memory: S5b merged on the operator's word; the r2_flow C3 amendment joins what booth-dev owes

persistent-memory.md

@@ -30,10 +30,14 @@ _As of 2026-09-27:_
   (a) the embed's clean-batch reload can take text typed into the HOST page;
   (b) carry() loses an edit set back to its original default mid-flight;
   (c) two r2b.toml anchors match twice ("D3 a stored theme…", "D3 forced
-  light…") and prove only by where the first match falls. **S5b** (the
-  in-place client: focus, floating status line, beforeunload) is design-dev's
-  next ref; it needs booth-dev's gate AND a seam pass against the batch/carry
-  code before it goes to Prime.
+  light…") and prove only by where the first match falls; (d) the r2_flow
+  contract's C3 steps 2-4 are stale against S5b (status line never hidden,
+  one-form failure no longer reloads over a draft) — amend to point at
+  as_antislop S5b. **S5b MERGED AND PUSHED** (`0233ca6`, operator "merge and
+  push", 2026-09-28; booth-dev gate 1068 / 428/428 + seam pass clean): focus
+  survives a swap, a floating status line, and leaving with an unsent answer
+  ASKS FIRST (beforeunload, both surfaces). S5c (keys, doc bar, and the
+  refused-batch words that a later "Saved." can bury) is design-dev's next.
 - ✅ **DOCS CAN NO LONGER RUN SCRIPT** (2026-09-28, Prime ruled "escape"): raw HTML in a `.md` renders as text, doc hrefs go through `is_safe_href`, the board's backslash twin of `//host` is closed. design-dev's anti-slop fix slices (`design-dev/antislop-sN`, Prime said GO in design-dev's session) arrive one ref at a time for booth-dev's gate.
   → `persistent-memory.d/2026-09-28-a-posted-doc-could-run-script.md`
 - ✅ **ONE SUBMIT SAVES EVERY ASK ON A PAGE** (2026-09-27, Prime's bug via