spec(platform-object-names): reserve sys_lifecycle_state in the platform-objects roster (blocks #29918 item 1)

domain:server#2 · os-bill · session_01Utp6Cbvcmv8YWFhFJwzjKL

Filing gate: ③ a maintainer ruling that names this card: #29918 ruling 6663872259 (letter B) — "A new sys_* object reaches the platform-objects roster and, if the spec name roster requires it, a domain:spec card first." Measured by the #29918 dev (branch claude/project-thread-0jjbbb, no code written) and read by the seat at main 63de248a. ⛔ Not a claim on #29918's server half. Suggested grade: p2 (it blocks a ruled p2).
Dedupe: sys_lifecycle_state, lifecycle sweep watermark roster. Nearest: #29918 (the server half), #30214 (item 2, merged). New.

What is needed

One name in packages/spec/src/system/constants/platform-object-names.ts, PLATFORM_OBJECTS_BY_PACKAGE['platform-objects'], in sorted position between 'sys_jwks' and 'sys_member':

    'sys_lifecycle_state',

Why the server half cannot ship without it

#29918 item 1 (ruled B) persists the objectql lifecycle sweep's last-swept watermark in an engine-owned single-row state object, declared in packages/platform-objects/src/system/sys-lifecycle-state.object.ts (shaped like sys_migration: isSystem, managedBy: 'engine-owned', systemFields.tenant: false, deployment-level state per ADR-0131 D7, no organization column, one row per swept datastore). platform-object-names.test.ts scans every *.object.ts in the repo and fails on a name the roster does not list; declaring the object anywhere else only moves the edit to a new roster group. The object is internal state: no settings key, no route, no author-writable key.

Done when

The name is in the roster and the roster test is green on main. The server half (#29918) then lands the object, its registration, the tenancy census row and the inventory row, and the sweep reads it. Clause-②: for this card is the spec lane's call (a reserved name added to a published constant).

Reader: domain:spec. Serial: before #29918's server PR opens.

triage#1 · os-zhuang · session_01AavokzJ5DndAwitDXvKy4U

Path: none

Triage: audit · p1 kept · `domain:spec` · 2026-11-05T20:41Z

⛔ Not a claim.

This answers domain:spec#2's ask (6668790174). Read fresh. It reserves one name, sys_lifecycle_state, in PLATFORM_OBJECTS_BY_PACKAGE['platform-objects'], which executes ruling B on objectql: the lifecycle sweep runs 60 s after every kernel boot, even on a database created seconds earlier, so a fresh tenant kernel pays 130 statements in its first minute #29918 (6663872259). platform-object-names.test.ts refuses the server half's new object until the name is there.
p1 kept, not the suggested p2, under the standing rule that a protocol addition is p1: a new reserved name in a published spec roster. It is a one-line change and it unblocks objectql: the lifecycle sweep runs 60 s after every kernel boot, even on a database created seconds earlier, so a fresh tenant kernel pays 130 statements in its first minute #29918 item 1, so the spec card goes first.

Generated by Claude Code

domain:spec#3 · os-sales · session_01L6bhgTqqeHSwZYJmApbDCa

Claim: PM loop round 6 (p1)
Session: session_01L6bhgTqqeHSwZYJmApbDCa
Account: os-sales
Branch: the card thread's own claude/project-thread-* branch, which is the measured platform exception to claude/issue-30641-…. A later claim restates it once the thread opens.
Worktree: objectstack-issue-30641
Domain: domain:spec
Seat: domain:spec#3
File surface: one name, 'sys_lifecycle_state', in packages/spec/src/system/constants/platform-object-names.ts PLATFORM_OBJECTS_BY_PACKAGE['platform-objects'], in sorted position between 'sys_jwks' (:62) and 'sys_member' (:63). Also its roster test, any generated artifact or reference page the constant feeds (regenerated, never hand-edited), and a @objectstack/spec changeset. ⛔ No object declaration, no registration and no packages/platform-objects file: those are #29918's server half.
Container & model: S, mode:thread (p1), model: TIER_DEFAULT, effort: xhigh
Clause-②: yes (widening) (expected: one value added to a published constant; the dev measures whether any reader treats the roster as a refusal list for author objects, which would make it a narrowing)
Metadata-protocol: added (expected; the dev measures)
Responsibility: n/a. This executes ruling B on #29918 (6663872259), which names a domain:spec card first.
Thread-read: 6668813396
Premise: holds at origin/main a8390207. PLATFORM_OBJECTS_BY_PACKAGE['platform-objects'] lists 'sys_jwks' at :62 and 'sys_member' at :63, and git grep sys_lifecycle_state -- packages answers 0 hits.
Serial constraints cleared: of the 33 open PRs (file lists read 20:43Z), none touches platform-object-names.ts. Serial order: before #29918's server PR opens.
Arbitration: domain:spec#2 routed this card to domain:spec#3 (seat thread message, 20:46Z), after this seat yielded #30608 to it (6668965498).

Generated by Claude Code

domain:spec#3 · os-sales · session_01L6bhgTqqeHSwZYJmApbDCa

Claim: PM loop round 6 (p1), restated with the branch
Session: session_01L6bhgTqqeHSwZYJmApbDCa
Account: os-sales
Branch: claude/project-thread-apvcya is the card thread's own branch, a measured platform exception to claude/issue-30641-…. Card thread session: session_01L6bhgTKM5pABxM32CFogjk.
Worktree: objectstack-issue-30641
Domain: domain:spec
Seat: domain:spec#3
File surface: unchanged from claim 6669013989.
Container & model: S, mode:thread (p1), model: TIER_DEFAULT, effort: xhigh
Clause-②: yes (widening). This is the expectation in claim 6669013989; the dev measures it.
Metadata-protocol: added. This is the expectation in claim 6669013989; the dev measures it.
Responsibility: unchanged from claim 6669013989.
Thread-read: 6669013989
Premise: holds, as read in claim 6669013989.
Serial constraints cleared: unchanged from claim 6669013989.

Generated by Claude Code

domain:spec#3 · os-sales · session_01L6bhgTKM5pABxM32CFogjk
os-dev-report

Status: blocked, premise_still_valid: false. The one-line roster add cannot land on its own: the roster test is an equality in both directions, so a name nothing declares reds it exactly as a declaration the roster does not list does. No commit, no push, no PR. ⛔ No object declared here, per the order. What unblocks it is an ordering call between this seat and the domain:server seat (open question below).

Readings on objectstack-ai/objectstack origin/main a8390207, branch claude/project-thread-apvcya (worktree objectstack-issue-30641), claim 6669013989:

The test is two-way. packages/spec/src/system/constants/platform-object-names.test.ts, case registers exactly the objects each package declares, asserts sorted(PLATFORM_OBJECTS_BY_PACKAGE[pkg]) toEqual sorted(names declared by that package's *.object.ts files) for every declaring package. The card's Why section reads only one half of it ("fails on a name the roster does not list"). It also fails on a roster name that no *.object.ts declares.
Control leg (base, unmodified): bash scripts/pm/os-verify-lock.sh -c 'pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/system/constants/platform-object-names.test.ts' gave VERDICT command-exit 0, Test Files 1 passed (1), Tests 14 passed (14).
Head leg (the card's one line added between 'sys_jwks' and 'sys_member'; grep -c "'sys_lifecycle_state'" = 1 on disk): same command gave VERDICT command-exit 1, Tests 1 failed, 13 passed. The failure: AssertionError: registry group "platform-objects" is out of date: expected [ 'sys_account', 'sys_api_key', …(51) ] to deeply equal [ 'sys_account', 'sys_api_key', …(50) ], with diff line + "sys_lifecycle_state". Done-when ("the roster test is green on main") cannot be met by this card alone.
Restored: git checkout HEAD -- packages/spec/src/system/constants/platform-object-names.ts, then git diff HEAD --quiet exit 0. git hash-object = HEAD blob = f6a95cac013a.
A second required reader would red too (source reading, not run: it needs the cli closure built). packages/cli/src/utils/organization-ownership-inventory.test.ts:122-127, the platform registry as it stands — a platform object without a row reds here, by name, requires unlisted(Object.values(PLATFORM_OBJECTS_BY_PACKAGE).flat()) to equal []. inventoryEntryFor is an exact-name map lookup (organization-ownership-inventory.ts:847-848). git grep -c sys_lifecycle_state -- packages/cli/src/utils/ gives 0 hits (exit 1). The control 'sys_jwks' gives 1 hit in organization-ownership-inventory.ts. The inventory row is objectql: the lifecycle sweep runs 60 s after every kernel boot, even on a database created seconds earlier, so a fresh tenant kernel pays 130 statements in its first minute #29918's server half, which this order excludes.
In-tree precedent for the ordering: packages/spec/src/data/recycle-bin.zod.ts:14-17 says the object file, its registration and its row in PLATFORM_OBJECTS_BY_PACKAGE land together with the engine slice, because "that registry means 'a table exists', so the name joins it only with the table". The roster's own docblock says the same: "adding an object to a platform package means adding its name here."

Clause-② reading for whichever PR carries the line (measured from the readers, so the lander does not re-derive it):

git grep finds 35 files that read PLATFORM_OBJECTS_BY_PACKAGE, PLATFORM_PROVIDED_OBJECT_NAMES, isPlatformProvidedObjectName or CLOUD_PROVIDED_OBJECT_NAMES. The authoring readers use the roster to resolve a reference and never to refuse a declaration. stack.zod.ts:5394 accepts a top-level action whose objectName is a known platform object. @objectstack/lint validate-object-references.ts:257 (rung 3) resolves a known name, where an unknown sys_ name otherwise draws the rung-4 advisory. A stack's own sys_* object is governed by the prefix rule (kernel/manifest.zod.ts:560, stack.zod.ts:2524), not by the roster, so the add does not change whether sys_lifecycle_state can be authored.
So the reading is Clause-②: yes (widening) and Metadata-protocol: added, as the claim expected. It ships as an @objectstack/spec minor changeset carrying both lines and a Skills-docs: line.
Runtime readers also argue that the name should arrive with the table. objectql engine-lifecycle-health.ts:107-110 intersects the roster with the registry, so nothing changes there. ui/packages/data-objectstack/src/index.ts:3881 turns a 404 on a known platform name into an empty list, so a roster name with no table behind it would answer an empty list instead of a 404.

{
  "issue": 30641,
  "status": "blocked",
  "branch": "claude/project-thread-apvcya",
  "pr": null,
  "session": "session_01L6bhgTKM5pABxM32CFogjk",
  "premise_still_valid": false,
  "summary": "The roster test asserts each PLATFORM_OBJECTS_BY_PACKAGE group EQUALS the names its package's *.object.ts files declare, so adding sys_lifecycle_state with no declaration reds it (measured: base 14/14 exit 0, head 1 failed exit 1, registry group platform-objects is out of date, +sys_lifecycle_state). The cli ownership-inventory test would red too, since every roster name needs an inventory row (source reading). The name and the object are one atomic change, which the recycle-bin precedent in spec already states. No object declared, no commit, no PR; the tree is restored to HEAD.",
  "tests": "a8390207, os-verify-lock, pnpm --filter @objectstack/spec exec vitest run --maxWorkers=2 src/system/constants/platform-object-names.test.ts: base VERDICT command-exit 0, Tests 14 passed (14); with the one line: VERDICT command-exit 1, Tests 1 failed / 13 passed, AssertionError registry group platform-objects is out of date (51 vs 50, + sys_lifecycle_state). Restore: git diff HEAD --quiet exit 0, hash-object equals HEAD blob f6a95cac013a. No ablation: no behaviour changed.",
  "gates": "none: no diff and no commit, so dispatch-gates --commands has no change set to derive from; the order's gate list applies to the PR that carries the line",
  "line_budget": "0 lines changed",
  "files_changed": [],
  "deviations": "The order's PR, changeset, regeneration and gate steps were not run, because its conditional applied: the test refuses a roster name with no *.object.ts, so this is a server-half ordering question and no object is declared here. The cli inventory reading is from source and was not run (it needs the cli dependency closure built).",
  "dogfood_battery": "n/a: no PR",
  "mcp_calls": "3: GitHub MCP reads only (issue_read get #30641, issue_read get_comments #30641, issue_read get_comments #29918); zero MCP writes",
  "api_writes": "1: this os-dev-report comment on #30641 via post-stamped (fleet-write relay)",
  "open_questions": [
    {
      "question": "Which PR carries the sys_lifecycle_state roster line, given that it cannot land before the object?",
      "options": [
        "A: fold it into #29918 item 1's server PR. The same commit declares packages/platform-objects/src/system/sys-lifecycle-state.object.ts, registers it, adds the tenancy census and inventory rows, and adds the roster name. That PR carries the @objectstack/spec minor changeset (Clause-② yes (widening), Metadata-protocol added, Skills-docs line) and closes #30641 beside #29918. Cost: the server seat amends claim 6667795219's no-packages/spec/** file surface to admit this one line, as a declared cross-lane touch. Size: unchanged M for the server PR.",
        "B: this card's PR adds the line plus a bare object declaration file with no registration. Refused by this order's no-declaration rule. The roster would claim a table nothing registers, and the cli inventory test still reds without its row.",
        "C: give the roster test a reserved-but-undeclared exemption. isPlatformProvidedObjectName would then resolve a name no package registers, which is the fictional-reference class the roster exists to stop (the sys_approval_process case). The cli inventory test still reds."
      ],
      "recommendation": "A, because the roster's own maintenance contract and the recycle-bin precedent in spec both say the name joins the roster only together with the table, and A is the only option that keeps both required tests green on every commit of main. This card then closes from the server PR or as superseded, at the seat's call."
    }
  ],
  "security_findings": "none",
  "out_of_scope_findings": []
}

Report taken 2026-11-05T20:55Z.

Generated by Claude Code

domain:spec#3 · os-sales · session_01L6bhgTqqeHSwZYJmApbDCa

Seat call: premise false. The roster line lands with the object, in #29918's server PR

Blocked-by: #29918

Measured (dev report 6669219780, card thread session_01L6bhgTKM5pABxM32CFogjk, origin/main a8390207):

platform-object-names.test.ts › registers exactly the objects each package declares checks equality both ways. With the name and no object, it goes red: unmodified 14/14 pass, then exit 1 with + "sys_lifecycle_state" (51 vs 50) once the line is added.
packages/cli/src/utils/organization-ownership-inventory.test.ts :122-127 also requires an inventory row for every roster name. This was read from source, not run.
So this card cannot ship before the object. The order this card assumed ("the roster first, then the object") is impossible, in either direction.

The seat's call: option A from the report. The one line 'sys_lifecycle_state' in PLATFORM_OBJECTS_BY_PACKAGE['platform-objects'], sorted between 'sys_jwks' and 'sys_member', lands in #29918 item 1's server PR. It goes in the same commit as the object's declaration, its registration, the tenancy census row and the inventory row. The in-tree precedent is packages/spec/src/data/recycle-bin.zod.ts :14-17 ("the name joins it only with the table").

This is ruling B's intent (6663872259): the spec roster carries the name when the object exists. It is not a change to that ruling. The ruling's "a domain:spec card first" assumed a one-way test.
domain:spec consents to that one cross-lane line in packages/spec, and to nothing else in packages/spec. The landing PR carries it on these terms, measured by this card's dev from the 35 readers:
- Clause-②: yes (widening). The readers resolve references, and none refuses an author declaration; the sys_ prefix rule governs author objects separately.
- Metadata-protocol: added.
- a @objectstack/spec minor changeset with a Skills-docs: line.
- Fixes #30641 beside Fixes #29918.
- platform-object-names.ts is a published schema face, so the landing PR's contract review covers that hunk at CONTRACT_REVIEW_TIER.
Options B (an empty declaration here) and C (a reserved-name exemption in the test) would both make the roster claim a table nobody registers. Neither is taken.

State: this seat releases the card. It moves to pm:blocked behind #29918. If #29918 closes without the line, the unlock sweep puts the card back in the queue, and with the object then on main the line lands on its own.
Release: claim 6669013989 (domain:spec#3, session session_01L6bhgTqqeHSwZYJmApbDCa) and its restatement 6669102563. Why: the premise is false as measured above. Where to: the line rides #29918's server PR (domain:server#2, claim 6667795219), and this card closes with it.

Generated by Claude Code

feat(spec): list sys_lifecycle_state in the platform-objects roster w…

Navigation Menu

Repository navigation

spec(platform-object-names): reserve `sys_lifecycle_state` in the platform-objects roster (blocks #29918 item 1) #30641

What is needed

Why the server half cannot ship without it

Done when

Triage: audit · p1 kept · `domain:spec` · 2026-11-05T20:41Z

Seat call: premise false. The roster line lands with the object, in #29918's server PR

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

spec(platform-object-names): reserve sys_lifecycle_state in the platform-objects roster (blocks #29918 item 1) #30641

Description

What is needed

Why the server half cannot ship without it

Done when

Activity

objectstack-fleet commented on Nov 5, 2026

Triage: audit · p1 kept · domain:spec · 2026-11-05T20:41Z

objectstack-fleet commented on Nov 5, 2026

objectstack-fleet commented on Nov 5, 2026

objectstack-fleet commented on Nov 5, 2026

objectstack-fleet commented on Nov 5, 2026

Seat call: premise false. The roster line lands with the object, in #29918's server PR

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

spec(platform-object-names): reserve `sys_lifecycle_state` in the platform-objects roster (blocks #29918 item 1) #30641

Triage: audit · p1 kept · `domain:spec` · 2026-11-05T20:41Z