arch-performance-optimization

```

Performance Issue?

├── Database (slow queries, N+1, indexes, pagination)

│ → Invoke database-optimization skill (covers all DB patterns)

├── Frontend (rendering, bundle size, change detection)

│ → ⚠️ MUST READ docs/claude/frontend-patterns.md

│ → Key: OnPush, trackBy, lazy loading, virtual scroll, tree-shaking

├── API/Endpoint (response time, payload, serialization)

│ → ⚠️ MUST READ references/performance-patterns.md (parallel queries, caching, DTOs)

├── Background Jobs (throughput, batch processing)

│ → ⚠️ MUST READ references/performance-patterns.md (bounded parallelism, batch ops)

└── Cross-Service (message bus, eventual consistency)

→ ⚠️ MUST READ references/performance-patterns.md (payload size, idempotency)

```

1. Identify bottleneck type using decision tree above
2. Measure baseline (response time, query count, bundle size)
3. Route to correct sub-tool or reference
4. Apply patterns from the routed resource
5. Verify improvement against baseline
6. Monitor for regressions

• Parallel tuple queries: var (a, b) = await (queryA, queryB);
• Eager loading: repo.GetAllAsync(filter, ct, e => e.Related)
• Projections: .Select(e => new { e.Id, e.Name }) instead of full entity
• Full-text search: searchService.Search(q, text, Entity.SearchColumns())
• Batch updates: repo.UpdateManyAsync(items, dismissSendEvent: true, checkDiff: false)
• Paged processing: PageBy(skip, take) at database level

For detailed patterns, profiling commands, and anti-patterns:

⚠️ MUST READ: .claude/skills/arch-performance-optimization/references/performance-patterns.md

Present findings and optimization plan. Wait for explicit user approval before making changes -- performance optimizations can have wide-reaching side effects.

• Always plan and break many small todo tasks
• Always add a final review todo task to review the works done at the end to find any fix or enhancement needed