diff --git a/.changeset/10150-author-shape-header-ruling.md b/.changeset/10150-author-shape-header-ruling.md
new file mode 100644
index 0000000000..599e585381
--- /dev/null
+++ b/.changeset/10150-author-shape-header-ruling.md
@@ -0,0 +1,6 @@
+---
+---
+
+Comment-only in `@object-ui/app-shell`: the docblock above `clientValidation`'s `AUTHOR_SHAPE_ONLY_TYPES` ended on a plan, to switch the `sharing_rule` edit gate on once the `_diagnostics` ingress was closed. That condition was met (`extractDraftBody` strips the read decorations through the spec's `stripReadDecorations`), the switch was put to the maintainer on objectui#7612, and the ruling was option A: the edit door stays with the server. The docblock now records that ruling beside the condition, and keeps the measurement above it. A source-scan pin requires the docblock to name the ruling's card and the function that closed the ingress.
+
+Declared as releasing nothing because the emit was measured rather than assumed: `AUTHOR_SHAPE_ONLY_TYPES` is not exported, so neither it nor its docblock reaches the emitted `clientValidation.d.ts`; the emitted `.js` carries the comment, and its comment-stripped emit is identical to the base. No published behaviour changes (objectui#10150).
diff --git a/packages/app-shell/src/views/metadata-admin/clientValidation.authorShapeRuling-10150.test.ts b/packages/app-shell/src/views/metadata-admin/clientValidation.authorShapeRuling-10150.test.ts
new file mode 100644
index 0000000000..b872da54a7
--- /dev/null
+++ b/packages/app-shell/src/views/metadata-admin/clientValidation.authorShapeRuling-10150.test.ts
@@ -0,0 +1,101 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * objectui#10150 — the `AUTHOR_SHAPE_ONLY_TYPES` header carries the ruling that
+ * keeps `sharing_rule` in the set, at the place its reader meets the condition.
+ *
+ * ## The trap this keeps shut
+ *
+ * The header in `clientValidation.ts` used to end on a plan: switch the
+ * `sharing_rule` edit gate on once the `_diagnostics` ingress is closed. The
+ * ingress was closed (objectui#7603, hoisted by objectui#8181), the switch was
+ * put to the maintainer on objectui#7612 with that already measured, and the
+ * ruling was option A: the edit door stays with the server. A header that still
+ * reads as the plan sends its next reader, who checks the condition and finds
+ * it met, straight back to re-filing objectui#7612.
+ *
+ * ## What this pins
+ *
+ * NAMED SUBJECTS inside ONE docblock, the one directly above the set: the
+ * ruling's card, and the function that closed the ingress. Inside that docblock
+ * and not merely somewhere in the file, because a citation moved to another
+ * comment is one the reader of the condition never meets.
+ *
+ * ## What it cannot pin, said here so it is not assumed
+ *
+ * - ⛔ Not the ruling's wording. The header quotes it verbatim; nothing parses
+ *   that quotation, so it is not pinned as text.
+ * - ⛔ Not prose intent. A header that cites objectui#7612 AND re-states the
+ *   old plan passes here. Catching that is review's job.
+ * - ⛔ Not whether the ingress is still closed. That is behaviour, and
+ *   `ResourceEditPage.readDecorationStrip.test.tsx` is its instrument.
+ *
+ * The first case is the lit CONTROL. The same reader finds the measurement the
+ * header must keep, and finds it only between the docblock's own bounds, so an
+ * empty or misplaced read cannot let the second case pass as a scan of nothing.
+ *
+ * If a later ruling takes `sharing_rule` out of the set, the header changes in
+ * that pull request, and this file changes with it.
+ */
+
+import { describe, it, expect } from 'vitest';
+import { readFileSync } from 'node:fs';
+import path from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const here = path.dirname(fileURLToPath(import.meta.url));
+const SOURCE = readFileSync(path.join(here, 'clientValidation.ts'), 'utf8');
+
+/** The declaration the header documents. */
+const ANCHOR = '\nconst AUTHOR_SHAPE_ONLY_TYPES = ';
+
+/**
+ * The docblock that ends directly above `ANCHOR`. A lookup, not a comment mask:
+ * nothing is removed from the source, the block is only located. Answers '' when
+ * the declaration is gone or is not directly preceded by a docblock, which every
+ * case below reads as a failure.
+ */
+function headerAboveSet(src: string): string {
+  const at = src.indexOf(ANCHOR);
+  if (at < 0) return '';
+  const before = src.slice(0, at);
+  const close = before.lastIndexOf('*/');
+  if (close < 0 || before.slice(close + 2).trim() !== '') return '';
+  const open = before.lastIndexOf('/**', close);
+  return open < 0 ? '' : before.slice(open, close + 2);
+}
+
+const HEADER = headerAboveSet(SOURCE);
+
+describe('objectui#10150 — the AUTHOR_SHAPE_ONLY_TYPES header records the objectui#7612 ruling', () => {
+  it('CONTROL — the reader returns that header, bounded, and it still carries the measurement', () => {
+    expect(HEADER, 'no docblock directly above `const AUTHOR_SHAPE_ONLY_TYPES`').not.toBe('');
+
+    // The measurement is why the type was listed, and the card that added the
+    // ruling kept it on purpose: the envelope family, the read decoration, and
+    // the inversion both would cause.
+    for (const subject of ['ADR-0010', '_diagnostics', 'objectstack#5316']) {
+      expect(HEADER, `the header no longer names ${subject}`).toContain(subject);
+    }
+
+    // Bounded on both sides: the code just above the docblock and the comment
+    // just below the set are in the file, and neither is in what was read.
+    for (const outside of ['function expandViewIssues', 'Map metadata-type name']) {
+      expect(SOURCE).toContain(outside);
+      expect(HEADER).not.toContain(outside);
+    }
+  });
+
+  it('cites the ruling and the closed ingress by name, in the header itself', () => {
+    for (const subject of ['objectui#7612', 'extractDraftBody']) {
+      expect(
+        HEADER,
+        `the header above AUTHOR_SHAPE_ONLY_TYPES must name ${subject}: the ` +
+          '`_diagnostics` ingress it describes was closed, the edit-door switch ' +
+          'was put to the maintainer on objectui#7612, and the ruling kept that ' +
+          'door with the server. Without it the header reads as a plan whose ' +
+          'condition is met (objectui#10150).',
+      ).toContain(subject);
+    }
+  });
+});
diff --git a/packages/app-shell/src/views/metadata-admin/clientValidation.ts b/packages/app-shell/src/views/metadata-admin/clientValidation.ts
index d7ba23ed78..4181dbb619 100644
--- a/packages/app-shell/src/views/metadata-admin/clientValidation.ts
+++ b/packages/app-shell/src/views/metadata-admin/clientValidation.ts
@@ -562,11 +562,33 @@ function expandViewIssues(
  * where a permissive match-all sharing condition gets written — and deliberately
  * not on `edit`. This is NOT a tolerant fallback: nothing is coerced and no
  * draft is waved through; one door has a client gate and the other keeps the
- * server's. Turning this gate on is gated on the `_diagnostics` ingress being
- * closed first (the strip belongs where the draft is assembled, not here —
+ * server's.
+ *
+ * ── The ingress condition was MET, and the answer was still no (objectui#7612) ──
+ *
+ * This block used to end on a plan: switch the edit gate on once the
+ * `_diagnostics` ingress is closed where the draft is assembled (not here —
  * reconstructing the decoration list in this file would be a second de-facto
- * contract). Until then a gate here would refuse legitimate author input, which
- * is worse than the defense-in-depth gap it closes.
+ * contract), and not before, since until then a gate here would refuse
+ * legitimate author input. That condition was met. objectui#7603 stripped the
+ * read decorations off the pending draft, and objectui#8181 hoisted the strip
+ * into `extractDraftBody`, which removes the spec's own list through the spec's
+ * `stripReadDecorations` before the pending draft is merged. So the
+ * `getDraft().item` assembly quoted above is the edit path as objectui#6982
+ * measured it, not as it stands; for the live answer read
+ * `ResourceEditPage.readDecorationStrip.test.tsx`, not this paragraph.
+ *
+ * The switch was then put to the maintainer on objectui#7612, with the met
+ * condition measured on that card and a finished implementation beside it
+ * (PR objectui#10054). The maintainer ruled option A on 2026-09-20 (restated on
+ * objectui#10150), verbatim 「7612 只需要服务端校验」 ("7612 needs server-side
+ * validation only"): the edit door stays ungated, because the server is
+ * authoritative on it and that is enough. The PR closed unmerged. The create
+ * door was not in question and keeps its client gate.
+ *
+ * ⇒ A closed ingress is NOT a reason to take `sharing_rule` out of this set:
+ * that question was asked with the condition already met, and answered. Only a
+ * new maintainer ruling re-opens it, never a re-reading of the ingress.
  */
 const AUTHOR_SHAPE_ONLY_TYPES = new Set<string>(['sharing_rule']);