Docker Doctor

Diagnostics your infrastructure with Docker and Docker Compose
$
npx @docker-doctor/cli@latest
View on GitHub
Terminal - 592x648
docker-doctor
$
Tick:0

every pull request

Add PunGrumpy/docker-doctor@v1 to your workflow and every pull request gets one sticky summary comment — advisory by default, gate when you trust it.

docker-doctorBotreviewed 2 days ago
Dockerfile
FROM node:latest
ENV DB_PASSWORD=hunter2
Docker Doctorno-secrets-in-env(error)

Fix -> use Docker Secrets or runtime environment variables instead of baking credentials into the image.

See docs

every sandbox

Docker Sandboxes run coding agents unattended in isolated microVMs. The Docker Doctor kit rides along — preinstalled in every sandbox, so the agent lints its own Dockerfile and Compose changes before committing them.

microVM

$sbx run --kit docker.io/pungrumpy/docker-doctor-kit:latest claude
microVM started — private kernel and Docker daemon
@docker-doctor/cli installed (pinned)
agent skill → ~/.claude/skills/docker-doctor
agent memory: lint Docker changes before committing
Committed clean:Dockerfile scored 100/100 before it left the sandbox

Agent

Containerize this app
Write Dockerfile
  • Dockerfile
Run docker-doctor
  • 3 issues found
  • pin-image-version → fixed
  • no-root-user → fixed
Commit
  • 0 issues · score 100/100
$
sbx run --kit docker.io/pungrumpy/docker-doctor-kit:latest claude

Read the Docker Sandboxes guide

every scan

The engine discovers the Dockerfiles and Compose files in your project, runs every rule over them, and collapses the result into one health score — warnings inform you, errors fail the build.

my-app
Error-level issue?
no
No errorsscore 94 · exit 0
yes
2 errors foundscore 71 · exit 1
25 rules across five categories. Errors fail the build; --score fails below 50. How scoring works