What relationships can a generated evidence release contain?

MirrorFirm could generate an evidence graph alongside the records an agent reads. The graph can connect a review request to its assertions, each assertion to applicable controls, and each control to the cases and evidence items that support assessment.

The relationships should express meaning. `supports`, `contradicts`, `supersedes` and `applies_to` serve different purposes. A policy may apply to a case without proving that anyone followed it. An approval may support an action while a later exception note contradicts the assumption that its conditions were met.

Many relationships are not one to one. A batch reconciliation can cover several transactions, while one transaction may require both a reconciliation and an authorised adjustment. Shared evidence creates useful dependencies: rejecting one batch record can affect several assertions without making every case deficient.

A generated release can include readable source material, explicit relationship tables and separate assessment labels for evaluation. Buyers can ask how the relationships were constructed and how they reflect business operating patterns. The commercial output remains generated material, rather than underlying business records. Its value depends on whether the graph captures meaningful support and contradiction, not merely whether every record has links.