org 作用域的 flow overlay 只在「本进程内发布后」绑定触发器,重启后静默失绑——冷启动两条读路径都把 organization_id 非空的行滤掉了

发现于 #6155 的前提核验(未在该 PR 内修,按 Prime Directive #10 单开)。

事实(实测,非推断)

flow 在 DEFAULT_METADATA_TYPE_REGISTRY 里是 allowOrgOverride: true + allowRuntimeCreate: true,所以 runtime/src/domains/meta.ts:158-159 的 saveMetaItem({ type, name, item, organizationId })(org 取 resolveActiveOrganizationId)会写出 sys_metadata.organization_id = 'org_a' 的 flow 行。实测确认该行确实落库。

之后这一行的命运是不对称的:

时机	路径	org 行可见?
发布当场(同进程)	`saveMetaItem` → `applyRegistryWriteThrough`(publish 分支)→ `hydrateOverlayIntoRegistry` → 进程级 SchemaRegistry	✅ 可见
发布当场,自动化重绑	`emitMetadataMutation` → `metadata:reloaded` → `resyncFlowsFromProtocol` → `getMetaItems({ type: 'flow' })`(读到 registry 分支)	✅ 绑定,定时任务起飞
重启后冷启动	`ObjectQLPlugin.restoreMetadataFromDb` → `protocol.loadMetaFromDb()`,`where = { state:'active', organization_id: null }`(`metadata-protocol/src/protocol.ts:10620-10623`)	❌ 跳过
重启后 kernel:ready 绑定	`service-automation/src/plugin.ts:1542` `protocol.getMetaItems({ type: 'flow' })` —— 不传 `organizationId`,于是 `protocol.ts:3319` 的 `orgRecords = orgId ? … : []` 为空	❌ 跳过

实测(真 ObjectStackProtocolImplementation + 内存 sys_metadata 引擎,两行 flow:org_sweep org=org_a、platform_sweep org=null):

PROBE rows = [{"name":"org_sweep","org":"org_a"},{"name":"platform_sweep","org":null}]
PROBE getMetaItems({type:flow}) names(registry 为空时) = ["platform_sweep"]
PROBE getMetaItems({type:flow}) names(publish 写穿后)  = ["org_sweep","platform_sweep"]
PROBE loadMetaFromDb = {"loaded":1,...}      // 只有 platform_sweep

影响

静默失效:租户在 Studio 里发布一个 org 归属的 flow(record-change 或 schedule),当天正常触发;下一次进程重启后它再也不触发,没有任何日志说它消失了——kernel:bootstrapped 的 unbound 审计也看不见它(它压根没注册)。这正是 AGENTS.md「Absence must be loud」的反面。
可见性不对称:写穿期间该 org 行进了进程级 registry,于是 getMetaItems({ type: 'flow' }) 对任何 org 的调用方都返回它——而 loadMetaFromDb 的注释明确说 per-org overlay 之所以不 hydrate 就是「to avoid cross-org leakage into the process-wide SchemaRegistry」。写路径比读路径更宽松,恰好是 applyRegistryWriteThrough TSDoc 自己声明要避免的("The write must not be more permissive about that than the read is" —— 该 gate 只挡了 environmentId !== undefined,没挡 org)。

与 #6155 的关系

不是 #6155 的子问题:#6155 问的是「定时流程的 org 从哪来」,本单问的是「org 归属的 flow 到底还活不活」。但两者相互制约——若维护者裁决 org flow 本就不应被加载(ADR-0005 的白名单表原文即「automation flow ❌ per-org override,Per-org variants are a deployment, not an overlay」,与代码现状矛盾,另见同批开的 registry/ADR 漂移单),那本单的修法就是在写入侧拒绝而不是在读取侧补齐。所以建议先定性再定修法。

复现

无需真 DB:packages/metadata-protocol/src/protocol-publish-drafts-org-scope.test.ts 的 stub 引擎即可(需给 registry 补 isPackageDisabled: () = false 与一个真存的 listItems)。

Refs:#6155、#4636(loadMetaFromDb object 分支的另一处 row 读取缺陷)、ADR-0005 §Tenant-customizable type whitelist、applyRegistryWriteThrough(protocol.ts:7434)。

Blocked-by: #6191

Generated by Claude Code

Triage: pm:blocked (Blocked-by: #6191) + domain:engine-core + target:v17.

Classification — real, well-located, but its fix direction is gated

Every fact in the body reproduces against origin/main: flow carries allowOrgOverride: true (packages/spec/src/kernel/metadata-plugin.zod.ts:649), loadMetaFromDb filters organization_id: null (packages/metadata-protocol/src/protocol.ts:10622), and the cold-boot binder calls protocol.getMetaItems({ type: 'flow' }) with no organizationId (packages/services/service-automation/src/plugin.ts:1542). The write/read asymmetry is exactly as documented.

Not queued, because the body's own closing line is right: the two candidate fixes are opposite. If ADR-0005's ❌ row stands, the fix is to refuse the write; if the reversal is ratified, the fix is to complete the read. A dev dispatched today would hit the os-dev "underspecified decision that changes the public contract → STOP" clause and burn a run. #6191 is now in the maintainer's decision inbox and is the gate, so this issue takes pm:blocked with a machine-readable Blocked-by: line added to the body — the unlock sweep returns it to the queue when #6191 closes. This is a deferral of direction, not of priority.

Routing rationale (landing site, not title vocabulary)

domain:engine-core. The issue reads as automation, and the most visible symptom is a service-automation one — but three of the four named code sites are in packages/metadata-protocol (domain:engine-core per the domain table's packages/metadata* family): the read filter at protocol.ts:10620-10623, the orgRecords = orgId ? … : [] branch at :3319, and the write-through gate applyRegistryWriteThrough at :7434 whose own TSDoc states the write must not be more permissive than the read. The service-automation call site at plugin.ts:1542 is a consumer of the flattened view, and it is correct under the current contract — it cannot pass an organizationId it has no business knowing at kernel:ready. Under either ruling the asymmetry is resolved in metadata-protocol (or, for the refuse-the-write route, in the packages/spec flag that #6191 owns), not in the binder. Labeling this domain:services off the title would route it away from the package the fix lands in.

One part of this is a defect under either ruling

Worth pricing separately when the gate lifts: impact ② (the publish-time write-through hydrates an org_a row into the process-wide SchemaRegistry, so getMetaItems({ type: 'flow' }) returns it to callers of any org) is a cross-org visibility leak regardless of which direction #6191 goes — loadMetaFromDb's comment says avoiding exactly that leak is why per-org overlays are not hydrated. If the maintainer's ruling is slow, that half is separable and dispatchable on its own.

Dedup

Full-text scan of all open issues and PRs across the three repos (loadMetaFromDb, OVERLAY_ALLOWED_TYPES, resyncFlowsFromProtocol, applyRegistryWriteThrough, allowOrgOverride): no existing dispatch entry. Neighbours checked and distinct:

loadMetaFromDb object branch reads record.packageId from a snake_case row — always undefined, every object overlay registers under the 'sys_metadata' sentinel at boot #4636 (pm:dispatched, domain:engine-core) — also loadMetaFromDb, but the record.packageId snake_case read on the object branch. Different line, different failure; same function, so whoever takes this one should expect a same-file serial constraint against loadMetaFromDb object branch reads record.packageId from a snake_case row — always undefined, every object overlay registers under the 'sys_metadata' sentinel at boot #4636's PR.
Schedule 触发器不携带任何 org 上下文——定时 runAs:'system' 清扫在多组织部署里仍会产出 organization_id=NULL 的行(#5494 的用户缺席半边) #6155 (needs-user-decision, domain:services, target:v17) — where a scheduled flow's org comes from, not whether an org-scoped flow survives a restart. Correctly kept separate by the author.
ADR-0005 的租户可定制白名单表说 flow/workflow/approval 一律 ❌ per-org override,注册表却写着 allowOrgOverride: true——反转没有走 ADR(Prime Directive #13) #6191 — the gate; cross-linked both ways.

Cross-repo: nothing in objectui/cloud.

Release board

target:v17 on criterion ① and ④: a tenant publishes an org-scoped flow in Studio, it fires all day, and after the next process restart it never fires again with no log line anywhere — silent loss on a shipped path, and the kind of thing release notes would have to apologise for. Note the label is applied on the symptom, independent of which fix direction #6191 picks; if the ruling is "org flows should never have been writable", the board item becomes the write-side rejection plus whatever migration the already-stored rows need, which is still v17 work.

Body edit

Appended one line to the body — Blocked-by: objectstack-ai/objectstack#6191 — per the label discipline that pm:blocked carries its machine half in the body. No other text touched.

本评论来自分诊座位 Routine(#5474 试点),不构成认领。

Generated by Claude Code

迁移:domain:engine-core → domain:metadata(维护者 2026-08-07 批准拆分;座位贴 #6367,SKILL PR #6370)。落点证据:冷启动两条读路径滤掉 organization_id 非空行 —— loadMetaFromDb(protocol.ts:10620)+ getMetaItems org 分支,metadata 冷启动加载/水合。pm:blocked 状态与正文 Blocked-by: 行随卡走。全量分类审计 49 卡三判,本卡 MOVE/high。分诊会话:session_01BickTBKm2JYSNnrtPT8ysa。

Generated by Claude Code

Unblocked — pm:blocked → pm:queue (domain:metadata seat unlock sweep, session session_01KDU3qAuJyajAQm3GkUXdfA).

Upstream #6191 closed 2026-08-08 01:58Z via PR #6478: flow's allowOrgOverride reverted to false per the maintainer ruling chain (#6155 Q1=B → #6283 → PR #6478), with write-side declared=enforced proven (org-scoped overlay of a packaged flow → 403 NOT_OVERRIDABLE, refusal before persistence).

This materially narrows the card. The body's own framing said: if the ruling is that org flow overlays should not exist, "修法是在写入侧拒绝而不是在读取侧补齐" — that write-side rejection has now shipped. Remaining scope to verify at dispatch, against current origin/main:

Are there any paths that can still mint sys_metadata flow rows with organization_id set (runtime create was part of the old surface — re-check after fix(spec): flow 的 allowOrgOverride 回滚为 false —— ADR-0005 原判,写路径响亮拒绝 #6478)?
Legacy stored rows: org-scoped flow rows written before fix(spec): flow 的 allowOrgOverride 回滚为 false —— ADR-0005 原判,写路径响亮拒绝 #6478 — do they still silently unbind on restart (the original symptom), and does any migration/cleanup or loud absence log exist for them? (ADR-0005 的租户可定制白名单表说 flow/workflow/approval 一律 ❌ per-org override,注册表却写着 allowOrgOverride: true——反转没有走 ADR(Prime Directive #13) #6191's ruling ordered a stored-row migration assessment for its half-1 types; whether [转移·spec] flow 的 allowOrgOverride:true 回滚为 false——#6155 Q1=B 裁决的契约半边,同步了结 #6191 #6283/fix(spec): flow 的 allowOrgOverride 回滚为 false —— ADR-0005 原判,写路径响亮拒绝 #6478 covered flow's stored rows needs checking.)
If both answers are "closed/none", this card closes by verification with evidence — a legitimate outcome.

Dispatch will be verification-first accordingly.

Generated by Claude Code

Claim: PM loop round 3 (domain:metadata seat, sticker #6367)
Session: session_01KDU3qAuJyajAQm3GkUXdfA
Branch: claude/issue-6190-org-flow-legacy-rows
Worktree: objectstack-issue-6190
Domain: domain:metadata
File surface (verification-first; the fix half only if verification demands it): packages/metadata-protocol/src/protocol.ts — loadMetaFromDb where-filter region and/or getMetaItems org-record region — plus tests; possibly a legacy-row remediation whose shape is TBD by the verification. ⛔ packages/runtime/src/domains/meta.ts is #3682's in-flight surface — if the fix needs it, STOP and report (serialize). ⛔ Stored-data destructive migration requires needs_decision, not implementation. Stop on breach; explain in the report.
Serial constraints cleared: same-seam predecessor #4513 → PR #6561 MERGED (05:19Z); #5839 in flight (engine-core seat, protocol.ts ensureOverlayIndex region — different function region, merge-main discipline applies); #6215 → PR #6574 and #5309 → PR #6576 both ACCEPT'd and in the merge queue (no file overlap with this card); #3682 in flight (mode:cloud, rest-server.ts + runtime meta.ts + plugin-security — this card must not touch those files); #5079 (same listing seam) deliberately NOT co-dispatched — next slot, after this card's verification lands.
Container class: S/M (verification-first) — mode:subagent, shared container.

Generated by Claude Code

Verification result: the premise SURVIVES #6478 — narrowed, not closed. Draft PR #6600 ships the implementable half (the loud cold-boot log); the contract half needs a maintainer ruling and is stated at the bottom.

Session session_01KDU3qAuJyajAQm3GkUXdfA, branch claude/issue-6190-org-flow-legacy-rows. Everything below is measured against origin/main at a36db28b7 with the real ObjectStackProtocolImplementation over a stub sys_metadata engine, not inferred.

1. Minting paths — NOT closed

PR #6478 closed the overlay tier. Re-measured, that half holds:

PROBE P4 err  = {"code":"NOT_OVERRIDABLE","status":403,
                 "message":"[NOT_OVERRIDABLE] 'flow' is not allowOrgOverride in the registry..."}
PROBE P4 rows = []                       // refused BEFORE persistence

The runtime-create tier is open by design — PR #6478's own test says so ("a BRAND-NEW org flow still saves — allowRuntimeCreate is a different tier"), and that is ADR-0005's "a deployment, not an overlay". But that is the tier the tenant scenario in this issue actually uses: authoring a new flow in Studio, not overlaying a packaged one. SysMetadataRepository.put stamps organization_id: this.organizationId for every type, and runtime/src/domains/meta.ts (inspected read-only) still threads resolveActiveOrganizationId into saveMetaItem on PUT /metadata/:type/:name. So the row is still minted today:

PROBE P1 save result = {"success":true,
  "message":"Saved flow 'org_sweep' (org=org_a, state=active) [seq=1]"}
PROBE P1 rows = [{"type":"flow","name":"org_sweep","org":"org_a","state":"active"},
                 {"type":"flow","name":"platform_sweep","org":null,"state":"active"}]

A second path mints one too — draft + publishPackageDrafts under an active org:

PROBE P5 after draft rows   = [{"name":"draft_sweep","org":"org_a","state":"draft"}]
PROBE P5 publish            = {"success":true,"published":["draft_sweep"],"failed":[]}
PROBE P5 after publish rows = [{"name":"draft_sweep","org":"org_a","state":"active"}]

This is not a declared≠enforced hole in #6478's gate — the gate does exactly what it declares. The gap is that allowOrgOverride governs whether a packaged item may be overlaid, while the row's org scoping is decided by the caller and applied to every type regardless. Closing it is a contract decision, not a bug fix; see §4.

2. Legacy stored rows — the original symptom reproduces unchanged, and silently

(a) They still unbind:

PROBE P2 rows in store = [{"name":"org_sweep","org":"org_a"},{"name":"platform_sweep","org":null}]
PROBE P2 loadMetaFromDb = {"loaded":1,"errors":0,"invalid":0,"storeUnavailable":false}
PROBE P2 getMetaItems({type:flow}) no org       = ["platform_sweep"]
PROBE P2 getMetaItems({type:flow, org_a})       = ["platform_sweep","org_sweep"]

(b) Nothing logs the absence — this is the measurement that mattered most:

PROBE P2 logs during cold boot = []

console.warn/error/info were all captured across loadMetaFromDb. Zero lines. kernel:bootstrapped's unbound audit cannot see it either, because the flow was never registered.

(c) No stored-row migration shipped for flow. PR #6478's file list is: one changeset, metadata-plugin.zod.ts, three test files, and overlay-precedence.test.ts — no migration. packages/metadata-protocol/src/migrations/ holds only recorded-by-sentinel and view-definition-active-index, neither related. So #6191's migration assessment was not extended to flow's stored rows, and — per §1 — a migration alone would not settle it anyway, since new rows are still being written.

3. Impact ② — real, wider than flow, filed separately as #6602

The write-through gate was not hardened; it still reads if (this.environmentId !== undefined) return; with nothing about org. Measured on a still-org-overridable type:

PROBE P3 rows = [{"type":"view","name":"org_grid","org":"org_a"}]
PROBE P3 registry write-through = [{"type":"view","name":"org_grid"}]

and the read side has the same hole (getMetaItems hydrates orgRecords into the process-wide registry whenever environmentId === undefined), so the TSDoc's "the write must not be more permissive about that than the read is" is no longer the right framing — both are permissive, and only loadMetaFromDb holds the line. Filed as #6602 (unassigned, search-first: no twin; #6483 is orthogonal — it asks which types should carry the flag, this asks what happens to the rows of the ones that legitimately do).

4. What shipped, and what still needs a ruling

Shipped (draft PR #6600, packages/metadata-protocol only): cold boot now prints one aggregated warn naming the org-scoped rows it walked past, for the types the registry declares NOT per-org overridable — with counts, a sampled name@org, the consequence ("A 'flow' listed here will NOT bind its triggers in this process") and the remedy. Silent for view-class types, where the skip is the ADR-0005 design. Hydration behaviour is unchanged. This is impact ①'s "Absence must be loud" half, and it is implementable without a ruling.

Needs a ruling (three options, none guessed at):

Refuse the write. saveMetaItem rejects an org-scoped write for any type whose registry entry is allowOrgOverride: false, so the tenant is told at the moment of the write instead of at the next restart. Contract-first and symmetric with fix(spec): flow 的 allowOrgOverride 回滚为 false —— ADR-0005 原判,写路径响亮拒绝 #6478's direction; it also removes org-scoped object/field/hook rows, which the same path mints today, so its blast radius is much larger than flow.
Force it env-wide. Strip the org for non-org-overridable types, so a tenant-authored flow lands organization_id = NULL and binds after every restart. Preserves the capability but changes visibility — the flow becomes environment-wide, which in a multi-org deployment is a different tenancy statement than the author made.
Teach the binder to read per-org. Ruled out by this dispatch as dead by Schedule 触发器不携带任何 org 上下文——定时 runAs:'system' 清扫在多组织部署里仍会产出 organization_id=NULL 的行(#5494 的用户缺席半边) #6155 Q1=B, and recorded here only so the option set is complete.

My recommendation is 1, on this project's three axes: it is the only option that keeps declared = enforced (the registry says the type has no per-org channel, so the store should not hold a per-org row of it); it makes AI-authored metadata fail loudly at authoring time rather than silently at the next restart; and it needs no business pull to justify, since it removes a surface rather than adding one. Option 2 buys back a capability nobody has been measured to want and silently re-scopes a tenant's data. But 1 is a real behaviour tightening across several types — the maintainer's call, not mine.

Generated by Claude Code

fix(metadata-protocol): scan the live metadata-type registry in the c…

fix(metadata-protocol): resolve the draft's own org scope on a per-it…

fix(metadata-protocol): per-item publish re-binds runtime consumers a…

docs(adr): ADR-0126 — the platform customization model for packaged m…

fix(rest): the untyped /meta/diagnostics sweep forwards the caller's …

fix(metadata-protocol): getMetaItemLayered gates the org read after t…

docs(rest): state the direction the read predicate rules on at the /p…

fix(metadata-protocol): gate `getMetaItem`'s overlay read on the meta…

rest/meta: state the caller's organization on the `/meta/diagnostics`…

fix(rest): the untyped /meta/diagnostics sweep forwards the caller's …

fix(metadata-protocol): `getMetaItemLayered` gates the org read, boun…

docs(rest): state the direction the read predicate rules on at the /p…

fix(runtime): a self-hosted restart reads the kernel's own sys_metada…

fix(metadata-protocol)!: the ADR-0010 _lock gate reads the row the re…

Navigation Menu

Repository navigation

org 作用域的 flow overlay 只在「本进程内发布后」绑定触发器,重启后静默失绑——冷启动两条读路径都把 organization_id 非空的行滤掉了 #6190

事实(实测,非推断)

影响

与 #6155 的关系

复现

Classification — real, well-located, but its fix direction is gated

Routing rationale (landing site, not title vocabulary)

One part of this is a defect under either ruling

Dedup

Release board

Body edit

1. Minting paths — NOT closed

2. Legacy stored rows — the original symptom reproduces unchanged, and silently

3. Impact ② — real, wider than flow, filed separately as #6602

4. What shipped, and what still needs a ruling

26 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

org 作用域的 flow overlay 只在「本进程内发布后」绑定触发器,重启后静默失绑——冷启动两条读路径都把 organization_id 非空的行滤掉了 #6190

Description

事实(实测,非推断)

影响

与 #6155 的关系

复现

Activity

claude commented on Aug 7, 2026

Classification — real, well-located, but its fix direction is gated

Routing rationale (landing site, not title vocabulary)

One part of this is a defect under either ruling

Dedup

Release board

Body edit

hotlong commented on Aug 7, 2026

baozhoutao commented on Aug 8, 2026

baozhoutao commented on Aug 8, 2026

baozhoutao commented on Aug 8, 2026

1. Minting paths — NOT closed

2. Legacy stored rows — the original symptom reproduces unchanged, and silently

3. Impact ② — real, wider than flow, filed separately as #6602

4. What shipped, and what still needs a ruling

26 remaining items

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions