Reference provenance from earlier trusted results in follow-up answers

(not necessarily intending that this be tackled right now)

Not attaching a provenance marker when no data tool was used in the current exchange makes sense, since tool usage is the basis for assigning provenance.

However, the absence of a marker kind of makes answers appear less trustworthy than even the "Untrusted" tagged answers. I think people might read it as, “we have no idea where this answer came from.”

This feels especially misleading when the agent is directly referring to the output of an earlier trusted calculation. In this screenshot, the follow-up immediately follows the measure result, so the relationship is pretty clear. If the follow-up happens much later in the conversation, though, it would be useful for the answer to reference that earlier measure result and its provenance.

I'm not thinking that follow-ups would get a trusted tag, but that they would somehow reference the turn that included the data tool with that tag.

Start by reviewing how provenance markers are currently assigned to data-tool results and how later follow-up answers are represented. Clarify how a follow-up should reference an earlier trusted result without receiving its own trusted tag, including behavior when the reference is much later in the conversation. Done means the behavior and scope are specified well enough to implement and test.