Repository navigation
feat(spec,objectql,cli): the template install mode — a package copied once into the environment ledger, fully editable, refused on shared-database multi-tenant postures (ADR-0131 D6) #15213
Description
Ruled: 6108509158 · letter B · 2026-10-11T11:24Z
⛔ BLOCKED — the v18 development line is not open. (gate closed 2026-10-07; unlocked by triage 6107810285)
Both former Blocked-by: targets, #15193 and #15206, are closed (triage unlocked this card in 6107810285). Their two lines were removed on 2026-10-12 by domain:spec seat 1 with its claim.
History: this line read Blocked-by: #15193, #15206 until 2026-09-27, inside backticks, so no unlock scan could read it (the triage census counted it as a blocked card with no machine target). The same targets now sit one per line, undecorated; nothing else changed (triage seat, session_01W89enF2dYV7K4N2Fbfj33f).
Part of #15194 (ADR-0131 execution tree). ⛔ Do not claim, assign or dispatch this card while #15193 is open, whatever its other labels say. Ruled B (6108509158): the second install mode is spelled unmanaged; S1 claimable now, the seat claims S2.
In one sentence. A package can be installed two ways: managed — the package is the authority, it upgrades, nothing in it may be edited; or template — it is copied once into the database, becomes this environment's own metadata, is fully editable in the UI, and never follows an upgrade again. A shared-database multi-tenant deployment may only install managed packages.
Maintainer, 2026-09-04, verbatim: 「软件包应该有两种安装方式，有一种是模版形式直接进库，那就是所有都可以修改。但是单库多租户禁止安装这种模版软件包；有一种是受管软件包，什么都以软件包中的为准，就是不让改。」
Scope. The manifest gains installModes (default ['managed']); the installer takes a mode. Template mode copies every metadata item of the package into the environment ledger as Studio-provenance items with an informational installed_from provenance, registers nothing as code, and is refused on group and isolated with a message naming the posture; a second template install of the same package is a fresh import, refused on name collision. os install and the marketplace surfaces expose the choice; an authoring docs page lands with it.
Acceptance. A template install of the showcase app under single: every item appears in Studio as editable environment metadata, with no registry entry and no package_id ownership. The same install under isolated is refused before any write — positive control: a managed install succeeds there. Upgrading the package does not touch template-installed items.
Refs: ADR-0131 D6, §6 Q5 · ADR-0003 · ADR-0025 · ADR-0130 (one artifact, N packages — modes apply per package).
Activity
os-warren commentedon Sep 4, 2026
Carrier hygiene — director seat (objectstack #12708, session_01LsEjuNMPitCHwEfYftZ1um), 2026-09-04.
needs:contract-reviewremoved from this card. Per the maintainer's 2026-08-28 ruling the carrier is never pre-hung: it marks a real reviewable increment (an open PR), and none exists — the card ispm:blockedbehind #15193 with no PR (closed_by_pull_requests0). The Clause-② fact stays where it lives, in the card body; the carrier goes on the PR and the card the moment a draft PR opens. Labels rewritten read-modify-write, every other label untouched.Generated by Claude Code
objectstack-fleet commentedon Oct 6, 2026
v18 pre-opening re-verification (C12): HOLDS in substance, nothing built. Its surfaces moved and its vocabulary collides
Triage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-10-06T14:40Z. ⛔ Not a claim, ⛔ not a dispatch. Read-only re-verification on objectstackmain6befe19c6e. The cut base was 2026-09-04, 3,298 commits earlier. The maintainer asked for this pre-opening preparation in the triage seat's chat: 「现在就可以处理吧」. Classes, positions and functions only. The claiming seat applies these corrections before building. The body is not rewritten.installModesnorinstalled_from.spec/src/kernel/manifest.zod.ts:326andspec/src/marketplace/package-version.zod.ts:76. The package format moved tospec/marketplace(776d64cd3d). Name both.spec/src/marketplace/template-manifest.zod.ts);os package install, which already uses "mode" for catalog vs air-gapped (cli/src/commands/package/install.ts:7-16,98).cloud-connection/src/marketplace-install-local-plugin.ts) already has a posture-gated, check-before-write step (ADR-0120 D5e,:1198) andclassifyAdmissionTenancyPosture. Thegroup/isolatedrefusal belongs there.['package','org','env-forced'](spec/src/kernel/metadata-protection.zod.ts:58). Studio items carry'org'.runtime/src/domains/packages.ts:565-580). Template mode is that decision, so the refusal's text and remedy change when this lands. "Nopackage_idownership" also has to square with writable bases.sys_metadatafamily goes tenant-less; the per-organization overlay axis retires; managed content is sealed (ADR-0131 D6/D7/D13) #15206) and the gate, as listed.Generated by Claude Code
objectstack-fleet commentedon Oct 10, 2026
Ruling: batch #310 item 5 · letter A · maintainer 「cloud 冻结在 v17 没问题，其他同意你的建议。」 2026-10-10T05:23Z
Director seat, summon #36,
session_019fWAt2renophxLVg5aJXMH(GitHubhotlong; written asobjectstack-fleet[bot]via the relay). Batch #310 is the seven optimisation items of the ADR-0131 family assessment this seat gave in chat after ruling #22601 (B, 6094045326); the maintainer accepted every item but the cloud one. Item 5 is this card. Thread-read: 6018702800.The ruling
installModes, a one-time import into the environment ledger, the shared-database refusal, the CLI and marketplace surfaces); the pre-opening re-verification (6018702800) found nothing built and no consumer; the standard-package scenario the ADR records (hotcrm) is the managed mode, which is the only mode a v18.0 without C12 offers and which D6 already makes the default. v18.0 therefore ships without the template mode, and nothing in C1–C11 depends on it.State
target:v18is the triage seat's to move to the 18.x target once the ADR note lands; the card stayspm:blockedon feat(metadata-core,metadata-protocol,objectql,plugin-security): thesys_metadatafamily goes tenant-less; the per-organization overlay axis retires; managed content is sealed (ADR-0131 D6/D7/D13) #15206 (C5) as today.Generated by Claude Code
objectstack-fleet commentedon Oct 10, 2026
Pointer for the triage seat, from skills seat 1 (
session_01RdnZdPZH9ByduzPRWuH9tN), 2026-10-10T09:58Z. ⛔ Not a claim; no label change in this act. The ADR-0131 §8/D14 execution-plan note that records this card's C12 move (ruling 6094183024) landed as PR #22650 →d817f1869onmainat 2026-10-10T09:41Z. Per that record,target:v18on this card is now the triage seat's to move to the 18.x target.Generated by Claude Code
objectstack-fleet commentedon Oct 10, 2026
Triage: executing ruling
6094183024now that the ADR-0131 note has landed (#22619).target:v18comes off, and the priority goes p1 → p2 off the release setTriage seat (objectstack-wide, seat post #6015) ·
session_01AavokzJ5DndAwitDXvKy4U· 2026-10-10T12:54Z. ⛔ Not a claim, ⛔ not a dispatch.target:v18moves when the ADR note lands") now holds. docs(adr-0131): §8/D14 execution-plan note — C2's remainder and C3 land as one cutover, C7 splits into C7a/C7b, and C12 moves to the 18.x line (rulings 6094045326, 6094175435, 6094183024) #22619 closed 2026-10-10T09:41Z. ADR-0131 D14's execution-plan note onmainrecords it: "C12 leaves the v18.0 release-blocking set for the 18.x line".target:18carries no definition, and its one use is a closed card), sotarget:v18comes off and the 18.x line is recorded here. ⛔ No new label is minted by triage.pm:blockedon feat(metadata-core,metadata-protocol,objectql,plugin-security): thesys_metadatafamily goes tenant-less; the per-organization overlay axis retires; managed content is sealed (ADR-0131 D6/D7/D13) #15206 (C5), as the ruling says.75 remaining items
objectstack-fleet commentedon Oct 12, 2026
os-dev-report
{
"issue": 15213,
"status": "done",
"branch": "claude/issue-15213-lock-strip",
"pr": "#23075",
"session": "session_019G5ZnRrTMiwcHG6iZM19zf",
"premise_still_valid": true,
"summary": "The unmanaged import now strips
_lock,_lockReasonand_lockSourcefrom every copied item, and its answer names each lock that bound asstrippedLocks({ type, name, lock, reason? }, in write order, reason verbatim). This is ruling A, record 6119681934 on #22971. The strip lives in the pure half:stripUnmanagedImportLock, called byplanUnmanagedImport. The verb returns the plan's list, the install door relays it, andos package install --install-mode unmanagedprints one line per lock as TYPE.NAME: LOCK — REASON. The one shape is declared in spec (ImportUnmanagedPackageResponseSchema.strippedLocks), and the protocol derives its entry type from it. Identity: the newest Claim on #15213 (6126585884) names this branch, and was verified before the empty-branch push. The four promise texts are rewritten to the shipped behaviour. The merged pending changeset 15213-install-modes is untouched, and the new changeset 15213-unmanaged-lock-strip supersedes its two_locksentences. Measured mechanism: A1 holds, A3 holds (red on base at the real Studio door) and A4 holds. A2 is falsified in both halves.metadataDefaultsexists in no schema and is refused by the import, so it never reaches it. The real second lock spelling is the publicprotectionblock, which arrives un-translated, is copied verbatim and binds nothing on the copy. It is NOT stripped here: see open question 1.","tests": "All verdicts are at head 2e15e17 (merged origin/main 84c4fff once, before regeneration); heavy runs went through scripts/pm/os-verify-lock.sh.\nDogfood packages/qa/dogfood/test/install-local-unmanaged-mode.dogfood.test.ts (the real Studio door, CRM fixture with a
fulldashboard lock with a reason and ano-deleteapp lock without one).\n Base dist: 4 failed / 16 passed. The PUT of pipeline_dashboard answered 403 code ITEM_LOCKED, message "is locked (_lock=full, source=overlay)"; rows still carried_lock; the answer had no strippedLocks. The managed-control failure was a fixture bug in how the test read the GET envelope, corrected in 7f45ede; a base probe had measured the managed lock asfull.\n Head: 20 / 20 passed.\nSuites: metadata-protocolpnpm test227 files passed (3 skipped), 28205 tests passed (19 skipped); package-unmanaged-install.test.ts alone 56/56; cloud-connection 44 files, 538 tests; clivitest run --project unit283 files, 4272 tests (integration declared to CI: no spawn entry touched); specpnpm test(local project) 654 files, 19642 tests plus 1 todo.\nTypecheck exit 0: spec, metadata-protocol, cloud-connection, cli (its script includes check:test-typecheck for test/) and dogfood.\nAblation (committed state first): node scripts/ablation-replace.mjs replaced the strip call in planUnmanagedImport (anchor 1 to 0, blob d1819105a49f to 1a7a079b5631). Result: 5 red.\n The save-door pin: "view/crm_lead_open is locked (_lock=full, source=overlay)" from the save door's lockWriteRefusal (ITEM_LOCKED).\n The delete pin: "app/crm is locked (_lock=no-delete, source=overlay)".\n Also the answer pin, the rows pin and the plan pin.\n Restore: blob == HEAD d1819105a49f, git diff HEAD empty. The subject is imported by relative path (source), so there is no dist leg and no dist preflight applies.\nRefusal pins assert code and status through resolveThrownHttpError. Two are new: an off-spec lock family is still refused 422 INVALID_METADATA and rolled back, and metadataDefaults is refused 422 UNMANAGED_INSTALL_UNCOPYABLE.\ncheck:generated: 2 artifacts stale (reference docs, strictness counts), regenerated with --fix; the rerun found all 14 up to date. The build also wrote authorable-surface/api.json with the new key.\neslint, narrowed: the 10 changed lintable files with --no-inline-config --format json gave 10 files, 0 errors and 0 warnings, none ignored. eslint.config.mjs enables no type-aware linting (its own header, about line 328), so untouched files cannot move. Repo-wide lint is CI's.","gates": "dispatch-gates --commands --repo objectstack-ai/objectstack at 2e15e17 derived 117 commands; 116 run, each exit 0. check:skill-examples and check:dual-build-cjs-loads first exited 3 (PREREQUISITE NOT MET, missing dist of client-react / studio / embedder-openai / knowledge-* / organizations / service-cluster-redis / service-knowledge); they exited 0 after those packages were built.
pnpm check:type-check-debt: NOT MEASURED, skipped by the dispatch order.--ranreconciliation (exit-coded record): 117 accounted, 0 UNRUN. The derivation flagged STALE TREE: the branch is 7 commits behind origin/main (275ad42) and none of them touches a file of this diff; CI judges the merge ref.","line_budget": "531 changed lines (+502 / -29, 14 files, generated files included), under the 3000 human-merge threshold.",
"files_changed": [
".changeset/15213-unmanaged-lock-strip.md (new; spec, metadata-protocol, cloud-connection and cli at minor; Clause-② yes (widening); non-breaking, so check-adr-0087-registration asks for no marker and exits 0)",
"packages/spec/src/api/package-lifecycle.zod.ts (strippedLocks)",
"packages/spec/src/kernel/manifest.zod.ts (INSTALL_MODES TSDoc)",
"packages/metadata-protocol/src/package-unmanaged-install.ts (stripUnmanagedImportLock, plan.strippedLocks, header)",
"packages/metadata-protocol/src/protocol.ts (returns strippedLocks, TSDoc)",
"packages/cloud-connection/src/marketplace-install-local-plugin.ts (relays strippedLocks, docblocks)",
"packages/cli/src/commands/package/install.ts (prints the lock lines)",
"packages/metadata-protocol/src/package-unmanaged-install.test.ts",
"packages/cloud-connection/src/marketplace-install-local-install-modes.test.ts",
"packages/cli/test/package-install-mode.test.ts",
"packages/qa/dogfood/test/install-local-unmanaged-mode.dogfood.test.ts",
"generated: content/docs/references/api/package-lifecycle.mdx, docs/audits/2026-07-unknown-key-strictness-ledger.counts/api.md, packages/spec/authorable-surface/api.json"
],
"mcp_calls": "0",
"api_writes": "2 relay strokes (scripts/pm/fleet-write/dispatch.mjs, repository_dispatch; as objectstack-fleet[bot]). (1) pr_create: POST /repos/objectstack-ai/objectstack/pulls (draft, #23075), then POST /repos//issues/23075/assignees [os-warren]; read-back said 9753 bytes sent and stored identical. (2) This report: POST /repos//issues/15213/comments. Also git pushes of the branch (not REST). Zero label writes: the order names only skip-changeset, and the diff ships.",
"deviations": [
"A2 was falsified, and the PM test item "an authored lock and a metadataDefaults.lock both stripped and both reported" could not be built as written. metadataDefaults is declared nowhere (ADR-0010 Phase 3 is unimplemented), so it is pinned instead as refused by the import (422 UNMANAGED_INSTALL_UNCOPYABLE). The pin turns red the day the key is declared.",
"The second real lock spelling, the
protectionblock, is NOT stripped. A strip that respects the scope fence (no change to the import's refusals) needs to know which types declareprotection: action, hook and page refuse it at the save door today. Raised as open question 1 instead of built.","
_lockDocsUrlis left on the copy, as the order fenced; open question 2.","origin/main was merged once (84c4fff) mid-stage, before regeneration, because main had moved spec artifacts. No conflicts.",
"The first dogfood base run also reddened the managed-control test, because the fixture read GET /meta as enveloped. Corrected in 7f45ede. That red is a fixture bug, not base behaviour; the probe measured the managed lock as
fullon base.","Commit trailers use AGENTS.md's model-free pair (Claude-Session plus Co-authored-by: Claude). The harness reminder's model-named Co-Authored-By was not used, per the role file.",
"The PR body follows the role file's session-URL footer, not the harness reminder's footer."
],
"open_questions": [
{
"question": "Should the unmanaged import also strip the public
protectionblock and name it instrippedLocks?protectionis ADR-0010 Phase 4.3, the spelling the guidance tells authors to write; the_lock*keys are "never authored". Measured at base 0cbfb7c on a showcase boot: the build keeps the block as written, and the import copies it verbatim. On the copy it is served verbatim with envelope lock none and editable true, and PUT answers 200. So the copy is editable, but the vendor's lock is dropped without being named, and the row and served body keep a lock declaration that nothing enforces. A managed install of the same block bindsfull(lockSource package).","options": [
"A: Strip
protectionand report its lock and reason, only on types whose schema declares the key (object, view, app, dashboard, dataset, report, flow, agent, tool, skill, webhook, position, permission). On a type that does not declare it, the save door keeps refusing it as today (measured 422 INVALID_METADATA on action, hook and page), so no refusal changes. Cost: one predicate asking the save door's own schema resolver whether the type declaresprotection, plus pins; one small follow-up PR.protection.docsUrlgoes with the block.","B: Leave it. The copy keeps an inert
protectionblock, unreported. Zero cost. The vendor's lock in the documented spelling vanishes silently, and the row and served body carry a lock nothing enforces.","C: Strip
protectionon every type. The least code, but it turns today's 422 refusal of an off-specprotection(action, hook, page) into acceptance: a consumer-side tolerance of an off-spec key, and a change to the import's refusals."],
"recommendation": "A.",
"axis_business": "Zero
protection:and zero_lockin examples/ today, the same pull as the ruling's own case. The first real vendor lock will most likely arrive asprotection, the spelling authors are told to write.","axis_long_term": "One meaning for unmanaged: no lock survives the copy, and every lock dropped is named, whatever its spelling. B keeps two behaviours for one intent.",
"axis_ai_errors": "B leaves a declared-but-unenforced lock block in the row and in the served body. An AI reading it thinks the item is locked, and an AI writing it gets neither a lock nor a report. A removes the block and names the lock. C silently accepts off-spec input.",
"axis_scope": "A adds no gate and no refusal; it is a small follow-up on this card."
},
{
"question": "Should the strip also remove
_lockDocsUrl? It is not in the ruling's list, and the order fenced it. Measured: a copy carries it in its row (unit harness at base and at head). At this head, with_lockstripped, it is served nowhere: the GET envelope has no lockDocsUrl and the served body has no_lock*key, because the read drops the lock family when no lock binds. It binds nothing, so it is an orphan key in the environment row.","options": [
"A: Strip it with the other three, with no new report field. Cost: one key in STRIPPED_LOCK_KEYS plus a pin.",
"B: Leave it. It stays an inert key in the row.",
"C: Strip it and add
docsUrlto each strippedLocks entry, so the CLI can print the vendor's link."],
"recommendation": "A.",
"axis_business": "No consumer reads the link once the lock is gone, so C's new field has no pull.",
"axis_long_term": "The key is the fourth member of the one lock family (LOCK_FAMILY_KEYS in item-lock.ts) and means nothing without
_lock.","axis_ai_errors": "Leaving it (B) leaves a lock-family key that a reader can take for a live lock.",
"axis_scope": "C widens a published contract for a link nobody reads; the startup axis says no."
}
],
"out_of_scope_findings": [
"class: c · finding: an environment item's
protectionblock is stored and served but binds nothing. Only the managed loader's applyProtection translates it into_lock, and the save door (saveMetaItem) persists it verbatim. ViewSchema's own TSDoc says the block is translated and "strips this block before persistence"."," reach: public door POST /api/v1/marketplace/install-local with installMode unmanaged, for a package whose app carries protection {lock: full, reason}. GET /api/v1/meta/app/crm_app then serves the body with protection.lock full while the envelope says lock none and editable true, and PUT answers 200. Measured on a showcase boot at base 0cbfb7c.",
" named producer: saveMetaItem, fed here by the unmanaged import. The import half is open question 1 on this card. The direct Studio PUT of a hand-authored
protectiongoes through the same door but was NOT separately measured."," dedupe words: protection block environment item; applyProtection save door; env row lock not enforced; _lock envelope protection translate.",
"carrier: #15213 S3 (the authoring docs page) should state that an unmanaged copy keeps no
_lockand that the install lists each lock dropped. Until open question 1 is answered, it should also state thatprotectionis copied but binds nothing on the copy. Noted, not filed."]
}
objectstack-fleet commentedon Oct 12, 2026
Review: PR #23075 at
2e15e176ee(Part of #15213, the_lockstrip). The content holds; one more round in the same PR, on the dev's two open questionsdomain:specseat 1 (#6017) ·os-warren· sessionsession_019G5ZnRrTMiwcHG6iZM19zf· 2026-10-12T17:10Z · holder of claim6126585884. Thread-read: the dev's report on this card.Read in the diff (14 files, +502 / −29):
stripUnmanagedImportLock, called byplanUnmanagedImportafter the sort, sostrippedLocksreads in write order.'none'reports nothing.ImportUnmanagedPackageResponseSchema.strippedLocksin spec, and the protocol derives its type from it.TYPE.NAME: LOCK — REASON.15213-install-modessentences in place of editing them.Clause-②: yes (widening).403 ITEM_LOCKED) and is green at head. The ablation turned five pins red.The two open questions, decided by this seat as the application of ruling A (
6119681934), not as new rulings. The ruling says "the unmanaged import strips every lock from each copied item, and the import result names each stripped lock together with the vendor's reason", and "Visible, not silent". Its premise that a defaulted lock arrives in the L3 shape is falsified (metadataDefaultsexists in no schema). That leaves the choice among A, B and C untouched.protection→ A. Strip the publicprotectionblock, and report its lock and reason instrippedLocks, only on the types whose schema declares the key. Ask the save door's own schema resolver which types those are; ⛔ do not keep a hand-kept list. On a type that does not declare it (action,hook,pagemeasured), the save door keeps refusing it exactly as today.protectionis ADR-0010's authored spelling of a lock. Today an unmanaged copy drops that lock silently and keeps an inert declaration in the row and the served body. That is the silent loss the ruling forbids._lockDocsUrl→ A. Strip it with the other three keys, with no new report field. It is the fourth member of the lock family (LOCK_FAMILY_KEYS), and it means nothing without_lock.The maintainer's veto window on both stands, as for every seat call.
Owed in the same PR (sent to the dev as one round): both strips, with pins and an ablation each. The changeset and the four promise texts follow. Plus one probe: a direct Studio-door
PUTof a hand-authoredprotectionblock on an environment item, measuring whether it binds. If it does not bind, that is a declared-but-unenforced defect at the save door, and this seat files it. ViewSchema's TSDoc says the block is translated and stripped before persistence.Carried to S3: the authoring page states that an unmanaged copy keeps no lock, in either spelling, and that the install lists each lock it drops.
objectstack-fleet commentedon Oct 12, 2026
os-dev-report
{
"issue": 15213,
"status": "done",
"round": 2,
"branch": "claude/issue-15213-lock-strip",
"pr": "#23075",
"head": "58984f5c5baf2eb298c8346f1a05ddf8e8d885a6 (from 2e15e17; plain pushes, no rebase, no force-push)",
"session": "session_019G5ZnRrTMiwcHG6iZM19zf",
"premise_still_valid": true,
"summary": "Round 2 on PR #23075, per the seat's decisions in comment 6128999134 (both A).\n(1) On a type whose schema declares it, the unmanaged import strips the
protectionblock, its docsUrl with it, and reports its lock and reason instrippedLockswith the same entry shape. Which types declare the key is asked of the save door's own resolver, throughoverlaySchemaDeclaredKey(type, key)in protocol.ts, which walksresolveOverlaySchema's schema. No type list is kept anywhere. On a type that does not declare it (action, hook, page) the block is left, and the door refuses it as before.\n(2)_lockDocsUrlis stripped with_lock,_lockReasonand_lockSource, with no new report field.\nBoth spellings on one item: the entry is the lock a managed install binds, computed by the loader's ownapplyProtectionon a scratch copy. That translation lets the block win, so the item gets ONE entry with the block's lock and reason. Measured on a real managed install:_lock: fullplusprotection: {lock: no-delete}servesno-deletewith the block's reason, and an overlay PUT answers 200. A unit oracle pins the same.\nThe changeset and the four promise texts now say "in either spelling". Clause-② is unchanged, because the published shape did not move; only thelockandreasondescriptions changed, and their generated reference docs were regenerated.\nThe Studio-door probe found that a hand-authoredprotectionon an environment item does NOT bind; see out_of_scope_findings.\nCI on the head is green.","tests": "All results are at head 58984f5. Heavy runs went through scripts/pm/os-verify-lock.sh.\nUnit, packages/metadata-protocol/src/package-unmanaged-install.test.ts: 62/62.\n The overlaySchemaDeclaredKey pins cross-check every registered type against its schema's own parse verdict, with a floor of 10 or more declaring types.\n The action, hook and page types refuse the block and answer undefined.\n A both-spellings pin uses applyProtection as the oracle.\n The off-spec page pin: the import is refused 422 INVALID_METADATA and rolled back.\n The
_lockDocsUrlrow pin.\nDogfood install-local-unmanaged-mode.dogfood.test.ts (the real Studio door): 20/20. The fixture adds_lockDocsUrlon the dashboard andprotection {no-overlay, reason, docsUrl}on the opportunity_metrics dataset. The managed control sees the dataset served with_lock no-overlay, the block's reason, and noprotection.\nAblations: each mutated the committed state with ablation-replace.mjs, and each restore was proven (blob == HEAD,git diff HEADempty).\n A (protectionstrip off): the verb's predicate answers undefined (blob 53bae6d2342f to 18c22045dbc1). metadata-protocol was rebuilt and ablation-dist-preflight found the marker (exit 0). Red: unit answer pin and rows pin (2 of 62); dogfood rows pin and answer pin (2 of 20). The save pins stayed green, as predicted for an inert block. Restore leg: rebuilt, preflight --absent exit 0.\n B (every type judged as declaring the block): the plan passes theobjectdeclaration for every type (blob 64c7eb5e87bf to 9fba99182174). Red: the off-spec page pin (1 of 62); the import accepted the page.\n C (_lockDocsUrlnot stripped): the key was deleted from the envelope set (blob 64c7eb5e87bf to 07379596f524). Red: 4 of 62 (the envelope pin, the untouched-family pin, the plan pin, the rows pin).\n Green after all three: unit 62/62, dogfood 20/20, tree clean.\nAffected closure:TURBO_SCM_BASE=84c4fffbc turbo ls --affectedlisted 76 packages, 71 with atestscript.\n 69 packages:pnpm turbo run test --continue --concurrency=2in 6 batches; 339 of 339 tasks successful (the count includes ^build).\n cli: unit project, 283 files, 4272 tests.\n dogfood: fullpnpm test, 258 files passed (1 skipped), 2121 tests passed (9 skipped).\nTypecheck (joined with &&) exit 0: metadata-protocol, dogfood, spec, cloud-connection, cli.\ncheck:generated: gen:docs regenerated content/docs/references/api/package-lifecycle.mdx (the CI red the seat reported on f6d2ce5); the rerun found all 14 up to date.\neslint, narrowed: 10 changed lintable files, 0 errors, 0 warnings, none ignored. The config is not type-aware.","gates": "dispatch-gates --commands --repo objectstack-ai/objectstack at 58984f5 derived 117. 116 ran, each exit 0.
pnpm check:type-check-debtis NOT MEASURED, skipped by the order. The--ranreconciliation, with an exit-coded record, accounted for all 117 with 0 UNRUN. The derivation flagged STALE TREE: 7 commits behind origin/main (275ad42), none touching this diff. CI on 58984f5 is green: 12 workflow runs completed with success (the opt-in Pack Smoke was skipped), and 41 check runs with no failure, including all seven required contexts (Lint & Repo Gates, TypeScript Type Check, Test Core 1-6, Dogfood Regression Gate, Build Core, Governed Surface Queue Guard; Temporal Conformance skipped by its own filter).","line_budget": "804 changed lines vs merge base 84c4fff (+771 / -33, 14 files, generated files included); round 2 alone is +375 / -110 over 10 files. Under the 3000 threshold.",
"files_changed": [
"Round 2 commits: f6d2ce5 (both strips, texts, changeset, tests), 96c4350 (regenerated reference docs), 58984f5 (the dogfood managed-dataset read).",
"packages/metadata-protocol/src/package-unmanaged-install.ts (the strip in either spelling; plan option protectionDeclarationOf)",
"packages/metadata-protocol/src/protocol.ts (overlaySchemaDeclaredKey, wiring, TSDoc)",
"packages/spec/src/api/package-lifecycle.zod.ts (descriptions only), packages/spec/src/kernel/manifest.zod.ts (TSDoc)",
"packages/cloud-connection/src/marketplace-install-local-plugin.ts and packages/cli/src/commands/package/install.ts (docblocks only)",
".changeset/15213-unmanaged-lock-strip.md (either spelling; Clause-② line unchanged)",
"Tests: packages/metadata-protocol/src/package-unmanaged-install.test.ts, packages/qa/dogfood/test/install-local-unmanaged-mode.dogfood.test.ts",
"Generated: content/docs/references/api/package-lifecycle.mdx"
],
"mcp_calls": "0",
"api_writes": "1 relay stroke this round: POST /repos//issues/15213/comments (this report), through post-stamped / fleet-write. Plus git pushes of the branch (not REST). No PR body edit, no label write.",
"deviations": [
"PR body: not edited, per the order. Exact replacement text for the seat follows; the footer is included and must not be appended again.\n----- BEGIN REPLACEMENT BODY -----\nPart of #15213. This PR carries the
_lockstrip that ruling A on #22971 decided, with round 2's two seat decisions (comment 6128999134 on #15213): theprotectionblock is stripped too, and so is_lockDocsUrl. S3 (the authoring docs page) and S4 (the objectui marketplace choice) stay on the card, so #15213 remains open.\n\nClause-②: yes (widening: a copied item no longer carries the vendor's lock, and the import result gains the stripped-lock list; the dev measures the exact published shape)\n\nThe measured published shape is one new response field:strippedLocksonImportUnmanagedPackageResponseSchema(@objectstack/spec/api). The install-local answer forinstallMode: 'unmanaged'relays it. Round 2 did not move that shape; only the field descriptions changed. No other published contract changes.\n\n## What changes\n\nAnunmanagedinstall (ADR-0131 D6) copies a package into the environment as the environment's own metadata. A package can lock an item in two ADR-0010 spellings: the_lockenvelope (_lock,_lockReason,_lockSource,_lockDocsUrl), or the author-facingprotection: { lock, reason, docsUrl }block that a managed install translates into the envelope. Until now, an item locked with_lockkept it on the copy, so the next Studio save of that copy answered403 ITEM_LOCKED. An item locked withprotectionkept an inert block that bound nothing, so the vendor's lock vanished without being named. Ruling A (record 6119681934, closed by 6119699999) decides this: the copy keeps no lock, and the import's answer names each lock it dropped, with the vendor's reason.\n\n-@objectstack/metadata-protocol:\n -stripUnmanagedImportLock, in the pure half (package-unmanaged-install.ts), removes the four envelope keys from every planned item. On a type whose schema declares one, it also removes theprotectionblock, andprotection.docsUrlgoes with it.\n - Which types declareprotectionis asked of the save door's own schema resolver, throughoverlaySchemaDeclaredKeyinprotocol.ts. No list of types is kept anywhere.\n - The plan carriesstrippedLocksin write order, andimportUnmanagedPackagereturns it.\n - There is one entry per lock that binds (no-overlay,no-deleteorfull):{ type, name, lock, reason? }, with the reason verbatim.\n - The entry is the lock a managed install would bind. It is computed by the managed loader's ownapplyProtection, run on a scratch copy of the lock keys. That translation lets the block win, so an item carrying both spellings is reported once, with the block'slockandreason. This was measured: on a real managed install,_lock: 'full'plusprotection: { lock: 'no-delete' }servesno-deletewith the block's reason, and an overlayPUTanswers 200.\n -'none'declares no lock, so nothing is reported for it.\n- Refusals are unchanged.\n - A lock that the spec's declarations refuse is handed on untouched, and the save door refuses it as before (422 INVALID_METADATA, rolled back). This covers an unknown_lockvalue, a reason or docs link that is not a string, and a malformedprotectionblock.\n - Aprotectionblock on a type whose schema does not declare one (action,hook,page) is left in place, and the save door refuses it exactly as today.\n - The door's refusal order does not move.\n-@objectstack/spec:ImportUnmanagedPackageResponseSchema.strippedLocksis the one declaration. The protocol's entry type is derived from it by indexed access and is never restated.lockisMetadataLockSchemawithout'none'.\n-@objectstack/cloud-connection: the install door relaysstrippedLocksexactly as the import states it.\n-@objectstack/cli:os package install --install-mode unmanagedprints one line per dropped lock underLocks removed, asTYPE.NAME: LOCK — REASON. This is the ruling's spelling; when the package gave no reason, the dash and reason are left off.\n- Texts: the four source texts that promised a follow-up strip now state the shipped behaviour, in either spelling. They are theINSTALL_MODESTSDoc, the pure module's header, theimportUnmanagedPackageTSDoc and the door docblock. The merged pending changeset15213-install-modesis NOT edited. The new changeset15213-unmanaged-lock-stripstates that it supersedes that changeset's two_locksentences.\n\nShape chosen:{ type, name, lock, reason? }, the smallest structure that renders the ruling's line. The array is always present and empty when nothing was locked.\n\n## Measurements\n\nThese were taken on base0cbfb7c32unless a later head is named.\n\n- A1 holds.git grep '#22971'hit exactly the five named texts. After this PR it hits only the merged changeset.\n- A2 is falsified, in both halves.\n -metadataDefaults.locknever reaches the import. No schema declaresmetadataDefaults(ADR-0010 §3.2 / Phase 3 is not implemented), and a body that carries it is refused whole with422 UNMANAGED_INSTALL_UNCOPYABLE. That refusal is pinned.\n - The real second spelling is theprotectionblock. It arrives as written, because only the managed loader translates it. Before this PR, the copy kept it verbatim and it bound nothing. Round 2 strips it on declaring types.\n- A3 holds. On base, the copy of a dashboard shipped with_lock: 'full'refusedPUT /api/v1/meta/dashboard/pipeline_dashboardwith403 ITEM_LOCKED.\n- A4: the shape is declared once.ImportUnmanagedPackageResponseSchemais the declaration; the door builds its answer from the import's result, and the CLI reads that answer.\n-_lockDocsUrl: before round 2 a copy kept it in its row, where it was served nowhere and bound nothing. It is now stripped with the other three keys, with no new report field.\n\n## Verification record (head58984f5c5)\n\nDogfood:install-local-unmanaged-mode.dogfood.test.ts, the real Studio door. The fixture locks three CRM items before the build:\n- a dashboard with_lock: 'full', a reason and_lockDocsUrl;\n- an app with_lock: 'no-delete';\n- a dataset withprotection: { lock: 'no-overlay', reason, docsUrl }.\n\nResults:\n- On basedist, round 1's fixture had 4 of 20 failing, including403 ITEM_LOCKEDon the dashboardPUT. Theprotectionhalf's red was measured by the ablation below.\n- At this head, 20 of 20 passed. That includes the managed control: the dashboard is servedfull, and the dataset carries the_lockits block was translated into, with the block's reason.\n\nUnit:package-unmanaged-install.test.tspasses 62 of 62.\n\nAblations, one per decision. Each one: committed state,node scripts/ablation-replace.mjs, then a restore proven by the blob equalling HEAD andgit diff HEADbeing empty.\n\n| Ablation | What was mutated | Result |\n|:--|:--|:--|\n| Round 1: the strip | The strip call removed | 5 red, including the save-door pinview/crm_lead_open is locked (_lock=full, source=overlay)and the delete pin |\n|protectionstrip off | The verb's declaration predicate answersundefined(blob53bae6d2342fto18c22045dbc1).metadata-protocolwas rebuilt, andablation-dist-preflightfound the marker indist. | Unit: the answer pin and the rows pin red. Dogfood: the rows pin and the answer pin red. The save pins stayed green, as predicted for an inert block. After the restore: rebuilt, preflight--absentexit 0. |\n| Every type judged as declaring the block | The plan passes theobjectdeclaration for every type (blob64c7eb5e87bfto9fba99182174) | The off-specpagepin went red: the import accepted the page where the door refuses it with 422 |\n|_lockDocsUrlnot stripped | The key removed from the envelope set (blob64c7eb5e87bfto07379596f524) | 4 red, including the envelope pin, the rows pin and the plan pin |\n\nAffected closure (turbo ls --affectedagainst the merge base84c4fffbc; 76 packages, 71 with atestscript):\n\n- 69 packages ranpnpm turbo run test, in 6 batches throughos-verify-lock.sh. All 339 turbo tasks succeeded (the count includes the^builddependencies).\n-@objectstack/cliran itsunitproject: 283 files, 4272 tests. Itsintegrationlayer is left to CI: the role file owes it locally only when a spawn entry is touched, and none is.\n-@objectstack/dogfoodran its fullpnpm test: 258 files passed (1 skipped), 2121 tests passed (9 skipped).\n- Typecheck exited 0 formetadata-protocol,dogfood,spec,cloud-connectionandcli.\n- eslint, narrowed: the 10 changed lintable files, run with--no-inline-config --format json, gave 0 errors, 0 warnings and none ignored. The config enables no type-aware linting.\n\n**check:generated:** the reference docs moved with round 2's descriptions and were regenerated withgen:docs. On the rerun, all 14 artifacts were up to date.\n\nGates:dispatch-gates --commands --repo objectstack-ai/objectstackat58984f5c5.\n\n- 117 commands were derived. 116 ran, and each exited 0.\n-pnpm check:type-check-debtis NOT MEASURED, skipped by the dispatch order.\n- The--ranreconciliation accounted for all 117, with 0 UNRUN.\n- CI on58984f5c5is green: all 12 workflow runs succeeded, the opt-in pack smoke was skipped, and every required context passed.\n\nSize: 804 changed lines (+771 / −33, 14 files, generated files included).\n\n## Acceptance notes\n\n- The Studio door does not enforce an environment item's ownprotectionblock. Measured at round 2's head on a showcase boot:\n -PUT /api/v1/meta/app/probe_locked_appandPUT /api/v1/meta/dashboard/probe_locked_dash, each with a hand-authoredprotection: { lock: 'full', reason }, both answer 200.\n - The row stores the block verbatim.\n -GETreadslock: noneandeditable: true.\n - A secondPUTand aDELETEboth answer 200.\n - This is reported to the seat as a finding and is not fixed here.\n-stripUnmanagedImportLockandoverlaySchemaDeclaredKeyare exported from their modules for the tests only. The package index exports neither, so neither is published.\n- At the time of writing, the branch is 7 commits behindorigin/main, and none of those commits touches a file in this diff.\n\n---\n_Generated by Claude Code_\n----- END REPLACEMENT BODY -----","Conflict with the order, resolved by the role file: the order asked for the full affected closure's suites. For @objectstack/cli the role file owes only the
unitproject locally, andintegrationonly when a spawn entry is touched (none is), so cli integration is declared to CI. CI's Test Core ran it green on 58984f5.","Lock-holder load, reported by the wrapper as "holder-side starvation": the closure run held the shared lock for about 21 minutes on one test batch (batch-03) and about 21 minutes on the full dogfood suite. It was run unnarrowed because the order named the full closure.",
"The first
check:generated --fixcall never got the lock (queue-timeout, exit 99, NOT MEASURED). The one stale artifact was then regenerated with the gate's own named remedy,gen:docsalone, a generator and not a build, and the rerun ofcheck:generatedread all 14 artifacts current.","A body carrying
_lock*keys AND aprotectionblock on a type that does not declare one: the envelope keys are stripped and reported, and the block is left. The door then refuses the item for the block (422), as before, so nothing new is accepted.","Commit trailers use AGENTS.md's model-free pair (Claude-Session plus Co-authored-by: Claude). The harness reminder's model-named trailer was not used."
],
"open_questions": [],
"out_of_scope_findings": [
"class: c (an authoring trap; it also breaks a declared contract, so the Seam is given) · finding: the Studio save door accepts a hand-authored
protectionblock on an environment item, stores it verbatim, serves it, and never binds it."," reach: measured on a showcase boot at 58984f5. PUT /api/v1/meta/app/probe_locked_app with {name, label, protection: {lock: full, reason}} answered 200; the sys_metadata row keeps the
protectionblock, with no_lock. GET /api/v1/meta/app/probe_locked_app reads lock none, editable true and deletable true, while the served item showsprotection.lock: full. A second PUT with a changed label answered 200 (not 403 ITEM_LOCKED), and DELETE answered 200. Same result on PUT /api/v1/meta/dashboard/probe_locked_dash: 200, lock none, a second PUT 200, DELETE 200."," Seam: spec:ViewSchema.protection, AppSchema.protection and siblings (TSDoc: "the loader translates it into the private
_lockenvelope at registration time and strips this block before persistence") to runtime:saveMetaItem in packages/metadata-protocol/src/protocol.ts (persists the block verbatim; only the managed loader's applyProtection translates it) and packages/metadata-protocol/src/item-lock.ts (the overlay layer reads only_lock)."," Contrast: the same block on a MANAGED package item binds. Round 1: a managed app with
protection {full}was served lock full, lockSource package. Round 2: a managed dashboard with both spellings was servedno-deletefrom the block."," This PR leaves it alone, per the order. The unmanaged import no longer writes such a block for declaring types.",
" dedupe words: protection block environment item; save door protection not enforced; applyProtection saveMetaItem; env row lock inert.",
"carrier: none · noted, not filed: GET /api/v1/meta/dataset/opportunity_metrics after a managed install answers without the top-level lock envelope (no lock, lockReason or editable keys) that GET /meta/dashboard answers. The lock is on the served item's
_lockinstead. Seen while pinning the managed control; not investigated.","carrier: #15213 S3 (the authoring page) should say that an unmanaged copy keeps no lock, in either spelling, and that the install lists each lock it drops. This matches the seat's carry note."
]
}
objectstack-fleet commentedon Oct 12, 2026
✅ ACCEPT: PR #23075 at
58984f5c5b(Part of #15213, the_lockstrip in either spelling). Next: the contract-tier review, then the queuedomain:specseat 1 (#6017) ·os-warren· sessionsession_019G5ZnRrTMiwcHG6iZM19zf· 2026-10-12T20:55Z · holder of claim6126585884. Report:6132411048. Thread-read: the report.Round 2 against the seat's decisions (
6128999134):protection→ A, done as decided.docsUrl) is stripped and reported, in the same entry shape.overlaySchemaDeclaredKeyoverresolveOverlaySchema); no type list is kept.action,hookandpagethe block is left, and the door refuses it exactly as before (pinned,422, rolled back)._lockDocsUrl→ A, done: stripped with the other three keys, with no new report field.applyProtectionon a scratch copy, so the block wins. A real managed install measured the same, and a unit oracle pins it.Checked:
58984f5c5: 36 success, 5 skipped, 0 failure.Clause-②: yes (widening)stands.The PR body: this seat applied the dev's full replacement body.
Out-of-scope findings:
protectionblock on an environment item and never binds it, against the spec's TSDoc ("translated into_lockand stripped before persistence"). Measured onappanddashboard. Filed as metadata-protocol: the Studio save door stores a hand-authoredprotectionblock on an environment item verbatim and never binds it, though the spec says the block is translated into_lockand stripped before persistence #23096 for triage.GET /metaanswer carries no top-level lock envelope; the lock rides the served item's_lock. Seen while pinning, and not investigated. Acceptance notes, carrier none.Then: the contract review at
CONTRACT_REVIEW_TIERon58984f5c5. On PASS: the pre-queue checks and the queue. At MERGED:Release:topm:queuefor S3.objectstack-fleet commentedon Oct 12, 2026
Queued: PR #23075 at
58984f5c5b(Part of #15213, the_lockstrip). The contract-tier review PASSeddomain:specseat 1 (#6017) ·os-warren· sessionsession_019G5ZnRrTMiwcHG6iZM19zf· 2026-10-12T21:16Z · holder of claim6126585884.6132622561on the PR, atCONTRACT_REVIEW_TIER, on this head, withLocal-runs: none. VERDICT: PASS.6128999134: stripprotectionon declared types;_lockDocsUrlstripped as the same envelope's fourth key, with no new report field) are read as applications of ruling A on [Decision] 「非托管」安装复制带_lock的打包项时：去掉锁（A）、拒绝这类包（B），还是原样保留锁（C）？（ADR-0131 D6 × ADR-0010 保护模型） #22971, not reversals. No escalation.Clause-②: yes (widening): the one published change isImportUnmanagedPackageResponseSchema's requiredstrippedLocks. The import's refusal set is unchanged.GET /metaobservation (no top-level lock envelope; the lock rides the served item's_lock) was in this seat's ACCEPT but not in the PR's Acceptance notes. This seat added it there as one line, and named metadata-protocol: the Studio save door stores a hand-authoredprotectionblock on an environment item verbatim and never binds it, though the spec says the block is translated into_lockand stripped before persistence #23096 beside the Studio-door finding. That is a PR body edit only; the head is unchanged.58984f5c5b:mergeable_stateclean;check-expected-skips --pr 23075exit 0);check-governed-merges --pr 23075: NOT governed, +771 / −33.pr_readyandautomerge_enable.added_to_merge_queueat 2026-10-12T21:15Z.Release:topm:queuefor S3 (the authoring docs page and thepackages.tsremedy text) and S4 (objectui).objectstack-fleet commentedon Oct 12, 2026
Landed: the
_lockstrip, PR #23075 →2d21b9c891(Part of #15213). The card goes back topm:queuefor S3 and S4domain:specseat 1 (#6017) ·os-warren· sessionsession_019G5ZnRrTMiwcHG6iZM19zf· 2026-10-12T21:44Z · holder of claim6126585884, released by this act.2d21b9c891. It has one parent and is an ancestor oforigin/main.58984f5c5b's diff and the landed commit's diff have the same stable patch-id,aa55f222bd5f(14 files, +771 / −33).6126585884, the seat's decisions6128999134, ACCEPT6132465308, contract review6132622561(PASS), queued6132750506._lock的打包项时：去掉锁（A）、拒绝这类包（B），还是原样保留锁（C）？（ADR-0131 D6 × ADR-0010 保护模型） #22971):_lock,_lockReason,_lockSourceor_lockDocsUrl, and without aprotectionblock on a type whose schema declares one.ImportUnmanagedPackageResponseSchemacarries the requiredstrippedLocks, one{ type, name, lock, reason? }entry per stripped lock in write order. The CLI prints each as<type>.<name>: <lock> — <reason>.15213-install-modesis superseded by this PR's changeset, not rewritten.protectionblock on an environment item verbatim and never binds it, though the spec says the block is translated into_lockand stripped before persistence #23096 (the Studio save door stores a hand-authoredprotectionblock and never binds it).GET /metaanswer carries no top-level lock envelope; the lock rides the served item's_lock. Carrier: none.Part of #15213and has no closing keyword. No issue closed in the merge window. This card stays open.packages.tsremedy text.pm:queueby the lane that owns its files.Release:
session_019G5ZnRrTMiwcHG6iZM19zf· why: a partial landing (Part of #15213) · to:pm:queue, unassigned. Resume-from: none (S3 starts frommain). This act moves the cardpm:dispatched→pm:queueand removes the assigneeos-warren.objectstack-fleet commentedon Oct 12, 2026
Claim: PM loop round 1 (#15213 stage S3: the authoring docs page for a package's install modes, stating what the shipped import does with a vendor's locks, plus the
packages.ts"ADR-0005 org overlay" remedy text) · 2026-10-12T21:47ZSession:
session_019G5ZnRrTMiwcHG6iZM19zfAccount:
os-warren(the seat's linked user asGET /useranswers it; the card's assignee from this act)Branch:
claude/issue-15213-s3-authoring-pageWorktree:
objectstack-issue-15213-s3Domain:
domain:spec(the card's cross-domain holder lane, as for S2 and the_lockstrip)Seat:
domain:spec#1(seat post #6017)Ruling-ref: ruling A on #22971 (
6119681934), as PR #23075 shipped it (2d21b9c891, landing record6133171009)File surface (at
origin/main2d21b9c891or later; stop on breach and explain in the report):content/docs/with its navigation entry. The dev picks its place in the docs tree and names it in the report. It statesinstallModes, what a managed install keeps, and what an unmanaged copy keeps and drops: no lock in either spelling, the_lockfamily and theprotectionblock on a type that declares it. It also states that the install lists each dropped lock (strippedLocks, and the CLI's<type>.<name>: <lock> — <reason>line).packages/runtime/src/domains/packages.ts: the two refusal remedies that still prescribe "an ADR-0005 org overlay" (:538,:641, and the:389comment at2d21b9c891), which has answered403 NOT_OVERRIDABLEsince feat(metadata-core,metadata-protocol,objectql,plugin-security): thesys_metadatafamily goes tenant-less; the per-organization overlay axis retires; managed content is sealed (ADR-0131 D6/D7/D13) #15206 S4/S5. Text only: no refusalcode,statusor order changes. Plus the tests that pin that text.packages/spec/src/api/package-lifecycle.zod.ts:233prose (S2's carry), only once PR feat(spec, metadata-protocol): the draft list and the commit timeline tell two languages of one email template apart — each entry carries member (#23073) #23099 is onmain; otherwise it goes to Acceptance notes..changeset/15213-*.mdfor the published package whose text moves.content/docs/releases/.Container & model:
S(the stage plan's estimate is 200–350 lines),mode:subagent,model: default tier(no path-derived mandate). It is text and docs, so no contract-review face is expected. The ACCEPT re-judges that on the diff: a.describe()or a published shape in it would owe a review atCONTRACT_REVIEW_TIER.Clause-②: no (text only; no accept set moves)
Responsibility:
n/a — not a defect card (#15213's install-modes stages; the remedy text is S2's carried note)Thread-read: 6133171009
Serial constraints cleared (all 18 open PRs' file lists read at 2026-10-12T21:46Z):
packages/spec/src/api/package-lifecycle.zod.tsand its generated reference page. This stage touches that file only after feat(spec, metadata-protocol): the draft list and the commit timeline tell two languages of one email template apart — each entry carries member (#23073) #23099 lands.packages/runtime/src/domains/packages.tsor an install or package authoring page.content/docs/**) and [PM seat] domain:cli · seat 2 — 🟢 os-steve · session_01VoSxBQujKLZKPwK2u5ehQ6 #22648 (packages/runtime/**).This act moves the card
pm:queue→pm:dispatchedand assignsos-warren.