chainscore.net
Desk
For an auditor
An auditor asks whether a competent reader could take silence, or a printed list name, as a pass this origin did not run. Cite the sealed copy and the overlay that actually ran. A rerun is a new measurement.
What you can sample
Ask the house for the sealed PDF, the engine, the report id and the overlay publish date. Do not re-run the screen and treat the new figure as the copy they filed. A rerun is a new measurement.
generated_at is now. Prices, labels and the sanctions snapshot all move. The SHA-256 seal is integrity of this copy, not a 17a-4 WORM store. An audit sample cannot re-strike the same dollars without pinning those inputs. Keep the PDF you relied on.
Ask which overlay ran. ScoreGuard prints a CIP-style log line: this screen against OFAC SDN at a published date, not a finding that the customer was identified. A no-match page that names only the lists that loaded is usable. A page that names lists that did not load is not.
How to read coverage
Value coverage and count coverage both print. The gate refuses to auto-clear on a thin screen. If the Assessment disagrees with the gate, the artefact disagrees with itself — quote both.
ScoreGuard runs 1–100, lower is riskier. ChainTrace runs 0–100, higher is riskier. Holistix runs 0.0–10.0, higher is riskier. Sampling across engines is three measurements, not one.
What you will still test at the house
CIP of the customer, 314(a), a signed-message proof that the fund controls the key, and the house’s retention of the PDF are not this origin. Named handoffs and not-established blocks are the honest absences. A match is never invented. chainscore.net does not e-file.
Designated addresses on chains this origin does not screen are a miss unless the report says so. Count unscreened chains as unnamed, not as cleared.
