deployment_skill

Overview

This skill deploys applications to DigitalOcean App Platform using GitHub Actions and the digitalocean/app_action/deploy@v2 workflow. It emphasizes PUSH-mode deployments where the app spec (.do/app.yaml) lives in the repository and CI drives creation, updates, and deployments. The workflow patterns support staging, production, PR previews, and controlled rollouts.

How this skill works

You store an App Platform spec (.do/app.yaml) in the repo and use GitHub Actions to run the app_action deploy step. The action can create or update the app, apply spec changes, and trigger code rebuilds in one step. Environment secrets and repo variables (DIGITALOCEAN_ACCESS_TOKEN, DO_PROJECT_ID, DATABASE_URL, etc.) are set per GitHub environment to enforce isolation between staging and production.

When to use it

• Setting up a CI/CD pipeline to deploy to DigitalOcean App Platform
• Maintaining separate staging and production environments with protected approvals
• Managing secrets and project IDs per environment securely in GitHub
• Running PR preview deployments for testing pull requests
• Automating spec changes and code deployments with a single workflow

Best practices

• Keep .do/app.yaml in the repo and set deploy_on_push: false so GitHub Actions controls deployments
• Use digitalocean/app_action/deploy@v2 to handle both spec updates and code deployments
• Create GitHub environments (staging, production) and set secrets per environment, not globally
• Store DO_PROJECT_ID as a repo variable per environment and use secrets for tokens and DB URLs
• Add a debug container entry to the spec for troubleshooting complex integrations

Example use cases

• Single-environment automatic deploy on main branch using a minimal workflow
• Two-stage pipeline: automatic staging deploys plus protected production deploys with approval
• PR preview environment automatically created for pull request testing
• Rollback workflow that triggers a previous deployment via doctl or the action
• Debugging integration issues by deploying a temporary debug worker container and running diagnostics