diff --git a/.changeset/18977-orderby-dual-declaration-cross-reference.md b/.changeset/18977-orderby-dual-declaration-cross-reference.md
new file mode 100644
index 00000000000..128263f0386
--- /dev/null
+++ b/.changeset/18977-orderby-dual-declaration-cross-reference.md
@@ -0,0 +1,23 @@
+---
+"@objectstack/spec": patch
+---
+
+`$orderby` is declared twice — `ODataQuerySchema.$orderby` and `QueryTransportParamsSchema.$orderby` now cross-reference each other, and a pin holds the two accept sets apart (#18977).
+
+Clause-②: no. No accept set moves and no export is added, removed or renamed: the change is two docblocks in published source (`src/api/odata.zod.ts`, `src/data/data-engine.zod.ts`) plus a new pin test. Measured — `check:generated` reports all 16 generated artifacts up to date, `check:api-surface` and `check:authorable-surface` included.
+
+The two declarations are **complementary refusals**: each accepts exactly what the other rejects, and neither pointed at the other, so reading one of them carefully and completely still produced the wrong answer about the other.
+
+| `$orderby` value | `ODataQuerySchema` | `QueryTransportParamsSchema` (`DataEngineSortSchema`) |
+|:---|:---|:---|
+| `'name desc'` / `'-created_at'` | accepted | REFUSED |
+| `['name desc', 'email asc']` | accepted | REFUSED |
+| `[{field, order}]` | REFUSED | accepted |
+| `{name: 'asc'}` / `{name: 1}` | REFUSED | accepted |
+
+- **Which one grades a query bag**: `QueryTransportParamsSchema`, reached from `FindDataRequestSchema.query` through `QueryWithTransportSchema` — the schema `POST /data/:object/query` parses its body against. `ODataQuerySchema` grades no runtime door: measured on this tree, its only consumers are the `OData.buildUrl` helper in its own file and its own unit test.
+- **The refusal on the transport side is deliberate and stays** — `#18704` settled it: lowering an OData sort *expression* means PARSING, and a second parser beside the door's is how one rule gets two implementations that disagree. Widening either side to close the gap is a decision, not a tidy-up, so this change closes the **reader's** half only.
+- **The string forms are not unserved.** `normalizeSortNodes` (`@objectstack/metadata-protocol`) reads `'name desc'`, `'-created_at'` and the `string[]` form at the shared ingress behind `GET /data/:object`, the export route and in-process `findData`. A querystring spelled the OData way works; the same bag sent as a `POST /data/:object/query` body answers `400 VALIDATION_FAILED`. The difference is the door, and neither door is `ODataQuerySchema`.
+- **The cost this repairs was already paid.** objectui#9554 was filed, triaged, graded and dispatched against a shipped `object-grid` producer that had been sending the canonical shape all along, because the filing seat read the OData declaration and quoted it correctly.
+
+`src/api/odata-orderby-dual-declaration.test.ts` is the mechanical half: 25 cases pinning each side's accept set, their disjointness (with the lit control that neither set is empty), and which of the two `FindDataRequestSchema.query` is graded by. Widening or narrowing either declaration turns it red and lands the author on the cross-reference.
diff --git a/packages/spec/src/api/odata-orderby-dual-declaration.test.ts b/packages/spec/src/api/odata-orderby-dual-declaration.test.ts
new file mode 100644
index 00000000000..b9ca5b27579
--- /dev/null
+++ b/packages/spec/src/api/odata-orderby-dual-declaration.test.ts
@@ -0,0 +1,149 @@
+// Copyright (c) 2026 ObjectStack. Licensed under the Apache-2.0 license.
+
+/**
+ * [#18977] `$orderby` is declared TWICE, and the two declarations are
+ * COMPLEMENTARY REFUSALS — each accepts exactly what the other rejects.
+ *
+ * | declaration | accepts | refuses |
+ * |:---|:---|:---|
+ * | `ODataQuerySchema.$orderby` (`api/odata.zod.ts`) | `string`, `string[]` | the record maps, `SortNode[]` |
+ * | `QueryTransportParamsSchema.$orderby` = `DataEngineSortSchema` (`data/data-engine.zod.ts`) | the record maps, `SortNode[]` | `string`, `string[]` — deliberately, #18704 |
+ *
+ * Neither file pointed at the other, so reading one of them carefully and
+ * completely still produced the wrong answer about the other — measured, at a
+ * price: objectui#9554 was filed, triaged, graded and dispatched against a
+ * shipped `object-grid` producer that had been sending the canonical shape all
+ * along, because the filing seat read the OData declaration and correctly
+ * quoted it. The cross-references landed in both files with this card; this
+ * file is their MECHANICAL half.
+ *
+ * ## What this pins, and what it deliberately does not
+ *
+ * It pins the two accept sets AS THEY ARE, and their disjointness. It is not
+ * an argument that either set is right:
+ *
+ *  - §1/§2 hold each side's accept set, so widening or narrowing either one
+ *    turns this red and lands the author on the cross-reference that explains
+ *    why the gap is a decision rather than a defect. Widening the TRANSPORT
+ *    side is the one the source argues against in its own words — lowering a
+ *    sort expression means PARSING, and a second parser beside the door's is
+ *    how one rule gets two implementations that disagree.
+ *  - §3 is the disjointness itself: no value parses under both. That is the
+ *    property a reader cannot get from either file alone, and the one that
+ *    makes "I read the declaration" insufficient.
+ *  - §4 says which of the two grades a query bag, through the slot the REST
+ *    door actually parses (`FindDataRequestSchema.query`).
+ *
+ * ⛔ It pins nothing about what the RUNTIME serves. The OData string forms are
+ * not unserved — `normalizeSortNodes` (`@objectstack/metadata-protocol`) reads
+ * them at the GET querystring ingress and for in-process `findData`. What this
+ * file measures is the two SCHEMAS, which is where the card's trap lives.
+ */
+
+import { describe, it, expect } from 'vitest';
+import { ODataQuerySchema } from './odata.zod';
+import { FindDataRequestSchema } from './protocol.zod';
+import { DataEngineSortSchema, QueryTransportParamsSchema } from '../data/data-engine.zod';
+
+/** The two shapes `ODataQuerySchema` declares and the transport schema refuses. */
+const ODATA_SPELLINGS: ReadonlyArray<readonly [string, unknown]> = [
+  ["the 'field direction' expression", 'name desc'],
+  ["the '-field' shorthand", '-created_at'],
+  ['the expression array', ['name desc', 'email asc']],
+  ['a single-element expression array', ['name']],
+];
+
+/** The three shapes `DataEngineSortSchema` declares and the OData schema refuses. */
+const TRANSPORT_SPELLINGS: ReadonlyArray<readonly [string, unknown]> = [
+  ['the asc/desc record map', { name: 'desc' }],
+  ['the 1/-1 record map', { name: 1 }],
+  ['the SortNode array', [{ field: 'name', order: 'desc' }]],
+];
+
+describe('[#18977] $orderby is declared twice — the two accept sets', () => {
+  describe('§1 ODataQuerySchema.$orderby — the OData URL-convention vocabulary', () => {
+    it.each(ODATA_SPELLINGS)('accepts %s', (_label, value) => {
+      const parsed = ODataQuerySchema.safeParse({ $orderby: value });
+      expect(parsed.success).toBe(true);
+    });
+
+    it.each(TRANSPORT_SPELLINGS)('refuses %s, at the $orderby member', (_label, value) => {
+      const parsed = ODataQuerySchema.safeParse({ $orderby: value });
+      expect(parsed.success).toBe(false);
+      if (parsed.success) return;
+      expect(parsed.error.issues.map((i) => i.path.join('.'))).toContain('$orderby');
+    });
+  });
+
+  describe('§2 DataEngineSortSchema — what QueryTransportParamsSchema.$orderby is', () => {
+    it.each(TRANSPORT_SPELLINGS)('accepts %s', (_label, value) => {
+      expect(DataEngineSortSchema.safeParse(value).success).toBe(true);
+      expect(QueryTransportParamsSchema.safeParse({ $orderby: value }).success).toBe(true);
+    });
+
+    it.each(ODATA_SPELLINGS)('refuses %s, at the $orderby member', (_label, value) => {
+      expect(DataEngineSortSchema.safeParse(value).success).toBe(false);
+      const parsed = QueryTransportParamsSchema.safeParse({ $orderby: value });
+      expect(parsed.success).toBe(false);
+      if (parsed.success) return;
+      expect(parsed.error.issues.map((i) => i.path.join('.'))).toContain('$orderby');
+    });
+
+    it('refuses the OData spelling on `sort` too — the bare transport alias of the same slot', () => {
+      // A reader who works around the `$orderby` refusal by re-spelling the key
+      // gets the same answer: the refusal is on the VALUE, not on the spelling.
+      for (const [, value] of ODATA_SPELLINGS) {
+        expect(QueryTransportParamsSchema.safeParse({ sort: value }).success).toBe(false);
+      }
+    });
+  });
+
+  describe('§3 the two accept sets are DISJOINT — this is the trap', () => {
+    it('no declared $orderby value parses under both', () => {
+      const every = [...ODATA_SPELLINGS, ...TRANSPORT_SPELLINGS];
+      const bothAccept = every.filter(([, value]) =>
+        ODataQuerySchema.safeParse({ $orderby: value }).success
+        && DataEngineSortSchema.safeParse(value).success);
+      expect(bothAccept.map(([label]) => label)).toEqual([]);
+    });
+
+    it('every declared $orderby value parses under exactly one of them — neither set is empty', () => {
+      // The lit control for the emptiness above: a disjointness assertion is
+      // also satisfied by two schemas that accept nothing at all.
+      const every = [...ODATA_SPELLINGS, ...TRANSPORT_SPELLINGS];
+      const accepted = every.map(([, value]) =>
+        Number(ODataQuerySchema.safeParse({ $orderby: value }).success)
+        + Number(DataEngineSortSchema.safeParse(value).success));
+      expect(accepted).toEqual(every.map(() => 1));
+    });
+  });
+
+  describe('§4 which one grades a query bag', () => {
+    // `POST /data/:object/query` parses its body through this schema and answers
+    // `400 VALIDATION_FAILED` on a refusal (`rest-server.ts`), so this is the
+    // declaration an author's stored query bag is actually judged against.
+    const findInput = (query: Record<string, unknown>) => ({ object: 't', query: { ...query, object: 't' } });
+
+    it.each(TRANSPORT_SPELLINGS)('FindDataRequest.query accepts %s on $orderby', (_label, value) => {
+      expect(FindDataRequestSchema.safeParse(findInput({ $orderby: value })).success).toBe(true);
+    });
+
+    it.each(ODATA_SPELLINGS)('FindDataRequest.query refuses %s on $orderby', (_label, value) => {
+      const parsed = FindDataRequestSchema.safeParse(findInput({ $orderby: value }));
+      expect(parsed.success).toBe(false);
+      if (parsed.success) return;
+      expect(parsed.error.issues.map((i) => i.path.join('.'))).toContain('query.$orderby');
+    });
+
+    it('ODataQuerySchema grades nothing here — it is not on any path into FindDataRequest', () => {
+      // The canonical AST key is the one the output carries, whichever declared
+      // spelling arrived: the transport spelling folds onto `orderBy`.
+      const parsed = FindDataRequestSchema.safeParse(findInput({ $orderby: { created_at: 'desc' } }));
+      expect(parsed.success).toBe(true);
+      if (!parsed.success) return;
+      expect((parsed.data as { query: { orderBy?: unknown } }).query.orderBy)
+        .toEqual([{ field: 'created_at', order: 'desc' }]);
+      expect((parsed.data as { query: Record<string, unknown> }).query.$orderby).toBeUndefined();
+    });
+  });
+});
diff --git a/packages/spec/src/api/odata.zod.ts b/packages/spec/src/api/odata.zod.ts
index 30399fd71ab..b25169449f6 100644
--- a/packages/spec/src/api/odata.zod.ts
+++ b/packages/spec/src/api/odata.zod.ts
@@ -71,6 +71,47 @@ import { z } from 'zod';
  * System query options defined by OData v4 specification.
  * These are URL query parameters that control the query execution.
  * 
+ * ## [#18977] This schema grades no runtime door — and a SECOND declaration of
+ * the same parameter names does
+ *
+ * The declaration that grades a query bag is `QueryTransportParamsSchema`
+ * (`../data/data-engine.zod.ts`), reached from `FindDataRequestSchema.query`
+ * through `QueryWithTransportSchema`. This one is the OData v4 URL-convention
+ * VOCABULARY and nothing parses through it: measured on this tree, its only
+ * consumers are the `OData.buildUrl` helper at the foot of this file and this
+ * file's own unit test — no route, no ingress, no normalizer.
+ *
+ * ⚠️ On `$orderby` the two declarations are COMPLEMENTARY REFUSALS: each
+ * accepts exactly what the other rejects, so reading either one carefully and
+ * completely still produces the wrong answer about the other. Measured with
+ * `safeParse` on both, one tree:
+ *
+ * | `$orderby` value | here | `QueryTransportParamsSchema.$orderby` |
+ * |:---|:---|:---|
+ * | `'name desc'` / `'-created_at'` | accepted | REFUSED |
+ * | `['name desc', 'email asc']` | accepted | REFUSED |
+ * | `[{field, order}]` | REFUSED | accepted |
+ * | `{name: 'asc'}` / `{name: 1}` | REFUSED | accepted |
+ *
+ * Both halves are pinned in `odata-orderby-dual-declaration.test.ts`, which is
+ * the mechanical half of this cross-reference: widening or narrowing either
+ * side turns it red and lands the author here.
+ *
+ * ⛔ The gap is NOT closed by widening one side to match the other. #18704
+ * settled which spelling is canonical and why the transport schema refuses
+ * these two: lowering a sort EXPRESSION means PARSING, and a second parser
+ * beside the door's is how one rule gets two implementations that disagree
+ * (the paragraph above `QueryTransportParamsSchema` states it verbatim).
+ * Changing either accept set is a decision, not a tidy-up.
+ *
+ * ⭐ What the string forms are not is unserved. `normalizeSortNodes`
+ * (`@objectstack/metadata-protocol`) is the one shared ingress normalizer
+ * behind `GET /data/:object`, the export route and in-process `findData`, and
+ * it reads `'name desc'`, `'-created_at'` and the `string[]` form — so a
+ * querystring spelled the OData way works, while the same bag sent as a
+ * `POST /data/:object/query` body answers `400 VALIDATION_FAILED`. The
+ * difference is the DOOR, and neither door is this schema.
+ *
  * @see https://docs.oasis-open.org/odata/odata/v4.01/odata-v4.01-part2-url-conventions.html#sec_SystemQueryOptions
  */
 import { lazySchema } from '../shared/lazy-schema';
@@ -115,6 +156,15 @@ export const ODataQuerySchema = lazySchema(() => z.object({
    * @example "name"
    * @example "revenue desc"
    * @example "country asc, revenue desc"
+   *
+   * ⛔ [#18977] These two shapes are exactly the ones
+   * `QueryTransportParamsSchema.$orderby` (`DataEngineSortSchema`,
+   * `../data/data-engine.zod.ts`) DELIBERATELY refuses, and that is the
+   * declaration a query bag is graded against. Sent as a
+   * `POST /data/:object/query` body this spelling answers
+   * `400 VALIDATION_FAILED`; sent on the querystring it is parsed by
+   * `normalizeSortNodes` at the ingress and works. See the cross-reference on
+   * the schema above before writing either shape into stored metadata.
    */
   $orderby: z.union([
     z.string(),           // "name desc"
diff --git a/packages/spec/src/data/data-engine.zod.ts b/packages/spec/src/data/data-engine.zod.ts
index 1d95c4acd37..683e600b5a4 100644
--- a/packages/spec/src/data/data-engine.zod.ts
+++ b/packages/spec/src/data/data-engine.zod.ts
@@ -830,6 +830,25 @@ const TransportCountValueSchema = lazySchema(() => z.union([z.boolean(), z.enum(
  * its `string[]` form. They fail the parse at the member that carries them
  * rather than reaching the AST as a string the engine would have to re-read.
  *
+ * ⚠️ [#18977] `$orderby` is declared a SECOND time, and the other declaration
+ * accepts exactly the two shapes this one refuses: `ODataQuerySchema.$orderby`
+ * (`../api/odata.zod.ts`) is `string | string[]` and refuses this one's record
+ * maps and `SortNode[]`. It cannot contradict this schema at a DOOR — measured
+ * on this tree, nothing parses through it: its only consumers are its own
+ * `OData.buildUrl` helper and its own unit test. It contradicts it in a READER,
+ * which is the cost already paid: objectui#9554 was filed, triaged and
+ * dispatched against a shipped producer that had been sending the canonical
+ * shape all along, because a competent seat read the OTHER declaration, quoted
+ * it correctly, and had no signal that this one exists. The two accept sets are
+ * disjoint and pinned as such in
+ * `../api/odata-orderby-dual-declaration.test.ts`.
+ *
+ * ⛔ Closing that gap by widening either side is a decision, not a tidy-up —
+ * and widening THIS one is precisely the second parser the paragraph above
+ * refuses. What serves the string forms is `normalizeSortNodes` at the
+ * `@objectstack/metadata-protocol` ingress (the GET querystring path, the
+ * export route and in-process `findData`), not a schema.
+ *
  * ⛔ Declaring the narrower structured form ALONE would have turned live
  * traffic into a `400` — measured: the body-form AST array on
  * `POST /data/:object/query`, pinned by `#7390 §3` in