Say for example someone setting one up on someone else's phone (without their knowledge) to run every X minutes and using it to track someone's location, without them ever knowing about it.
I'm not entirely sure that's the real reason behind these notifications, but it does seem plausible.
"The shortcut 'Foo' wants to run. Do you want to allow it?
- Never
- Not this time
- Run once
- Always"
Put an icon next to each shortcut showing whether it's always (or never) allowed, and it seems like the problem would be largely solved without asking much from the end user.
What you want to do is periodically (at a random time each time) ask the user if they still want it to do the privacy sensitive action still. This way, it vastly increases the chances of detection if the action was added by a third party.
Location accuracy isn't always the greatest, particularly indoors; in-and-out connections to wifi/bluetooth are possible when you're on the fringe of its range.
These issues need to be addressed in some form before _true_ automation is given to them.
It would be nice if these automations could run automatically but perhaps, notify the user that said automation has run and allow them to easily disable the automation.
They could also do what Gmail and possibly others do when forwarding is added. An obnoxious red colored warning at top of Gmail for 7 days about new forwarding. iOS could do that in some fashion.