My understanding of the frame field in GFF is that it is supposed to report the reading frame of exons relative to a gene, which only changes in the case of splice sites interrupting codons in genes with introns.
So this field is always '0' for microbial genes. The one exception might be if Prodigal reported partial genes as (1..100), but it does not; it reports them as (<2..100), so you can always deduce the frame from
the coordinates.
regards,
doug