frontend-sentinel

Use the DIFF_SCAN mode to perform a PR-aware review of changed files. It categorizes findings across 17 bug types such as hydration mismatch and stale closures, each rated by severity, confidence, user impact, and scale risk.

The scan detects render cycle bugs, useEffect antipatterns, broken memoization, state ownership leaks, server/client boundary violations, cache invalidation bugs, bundle bloat, accessibility gaps, and AI-generated code smells.

Yes, the pre-scan inference step identifies the Next.js version and router type before analysis. It then applies router-specific rules such as use client placement, caching directives, and server action boundaries for App Router projects.

Yes, FILE_SCAN analyzes one component or file, while FOLDER_SCAN covers a directory and REPO_SCAN covers the full frontend codebase. For large repositories it proposes a phased scan plan starting with routing, data, state, and rendering.

Findings include a confidence rating because static analysis cannot measure actual runtime behavior such as real render counts or memory retention. Performance claims like memoization benefit still require measurement in real browsers to confirm.