How to use the Deployment Patterns agent skill

What the deployment-patterns skill makes the agent do

An agent skill is a folder of instructions that your AI agent reads when a task matches it. This one is a single file, SKILL.md, that contains a guide to deployment workflows, CI/CD pipeline patterns, Docker containerization, health checks, rollback strategies and production-readiness checklists for web applications.

The guide is written in Spanish. The agent uses it to advise you or to draft deployment files for your project. The file contains templates and advice, not executable deployment scripts, so the agent is not deploying anything on its own.

• Explains rolling, blue-green and canary deployment strategies, with when each one fits.
• Provides multi-stage Dockerfile templates for Node.js, Go and Python/Django, each with a health check and a non-root user.
• Outlines a GitHub Actions pipeline that tests, builds a container image and deploys on merge to main.
• Includes a health check endpoint pattern and Docker good and bad practices.

Install the deployment-patterns skill

You can install the skill for yourself across all projects, or for a single project only. Download the ZIP file from this page first. Then choose one of the options below.

For a personal install, unzip the file into ~/.claude/skills/. For a project-only install, unzip it into the .claude/skills/ folder inside your project. In the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

• Personal: unzip into ~/.claude/skills/
• Project only: unzip into .claude/skills/ in your project folder
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

How to trigger the skill with natural requests

You do not need a special command. Describe your task in plain language and mention the deployment work you need. The skill is meant to apply when you ask for help with one of these tasks: configuring CI/CD pipelines, containerizing an application with Docker, planning a deployment strategy (blue-green, canary or rolling), implementing health checks and readiness probes, preparing a production release, or setting up environment-specific settings.

Be specific about your stack and goal. The more the agent knows about your app, the more useful the draft will be.

• "Containerize my Node.js API with Docker, using a multi-stage build and a non-root user."
• "Set up a GitHub Actions pipeline that runs lint and tests, then builds an image on merge to main."
• "Should I use canary or blue-green for a high-traffic checkout service? Explain the trade-offs."
• "Add a /health endpoint and a detailed health check that reports database status."
• "Give me a production readiness checklist before we launch this Django app."

Understand the deployment strategies

The skill compares three approaches. Rolling deployment replaces instances gradually, so old and new versions run at the same time and your changes must be backward compatible. Blue-green runs two identical environments and switches traffic at once, which gives an instant rollback but needs twice the infrastructure during the release. Canary sends a small share of traffic to the new version first and widens it if metrics look good, which catches problems with real traffic but requires traffic-splitting infrastructure and monitoring.

• Rolling: default choice for standard releases with backward-compatible changes.
• Blue-green: for critical services where zero tolerance for problems matters.
• Canary: for high-traffic services, risky changes, or work behind feature flags.

Use the Docker and health check templates

The skill includes multi-stage Dockerfile templates for three stacks. Each uses pinned image tags such as node:22-alpine, runs as a non-root user, and includes a HEALTHCHECK instruction. The Node.js template expects a /health route, and the Django template checks /health/. The skill also suggests a simple health endpoint that returns status ok, plus a detailed one that checks dependencies such as the database and returns 503 when a check fails.

• Node.js template: builds with npm, serves dist/server.js on port 3000.
• Go template: builds a static binary and serves on port 8080.
• Python/Django template: uses uv to install requirements and runs gunicorn on port 8000 with 4 workers.
• Avoid :latest tags, running as root, and storing secrets in the image.

Review the GitHub Actions pipeline

The pipeline example has three jobs. The test job runs lint, typecheck and tests with coverage. The build job runs only on main, builds the image and pushes it to the GitHub Container Registry (ghcr.io), tagged with the commit SHA. The deploy job runs only on main and uses the production environment.

The deploy step is a placeholder. It only echoes the commit SHA. The file includes commented examples for Railway, Vercel and kubectl, but you must replace the placeholder with the command for your own platform.

• Test job: lint, typecheck, tests with coverage, upload coverage artifact.
• Build job: needs test, pushes to ghcr.io using GITHUB_TOKEN.
• Deploy job: needs build, uses the production environment, currently only a placeholder.