mcpField
Q
qe-test-environment-management

qe-test-environment-management

github.com/proffesor-for-testingTesting & quality51983 installsUpdated today
Repository

About qe-test-environment-management

A testing skill for planning and operating local, CI, staging, and production-like test environments. It covers reproducibility, production parity, service mocks, and infrastructure cost controls.

Key features

  1. 01Define environment types
  2. 02Use Docker for reproducibility
  3. 03Check production parity
  4. 04Mock external services
  5. 05Reduce infrastructure costs

Use cases

  1. 01Set up test infrastructure
  2. 02Investigate environment-specific failures
  3. 03Align development and production environments
  4. 04Reduce test environment expenses

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/proffesor-for-testing/agentic-qe --skill qe-test-environment-management

After installing, type/qe-test-environment-managementin Claude Code to confirm it loaded.

Details

From the project README

Test Environment Management

<default_to_action> When managing test environments:

  1. DEFINE environment types (local, CI, staging, prod)
  2. CONTAINERIZE with Docker for consistency
  3. ENSURE parity with production (same versions, configs)
  4. MOCK external services (service virtualization)
  5. OPTIMIZE costs (auto-shutdown, spot instances)

Quick Environment Checklist:

  • Same OS/versions as production
  • Same database type and version
  • Same configuration structure
  • Containers for reproducibility
  • Auto-shutdown after hours

Critical Success Factors:

  • "Works on my machine" = environment inconsistency
  • Infrastructure as Code = repeatable environments
  • Service virtualization = test without external dependencies </default_to_action>

Quick Reference Card

When to Use
  • Setting up test infrastructure
  • Debugging environment-specific failures
  • Reducing test infrastructure costs
  • Ensuring dev/prod parity
Environment Types
TypePurposeLifetime
LocalFast feedbackDeveloper session
CIAutomated testsPer build (ephemeral)
StagingPre-prod validationPersistent
ProductionCanary/syntheticContinuous
Dev/Prod Parity Checklist
ItemMust Match
OSSame version
DatabaseSame type + version
DependenciesSame versions
ConfigSame structure
Env varsSame names

Docker for Test Environments

# docker-compose.test.yml
version: '3.8'

services:
  app:
    build: .
    ports:
      - "3000:3000"
    environment:
      NODE_ENV: test
      DATABASE_URL: postgres://postgres:password@db:5432/test
    depends_on:
      - db
      - redis

  db:
    image: postgres:15
    environment:
      POSTGRES_DB: test
      POSTGRES_PASSWORD: password

  redis:
    image: redis:7

Run tests in container:

docker-compose -f docker-compose.test.yml up -d
docker-compose -f docker-compose.test.yml exec app npm test
docker-compose -f docker-compose.test.yml down

Infrastructure as Code

# test-environment.tf
resource "aws_instance" "test_server" {
  ami           = "ami-0c55b159cbfafe1f0"
  instance_type = "t3.medium"

  tags = {
    Name         = "test-environment"
    Environment  = "test"
    AutoShutdown = "20:00" # Cost optimization
  }
}

resource "aws_rds_instance" "test_db" {
  engine                  = "postgres"
  engine_version          = "15"
  instance_class          = "db.t3.micro"
  backup_retention_period = 0 # No backups needed for test
  skip_final_snapshot     = true
}

Service Virtualization

// Mock external services with WireMock
import { WireMock } from 'wiremock-captain';

const wiremock = new WireMock('http://localhost:8080');

// Mock payment gateway
await wiremock.register({
  request: {
    method: 'POST',
    url: '/charge'
  },
  response: {
    status: 200,
    jsonBody: { transactionId: '12345', status: 'approved' }
  }
});

// Tests use mock instead of real gateway

Cost Optimization

# Auto-shutdown test environments after hours
0 20 * * * aws ec2 stop-instances --instance-ids $(aws ec2 describe-instances \
  --filters "Name=tag:Environment,Values=test" \
  --query "Reservations[].Instances[].InstanceId" --output text)

# Start before work hours
0 7 * * 1-5 aws ec2 start-instances --instance-ids $(aws ec2 describe-instances \
  --filters "Name=tag:Environment,Values=test" \
  --query "Reservations[].Instances[].InstanceId" --output text)

Use spot instances (70% savings):

resource "aws_instance" "test_runner" {
  instance_type = "c5.2xlarge"
  instance_market_options {
    market_type = "spot"
    spot_options {
      max_price = "0.10"
    }
  }
}

Agent-Driven Environment Management

// Provision test environment
await Task("Environment Provisioning", {
  type: 'integration-testing',
  services: ['app', 'db', 'redis', 'mocks'],
  parity: 'production',
  lifetime: '2h'
}, "qe-test-executor");

// Chaos testing in isolated environment
await Task("Chaos Test Environment", {
  baseline: 'staging',
  isolate: true,
  injectFaults: ['network-delay', 'pod-failure']
}, "qe-chaos-engineer");

Agent Coordination Hints

Memory Namespace
aqe/environment-management/
├── configs/*            - Environment configurations
├── parity-checks/*      - Dev/prod parity results
├── cost-reports/*       - Infrastructure costs
└── service-mocks/*      - Service virtualization configs
Fleet Coordination
const envFleet = await FleetManager.coordinate({
  strategy: 'environment-management',
  agents: [
    'qe-test-executor',       // Provision environments
    'qe-performance-tester',  // Environment performance
    'qe-chaos-engineer'       // Resilience testing
  ],
  topology: 'sequential'
});

Related Skills


Remember

Environment inconsistency = flaky tests. "Works on my machine" problems come from: different OS/versions, missing dependencies, configuration differences, data differences.

Infrastructure as Code ensures repeatability. Version control your environment configurations. Spin up identical environments on demand.

With Agents: Agents automatically provision test environments matching production, ensure parity, mock external services, and optimize costs with auto-scaling and auto-shutdown.

Comments

Markdown supported