Can you add an evidence layer to your existing screening provider?

• A coverage state per decision, which sources answered, which were stale, which were unavailable or excluded, instead of a bare alert.
• Candidate-set completeness and screening mode, so a capped or exact-only run cannot pass as a full screen.
• A retained, retrievable decision record with a content-integrity hash, keyed to your case or request ID.
• A recomputation check, validating the decision against the open reference so a false clear is caught rather than filed.

1. 1Confirm it augments, not replaces: your incumbent keeps producing matches; the layer captures and structures the decision around them.
2. 2Check what it can actually read from upstream, coverage, source list and versions, candidate completeness. It can only record what the incumbent exposes.
3. 3Confirm it fails closed: if upstream signals a source was unavailable, the layer records partial coverage rather than a clean record.
4. 4Check retrieval and integrity: can you fetch a record by ID for the full retention period, and prove it is unchanged?
5. 5Test the boundary: feed it a case the incumbent screened with a stale list, and confirm the layer surfaces the staleness rather than hiding it.

The clean pattern is a thin capture step in your screening workflow: when your incumbent returns a result, pass the result plus its source and coverage metadata to the evidence layer, which structures, validates and retains the decision record and returns a capsule reference you store on the case.

Nothing about the matching changes. What changes is that six months later you can produce the record, and show that a no-hit was, or was not, eligible to be treated as clear.