Fxa Check React
Testing & Qualityfxa-check-reactFxa Check React is an Agent Skill for testing & quality published in the mozilla/fxa repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Reviews changed React/TSX code for component design, hooks misuse, performance, accessibility, and state management issues. Reports findings with severity and concrete fix recommendations. Operates on files changed vs main.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/mozilla/fxa/blob/HEAD/.claude/skills/fxa-check-react/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/fxa-check-react, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
You are a senior React engineer. Review all React/TSX files changed in the current branch. For each finding, report:
- Severity: High / Medium / Low
- Location: file:line
- Issue: concise description of the problem
- Recommendation: concrete, actionable fix
If a category is clean, say so briefly. Do not skip categories.
How to gather the diff
Run:
git diff main...HEAD
Focus on .tsx, .ts, .jsx, .js files. Read additional context from related files as needed.
Review Checklist
1. Component Design & Hooks
Component responsibilities
- Components doing too much — rendering, data fetching, business logic, and formatting all in one
- Recommendation: split into a container (data/logic) and a presentational (render-only) component
Prop drilling
- Props passed through 3+ levels of components that don't use them
- Recommendation: lift to context, a store, or colocate the consuming component closer to the data
Component size
- Components exceeding ~150 lines or containing more than one distinct visual section
- Recommendation: extract sub-components with descriptive names
useEffect misuse
- Effects used to sync derived state that could be computed inline or with
useMemo - Effects that run on every render due to missing or incorrectly specified dependency arrays
- Multiple unrelated concerns in a single
useEffect - Missing cleanup for subscriptions, timers, or event listeners
- Recommendation: compute derived values inline; split effects by concern; always return cleanup functions
Stale closures
- Variables captured in callbacks or effects that don't reflect the latest state/props
- Common in
setInterval, event listeners, or async callbacks not listed in deps - Recommendation: add to dependency array; use
useReffor values that should not re-trigger effects
Hook ordering violations
- Hooks called conditionally or inside loops
- Recommendation: move hooks to top level; use conditional logic inside the hook body
Custom hook opportunities
- Repeated stateful logic across multiple components that could be extracted
- Recommendation: extract to a
use*custom hook in a shared location
2. Performance
Unnecessary re-renders
- Components re-rendering because a parent passes a new object/array literal or inline function on every render
- Recommendation: memoize with
useMemo/useCallback; stabilize references
Missing React.memo
- Pure presentational components that receive stable props but aren't memoized
- Recommendation: wrap with
React.memowhere the component is expensive to render
Expensive computations in render
- Heavy calculations or data transformations done inline during render without
useMemo - Recommendation: memoize with
useMemo; move to a selector or derived query
Large component trees without lazy loading
- Heavy components or routes not using
React.lazy/Suspense - Recommendation: code-split at route or modal boundaries
Key prop issues
- Missing
keyprops in lists; using array index askeyfor reorderable lists - Recommendation: use stable, unique IDs as keys
3. Accessibility (a11y)
Missing or incorrect ARIA
- Interactive elements lacking
aria-label/aria-labelledbywhen no visible text is present roleattributes used incorrectly or unnecessarily (avoid overriding native semantics)- Dynamic content changes not announced via
aria-live
Keyboard navigation
- Click handlers on non-interactive elements (
div,span) without a correspondingonKeyDown/onKeyPressandtabIndex - Focus not managed after modal open/close or route transitions
- Custom dropdowns, dialogs, or menus that don't implement expected keyboard patterns (Escape to close, arrow key navigation)
Focus management
- Modals that don't trap focus or don't return focus to the trigger on close
- Skip-to-content links missing on new pages
Semantic HTML
- Using
<div>or<span>where a semantic element (<button>,<nav>,<main>,<section>) is appropriate - Form inputs missing associated
<label>elements
Color & contrast
- Inline styles or new CSS that introduce text/background color combinations — flag for manual contrast check (tool cannot verify ratios, but can identify new color declarations to review)
4. State Management
Unnecessary state
- State that is fully derivable from other state or props and doesn't need to be stored
- Recommendation: compute inline or with
useMemo; remove theuseState
Local vs global state
- UI-only state (open/closed, hover, form input) stored in global state/context unnecessarily
- Shared cross-component state managed locally, causing sync bugs
- Recommendation: keep UI state local; lift or globalize only when multiple disconnected components need it
Apollo / GraphQL cache
- Queries that bypass the cache with
fetchPolicy: 'network-only'without clear justification - Manual cache writes (
writeQuery,writeFragment) that could cause stale data - Missing
optimisticResponseon mutations where latency would hurt UX - Over-fetching: queries selecting more fields than the component uses
Stale or redundant state after data fetches
- Local state that mirrors remote data and can get out of sync
- Recommendation: derive from the query result directly; avoid duplicating server state locally
5. Testing Patterns (React/TSX)
When reviewing *.test.tsx changes, apply the React testing rules in .claude/rules/testing/react.md in addition to the general FXA rules in .claude/rules/testing/base.md. Both rules auto-load when Claude reads a matching test file; this skill's pre-flight should explicitly Read them too so reviews cover test changes consistently. Headline triggers to flag: implementation-detail queries (className/CSS/data-testid where semantic queries fit), fireEvent where userEvent belongs, asserting on instances/refs/private state, missing or misused act(), whole-tree snapshots for non-trivial components, and duplicated provider boilerplate.
Output Format
Lead with a summary table of all findings (severity, category, file:line, one-line description). Follow with detailed write-ups for High severity items. End with a "Clean categories" list for anything with no issues found.