I was just working on something similar, so I gave it a try.

Here is some feedback:

It does not adhere to the eutils user guideline It is very likely that it will make more than 3 requests per second.

It does not work for all entries (e.g. this one).

It should give some indication if it can not find a DNA sequence for a protein and not fail silently like it is now.

Here's a script that utilizes Entrez Direct and protein GIs:

#!/bin/bash
IFS=$'\n'

if [ -n "$1" ]
then
  split -l 500 $1 input.

  for f in input.*
  do
    query=$(cat $f | tr "\n" ",")
    efetch -db protein -id $query -format fasta > $f.faa.tmp
    rm $f
  done

  cat *.faa.tmp | grep . > $1.faa
  rm *.faa.tmp

else
  echo "Usage: fetchSeqs listOfNcbiGis"
  echo "Requires Entrez Direct in \$PATH: http://www.ncbi.nlm.nih.gov/books/NBK179288/"
fi

edit. Actually, it doesn't do what OP's script does. Just proteins with protein GIs. If you want nucleotides with protein GIs, replace the efetch line with:

epost -db protein -id $query | elink -target nuccore | efetch -format fasta_cds_na > $f.faa.tmp

edit2. actually, I think this fetches all CDS of the related nt molecule. I linked gi to cds_nt before but can't remember right now how to do it :(

Here is python script that can also be used to download GenBank files given accession numbers.

Copy the python code below and save it as fetch_accession.py. Create a text file and give it a name say accessions.txt. Write these accession numbers (AP010935.1, AE014074.1, CP001129.1) to the file and save.

Run it as follows:

./fetch_accession.py -a accessions.txt

The output should files with the accession id followed by .gb and .fasta extensions.

#!/usr/bin/env python
import os, sys, glob
from Bio import SeqIO, Entrez
import argparse

def main():
    opts=argparse.ArgumentParser(sys.argv[0],description="fetch genbank files given accessions",
        prefix_chars="-",add_help=True,epilog="2014")
    opts.add_argument("-a",action="store",nargs=1,metavar="ACCESSION FILE",dest="ACCESSION",help="specify accessions file",required=True)
    options=opts.parse_args()
    input_file=options.ACCESSION[0]
    indata=""
    ncounter=0
    try:
        indata=[str(j).strip() for j in open(str(input_file),"rU")]
    except IOError as inError:
        print "Error: Failed to open input file "+str(input_file)+"..."
        print "\n\nException caught: \n",inError
        sys.exit()

    try:
        for I in list(indata):
            ncounter=ncounter+1
            print "Parsing...\t"+str(i)+"\t"+str(ncounter)+"/"+str(len(indata))
            seqfile=Entrez.efetch(email="example@email.com",db="nucleotide",id=str(i),retmode="text",rettype="gb")
            gb_record=SeqIO.read(seqfile,"genbank")
            SeqIO.write(gb_record,str(i)+".gb","genbank")
            SeqIO.write(gb_record,str(i)+".fasta","fasta")
    except Exception as exError:
        print "Error: Problem encountered while downloading files..."
        sys.exit()
if __name__ == "__main__":
    main()

Thanks guys for your insightful comments and contributions to this issue!