I don't think you can just give a tool a random genomic sequence and expect it to identify protein functional domains. The number of permutations would be huge in identifying TSS, intron/exon/UTR boundaries, reading frames, etc... Even for a simple genomic sequence.

To my knowledge, the minimum input requirement would be an mRNA transcript which can be used to infer protein sequence. Then there are plenty of tools if you have transcript coordinates or mRNA sequences. TRAPID is a useful tool designed for use with RNA experiments to annotate sequences with protein domains, GO, KEGG, etc...

If you have mRNA sequences already interproscan https://github.com/ebi-pf-team/interproscan is probably one of the classic tools for this.

https://interproscan-docs.readthedocs.io/en/latest/Introduction.html#what-is-interproscan

It is all rather well integrated and works well in my experience.

Others might be Maker/Augustus to produce mRNA or protein sequences (ie, gene prediction), then Blast2go or you can check out further decent functional annotation tools from this competition (past publications) https://biofunctionprediction.org/cafa/