First, turn your idea into something expressed as prometheus metrics. What condition do you want to test on? Is it when a particular metric exists with a particular value? (Like a metric which is 1 when mounted and 0 when not mounted)? Or when a particular metric with a particular set of labels
does not exist at all?
In the latter case, the absent() function may be what you're looking for. e.g.
expr: absent(node_filesystem_avail_bytes{instance="nuc2", mountpoint="/boot/efi"})
Note that any "expr" you use for alerting can be tested simply by pasting it into the PromQL browser in the Prometheus web interface. If it returns an empty result set, then the alert is not firing. If it returns a non-empty result (i.e. 1 or more timeseries), then this is the set of alerts which will fire, regardless of their value.