React review
Turn any topic into an expert-level AI skill — for Claude, ChatGPT, Gemini, or any LLM. A meta-skill that researches a domain and writes the skill for you.
npx -y skills add jacobjustin8/skill-forge --skill react-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 12 days oldThe repository was created 12 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 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.
What its author says it does
Copied from the file, not written here
Review React + TypeScript code — components, hooks, a PR, or a pasted file — the way a senior frontend engineer would, prioritizing correctness, hooks bugs, performance, accessibility, and type safety over style nits. Use whenever the user asks you to review, critique, or check React/TSX/JSX code or a frontend PR, or asks "is this component right?" — even if they don't say "code review".
SKILL.md
4.2 KB, 892 tokens by cl100k_base, as published. Nobody here has run it
React + TypeScript review
Review the way a senior frontend engineer whose reviews people trust would: catch the bugs that actually ship broken UI, and skip the bikeshedding. The goal is a change that leaves the frontend healthier, with the author's time respected. Prettier/ESLint already handle formatting — don't spend review on it.
Review in priority order
Read every line you approve, but spend attention where the risk is highest.
-
Hooks correctness — the #1 source of real React bugs.
- Hooks must be called unconditionally at the top level — never inside conditionals, loops, or nested functions. Flag any conditional hook.
useEffect/useCallback/useMemodependency arrays must include every reactive value they read. A missing dependency is a stale-closure bug waiting to happen; aneslint-disableon the deps rule is only acceptable with a comment explaining why. Watch for effects that should have a cleanup function (subscriptions, timers, listeners) and don't.
-
Type safety. Flag
any(it disables the very protection TS provides — prefer a real interface orunknown+ narrowing). Questionasassertions without justification; prefer runtime validation (e.g. Zod) for external data. Check optional values are actually guarded (?.,??, narrowing) rather than accessed blindly. -
Correctness & data flow. Does it handle loading, empty, and error states, not just the happy path? Are async calls awaited and their rejections handled? Does state update logic avoid mutating state directly?
-
Performance — but only where it's real. Inline object/array/function props defeat
React.memoand cause re-renders; flag them when the child is memoized or the list is large. Flag expensive work on every render that belongs inuseMemo. Reject index-as-key (and missing keys) in dynamic lists — it causes subtle state/DOM bugs. Don't demand memoization everywhere, though — prematureuseMemo/useCallbackadds noise and its own bugs. -
Accessibility. Real elements over
div onClick(<button>,<a>); icon-only buttons need anaria-label; form inputs need associated labels; interactive things must be keyboard-reachable. These are correctness issues for a share of your users, not polish. -
Security. Block
dangerouslySetInnerHTMLon unsanitized input (XSS). No secrets/API keys in client code. No logging of tokens or PII. -
Structure. Overgrown components (~200+ lines doing many things) want splitting; deep prop-drilling suggests context or a custom hook; business logic and data-fetching belong out of JSX in pure functions / hooks.
What NOT to flag
Formatting (Prettier's job), personal style (map vs. loop, naming aesthetics), and full rewrites. Suggest approaches; don't rewrite the author's block for them. Blocking on preference is what makes a reviewer dreaded rather than valued.
Tone
Ask questions where you might lack context ("is there a reason the empty state isn't handled?"), explain the why so the author learns the rule, offer the fix, and note something that's genuinely good. Distinguish blocking issues (a hooks bug, a type hole, an XSS) from optional suggestions — prefix the latter with "Nit:".
Output
## Verdict
<Approve / Approve with nits / Needs changes> — one line on the headline.
## Blocking
<Numbered: where, what's wrong, why it matters, suggested fix. Omit if none.>
## Non-blocking
<Suggestions and questions; prefix optional polish with "Nit:".>
## What's good
<Briefly reinforce what was done well.>
If you were given a snippet without its surroundings, review what's visible and say what you'd need (the parent, the types, the tests) to be sure — don't invent context.