It took me a while to see what was actually happening. Initially, I was focused on a [1,5] suprafacial sigmatropic proton shift followed by a diaza-electrocyclic ring opening reaction. However, after drawing the corresponding orbitals I realised that I would have a eight-electron antiaromatic π system, in which a proton shift would need to occur antarafacially — not possible.
It still is clear that it is some kind of pericyclic reaction, and going through the list we can exclude everything except for a (retro-) group transfer reaction for which the (retro-) ene reaction is the most well-known example. (We are transforming one σ bond into one π bond which excludes all the others.)
It took me a while to actually see the retro-ene system in the substrate, but finally I did, and I took the liberty of highlighting it in red in the scheme below. Note that you have to ignore the carbon-nitrogen bond which belongs to the lowest five-membered ring from the ene-system. It is not taking part, if anything it is scaffolding.
Scheme 1: Retro-ene reaction of the title compound with the pericyclic six-membered ring system highlighted. Dashed bonds represent those that are broken/formed within the reaction.
Counting electrons, I realised that my initial assumption of a proton shift was wrong and that it is indeed a hydride shift.
 [1,n]-rearrangements are classified as suprafacial or antarafacial depending on whether the group is on the same side or the other side of a π system after migration. Technically, there is a further difference between retention and inversion, but hydrogen cannot rearrange under inversion so that is moot. The general rule for proton transfers is: If the number of bonds is $4n + 1$, the mechanism is suprafacial, for $4n + 3$ it is antarafacial.