diff --git a/.changeset/30824-formula-no-group-key.md b/.changeset/30824-formula-no-group-key.md
new file mode 100644
index 00000000000..494149e6917
--- /dev/null
+++ b/.changeset/30824-formula-no-group-key.md
@@ -0,0 +1,13 @@
+---
+'@objectstack/skills': patch
+---
+
+docs(skills): the `objectstack-data` field-types formula row says a formula is no dataset or cube dimension (group key) either
+
+Clause-②: no
+Metadata-protocol: none
+
+- **Before.** The `formula` row of `skills/objectstack-data/rules/field-types.md` said a formula is virtual and not aggregatable, and that a dataset measure needs a stored field. It said nothing about grouping by one, so an AI author reading the row had nothing telling it not to declare a dataset or `analyticsCubes` dimension over a formula.
+- **Now.** The row says a formula is neither aggregatable nor a dataset or cube dimension (group key), and that both need a stored field — what the platform refuses today: the authoring rule `dimension-virtual-field-refused`, the dataset compile's `400 DATASET_INVALID`, and every analytics query grouping by one, `400 INVALID_FIELD`.
+- `skills/objectstack-ui/rules/dashboards.md` (Dataset-Bound Widgets, the row's pointer): the multi-value breakdown bullet now names a `formula` beside the multi-value types as refused for a dataset or cube dimension and `groupBy`, so the reader lands on the dimension refusal too.
+- No schema, key or behaviour changes; the skill text only states refusals that already ship.
diff --git a/scripts/check-skills-token-ratchet.mjs b/scripts/check-skills-token-ratchet.mjs
index 2d2a9b19146..298cc761aa7 100644
--- a/scripts/check-skills-token-ratchet.mjs
+++ b/scripts/check-skills-token-ratchet.mjs
@@ -697,7 +697,10 @@ export const CEILINGS = new Map([
   // headroom, ceiling +49, pinned AT the measurement. Standing ruling (2026-10-27, tap A,
   // verbatim and untranslated): 「A:先削重复,余额按实测提高」; recognised path per ruling
   // record 6513306068.
-  ['skills/objectstack-data/rules/field-types.md', 3207],
+  // 3207 -> 3199 (#30824): RE-LOCK at the landed count. The formula row now also says
+  // a formula is no dataset or cube dimension (group key), paid by the Formula example's
+  // remaining `// see the Calculated table` pointer; 12794 bytes -> 3199 tokens, zero headroom.
+  ['skills/objectstack-data/rules/field-types.md', 3199],
   ['skills/objectstack-data/rules/indexing.md', 3241],
   ['skills/objectstack-data/rules/lifecycle.md', 1590],
   ['skills/objectstack-data/rules/naming.md', 773],
@@ -853,7 +856,12 @@ export const CEILINGS = new Map([
   // The file is 25047 bytes -> 6262 tokens, zero headroom, ceiling +41, pinned AT the
   // measurement. Standing ruling (2026-10-27, tap A, verbatim and untranslated):
   // 「A:先削重复,余额按实测提高」; recognised path per ruling record 6513306068.
-  ['skills/objectstack-ui/rules/dashboards.md', 6262],
+  // 6262 -> 6232 (#30824): RE-LOCK at the landed count. The multi-value breakdown bullet
+  // adds a `formula` as a refused dimension and `groupBy` (a dataset dimension over one is
+  // `DATASET_INVALID` at compile); paid by the `compareTo` paragraph's restated "no second
+  // widget-side vocabulary" sentence and the `DatasetSchema` source pointer that
+  // `references/_index.md` already carries; 24926 bytes -> 6232 tokens, zero headroom.
+  ['skills/objectstack-ui/rules/dashboards.md', 6232],
   // 3011 -> 3141 (#19715, card item 4): a 5-line table of contents at the top,
   // anchors to the file's existing headings, nothing else changed. +132 tokens,
   // of which the row's 2 tokens of headroom absorb 2; the CEILING moves 130.
diff --git a/skills/objectstack-data/rules/field-types.md b/skills/objectstack-data/rules/field-types.md
index 36a333d3bc0..626cbff0d11 100644
--- a/skills/objectstack-data/rules/field-types.md
+++ b/skills/objectstack-data/rules/field-types.md
@@ -128,7 +128,7 @@ Stored as JSON on the parent row — no separate table / FK:
 
 | Type | When to Use | Config |
 |:-----|:------------|:-------|
-| `formula` | Computed from an expression referencing other fields — virtual: not aggregatable, a dataset measure needs a stored field (objectstack-ui → Dashboards → Dataset-Bound Widgets) | `expression` (CEL, `record.` prefixes), `returnType` (`'number' \| 'text' \| 'boolean' \| 'date' \| 'currency'`); `'currency'` carries its own `currencyConfig` |
+| `formula` | Computed from an expression referencing other fields — virtual: neither aggregatable nor a dataset or cube dimension (group key); both need a stored field (objectstack-ui → Dashboards → Dataset-Bound Widgets) | `expression` (CEL, `record.` prefixes), `returnType` (`'number' \| 'text' \| 'boolean' \| 'date' \| 'currency'`); `'currency'` carries its own `currencyConfig` |
 | `summary` | Roll-up aggregation from child records | `summaryOperations` ({ object, field? (omit for `count`), function, relationshipField?, filter? }) |
 | `autonumber` | Auto-incrementing display format ({0000} counter + optional date / {field} tokens, resets per scope) | `format` (shorthand) / `autonumberFormat` (canonical) — e.g., `"CASE-{0000}"`, `"AD{YYYYMMDD}{0000}"` |
 
@@ -201,7 +201,7 @@ import { F } from '@objectstack/spec';
 const total = {
   type: 'formula',
   expression: F`record.amount * record.tax_rate`,
-  returnType: 'number',   // see the Calculated table
+  returnType: 'number',
 };
 ```
 
diff --git a/skills/objectstack-ui/rules/dashboards.md b/skills/objectstack-ui/rules/dashboards.md
index d7165d45aaf..c6f898a9b97 100644
--- a/skills/objectstack-ui/rules/dashboards.md
+++ b/skills/objectstack-ui/rules/dashboards.md
@@ -38,8 +38,7 @@ and select named `dimensions` + `values`**; the dataset owns the base object,
 allowed joins, intrinsic filter, dimensions, and certified measures. A widget
 **requires `dataset` + `values`**; the closed schema rejects `object` /
 `categoryField` / `valueField` / `aggregate` by name, and one lacking `dataset`
-fails `os validate`. The dataset shape is
-`DatasetSchema` — see `node_modules/@objectstack/spec/src/ui/dataset.zod.ts`.
+fails `os validate`.
 
 A widget's presentation-scope `filter` flows into the query as the runtime
 filter; keep `filter` on the widget when binding a dataset.
@@ -102,8 +101,9 @@ Standardized answers to the recurring ambiguous cases:
   `beforeInsert`/`beforeUpdate` hook writes, then `{ name: 'days_sum', aggregate: 'sum', field: 'days_count' }`.
 - **Breakdown by a multi-value field.** `multiselect`/`checkboxes`/`tags` and
   any `multiple: true` field is refused as a dataset or cube dimension and as
-  `groupBy` (`INVALID_FIELD`; no per-member buckets). One widget per member,
-  filtered `{ tags: { $contains: 'urgent' } }`, or a single-value axis.
+  `groupBy` (`INVALID_FIELD`; no per-member buckets); so is a `formula` (at a
+  dataset's compile, `DATASET_INVALID`). One widget per member, filtered
+  `{ tags: { $contains: 'urgent' } }`, or a single-value axis.
 - **Filter by a parent's attribute** → a **dimension** (guaranteed to join); a
   lookup-path *filter* is not a reliable analytics-path construct.
 - **A dashboard filter driving several charts** → **not** a dataset: `globalFilters`
@@ -228,7 +228,7 @@ no second `filter` is required.
 
 `compareTo` is `{ kind, dimension? }` — the same shape the analytics executor
 reads (`DatasetSelection.compareTo`), so what a widget declares is exactly what
-runs. There is no second widget-side vocabulary.
+runs.
 
 | Key | Value | Behaviour |
 |:--|:--|:--|