@@ -0,0 +1,9 @@
+---
+---
+
+Correct the Live E2E lane's claim about what a failing run leaves behind, and pin it to the
+config that decides it (objectui#8238). `playwright.live.config.ts` declares
+`reporter: [['list']]`, which writes to stdout only, so no run of this lane could ever produce
+the `playwright-report/` the workflow header, the job summary, the upload glob and
+`content/docs/guide/ci-cd-pipeline.md` all promised. CI, docs and tests only; no package is
+released by this change.
@@ -0,0 +1,8 @@
+---
+---
+
+Record the `#6150` ledger's own blind spot in its header: the census it pins is
+complete only for the un-cast `schema.KEY` spelling, and the cast-aware re-run
+(objectui#8327) finds 112 `(schema as any).KEY` reads across 23 files that it
+could never have seen. Comment only, in a test file; no package is released by
+this change.
@@ -0,0 +1,88 @@
+---
+'@object-ui/types': minor
+---
+
+**Breaking for already-authored metadata:** `ToastSchema.action` is RETIRED on both
+published faces (objectui#8338, ADR-0049 enforce-or-remove).
+
+**What was wrong.** The two faces shared nothing, and one of them had no JSON
+inhabitant at all. `packages/types/src/feedback.ts` declared
+`action?: { label: string; onClick: () => void }` — both members REQUIRED and
+`onClick` a function, so a JSON document could omit the key but never author it —
+while `packages/types/src/zod/feedback.zod.ts` declared
+`z.union([SchemaNodeSchema, z.array(SchemaNodeSchema)])`, a node or a list of nodes.
+Disjoint accept sets, one of them empty: the same document got a green `safeParse`
+and a `tsc` refusal, and no spelling satisfied both. `renderers/feedback/toast.tsx`
+read NEITHER — it reads `variant`, `title`, `description`, `duration`,
+`buttonVariant`, `className` and `buttonLabel`, and the file has exactly one
+`ComponentRegistry.register` call, so the zero is a reading and not a failed scan.
+
+This is objectui#6496's Direction 2, the half its `completed` close left behind:
+that card measured the key, prescribed enforce-or-remove, and was closed by PR #6542,
+which is Direction 1 only (declaring `buttonLabel` / `buttonVariant`). The sibling
+`onDismiss` was finished separately under objectui#6124 and sits four lines down as
+an ADR-0049 tombstone whose prose gives, word for word, the reason `action` should
+have gone with it. `action` survived only because that sweep was over TOP-LEVEL
+function-valued keys and this key's function is one level down.
+
+**The accept set moves in one direction only — on the MIRROR.** Measured by ablating
+the tombstone back to the base tree's two declarations (`3bc187b7f`) and re-running
+the pins, then restoring and proving both files byte-identical to `HEAD`:
+
+| `{ "type": "toast", "action": … }` | before | after |
+| --- | --- | --- |
+| `{ "type": "button", "label": "Undo" }` — a node | **accepted** | refused, `invalid_type` at `action` |
+| `[{ "type": "button" }]` — a list of nodes | **accepted** | refused, `invalid_type` at `action` |
+| `{ "label": "Undo", "onClick": … }` — the TS face's own shape, and `{}` | refused, `invalid_union` at `action` | refused, `invalid_type` at `action` |
+| `"Undo"`, `1`, `true`, `null`, `[]`, `["a", 1]` — bare primitives, and lists of them | **accepted** | refused, `invalid_type` at `action` |
+| `[[{ "type": "button" }]]` — a nested list | refused, `invalid_union` at `action` | refused, `invalid_type` at `action` |
+| the key omitted, or an explicit `undefined` | accepted | accepted |
+
+⭐ The primitive row is the LARGER half of this narrowing and is the easy one to miss: `SchemaNodeSchema` is `z.union([BaseSchemaCore, z.string(), z.number(), z.boolean(), z.null(), z.undefined()])` (`zod/base.zod.ts:84`), so `action: "Undo"` parsed green under the old union exactly as a node object did. Re-derived by re-forming the old spelling around the SHIPPED `SchemaNodeSchema` rather than a hand-rebuilt one, with a non-vacuity control: the two envelopes must disagree on the node case, else the probe is comparing a schema with itself.
+
+Nothing went from refused to accepted. The key stays DECLARED rather than deleted,
+because `BaseSchema` is `.passthrough()`: removing the member would KEEP an authored
+value unvalidated and silently inert instead of refusing it. `retirementTombstone()`
+writes ONE guidance string into BOTH author-facing channels — the parse-time issue
+message and `.describe()` — so they cannot drift.
+
+**The compile-time face, which a TypeScript author meets first.** A `.tsx` author
+COULD write `action: { label, onClick }` against the old declaration and it compiled;
+it is now `action?: never`, so the same literal fails type-check. That is the only
+place this narrowing removes something that ever worked, and it worked only in the
+programmatic channel — no JSON document could ever hold it, and no renderer read it.
+
+**Not `handlerKeyRefusal()`.** `action` is ⛔ not a handler key: it is a VALUE key
+whose NESTED member was a function. So it takes the ADR-0049 retirement helper
+(`invalid_type`, and `?: never` on the declaration) and not objectui#6124's
+named-refusal arm (`custom`, with a TS twin that stays callable for a runtime slot).
+The two helpers are pinned apart in `handler-keys-json-refusal-6124.test.ts`, whose
+census counts 45 runtime slots + 22 retired `on*` sites — a non-handler key has no
+seat in it.
+
+**Migration: there is NONE, and the capability was never fulfilled.** ⛔ Not "not yet
+supported" and ⛔ not a pointer at a future shape. objectui#6250's close moved the
+seven toast demos off an in-toast action entirely; an in-toast action button remains
+a capability expansion with zero runtime and would need its own card. Raise the toast
+from the node itself (`title`, `description`, `variant`, `duration`) and label its
+trigger with `buttonLabel` / `buttonVariant`. Measured across this repository: of the
+7 authored `toast` nodes in tracked JSON, **0** carry `action` (control: the same walk
+finds all 7 nodes), and no source file authors one **on a `toast` node**. ⚠️ Stated that narrowly on purpose: `app-shell/src/chrome/notificationToast.tsx:76` and `chrome/toast-helpers.ts:71` DO write `action: { label, onClick }` — for **sonner's** runtime API (`import { toast } from 'sonner'`), a different interface this key never fed. A flat "no source authors one" would be literally false.
+
+**⛔ `EmptySchema.action` is untouched** (objectui#7105 / PR #8330) — the same word on
+a sibling schema in the same file, with the deliberately opposite disposition, and its
+own comment block explains that it refuses the `{ label, onClick }` shape this key
+spelled. Two different interfaces, on purpose.
+
+**The parity ledgers drain with it,** every figure re-derived by
+`zod-mirror-parity.test.ts`'s own AST and mirror instruments rather than stepped by
+hand: `KnownDrift` 42 entries / 64 keys → **41 / 63** (the entry's whole content, so
+the entry went too — the ledger's first loss by RETIRING a key rather than by moving
+either face toward the other), and `WiderThanDeclared` 23 / 36 / 47 arms, split
+6 / 30 / 0 / 11 → **22 / 35 / 45**, split **6 / 29 / 0 / 10**. The pair itself stays
+registered, so `EXPECTED_MIRROR_PAIRS` does not move.
+
+Graded `minor`, not `patch`: a published accept set narrows. Not `major` per this
+repo's fixed-group convention — objectui's own breaking changes ship as `minor` and
+the group's major tracks `@objectstack` (AGENTS.md 版本号策略, enforced by
+`scripts/check-changeset-no-major.mjs`).
@@ -0,0 +1,12 @@
+---
+---
+
+Pin the i18n call-site gate's factory abstention counters on `main`
+(objectui#8423). The factory half of `scripts/check-i18n-call-site-keys.mjs`
+prints `factoryRowsNoEnKey`, `factoryUnjudgedRows` and its matching/compared
+pair on every run and held none of them to anything, so `factoryRowsNoEnKey`
+moved 5 to 0 when objectui#7887 retired the `timeline.relative.*` rows and
+nothing failed. The main-line case now mirrors the three lines the hand-rolled
+half already pins, keeping the existing `factoryTables > 25` and
+`factoryComparedRows > 500` lower bounds so the zeros stay readings rather than
+"the scan found nothing". Test only; no package is released by this change.
@@ -0,0 +1,27 @@
+---
+'@object-ui/app-shell': patch
+---
+
+Make `InspectorSelectField`'s `placeholder` reach the rendered trigger
+(objectui#8450). Every empty select in the metadata-admin designer drew a BLANK
+trigger instead of its hint.
+
+The field bridges a caller's `''` through an internal sentinel so a "— None —"
+option can exist at all (Radix `<Select.Item value="">` throws). That bridge also
+guaranteed the value handed to Radix was never `''` or `undefined` — the only two
+values for which Radix renders `SelectValue`'s placeholder — and a controlled
+value matching no `SelectItem` renders as nothing. So the declared `'—'` default
+was unreachable at all 45 call sites, none of which passes a placeholder of its
+own.
+
+The trigger now renders the placeholder itself, on the narrow state that means
+"nothing is selected": no value AND no option standing for none. Where the caller
+DOES offer a `''` row, that row is a selection and its label still wins,
+unchanged. A non-empty value matching no option also still renders blank —
+that is a stale value, not an empty one, and is out of this change's scope.
+
+Measured effect on the designer: 13 of the 45 call sites can be empty without
+offering a "none" row, and those now show `'—'` where they showed nothing — the
+flow-node config selects, the app-nav item type, the curated page-block props,
+the report dataset/chart-axis pickers and the action target/variant/mode/component
+pickers. The other 32 render exactly as before.
@@ -0,0 +1,20 @@
+---
+"@object-ui/core": patch
+---
+
+`ValueDataSource`: a stored value that is not a string now satisfies `not_contains`
+/ `$notContains` instead of failing the operator and its negation both.
+
+Both arms were written as `typeof value === 'string' && !value.includes(target)` — a
+type test standing in for the predicate. A row whose column held the number `5`
+failed `contains '5'` (correct: a number cannot contain a substring) and also failed
+`not_contains '5'` (wrong: for that same reason it does not contain it), so the row
+appeared in **no** filter answer, and the opposite filter — the one thing a user has
+to debug a missing row with — was silent too. Measured on a mixed fixture,
+`not_contains '5'` returned 1 of 8 rows where it now returns 7.
+
+The arms are now the exact complement of `contains` / `$contains`. The positive text
+operators are unchanged and keep their type gate: that is the other half of
+objectstack#14079 (maintainer ruling 2026-09-05, option A) — a non-string never
+satisfies a positive text operator and always satisfies `$notContains`, so
+complementarity holds on every face.
@@ -0,0 +1,15 @@
+---
+---
+
+Test-only: `packages/types/src/__tests__/zod-mirror-parity.test.ts`. It sits under
+`src/`, so the presence gate counts it, but nothing published moves: the package's
+build `tsconfig.json` excludes `**/__tests__/**` and its `files` list is `dist` only,
+so the file never reaches a consumer. Declared as releasing nothing.
+
+- objectui#8458: the WIDER ledger's objectui#7760 movement sentence said "16 pairs",
+  the one figure in that paragraph no `headerFigures` pin reached. Re-derived as 18
+  (13 emptied + 5 reduced), defined, and pinned.
+- objectui#8248: objectui#8222's exclusion note called the seed decomposition a LIVE
+  figure, which objectui#8243 made historical. The guidance clause is amended per the
+  maintainer ruling (decision batch #71); the two measurement-record clauses are
+  untouched, and the record-vs-guidance rule is now written once in the file header.
@@ -0,0 +1,44 @@
+---
+'@object-ui/plugin-detail': minor
+---
+
+**An empty array is no longer a cell value on the record page (objectui#8474).**
+
+`hasCellValue` — THE definition of emptiness for `record:details` (objectui#8376,
+widened to the package by objectui#8394, extracted to `emptiness.ts` by
+objectui#8457) — opened with
+`if (value !== null && typeof value === 'object') return true;`. `typeof []` is
+`'object'`, so an **empty array was a value** and nothing further was asked of
+it.
+
+The object half's reasoning is sound and stays: an object value is handed to a
+type-aware renderer that knows how to draw it. But every example it reasons
+about is a *populated* object, and for `[]` the renderer has nothing to draw —
+`SelectCellRenderer` tests `value == null || value === ''`, which `[]` passes,
+then maps it over zero entries. The result was the blank cell the em-dash exists
+to prevent, produced by the function that exists to prevent it, plus the full
+objectui#8376 triple: the row escaped `emptyCount`, `canCopy` offered to copy it,
+and because `shouldAutoHideEmpty` needs only `filledCount > 0` a section whose
+one non-null value was `[]` armed auto-hide by itself and buried every genuinely
+empty row around it.
+
+**The change.** One arm inside the object branch:
+`if (Array.isArray(value) && value.length === 0) return false;`. All five readers
+of the shared authority get the corrected answer — `DetailSection`'s affordance,
+`emptyCount` and `canCopy`; `HeaderHighlight`'s highlight strip; `DetailView`'s
+summary chips; `HistoryTimeline`'s diff placeholder; `RecordMetaFooter`'s actor
+test.
+
+**What did NOT move, and why it is a measurement.** `{}` is still a value: on
+`json`, `object` and `location` fields it draws the literal `{}` through
+`JsonCellRenderer`, so there is no blank cell to fix. The shape that would have
+swept it in — `Object.keys(value).length === 0` — is also true of a `Date`, a
+populated `Map`, a populated `Set` and a getter-backed class instance, which
+would be a false-empty on values that render.
+
+**Declared cost.** A `json`-family field holding `[]` previously rendered the
+literal two-character text `[]`; it now draws the `No value` placeholder. That is
+intended and pinned.
+
+`RelatedList.isValueEmpty` is untouched: it already drew this line, and the
+shared authority moved toward it — never the reverse.
@@ -0,0 +1,39 @@
+---
+'@object-ui/fields': patch
+---
+
+An empty array no longer makes a cell renderer fabricate a value (objectui#8490).
+
+objectui#8481 fixed the shared read renderers whose output for `[]` was *blank*. These
+are the other half of that census: the renderers whose output for `[]` was not blank
+but **wrong** — a value the record does not hold. Measured by rendering on `e411c3e58`:
+
+| field types | renderer | rendered for `[]` before |
+|---|---|---|
+| `boolean`, `toggle` | `BooleanCellRenderer` | a **checked**, disabled checkbox (`aria-checked` true); a completion field drew the green "Completed" indicator |
+| `number`, `slider`, `rating` | `NumberCellRenderer` | the digit `0` |
+| `currency` | `CurrencyCellRenderer` | the digit `0` |
+| `percent`, `progress` | `PercentCellRenderer` | a 0% progress bar with `aria-valuenow` of 0 |
+| `email` | `EmailCellRenderer` | a live anchor whose `href` was `mailto:` with no address, plus a copy button |
+| `url` | `UrlCellRenderer` | a live `_blank` anchor with an empty `href` |
+| `phone` | `PhoneCellRenderer` | a live anchor whose `href` was `tel:` with no number |
+| `color` | `ColorSwatchCellRenderer` | a bordered swatch box with no colour and an empty hex span |
+| `date` | `DateCellRenderer` | a hand-rolled em-dash with no accessible name (not the shared placeholder) |
+
+All of them now render the shared `EmptyValue` affordance — the muted em-dash with a
+`No value` accessible name — exactly as they already did for `null`.
+
+**The ruling is per renderer, not one predicate.** A boolean cell is not a text cell:
+`[]` holds no boolean, so the column holds *no value*, not `false` — a real `false` is
+still an unchecked box, and only `[]` moves (scalar truthiness coercions are untouched).
+The number family's `0` was `Number('')`, a coercion artefact rather than a stored zero,
+so its guard is now on the coerced *text* — which **deliberately also sweeps a stored
+`''`** (the same fabrication one input shape over); a real stored `0` still prints. The
+link family draws no anchor when there is nothing to link to; `color` draws no swatch
+without a colour string; `date` reaches the shared affordance instead of `formatDate`'s
+private dash.
+
+**Deliberately narrow.** `{}` is untouched everywhere, `json` still draws the array
+literal and `file` still states `0 files` (the objectui#8481 fence), and whether a
+non-boolean *scalar* in a boolean column should surface as a coercion error is a separate
+question this change does not answer.
@@ -0,0 +1,37 @@
+---
+'@object-ui/plugin-detail': patch
+'@object-ui/types': patch
+---
+
+`record:details` — implement the `sections[].group` reference form, stop one
+section taking its siblings down, and resolve enumerated field labels through
+the object's declared `label` (objectui#8497).
+
+**An authored `{ group: 'parties' }` used to blank the whole record page.**
+`@objectstack/spec` 17.3.0 declares the group-reference form in full
+(`RecordDetailsProps.sections[].group`, objectstack#13855, ADR-0085 §5), but
+this renderer read the key nowhere, and its own props description told authors
+so. That was not a no-op: a section carrying `group` carries no `fields`,
+`DetailView` mapped every section through `s.fields` unguarded, and `flatMap`
+kept the resulting `undefined` as an element whose `.name` the next line read.
+The throw arrived with `e99770841` (2026-05-01), months before the key it
+fires on was declared, so it was never a deliberate refusal — the declaration
+was right and the runtime had never honoured it. `group` is now resolved
+through `deriveFieldGroupDetailSections`, the same adapter the synthesized
+default record page uses, so the group-referenced body and an equivalent
+enumerated body render identically. A `group` naming no declared group renders
+nothing and says so on the console (`@objectstack/lint` reports it as
+`page-section-group-unknown`); it is no longer silent, and it no longer throws.
+
+**A malformed section now degrades to that section.** Two layers, because one
+is not enough: the field-collecting reads are tolerant of a section with no
+`fields` array (the crash above fired in a `useMemo` above the section loop,
+where no boundary could reach it), and each section is additionally wrapped in
+the per-section error boundary, so a section that throws while rendering shows
+the failure in its own place instead of replacing the whole component.
+
+**Enumerated field labels reach the object's declared `label`.** The ladder is
+`i18n bundle key -> the object's declared label -> the field name`. Rung 2 was
+missing, so an app with no translations configured showed raw snake_case field
+names on every detail page. An authored entry `label` still wins, and an app
+with a full bundle is unchanged.