In that case, you should use the the --p-adapter to remove the reverse compliment sequences from the 3' end and any sequence after it, whereas the --p-front flag is intended to start searching from the 5' end and remove any extra sequence before it. The latter case is why you had too many short reads as it is finding the reverse compliment towards the 3' end and then removing it and any sequence before it.
If you read the help text you'll see that you should set up your command like :
qiime cutadapt trim-single \
--i-demultiplexed-sequences single-end-demux.qza
--p-front CCTGTCAGATCAAAAACGAC CAGGAACTACATCACAATAAC \
--p-adapter GTCGTTTTTGATCTGACAGG GTTATTGTGATGTAGTTCCTG \
--p-overlap 3 \
--p-error-rate 0.1 \
--p-match-read-wildcards \
--p-match-adapter-wildcards \
--p-discard-untrimmed \
--o-trimmed-sequences primer-trimmed-demux.qza \
--verbose > cutadapt-log-3.txt