The institution should be able to prove that it preserved its own record.

This engine evaluates the structured public corpus against known source inventories and record-level integrity rules. It is designed to expose missing source records, duplicate identities, weak provenance fields, and inconsistent publication states before they silently become institutional memory.

Known source inventories versus the public corpus.

A source family passes only when every structured source record in that inventory is represented in the merged institutional ledger.

No unresolved integrity findings.

These findings are generated from the structured corpus itself. They do not claim that an external URL is reachable in real time; network link verification should be added as a separate scheduled evidence check.

Integrity is a continuing condition, not a one-time cleanup.

This first engine audits structured identity, source coverage, required provenance fields, and state consistency. The next layer should perform scheduled external-link verification, DOI resolution checks, repository reachability, patent-source reconciliation, and historical snapshots so corpus health itself becomes replayable evidence.

No admissible evidence. No admissible execution.