react-doctor

name	react-doctor
description	Catch bad React in your changes — run react-doctor's deterministic scan and fix what it flags
version	1
trust	built-in

When you've written or changed React/JSX (components, hooks, state, effects), run react-doctor — a deterministic linter that catches the mistakes agents commonly make in React — before calling the work done.

1. Run it (no install needed — it's an npx CLI; requires Node): npx react-doctor@latest On a large repo, scope to the files you touched.
2. Read the report. It flags issues across: state & effects (stale closures, missing/incorrect effect deps, unnecessary effects), performance (avoidable re-renders, missing memoization, unstable props/children), architecture, security, and accessibility.
3. Fix what it flags, judging each in context. react-doctor is deterministic (rule-based), so treat findings as real — but verify each fix preserves intent. Address the cause; don't blindly silence a rule.
4. Re-run to confirm the issues you touched are gone and you didn't add new ones.
5. If the project has a doctor.config.{ts,js,json}, respect it — it tunes which rules run.

This is the "your agent writes bad React → this catches it" check: cheap, fast, and it catches the bug classes that slip past a quick read. (react-doctor also ships an install command that drops editor skills/rules, and an experimental LSP server — but for an agent loop, running the CLI on your diff is the direct path.)