All of the RIS data is imported into the ReferenceRIS attribute; you just need to add code that copies the relevant data from ReferenceRIS into your chosen user attribute (or maybe $Text in the case of an abstract).
There was a related thread here, in the context of Bookends that died because of an apparent bug in Bookends or maybe Tinderbox.
In the case of Zotero this actually works. Here is adapted code for Lew Friedland that reads the ReferenceRIS , extracts the relevant information from the RIS tag AB and puts it into the user-defined attribute Abstract.
(1) You need to create a new User attribute to hold the data. Call it Abstract, of type string.
(2) I'll assume you've already got a Reference prototype in your Prototype container -- one is created automatically when you first drag in a Reference. Modify the rule for the Reference to:
$ReferenceRIS.contains("AB - ([^\n]*)\n");
$Abstract=$1.replace("AB - ","").replace("\n","")}
(3) Add the user-defined Abstract as a KA to the Reference prototype.
Now, assuming you have a Zotero example with an abstract, dragging it into Tinderbox gives you a reference with the Abstract KA populated with the text from the Abstract field in Zotero -- and if there is no AB field then the abstract is empty.
This works with BibDesk too, but, again, not with Bookends.
However, in the case of Zotero there is a problem with journal articles having Journal Name mapped into the BookTitle KA. The problem here is that Zotero is using the RIS tag T2 to hold the Journal Name, while Tinderbox is expecting it be contained in RIS tag JA. RIS contains ambiguities. Again, you can add code that maps RIS tags into the appropriate TBx KAs.
Let me know if this works for you..