| name | docker |
| description | Use when writing or reviewing Dockerfiles, docker-compose files, or .dockerignore for any backend service. Covers multi-stage builds, container security hardening, compose orchestration, advanced build patterns, and a Docker review checklist for all 4 stacks: NestJS/Node 24, Spring Boot WebFlux 3.5.x/Java 21, Python FastAPI 3.14, and TypeScript/Fastify/Node 24. |
Iron Law
NO DOCKERFILE WITHOUT READING reference/dockerfiles.md FIRST — each stack has specific base images, build tools, and health check patterns that must be used exactly
Docker
Production-ready containerization for NestJS/Node 24, Spring Boot WebFlux 3.5.x/Java 21, Python FastAPI 3.14, and TypeScript/Fastify/Node 24. All patterns follow: multi-stage build, non-root user, HEALTHCHECK, .dockerignore.
When to Use
- Writing a Dockerfile for any backend service
- Writing or updating docker-compose files for local dev or production
- Reviewing an existing Dockerfile for security, size, or correctness
- Adding build cache optimization (
--mount=type=cache)
- Setting up Docker Compose secrets for production
- Building multi-architecture images (linux/amd64 + linux/arm64)
- Debugging container build failures, slow builds, or large image sizes
How This Skill Relates to Others
| Skill | Scope |
|---|
| docker (this skill) | Container definitions, compose files, build patterns |
| java-spring-api | Java application code — loads docker skill for Dockerfile |
| nestjs-api | NestJS templates include Dockerfile — this skill extends those patterns |
| python-dev | FastAPI code — loads docker skill for Dockerfile |
| agentic-ai-dev | Python agent code — has deployment reference that this skill augments |
Process
Step 1: Choose the Right Dockerfile Pattern
Read reference/dockerfiles.md for the complete, copy-ready Dockerfile for your stack:
| Stack | Base Image | Build Tool | Port |
|---|
| NestJS 11.x / Node 24 | node:24-alpine | pnpm | 3000 |
| Spring Boot WebFlux 3.5.x | maven:3.9-eclipse-temurin-21-alpine → eclipse-temurin:21-jre-alpine | Maven | 8080 |
| Python FastAPI 3.14 | python:3.14-slim | uv | 8000 |
| Python Agentic AI | python:3.14-slim | uv + gunicorn | 8000 |
| TypeScript/Fastify / Node 24 | node:24-alpine | npm | 3000 |
Step 2: Set Up Docker Compose
Read reference/compose-patterns.md for:
- Development compose (PostgreSQL 17 + Redis 7)
- Production compose with Docker secrets (
_FILE env pattern)
- Health check dependency ordering (
condition: service_healthy)
- Network isolation (internal backend network)
- Resource limits and restart policies
Step 3: Apply Advanced Patterns (if needed)
Read reference/advanced-patterns.md for:
- Build cache mounts (
--mount=type=cache) for Maven, npm, pip/uv — cuts rebuild time significantly
- Multi-architecture builds (
docker buildx) for linux/amd64 + linux/arm64
- Distroless runtime images for zero-CVE base
- Build-time secrets (BuildKit
--mount=type=secret)
- Diagnostics: slow builds, large images, networking failures, hot-reload issues
Step 4: Review with Checklist
Read assets/docker-review-checklist.md before committing any Dockerfile or compose file. Covers 6 categories: optimization, security, compose, image size, dev workflow, networking.
Reference Files
| File | Content | Load When |
|---|
reference/dockerfiles.md | Copy-ready multi-stage Dockerfiles + .dockerignore for all 5 stack variants | Writing any new Dockerfile |
reference/compose-patterns.md | Dev compose (PostgreSQL+Redis), production compose with Docker secrets, health checks, networks, resource limits | Writing or reviewing docker-compose files |
reference/advanced-patterns.md | Build cache mounts (all stacks), multi-arch buildx, distroless images, build secrets, diagnostics | Optimizing builds or debugging container issues |
assets/docker-review-checklist.md | 30-item Docker review checklist (6 categories) with stack-specific items | Before committing Dockerfile or compose changes |
Error Handling
Build fails with "not found" or permission error: Check .dockerignore is not excluding required files. Verify COPY paths match actual project structure.
Image too large (>500MB for Node/Python, >300MB for Java runtime): Read reference/advanced-patterns.md section "Image Size Optimization" — switch to multi-stage if single-stage, or distroless runtime.
Container exits immediately: Check HEALTHCHECK target path matches the actual health endpoint. Verify USER directive comes after all COPY operations. Check CMD vs ENTRYPOINT usage.
docker-compose: service fails to connect to database: Verify depends_on: condition: service_healthy is set. Verify DATABASE_URL credentials match docker-compose environment values exactly.
