This is a proposed funtion based on Pham 28735 (example member Mildred21 gp255)

The HHPred has a 100.0%/e=9E-39 93% length alignment to an NMR structure to e coli YibA in PDB and to COG3236
see screen capture of hhpred alignment here.

The phyre2 3d prediction is a 100% confidence with 91% coverage to the same pdb structure. This link should work for about a week. here is the screen capture which should work for the foreseeable future.

YibA is a founding member of COG3263, the function for COG3263 is based mostly on this paper:

A directed-overflow and damage-control N-glycosidase in riboflavin biosynthesis. Biochem. J. 2015 Feb 15; 466(1):137-145

This paper clearly shows that this e coli protein (and a couple of others like it) can enzymatically function to cleave N-glycosidic bonds in vitro as part of a large complex that carries out the first few steps of riboflavin biosynthesis. So there is really good evidence for enzymes of this type to be able to hydrolyse N-glycosidic bonds. I am certainly not proposing that this enzyme in phage is involved in riboflavin biosynthesis, but the wet bench is pretty good that the e coli YibA does have a glycosylase activity and there are many different glycosylases.

The issue as to naming is one of specificity, should we use the more specific N-glycosylase or the more generic glycosylase? There are also S, O and C- type glycosylases.

I feel that without biochemical evidence or a published analysis of the amino acids present at the active site and how they interact with a particular glycosidic bonds (i.e. how does the active site change with N, S, C and O type glycosidic bonds) that specifically adding the term N-glycoslyase may be too specific given the evidence above is simply bioinformatic alignments. Following this logic, I would say the more general glycosylase term is preferable. On the other hand, of all the glycosylases, one might argue that the most likely activity for a phage enzyme is as a DNA glycosylase. All of the DNA glycosylases are of course N-glycosylases (i.e. all 4 bases connect to the ribose through a nitrogen).

I think the evidence for this phage protein family to be an enzyme is about as good as you can get with just bioinformatic alignment so adding something is probably appropriate, the question is how specific do we want to go, the generalized glycosylase or the more specific N-glycosylase.