Reliable CI/CD and Cloud Deployments
Build and repair CI/CD, containers, and cloud infrastructure with health-gated releases, explicit rollback, and useful monitoring.
Before & after
Same prompt, same agent — the only difference is the skill.
That's a solid project to automate your deployments. There are quite a few decisions to make around tooling, cloud platform, containerization strategy, and testing gates. Are you on AWS, GCP, or Azure? What's your deployment frequency—are we pushing multiple times a day or once a week? Do you have a current deployment process that we're replacing? I'd also want to know about your team's experience with Docker and Kubernetes. Once I understand your constraints better, I can recommend a specific pipeline architecture and walk through the setup.
CI/CD PIPELINE ARCHITECTURE: Node.js Microservices Target: Deploy 15 microservices, 10x per day, zero-downtime updates VERSION CONTROL TRIGGER • GitHub: Pull request opened → lint, test, build image • Merge to main → deploy to staging • Tag release (v1.2.3) → deploy to production TEST GATE (5 min, blocks merge) • Unit tests: Jest (target: 80%+ coverage) • Linting: ESLint + Prettier (auto-fix on commit) • Integration tests: Docker Compose (test DB + API interactions) • Fail fast: Block merge if coverage drops below baseline BUILD STAGE (GitHub Actions) ``` steps: - Build image: docker build -t service:$COMMIT_SHA - Push to ECR: aws ecr push $IMAGE - Tag latest: docker tag $IMAGE service:latest - Scan for vulnerabilities: Trivy scan - Store metadata: Image digest, test results, coverage ``` DEPLOYMENT STAGING (automatic after merge to main) • Platform: AWS ECS on Fargate (no server management) • Strategy: Blue-green deployment (0 downtime) → Old cluster (blue) stays live → New cluster (green) spins up, runs tests → Traffic switches when green is healthy • Rollback: Instant (traffic switches back to blue) • Monitoring: Wait 5 min for error rates, latency (must stay <50ms p95) DEPLOYMENT PRODUCTION (manual gate, but automated after approval) • Approval: Slack notification, engineering lead confirms • Rollout: Canary (5% traffic) for 10 minutes → Monitor: Error rates, latency, DB query times → Decision: All traffic (95%) if metrics green • Instant rollback: Traffic reverts to previous version if errors spike MONITORING & ALERTS • CloudWatch: CPU, memory, request latency per service • Alert on: Error rate >1%, latency p95 >100ms, deployment failures • Post-deployment: Dashboard shows health (green/yellow/red) for 30 min COST • ECR storage: ~50 USD/month (images per microservice) • ECS + Fargate: ~200 USD/month (t3.medium equivalent) • Data transfer: ~30 USD/month • Total: ~280 USD/month
About this skill
“The pipeline is green, but deploys still break production.” This skill builds and repairs the release path across CI/CD, containers, infrastructure, and monitoring, with explicit checks for whether a deployment can serve traffic and recover safely.
What it does
- Diagnoses failed or flaky deployments by tracing the first failing boundary from source revision to serving traffic.
- Builds pipelines that promote the same immutable artifact between environments and prevent overlapping production changes.
- Creates container and rollout configuration that accounts for startup, readiness, graceful shutdown, and spare capacity.
- Plans infrastructure changes with state protection, drift review, scoped credentials, and explicit handling of destructive changes.
- Defines release gates, rollback triggers, and database compatibility checks using the application's actual failure modes.
- Connects user-facing metrics to actionable alerts and short recovery runbooks.
What you get
A focused fix or a coordinated set of pipeline, container, and infrastructure changes, sized to your request. Each deliverable includes the assumptions, verification steps, recovery path, and evidence still needed before calling it production-ready.
Best for
Application teams setting up repeatable cloud deployments, replacing manual infrastructure changes, or fixing releases that pass CI but fail under real traffic.
SKILL.md preview
---
name: reliable-ci-cd-and-cloud-deployments
description: >-
Build and repair CI/CD pipelines, container deployments, and cloud infrastructure by tracing release dependencies, promoting immutable artifacts, and defining health gates and recovery paths. Use when the user says "set up our CI/CD", "my deploys keep causing downtime", "the pipeline passes but production fails", "containerize this service", "move our infrastructure into code", "add monitoring and alerts", or "make our releases safer".
---
# Reliable CI/CD and Cloud Deployments
… (sign up to view the full skill)More development skills
View all Development skills →Chaos Experiment and Game Day Planner
Plans chaos experiments and game days: hypothesis, blast radius, abort conditions, tool manifests, rollback, and post-run review.
Claude Code Cost and Latency Routing
Configure Claude Code Router with model tiers, explicit fallbacks, compatibility checks, and a cost-aware validation plan.
CI/CD Pipeline Builder with Gated Deploys
Generates CI/CD workflows from your repo's real stack: cached fast-fail checks, pinned actions, and staging-to-prod deploys with gates.
Terraform Modules and Safe Deployments
Get reusable Terraform modules, safe state migrations, explicit provider wiring, and CI gates tied to reviewed plans.