"It was our secret."
Translation:To była nasza tajemnica.
"Tajemnica" can indeed mean a mystery, but also a secret. Both "tajemnica" and "sekret" are considered 'best answers' (so that's why I see 'tajemnica' as the translation, the algorithm just chooses one randomly), but if you get the PL->ENG exercise, Duo will ask you to translate "sekret".
I gave it a bit of thought and not really, IMHO – if you have it entered as [nasza tajemnica/nasz sekret], the system matches „nasza”, then checks „sekret” against „tajemnica”, returns false and then shows „nasza tajemnica” as the closest match, because „nasza” matched – „nasz sekret” is part of different short notation option, after all, so it looks like further apart from its perspective – Polish doesn't have all that fancy lexeme tagging, so the system doesn't even know that „nasz” and „nasza” are different genders of the same lexeme, it just thinks these two are completely separate words…
This is pretty much, why it is not so good that not-in-house languages do not have lexeme tagging – I'm pretty sure that if we had it for Polish, it would correct the gender rather than show a wholly different noun.
No idea, no plans(as far as I know) and yes, precisely why there isn't a words tab for languages that aren't in-house – it also affects other things too, like lack of conjugation and gender information in hints and so on.
Yup, it is a pretty big deal – Incubator(and therefore the courses too) for not-in-house languages works like a dump take-XXX-match-it-to-YYY finite state machine and knows nothing about grammar, which makes it pretty hard to teach it, as these courses can't even measure whether you understand these concepts or even differentiate words that are spelled the same but are different(like say genitive singular vs. accusative plural)…