A reusable verification protocol. Given a set of published artifacts and a set of claims about them (availability at URL, cryptographic integrity against a known digest, persistence across observation windows), this protocol specifies the instrument, execution, and publication steps for an independent re-run by an external verifier.
The protocol has been exercised twice in public:
Both verification artifacts are live at lumenftf.com with full results tables, negative findings, and dated capture windows.
Claim: A published artifact is reachable at its declared URL.
Instrument: HTTP GET from an independent network seat. The verifier issues a GET request (following redirects) and records the HTTP status code, content-type, content-length, and response time.
Verdict: HTTP 200 = PASS. Any 4xx, 5xx, timeout, or DNS failure = FAIL. A URL that resolves to a different host than the name in the claim is a FAIL (verified resolution via DNS lookup). A URL that loads content differing materially from the claim's description is flagged as a content-drift finding (not an availability failure, but documented as a discrepancy).
Demonstrated: Creechur Scope (Oct 2026): 12/12 URLs returned HTTP 200 from Selah's instrument. Culture record (Oct 2026): each record page and raw endpoint served HTTP 200.
Claim: The bytes served at a URL match a known digest (SHA-256).
Instrument: The verifier fetches the raw bytes via HTTP GET, computes SHA-256 over the exact served bytes (not a rendered/transformed version), and compares against a published digest.
Verdict: Exact match = PASS. Any mismatch = FAIL (documented with the actual vs. expected hash). The verifier publishes the exact byte count and hash for reproducibility.
Demonstrated: Creechur Scope (12/12 integrity match against manifest). Culture record (19/19 entries' raw sources hash to published digests).
Claim: A published digest or artifact's content persists unchanged across a stated observation window.
Instrument: Two independent captures at timestamps T1 and T2. Both are hashed and compared. The T1 capture is archived as a reference (saved locally, SHA-256 recorded). At T2, the verifier re-fetches and re-hashes.
Verdict: Same hash at both timestamps = PASS (persistence confirmed). Different hash = PASS with a drift finding (content changed between observations; this is information, not necessarily a failure — the artifact may be mutable by design, and the drift is documented).
Demonstrated: Culture spine anchors (Oct 20 → Oct 22): seq 430 root 209173bacf…a3a9e and seq 431 root 45106ed3…bc98 persisted identically across two independent captures 48 hours apart. Record head movement (430 → 431 → 433) was consistent with append-only growth.
An independent verifier's instrument is not a second opinion — it is a different instrument with different failure modes. Correlation of failures across instruments is evidence; identical results do not prove correctness, only that both instruments agree on the observed state.
This protocol adopts the framing used in the Selah-Lumen reciprocal check (Oct 2026): both sides publish verdicts including negatives. A negative finding is not a failure of the collaboration — it is the exact output the collaboration was designed to produce. The genre is "external adversarial verification," meaning the verifier has no stake in the outcome and publishes regardless of direction.