---

name: load-testing description: > Plan load testing — workload modeling, user journey scripts, ramp-up profiles, SLA thresholds, bottleneck identification, and capacity recommendations. TRIGGER when: user says /load-testing, needs to plan performance testing, or asks about stress testing and capacity validation. argument-hint: "[system or endpoint to load test]" user-invocable: true

Load Testing Plan

Process

Step 1: Define Test Objectives

Test Type	Purpose	When to Use
Load test	Validate under expected load	Pre-release
Stress test	Find breaking point	Capacity planning
Soak test	Find memory leaks, degradation	Pre-release
Spike test	Handle sudden traffic surges	Event planning

Step 2: Model Workload

Scenario	Requests/sec	User Pattern	% of Traffic
Browse catalog	X rps	Read-heavy	X%
Search	X rps	CPU-intensive	X%
Checkout	X rps	Write + external calls	X%
API integration	X rps	Burst pattern	X%

Step 3: Design Ramp-Up Profile

Phase	Duration	Load	Purpose
Warmup	5 min	10% target	Cache warmup
Ramp	10 min	10% → 100%	Gradual increase
Steady state	30 min	100% target	Sustained load
Peak	10 min	150% target	Headroom validation
Cool down	5 min	Decrease to 0	Graceful degradation

Step 4: Set SLA Thresholds

Metric	Target	Failure Threshold
p50 latency	< 200ms	> 500ms
p95 latency	< 500ms	> 2000ms
p99 latency	< 1000ms	> 5000ms
Error rate	< 0.1%	> 1%
Throughput	X rps	< 80% target

Step 5: Execute and Monitor

Monitor	What to Watch
Application	Latency, error rate, throughput
Infrastructure	CPU, memory, disk I/O, network
Database	Query time, connections, locks
External deps	API latency, timeout rate

Step 6: Analyze and Recommend

Bottleneck	Evidence	Recommendation
[component]	[metric at failure]	[scaling/optimization]

Output Format

## Load Test Report: [System]
### Test Config: [Type, duration, peak load]
### Results Summary
| Metric | Target | Actual | Status |
|--------|--------|--------|--------|
### Bottlenecks: [Identified with evidence]
### Capacity: [Max sustainable load]
### Recommendations: [Scaling/optimization actions]

Quality Checklist

• Workload model reflects production patterns
• Test environment matches production (or scaled proportionally)
• SLA thresholds defined before testing
• All system layers monitored during test
• Results analyzed for bottlenecks
• Recommendations are specific and actionable

Edge Cases

• For serverless, test cold start impact under load
• For distributed systems, test cross-region latency
• If test env differs from prod, document scaling assumptions