Read a Dockerfile and get findings on layer-cache ordering, running as root, unpinned tags and the secrets that persist in a layer even after deletion.
Waiting for a Dockerfile
Nothing is built and no daemon is contacted — the file is read as text.
The findings appear here
Security, image size, build cache and correctness, each with the line it is on.
Nothing here knows what is inside your base image, so several notes are questions rather than corrections — whether the image already defines a non-root user, whether curl exists for a healthcheck. It also cannot see your .dockerignore, which decides what COPY . . actually copies and is frequently the real reason a build context is enormous.
Dockerfile Inspector runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
The findings are ordered by what they cost you. A secret baked into a layer is top of the list, because of the thing people get wrong about it: deleting the file in a later instruction does not remove it. Every layer is kept, and anyone with the image can read the layer where the file existed. The fix is a build secret or a multi-stage build, not an `rm`.
Cache ordering is the most common finding and the one with the clearest payoff. Copying the whole source tree before installing dependencies makes every build reinstall everything, and the inspector says which instruction to move where rather than just naming the problem.
`ADD` with a URL is called out because it does two things people do not expect: it fetches over the network at build time, and with a local archive it extracts automatically. `COPY` plus an explicit fetch is clearer about both.
Findings are classified rather than listed flat, because a missing `USER` and a missing `.dockerignore` are not the same size of problem. The classification says what the consequence is, not how severe a linter thinks it is.
Nothing is built and the Docker daemon is never contacted. The file is parsed in your browser, including the continuation lines and here-docs that make Dockerfiles awkward to read with a regex.
No, and this is the single most costly Dockerfile misunderstanding. Layers are immutable and all of them ship with the image, so the file is still readable in the layer where it existed. Use a build secret mount or a multi-stage build where the secret never reaches the final stage.
Because ADD does two surprising things: it fetches URLs over the network during the build, and it auto-extracts local archives. COPY only copies. For a remote file, COPY plus an explicit RUN that fetches it is clearer about what is happening and when.
Keeping things out of the build context. Without one, your .git directory, node_modules and any local .env are sent to the daemon and can end up in the image via a broad COPY. It also makes builds noticeably faster on a large repository.
No. This is a parser and a set of findings written for this tool, running entirely in your browser. It overlaps with hadolint on the common rules and is not a replacement for it in CI.
No. No Docker command is run and the local daemon is never contacted. The file is parsed as text.