@@ -11,11 +11,12 @@ Clause-②: no (narrowing)
 
 **BREAKING**: an accept-set narrowing on a published authoring surface (`objectstack validate`'s `expression-invalid` rule).
 
-**Why.** Flow CEL evaluates in one scope per run: the flow's variables spread to top level, the trigger record's fields flattened beside them, and `record`, `previous`, `vars` and `current_user` bound by the engine. Formulas, row-level security and the client also bind the run's user as `user`, `ctx.user` and `os.user`; flow CEL does not, because a top-level `user` would collide with a variable or a lookup field of that name. So an author who wrote the alias by habit shipped a flow that faulted at its first evaluation, and nothing warned.
+**Why.** Flow CEL evaluates in one scope per run: the flow's variables spread to top level, the trigger record's fields flattened beside them, `record` bound to the record the run was handed (where an entrance hands it one), and `previous`, `vars` and `current_user` bound by the engine. Formulas, row-level security and the client also bind the run's user as `user`, `ctx.user` and `os.user`; flow CEL does not, because a top-level `user` would collide with a variable or a lookup field of that name. So an author who wrote the alias by habit shipped a flow that faulted at its first evaluation, and nothing warned.
 
 **What is refused.** At every flow CEL site the build door reads — a node's `config.condition` (the start node's trigger gate included), an edge `condition`, a `decision` branch `expression`, a screen field `visibleWhen`, and the CEL value envelopes of an `assignment` and of a `create_record` / `update_record` `fields` map — a root that is none of:
 
-- the engine's roots: `record`, `previous`, `vars`, `current_user`;
+- the engine's roots: `previous`, `vars`, `current_user`;
+- `record`, where an entrance the stack declares hands the flow a record — a record trigger (a `record-*` `triggerType`), a time-relative sweep, the inbound hook, a `type: 'flow'` action, a `map` node with an `itemObject`, or a `subflow` / `map` parent handed one — or where the flow binds a variable named `record`;
 - a name the flow binds: a declared variable, an `outputVariable`, an `iteratorVariable` (a `loop`'s default `item` included), an `indexVariable`, an `errorVariable`, an `assignment` target, a node id, a screen field's `name`;
 - a field (or a registry-injected column) of any object the stack declares — the run flattens whichever record its entrance hands it (an object-less flow action hands only the row `id`).
 
@@ -45,7 +46,7 @@ A `$` name has no CEL spelling (`$runId == "x"` does not parse), so the `$` root
 
 **The one-line fix: write `current_user` for the run user, and bind or correct every other root the refusal names.**
 
-**Who is affected, measured.** The last published spec line, 17.x, judged a flow expression's syntax only, so it accepts every refused source. Measured with the judge over every flow CEL site: the example apps (`app-crm`, `app-todo`, `app-multi-package`, `app-showcase`: 35 flows, 45 sites), `packages/platform-objects` (no flows) and `objectstack-ai/hotcrm` at `f0afcbda07` (32 flows, 65 sites) — 0 refusals; every root those flows read is an engine root, a binding of the flow's own or a field of its trigger object. One flow of the 67 is left unjudged by an entrance (`showcase_inbound_task_webhook`, an `api` trigger). Deployed metadata and other repositories were not measured.
+**Who is affected, measured.** The last published spec line, 17.x, judged a flow expression's syntax only, so it accepts every refused source. Measured with the judge over every flow CEL site: the example apps (`app-crm`, `app-todo`, `app-multi-package`, `app-showcase`: 35 flows, 45 sites), `packages/platform-objects` (no flows) and `objectstack-ai/hotcrm` at `f0afcbda07` (32 flows, 65 sites) — 0 refusals; every root those flows read is an engine root, `record` on a flow an entrance hands a record, a binding of the flow's own or a field of its trigger object. One flow of the 67 is left unjudged by an entrance (`showcase_inbound_task_webhook`, an `api` trigger). Deployed metadata and other repositories were not measured.
 
 ### The kit
 
 
@@ -35,4 +35,4 @@ In a flow whose run holds no record, read a variable by its name:
 - this repository at `0ec4268972` (`examples/**`, `packages/platform-objects`, `packages/qa/dogfood`): 64 flows, 51 CEL slots, no flow in `packages/platform-objects`; the 2 flows that read `record` are an `api` hook flow and a record-change flow;
 - `objectstack-ai/hotcrm` at `f0afcbda07` (`src/`, `test/`): 45 flows, 57 CEL slots; the 12 flows that read `record` are all record-change flows.
 
-Deployed metadata and other repositories were not measured. `objectstack validate` does not refuse such a `record` read yet; the run fails it, naming the root and the source.
+Deployed metadata and other repositories were not measured. `objectstack validate` refuses such a `record` read too, naming the root (see `@objectstack/lint`'s note).
@@ -0,0 +1,31 @@
+---
+'@objectstack/lint': major
+---
+
+A flow CEL expression that reads `record` on a flow no entrance hands a record is refused at `objectstack validate`. Since #22642 the run binds `record` only to the record it was handed, so such a read fails the run with `Unknown variable: record`, and the build door still counted `record` as bound on every run and passed it.
+
+Clause-②: no (narrowing)
+
+<!-- adr-0087: not-required (already-registered flow-cel-unbound-root-refused, flow-cel-record-variables-alias-retired) A record read on a flow no entrance hands a record is a root the flow does not bind, which is the surface of the D3 entry flow-cel-unbound-root-refused (protocol 18, not yet released); this diff corrects that entry's reason to say so. The runtime half is the D3 entry flow-cel-record-variables-alias-retired (#22642). Both exist at the merge base and neither has shipped, so this narrowing adds no entry. -->
+
+**BREAKING**: an accept-set narrowing on a published authoring surface (`objectstack validate`'s `expression-invalid` rule).
+
+**What is refused.** `record` is no longer one of the roots the build door binds on every run (`previous`, `vars` and `current_user` still are). It is bound where:
+
+- an entrance the stack declares hands the flow a record: a record trigger (start `config.triggerType` is a `record-*` token), a time-relative sweep (start `config.timeRelative`), the inbound hook (trigger kind `api`), a `type: 'flow'` action that targets the flow (on any object, or on none), a `map` node with a `config.itemObject` that names the flow, or a `subflow` / `map` parent that is itself handed a record;
+- or the flow binds a variable named `record` itself (a declared variable or an `assignment` target).
+
+A start `config.objectName` alone is no entrance: with no `record-*` trigger nothing hands the run a row. A parent's own `record` variable is not handed on either: a child gets its parent's context, not its variables. Everywhere else `record.X` is refused as `expression-invalid`, `error`, with the remedy below.
+
+**Not judged.** The stand-downs of the unbound-root rule are unchanged. At the runtime publish gate this judgment still stands down: its per-write snapshot carries no actions and no other flows, so it cannot see an entrance, and a flow write there is not refused for `record` until that snapshot carries them.
+
+## FROM → TO
+
+| you wrote | write instead |
+|:--|:--|
+| `record.assignee` on a flow no entrance hands a record, meaning a flow variable `assignee` | `assignee`, or `vars.assignee` |
+| `record.status == "open"` on a flow meant to run when a row is written | the trigger that hands it the row: start `config: { objectName: 'acct', triggerType: 'record-after-update' }` |
+
+**The one-line fix: read a flow variable by its name or through `vars`, and read `record` only on a flow an entrance hands one.**
+
+**Who is affected, measured.** With this judge over every flow CEL site: the example apps (`app-crm`, `app-todo`, `app-multi-package`, `app-showcase`: 35 flows, 14 that read `record`), `packages/platform-objects` (no flows) and `objectstack-ai/hotcrm` at `1d7148b` (32 flows, 14 that read `record`): 0 newly refused. Every flow that reads `record` there is a record-change, time-relative or inbound-hook flow. Deployed metadata and other repositories were not measured.