What could be better, before anything breaks.
The app already tells you when something goes wrong. The checkup tells you what could be better while everything still works: a container that runs without a restart policy, a port open to your whole network, a volume nobody ever backed up. With the button that fixes it next to every finding.
Everything works. Now what?Read the comic: an inspector walks round before anything breaks →What you can do
Thirty-two rules, three weights
Worth a look, better changed, could be tidied. No score: a number invites polishing instead of thinking.
More power than it needs
A container running --privileged, the docker socket mounted, or a folder on the host it can write to.
Known vulnerabilities
A running image with critical vulnerabilities that already have a fix, from its last scan on the Images page. A finding you marked as not applicable or accepted no longer counts.
Reachable from your whole network
A published port listening on every address, where 127.0.0.1 would have been enough.
It has to survive a reboot
No restart policy, no health check, no memory limit — counted only for containers that have been up for a day.
A memory limit that fits
A limit close to the peak of the last days, one far above it, and limits that together are more than the machine has, each with a proposal from the measurements.
The daemon itself
From the CIS benchmark, what you can change here: logs without rotation, live-restore off, and containers without no-new-privileges, as a tip.
And on a swarm
One manager or an even number, a WSL node that is not drained, a port open to every node through ingress, a service without a restart policy or health check, a node drained for days, a secret nobody uses.
And in WSL
A credsStore that Docker Desktop left behind, containers that cannot resolve names, a daemon that does not start with the distro, bind mounts from /mnt/c, and programs on Windows that find no Docker at all. Only when the engine runs in WSL on this computer.
One click to the fix
Every finding links to the dialog that changes it. The app never changes anything by itself.
On a schedule
Every day or every week at a fixed time, over every machine you follow, one after the other. A notification only when something changed: what is new, what is solved, and how much stays open.
Explain and propose
Next to every finding, the AI explains what it means on this machine and proposes a change. Nothing changes until you apply it.
This is on purpose
Dismiss a finding and it stays dismissed, greyed out at the bottom of its group so you can undo it. A list you cannot empty is a list nobody reads.
Passwords stay covered
Environment variables that look like secrets are hidden until you click them, and so are the ones in the docker run command you copy. Copying still gives you the real thing.
Secrets where they should not be
A private key or an .env mounted into a container, or a password sitting in a compose file while an .env lies next to it.
On its own
A checkup that runs itself, and only tells you what changed.
Under Scheduled checkup… in the menu of the Checkup page, pick every day or every week and a time. The app then checks every machine you follow, one at a time and each with a time limit, and compares with the previous round: two new, one solved, twelve open. A list that says the same every week teaches you to ignore it; a difference you read. The last round stays on the page as a card, and Trivy can scan the images on the machine of the app first.
A report you can pass on
What runs on a machine, in one page of Markdown or HTML.
Containers with their ports and projects, compose projects, images, volumes, networks, and what the checkup found. Also of a machine you are not on: the app opens a short-lived connection to it and closes it again. Environment variables are left out, so the report is safe to share.
Since the last report
What came, what went, and what moved to another image.
The app remembers only the container names and their images from the previous report. The next one puts the difference underneath, so you can see at a glance what changed on that machine since you last looked.
Nothing leaves the app
The checkup reads what is already there.
The lists the app fetches anyway, plus one inspect per running container, on request and not continuously. No agent, nothing installed, no call to any service.
All features