planner_skill

Overview

This skill generates staged project plans that break projects into resumable steps from design through deployment. It detects whether a project is greenfield or brownfield and classifies complexity into three tiers to produce appropriate stage sequences. Plans are file-based, resumable, and designed to validate locally before cloud deployment. It focuses on App Platform projects, infrastructure setup, and multi-step deployments.

How this skill works

The skill inspects the project workspace and user description to detect greenfield vs brownfield state and to classify the project as Tier 1 (simple), Tier 2 (database-backed), or Tier 3 (complex). It then emits a sequence of stage files (Plan/01-.. through final) with tasks, verification commands, and next-step guidance, storing runtime secrets in an ignored .env.secrets file. Critical validation gates such as local builds (doctl app dev build) are included early so cloud deploys are performed only after local verification.

When to use it

• Planning a new App Platform deployment or migration
• Breaking a complex deployment into resumable, verifiable stages
• Tracking multi-step infrastructure setup (databases, networking, workers)
• Converting an existing .do/app.yaml and Dockerfiles into an enhanced deployment plan
• Enforcing local-first build and validation before cloud pushes

Best practices

• Always run local builds and health checks (doctl app dev build) before cloud deploys
• Do not recreate valid existing artifacts; detect brownfield and enhance instead
• Keep secrets out of plan files; store sensitive values in .env.secrets (gitignored)
• Use the tier decision tree: map user requirements to Tier 1/2/3 to pick the right stage set
• For Tier 3 apps, deploy via GitHub Actions and set deploy_on_push: false in app spec

Example use cases

• Greenfield static site: generate a 5-stage plan for local design → cloud validate
• Node.js API + Postgres: produce a 7-stage plan that provisions DB, configures services, then deploys
• Event-driven microservices with Kafka/OpenSearch: create a 9-stage plan with clusters, topics, and GitHub Actions deployment
• Brownfield enhancement: detect existing Dockerfiles and .do/app.yaml, validate with doctl, and add verification and CI stages
• Migration: break a migration into isolated stages to allow safe rollback and resumable progress