Can this break my estate?
You are considering giving an automated process permission to stop and delete things. Five straight answers, then the full detail if you want it.
Who holds the credentials?
Nobody but you. It runs inside your own Azure account, using an access identity you create there. That identity never leaves Azure and we never receive it, so there is no password of yours for us to lose.
What can it touch?
Only the things you have labelled as non-production. Anything labelled production is off limits, and so is anything with no label at all, because if nobody has said what it is for, nobody has said it is safe to touch. Anyone can also mark a single item as hands-off and it will be left alone.
Can it reach into my network?
No. Nothing connects in to you. It reaches out to us, never the other way around, and if you block even that it keeps working and simply stops acting on its own.
What if it gets something wrong?
It writes down how to undo something before it does it, and takes a backup before deleting a disk. If the backup fails, the delete does not happen. When it is not sure, it does nothing and tells you why.
How do I stop it?
Two ways, and neither needs our permission. Shut it down, or take away its access in Azure. Taking away access is the softer option: it carries on watching and reporting, and anything it tries to change simply fails and gets written down. It never stops halfway through something.
The part we will not dress up
Two of the permissions it asks for are broader than we would like, because Azure does not offer narrower versions of them. One is needed to resize a machine, the other to back up a disk before deleting it. We are not going to pretend they are harmless. What limits them is that the agent never uses them for anything else, and that you decide which parts of your estate it can reach, which Azure enforces rather than us. Your security team sees the exact list before anything is installed.
See what it would change