I have not run this script but you may try if it gives the desired result: https://github.com/jorvis/biocode/blob/master/sandbox/jorvis/convert_exonerate_gff_to_gff3.py
2nd way:
The GFF2 file contains information about features' positions in the genome, and you can extract the sequences from the reference genome based on this information. To create a GFF3 file, you typically need information about the features (e.g., genes, exons, CDS regions) and their positions on a reference genome. If you have this annotation information, you can create a GFF3 file. To create a GFF3 file from your FASTA data ensure you have two key files: A FASTA file with your sequences and A BED file containing feature coordinates and names. If you don't have a BED file, you may need to generate one, as it's a common intermediate format for feature annotations. then utilize BEDTools to convert your BED file to GFF3 format.