This may be a duplicate of How to create kallisto index that does not include ribosomal genes, or the thread may be helpful anyways. Main question is: why do you want to filter rRNA at this stage - it is likely better to do this at a later stage in your experiment if required.

When creating a kallisto index using the recommended approach (which is kb-python), when running kb ref you can use --exclude-attribute to exclude certain features from the reference (e.g. --exclude-attribute gene_biotype:lincRNA will exclude lincRNAs).

You should filter out things that are causing multimapping (like read-through transcripts) and/or things that should never be found in your RNA-seq reads (since kallisto's index grows a lot in memory with the more things you add in it).

If there is a lot of rRNA contamination, it's probably not the best idea to filter them out at this stage as explained in the other thread.

Multiple people in this and the previous thread suggested not to remove rRNA, so it is on you to either follow that or not. For technical purposes, since you mention R, reading a fasta is best done with readDNAStringSet from Biostrings which creates an easy-to-manipulate object. Easiest is probably to get the names of transcripts you aim to remove from a GTF as this explicitely contains the biotype, and then remove that from the DNAStringSet, see also https://www.rdocumentation.org/packages/Biostrings/versions/2.40.2/topics/XStringSet-io