You need to know the gene accession/bank id of your protein. Example NIF [Xenopus laevis] GenBank: AAP43959.1. Then its easy to get sequence. Other wise download the file in GenPept format and you will see a section (/product="NIF) in your file, just search (/product="NIF) in your text file. Next to /product="NIF section will be your sequence. Let me know did you get the results or not.
enter in the query box "sp=taxon_you_want_sequences of", can be "bacteria", a given genus or species ==> build list2
enter "list1 and list2" in querybox =>list3
then possibly filter on length (recommended to filter out short sequences)
select list3, then clic "export you sequences" (select fasta format). Be careful the graphic interface is often ugly and sometimes strange. You may need trials before you master it...
clic on replace by parents => list4
export list4 as genbank or embl format
fasta files are sequences extracted exactly at the boundaries indicated in the entries at the corresponding feature.
This can also be done directly at pbil using the online interface.
The above explanation retrieves gene sequence present in EMBL or GenBank but not from complete genomes.
To retrieve gene sequences from complete genomes, select one of the complete genome database (see explanations on pbil for which to select)
One more note:
If you do several export and choose the same filename, by default sequences are added (appended) to this file.
No pb for fasta, but if database entries have different format (for example picking genbank and then ensembl) then you end up mixing two format in a single file, which is not recommended :-). Be carefull complete genomes entries may be quite large (well annotated eukaryotic genomes).