fix(types,components,docs): correct the texts objectui#6771 left false about body (objectui#8284)

.changeset/6773-aspect-ratio-demo-content.md

@@ -14,6 +14,13 @@ nested `card` never read `content` either, so moving only the outer key would ha
 an empty box for an empty card. The page's Schema block published `content` as contract
 while omitting `image`/`alt`; it now documents what the renderer reads.
 
+⚠️ **Dated note, 2026-09-28 — the child list is `children` alone — objectui#6771.** Since
+this change, objectui#6771 dropped the `body` arm of that read, in `aspect-ratio.tsx` and in
+the nested `card` alike: both read `children` and never `body`, both zod mirrors refuse an
+authored `body` by name, and both TypeScript faces declare it `never`. The four card demos
+already author `children` at both levels, so they are unaffected. The rest of this entry is
+kept as the reading of this change.
+
 Nothing publishes from this change — a docs page plus `@object-ui/example-schema-catalog`
 fixtures, both outside the release — hence the empty frontmatter. The regression control is
 `examples/schema-catalog/test/aspect-ratio-demo-content-6773.test.tsx`: category scope, not

.changeset/6788-context-menu-demo-content.md

@@ -18,4 +18,11 @@ dialect, and objectui#6773 authored `children` in the four sibling
 `aspect-ratio` card demos. The renderer was NOT widened to read `content` —
 that would add a second dialect for one slot to a published surface.
 
+⚠️ **Dated note, 2026-09-28 — `card` reads `children` alone — objectui#6771.** Since this
+change, objectui#6771 retired `body` as a child-list spelling: the `children || body` read
+above is now `children` alone, `CardSchema` declares `body` as `never` and its zod mirror
+refuses it by name, and `BaseSchema.body` is `never` rather than a legacy member. So the
+demo's `children` is the one spelling that renders, not the preferred one of two. The rest
+of this entry is kept as the reading of this change.
+
 No package source changed, so this declares no release.

.changeset/6877-div-guidance-names-box.md

@@ -15,3 +15,11 @@ exactly the conversions objectui#3965 measured and rejected.
 Both statements now name `box` first for the mechanical swap, keep the layout components for the
 cases where their layout is actually wanted, and state the `body` → `children` move that every
 option except `card` requires. `span`'s guidance is deliberately unchanged.
+
+⚠️ **Dated note, 2026-09-28 — every option, and `div` itself, reads `children` only —
+objectui#6771.** Later in this same release objectui#6771 retired `body` as a child-list
+spelling, and `card` and `div` dropped their `body` arms with it. So the `body` → `children`
+move is one every option requires, `card` included. A retype no longer drops `body` content
+either: `div` does not draw it, and validation refuses the key by name. The notice's `body`
+bullet now says that (objectui#8284); `deprecated.replacement` is unchanged. The rest of this
+entry is kept as the reading of this change.

.changeset/8284-content-channel-per-component.md

@@ -27,6 +27,19 @@ that names the channel to write instead:
 | `children` | `box`, `span`, `container`, `flex`, `stack`, `grid`, `scroll-area`, `form`, `toggle` | `body` |
 | `body` (at the time of this change; objectui#6771 has since retired that spelling and these three read `children`) | `alert`, `badge`, `tooltip` (which reads `content` first, the child list as its fallback) | the other channel |
 
+⚠️ **Dated note, 2026-09-28 — `BaseSchema` declares one content channel, and the
+`alert` / `badge` / `tooltip` row is inverted at release — objectui#6771.** Later in this same
+release objectui#6771 retired `body` as a child-list spelling: `BaseSchema.body` is `never` on
+the TypeScript face and refused by name on the zod mirror, and `children` is the one
+child-list key. So the paragraph that opens "`BaseSchema` declares two optional content
+channels" no longer describes `BaseSchema`, and for all twelve components in the table the
+renderer reads `children` and the refused channel is `body`. That includes `alert`, `badge`
+and `tooltip`, which accept `children` and refuse `body`, not the other way round. For those
+three the Migration paragraph below does not hold as written: in the previous release their
+renderers drew an authored `body`, and this release refuses it. objectui#6771's entry states
+that migration: author `children`. The rest of this entry is kept as the reading of this
+change.
+
 Which channel each renderer reads was measured with the TypeScript type checker over
 every `ComponentRegistry.register(...)` call in `packages/components` — a read site is a
 property access filed under the type of the object it is read from, so a docblock mention

.changeset/8284-refusal-and-notice-text.md

@@ -0,0 +1,23 @@
+---
+'@object-ui/types': patch
+'@object-ui/components': patch
+---
+
+Correct author-facing text that objectui#6771's retirement of the `body` child-list spelling
+left false (objectui#8284). Only wording moves; no key, value or rendered result does.
+
+- `@object-ui/types` — the `body` refusal on `box`, `span`, `container`, `flex`, `stack`,
+  `grid`, `scroll-area`, `form` and `toggle` explained that `body` "is inherited from
+  `BaseSchema`, so an authored `body` parsed green here". Since objectui#6771, `BaseSchema`
+  refuses `body` by name itself, so that explanation no longer holds. The message now says that
+  `body` is the child-list spelling objectui#6771 retired, refused by name, and that `children`
+  is the one child-list key. The same refusal on `alert` and `badge` said that an authored
+  `body` "now parses green through `.passthrough()`", which that refusal itself contradicts;
+  like the other nine, it now says that `body` is refused here by name and that `children` is
+  the one child-list key. The refused key, the replacement it names, the issue code and the
+  issue path are unchanged; the same string is still the `.describe()` metadata.
+- `@object-ui/components` — the `div` deprecation notice said that every replacement except
+  `card` reads `children` only, so a blind retype drops `body` content. Since objectui#6771,
+  `card` and `div` itself read `children` only as well. The notice now says that validation
+  refuses `body` by name and that neither `div` nor any replacement draws it. The replacements
+  it offers are unchanged.

content/docs/components/basic/div.mdx

@@ -13,9 +13,9 @@ description: "Generic container element - use Shadcn components instead"
     <li>When you want their layout: use <code>card</code>, <code>flex</code>, <code>container</code>, <code>stack</code>, or <code>grid</code> — each injects classes of its own, and <code>card</code> also moves children into an extra element</li>
   </ul>
   <p className="mt-2 text-sm text-yellow-700 dark:text-yellow-300">
-    Move any content authored under the legacy <code>body</code> key into <code>children</code> first: every
-    replacement listed above except <code>card</code> reads <code>children</code> only, so a blind retype drops
-    that content silently, at an unchanged element count.
+    Move any content authored under the retired <code>body</code> key into <code>children</code> first:
+    objectui#6771 retired that child-list spelling, so validation refuses it by name, and neither
+    <code>div</code> nor any replacement listed above draws it.
   </p>
 </div>
 
@@ -47,9 +47,10 @@ the page; a `div` → `box` swap does not.
 
 One caveat for a node still authored in the retired `body` spelling: every
 option reads `children` and never `body` (objectui#6771 retired it across the
-protocol, `card` included). Such a node must move that content into `children`
-as part of the swap, or it disappears from the page without changing the element
-count — silently.
+protocol, `card` included), and so does `div` itself. Such a node must move that
+content into `children` as part of the swap. Until it does, validation refuses
+the key by name, and the content is missing from the page under `div` and under
+every replacement alike.
 
 ### For a Plain Wrapper → Use `box`
 
@@ -83,8 +84,7 @@ interface DivSchema {
   type: 'div';
 
   // Content
-  children?: SchemaNode | SchemaNode[];  // Child components
-  body?: SchemaNode[];                    // Alternative content prop
+  children?: SchemaNode | SchemaNode[];  // Child components (the child key this component reads)
 
   // Styling
   className?: string;                     // Tailwind CSS classes

content/docs/components/layout/page.mdx

@@ -25,8 +25,7 @@ interface PageSchema {
   description?: string;                 // Page description
 
   // Content
-  body?: SchemaNode | SchemaNode[];     // Main content — one node or a list
-  children?: SchemaNode | SchemaNode[]; // Alternative content prop
+  children?: SchemaNode | SchemaNode[]; // Main content — one node or a list
 
   // Styling
   className?: string;                   // Tailwind CSS classes

packages/components/src/renderers/basic/div.tsx

@@ -80,7 +80,10 @@ function warnDeprecatedOnce(type: string, message: string): void {
  * and four of those five — `flex`, `container`, `stack`, `grid` — read
  * `children` ONLY, so a node that authored `body` loses its content SILENTLY,
  * at an unchanged element count. `box` is the one class-transparent swap, and
- * the old text never named it.
+ * the old text never named it. (That was the measurement at the time. Since
+ * objectui#6771 retired the `body` spelling, `card` and `div` itself read
+ * `children` only as well, so `body` content is drawn by none of them and the
+ * notice's `body` bullet says so.)
  *
  * That is why this is worth a re-ruling rather than a nice-to-have. Deprecation
  * guidance is followed LITERALLY, by humans and by generating models reading
@@ -116,7 +119,7 @@ const DIV_DEPRECATION_NOTICE =
   '[ObjectUI] The "div" component is deprecated on every authoring surface. Please use Shadcn components instead:\n' +
   '  - For a plain wrapper the drop-in swap is "box": same element, your `className` verbatim, no layout of its own.\n' +
   '  - Reach for "card", "flex", "container", "stack", or "grid" only when you want their layout — each injects classes of its own, and "card" also moves children into an extra element.\n' +
-  '  - Move any `body` content into `children` first: every replacement above except "card" reads `children` only, so a blind retype drops it silently at an unchanged element count.\n' +
+  '  - Move any `body` content into `children` first: `body` is the child-list spelling objectui#6771 retired, so validation refuses it by name, and neither this component nor any replacement above draws it.\n' +
   '  This applies to JSON-authored nodes and to kind:\'html\' pages alike: an html page refuses the\n' +
   '  tag when it compiles, naming the same replacement.\n' +
   'See documentation at https://www.objectui.org/docs/components for alternatives.';

packages/components/src/renderers/layout/card.tsx

@@ -62,13 +62,16 @@ const CardRenderer = forwardRef<HTMLDivElement, { schema: CardSchema; className?
           {header}
         </CardHeader>
       )}
-      {/* `||` here is ALIAS RESOLUTION — `body` is the legacy spelling of
-          `children` — and ⛔ it is no longer the guard. That distinction is
-          the whole point: `children: 0` used to be converted away by the
-          accident of `0 || undefined === undefined`, which protected nothing,
-          because the sibling `body: 0` went through `undefined || 0 === 0`
-          and leaked. `renderNodeSlot` covers both; the alias order is
-          unchanged. */}
+      {/* `children` is the one child-list key: this slot used to fall back
+          from `children` to `body` through an `||`, and objectui#6771 dropped
+          that `body` arm when it retired the spelling (the callback's `body`
+          below is only a local name for the slot's content). ⛔ No `||` is
+          the guard here. That distinction was the point of objectui#9162:
+          `children: 0` used to be converted away by the accident of
+          `0 || undefined === undefined`, which protected nothing, because
+          the sibling `body: 0` went through `undefined || 0 === 0` and
+          leaked. `renderNodeSlot` is the guard, so an empty slot, `0`
+          included, renders no `CardContent` at all. */}
       {renderNodeSlot(schema.children, (body) => (
         <CardContent>{renderChildren(body)}</CardContent>
       ))}

packages/types/src/__tests__/content-channel-per-component-8284.test.ts

@@ -7,37 +7,45 @@
  */
 
 /**
- * objectui#8284 — the `body` / `children` duality on `BaseSchema` is resolved
- * PER COMPONENT: each component schema narrows to the channel its renderer
- * actually reads and TOMBSTONES the other, on both published faces
+ * objectui#8284 — each component schema narrows to the content channel its
+ * renderer actually reads and TOMBSTONES the other, on both published faces
  * (maintainer ruling, summon #17 decision batch #2, 2026-09-07, verbatim
- * 「同意」).
+ * 「同意」). This file pins that rule on the twelve dedicated declarations
+ * listed in `ROWS`.
  *
  * ## The defect this pins closed
  *
- * `BaseSchema` declares TWO optional content channels and its own docblock
- * admits that "some components use `children` instead of `body`" WITHOUT
- * saying which. The zod base is `.passthrough()` and both keys are optional,
+ * `BaseSchema` used to declare TWO optional content channels, and its docblock
+ * admitted that "some components use `children` instead of `body`" WITHOUT
+ * saying which. The zod base is `.passthrough()` and both keys were optional,
  * so a node carrying the WRONG channel type-checked, parsed green, was
  * PRESERVED by the parse, and then rendered an EMPTY element — no error at
  * authoring time, none at validation time, none at render time. Seven cards
  * repaired one page of that each (#5027 · #3900 · #6773 · #6806 · #8197 ·
  * #8234 · #6939) before the declaration itself was named.
  *
+ * objectui#6771 has since retired `body` as a child-list spelling on every
+ * node: `BaseSchema.body` is `never` on the TypeScript face and refused by name
+ * on the zod mirror, and `children` is the one child-list key. So on every row
+ * below the renderer reads `children` and the refused channel is `body`. What
+ * this file still pins is objectui#8284's own refusal on each row's DEDICATED
+ * declaration: it is installed there, it names both keys and this card, it is
+ * the `.describe()` metadata too, and it reaches a nested node.
+ *
  * ## The population here is FAMILY A + B of the measured table, not "all"
  *
  * The table posted on objectui#8284 is derived from every
  * `ComponentRegistry.register(...)` call under
- * `packages/components/src/renderers/**` (114 registrations) and the read side
- * is measured with the TypeScript TYPE CHECKER — each `body` / `children`
- * property access is filed under the TYPE of the object it is read from, so a
- * docblock mention cannot score.
+ * `packages/components/src/renderers/**` (114 registrations when it was taken)
+ * and the read side is measured with the TypeScript TYPE CHECKER — each
+ * `body` / `children` property access is filed under the TYPE of the object it
+ * is read from, so a docblock mention cannot score.
  *
  * ⚠️ THE CONTROL MOVED, because objectui#6771 SPENT the one that stood here.
  * It read: the `BoxSchema` docblock in `renderers/layout/box.tsx` SAYS
  * `schema.body` in prose and `grep -l` counts it, the checker does not — and
- * the card's own `17 / 17` figure came from that query. This card rewrote that
- * docblock, so `box.tsx` now answers 0 to BOTH halves and a reader replaying it
+ * the card's own `17 / 17` figure came from that query. objectui#6771 rewrote
+ * that docblock, so `box.tsx` now answers 0 to BOTH halves and a reader replaying it
  * gets no divergence at all. ⛔ A control that cannot fire is worse than none:
  * it reads as evidence and is not.
  *
@@ -51,24 +59,27 @@
  * `renderers/action/action-icon.tsx` is a REAL property access and survives
  * stripping at 1, so the 0 above is a reading and not a blanked file.
  *
- * The twelve rows below are the ones where the renderer reads EXACTLY ONE
- * channel, the component owns a dedicated declaration, and exactly one
- * registration claims its `type`. `sidebar` is measured into family B and held
- * BACK from it, because two registrations claim `sidebar` and the second types
- * its schema prop `any`. Families C (reads both — a live `children || body`
- * fallback), D (reads neither) and E (no dedicated declaration) are named on
- * the card with the measurement each still needs.
+ * The twelve rows below are the ones that table measured as reading EXACTLY ONE
+ * channel, owning a dedicated declaration, and claimed by exactly one
+ * registration. `sidebar` was measured into family B and held BACK, because two
+ * registrations claim `sidebar` and the second types its schema prop `any`; it
+ * is not a row here. The other families are not pinned in this file:
+ * family C (read both, through a live `children || body` fallback) lost its
+ * `body` arm to objectui#6771 and reads the `children` channel only; family D
+ * (reads neither) is pinned by `content-channel-family-d-9256.test.ts`
+ * (objectui#9256); family E (no dedicated declaration) was ruled done or
+ * superseded by objectui#6771 and objectui#9910, with its residual moved to
+ * objectui#9256 (ruling 5861449497 on objectui#8284).
  *
  * ## What is NOT pinned here, and why
  *
- * ⛔ Not the renderers. Nothing about rendering changes: a document that
- * authors the channel its renderer reads is byte-identical through both faces,
- * and a document that authors the other one rendered nothing before and
- * renders nothing now — it is merely REFUSED first. The counter-probes in
+ * ⛔ Not the renderers. This file pins the two authoring faces only: a
+ * document that authors the channel its renderer reads parses and compiles,
+ * and a document that authors `body` is refused by both.
+ * The render side is pinned elsewhere, by counter-probes that deliberately
+ * author the dead channel (`body`) and assert the empty render:
  * `examples/schema-catalog/test/badge-demo-label-6829.test.tsx` and
- * `packages/components/src/__tests__/span-children-rendering.test.tsx`, which
- * deliberately author the dead channel and assert the empty render, therefore
- * keep passing.
+ * `packages/components/src/__tests__/span-children-rendering.test.tsx`.
  *
  * ## ⚠️ Half of this file is a COMPILE-TIME assertion and vitest CANNOT read it
  *

packages/types/src/zod/data-display.zod.ts

@@ -77,8 +77,8 @@ export const AlertSchema = BaseSchema.extend({
     'children',
     'this alert node',
     '`alert` reads `children`, never `body` (READ SITE, measured with the TypeScript type checker: `packages/components/src/renderers/data-display/alert.tsx`). '
-    + '`body` was this node\'s only child-list key until objectui#6771 retired the spelling; an authored `body` now parses green through '
-    + '`.passthrough()` and renders an EMPTY element — no error, no warning. objectui#8284.',
+    + '`body` was this node\'s only child-list key until objectui#6771 retired the spelling — one concept, one spelling — so it is '
+    + 'refused here by name; write the content under `children`, the one child-list key. objectui#8284.',
   ),
 });
 
@@ -124,8 +124,8 @@ export const BadgeSchema = BaseSchema.extend({
     'children',
     'this badge node',
     '`badge` reads `children`, never `body` (READ SITE, measured with the TypeScript type checker: `packages/components/src/renderers/data-display/badge.tsx`). '
-    + '`body` was this node\'s only child-list key until objectui#6771 retired the spelling; an authored `body` now parses green through '
-    + '`.passthrough()` and renders an EMPTY element — no error, no warning. objectui#8284.',
+    + '`body` was this node\'s only child-list key until objectui#6771 retired the spelling — one concept, one spelling — so it is '
+    + 'refused here by name; write the content under `children`, the one child-list key. objectui#8284.',
   ),
 });