When a decision record begins from a public record of the past, it has to say how much of that record may pass into the present decision. This paper asks what one set of records says about that passage. It reads fifteen authored cases from AEGIS, a studio of the RISCON family: every case of AEGIS's new-domain test programme whose situation names a public or declassified record, among them a war history, an old manual, two classics, a manufacturer's release, declassified files, a memoir, a wargame and the minutes of a meeting. All fifteen are synthetic training cases written in Korean, not real events, and AEGIS says that the programme refers to no public record at all; the English quoted is this study's own translation.
Three findings follow. Every case separates the record from what is drawn from it, and every recommended option carries the record over only on stated conditions. The option the author rules out is more often the neglect of the record than its literal copy, which survives as the alternative and is softened by its own reasons. And the original that fourteen of the cases ask for first is nowhere in the record set. Read beside the debate on lesson-drawing, analogies and the evidence that what worked there will work here, the set takes a middle position that it can state but cannot yet perform.