docs(skills): correct two false claims and add three routing bullets in skills/README.md (#14795)

Settle `skills/README.md`'s CEL conventions and cross-skill routing list
against source.

1. `:87` — "Legacy `OLD` / `NEW` evaluate to `null` since M9.5" was false in
   both halves. `M9.5` is anchored nowhere in the repo, and `OLD` / `NEW` are
   not values that evaluate to `null`: they are undeclared identifiers.
   `SCOPE_ROOTS` (packages/formula/src/cel-engine.ts:95) is the published
   closed set of record-scope roots and contains neither, so
   `firstUndeclaredReference` (:186) reports them as bare undeclared
   references — the class `@objectstack/lint` raises as an error, not a
   silent null. Replaced with the true clause plus the migration objectstack-
   formula's own table prescribes (`OLD.x` -> `previous.x`, `NEW.x` ->
   `record.x`).

2. `:105` — "AI tool params" is not a CEL site. `packages/spec/src/ai/tool.zod.ts`
   carries no expression field of any kind, and the AI domain's only expression
   site is `model-registry.zod.ts:121-122`, which is `TemplateExpressionInputSchema`
   — the `template` dialect (`{{var}}` interpolation), not CEL. Dropped the item;
   no true replacement exists, because the AI domain has no CEL predicate.

3. `:94-110` — three routing bullets added for measured load-failure modes:
   labels to objectstack-i18n, protocol-major moves to objectstack-upgrade,
   and the consuming UI to the objectui skills catalog.

Items 1-2 net exactly zero bytes; item 3 costs 75 tokens, at the size ruling.
The generated region and the anatomy tree are untouched.


Claude-Session: https://claude.ai/code/session_01LraLgQVGq8egUwfYZpbYt1

Co-authored-by: Claude <noreply@anthropic.com>

Author: os-litant

skills/README.md

@@ -84,7 +84,8 @@ carries `SKILL.md` alone — `gen:skill-refs` only visits skills listed in its
 - **Short object names** (`account`, `task`); no `namespace`, no `tableName`.
 - **CEL for all expressions** — predicates, conditions, schedules. Use the
   `F\`\``, `P\`\``, `cel\`\``, `cron\`\``, `tmpl\`\`` tagged templates from
-  `@objectstack/spec`. Legacy `OLD` / `NEW` evaluate to `null` since M9.5.
+  `@objectstack/spec`. Legacy `OLD` / `NEW` are not CEL — use
+  `previous.` / `record.`.
 - **v5.0 vocabulary** — runtime workspace is `environment`, not `project`.
 - **Singular metadata type names** (`agent`, `view`, `flow`, …); REST resource
   collections are plural (`/api/v1/ai/agents`).
@@ -102,10 +103,14 @@ A few common decision points where the right skill isn't obvious:
   **automation** (screen flows). Static record / list / dashboard surfaces are
   **ui**.
 - **Any CEL expression** — load **objectstack-formula** alongside the host
-  skill (data validations, automation guards, UI visibility, AI tool params).
+  skill (data validations, automation guards, UI visibility).
 - **Kernel / plugin events vs. data lifecycle** — `PluginContext` lifecycle and
   `EventBus` belong to **objectstack-platform**; record-level hooks belong to
   **objectstack-data**.
+- **Labels vs. bundles** — a `label` you want translated is **objectstack-i18n**.
+- **Upgrade vs. platform** — a protocol-major move is **objectstack-upgrade**.
+- **Rendering this metadata** — the consuming UI is
+  [objectui `skills/objectui/`](https://github.com/objectstack-ai/objectui).
 
 ---