[objectstack-fleet]

Requested by Bill · project thread

domain:server#2 · os-bill · session_01Utp6CYSw4nSVJDr6XpcWLJ

Fixes #30319

Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么： 复制包（duplicatePackage）时，复制出的权限集把源权限集在库中存储的授权对象名作为 inheritedGrantObjects 交给保存门，与 main red: Dogfood Regression Gate (2/3) fails permission-set-clone-door after PR #30222 (member_default grants sys_comment) #30255 的权限集克隆门同一条规则。没装 plugin-audit 的部署里，授权 sys_comment 的权限集现在能随包一起复制；非托管导入按席位裁定 A 保持拒收，未改。
• 为什么改： 修 fix(runtime, metadata-protocol): the permission-set clone door copies a set that grants on an object this deployment does not install, and member_default's feed grants re-land #30310 时只改了克隆门，复制包这扇门仍把这类权限集放进 failed[]，整次复制回 success: false，用户复制一个正常的包却丢了权限集。
• 风险与回滚： 只放行源行本来就持有的授权键；被命名空间改写过的键（源包自己的对象改名为副本的对象）不算继承，仍照常判；键只从库中的源行取，不读请求。非托管导入的包是请求体，不是存储行，所以仍逐条判（裁定 A）。回滚即撤 PR。
• 先看： packages/metadata-protocol/src/protocol-package-commits.ts 的复制门 saveMetaItem 调用（inheritedGrantObjects 一行）与文件头的 storedGrantObjectKeys；再看 protocol.copy-doors-inherited-grants.test.ts 的第二个用例（改名后的授权仍拒）。
• 两分钟验证： pnpm --filter @objectstack/metadata-protocol exec vitest run src/protocol.copy-doors-inherited-grants.test.ts，预期 4 个用例全过。
• 需要你做的： 无，席位落地。

What changes

• duplicatePackage (protocol-package-commits.ts): each copied permission item is saved with inheritedGrantObjects = the keys of the SOURCE row's objects map as stored (storedGrantObjectKeys(body), the raw parsed row, before conversion and rewrite). The save door's unknown-object refusal (metadata: a runtime permission-set save accepts a grant on an object that does not exist (200) — the build refuses the same grant (#18202), so a typo silently grants nothing #28990) already skips inherited keys since PR fix(runtime, metadata-protocol): the permission-set clone door copies a set that grants on an object this deployment does not install, and member_default's feed grants re-land #30310 (protocol-save.ts, unknownGrantObjectSaveRefusal), so a set granting on an optional plugin's object this deployment does not install (sys_comment / sys_activity, @objectstack/plugin-audit) is copied instead of landing in failed[].
  • The keys come from the stored row only; duplicatePackage's request has no such field, exactly as the clone door's (triage 6651966154: "server-derived from the SOURCE set as stored, never read from request input").
  • A key the copy's namespace rewrite renames (leave_ticket → leave_copy_ticket) is not the source's key, so it is judged as before: when the copied object is not there, the set is still refused, naming that grant alone.
  • storedGrantObjectKeys restates the clone door's inheritedGrantObjectsOf (two lines): that helper lives in @objectstack/runtime, which depends on this package, so it cannot be imported here.
• importUnmanagedPackage: unchanged in behaviour, with a comment at its save call. By the seat's ruling on Package duplicate and unmanaged import refuse a copied permission set that grants on an optional plugin's object not installed here (same 422 the clone door had, #30255) #30319 (option A), its package is the request's body, not a stored row, so its permission sets are new writes and every grant is judged; deriving inherited keys from the body would read request input and turn the refusal off for this door.
• One changeset, patch for @objectstack/metadata-protocol (Metadata-protocol: none). No published schema moves; a save that was refused now succeeds (Clause-②: no).

Measured first (on main at 870ebce3, before the fix)

Harness: the stateful sys_metadata stub of the new test file, registry registering no object (no plugin-audit), a source package com.example.leave holding object leave_ticket and set leave_staff granting leave_ticket + sys_comment.

Duplicate door — refused (200, success: false, the set in failed[])

"success": false, "copiedCount": 1, "failedCount": 1,
"copied": [{ "type": "object", "name": "leave_copy_ticket" }],
"failed": [{ "type": "permission", "name": "leave_staff",
  "error": "permission/leave_copy_staff failed spec validation: objects.sys_comment: Permission 'leave_copy_staff' grants on object 'sys_comment' which is not defined in objects: no package registers an object of that name and none is saved here, so the grant applies to nothing. …" }]

Unmanaged import — refused (422 INVALID_METADATA, rolled back whole)

422 INVALID_METADATA The unmanaged install of 'com.example.crm' was rolled back: permission 'crm_rep' could not be written — permission/crm_rep failed spec validation: objects.sys_comment: … the new base 'com.example.crm-copy' was removed, so nothing of the package remains.
issues: [{"path":"objects.sys_comment", "code":"custom", …}]
sys_metadata rows after: []

Tests

packages/metadata-protocol/src/protocol.copy-doors-inherited-grants.test.ts, 4 cases:

1. Duplicate: a set granting on sys_comment (absent) and the source's own object is copied; the copy grants leave_copy_ticket + sys_comment.
2. Duplicate: a grant the rewrite renames onto an object that did not copy (leave_copy_ticket, its source row unparseable) is still refused — the exact failed[] entry, one finding at objects.leave_copy_ticket, sys_comment not judged; nothing stored.
3. Duplicate: a grant the source holds outside its namespace on an absent object (other_gone, not renamed) is the source's and is copied.
4. Unmanaged import: a set granting on sys_comment is refused 422 INVALID_METADATA at objects.sys_comment (code, status, path, first sentence) and the import is rolled back (no row, no base) — pinned as intended by ruling A.

Ablations (each through scripts/ablation-replace.mjs, mutation proven on disk, restored to the HEAD blob with git diff HEAD empty; the test imports ./protocol.js from source, so no dist/ is involved):

Mutation	Expected	Observed
A. drop the inheritedGrantObjects argument at the duplicate door	1, 3 red	1, 2, 3 red (2 gains the sys_comment finding)
B. derive inherited keys from the REWRITTEN copy instead of the stored row	2 red	2 red (renamed grant passes), 1/3/4 green
C. pass the import body's own keys as inheritedGrantObjects at the import door	4 red	4 red (import resolves, no 422)

A first attempt at C was a no-op refused by the tool (replacement contained the anchor; anchor count 1 → 1); re-run with a distinct replacement.

Verification

All on this branch; the code under test is identical from 83a51faa to the head (later commits: changeset wording, the engine-double ledger row, a merge of main).

• New pins: pnpm --filter @objectstack/metadata-protocol exec vitest run src/protocol.copy-doors-inherited-grants.test.ts src/protocol.unknown-grant-object-refusal.test.ts on c718acd3 (after merging main at 7d7babb1): 23/23.
• Package suite: pnpm --filter @objectstack/metadata-protocol test on c718acd3: Test Files 315 passed, 4 skipped; Tests 22410 passed, 26 skipped. pnpm --filter @objectstack/metadata-protocol typecheck: exit 0 (--listFiles includes the new test file).
• Full dogfood (pnpm --filter @objectstack/dogfood test under the verify lock, closure built first, no shard): Test Files 419 passed, 3 skipped (422); Tests 3492 passed, 36 skipped (3528). Lock VERDICT command-exit 0, held 29m40s.
• Derived gates (dispatch-gates.mjs --commands, 67 families on 4463c9f7): 66 exit 0, 1 NOT MEASURED — check:dual-build-cjs-loads exit 3, PREREQUISITE NOT MET (studio, client-react and five plugin/service packages have no dist/ in this worktree); left to CI. --ran reconciliation: "67 derived famil(ies) accounted for — 66 run, 1 NOT-MEASURED". check:engine-double-contract asked for the new test file's doubles to be recorded: --write added one seam row set, committed. After the merge, node scripts/check-engine-double-contract.mjs exit 0.
• pnpm lint: exit 0 on 4463c9f7.

Main's known red (check-liveness query/searchFields, #30367 / PR #30368) is not touched by this diff.

Human Minimum

• 1. Schemas start as Zod; packages/spec holds no runtime logic. (No spec change.)
• 2. Every name I added uses the casing for its kind.
• 3. No formatter ran; pnpm lint passes.
• 4. One issue, one branch off an up-to-date main, one pull request.
• 5. A change that publishes carries a changeset, and both declaration lines are filled in here and in the changeset.
• 6. packages/spec touched: n/a.
• 7. No hand edits to content/docs/references/, packages/*/CHANGELOG.md or content/docs/releases/.
• 8. Title, body and comments are in English (the 维护者速读 block is the dispatch contract's Chinese summary).
• The Before Submitting list in CONTRIBUTING.md is done.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Utp6CYSw4nSVJDr6XpcWLJ

---

Generated by Claude Code

[objectstack-fleet]

domain:server#2 · os-bill · session_01Utp6Cbvcmv8YWFhFJwzjKL

维护者速读

• 改了什么： 复制包时，包里的权限集如果授权了本部署没装的可选插件对象（如没装审计插件时的 sys_comment），现在照样复制，规则与权限集「克隆」门一致（main red: Dogfood Regression Gate (2/3) fails permission-set-clone-door after PR #30222 (member_default grants sys_comment) #30255）；非托管导入按裁定仍拒收。
• 为什么改： 以前复制包时这类权限集会进 failed[]，复制结果报失败（Package duplicate and unmanaged import refuse a copied permission set that grants on an optional plugin's object not installed here (same 422 the clone door had, #30255) #30319，fix(runtime, metadata-protocol): the permission-set clone door copies a set that grants on an object this deployment does not install, and member_default's feed grants re-land #30310 复审的跟进项）。
• 风险与回滚： 放行的只限源存储行本来就有的授权键，从库里读，从不读请求；复制时改名到新命名空间的授权照判；导入门不变。回滚即撤本 PR。
• 先看： packages/metadata-protocol/src/protocol-package-commits.ts 的 storedGrantObjectKeys 与复制门那一行。
• 两分钟验证： pnpm --filter @objectstack/metadata-protocol exec vitest run src/protocol.copy-doors-inherited-grants.test.ts。
• 需要您做的： 无，席位经合并队列落地（非受管面，308 行）。
• 席位意见： ACCEPT。声明面复审由席位在 TIER_DEFAULT 完成（见下），PASS。

席位核对 — ACCEPT · domain:server#2 · head c718acd3

Served-tier: TIER_DEFAULT (declaration face: Clause-②: no, Metadata-protocol: none, no packages/spec/** and no governed path). Implemented-by session_01Utp6CYSw4nSVJDr6XpcWLJ; Reviewed-by session_01Utp6Cbvcmv8YWFhFJwzjKL.

• 形态: draft, author objectstack-fleet[bot] (relay pr_create), no assignee, no review request; 4 files, +308 −0, merged with main 7d7babb1; check-governed-merges --pr 30414: NOT governed (0 of 4 paths), 308 lines ≤ 3000.
• Diff read (seat): at the package-duplicate door, inheritedGrantObjects is the objects keys of body, the SOURCE row as read from the store (before deepRewrite), passed only for singular === 'permission'; a key the rewrite renames onto the target namespace is not in that list and is judged by the save door as before. The unmanaged-import door passes none, with a comment naming the ruling (seat ruling on the card thread, 08:58Z: the import package is the request's body, so its sets are new writes). storedGrantObjectKeys restates the clone door's inheritedGrantObjectsOf because @objectstack/runtime depends on this package (no reverse import). scripts/engine-double-contract.pinned.json adds the new test file's three double verbs, as that gate requires (a file outside the claim's surface, accepted).
• Declarations: Clause-②: no holds: a copy the clone door already accepts stops being refused at the sibling door; no export, error code or schema moves. Metadata-protocol: none; changeset patch for @objectstack/metadata-protocol, same two lines in body and changeset.
• Pins: three duplicate-door cases (absent optional object copied; renamed grant onto a missing object still refused alone; source grant outside its namespace copied) and the import-door refusal with rollback. Dev: ablations A (argument dropped), B (keys from the rewritten copy), C (import fed the body) red then restored; package tests 22410 passed, typecheck 0, lint 0, full dogfood 419 / 3492 passed; derived gates 66/67 (dual-build-cjs-loads left to CI).
• 门禁 check-runs: 9 success, 13 skipped, 15 in progress, 0 failed at ACCEPT; auto-merge waits on the required contexts.
• 落地条款: Fixes Package duplicate and unmanaged import refuse a copied permission set that grants on an optional plugin's object not installed here (same 422 the clone door had, #30255) #30319 (both doors measured; the import door stays refused by ruling and is pinned). Ordinary queue landing by this seat: pr_ready + automerge_enable on this head.

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/metadata-protocol, touching 3 documentable anchor(s).

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

• content/docs/releases/v18/18-3.mdx (via duplicatePackage (symbol, a method of class ProtocolPackageCommitsFacet))

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

• 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 — 10 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 6fa842af1e3137c93d7de2c823965c52fc5d524c → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 8b82348cac3ed90b31c5d180783052312ba44e0d — the merge of head c718acd35546ac254449a4966e3f9616a1e2e8b3 into base 6fa842af1e3137c93d7de2c823965c52fc5d524c, 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 8b82348cac3ed90b31c5d180783052312ba44e0d && git checkout 8b82348cac3ed90b31c5d180783052312ba44e0d
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin 6fa842af1e3137c93d7de2c823965c52fc5d524c c718acd35546ac254449a4966e3f9616a1e2e8b3 && git checkout -B drift-repro 6fa842af1e3137c93d7de2c823965c52fc5d524c && git merge --no-ff c718acd35546ac254449a4966e3f9616a1e2e8b3

node scripts/docs-audit/affected-docs.mjs --json 6fa842af1e3137c93d7de2c823965c52fc5d524c

⚠️ 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 6fa842af1e3137c93d7de2c823965c52fc5d524c → pass the list as
args.docs, on the commit named under Which tree this was computed on.