millionco/react-doctor is a meta-tooling agent built on the premise that 'your agent writes bad React, and this catches it.' Distinct from the basic harness-or-skill category, React Doctor sits one level up: it reads the output of other coding agents and flags React-specific anti-patterns, accessibility regressions, performance smells (unnecessary re-renders, missing memoization, prop-drilling) and architectural drift. The repo trended at #7 on GitHub on 2026-05-09 with 636 stars in 24h, marking it as one of the canonical entries in the emerging AI-on-AI tier — alongside AI-Trader (agent auditing trading agents) and the broader 'react-doctor catches bad agent code' category. Pairs with the Convergence #6 cluster (HN's 'AI is breaking two vulnerability cultures' + Mozilla's 271-bug Mythos run) where AI-on-AI security tooling is consolidating into a recognized news beat.
millionco/react-doctor is a meta-tooling agent that operates on the output of other coding agents rather than producing code itself. The premise is honest: AI-generated React code looks plausible at first glance but quietly accumulates anti-patterns — unnecessary re-renders, missing keys, accessibility regressions, prop-drilling instead of composition. React Doctor reads that output and flags it before it ships.
The agent specializes in React-specific drift that generic linters miss. It understands JSX/TSX semantics, hooks rules, the React rendering lifecycle, and modern patterns (Server Components, Suspense boundaries, concurrent features). Crucially, it operates as a post-generation review step — slot it into your harness after Codex or Claude Code finishes a task, and it produces a critique-style report against the diff.
Teams using AI coding agents for frontend work — particularly when shipping at the velocity tokenmaxxing implies — hit a quality wall. The first 50 features land fast; the next 50 introduce subtle React-specific bugs that don’t surface until a Lighthouse run or an a11y audit much later. React Doctor is positioned as the catch-fast, fix-fast layer between agent output and merge.
React Doctor is one of the canonical entries in the AI-on-AI meta-tooling tier that materialized on the GitHub trending board in May 2026. Together with AI-Trader (auditing trading-agent output) and the broader “agent-that-audits-agents” category, the tier sits one level above the basic harness/skills layer and represents a higher-leverage role with cleaner ROI than basic code generation.
If you’ve adopted agent-skills or obra/superpowers and run Claude Code or Codex daily against a frontend codebase, React Doctor is the natural complement — a final review pass that catches the long-tail of React-specific issues your harness’s general code-review skill misses.
Persistent memory layer for AI coding agents — benchmark-backed (95.2% on LongMemEval-S), 92% fewer tokens per session vs full-context pasting, zero manual memory.add() calls.
Open-source AI pair programming tool that works in your terminal to edit code across your entire repository.
AWS's AI-powered coding assistant that helps developers build, deploy, and optimize applications on AWS with code generation and transformation.