Maybe this would do the trick!

esearch -db genome -query "txid543 [Organism]"|elink -target nuccore|efilter -query "RefSeq"|efetch -format fasta

From speaking with a few other pros, this was the solution in the end (though only very rough at the mo):

Use the ete3 toolkit to get a list of IDs:

from ete3 import NCBITaxa
import sys

taxon_name = sys.argv[1]

ncbi = NCBITaxa()
ncbi.update_taxonomy_database()
ebact = ncbi.get_descendant_taxa(taxon_name)

with open('./taxids', 'w') as ofh:
    for i in ebact:
        ofh.write("%s\n" % i)

# At this point, one could import ncbi-genome-download as a python method and continue

Which gave me a list of IDs (though this includes ALL descendent taxa, even ones without complete genomes etc).

I passed these to the latest version of ncbi-genome-download which accepts a --taxid 12345,65890 format for specifiying the IDs.

So I just ran:

for file in * ; 
do python ~/bin/ncbi-genome-download/ncbi-genome-download-runner.py -l complete -v -p 10  --taxid $(paste -s -d ',' "$file") bacteria ; 
done

I had to run this iteratively on many files after I split my taxids file up as there is a limit to how many args can be passed to --taxid at once.

EDIT Sept 2018:

I contributed a script to the ncbi-genome-download repo to make getting the TaxIDs nice and easy. It uses the approach above, but there’s no need to rewrite it for oneself now.