It would have been helpful, if you had provided a sample, but I presume you are thinking of something like this:

>chr1
AAAAAaatttaccCCCtagatgaCCCCCCCGCTACTGGGGGGGGGGGGGGgggtaacatcaaat

And now you would like to count exclusively the number of softmasked based at the 3' prime end?

Your idea using seqkit was actually pretty good...

seqkit locate -p "[acgtn]+$" -r -P example.fasta

will return

seqID   patternName pattern strand  start   end matched
chr1    [acgtn]+$   [acgtn]+$   +   51  64  gggtaacatcaaat

which you can pipe to awk to count the length:

seqkit locate -p "[acgtn]+$" -r -P example.fasta | awk 'NR>1{print $1,$6-$5}'