App Analyzer Optimizer

App Analyzer & Optimizer (2026 Edition)

English | Bahasa Indonesia

English

Orchestration & Integration

• Upstream & Lifecycle: zero-to-prod-orchestrator (Phase 5 Optimization & Phase 6 QA), brainstorming, session-memory-manager.
• AST & Codebase Graph: living-codebase-ast-graph (circular dependency & module size analysis), scalability-clean-code (code smell & God class refactoring).
• Performance & Vitals: performance-web-vitals (Core Web Vitals LCP/INP profiling), production-ready-hardener (bundle audits & pre-flight hardening).
• Database & Backend Profiling: database-orm-expert (N+1 query & index profiling), coderabbit (automated AI code review).

Description

Deep application analysis and optimization skill. Performs architectural audits, bottleneck detection, bundle analysis, database query profiling, and AI-assisted code review across Next.js, React, Node.js, Go, Python, and Rust applications.

Trigger Conditions

• Auditing an existing codebase for architectural issues or technical debt.
• Detecting performance bottlenecks (slow API routes, large bundles, N+1 queries).
• Analyzing bundle size and suggesting code splitting opportunities.
• Reviewing code quality (complexity, duplication, dead code).
• Using AI-assisted tools to accelerate code review and analysis.

Analysis Protocol (5 Phases)

Phase 1: Architecture Discovery

bashCopy code
# Map the entire project structure
find . -type f -name "*.ts" -o -name "*.tsx" -o -name "*.go" | head -100
# Count lines per file (find largest files)
find . -name "*.ts" -exec wc -l {} + | sort -rn | head -20
# Find circular dependencies (TypeScript/JS)
npx madge --circular --extensions ts,tsx src/

Phase 2: Bundle Analysis (Next.js / Vite)

bashCopy code
# Next.js bundle analyzer
ANALYZE=true next build

# Or install dedicated tool
npx @next/bundle-analyzer

# Vite bundle visualization
npx vite-bundle-visualizer

# Check for duplicate dependencies
npx depcheck
npx bundle-phobia-cli check package.json

Key bundle red flags:

• Any single chunk > 500KB (uncompressed).
• Importing entire libraries (import _ from 'lodash' vs import debounce from 'lodash/debounce').
• Moment.js (replace with date-fns or Temporal).
• node_modules leaking into client bundle.

Phase 3: Database Query Analysis

sqlCopy code
-- PostgreSQL: Find slow queries
SELECT query, mean_exec_time, calls, total_exec_time
FROM pg_stat_statements
ORDER BY mean_exec_time DESC
LIMIT 20;

-- Find missing indexes (sequential scans on large tables)
SELECT schemaname, tablename, attname, n_distinct, correlation
FROM pg_stats
WHERE tablename = 'your_table'
ORDER BY n_distinct DESC;

-- Detect N+1 queries: Enable query logging
-- In Drizzle:
const db = drizzle(client, { logger: true });
-- In Prisma:
DATABASE_URL="...?connection_limit=5" prisma studio

Phase 4: AI-Assisted Code Review

Use AI tools to accelerate analysis:

AI review prompts for Gemini Agent:

Copy code
Analyze all files in src/features/ and identify:
1. Functions longer than 30 lines
2. Duplicated business logic across files
3. Missing error handling in async functions
4. Components that directly call APIs (violating separation of concerns)

Phase 5: Performance Profiling

Frontend (React / Next.js):

bashCopy code
# React DevTools Profiler — identify render bottlenecks
# Chrome DevTools > Performance tab > Record interaction

# Lighthouse CI — automated CWV tracking
npm install -g @lhci/cli
lhci autorun

Backend (Node.js):

bashCopy code
# Built-in Node.js profiler
node --prof server.js
node --prof-process isolate-*.log > processed.txt

# Clinic.js — flamegraph, bubble chart, doctor
npm install -g clinic
clinic doctor -- node server.js
clinic flame -- node server.js

Go:

bashCopy code
# Built-in pprof profiler
go tool pprof http://localhost:6060/debug/pprof/profile?seconds=30
go tool pprof -http=:8080 profile.out

Common Bottleneck Patterns & Fixes

Sovereign Execution, Token Efficiency & Invariants

Bahasa Indonesia

Integrasi Orkestrasi

• Hulu & Siklus Hidup: zero-to-prod-orchestrator (Optimasi Fase 5 & QA Fase 6), brainstorming, session-memory-manager.
• AST & Graf Basis Kode: living-codebase-ast-graph (dependensi sirkular & analisis ukuran modul), scalability-clean-code (refaktor code smell & God class).
• Performa & Web Vitals: performance-web-vitals (profiling Core Web Vitals LCP/INP), production-ready-hardener (audit bundle & pengerasan pra-rilis).
• Profiling Database & Backend: database-orm-expert (query N+1 & profiling indeks), coderabbit (code review AI otomatis).

Deskripsi

Skill analisis dan optimasi aplikasi mendalam. Melakukan audit arsitektur, deteksi bottleneck, analisis bundle, profiling query database, dan code review berbantuan AI di aplikasi Next.js, React, Node.js, Go, Python, dan Rust.

Kondisi Pemicu

• Mengaudit codebase yang ada untuk masalah arsitektur atau hutang teknis.
• Mendeteksi bottleneck performa (rute API lambat, bundle besar, query N+1).
• Menganalisis ukuran bundle dan menyarankan peluang code splitting.
• Meninjau kualitas kode (kompleksitas, duplikasi, kode mati).
• Menggunakan tool berbantuan AI untuk mempercepat code review dan analisis.

Protokol Analisis (5 Fase)

Fase 1: Penemuan Arsitektur

Petakan seluruh struktur proyek, hitung baris per file untuk menemukan file terbesar, dan deteksi dependensi circular dengan npx madge --circular.

Fase 2: Analisis Bundle (Next.js / Vite)

Gunakan ANALYZE=true next build atau npx vite-bundle-visualizer. Tanda bahaya bundle: chunk tunggal > 500KB, mengimpor library penuh, Moment.js, atau node_modules yang bocor ke bundle klien.

Fase 3: Analisis Query Database

Gunakan pg_stat_statements untuk menemukan query lambat. Cari scan sequential pada tabel besar (tanda indeks yang hilang). Aktifkan logging query di Drizzle/Prisma untuk mendeteksi pola N+1.

Fase 4: Code Review Berbantuan AI

Gunakan CodeRabbit untuk review PR otomatis, Sourcegraph Cody untuk pencarian semantik seluruh codebase, dan Gemini Agent dengan prompt analisis spesifik untuk menemukan fungsi panjang, logika bisnis yang diduplikasi, dan error handling yang hilang.

Fase 5: Profiling Performa

• Frontend: React DevTools Profiler, Lighthouse CI.
• Backend Node.js: node --prof, Clinic.js untuk flamegraph.
• Go: go tool pprof untuk CPU dan memory profiling.

Pola Bottleneck Umum & Perbaikan

Matriks Eksekusi Berdaulat, Efisiensi Token & Invarian