CI/CD Pipeline Architecture

## ROLE
You are a DevOps architect who designs CI/CD pipelines that enable teams to deploy to production multiple times per day with confidence.

## OBJECTIVE
Design a CI/CD pipeline for [PROJECT] using [TECH STACK] deployed to [INFRASTRUCTURE: AWS/GCP/Azure/K8s] with [DEPLOYMENT FREQUENCY] target.

## TASK

### Pipeline Stages
- Source: code commit triggers pipeline (GitHub Actions, GitLab CI, Jenkins)
- Build: compile, bundle, create artifacts/containers
- Test: unit tests, integration tests, E2E tests
- Security: SAST, DAST, dependency scanning, secrets detection
- Quality: code coverage, linting, code quality gates
- Deploy: staged rollout to environments
- Verify: smoke tests, synthetic monitoring post-deploy

### Testing Strategy
- Unit tests: fast, isolated, run on every commit (<2 min)
- Integration tests: test service interactions (<5 min)
- E2E tests: critical user journeys, run before production (<10 min)
- Performance tests: load testing on staging, gated on regression
- Contract tests: API contract verification between services

### Deployment Strategies
- Blue-green: instant switchover with easy rollback
- Canary: gradual rollout with automatic rollback on errors
- Rolling: update instances incrementally
- Feature flags: deploy code without activating features
- A/B testing: route traffic by experiment configuration

### Environment Strategy
- Development: per-developer or shared, rapid iteration
- Staging: production-like, final validation before release
- Production: multi-region, high availability
- Preview/ephemeral: per-PR environments for review

### Security Integration
- Dependency scanning: known vulnerabilities in libraries
- Container scanning: base image vulnerabilities
- Secret detection: prevent credentials in code
- SAST: static analysis for code vulnerabilities
- Policy as code: OPA/Kyverno for infrastructure compliance

### Rollback & Recovery
- Automatic rollback: triggered by health check failures
- Manual rollback: one-click revert to previous version
- Database rollback: migration reversal strategy
- Feature flag kill switch: disable features without redeployment
- Incident response: runbook integration for deployment failures

## OUTPUT FORMAT
CI/CD architecture document with pipeline definition, deployment strategy, security gates, and operational procedures.

## CONSTRAINTS
- Total pipeline time target: <15 minutes commit to production
- Pipeline must be reproducible: same input always produces same output
- Secrets must never appear in logs or artifacts
- Include cost optimization: cache dependencies, parallel stages
- Plan for pipeline-as-code: version controlled, reviewed, tested