[objectstack-fleet]

domain:server#3 · os-bill · session_01Utp6CXX2qjqm2uXbJA8Jcv

Requested by Bill · project thread
Fixes #30055

Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么：engine.ts 里被 lint 规则:禁止对引擎/驱动查询选项做 as any / : any 擦除(#4721 的顺带项,已实测残余量) #4918 擦除钉住的六个 ObjectQL 方法，先把查询选项补上类型（engine.ts 的擦除点 6 → 0，退出 erasure baseline），再用 codemod 新加的第四遍 --members 把方法体搬进三个已有的 facet 模块。engine.ts 11,771 → 11,316 行。PIN 表和 facade 例外表各退六行，referenceExists 的 read-invention 台账行跟着改指向。
• 为什么改：维护者在 refactor(objectql): generate the engine.ts section split with a deterministic codemod #30012 上说「我觉得文件还是太长，还有进一步的拆分优化空间吗」「可以后续立卡处理」「合并之后再继续」。第一遍 codemod 只能在未拆分的文件上跑，所以加了第四遍，它在已拆分的文件上按第一遍的规则搬指定方法。
• 风险与回滚：行为不变。四处直接删掉 as any，另外两处保留具名断言（见 Acceptance notes）。搬移由撤销证明逐字节还原 base；.d.ts 声明不变，只少了随方法体搬走的私有 docblock；objectql 测试 base 550 文件 / 10,979 = head。回滚就是 revert。
• 先看：① engine.ts 里六个擦除点的补类型提交 c358b417、162cb326；② scripts/codemod-split-engine.mjs 新的第四遍一节（MEMBERS_CARD）；③ engine-references-and-master-detail.ts 新增的四个成员。
• 两分钟验证：node scripts/codemod-split-engine.mjs --members --verify --base baf1a536^，应看到 6 moved method(s) 和 does rebuild the base engine.ts and its modules byte-for-byte，退出 0。
• 需要你做的：无，席位落地。

What changes

1. The six lint 规则:禁止对引擎/驱动查询选项做 as any / : any 擦除(#4721 的顺带项,已实测残余量) #4918 sites are typed (commits c358b417, 162cb326, no behaviour change). referenceExists, resolveMasterDetailParent, resolveMasterDetailParents and probeReferencingRows drop the as any outright: their literals (where / fields / context) already satisfy EngineQueryOptions, so tsc now checks them. Two sites keep a named assertion instead of an erasure:
   • seedAutonumber: the keyset walk's page query types its where as unknown (it is this scan's filter AND-ed with the seek predicate), so the spread names that one slot as EngineQueryOptions['where'] (an Omit of where from KeysetPageQuery, intersected with that slot). The runtime object is the same spread.
   • inspectDanglingReferences: the published DanglingReferenceAuditPort.find spells its options as a string-keyed record of unknowns, so the hand-over is opts as EngineQueryOptions. Changing the port would change the public .d.ts, which this card rules out.
     The rule's sites in engine.ts go from 6 to 0, so engine.ts leaves scripts/query-options-erasure-baseline.json and the lint rule now covers it like any other file.
2. The codemod gains a fourth pass, --members (scripts/codemod-split-engine.mjs). The first pass refuses an engine.ts that already carries its stubs, and main has been split since Codemod split: generate the objectql engine.ts split with a deterministic script, verified byte-identical #29986, so "re-run the codemod" needs a pass that works on the split file. --members A,B moves the named methods exactly as the first pass would: same stay rules (any reason refuses), same stub, same facet member, same docblock rule, into the facet module of the method's section that the first pass wrote, in class order. That module's header list, host interface and import lines grow by what the moved bodies read, and so do the class's statics route and that facet's host factory. The pass never starts a facet, reroutes a module or widens a generated export list; each of those refuses with exit 2. --members --verify --base REF undoes the moves from the head alone and requires the base engine.ts and modules byte for byte, every other generated module unchanged, and each moved member equal to its base method's text. Pass 1's renderers are shared, not copied: five pure helpers were extracted from it (classStaticsOf, specifierImportLines, moduleHeaderLines, facetMethodText, hostInstallLines), and its 105 self-test cases still pass unchanged.
3. The six bodies move (commit baf1a536, generated, not hand-edited): node scripts/codemod-split-engine.mjs --members seedAutonumber,referenceExists,resolveMasterDetailParent,resolveMasterDetailParents,inspectDanglingReferences,probeReferencingRows --write.
   • engine-defaults-and-autonumbers.ts takes seedAutonumber; engine-references-and-master-detail.ts takes referenceExists, resolveMasterDetailParent, resolveMasterDetailParents and inspectDanglingReferences; engine-cascade-planning.ts takes probeReferencingRows.
   • The private five move with their docblocks; inspectDanglingReferences is public, so its docblock stays on the class and the module carries the placeholder.
   • ObjectQLStatics gains referenceCheckContext. The references facet's host gains find, findOne, materializeParentHeader, resolveEnginePosture. Nine import bindings only the moved bodies read leave engine.ts.
4. The pin and allow-list rows retire: the six PATH_READER_PINS rows (the table is now empty), the six tooling: a structural gate that refuses any method body in a facade file once protocol.ts and engine.ts land in facade form (ruling 6624258595) #30019 no-body exceptions in scripts/facade-delegation-registry.json, and the referenceExists entry in scripts/durability-read-invention.baseline.json now names engine-references-and-master-detail.ts, where its body lives.

Size: 9 files, +1,479 / −609 = 2,088 changed lines (under 3,000).

Tests

• node scripts/codemod-split-engine.mjs --self-test: 124 cases in 9 batteries pass. The new battery named methods onto the split class (fourth pass) has 18 cases: the first pass keeps two pinned fixture methods; the fourth pass moves both into the right module in class order; the proof holds and two runs write identical trees; the stub is the first pass's with async dropped; a private docblock moves and a protected one stays; the header, host interface and factory, statics route and imports each grow by exactly what the bodies read; engine.ts prunes what only the bodies read and keeps a type a stub's signature still names; unknown, already-moved, stay-rule and still-pinned names refuse; and the proof fails on an edited body character, a host member nothing reads, an unread import, a stub that kept async, an unrelated edit in engine.ts or another module, and a head that moved nothing.
• pnpm --filter @objectstack/objectql test: base (f7afab86 sources) 550 files / 10,979 tests passed; head 550 / 10,979 passed.
• pnpm --filter @objectstack/objectql typecheck: exit 0 (tsc and check:test-typecheck).

Acceptance notes

• Two named assertions stay (the two sites in What changes 1). They name the contract the value is read under, which is what the lint 规则:禁止对引擎/驱动查询选项做 as any / : any 擦除(#4721 的顺带项,已实测残余量) #4918 rule asks for when the value crosses a boundary typed looser than the engine. Neither is counted by the rule.
• homeAmountReadsFor (added by hand in fix(objectql): an uninstalled package no longer reads as read-only until a restart #30280) sits in engine-references-and-master-detail.ts out of class order and missing from the module header's list. The fourth pass reads the header's own list and adds its names to it, leaving that drift as it stands. Not a defect; noted for whoever next regenerates the module.
• .d.ts: the private methods' docblocks leave the bundled .d.ts with their bodies (211 lines in each of util-HASH.d.ts and util-HASH.d.mts). Ruling 6624345766 item 1 says a private docblock moves with its body. This is the same reading refactor(objectql): move 36 hand-series private docblocks out of engine.ts with the codemod's third pass (slice 1 of 2) #30341 / refactor(objectql): move the last 28 hand-series private docblocks out of engine.ts with the codemod's third pass (slice 2 of 2) #30415 gave.

Verification

Undo proof on the real tree

node scripts/codemod-split-engine.mjs --members --verify --base baf1a536^ exit 0:

undo proof: 6 moved method(s) (seedAutonumber, referenceExists, resolveMasterDetailParent, resolveMasterDetailParents, inspectDanglingReferences, probeReferencingRows); engine.ts 11776 -> 11316 lines; undoing the moves does rebuild the base engine.ts and its modules byte-for-byte once the 9 import binding(s) only moved bodies read are pruned from the base (AUTONUMBER_SEED_PAGE_SIZE, AuditableObject, KeysetPageQuery, declaredMultiValued, isMissingTableError, isStoredMetadataBodyObject, keysetWalk, materializeDeclaredFields, readStoredAutonumberCounter)

Against main (f7afab86) the same command refuses all six (the facet member is not its base method's text), because the typing commits changed those bodies. The move is proved against the commit just before it; the typing is the reviewed diff.

Declarations, base vs head

pnpm --filter @objectstack/objectql build at f7afab86 sources and at head. The chunk name hash (util-HASH) was normalised. core.d.ts, core.d.mts, index.d.ts and index.d.mts are byte-identical once the hash is normalised. The two util-HASH files differ by 422 lines, every one a docblock line. With docblocks stripped, all six declaration files are byte-identical.

Ablation

Re-erasing one moved site (as any back on referenceExists's options in engine-references-and-master-detail.ts, through scripts/ablation-replace.mjs): pnpm check:query-options-erasure exits 1 with engine-references-and-master-detail.ts: NEW engine query-options erasure (1 site(s)). Restored to HEAD; the hash matches and git diff HEAD is empty.

Gates

node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 74 families, 80 commands. Among them are the fourth-pass verify, the codemod self-test, eslint on the touched files, and pnpm --filter @objectstack/spec check:liveness. All 80 recorded exit 0, and --ran reconciles: 74 derived famil(ies) accounted for — 74 run, 0 NOT-MEASURED.

Two first ran exit 3 (prerequisite, not measured) and were rerun green: check-plugin-teardown-shape --self-test needed its pinned fixture commit fetched into the shallow clone, and check:dual-build-cjs-loads needed pnpm build (119 of 119 tasks; then 116 entry points across 71 packages load).

Named here because the card asks: check-facade-delegation-only passes with 5 facade class(es) in 3 file(s) hold delegations only, 62 named exception(s) (six fewer); check:query-options-erasure passes with 30 unswept … 13 file(s), none new; check:liveness passes.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01Utp6CXX2qjqm2uXbJA8Jcv

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/objectql, touching 22 documentable anchor(s).

20 hand-written doc(s) name something this change touched — list omitted above 15 rows. Re-derive on the tree named below: node scripts/docs-audit/affected-docs.mjs --json f273c9fe04d4798f8ed3dcf41822817588fe1ac1.

⛔ 7 release-owned page(s) also affected — read-only, see AGENTS.md Documentation Guardrails.

What this run could not see

• 1 anchor(s) matched too much of the corpus to be a work list: ObjectQL (symbol, 80 pages)
• 1 name(s) were too generic to anchor anything (single lowercase words)
• the SDK route bridge reached 61 of 219 client-bound route-ledger rows — the other 158 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 158: 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; 103 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 — 21 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 f273c9fe04d4798f8ed3dcf41822817588fe1ac1 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 146b16c98623c351b89a84205daedf279245b7d5 — the merge of head 293b12ffec15abd89ec4f36b9fbf7b995e0ec481 into base f273c9fe04d4798f8ed3dcf41822817588fe1ac1, 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 146b16c98623c351b89a84205daedf279245b7d5 && git checkout 146b16c98623c351b89a84205daedf279245b7d5
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin f273c9fe04d4798f8ed3dcf41822817588fe1ac1 293b12ffec15abd89ec4f36b9fbf7b995e0ec481 && git checkout -B drift-repro f273c9fe04d4798f8ed3dcf41822817588fe1ac1 && git merge --no-ff 293b12ffec15abd89ec4f36b9fbf7b995e0ec481

node scripts/docs-audit/affected-docs.mjs --json f273c9fe04d4798f8ed3dcf41822817588fe1ac1

⚠️ 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 f273c9fe04d4798f8ed3dcf41822817588fe1ac1 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[objectstack-fleet]

domain:server#3 · os-bill · session_01Utp6CbPxYq1wQmUNWvMZkC

维护者速读（席位终稿）

• 改了什么：engine.ts 里被 lint 规则:禁止对引擎/驱动查询选项做 as any / : any 擦除(#4721 的顺带项,已实测残余量) #4918 擦除钉住的六个 ObjectQL 方法先补上查询选项的类型（四处删掉 as any，两处改成具名断言），再用 codemod 新加的第四遍 --members 把方法体搬进三个已有的 facet 模块；engine.ts 11,771 → 11,316 行，退出 erasure baseline。
• 为什么改：维护者在 refactor(objectql): generate the engine.ts section split with a deterministic codemod #30012 上要求继续拆 engine.ts（「合并之后再继续」）；这六个方法是 lint 规则:禁止对引擎/驱动查询选项做 as any / : any 擦除(#4721 的顺带项,已实测残余量) #4918 钉住、前三遍搬不动的最后一批。
• 风险与回滚：行为不变。搬移由 --members --verify 从 head 撤销后逐字节还原 base 证明；私有方法的声明不进公开 .d.ts；objectql 测试 base 与 head 同为 550 文件 / 10,979 条。回滚即 revert。
• 先看：补类型的提交 c358b417、162cb326；scripts/codemod-split-engine.mjs 的第四遍（MEMBERS_CARD）；engine-references-and-master-detail.ts 新增的四个成员。
• 两分钟验证：node scripts/codemod-split-engine.mjs --members --verify --base baf1a536^，退出 0。
• 需要你做的：无，席位落地（非 BREAKING，非受管面）；入队等 p0 4-hourly full run: red on main (Lint & Type Check) #30467（main 红）止血合入之后。

席位意见：同意落地。纯重构，类型收紧后 tsc 开始检查这六处读，撤销证明与测试计数一致；改动 2,088 行，未超 3,000。

席位核对

ACCEPT · head 293b12ffec15abd89ec4f36b9fbf7b995e0ec481 · card #30055 (claim on the card) · 2026-11-05T13:04Z

• Shape: draft PR to main, author objectstack-fleet[bot] (relay), attribution block then Fixes #30055; no other closing keyword. 9 files, +1,479 −609 = 2,088 lines. check-governed-merges.mjs --pr 30463 → exit 0, 0 of 9 paths governed, ≤ 3000. No merge=os-regen path.
• Premise measured: the six #4918 sites read on origin/main (6 erasures in engine.ts, baseline recorded 8); after: 0, and engine.ts leaves scripts/query-options-erasure-baseline.json. Ablation recorded (an as any put back in the references module reds check:query-options-erasure, restored).
• Gates (check-runs on this head): 48 runs, 35 success, 13 path-gated skipped, 0 failed; Lint & Repo Gates, TypeScript Type Check, Test Core, Build Core, Governed Surface Queue Guard, Test UI, Live E2E UI, Check Changeset, Lint · PM dispatch-gates all success. Dev's derived list: 80 commands / 74 families exit 0, --ran 74/74, incl. check:liveness and check-facade-delegation-only.
• Seat read of the diff: engine.ts keeps one-line delegation stubs (private seedAutonumber(…) drops async and returns the facet's promise; same Promise<number>); referenceExists moves verbatim minus as any (findOne options now typed as EngineQueryOptions); ObjectQLStatics gains referenceCheckContext. Private members carry no type in the emitted declarations, so the public surface is unchanged: Clause-②: no, Metadata-protocol: none, @objectstack/objectql patch — consistent. The .d.ts byte-identity reading judges the name set only (no positive control in the block), and the seat does not rely on it for the verdict.
• Accepted deviation: none. Noted (Acceptance notes): homeAmountReadsFor (hand-added by fix(objectql): an uninstalled package no longer reads as read-only until a restart #30280) sits outside the module's header list and class order; the fourth pass leaves it as is.
• Out of scope: none filed. refactor: give the inferred-return-type public methods on the engine.ts and protocol.ts facades explicit return types (.d.ts unchanged), so the codemod can move their bodies #30054 (seat 5) is serial behind this PR; Offboarding: banning a user says nothing about their open work, and there is no way to hand it over #30205's one-line findOne host addition in assertReferencesResolve does not meet these hunks.
• Landing: non-BREAKING, not governed ⇒ seat lands it. Held out of the queue until p0 4-hourly full run: red on main (Lint & Type Check) #30467 (main red since bc254be3, this seat's lane) lands, per the main-red rule.
• Writes: dev api_writes 2, mcp_calls 5 (reads).

[objectstack-fleet]

domain:server#3 · os-bill · session_01Utp6CbPxYq1wQmUNWvMZkC

Landing · 2026-11-05T13:36Z · head 293b12ffec15abd89ec4f36b9fbf7b995e0ec481 unchanged since ACCEPT 6658708556; the p0 main-red fix #30487 is in the queue ahead of it, so the main-red hold is lifted. Ready + auto-merge.

[objectstack-fleet]

domain:server#3 · os-bill · session_01Utp6CbPxYq1wQmUNWvMZkC

Queue kick 13:58Z, re-queued once · 2026-11-05T14:01Z

• Failing check: Live E2E UI in merge-group CI run 42037558126 (job 127279783258). One spec, studio-customize-journey.spec.ts:121, failed with apiRequestContext.get: read ECONNRESET on GET /api/v1/meta/view/showcase_task.in_progress. The other 16 specs passed, including the 6 that ran after it, so the server stayed up.
• Why not this PR: this diff changes engine docblocks, the split codemod and baseline files only. It touches no meta-view read and no Studio or marketplace path. Live E2E UI is success on this head (run 42028327503).
• Action: re-queued once with the same head 293b12ff and the same ACCEPT 6658708556. A second red on this spec is treated as real.

[github-actions]

⛔ merge queue 构建失败 — 先分诊，再决定要不要重排

队列构建 42037558126 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）：

• Live E2E UI — 失败步骤: Run the live e2e specs（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)