Fetch
A page loads data in a chain
Review whether independent requests can run in parallel, or whether an await can move into the branch that actually needs it.
01 / The skill · React · Performance guidance
A rule-based Skill for writing and reviewing React or Next.js code. It prioritizes data waterfalls and bundle size before lower-impact micro-optimizations.
Published by Vercel Labs · Type: Code-review and implementation Skill
Independent guide. Vercel Labs has not endorsed Floatboat.
Upstream installation
npx skills add vercel-labs/agent-skills --skill vercel-react-best-practicesInstalls the upstream Skill with the open Skills CLI; no Floatboat capability is implied.
The source organizes React and Next.js performance rules into eight impact-ranked categories. That makes it useful for a focused review of fetching, bundles, server rendering, client state, and re-renders—not a substitute for profiling a real page.
Ask the agent
“Review this Next.js product page for request waterfalls and unnecessary client-side JavaScript. Rank the fixes and show the exact files.”
A useful result
A useful review identifies the request chain or import responsible, proposes bounded edits, and separates measured findings from hypotheses.
Use it when React code is being written, reviewed, or refactored and you need a prioritized checklist.
Fetch
Review whether independent requests can run in parallel, or whether an await can move into the branch that actually needs it.
Bundle
Trace broad imports, heavy modules, and third-party scripts before adding another loading indicator.
Render
Inspect state placement, derived values, and re-render triggers before applying memoization everywhere.
The Skill directs attention to high-impact categories, then the agent inspects the relevant code path.
Name the page or interaction and gather a trace or reproducible symptom.
Read the relevant rule and inspect only the code that contributes to that path.
Compare behavior and measurements after the change; do not call a stylistic rewrite a performance win.
02 / Practical guide
The source ranks waterfalls and bundles above micro-optimizations. Tie each proposed edit to the page's actual fetch and render path.
Its eight groups span asynchronous waterfalls, bundle loading, server work, client fetching, re-renders, rendering, JavaScript cost, and advanced patterns. The categories are ordered so reviews start with likely larger wins.
No. The SKILL.md provides instructions and references for an agent. It can suggest code changes, while before/after traces, network timings, and user-facing metrics must be checked in the project.
Read another Skill guide and check its original source before using it.
03 / Put it to work · Review the change
Bring your React files and review notes into Floatboat, then verify the important claims with your own measurements.