fix(components)!: data-table reads only the declared accessorKey (retire the col.name alias)

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

.changeset/5120-retire-data-table-name-alias.md

@@ -0,0 +1,60 @@
+---
+'@object-ui/components': minor
+---
+
+**Breaking for authored metadata:** a `data-table` column spelled `name` no
+longer resolves its cells. Use the declared `accessorKey`.
+
+`data-table`'s column normalization used to read
+`accessorKey: col.accessorKey || col.name` — but `TableColumn`
+(`@object-ui/types`) declares only `accessorKey`, never `name`. The declared
+surface admitted one spelling while the runtime admitted two, which is the
+second de-facto contract AGENTS.md #0.1 forbids. The maintainer ruling of
+2026-08-20 settled the direction for the whole family: retire the consumer-side
+alias, translate at the producers. `label` → `header` retired first
+(objectui#5351); this retires `name` → `accessorKey` and closes the family.
+
+**Who is affected — a column authored DIRECTLY onto a `data-table` node:**
+
+```json
+{ "type": "data-table",
+ "columns": [{ "name": "email", "label": "Email" }] } // ← was tolerated
+```
+
+becomes
+
+```json
+{ "type": "data-table",
+ "columns": [{ "header": "Email", "accessorKey": "email" }] }
+```
+
+**Who is NOT affected.** Columns reaching the table through `object-data-table`,
+a detail view's `related[]` list, or `object-grid` are unchanged — the adapter
+never sees a legacy spelling from any of them. The reason differs by producer,
+and the difference matters if you are debugging one:
+
+- `object-data-table` and a detail view's `related[]` list **resolve** the
+ legacy spelling before delivery, stamping `accessorKey` from `name` (via
+ `columnIdentity`). A `name`-spelled column keeps working there.
+- `object-grid` **refuses** it instead: since objectui#5068 an authored column
+ must spell the declared `field`, and one that does not is dropped at intake
+ and never reaches the table. Its delivered columns carry `accessorKey`
+ stamped from `field`. So a `name`-spelled `object-grid` column does not
+ render today either — that is objectui#5352's open question, unchanged by
+ this release.
+
+Only the directly-authored `data-table` node narrows here.
+
+**How the break presents, so you can recognise it:** the column is not dropped
+and nothing is thrown — its header still renders over blank cells, and
+neighbouring columns are unaffected. If a table's header row looks right but one
+column's cells are empty, check that column's key spelling first.
+
+The two published skill guides that taught the `name` spelling
+(`skills/objectui/guides/data-integration.md`, `schema-expressions.md`) migrate
+in this same release, so the platform never refuses a spelling it still ships.
+
+Graded `minor`, not `patch`: this narrows the accepted input set, which is a
+breaking change for any author who used the tolerated spelling. It is not
+`major` per this repo's fixed-group convention (objectui's own breaking changes
+ship as `minor`; the group's major tracks `@objectstack`).

packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx

@@ -40,11 +40,24 @@
  * empty state. The correction is verified against the renderer rather
  * than against a reading of it.
  *
- * The `columns` entries in these examples are deliberately untouched: their
- * `{ name, label }` spelling is the separate open question on objectui#5120
- * (the undeclared `col.name` / `col.label` alias at `data-table.tsx:776-777`),
- * parked with the maintainer. This file pins the BINDING only, and stays true
- * whichever way that one lands — nothing below asserts a column key spelling.
+ * ⚠️ COLUMN SPELLING — this file is NOT spelling-independent, and an earlier
+ * revision of this docblock said it was. It claimed the file "pins the BINDING
+ * only … nothing below asserts a column key spelling". That was false and was
+ * filed as objectui#5479: the BEHAVIOUR assertions below are on rendered CELL
+ * TEXT, and a cell only has text when its column's accessor resolves — so they
+ * transitively pin the accessor spelling the guides use. A docblock asserting
+ * independence over assertions that were not independent is what made
+ * objectui#5120's last step invisible when the family was ruled on.
+ *
+ * Both aliases have since retired — `label` in objectui#5351, `name` in
+ * objectui#5120 — and the two guides' `data-table` columns migrated to the
+ * declared `{ header, accessorKey }` in the same commit as the `name`
+ * retirement, because these blocks are lifted and rendered at run time and
+ * would otherwise go blank. That coupling is the point, not a defect: the
+ * guides are executable fixtures, so the instruction corpus cannot drift from
+ * the runtime without this file going red. Keep it that way — if a future
+ * change makes the adapter's accepted key set narrower again, migrate the
+ * guides in the SAME commit.
  */
 
 import { describe, it, expect, vi, beforeEach, afterEach } from 'vitest';
@@ -162,6 +175,31 @@ describe('skill guides — no `data-table` example is bound with `bind` (#5126,
  expect(offenders).toEqual([]);
  });
 
+ it('no `data-table` example spells a column with the retired `name` (#5120)', () => {
+ // objectui#5120 retired `accessorKey: col.accessorKey || col.name` on the
+ // adapter. These blocks are RENDERED below, so a guide that regressed to
+ // `name` would already fail — but only as "expected [] to equal
+ // ['Ada Lovelace', …]", which names neither the key nor the card. This
+ // assertion is here to make the failure SAY what broke, and to hold the
+ // corpus even if the behaviour legs are ever narrowed.
+ for (const rel of GUIDE_PATHS) {
+ const md = readGuide(rel);
+ for (const node of blocksOfType(md, 'data-table')) {
+ const columns = (node.columns ?? []) as Record<string, unknown>[];
+ for (const col of columns) {
+ expect(
+ 'name' in col,
+ `${rel}: a data-table column spells the retired \`name\`; the declared key is \`accessorKey\` (objectui#5120)`,
+ ).toBe(false);
+ expect(
+ 'accessorKey' in col,
+ `${rel}: a data-table column is missing the declared \`accessorKey\` (objectui#5120)`,
+ ).toBe(true);
+ }
+ }
+ }
+ });
+
  it('no guide claims a table component calls useDataScope', () => {
  for (const rel of GUIDE_PATHS) {
  const md = readGuide(rel);

packages/components/src/__tests__/skill-guide-provider-envelope.test.tsx

@@ -57,9 +57,16 @@ const here = path.dirname(fileURLToPath(import.meta.url));
 const repoRoot = path.resolve(here, '../../../..');
 const skillsRoot = path.join(repoRoot, 'skills/objectui');
 
+// The DECLARED column keys (`TableColumn`: `header` + `accessorKey`). These
+// mirror the two published guides' `data-table` example, which is why they
+// moved when the guides did: objectui#5120 retired the adapter's undeclared
+// `col.name` alias, so the `{ name, label }` spelling this fixture used to
+// carry now resolves no accessor and every cell below would read ''. The rows
+// keep `name`/`email` as their own DATA keys — that is what `accessorKey`
+// points AT, and it is unrelated to the column vocabulary.
 const COLUMNS = [
- { name: 'name', label: 'Name' },
- { name: 'email', label: 'Email' },
+ { header: 'Name', accessorKey: 'name' },
+ { header: 'Email', accessorKey: 'email' },
 ];
 const ROWS = [
  { name: 'Ada Lovelace', email: 'ada@example.com' },

packages/components/src/renderers/complex/__tests__/data-table-declared-column-keys.test.tsx

@@ -7,8 +7,9 @@
  */
 
 /**
- * `data-table` reads the DECLARED `header`, not the undeclared `label`
- * (objectui#5351).
+ * `data-table` reads the DECLARED `header` and `accessorKey`, and NOTHING else
+ * — not the undeclared `label` (objectui#5351), not the undeclared `name`
+ * (objectui#5120).
  *
  * `TableColumn` (`packages/types/src/data-display.ts`) declares `header: string`
  * and `accessorKey: string`. It declares neither `label` nor `name`. The
@@ -26,14 +27,39 @@
  * purpose. Metadata vocabulary in, adapter vocabulary out; one translation, one
  * place — and that place is each producer, never here.
  *
- * SCOPE. The `label` alias retires here; the `name` alias is HELD, and the last
- * describe below pins it as still-read so the hold cannot be mistaken for the
- * retirement having happened. Two published skill guides teach a directly
- * authored `data-table` whose columns are spelled `{ name, label }`, and
+ * SCOPE. `label` retired in objectui#5351; `name` retires here, closing the
+ * family. The hold that kept `name` alive was never about the producers — all
+ * three resolve `accessorKey` themselves, measured — but about the INSTRUCTION
+ * CORPUS: two published skill guides taught a directly authored `data-table`
+ * whose columns were spelled `{ name, label }`, and
  * `skill-guide-data-table-binding.test.tsx` renders those blocks straight out of
- * the guide files — so `name` cannot retire until the instruction corpus moves.
- * That is objectui#5120's remaining step and is nobody's to take unbidden:
- * `skills/**` is a customer-published surface with its own owning seat.
+ * the guide files. Both guides migrated to `{ header, accessorKey }` in the same
+ * commit as this retirement, which is the only ordering in which the platform
+ * never refuses a spelling it still ships.
+ *
+ * WHAT NARROWS, precisely. A `{ name, label }` column arriving through
+ * `ObjectDataTable`, `RelatedList` or `ObjectGrid` is UNAFFECTED — but for two
+ * different reasons, and conflating them is a mistake this card's own ruling
+ * already made once:
+ *
+ * - `ObjectDataTable` and `RelatedList` RESOLVE the legacy spelling, stamping
+ * `accessorKey` from `columnIdentity(col)` before delivery.
+ * - `ObjectGrid` does NOT fold. Since objectui#5068 it REFUSES undeclared
+ * spellings at intake — `resolvesToDataColumn` requires a string `field`
+ * (`columnSpellingDiagnostics.ts`), so a `name`-spelled entry is dropped
+ * before delivery and never reaches this adapter at all (before AND after
+ * this change; that silence is objectui#5352's territory). What it does
+ * deliver carries `accessorKey` stamped from `field`.
+ *
+ * The 2026-08-20 ruling's item 2 told ObjectGrid to "connect to the same
+ * `columnIdentity` resolution"; objectui#5478 measured that following it would
+ * have RE-WIDENED what #5068 narrowed. Stated here because the claim has been
+ * restated wrongly more than once, and a third repetition would settle it.
+ *
+ * Either way the adapter never sees a legacy spelling from a producer. What
+ * stops resolving is `name` on a column authored DIRECTLY onto a `data-table`
+ * node. That is the whole blast radius, and it is pinned below in both
+ * directions.
  *
  * The two aliases were DIFFERENT failure classes, which is why the cards were
  * filed apart: an unresolved `accessorKey` gives blank cells under a live
@@ -79,24 +105,52 @@ describe('data-table columns — the declared keys render (unchanged)', () => {
  });
 });
 
-describe('data-table columns — the `name` alias is still read, and that is a HOLD (#5120)', () => {
- it('still resolves an accessor from the undeclared `name`', () => {
- // NOT an endorsement — a receipt. The 2026-08-20 ruling retires this limb;
- // what stops it today is that two published skill guides teach it and
- // `skill-guide-data-table-binding.test.tsx` renders their bytes. Pinning the
- // CURRENT behaviour means the day the guides move, this test goes red and
- // names itself as the thing to delete, instead of the retirement quietly
- // never happening.
+describe('data-table columns — the undeclared `name` alias is retired (#5120)', () => {
+ it('does not resolve an accessor from `name`', () => {
+ // The narrowing this card ships, and the receipt that replaces the HOLD pin
+ // that stood here while the instruction corpus still taught `name`.
+ // A DIFFERENT failure class from #5351's: the header is fine and the CELLS
+ // are what go blank.
  renderTable([{ header: 'Stage', name: 'stage' }]);
- expect(bodyCells()).toEqual(['Won', 'Lost']);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['', '']);
  });
 
- it('keeps an authored `accessorKey` ahead of a divergent `name`', () => {
+ it('keeps an authored `accessorKey` winning over a divergent `name`', () => {
  // Precedence, unchanged and load bearing: columns can arrive in the table
- // library's own shape, and those must not be second-guessed.
+ // library's own shape, and those must not be second-guessed. This passed
+ // before the retirement too — it is here so the two directions cannot drift.
  renderTable([{ header: 'Stage', accessorKey: 'stage', name: 'nonsense' }]);
  expect(bodyCells()).toEqual(['Won', 'Lost']);
  });
+
+ it('LEGIBILITY: an unresolvable column keeps its header and spares its neighbour', () => {
+ // Measured, not assumed, and deliberately pinned as the SHAPE OF THE BREAK:
+ // the column is not dropped and nothing throws — a live header sits over
+ // blank cells while the declared neighbour is untouched. This is the failure
+ // an author who kept the legacy spelling now gets, and it is the same
+ // illegible shape objectui#5349 measures against.
+ renderTable([
+ { header: 'Stage', name: 'stage' },
+ { header: 'Id', accessorKey: 'id' },
+ ]);
+ expect(headers()).toEqual(['Stage', 'Id']);
+ expect(bodyCells()).toEqual(['', '1', '', '2']);
+ });
+
+ it('a PRODUCER-resolved column is unaffected — the narrowing is adapter-only', () => {
+ // The blast-radius bound in one assertion: whatever a producer delivers
+ // already carries a declared `accessorKey`, so the retirement cannot reach
+ // it. ObjectDataTable and RelatedList get there by RESOLVING `name` through
+ // `columnIdentity`; ObjectGrid gets there by REFUSING undeclared spellings
+ // at intake and stamping from `field` (see the header — the two routes are
+ // not the same mechanism). Simulating the hand-off here keeps this file
+ // honest about what the retirement does and does not break, without
+ // importing a plugin package.
+ renderTable([{ header: 'Stage', name: 'stage', accessorKey: 'stage' }]);
+ expect(headers()).toEqual(['Stage']);
+ expect(bodyCells()).toEqual(['Won', 'Lost']);
+ });
 });
 
 describe('data-table columns — the undeclared `label` alias is retired (#5351)', () => {

packages/components/src/renderers/complex/data-table.tsx

@@ -832,9 +832,9 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  if (nonArrayDataMessage) console.warn(nonArrayDataMessage);
  }, [nonArrayDataMessage]);
 
- // The adapter reads the column keys `TableColumn` DECLARES. The `label`
- // alias is gone (objectui#5351); the `name` alias is HELD, and the hold is
- // deliberate and documented rather than an oversight.
+ // The adapter reads ONLY the column keys `TableColumn` DECLARES. Both
+ // undeclared aliases are now retired: `label` (objectui#5351) and `name`
+ // (objectui#5120, this change).
  //
  // These two lines used to normalize each column as
  // `header: col.header || col.label` and
@@ -846,28 +846,33 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  // ruling settled the direction for the whole family: retire the consumer-side
  // alias, unify the producers.
  //
- // `header` has retired. Where its translation went — `columnHeader` in
- // `@object-ui/core`, called by each producer before delivery:
+ // Where the translation went — `columnIdentity` / `columnHeader` in
+ // `@object-ui/core`, called by each producer BEFORE delivery:
  // `ObjectDataTable.normalizeColumns` (`@object-ui/plugin-dashboard`)
  // `RelatedList.normalizeColumn` (`@object-ui/plugin-detail`)
  // `ObjectGrid.generateColumns` (`@object-ui/plugin-grid`, since #5068)
- // Metadata vocabulary in, adapter vocabulary out; one translation, one place.
+ // Metadata vocabulary in, adapter vocabulary out; one translation, one place —
+ // and that place is each producer, never here. A `{ name, label }` column
+ // arriving through any of the three still renders: its producer resolved the
+ // identity into `accessorKey` before the adapter ever saw it. What no longer
+ // resolves is `name` on a column authored DIRECTLY onto a `data-table` node,
+ // which is the accepted-set narrowing objectui#5120 rules and ships.
  //
- // `accessorKey || col.name` STAYS, pending objectui#5120's remaining step. It
- // is not that the producers still need it — all three resolve `accessorKey`
- // themselves, measured — but that two PUBLISHED skill guides teach a directly
- // authored `data-table` whose columns are spelled `{ name, label }`:
+ // The instruction corpus moved in the SAME commit, which is the whole reason
+ // this step could be taken: `skill-guide-data-table-binding.test.tsx` lifts the
+ // fenced JSON out of the published guides at run time and renders it, so the
+ // guides are executable fixtures rather than prose. Both now teach
+ // `{ header, accessorKey }`:
  // skills/objectui/guides/data-integration.md
  // skills/objectui/guides/schema-expressions.md
- // `skill-guide-data-table-binding.test.tsx` lifts those blocks out of the real
- // files at run time and renders them, so retiring `name` here turns that gate
- // red until the guides move. Retiring the runtime ahead of the instruction
- // would leave the platform refusing a spelling it still ships, and the failure
- // it teaches into is the illegible one below: a header over blank cells.
+ // Retiring the runtime ahead of the instruction would have left the platform
+ // refusing a spelling it still shipped, and the failure it teaches into is the
+ // illegible one: a header over blank cells (pinned in
+ // `data-table-declared-column-keys.test.tsx`).
  const initialColumns = useMemo(() => {
  return rawColumns.map((col: any) => ({
  ...col,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  }));
  }, [rawColumns]);
 
@@ -876,10 +881,10 @@ const DataTableRenderer = ({ schema }: { schema: DataTableSchema }) => {
  const widths: Record<string, number> = {};
  // Spelled identically to `initialColumns` above — the auto-width pass must
  // measure the SAME columns the table renders, so the two reads move
- // together (objectui#5351 retired `header`'s alias; `name`'s is held).
+ // together (objectui#5351 retired `header`'s alias, objectui#5120 `name`'s).
  const cols = rawColumns.map((col: any) => ({
  header: col.header,
- accessorKey: col.accessorKey || col.name,
+ accessorKey: col.accessorKey,
  width: col.width,
  fitContent: col.fitContent,
  }));

skills/objectui/guides/data-integration.md

@@ -217,8 +217,8 @@ through is measured, with its open-question caveat, in
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```

skills/objectui/guides/schema-expressions.md

@@ -421,8 +421,8 @@ and is blank is the whole failure.
  { "name": "Grace Hopper", "email": "grace@example.com" }
  ],
  "columns": [
- { "name": "name", "label": "Name" },
- { "name": "email", "label": "Email" }
+ { "header": "Name", "accessorKey": "name" },
+ { "header": "Email", "accessorKey": "email" }
  ]
 }
 ```