Stable software and evaluation release with an append-only evidence model. The final v1 snapshot records two genuine same-day prospective runs, 7,010 current-vintage retrospective replay checks, 9/9 synthetic fault detections, and 0/1 clean-control false alerts. Same-day runs do not establish longitudinal revision-detection evidence; replay and synthetic validation are not prospective evidence. The automation, first-run bootstrap, and public evidence boundary are verified. See docs/v1-release-audit.md.