@@ -253,7 +253,7 @@ This replaces #30806 S3's name rewrite. Nothing is renamed, so nothing is rewrit
 ### D9 — What retires, and what stays
 
 - **Narrows per type, in that type's own trailing child, last:** phase 1's one-holder refusal (`BARE_NAME_ONE_HOLDER_TYPES`). It narrows from one package holder per environment to one holder per namespace (D2), only after that type's indexes, doors and stored references all carry the package. Until then it stays in force, so every intermediate `main` behaves as today plus optional package coordinates (the card's decision 6 and red line 3).
-- **Withdrawn with this record:** #30806 phase 3 (S1 spec refusal, S5 HotCRM renames). **Retired by child P:** phase 2's `naming/namespace-prefix` bare-name warning for buckets 2–4 (the intra-package duplicate warning stays), and the authoring skills' and templates' instruction to prefix non-object names (#30805's text). Studio prefixes object names only (`prefixObjectName` in `ui/packages/app-shell/src/views/studio-design/packages-io.ts`), so child U retires no prefix. **Narrowed to bucket 1, outside this epic:** the scenario scorer's `bare-name` Build check and its build-prompt instruction to prefix every item. Today `gradeItemNames` in `packages/qa/scenarios/src/score/scenario-naming.ts` fails every non-object item named without `NS_`, and `src/runner/prompts.ts` tells the agent to prefix. Both are narrowed to bucket 1 types (object, field, view and the rest keep the literal prefix), so a bare name of a bucket 2–4 type passes. The cloud#3600 epic seat raised it as a dependent of ruling C ([`6726694021`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6726694021) on #31527). That seat files it to the lane that owns the scorer, and it lands after P, so the skills and the scorer teach the same rule. It is a scoring rule, not a door: it narrows nothing and loosens no resolution.
+- **Withdrawn with this record:** #30806 phase 3 (S1 spec refusal, S5 HotCRM renames). **Retired by child P:** phase 2's `naming/namespace-prefix` bare-name warning for buckets 2–4 (the intra-package duplicate warning stays), and the authoring skills' and templates' instruction to prefix bucket 2–4 names (#30805's text). Studio prefixes object names only (`prefixObjectName` in `ui/packages/app-shell/src/views/studio-design/packages-io.ts`), so child U retires no prefix. **Narrowed to bucket 1 and positions, outside this epic:** the scenario scorer's `bare-name` Build check and its build-prompt instruction to prefix every item. Today `gradeItemNames` in `packages/qa/scenarios/src/score/scenario-naming.ts` fails every non-object item named without `NS_`, and `src/runner/prompts.ts` tells the agent to prefix. Both are narrowed to bucket 1 types and positions (D7: one environment-wide slot per name, so a bare position still collides at install). Of the collections the scorer walks, view, doc and capability keep the check, the capability by its `NS.SHORT` spelling; an api item's name is bucket 2. So a bare name of a bucket 2–4 type passes. The cloud#3600 epic seat raised it as a dependent of ruling C ([`6726694021`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6726694021) on #31527). That seat files it to the lane that owns the scorer, and it lands with or before P. It is a scoring rule, not a door: it narrows nothing and loosens no resolution.
 - **Stays:** ADR-0129 and the object rule, unchanged; the one-holder rule for platform-requested bare names (D3) and, as ruled (Q2 = A), for positions (D7); the capability's `NS.SHORT` spelling (bucket 1); the ADR-0048 §3.2 namespace install gate, which stays and gains the G1 and K refusals.
 - **Names already prefixed stay as they are.** The examples renamed in phase 2 (#30927) and HotCRM's already prefixed names (email templates) are just names, and ⛔ nothing un-renames them.
 
@@ -320,13 +320,13 @@ Each child PR proves it is releasable with four readings:
 | S3 | automation columns: R7, R8, R9, R10, R11, R14, R20 | server | service-automation, plugin-approvals, service-job, plugin-email | no | no |
 | S4 | the ambiguous-set upgrade report, the listing of package references whose only holder is the environment (D3 (a)), and the reviewed bind command (D10), with its ADR-0087 migration entry | server | cli (`os migrate`), metadata-protocol | no | no |
 | B1…B13 | **backstop narrowing, one child per type**, last: the one-holder refusal narrows from environment to namespace (D2), after that type's E, G and S children land | server, after each type's E/G/S | objectql `bare-name-one-holder.ts`, `security-catalog-namespace.ts` (permission sets) | no | **yes** for co-owners of one namespace (measured first-party: 0); a widening for everyone else |
-| A | ADR status lines (ADR-0048 addendum, ADR-0129 cross-reference) | epic seat | `docs/adr/` | no | no |
+| A | ADR status lines (ADR-0048 addendum, and §5's `[done]` lint bullet and §3.4's lint sentence, which go stale after P and the phase-3 withdrawal; ADR-0129 cross-reference) | epic seat | `docs/adr/` | no | no |
 
 **Count:** D, R, N, K, P, A, U and C (8), 8 E, 3 G and 4 S make 23 build children, N and K included. That is one above the evaluation thread's estimate of 18–22, because of K, which carries the Q3 ruling.
 
 The B children are one per covered type, 13 in all (phase 1's 12 types plus permission sets), and small. Several types of one family can ride one PR where their E, G and S children have all landed. The trailing B children are new, and come from red line 3.
 
-**Hand-off outside this epic** (no child, so the count is unchanged): the scenario scorer's `bare-name` Build check and build prompt narrowed to bucket 1 (D9), filed by the cloud#3600 epic seat ([`6726694021`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6726694021)) to the lane that owns the scorer, after P.
+**Hand-off outside this epic** (no child, so the count is unchanged): the scenario scorer's `bare-name` Build check and build prompt narrowed to bucket 1 types and positions (D9), filed by the cloud#3600 epic seat ([`6726694021`](https://github.com/objectstack-ai/objectstack/issues/31527#issuecomment-6726694021)) to the lane that owns the scorer, with or before P.
 
 **Critical path, at fleet pace** (about 2–6 hours per medium PR, lanes in parallel):