feat(plugin-security,platform-objects): grant forms pick a position or permission set by name from the catalog (#15204 B2)

[objectstack-fleet]

Part of #15204
Clause-②: no

B2, objectstack half (objectui#7611 names it). Setup's user page grants a position or a permission set by picking its NAME from the security catalog, instead of an add picker that listed sys_position / sys_permission_set rows (ADR-0131 D3/D4). This PR is one of the pair that gates #22900 (the boot stops writing sys_permission_set rows): once #22900 lands, a fresh database holds no such row, and the old pickers would offer nothing.

What changes

Surface	Before	Now
sys-user.page.ts Positions / Permission Sets lists	"Assign position" / "Grant permission set" add pickers over sys_position / sys_permission_set rows, beside New	the two add blocks are deleted. New opens the grant form. Remove is the host's row delete (RelatedRecordActionsBridge), unchanged. The Business Units list keeps its add
sys_user_position.position	a text input (type text)	widget: 'position-ref': the form lists GET /api/v1/meta/position and stores the chosen name
sys_user_permission_set.permission_set	a text input	widget: 'permission-ref': the form lists GET /api/v1/meta/permission and stores the chosen name
sys_user_permission_set.permission_set_id ("Permission Set (legacy id)")	an editable lookup on the grant form, though it grants nothing	hidden: true, so it is off the create and edit forms (maintainer, 2026-10-12, 「legacy id 字段隐藏，同意」). The column stays until ADR-0131 C7b drops its data

Until the console pin moves, the two fields render as text inputs (the type renderer: a text box for the name). position-ref / permission-ref ship in objectui#12175 (PR-5, merged as 088e0f705520), and this repo's .objectui-sha still pins 4997995022f8. An unregistered widget falls back to the text renderer, so nothing breaks. The pin follows #22874: .objectui-sha is that PR's single-writer path, and objectui from e5311165 on reads allowEnvironmentOverlay only, which main's server sends once #22874 lands. A follow-up PR then moves the pin to 088e0f705 (or later), re-measuring only that hop. main is unreleased, so no released build shows the gap.

This branch briefly carried that pin move (e486abb9c), reverted by 7de0f00c9. The net diff is phase 2 only: 9 files, tree-identical to d4c22c8fe.

Measured before the build (phase 1, #15204 6116275825)

On a booted showcase, as the admin:

• both grant lists already offered New;
• with the two add blocks deleted, every row still offers Delete, and a delete was exercised end to end (UI 4 → 3 rows; the API shows the row gone).

The renderer's onRowDelete prefers the host's onDelete; the add-based remove is only its fallback when no host bridge is mounted.

Pins

• packages/qa/dogfood/test/grant-name-pickers.dogfood.test.ts reads the SERVED GET /meta/object/<name> on a booted showcase:
  • position carries position-ref;
  • permission_set carries permission-ref;
  • permission_set_id is hidden, with permission_set not hidden as the control.
  • Ablation: stripping the two hints and hidden from the objects, then rebuilding, turned all 3 cases red (expected undefined to be 'position-ref', …); restoring turned them green.
• sys-user-permission-set-window-columns.test.ts asserts that neither grant list carries add, with the Business Units list's add as the control. Ablation: re-adding an add to the Permission Sets list turned that case red.
• format.server-ready-dev-admin-audience.test.ts: the CLI banner's "grant it a permission set" pin is re-anchored on the Permission Sets list (its New), since the "Grant permission set" label is gone.
• platform-page-i18n-parity.test.ts: the anti-vacuity floor of inline locale maps is 44 (was 45), because the two add labels went with their blocks.

Browser check against the widget (objectui#12175 head 8237a1cd886d, built locally before it merged; the merge 088e0f705520 carries the same widget files)

• The Position and Permission Set fields render as registry comboboxes (registry-ref-position / registry-ref-permission), listing 16 positions and 17 permission sets.
• Picking auditor and mcp_agent_data_read, then pressing Create, wrote both grant rows (verified through the data API).
• "Permission Set (legacy id)" is gone from the form, and both list headers show only New.

Verification (local, at this head)

• dogfood in full: 253 files passed, 1 skipped; 2,039 tests passed;
• plugin-security: 197 files, 4,045 passed;
• platform-objects: 69 files, 1,086 passed;
• cli unit tier: 282 files, all passed after the floor fix (platform-page-i18n-parity 27/27);
• typecheck green: platform-objects, plugin-security, cli, dogfood;
• dispatch-gates --ran: 72 derived families, all recorded with exit 0.

Changeset

@objectstack/platform-objects and @objectstack/plugin-security patch, Clause-②: no. This changes shipped metadata only; no key anyone authors moves.

Acceptance notes

• objectui finding (carrier: none yet; the seat routes it): the position-ref picker lists the everyone audience anchor among positions, and the server refuses it as a grant target. Not filtered here.
• Tier: no governed surface. feat(plugin-security)!: the boot writes no sys_permission_set row (#15204 stage 6b-2) #22900 waits for this PR and for the pin PR that follows feat(spec,metadata-protocol)!: rename the type registry's allowOrgOverride to allowEnvironmentOverlay, on the registry and on GET /meta/types #22874.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_013XKtmoR5QCUSTMjiQ3os49

[github-actions]

📓 Docs Drift Check

This PR changes 3 package(s): @objectstack/cli, @objectstack/platform-objects, @objectstack/plugin-security, touching 6 documentable anchor(s).

11 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/concepts/metadata-lifecycle.mdx (via sys_permission_set (literal, a string literal in slots))
• content/docs/data-modeling/objects.mdx (via SysUserPermissionSet (symbol, a top-level const object))
• content/docs/deployment/environment-variables.mdx (via sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/deployment/validating-metadata.mdx (via permission_set (literal, a string literal in a comment in slots; a string literal in slots))
• content/docs/permissions/authorization.mdx (via sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/permissions/delegated-administration.mdx (via sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/permissions/explain.mdx (via permission_set (literal, a string literal in a comment in slots; a string literal in slots))
• content/docs/permissions/permission-sets.mdx (via permission_set (literal, a string literal in a comment in slots; a string literal in slots), sys_permission_set (literal, a string literal in slots))
• content/docs/permissions/positions.mdx (via sys_position (literal, a string literal in slots))
• content/docs/permissions/system-context.mdx (via permission_set (literal, a string literal in a comment in slots; a string literal in slots), sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/protocol/backward-compatibility.mdx (via sys_position (literal, a string literal in slots))

⛔ 10 release-owned page(s) also name something this change touched. These are read-only:

• content/docs/releases/index.mdx (via sys_position (literal, a string literal in slots))
• content/docs/releases/v12.mdx (via sys_permission_set (literal, a string literal in slots))
• content/docs/releases/v13.mdx (via SysUserPosition (symbol, a top-level const object), sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/releases/v14.mdx (via sys_position (literal, a string literal in slots))
• content/docs/releases/v15.mdx (via sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/releases/v17/17-0.mdx (via sys_permission_set (literal, a string literal in slots))
• content/docs/releases/v17/17-1.mdx (via sys_permission_set (literal, a string literal in slots), sys_position (literal, a string literal in slots))
• content/docs/releases/v17/17-2.mdx (via sys_position (literal, a string literal in slots))
• content/docs/releases/v17/17-5.mdx (via sys_permission_set (literal, a string literal in slots))
• content/docs/releases/v17/17-6.mdx (via sys_permission_set (literal, a string literal in slots))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• 1 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 54 of 206 client-bound route-ledger rows — the other 152 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 152: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 55 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 97 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 41 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json 51b43a8dc45fef9d86229509ab6b2679ffa9d501 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from c6d4c0b294bba6e9748fc057aa9341c1ca3f46dc — the merge of head 7de0f00c91aaaecd2d70ba59695d373baa290bf1 into base 51b43a8dc45fef9d86229509ab6b2679ffa9d501, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin c6d4c0b294bba6e9748fc057aa9341c1ca3f46dc && git checkout c6d4c0b294bba6e9748fc057aa9341c1ca3f46dc
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 51b43a8dc45fef9d86229509ab6b2679ffa9d501 7de0f00c91aaaecd2d70ba59695d373baa290bf1 && git checkout -B drift-repro 51b43a8dc45fef9d86229509ab6b2679ffa9d501 && git merge --no-ff 7de0f00c91aaaecd2d70ba59695d373baa290bf1

node scripts/docs-audit/affected-docs.mjs --json 51b43a8dc45fef9d86229509ab6b2679ffa9d501

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs 51b43a8dc45fef9d86229509ab6b2679ffa9d501 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

Seat ACCEPT: #22939 (#22865 B2 phase 2: grant name pickers on the user page) at 7de0f00c91; queued on green

Epic PM session_01Rerax7QTjKMPCUZxQUtPFR (marchtian), seat epic:#15194, 2026-10-12T05:27Z. Ruling objectstack#22865 → B2 (6115690641). Phase 1 measured in 6116396441. The widget keys were accepted in 6116842485, and the widget is objectui PR-5, merged as 088e0f705 (Landed 6117301706). The developer is session_013XKtmoR5QCUSTMjiQ3os49.

The head's tree equals d4c22c8fe (git diff d4c22c8fe 7de0f00c91 is empty). The pin commit e486abb9c8 was folded in and then reverted by a new commit, without force-push. .objectui-sha is claimed by #22874, and objectui e5311165 (#12168/#12141, which reads allowEnvironmentOverlay only) is an ancestor of 088e0f705, so the pin follows #22874 (seat note 6118044018). CI on d4c22c8fe was green across all shards, including all 6 Test Core shards.

Checklist, read on GitHub, not from the report:

• PR form: draft, base main; line 1 Part of #15204, line 2 Clause-②: no; the whole body is scanned and no closing keyword sits next to a card number; assignee marchtian.
• Scope: 9 files, +133/−52. check-governed-merges --pr reads NOT governed. No CHANGELOG.md and no content/docs/releases/.
• The diff does exactly the ruled phase 2:
  • sys_user_position.position declares widget: 'position-ref'.
  • sys_user_permission_set.permission_set declares widget: 'permission-ref'.
  • permission_set_id is hidden: true, and its column stays for C7b (the maintainer agreed to hide it).
  • The user page's two add blocks are deleted, and their comments are rewritten to say why.
• Clause-②: no holds. No packages/spec file is touched. widget is a free-string field hint (FieldSchema.widget), and hidden is an existing field key. The removed add blocks were page content, not an accepted-metadata set.
• The changeset 15204-grant-name-pickers.md, patch for platform-objects and plugin-security. Each sentence was read against the diff:
  • the two widget declarations, and the stored name;
  • "a console without those widgets keeps the text input" (the record form degrades an unregistered hint to the type renderer, objectui#5254);
  • permission_set_id hidden, with its column staying;
  • the two buttons removed, because they listed rows a deployment need not hold;
  • grant with New;
  • "removing a grant is unchanged", which is phase 1's measurement in 6116396441 ("row-level Remove survives and works").
• Evidence:
  • Dogfood grant-name-pickers pins: the served GET /meta/object carries both hints and the hidden flag.
  • The page pin: no add on either list, with a business-unit control.
  • Ablations: stripping the hints turns the three served-metadata cases red; re-adding an add turns the page pin red.
  • Full suites: dogfood 253 files, plugin-security 197, platform-objects 69, cli unit green. The i18n parity floor went 45 → 44 because the two add labels left.
  • dispatch-gates --ran: 72 of 72 exit 0.
  • In the browser, against objectui PR-5's head: both fields render as registry pickers (16 positions, 17 sets), a pick wrote both rows, and the legacy-id field is gone.
• CI on 7de0f00c91: 42 check-runs: 37 success, 5 skipped, 0 red. The five skips are all on check-expected-skips's roster: Build Docs, Console Pin Gate (the net diff leaves .objectui-sha untouched), Auto Label, Check PR Size and Packed-tarball smoke (opt-in). All six Test Core shards and their aggregate are success. mergeable_state is clean. The developer re-ran the derived gates on this head as well: 72 of 72 exit 0, check:console-sha included.

Until the console pin moves, the two fields render as text inputs for the name. The pin goes 4997995 → e5311165 in #22874, then a follow-up hop to 088e0f705. main is unreleased.

Recorded finding, not filed here: the position picker also lists everyone, the audience anchor, which the server refuses as a grant target. This is an objectui filter question; the seat routes it.

Then: flipped ready and enqueued in this same stroke. On MERGED, a Landed record. This PR and the pin PR together are D's (#22900) landing gate.

---

Generated by Claude Code

[github-actions]

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 38208298792 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

分类：failure —— 按下面的日志分诊。

失败的 job（日志抽取，best effort）：

• Test Core — 失败步骤: Check the per-package suite-duration ceilings（日志不可读，点进 job 看）

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

⚠️ 断言这一侧有一类例外，判据是断言在测什么，不是它是不是 AssertionError。 断言的对象是产品行为（一个值、一个形状、一次拒收）⇒ 照上面读：真实的行为改变，去查，⛔ 不要重排掉；
断言的对象是这次实验自身的有效性前提（跑完的耗时、负载下的先后、任何只在时间预算内才成立的条件）⇒ 它跟超时是同一类，同样对负载敏感，重排一次是合法的判别手段。
识别是机械的：断言的消息或它比较的值本身点名了一段时长、一个时间戳、一个耗时计数。实测过的一对 —— AssertionError: SecurityPlugin.init() ran: expected false to be true 测的是产品行为（真回归）；
AssertionError: this run took over a second, so second-precision stamps could have differed too: expected 1006 to be less than 1000 测的是实验前提：它守护的那条不变式当时是绿的，同一个 head 原样重排一次即成功。
穿着 AssertionError 外衣的时间测量，仍然是时间测量。（⛔ 这只改「怎么读一次红」，不改「哪些测试可以重排」——后者由别处管。）

跨 PR 相同签名（24h，按失败测试文件聚合）：

• ⚠️ 本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
• ⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

• 本 PR 过去 24h 无队列失败记录（首次）。
• 过去 24h 队列共有 14 个失败构建（不含本次）。

分诊清单：

1. 失败测试在本 PR 改动的包里 → 真回归，修 PR。
2. 失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
3. 两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

---

Generated by Claude Code · merge-queue-triage workflow (#4859)

[objectstack-fleet]

Seat note: dequeued on one suite-duration ceiling, in a package this PR does not touch; re-queued once · epic PM session_01Rerax7QTjKMPCUZxQUtPFR (marchtian) · 2026-10-12T05:58Z

What red: queue build 38208298792, Test Core → "Check the per-package suite-duration ceilings" → @objectstack/client: 182.82s, over its 179.74s ceiling. That is 1.7% over, and the slowest file is src/auth-login-register-envelope.test.ts. No test failed.

Why it is not this PR's:

• The diff (9 files: platform-objects, plugin-security, dogfood) does not touch packages/client.
• client has crossed the same ceiling in other queue groups: the sightings on ci: merge-queue Test Core reds on the new suite-duration ceiling — @objectstack/cloud-connection measured 177.5 s and 200.3 s vs its 163.89 s ceiling, in two groups of PRs that do not touch it (live since 1bd612bd6) #22949, the anchor card for this check, which is priority:p1 and needs-user-decision.
• The ceiling measures runner load in a full-suite queue build, which is the load-sensitive class.

No fix exists yet. The seat spends this PR's one re-run as one re-queue on the unchanged head 7de0f00c91. If it reds again on a ceiling alone, the seat leaves the PR out of the queue until #22949 is ruled, and records the sighting there.

---

Generated by Claude Code

[github-actions]

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 38208631030 红了。队列跑的是全量套件（PR 侧 CI 只跑 affected 子集），
所以失败的测试可能在本 PR 没碰过的包里 —— 那不是重排能修的。每次盲目重排都会让排在后面的所有 PR 重建一轮。

分类：failure —— 按下面的日志分诊。

失败的 job（日志抽取，best effort）：

• Test Core — 失败步骤: Check the per-package suite-duration ceilings（日志不可读，点进 job 看）

↳ 失败原因 是判读的关键：超时（Test timed out in … / Hook timed out in …）多半是负载/时序，不是本 PR 的回归；
断言（AssertionError: …）才指向真实的行为改变。两者的 FAIL 行长得一模一样，只有这一行能区分。

⚠️ 断言这一侧有一类例外，判据是断言在测什么，不是它是不是 AssertionError。 断言的对象是产品行为（一个值、一个形状、一次拒收）⇒ 照上面读：真实的行为改变，去查，⛔ 不要重排掉；
断言的对象是这次实验自身的有效性前提（跑完的耗时、负载下的先后、任何只在时间预算内才成立的条件）⇒ 它跟超时是同一类，同样对负载敏感，重排一次是合法的判别手段。
识别是机械的：断言的消息或它比较的值本身点名了一段时长、一个时间戳、一个耗时计数。实测过的一对 —— AssertionError: SecurityPlugin.init() ran: expected false to be true 测的是产品行为（真回归）；
AssertionError: this run took over a second, so second-precision stamps could have differed too: expected 1006 to be less than 1000 测的是实验前提：它守护的那条不变式当时是绿的，同一个 head 原样重排一次即成功。
穿着 AssertionError 外衣的时间测量，仍然是时间测量。（⛔ 这只改「怎么读一次红」，不改「哪些测试可以重排」——后者由别处管。）

跨 PR 相同签名（24h，按失败测试文件聚合）：

• ⚠️ 本次没有可用的聚合签名（日志里没有能解析出测试文件名的 FAIL 行）—— 这不是「没有同签名的其他 PR」，是这一轮没测到。跨 PR 聚合本次不可用，请手工比对其他 PR 的同类评论。
• ⚠️ 24h 评论账本没读完（超过 5 页仍未读到窗口尽头），所以上面的「不同 PR 数」是下界，不是全量。

历史信号：

• ⚠️ 本 PR 过去 24h 已在队列失败 1 次（不含本次）。 内容未变而反复失败 ⇒ 高度怀疑 flaky 测试或与同组 PR 的语义冲突，重排不解决。
• 过去 24h 队列共有 15 个失败构建（不含本次）。

分诊清单：

1. 失败测试在本 PR 改动的包里 → 真回归，修 PR。
2. 失败测试与本 PR 无关 → 看上面的「跨 PR 相同签名」；已有汇总 issue ⇒ flaky/环境问题实锤，去那张 issue 上谈，修好前重排只会再烧一轮全队列。
3. 两者都不是 → 可能与同组 PR 语义冲突；等前面的 PR 落地或失败出队后再重排一次即可，不要连续重排。

---

Generated by Claude Code · merge-queue-triage workflow (#4859)

[objectstack-fleet]

Seat note: re-queued behind W2, whose queue group carries #22949's fix · epic PM session_01Rerax7QTjKMPCUZxQUtPFR (marchtian) · 2026-10-12T08:19Z

This PR left the queue on the suite-duration ceiling step (client), held per 6118649663. That hold has now lifted:

• ci: merge-queue Test Core reds on the new suite-duration ceiling — @objectstack/cloud-connection measured 177.5 s and 200.3 s vs its 163.89 s ceiling, in two groups of PRs that do not touch it (live since 1bd612bd6) #22949 is ruled C2 for both timing steps plus a floor (6119186511). Its fix is ci: Test Core's two timing gates exit-grade on schedule alone (C2), with a 20s ceiling floor #22973.
• W2 fix(security)!: under a wall a catalog-carrying package is the platform operator's — install-local and the /packages delete refuse tenants (ADR-0131 D3) #22935 carries the two ci: Test Core's two timing gates exit-grade on schedule alone (C2), with a 20s ceiling floor #22973 commits, ported byte-identical (ACCEPT 6120417056), and is in the merge queue ahead of this PR.
• So this PR's queue group is built on W2's and grades the timing steps under C2.

Re-queued on the unchanged head 7de0f00c91. ACCEPT 6118326033 stands, and mergeable_state is clean. This is a re-queue behind an existing fix, not a flake re-run.

---

Generated by Claude Code