What problem does it solve? Hand-written Dockerfiles often ship bloated images, leak secrets into layers, rebuild slowly, and fail on ARM machines. This Skill encodes opinionated rules for writing Dockerfiles that are minimal, reproducible, fast to build, and portable across CPU architectures. ## Core Features & Use Cases - Multi-Stage Build Patterns: Enforces named stages, --target builds for test/lint/debug, and dropping the build toolchain from the final image (e.g., Go apps from 800 MB down to 5-20 MB with distroless). - Build-Time and Size Optimization: Covers layer ordering, BuildKit cache mounts, bind mounts, secret mounts, SSH mounts, and CI cache backends (GitHub Actions, registry, local, S3). - Multi-Architecture Builds: Guides buildx setup, QEMU, manifest lists, TARGETPLATFORM/TARGETARCH ARGs, and the FROM --platform=$BUILDPLATFORM cross-compile pattern, plus GitHub Actions and GitLab CI pipeline templates. - Use Case: A developer asks to containerize a Go service for both Apple Silicon laptops and AWS Graviton. The Skill produces a pinned, multi-stage Dockerfile with cache mounts, a non-root distroless final stage, a .dockerignore, and a CI workflow that pushes a multi-arch manifest list. ## Quick Start Ask the AI to write a production Dockerfile for your application, for example: "Write an optimized multi-stage Dockerfile for my Node.js app that builds for both amd64 and arm64."