| name | bottleneck-patterns |
| description | This skill should be used when the user describes slow response times, high CPU usage, memory pressure or leaks, database slowness, network issues, frame drops, ANR errors, battery drain, app jank, a performance regression after deploy, or app lagging under load; or asks "my database queries are taking forever", "why does my Flutter app stutter", "what's causing my API to slow down under load", or describes any observable performance symptom. Provides domain-specific bottleneck pattern recognition and diagnostic guidance for web, mobile (Android/iOS/Flutter), desktop, and API/backend applications. |
| argument-hint | [domain] e.g. web | mobile | api | database | memory | cpu | network |
| allowed-tools | Read |
Bottleneck Patterns
Performance Bottleneck Patterns
Identify which domain matches the symptom, then apply the corresponding pattern. For the full symptom → pattern → fix lookup tables, see references/bottleneck-lookup.md.
Domain Quick Reference
Response Time: Tail latency (p99 >> p50), throughput ceiling, cold-path slowness, periodic spikes
Web: Render-blocking resources, large JS bundles, layout thrashing, missing CDN, high INP
API/Backend: N+1 queries, missing indexes, connection pool exhaustion, synchronous I/O, over-fetching
Memory: Monotonic heap growth (leak), burst allocation GC churn, premature promotion, unbounded cache
CPU: Busy-wait loops, hot codec (serialization), uncached regex, lock contention
Mobile (Android): ANR on main thread, UI jank, overdraw, WakeLock abuse, battery drain from background work
Mobile (iOS): Main thread I/O, deep autolayout graph, retain cycles, Core Data on main thread
Mobile (Flutter): Expensive build() methods, shader compilation jank, large isolate messages, heavy list cells
Networking: Chatty API, missing compression, no connection reuse, oversized images, sequential loading
Database: Missing indexes (EXPLAIN ANALYZE first), SELECT *, missing composite index, ORM over-fetching
Diagnostic Approach
- Match the symptom to a domain from the quick reference above
- Look up the specific pattern and fix in
references/bottleneck-lookup.md
- Confirm with data before implementing: run
EXPLAIN ANALYZE, attach a profiler, or measure with a benchmark
- Apply the
impact-matrix skill to prioritize multiple findings
Cross-Domain Signals
Some symptoms span multiple domains — check these cross-cutting patterns first:
- Periodic latency spikes → GC pause + background job + cache invalidation (all three at once)
- Slow under load only → thread/connection pool contention (not a code bug)
- Fast p50, slow p99 → lock wait or GC stop-the-world on a fraction of requests
- Memory growing then dropping → burst allocation, not a leak; reduce short-lived object creation
- First request slow, subsequent fast → cold-path (JIT, DNS, connection init) — warm up or cache
Additional Resources
references/bottleneck-lookup.md — Full symptom → pattern → fix lookup tables by domain (response time, web, API/backend, memory, CPU, mobile, networking, database)
