Go to the NCBI nuccore database, at https://www.ncbi.nlm.nih.gov/nuccore
Copy and paste the search term:
(bacterial virus[Organism]) AND (3000:110000[sequence length]) NOT bacteria[Organism]
You'll get 74583 results.
Click on the link send to and click the check box file
It will then download.
Check the file is in fasta format with ">" to ...
I think this could be close to what you're looking for: http://glad4u.zhang-lab.org/index.php
Paper: GLAD4U: deriving and prioritizing gene lists from PubMed literature
It works with gene names though, so I am not sure it's a 100% fit.
At a guess you need to install the SDK tools from GCloud. You would then use gsutil to access the file, within the 'bucket' (Google's terminology). Its quite easy.
gsutil allows searching the bucket and pull down relevant files.
There is no other way to access the file outside gsutil such as sftp, even within Gcloud.
The question is ...
Is there any particular reason why you want to download the original read files specifically from that platform?
If the sra toolkit is not good enough for you and you want a direct download link of name-standardised files, you can use the links on ENA: