Hi @pau,
Check out the actions in q2-quality-control
See also tutorials
And a real live example using a mock community at the end of this tutorial
Note that these methods were designed for use with mock communities or other datasets where the composition is known, since the outputs are accuracy metrics. However, it would be possible to use these methods to compare taxonomy classifications of samples with unknown compositions and the outputs essentially become measures of how similar those taxonomy classifications are.
I have reclassified this as "user support"
Good luck!