Skip to content

artubss/SKILLS-CLAUDE-CODE / favorites/tpl-situacao-deployment-devops

tpl-situacao-deployment-devops

Verified

Template do pack (situacao/10-deployment-devops.md). Orienta o agente em tarefas situacionais como debug, seguranca e refactor alinhado a esse contexto.

SITUATION: Deployment & DevOps Pipeline Setup

Skill gerado a partir do pack templates-claude-code. Arquivo de origem: situacao/10-deployment-devops.md. Use como baseline e adapte ao projeto antes de mudancas grandes.

Conteudo do template

CONTEXT

Use this configuration when setting up or improving the deployment pipeline — building CI/CD, configuring environments, managing secrets in production, setting up health checks, establishing rollback procedures, or preparing for a production go-live. The goal is deployments that are automated, observable, reversible, and boring. "Boring" is good — exciting deployments are emergencies.

OBJECTIVES

  • Achieve fully automated deployments with no manual steps
  • Establish environment parity (dev/staging/production behave the same)
  • Make rollbacks fast and practiced before they're needed in an emergency
  • Set up monitoring BEFORE going live, not after the first incident
  • Make secrets management systematic and auditable

APPROACH RULES

  1. Every environment is cattle, not pets. Environments must be reproducible from code and configuration. If a server needs manual steps to configure, that knowledge is lost when the server is replaced.

  2. Environment parity is non-negotiable. Staging must mirror production in: OS, runtime versions, database engine and version, memory/CPU class (within budget). Any difference is a potential "works in staging" failure.

  3. Deploy from artifacts, not from source. Build once, deploy the artifact. Never run npm install or build steps on production servers. The artifact that passes staging is the artifact deployed to production.

  4. Feature flags over long-lived branches. If a feature is too risky to deploy to all users, use a feature flag. Don't maintain a branch for months. Dark launch → canary → full rollout.

  5. Health checks are load-balancer ready before deployment. Every service must expose a /health endpoint that returns 200 when healthy. The load balancer must be configured to stop routing traffic to unhealthy instances.

  6. Monitoring is a prerequisite for go-live, not a follow-up task. If you cannot observe the service after deployment, you cannot safely deploy. Error rate, latency P95, and CPU/memory must be visible on a dashboard before the first production deployment.

  7. Rollback must be practiced. Run a rollback drill in staging before the first production deployment. The procedure must be documented and under 10 minutes.

ROUTING TABLE

If you encounterThen
Secrets in code or config filesStop everything. Move to secrets manager (AWS Secrets Manager, Vault, GitHub Secrets). Rotate all exposed secrets.
npm install running on production serverWrong pattern. Build Docker image in CI. Push to registry. Deploy the image.
No staging environmentCreate one before continuing. Deployments without staging are exploratory surgery without anesthesia.
Database migration as part of deploymentRun migrations BEFORE deploying new code, using expand-contract. Never migrate during traffic spike.
Manual approval gate before productionUse protected environments in GitHub Actions. Require explicit approval from named individuals.
"Works on my machine" reportsStandardize on Docker for local development. If it runs in Docker locally, it runs the same way in CI.
No rollback procedure documentedWrite rollback procedure. Test it in staging. Link it from deployment runbook.
Zero-downtime deployment neededUse: rolling update, blue-green deployment, or canary deployment. Configure readiness probes.
Long-running deploy causing downtimeImplement: graceful shutdown signal handling, connection draining (30s), health check that returns unhealthy before process exits
New team member deploying for first timePair program the first deploy. Runbook must be sufficient for them to do it alone the second time.

Deployment Checklist (Pre-Deploy)

Code Readiness

  • All CI checks passing (lint, tests, build, security scan)
  • PR reviewed and approved
  • Database migrations written and tested on staging
  • Feature flags configured (if applicable)
  • CHANGELOG.md updated

Environment Readiness

  • Secrets updated in secrets manager (if changed in this release)
  • External service credentials valid and tested
  • Staging deployment successful with E2E smoke tests passing
  • Database backup confirmed (before migration-heavy releases)

Monitoring Readiness

  • Dashboards displaying current baseline metrics
  • Alert thresholds defined: error rate > 1%, P95 latency > 2s
  • On-call person informed of deployment
  • Runbook link bookmarked for quick access

Rollback Readiness

  • Previous version artifact available in registry
  • Rollback procedure documented and accessible
  • Database rollback script prepared (if migration involved)
  • Decision criteria defined: what metric/threshold triggers rollback?

Zero-Downtime Deployment Strategies

StrategyUse WhenComplexityRollback Speed
Rolling UpdateStateless services, N instancesLowFast (redeploy old version)
Blue-GreenCritical services, need instant cutoverMediumInstant (switch load balancer)
CanaryHigh risk change, need gradual validationHighFast (remove canary)
Feature FlagCode shipped but not activatedLowInstant (toggle flag)

CI/CD Pipeline Template (GitHub Actions)

name: Deploy to Production

on:
  push:
    branches: [main]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - run: npm ci
      - run: npm run lint
      - run: npm test -- --coverage
      - run: npm run build

  deploy-staging:
    needs: test
    environment: staging
    runs-on: ubuntu-latest
    steps:
      - name: Deploy to staging
        run: ./scripts/deploy.sh staging ${{ github.sha }}
      - name: Run smoke tests
        run: npm run test:smoke -- --env staging

  deploy-production:
    needs: deploy-staging
    environment: production  # Requires manual approval
    runs-on: ubuntu-latest
    steps:
      - name: Deploy to production
        run: ./scripts/deploy.sh production ${{ github.sha }}
      - name: Verify health checks
        run: ./scripts/verify-health.sh production

DO NOT

  • DO NOT deploy on Fridays or before holidays unless it's an emergency hotfix
  • DO NOT deploy without a monitoring dashboard open and visible
  • DO NOT deploy during high-traffic hours — choose low-traffic windows
  • DO NOT skip staging — ever
  • DO NOT manually modify production servers — all changes through code
  • DO NOT store secrets in environment variables baked into Docker images
  • DO NOT deploy without notifying the on-call person
  • DO NOT run database migrations and deploy application code in the same step

OUTPUT FORMAT

For each deployment, produce a Deployment Record:

## Deployment Record

**Version:** v2.4.1
**Environment:** Production
**Date:** 2024-01-15 14:30 UTC
**Deployer:** [name]
**Changes:** [link to CHANGELOG / PR list]

### Pre-Deploy Checklist
[x] All CI checks passing
[x] Staging deployment successful
[x] Database migrations tested on staging
[x] On-call notified
[x] Rollback procedure accessible

### Deployment Timeline
- 14:30: Migration started
- 14:32: Migration complete. Row counts verified.
- 14:33: Rolling update started (3 of 5 instances updated)
- 14:36: Rolling update complete. All instances healthy.
- 14:37: Smoke tests passed.

### Post-Deploy Metrics (15 min observation)
- Error rate: 0.02% (baseline: 0.01%) ✅
- P95 latency: 245ms (baseline: 230ms) ✅
- CPU: 42% (baseline: 38%) ✅

**Status: Successful ✅**

QUALITY GATES

  • Deploy pipeline fully automated — zero manual SSH steps
  • /health endpoint returns 200 on healthy instance, 503 on unhealthy
  • Rollback tested in staging before first production deploy
  • Zero secrets in codebase, config files, or Docker images
  • Staging and production have matching runtime versions (verify: node --version)
  • Monitoring dashboard created and showing pre-deploy baseline
  • Deployment runbook accessible to all team members
  • Recovery time < 10 minutes for standard rollback scenario
  • CI build produces immutable artifact tagged with git SHA
  • Deployment notification sent to team channel on success/failure
finishing-a-development-branchobra/superpowersUse when implementation is complete, all tests pass, and you need to decide how to integrate the workStars: 319kUpdated Claude CodeCodexVerifiedInstall
systematic-debuggingobra/superpowersUse when encountering any bug, test failure, or unexpected behavior, before proposing fixesStars: 319kUpdated Claude CodeCodexVerifiedInstall
test-driven-developmentobra/superpowersUse when implementing any feature or bugfix, before writing implementation codeStars: 319kUpdated Claude CodeCodexVerifiedInstall
skill-creatoranthropics/skillsCreate new skills, modify and improve existing skills, and measure skill performance. Use when users want to create a skill from scratch, edit, or optimize an existing skill, run evals to test a skill, benchmark skill performance with variance analysis, or optimize a skill's description for better triggering accuracy.Stars: 197kUpdated Claude CodeCodexVerifiedInstall