Interface EvidenceComputedModels

Which project-varying sub-models this particular run actually computed (OUT-04).

Without this, a persisted data.evidence.json is ambiguous in exactly the way a governance artifact must never be: an absent dependency could mean "nothing in this project depends on anything" (a real, reportable finding) or "this run never scanned for dependencies" (no finding at all, just an un-run pass). Those are opposite conclusions and the JSON looked identical either way.

env-cap solves the same ambiguity by always computing all six sub-models, so absence never happens. data-cap deliberately does not follow that: the usage scan is the expensive pass here, and --docs alone is supposed to stay cheap. So instead of forcing every run to pay for a full dependency scan, each sub-model's presence is stated explicitly.

capability, lifecycle, and runtimeContract have no flag -- they are always populated by construction, so a flag could only ever read true.

A pass that ran and found nothing is true with an empty result, never false: "we looked and found none" is a genuine finding, and collapsing it into "we didn't look" is precisely the confusion this field exists to prevent.