How to determine which outputs to retrieve?

[carolynvs]

In the updated claims spec, we don't necessarily retrieve all the outputs anymore and store them in the claim. It's up to the tool to say which outputs it wants to persist. Additionally the claim result stores a map of the generated output names to an object that contains metadata about the outputs, currently we just have one piece of info we want: contentDigest.

I'm just thinking about the scenario where the tool does NOT want all the outputs persisted and if we should also avoid retrieving those outputs from the container too? If the tool doesn't want the output persisted, does that mean we don't pull it, generate the content digest and store that on the claim result?

Tagging in @jlegrone since this scenario is his use case.