only pluginsin 11 of 103, across 6 files
  • Set a TTL on every cache entryin 10 of 103, across 9 files
  • Check the official docs for your Next.js versionin 9 of 103, across 5 files
  • Ensure the cache is not cleared unnecessarilyin 8 of 103, across 4 files
  • Verify Turbopack is active when dev is slowin 8 of 103, across 4 files
  • Run next dev for local developmentin 7 of 103, across 6 files
  • Run next dev for local development with Turbopackin 6 of 103, across 2 files
  • Append static extensions to dynamic URLs to trigger cachingin 6 of 103, across 3 files

Said here and by no other author read

  • Defer render-blocking JavaScript
  • Use tree-shakeable named imports
  • Batch queries to avoid N+1
  • Add timeouts to queries and outbound calls
  • Serve static assets and images from a CDN
  • Monitor Web Vitals and endpoint latency in production

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 325,949. 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.