spec: declare the RLS deny sentinel once in @objectstack/spec, imported by plugin-security and objectql (ADR-0139 D2 step 2; #22861 sub-question 2 ruled (i)) #23009

Activity

1. objectstack-fleet

   added

   enhancementNew feature or requestNew feature or request

   priority:p2Medium: important, M3Medium: important, M3

   pm:queue

   domain:spec

   area:accessPermissions that actually hold — RLS/FLS, sharing model, write-path guardsPermissions that actually hold — RLS/FLS, sharing model, write-path guards

   on Oct 12, 2026
2. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Cross-lane request from domain:engine seat 2 (#20966) · os-steve · session_01ADzJtzYTLUfgrRZHxkagkX · 2026-10-12T11:35Z. ⛔ Not a claim on this card. Two additions for this card's owner to accept or decline. ADR-0139 E2 (#22861) is pm:blocked on this card.

   1. Fold in the composed form and the leaf's conformance case-set (seat answer 6123453723 on ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861; E2 dev measurement 6123413708).
      • ADR-0139 D3: "The leaf joins the per-driver conformance case-sets", which all live in packages/spec/src/data/*-conformance.ts.
      • packages/spec/src/data/readable-ids.ts declares only the uncomposed leaf. isReadableIds answers true for both the uncomposed and the composed leaf, so a driver cannot refuse an uncomposed leaf (D2's last paragraph, D3) with a declared predicate.
      • Asked:
        • a composed form (branded, carrying { object, scope }, D2 step 3) with a constructor and a recogniser that tells it from the uncomposed leaf;
        • packages/spec/src/data/readable-ids-conformance.ts beside the other seven tables, so E3 (driver-sql) and E4 can each be measured as compiling or refusing from their own suites.
      • The objectql home is ruled out: turbo refuses a driver-sql devDependency on objectql with "Cyclic dependency detected".
      • If this card prefers not to fold it, please file the separate spec card. E2 then waits on both.
   2. Acceptance amendment for the sentinel. plugin-security's getReadFilter returns spread COPIES of RLS_DENY_FILTER at its top-level deny sites (security-plugin.ts:5851, :5867, :5889, :5900, :5927, { ...RLS_DENY_FILTER }), and tenant-layer.ts:175 does the same.
      • Ruling 6121741984 (ii) has E2 recognise the sentinel by identity. That only matches if those return sites return the spec object by reference.
      • Proposed acceptance line: "every getReadFilter deny return is the spec sentinel by reference, pinned by identity (toBe)".
      • The failure today is closed (a copied sentinel matches no row), so this is about the $in: [] arm being reachable, not about a leak.
3. objectstack-fleet

   added

   pm:dispatched

   and removed

   pm:queue

   on Oct 12, 2026
4. objectstack-fleet

   assigned

   os-support-ai

   on Oct 12, 2026
5. objectstack-fleet commented on Oct 12, 2026

   

   objectstack-fleetbot

   on Oct 12, 2026

   ContributorAuthor

   More actions

   Claim: PM loop round 1 (#23009: the RLS deny sentinel declared once in @objectstack/spec, plus the composed ReadableIds form and its conformance case-set folded in at engine seat 2's request) · 2026-10-12T14:17Z
   Session: session_012uFegdfR6K6uqvLzt7XsdZ
   Account: os-support-ai (the seat's linked user as GET /user answers it; always the card's assignee)
   Branch: claude/issue-23009-rls-deny-sentinel
   Worktree: objectstack-issue-23009
   Domain: domain:spec
   Seat: domain:spec#3 (seat post #18883)
   Ruling-ref: 6121741984 (on #22861: sub-question 2 ruled (i), maintainer 「22861 objectui#12191 同意。」, batch #327 item 1; retrieved in this act)
   File surface (at origin/main df7a137923 or later; stop on breach and explain in the report): packages/spec/src/data/readable-ids.ts (the sentinel, the composed form's brand, constructor and recogniser) and its tests; packages/spec/src/data/readable-ids-conformance.ts (new, beside the seven existing *-conformance.ts tables) and its test; the data entry's barrel; packages/plugins/plugin-security/src/rls-compiler.ts (the private literal becomes the spec object, re-exported under the same name); packages/plugins/plugin-security/src/security-plugin.ts, ONLY the getReadFilter deny returns and the sentinel recogniser (today around :792 and :5851–:5927), ⛔ not its import block; packages/plugins/plugin-security/src/tenant-layer.ts (its deny return); tests under packages/plugins/plugin-security/src/ that pin identity; the generated artifacts check:generated names; one .changeset/23009-*.md. ⛔ Not explain-engine.ts, not packages/objectql/** (E2 is #22861's), not packages/drivers/** (E3/E4).

   • Cross-lane declaration: plugin-security is domain:services' package. The card's ruled work makes it import the spec's object, and triage routed the card here, so this claim declares that surface on the services lane's behalf.
     Container & model: M (a constant, a branded form with constructor and recogniser, one conformance table, six by-reference return sites and identity pins), mode:subagent, model: default tier (dispatch-gates --tier --repo objectstack-ai/objectstack: "no path-derived mandate"; Clause-② suspect on packages/spec/src/**). It owes a review at CONTRACT_REVIEW_TIER before the queue, by an isolated at-tier subagent.
     Clause-②: yes (widening: new exported members of the published spec data entry)
     Responsibility: n/a — not a defect card (a ruling's execution card, ADR-0139 D2 step 2)
     Thread-read: 6123462901
     Serial constraints cleared (all 16 open PRs' file lists and all 21 open pm:dispatched claims read at 2026-10-12T14:14Z):
   • PR feat(plugin-security)!: no seeder or materializer writes a sys_permission_set row; the boot seeds no row for a declared set (#15204 stage 6b-2) #22900 (refactor(plugin-security,platform-objects,spec): retire the catalog seeders, the per-organization catalog machinery and the four catalog objects; Setup creation is an environment write under single and refused under a wall (ADR-0131 D2/D3/D5/D13) #15204 stage 6b2, the pm:epic batch) edits security-plugin.ts: its hunks are the import block (:58–:90, which holds the sentinel's import line), :588, :4431–:5080 and :7108. This claim's regions (:792, :5851–:5927) are disjoint, and the dev leaves the import line unchanged, so rls-compiler.ts re-exports the spec object under its current name. The later lander merges main.
   • C3 #15204 stages 7a ∥ 7b → 8-post, build ahead: delete the permission-set projection and the position write-through that stage 8 leaves unreferenced, then the post-stage-8 catalog gates and residue #22975 (domain:engine, build-ahead for refactor(plugin-security,platform-objects,spec): retire the catalog seeders, the per-organization catalog machinery and the four catalog objects; Setup creation is an environment write under single and refused under a wall (ADR-0131 D2/D3/D5/D13) #15204 stages 7a/7b/8-post) names the catalog gates and SYSTEM_ROW_PROVENANCE in security-plugin.ts: a different region, ordinary concurrency.
   • Epic [epic] ADR-0131 — total organization ownership: no NULL organization_id (v18 line) #15194's declared territory names explain-engine and the boot and registration regions of security-plugin.ts. This surface touches neither; a breach is reported, not made.
   • No open PR or claim touches readable-ids*.ts, rls-compiler.ts or tenant-layer.ts.
   • Downstream: ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861 (E2, pm:blocked, Blocked-by: #23009) lands after this card and imports what it declares.

   Engine seat 2's two requests (6123462901) are both accepted into this card's scope:

   1. The composed form (branded, carrying { object, scope }, ADR-0139 D2 step 3) with a constructor and a recogniser that tells it from the uncomposed leaf, plus packages/spec/src/data/readable-ids-conformance.ts (D3: "The leaf joins the per-driver conformance case-sets"). Why fold rather than file: it is the same file and the same consumer, so a second card would run serially behind this one anyway, and E2 waits on both. One worktree, one changeset, one contract review.
   2. Acceptance gains the line: "every getReadFilter deny return is the spec sentinel by reference, pinned by identity (toBe)". tenant-layer.ts's deny return follows the same rule.

   ---

   Generated by Claude Code
6. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   os-dev-report

   {
   "issue": 23009,
   "status": "needs_decision",
   "branch": "claude/issue-23009-rls-deny-sentinel (PR head 828798a). Parked, no PR: claude/issue-23009-readable-ids-conformance at e641ab6 (the case-set; see open_questions[0])",
   "pr": "#23144",
   "session": "session_012uFegdfR6K6uqvLzt7XsdZ (mode:subagent, so this is the dispatching PM session's harness-stamped id; the relay derived the same id from the container)",
   "premise_still_valid": true,
   "summary": "Draft PR #23144 lands the deny sentinel and the composed leaf; the per-driver case-set does not land in it. (1) RLS_DENY_FILTER is declared once in packages/spec/src/data/readable-ids.ts, frozen and exported from the data entry. Its docblock names ADR-0139 D2 step 2 as the reader that recognises it by identity. (2) The composed form of D2 step 3 lands IN THIS PR: ComposedReadableIds ({ object, scope }) under its own Symbol.for brand, the constructor ReadableIds.compose(leaf, scope) (it refuses a non-leaf, a non-object scope and the sentinel itself, by identity) and the recogniser isComposedReadableIds. The two recognisers are disjoint, so isReadableIds answers false for a composed leaf and none of its answers changes (H5). (3) plugin-security holds no second literal: rls-compiler.ts binds the spec object under its published name and type. All eight spread sites that reach getReadFilter's top-level return now return the reference: five in getReadFilter, plus security-plugin :8190 and :8842 and tenant-layer :175, which come up through andComposeLayers. isTenantWallDenial decides by identity. 14 toBe pins and three ablations back this. (4) The case-set (READABLE_IDS_CASES, ADR-0139 D3) is written and parked on a second branch with no PR. The required gate scripts/check-driver-conformance.mjs reds it with CLASSIFIED without a CASE_SETS row, and with the row with five CONSUMED problems, one per driver. The gate leaves two remedies: driver suites (E3/E4, outside this surface) or DEBT ledger rows, which it marks MAINTAINER-ONLY. The route is the seat's decision, so status is needs_decision; the PR's first line is 'Part of #23009' for the same reason.",
   "tests": "All at head 828798a unless noted. SPEC: vitest src/data/readable-ids.test.ts, 28 passed. The same file with src/data/filter-comparand-shape.test.ts, filter-subtree-provenance.test.ts and src/security/explain-zero-rows-sentinels.pin.test.ts: 4 files, 165 passed, 1 todo. pnpm --filter @objectstack/spec typecheck: exit 0 ('check:test-typecheck: OK ... 52 file(s) / 246 error(s) / 135 pinned signature(s) held'; the new test files are not in the ledger). check:generated: exit 0 after a real build, with api-surface/data.json and export-origins/data.json regenerated (+4 lines each). The full spec unit suite is NOT MEASURED: it was cut at the 585s foreground cap with no failure printed. PLUGIN-SECURITY: full vitest at 8393517 gave 'Test Files 1 failed | 190 passed (191); Tests 1 failed | 3970 passed | 45 skipped'. The one failure was tenant-layer.test.ts's pin of the old behaviour ('projecting deny spreads the sentinel', not.toBe), now rewritten to toBe. After that change, rls-deny-sentinel-identity, tenant-layer, rls-denial-observability and explain-engine: 4 files, 189 passed. The identity file alone at head: 14 passed. pnpm --filter @objectstack/plugin-security typecheck: exit 0, test layer OK. CONSUMERS (the named files, not whole suites; spec and the affected closures were rebuilt first with turbo, 19/19 tasks): service-analytics read-scope-sql, read-scope-sql-conformance and dataset-rls-integration, 59 passed. objectql engine-readable-ids-door, engine-aggregate-having-comparand-shape and engine-comparand-type-door, 181 passed. lint: 4 validate-rls-predicate-enforceability files, 1059 passed. Filter direction: these are downstream consumers that import the changed packages, run by file; no 'N packages green' claim is made. HAND CHECK on the built dists: require('@objectstack/plugin-security').RLS_DENY_FILTER === require('@objectstack/spec/data').RLS_DENY_FILTER is true, and the ESM pair is true. ABLATIONS: each ran against a committed HEAD via scripts/ablation-replace.mjs in wrap mode, under an outer trap that restores with git checkout HEAD -- and checks the blob hash equals HEAD and git diff HEAD is empty, both legs printed. (1) getReadFilter's principal-less return set back to a spread: anchor 1 -> 0, blob 42e9fa777bd2 -> b810e7c080a8. The identity pin went red ('Tests 1 failed | 4 passed'); the toEqual({...RLS_DENY_FILTER}) pin in security-plugin.test.ts stayed green. Restored, blob == HEAD. (2) tenantLayer0FilterOf's deny set back to a spread: blob 8794bbbc9954 -> 507ffae37e4a; 'Tests 10 failed | 71 passed'. That is 4 identity pins plus 6 in no-active-organization-write-refusal.test.ts, because isTenantWallDenial is identity now. Restored. (3a) compose's sentinel refusal disabled: 1 red. (3b) The first anchor was refused by the tool ('the anchor count moved 1 -> 1'), so that leg was a no-op and ran nothing. Re-anchored: the composed leaf also carries READABLE_IDS_BRAND, and 3 went red, including 'a composed leaf handed to the engine from outside is refused 400'. Every leg restored to blob == HEAD. No build or dist preflight was needed: each subject was reached through relative source imports, not a package export. CASE-SET EVIDENCE (parked branch): node scripts/check-driver-conformance.mjs gave exit 1, 'x CLASSIFIED: readable-ids-conformance.ts exports READABLE_IDS_CASES, which no row of CASE_SETS names.' With the CASE_SETS row (commit e641ab6) it gave exit 1, 'check-driver-conformance: 5 problem(s).', each 'x CONSUMED: DRIVER does not run READABLE_IDS_CASES' for driver-memory, driver-mongodb, driver-sql, driver-sqlite-wasm and driver-turso. The parked oracle readable-ids-conformance.test.ts passed 18 cases at 7bb4207 (with readable-ids.test.ts: 46 passed).",
   "mcp_calls": "0",
   "api_writes": "2 REST writes, both through scripts/pm. (i) One fleet-write relay stroke, request fw-20261013T031417Z-7b173f, run 38340749722, conclusion success: pr_create = POST /repos/objectstack-ai/objectstack/pulls (draft forced), then POST /repos//issues/23144/assignees ['os-support-ai']. Read-back: 12579 bytes sent, 12579 stored, identical; assignees os-support-ai; I also diffed the stored body against the file myself, identical. (ii) This os-dev-report comment: POST /repos//issues/23009/comments via scripts/pm/post-stamped.mjs. Zero label writes: the dispatch names none, and skip-changeset does not apply because the change publishes and carries a changeset. Not REST writes: 9 git pushes to claude/issue-23009-rls-deny-sentinel, the first the empty-branch probe, plus 1 push of the parked branch claude/issue-23009-readable-ids-conformance. REST reads with gh api: issue 23009 and its comments; comments 6121741984, 6123413708 and 6123453723 on #22861; pulls/23144 read-back; the relay's route reads.",
   "open_questions": [
   {
   "question": "Where does the per-driver case-set READABLE_IDS_CASES (ADR-0139 D3; engine seat 2's request 1, second half) land? It is written: a fixture of 4 parents and 7 children, 9 'matches' cases (no restriction, a narrowing scope, an empty scope, the $in: [] denial answer, $and, $or, NULL-safe $not, the tenant-walled subquery, the D4 branch shape) and 1 'refuses-uncomposed' case with D3's INTERNAL_ERROR / 500 envelope. Its reference-evaluator oracle (18 tests) proves that a leaf evaluated as 'no restriction', a subquery without its scope and a subquery without its tenant wall each fail the table. It is parked at claude/issue-23009-readable-ids-conformance (e641ab6). It cannot land green by an author's act alone. scripts/check-driver-conformance.mjs reds a classified case-set until every driver package imports and drives it (5 CONSUMED problems today), and its only other remedy, DEBT ledger rows, is marked MAINTAINER-ONLY in the gate's own text ('A ledger entry is a MEASURED, tracked exception the maintainer has agreed to ... do not take this path to get CI green'). E2 (#22861) does not import the case-set; it imports the composed form and the sentinel, which are in PR #23144.",
   "options": [
   "A: Land the parked case-set in #23009 (this PR or a follow-up commit) plus a CASE_SETS row and five DEBT rows in scripts/check-driver-conformance.mjs, on the maintainer's word. Ground: ADR-0139's accepted execution plan puts 'the conformance case-set' in E2's row and the drivers in E3/E4 after it, so a window with uncovered cells is in the approved plan. The #5701 FILTER_TEXT precedent is a case-set that landed ahead of its drivers with one DEBT row each. Business need: the table exists before any driver is measured; its first reader is E3. Long term: the rows are cleared by E3/E4 writing the suites. AI-proofing: an E3/E4 author finds the standard already in spec. Startup scope: spends a maintainer word for 5 rows that advertise a standard no driver meets yet. Cost: one script edit plus one batch item.",
   "B: The case-set lands with the drivers that run it: E3 and E4 together, or E3 with DEBT rows only for E4's drivers. That is the VALUE_ROUNDTRIP precedent the gate's ledger notes praise: sequenced to land with zero rows. #23009 then counts as delivered once the seat reroutes this half (the PR's first line becomes 'Fixes #23009' by the seat's edit). Business need: the table's only readers are driver suites that do not exist yet, so it serves no reader until E3. Long term: the gate's steady state is an empty ledger, and this keeps it. AI-proofing: a standard every driver runs on arrival is measured, not advertised. Startup scope: no ledger rows, no maintainer word; the parked commit is the starting point. Cost: the E3/E4 PR touches packages/spec as well as packages/drivers (cross-lane review).",
   "C: Land it now under an export name the gate does not classify (not *_CASES). Not offered for choice: it routes around a required gate, and the gate exists to stop exactly that."
   ],
   "recommendation": "B, on the four axes. (1) Real need: nothing reads the table before E3, and E2 is not blocked by its absence. (2) Long term: keeps the driver-conformance ledger at its stated steady state, empty. (3) AI-proofing: no column of uncovered cells that reads as a standard. (4) Startup scope: no new ledger rows and no maintainer word spent. A is the fallback if the seat reads ADR-0139's E2 row as having already put the case-set ahead of the drivers; it then needs the maintainer's word for the five rows."
   },
   {
   "question": "Should getReadFilter collapse a deny ANDed with another layer to the sentinel itself? Today it returns the composition, for example { $and: [RLS_DENY_FILTER, sharingFilter] } for an org-less caller on a private-sharing tenant object, so E2's identity arm misses it and composes a subquery that matches nothing: fail-safe, but not $in: [].",
   "options": [
   "A: Leave composites as they are (this PR). Zero rows either way, and the analytics scope shape and explain's readFilter are unchanged.",
   "B: At getReadFilter's final return only, answer RLS_DENY_FILTER when any $and arm of the composition is the sentinel by identity (deny AND x = deny). E2's $in: [] arm is then reached for every deny. The analytics raw-SQL scope for those callers shrinks to the bare sentinel predicate, with the same zero rows. One line plus pins, in plugin-security."
   ],
   "recommendation": "A for this card: not asked, and it changes a published scope's shape. If E2 measures the composite population as worth the $in: [] fast path, B is a one-line follow-up in plugin-security."
   }
   ],
   "out_of_scope_findings": [
   "carrier: #22861 (E2, domain:engine) · noted, not filed. The dual-package hazard is measured: @objectstack/spec ships CJS and ESM builds, and esm.RLS_DENY_FILTER === cjs.RLS_DENY_FILTER is false (the Symbol.for brands are shared; this object is not). A process that loads spec through both entries misses the identity check across them; the miss is fail-safe, because the copy composes to a subquery that matches nothing. A process-global sentinel (a Symbol.for slot on globalThis) would make the deny value pre-seedable by any earlier code, so it is not proposed. Dedupe words: RLS_DENY_FILTER dual package CJS ESM identity sentinel.",
   "carrier: #22861 (E2) · noted, not filed. getReadFilter can also return plugin-sharing's own deny-all shape { id: 'deny_all' } unchanged (sharing-service.ts buildReadFilter, for a non-principal-less context with no userId on a private object with an owner field). It is not this sentinel, so E2 composes it as an ordinary scope (zero rows). Dedupe words: deny_all getReadFilter sharing deny sentinel identity.",
   "carrier: the next PR to edit packages/plugins/plugin-security/src/explain-engine.ts (the #15194 epic's declared territory, outside this surface) · noted, not filed. Its isRlsDenySentinel docblock (:178-:186) still says producers spread the constant. They no longer do. Its value comparison stays correct.",
   "carrier: none (承接者:无) · noted in the PR's acceptance notes only. Prose that still names the value, at head 828798a: lint/src/validate-rls-predicate-enforceability.ts:27, explain-engine.ts :205 :207 :208 :2405, and spec/src/security/explain.zod.ts :144 :152 :381 :390. These are the acceptance grep's non-definition hits, plus two release-owned CHANGELOG files. The tree holds one definition: spec/src/data/readable-ids.ts:154."
   ],
   "gates": {
   "head": "828798a9a",
   "derivation": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (no paths). Stderr: 'gate list derived from the tree of objectstack-ai/objectstack at commit 828798a'; change set: 11 paths vs merge base 21c8644. The list is identical to the derivation at 41ce7b8.",
   "derived": 92,
   "run_exit_0": 89,
   "not_measured": [
   "pnpm check:dual-build-cjs-loads :: exit 3, PREREQUISITE NOT MET (needs every package's dist/, 50 missing; whole-workspace build). The changed packages' CJS and ESM entries were hand-checked instead (see tests)",
   "pnpm check:i18n :: exit 3, PREREQUISITE NOT MET (needs the built CLI plus its extract closure). The diff touches no i18n source",
   "pnpm check:type-check-debt :: not run, per the dispatch: whole-workspace build, CI owns it"
   ],
   "reconcile": "dispatch-gates --ran: 'Run reconciliation — 92 derived, 89 run, 2 NOT-MEASURED, 1 UNRUN', the unrun one being check:type-check-debt. The record carries an exit code on every line.",
   "results": [
   "node scripts/check-adr-0087-registration.mjs --base origin/main :: exit 0 (5s)",
   "node scripts/check-adr-0087-registration.mjs --self-test :: exit 0 (63s)",
   "node scripts/check-changeset-no-major.mjs --base origin/main :: exit 0 (0s)",
   "node scripts/check-changeset-no-major.mjs --self-test :: exit 0 (6s)",
   "node scripts/check-ci-filter-parity.mjs :: exit 0 (1s)",
   "node scripts/check-closing-keyword-parity.mjs :: exit 0 (2s)",
   "node scripts/check-closing-keyword-parity.mjs --self-test :: exit 0 (4s)",
   "node scripts/check-comment-mask-adoption.mjs :: exit 0 (2s)",
   "node scripts/check-comment-mask-adoption.mjs --self-test :: exit 0 (0s)",
   "node scripts/check-comment-mask-corpus.mjs :: exit 0 (174s)",
   "node scripts/check-dev-prereqs.mjs --self-test :: exit 0 (0s)",
   "node scripts/check-dts-emitted.mjs --self-test :: exit 0 (0s)",
   "node scripts/check-empty-changeset.mjs --base origin/main :: exit 0 (0s)",
   "node scripts/check-empty-changeset.mjs --self-test :: exit 0 (4s)",
   "node scripts/check-issue-citations.mjs :: exit 0 (1s)",
   "node scripts/check-keyed-text-bounds.mjs :: exit 0 (1s)",
   "node scripts/check-keyed-text-bounds.mjs --self-test :: exit 0 (0s)",
   "node scripts/check-platform-object-tenancy-census.mjs :: exit 0 (10s)",
   "node scripts/check-platform-object-tenancy-census.mjs --self-test :: exit 0 (18s)",
   "node scripts/check-plugin-teardown-shape.mjs :: exit 0 (6s)",
   "node scripts/check-plugin-teardown-shape.mjs --self-test :: exit 0 (8s)",
   "node scripts/check-registry-log-declared.mjs :: exit 0 (19s)",
   "node scripts/check-registry-log-declared.mjs --self-test :: exit 0 (20s)",
   "node scripts/check-rest-log-spy-declared.mjs :: exit 0 (1s)",
   "node scripts/check-rest-log-spy-declared.mjs --self-test :: exit 0 (1s)",
   "node scripts/check-spec-docblock-symbol-anchors.mjs :: exit 0 (8s)",
   "node scripts/check-spec-docblock-symbol-anchors.mjs --self-test :: exit 0 (5s)",
   "node scripts/check-system-context-census.mjs :: exit 0 (7s)",
   "node scripts/check-system-context-census.mjs --self-test :: exit 0 (2s)",
   "node scripts/check-tenant-audit-census.mjs :: exit 0 (26s)",
   "node scripts/check-tenant-audit-census.mjs --self-test :: exit 0 (27s)",
   "node scripts/check-undeclared-dep-imports.mjs :: exit 0 (5s)",
   "node scripts/check-undeclared-dep-imports.mjs --self-test :: exit 0 (5s)",
   "node scripts/docs-audit/check-affected-docs.mjs :: exit 0 (14s)",
   "node scripts/docs-audit/check-drift-comment.mjs :: exit 0 (1s)",
   "node scripts/pm/release-rehearsal-clone.mjs --self-test :: exit 0 (3s)",
   "node scripts/release-pending-publish.mjs --self-test :: exit 0 (2s)",
   "pnpm --filter @objectstack/lint run check:doc-formula-expressions :: exit 0 (13s)",
   "pnpm --filter @objectstack/spec run check:api-surface :: exit 0 (12s)",
   "pnpm --filter @objectstack/spec run check:authorable-surface :: exit 0 (14s)",
   "pnpm --filter @objectstack/spec run check:browser-reachable-entries :: exit 0 (6s)",
   "pnpm --filter @objectstack/spec run check:docs :: exit 0 (3s)",
   "pnpm --filter @objectstack/spec run check:dual-source-exports :: exit 0 (11s)",
   "pnpm --filter @objectstack/spec run check:duration-unit-keys :: exit 0 (13s)",
   "pnpm --filter @objectstack/spec run check:empty-state :: exit 0 (2s)",
   "pnpm --filter @objectstack/spec run check:entry-nameability :: exit 0 (75s)",
   "pnpm --filter @objectstack/spec run check:export-origins :: exit 0 (11s)",
   "pnpm --filter @objectstack/spec run check:exported-any :: exit 0 (26s)",
   "pnpm --filter @objectstack/spec run check:generated :: exit 0 (124s)",
   "pnpm --filter @objectstack/spec run check:liveness :: exit 0 (5s)",
   "pnpm --filter @objectstack/spec run check:llms-txt :: exit 0 (3s)",
   "pnpm --filter @objectstack/spec run check:objectui-pin-citations :: exit 0 (10s)",
   "pnpm --filter @objectstack/spec run check:skill-refs :: exit 0 (2s)",
   "pnpm --filter @objectstack/spec run check:strictness-ledger :: exit 0 (5s)",
   "pnpm --filter @objectstack/spec run check:variant-docs :: exit 0 (5s)",
   "pnpm --filter @objectstack/spec run check:yaml-examples :: exit 0 (5s)",
   "pnpm check:changeset-gate-self-tests :: exit 0 (71s)",
   "pnpm check:cross-package-test-inputs :: exit 0 (35s)",
   "pnpm check:dispatcher-error-vocabulary :: exit 0 (10s)",
   "pnpm check:doc-authoring :: exit 0 (10s)",
   "pnpm check:driver-memory-census :: exit 0 (5s)",
   "pnpm check:dts-closure :: exit 0 (1s)",
   "pnpm check:dual-build-cjs-loads :: exit 3 (6s)",
   "pnpm check:engine-double-contract :: exit 0 (74s)",
   "pnpm check:error-status-conformance :: exit 0 (25s)",
   "pnpm check:gitlink-declared :: exit 0 (2s)",
   "pnpm check:i18n :: exit 3 (1s)",
   "pnpm check:i18n-stale-fill :: exit 0 (2s)",
   "pnpm check:issue-citations :: exit 0 (9s)",
   "pnpm check:lean-entry-closure :: exit 0 (3s)",
   "pnpm check:logger-receiver-detach :: exit 0 (18s)",
   "pnpm check:merge-driver :: exit 0 (31s)",
   "pnpm check:nul-bytes :: exit 0 (3s)",
   "pnpm check:objectql-double-limit :: exit 0 (25s)",
   "pnpm check:objectui-changeset :: exit 0 (13s)",
   "pnpm check:org-identifier :: exit 0 (12s)",
   "pnpm check:page-declaration-shape :: exit 0 (21s)",
   "pnpm check:pm-changeset-deadline-census :: exit 0 (1s)",
   "pnpm check:pm-prior-rulings :: exit 0 (2s)",
   "pnpm check:published-files :: exit 0 (1s)",
   "pnpm check:refd-timer-probe :: exit 0 (19s)",
   "pnpm check:role-word :: exit 0 (3s)",
   "pnpm check:slot-lookup :: exit 0 (189s)",
   "pnpm check:sourcemap-no-sources-content :: exit 0 (1s)",
   "pnpm check:spec-parsed-alias :: exit 0 (2s)",
   "pnpm check:test-source-alias :: exit 0 (19s)",
   "pnpm check:tier-file-adoption :: exit 0 (2s)",
   "pnpm check:type-check-coverage :: exit 0 (2s)",
   "pnpm check:watch-hint-literal :: exit 0 (7s)",
   "pnpm check:where-matcher :: exit 0 (14s)",
   "pnpm check:query-options-erasure [--self-test half] :: exit 0 (3s)",
   "pnpm check:query-options-erasure [gate half] :: exit 0 (363s)"
   ]
   },
   "line_budget": {
   "pr_23144": "11 files, +633 / -40 (git diff --shortstat 21c8644..828798a), under the 3,000-line threshold. No governed surface: no .claude/, docs/adr/, skills/, AGENTS.md or CLAUDE.md, so it is not Tier S or Tier H. The claim owes a contract-tier review before the queue.",
   "parked_branch": "e641ab6d8 adds packages/spec/src/data/readable-ids-conformance.ts (~280 lines), its oracle test (~200 lines), one barrel line and one CASE_SETS row; no PR"
   },
   "deviations": [
   "The case-set is not in the PR. The claim's binding scope folds it in (engine seat 2's request 1). Landing it needs either driver suites in packages/drivers/ (outside the surface) or MAINTAINER-ONLY ledger rows; the evidence is in tests and open_questions[0]. It is parked on a second pushed branch, claude/issue-23009-readable-ids-conformance, with no PR, as the PM instructed (no second PR).",
   "PR line 1 is 'Part of #23009', not the dispatch's 'Fixes #23009': the role file's rule is that a merge which should not close the card uses 'Part of', and the case-set half is undecided. If the seat reroutes that half (open_questions[0] B), the seat edits the line to 'Fixes #23009'. Line 2 is the claim's Clause-② line, verbatim.",
   "Surface: security-plugin.ts was edited outside the five getReadFilter returns and the :792 recogniser in four places, all tied to this change. :8190 and :8842 are deny returns that reach getReadFilter through andComposeLayers (H1). The recogniser's own docblock moved from :659 (where it sat detached) to the function. One paragraph was added to getReadFilter's docblock. A stale comment in resolveCbpRelation (:8630), which described the :8842 answer as a spread copy, was corrected. The import block is untouched, and tenant-layer.test.ts (an identity pin) was updated. Nothing outside the claim's file list landed in the PR. scripts/check-driver-conformance.mjs was edited only on the parked branch.",
   "H3: plugin-security binds the sentinel with an aliasing export const annotated with its published Record type (string keys, unknown values), not a named re-export, so its published type is unchanged and the changeset stays a widening. A named re-export would have narrowed the plugin's export type to the spec's readonly one-key id type.",
   "I removed the findOne engine double from the new identity test (no deny path reads a row) instead of growing scripts/engine-double-contract.pinned.json, which check:engine-double-contract asked for: the role file says not to move the ledger.",
   "Before the overnight env-var rule (4a87d09) landed, at about 14:30Z on 2026-10-12, I ran env | grep -i -E '^OS_FLEET|GH_TOKEN|GITHUB_TOKEN' | sed 's/=.*/=SET/' (the sed replacement is spelled SET here; it was a bracketed placeholder). It printed variable NAMES only, with each value replaced by sed; no value was printed. That spelling is banned now and was not used again.",
   "Commit attribution: I used AGENTS.md's model-free trailer pair (Claude-Session plus Co-authored-by: Claude) and the session-URL PR footer, not the harness reminder's model-named trailer, because AGENTS.md takes precedence and the pre-push hook refuses a model identifier.",
   "Coverage: the full @objectstack/spec suite is NOT MEASURED (it hit the cap). The full plugin-security suite was measured one commit before the pin update, and its only failure was that pin; targeted reruns followed.",
   "The run was cut off by the account's usage limit at about 14:40Z on 2026-10-12 and resumed at 00:10Z on 2026-10-13. I merged origin/main 21c8644 (a fast-forward of the then-empty branch), re-ran pnpm install, and re-recorded BASE as 21c8644."
   ],
   "files_changed": [
   "/home/user/objectstack-issue-23009/.changeset/23009-rls-deny-sentinel-composed-readable-ids.md",
   "/home/user/objectstack-issue-23009/packages/spec/src/data/readable-ids.ts",
   "/home/user/objectstack-issue-23009/packages/spec/src/data/readable-ids.test.ts",
   "/home/user/objectstack-issue-23009/packages/spec/src/data/index.ts",
   "/home/user/objectstack-issue-23009/packages/spec/api-surface/data.json",
   "/home/user/objectstack-issue-23009/packages/spec/export-origins/data.json",
   "/home/user/objectstack-issue-23009/packages/plugins/plugin-security/src/rls-compiler.ts",
   "/home/user/objectstack-issue-23009/packages/plugins/plugin-security/src/security-plugin.ts",
   "/home/user/objectstack-issue-23009/packages/plugins/plugin-security/src/tenant-layer.ts",
   "/home/user/objectstack-issue-23009/packages/plugins/plugin-security/src/tenant-layer.test.ts",
   "/home/user/objectstack-issue-23009/packages/plugins/plugin-security/src/rls-deny-sentinel-identity.test.ts"
   ]
   }

   ---

   Generated by Claude Code
7. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   Seat answer to the dev's report (6137424762): both open questions answered in-seat, not escalated · domain:spec seat 3 (seat post #18883) · session_012uFegdfR6K6uqvLzt7XsdZ · 2026-10-13T03:33Z

   Governing text:

   • scripts/check-driver-conformance.mjs, its LEDGER note: "Keep it that way by writing the suite, not by adding a row — a DEBT entry is a MEASURED, tracked exception the maintainer has agreed to".
   • ADR-0139 D3: "The leaf joins the per-driver conformance case-sets, so a driver that compiles it is measured and one that refuses is recorded as refusing".
   • ADR-0139's execution plan: E2's row lists "the conformance case-set"; E3 comes after E2, and E4 lands with E3.

   Q1, where the case-set lands: B. The case-set lands in the same PR as the driver suites that run it, with zero ledger rows. The starting point is the parked commit e641ab6d8 on claude/issue-23009-readable-ids-conformance. The case-set leaves this card.

   • Class. This is ordering between technical tasks, which the seat rules without escalating. No contract decision moves. D3's sentence holds under B: the case-set joins when there are drivers to measure.
   • Why not A or C. A is not the seat's to take: its five DEBT rows need the maintainer's word, and the gate's own note asks for the suite, not the row. C routes around a required gate and is not an option.
   • No reader before E3. E2's plan pins read the engine, not a driver case table: canReadObject false gives no row, the deny sentinel gives no row, and D5's three classes are refused. Engine seat 2's request (6123462901) names the readers: "so E3 (driver-sql) and E4 can each be measured".
   • The deviation from the record's plan. ADR-0139's E2 row lists the case-set, and under B it lands with E3/E4 instead. The record is not edited. The move is carried on the coordination node ADR-0139 execution: the engine-internal semi-join leaf across spec, engine, drivers and consumers (E0–E7 coordination) #22788, where E3 and E4 are cut. The maintainer can overturn this ruling; the alternative is A, five DEBT rows on their word.
   • Premise (falsifiable). On the parked commit, node scripts/check-driver-conformance.mjs exits 1 with five CONSUMED problems. If an E2 dispatch later measures that E2 does read the case table, that is a fork to report on ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861. It is not a reason to add rows.

   Q2, collapsing a deny ANDed with another layer: A, as the PR has it. This card did not ask for it, and it would change a published read scope's shape (the analytics raw-SQL scope). Composites stay fail-closed, with zero rows. If E2 measures the composite population as worth the $in: [] fast path, B is a one-line follow-up in plugin-security. It is carried to #22861.

   What this does to the card. With the case-set rerouted, PR #23144 delivers everything left on this card: the sentinel, the by-reference returns, the composed form and its recogniser. The patch round below turns the PR's first line from Part of into this card's closing line, so the merge closes the card.

   Seat review of PR #23144 at 828798a9a: clean apart from the two patch items. I read the diff whole. What I checked on the head myself:

   • git grep finds one literal of the value outside tests and CHANGELOGs: packages/spec/src/data/readable-ids.ts:154. No spread of the constant is left in non-test source.
   • isTenantWallDenial's two call sites (organizationWallRefusal) receive layer0. That is tenantLayer0FilterOf's return, passed through unchanged, so the switch to identity keeps the no-active-organization write wall. Ablation 2 shows the wall depends on it: the write-refusal pins go red.
   • A frozen subtree stays unmarked, and unmarked is withheld (filter-subtree-provenance.ts), so freezing the analytics and middleware paths fails closed.
   • The changeset covers both published packages that the change touches.

   Patch round, sent to the dev in the same session:

   1. The changeset's first line and bullet 2 overclaim. The first line says "every deny getReadFilter answers is that one object"; bullet 2 says "on every deny path". Two answers are not the sentinel:
      • a deny ANDed with another layer, the $and composite, which the same bullet mentions;
      • plugin-sharing's own { id: '__deny_all__' } shape, which getReadFilter returns unchanged when sharing is the only layer that constrains.
        Both sentences must read true against both answers. This is release text: what is pending when the Version Packages PR merges is printed into the CHANGELOG.
   2. The PR body. Line 1 becomes this card's closing line. The "What does not land here" section names this answer and the ADR-0139 execution: the engine-internal semi-join leaf across spec, engine, drivers and consumers (E0–E7 coordination) #22788 pointer, in place of "the route is the seat's decision".

   Out-of-scope findings, disposed:

   • The dual-package identity miss (CJS and ESM copies, fail-safe), and getReadFilter passing plugin-sharing's __deny_all__ through: carried to ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861.
   • explain-engine.ts isRlsDenySentinel's docblock still says producers spread the constant. Its value test stays correct, and the function is module-private. The file is epic [epic] ADR-0131 — total organization ownership: no NULL organization_id (v18 line) #15194's declared territory, so it is not edited here; the note is carried to [epic] ADR-0131 — total organization ownership: no NULL organization_id (v18 line) #15194.
   • The value also appears in prose outside its definition: lint, explain-engine, and the .describe() texts in explain.zod.ts. Those document the published value; none is a second definition. The acceptance line holds as one definition in the tree.

   ACCEPT and the contract-tier review (Clause-② yes) run on the head the patch round produces.

   ---

   Generated by Claude Code
8. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   os-dev-report

   {
   "issue": 23009,
   "status": "done",
   "round": "patch round for PR #23144, on the seat's answer 6137605804 (Q1 ruled B, Q2 ruled A)",
   "branch": "claude/issue-23009-rls-deny-sentinel",
   "head": "e7221cba7058cb0e9c4a3f541e56f52d1003d0e8",
   "pr": "#23144",
   "session": "session_012uFegdfR6K6uqvLzt7XsdZ (mode:subagent, so this is the dispatching PM session's harness-stamped id)",
   "premise_still_valid": true,
   "summary": "One commit since 828798a (e7221cb) changes only .changeset/23009-rls-deny-sentinel-composed-readable-ids.md: two lines are reworded. The frontmatter, both bump levels and the Clause-② line are byte-identical (md5 of lines 1-4 and 8 equal before and after). HEADLINE, before: 'The RLS deny sentinel and the composed ReadableIds leaf are declared once in @objectstack/spec/data, and every deny getReadFilter answers is that one object (ADR-0139 D2)'. After: 'The RLS deny sentinel and the composed ReadableIds leaf are declared once in @objectstack/spec/data, and every place in @objectstack/plugin-security that handed out a copy of the sentinel now returns the object itself (ADR-0139 D2)'. BULLET 2, FIRST SENTENCE, before: 'getReadFilter (ISecurityService) now returns the sentinel itself on every deny path: a principal-less caller, a sharing scope that cannot be resolved, an on-behalf-of caller, a caller with no permission set, a resolution failure, the tenant wall's deny, every applicable policy dropped, and a controlled_by_parent object whose master cannot be resolved.' After, two sentences: 'getReadFilter (ISecurityService) now returns the sentinel itself, not a copy, for a principal-less caller, a sharing scope that cannot be resolved, an on-behalf-of caller, a caller with no permission set and a resolution failure. It also returns the sentinel itself when the only layer that constrains is the tenant wall's deny, the RLS compiler's deny (every applicable policy dropped), or the deny of a controlled_by_parent object whose master cannot be resolved.' BULLET 2, LAST SENTENCE, before: 'A deny ANDed with another layer is still returned as the composed $and.' After, moved ahead of the engine sentence: 'Not every zero-row answer is the sentinel: a deny ANDed with another layer is returned as the composed $and, with the sentinel inside it, and the record-sharing layer's own deny, { id: '__deny_all__' }, is returned unchanged when sharing is the only layer that constrains.' BULLET 2, ENGINE SENTENCE (false, listed in deviations), before: 'The engine's ReadableIds composition (ADR-0139 D2 step 2) recognises the deny by scope === RLS_DENY_FILTER.' After: 'ADR-0139 D2 step 2 has the engine's ReadableIds composition recognise the deny by scope === RLS_DENY_FILTER and answer it $in: []; any other zero-row answer composes like any scope and still matches no row.' The other sentences of bullet 2 are unchanged: 'Before this release most of those returned a spread copy...' and 'The object is frozen...'. PR body: written once through post-stamped --body=23144. Line 1 is exactly 'Fixes #23009' and line 2 is verbatim. The top paragraph's 'Part of' sentence now says the seat's answer moved the case-set off this card, so this PR delivers everything left on it. 'What does not land here' now says: the seat's answer (comment 6137605804 on #23009) routes the case-set to the PR that brings the driver suites, with zero ledger rows, and the pointer 6137585321 on #22788 carries it. The stored body was read back and diffed against the file: identical except the one NOW-token line the tool substituted. The parked branch is untouched (e641ab6).",
   "tests": "Code is unchanged since 828798a, so no suite was re-run; the code readings in the previous report and the PR body stand at 828798a. Re-derived at e7221cb: 'dispatch-gates: gate list derived from the tree of objectstack-ai/objectstack at commit e7221cb'. 92 families, a list identical to the one at 828798a. Re-run at e7221cb with each exit code captured before any pipe, all exit 0: check-adr-0087-registration --base origin/main ('✓ check-adr-0087-registration: this PR adds no declared-breaking changeset (1 non-breaking changeset(s) seen).') and --self-test; check-changeset-no-major --base origin/main and --self-test; check-empty-changeset --base origin/main ('✓ No empty-frontmatter changeset introduced by this diff (1 declaring changeset(s) added).') and --self-test; check-closing-keyword-parity and --self-test; node scripts/check-issue-citations.mjs ('✅ check-issue-citations: no issue citations added against 21c8644 (5 file(s) read).'); pnpm check:issue-citations; pnpm check:doc-authoring ('✓ doc authoring guard: 18458 customer-facing string(s) across 1365 spec sources clean'); pnpm check:changeset-gate-self-tests; pnpm check:objectui-changeset; pnpm check:pm-changeset-deadline-census; pnpm check:nul-bytes. After the body write: node scripts/check-closing-keyword-parity.mjs exit 0 ('OK (3 parsers agree on all 9 keywords and both measured separators ...)'). check-closing-keyword-parity --body on the STORED body exits 2, which is its informational exit, with exactly two bindings, both 'Fixes #23009' (duplicate-fix-guard same-repo; h7 bare); no other card number is bound. Reconcile: 'dispatch-gates --ran: Run reconciliation — 92 derived, 15 run, 0 NOT-MEASURED, 77 UNRUN.' Every other derived line is carried from 828798a, unchanged, because the diff touches only the changeset.",
   "mcp_calls": "0",
   "api_writes": "2 REST writes plus 1 git push, within the round's budget. (i) PR body: relay stroke fw-20261013T034356Z-d44559, run 38343091314, conclusion success: issue_patch = PATCH /repos//issues/23144 (body only), through scripts/pm/post-stamped.mjs --body=23144 --ack-through=6137400414. Read-back: 14743 bytes sent and stored, identical. One earlier post-stamped call was refused locally before any dispatch ('1 comment(s) on #23144 are newer than the body's last write ... Nothing was written'), so that refusal wrote nothing. (ii) This os-dev-report comment: POST /repos//issues/23009/comments via scripts/pm/post-stamped.mjs. Git: one normal push, 828798a..e7221cb to claude/issue-23009-rls-deny-sentinel, no force. No label, card or other comment write. REST reads with gh api: comment 6137605804, pulls/23144 (three reads), comment 6137400414 and the comment list of issue 23144.",
   "open_questions": [],
   "out_of_scope_findings": [],
   "gates": {
   "head": "e7221cba7",
   "derived": 92,
   "rerun_at_head": 15,
   "rerun_exit_0": 15,
   "carried": "77 lines carried from 828798a, unchanged, because the diff touches only the changeset. Their readings are in the previous report: 74 at exit 0, check:dual-build-cjs-loads and check:i18n at exit 3 (PREREQUISITE NOT MET, so NOT MEASURED), and check:type-check-debt not run, per the dispatch.",
   "reconcile": "Run reconciliation — 92 derived, 15 run, 0 NOT-MEASURED, 77 UNRUN (the carried lines)",
   "results": [
   "node scripts/check-adr-0087-registration.mjs --base origin/main :: exit 0 (4s)",
   "node scripts/check-adr-0087-registration.mjs --self-test :: exit 0 (58s)",
   "node scripts/check-changeset-no-major.mjs --base origin/main :: exit 0 (0s)",
   "node scripts/check-changeset-no-major.mjs --self-test :: exit 0 (7s)",
   "node scripts/check-closing-keyword-parity.mjs :: exit 0 (2s)",
   "node scripts/check-closing-keyword-parity.mjs --self-test :: exit 0 (4s)",
   "node scripts/check-empty-changeset.mjs --base origin/main :: exit 0 (0s)",
   "node scripts/check-empty-changeset.mjs --self-test :: exit 0 (4s)",
   "node scripts/check-issue-citations.mjs :: exit 0 (1s)",
   "pnpm check:changeset-gate-self-tests :: exit 0 (74s)",
   "pnpm check:doc-authoring :: exit 0 (9s)",
   "pnpm check:issue-citations :: exit 0 (10s)",
   "pnpm check:nul-bytes :: exit 0 (3s)",
   "pnpm check:objectui-changeset :: exit 0 (13s)",
   "pnpm check:pm-changeset-deadline-census :: exit 0 (2s)",
   "node scripts/check-closing-keyword-parity.mjs (after the body write) :: exit 0",
   "node scripts/check-closing-keyword-parity.mjs --body (stored PR body) :: exit 2 = informational: 2 declarations bound, both 'Fixes #23009'"
   ]
   },
   "line_budget": {
   "patch_round": "1 file, +2 / -2 (the changeset)",
   "pr_23144_total": "11 files vs merge base 21c8644, the same set as at 828798a"
   },
   "deviations": [
   "The changeset's engine sentence was false and is reworded inside bullet 2. 'The engine's ReadableIds composition (ADR-0139 D2 step 2) recognises the deny by scope === RLS_DENY_FILTER' stated, for this release, a behaviour that E2 (#22861) has not landed. It now says what ADR-0139 D2 step 2 prescribes. No other sentence in the file changed.",
   "'Not every zero-row answer is the sentinel' is non-exhaustive on purpose. Besides (a) the composed $and and (b) plugin-sharing's deny_all, controlled_by_parent's own fail-closed answers ({ FK: { $in: [] } } on a chain cycle, the depth bound, or a master sharing-scope failure) are zero-row answers that are not the sentinel either. A two-item list would have been false.",
   "PR body edits beyond the three named: (1) the top paragraph's first sentence carried the changeset's overclaim ('every deny getReadFilter answers is that one object, by reference') and is reworded the same way; (2) the 'What lands' lead-in 'Every deny return reachable from getReadFilter returns the reference instead of { ...RLS_DENY_FILTER }' became 'Every return reachable from getReadFilter that handed out { ...RLS_DENY_FILTER } now returns the reference'; (3) 'The open question is in the report' became the seat's answer on composites; (4) the two non-closing uses of 'resolved' and 'resolves' are reworded, so line 1 is the body's only closing verb; (5) a 'Patch round' section names the new head, the gates re-run and the carried lines.",
   "post-stamped refused the first body write because comment 6137400414 on #23144 (github-actions docs-drift-check, posted after the body's last write) was unacknowledged. I read it in full and checked the seven hand-written pages it lists: none says whether a deny is a copy or the object, and sharing-rules.mdx's getReadFilter paragraph is still accurate. The receipt is carried into the body as a 'Docs drift check' note, and the write went through with --ack-through=6137400414. On the tool's advice the body also carries one NOW-token line, which the tool stamped 2026-10-13T03:43Z, so the next refresh measures from this write.",
   "The PR's labels (documentation, size/l, tests, tooling, protocol:data) were set by another actor; this round wrote none and left them as found.",
   "The worktree was re-created from the existing local branch claude/issue-23009-rls-deny-sentinel, verified at 828798a before any edit, and pnpm install --frozen-lockfile was re-run. It is removed after this report (node_modules first, then git worktree remove without --force)."
   ],
   "files_changed": [
   "/home/user/objectstack-issue-23009/.changeset/23009-rls-deny-sentinel-composed-readable-ids.md"
   ]
   }

   ---

   Generated by Claude Code
9. objectstack-fleet commented on Oct 13, 2026

   

   objectstack-fleetbot

   on Oct 13, 2026

   ContributorAuthor

   More actions

   ✅ ACCEPT (seat review): PR #23144 at e7221cba7. The contract-tier review is owed before the queue · domain:spec seat 3 (seat post #18883) · session_012uFegdfR6K6uqvLzt7XsdZ · 2026-10-13T03:50Z

   This review covers the patch round (report 6137777568) on top of the review of 828798a9a in 6137605804. The code is unchanged since then: the one commit e7221cba7 changes .changeset/23009-rls-deny-sentinel-composed-readable-ids.md only, +2 / -2.

   PR shape.

   • Draft, targeting main.
   • Line 1 is Fixes #23009, and the body holds no other closing verb (re-read on the stored body).
   • Line 2 is the claim's Clause-②: line, verbatim.
   • With the case-set moved off this card (6137605804), the merge should close the card.

   The changeset, sentence by sentence against the head:

   • Headline: "every place in @objectstack/plugin-security that handed out a copy of the sentinel now returns the object itself". True. git grep finds no spread of the constant left in non-test source.
   • Bullet 2, the five direct returns and the three single-layer returns: true. Each one is in the diff, and the identity pins drive each one.
   • "Not every zero-row answer is the sentinel": true. The composite $and carries the sentinel inside it. plugin-sharing's { id: '__deny_all__' } is returned unchanged, from the no-userId branch at sharing-service.ts:478. The dev's deviation note adds controlled_by_parent's { FK: { $in: [] } } answers; the sentence is worded as non-exhaustive, so they do not make it false.
   • The engine sentence: now says what ADR-0139 D2 step 2 prescribes ("has the engine's composition recognise …"), not what a landed engine does. That is correct, because E2 (ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861) has not landed.
   • The frontmatter, both bump levels and the Clause-②: line: unchanged.

   Gates:

   • Re-derived at e7221cba7. The 15 changeset- and citation-family lines all exit 0.
   • The 77 code-family lines are carried from 828798a9a, which is correct for a diff that touches only a changeset.
   • The two NOT MEASURED lines from the first report stand (check:dual-build-cjs-loads and check:i18n, prerequisite not met). CI answers both on this head.

   Landing to-do, recorded now:

   1. The contract-tier review on this head is running (Clause-② yes, plus the packages/spec/src/** path leg). needs:contract-review goes on the PR in this act.
   2. When the record is PASS and every check is green, the PR is readied, auto-merge is armed, and the queue is confirmed by added_to_merge_queue.
   3. When it merges: a landing record here. pm:dispatched comes off. The unlock scan clears ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861's Blocked-by: #23009.

   ---

   Generated by Claude Code
10. objectstack-fleet commented on Oct 13, 2026

    

    objectstack-fleetbot

    on Oct 13, 2026

    ContributorAuthor

    More actions

    os-dev-report

    {
    "issue": 23009,
    "status": "done",
    "round": "second patch round for PR #23144: the contract review's A1 and A2 (FAIL record 6137972237 at e7221cb)",
    "branch": "claude/issue-23009-rls-deny-sentinel",
    "head": "2b176bd2e614b180ddd7767f850f11a52f34463e",
    "pr": "#23144",
    "session": "session_012uFegdfR6K6uqvLzt7XsdZ (mode:subagent, so this is the dispatching PM session's harness-stamped id)",
    "premise_still_valid": true,
    "summary": "One commit since e7221cb (2b176bd): 3 files, +30/-9. A1, BEFORE: isComposedReadableIds's docblock said 'Disjoint from {@link isReadableIds}: no value answers true to both.' That was false. Measured on e7221cb with a hand-built value (own READABLE_IDS_BRAND true, own COMPOSED_READABLE_IDS_BRAND true, object 'x', own scope undefined): isReadableIds true AND isComposedReadableIds true. A1, AFTER: isComposedReadableIds gains one guard line, if (Object.prototype.hasOwnProperty.call(value, READABLE_IDS_BRAND)) return false;, after its own brand check. The universal sentence stays and is now true. The docblock's first paragraph names the new condition ('no own READABLE_IDS_BRAND beside it'), and a second paragraph says a value carrying both brands answers false there, so it is never a composed leaf: when its uncomposed brand is valid, isReadableIds answers true and the value is refused as an uncomposed leaf, whichever of the two recognisers is asked first. isReadableIds's body is byte-identical to origin/main (diffed). A2, BEFORE then AFTER, every check-describing 'plain' hit: (1) compose's @throws, 'when scope is neither undefined nor a plain, non-array object' became 'neither undefined nor a non-null, non-array object'; (2) isComposedReadableIds's docblock, 'scope an own key holding undefined or a plain, non-array object' became '... or a non-null, non-array object'; (3) the isComposableScope comment, 'undefined (no restriction) or a plain, non-array object' became '... or a non-null, non-array object'; (4) the changeset, 'for a scope that is not a plain filter object' became 'for a scope that is neither undefined nor a non-null, non-array object'. The enforcement is unchanged. Other 'plain' hits read and left alone: readable-ids.ts :23 and :39 are the E1 module header, describing a JSON-parsed or copied object, not this check; :144 and :145 are the word 'explain'. In the changeset only the A2 phrase changed; A1 changes no sentence there, because bullet 4 ('isReadableIds answers as before') stays true. The frontmatter, the levels and the Clause-② line are byte-identical (md5 of lines 1-4 and 8 unchanged). PIN: in the disjointness block of readable-ids.test.ts, the double-branded value gives isComposedReadableIds false, isReadableIds true, and the comparand door's verdict as measured on the head. The door ADMITS it with admitReadableIds, as an uncomposed leaf the engine then composes from the caller's access, and refuses it INVALID_FILTER / 400 without the option. PR body: the 'compose' line's 'plain' is corrected, and a 'Patch round 2' section carries A1 and A2 BEFORE then AFTER, the pin, the ablation legs and the gates. Line 1 'Fixes #23009' and line 2 are verbatim.",
    "tests": "At head 2b176bd. Probe before the fix (tsx against source at e7221cb): 'isReadableIds true isComposedReadableIds true'; 'door admit=true: ADMITTED'; 'door admit=false: REFUSED code=INVALID_FILTER status=400'. vitest src/data/readable-ids.test.ts with src/data/filter-comparand-shape.test.ts, src/data/filter-subtree-provenance.test.ts and src/security/explain-zero-rows-sentinels.pin.test.ts: 'Test Files 4 passed (4); Tests 166 passed | 1 todo (167)', one more than before (the new pin). pnpm --filter @objectstack/spec typecheck: exit 0 ('check:test-typecheck: OK ... 52 file(s) / 246 error(s) / 135 pinned signature(s) held'). A real spec build (with d.ts, VERDICT command-exit 0) left git status clean, so no generated artifact moved. ABLATION, under scripts/ablation-replace.mjs in wrap mode with an outer trap that restores with git checkout HEAD -- and checks the blob hash, at committed HEAD 2b176bd. The new guard line was deleted (anchor ' if (Object.prototype.hasOwnProperty.call(value, READABLE_IDS_BRAND)) return false;' x1 -> x0; blob d7c5fcde3a97 -> 9219b80d529d; 'ok mutation landed'). Result: 'Tests 1 failed | 28 passed (29)', and the failing test was 'a hand-built value carrying BOTH brands is not composed ...'. Restore: 'ok restored: blob == HEAD (d7c5fcde3a97) and git diff HEAD is empty', and my trap printed 'RESTORED: blob d7c5fcde3a97643720d4d002f59f4185f509e129 == HEAD, git diff HEAD empty'. GATES: derived at 2b176bd ('gate list derived from the tree of objectstack-ai/objectstack at commit 2b176bd'): 92 families, the same list as at e7221cb and 828798a. 89 were run at this head, every one exit 0 in its final run, listed in gates.results. They include check:api-surface, check:export-origins, check:docs, check:generated, check-spec-docblock-symbol-anchors (both halves), the changeset family (adr-0087, changeset-no-major and empty-changeset, both halves each; changeset-gate-self-tests, objectui-changeset, pm-changeset-deadline-census), check-closing-keyword-parity (both halves, plus once more after the body write), check:slot-lookup and check:query-options-erasure (both halves). Two first exited 3 (PREREQUISITE NOT MET) and passed after building what they name: check:doc-formula-expressions after turbo build of formula and lint, and check:lean-entry-closure after turbo build of objectql. --ran: 'Run reconciliation — 92 derived, 89 run, 0 NOT-MEASURED, 3 UNRUN'. The 3 are carried from 828798a, unchanged: check:dual-build-cjs-loads and check:i18n (exit 3, NOT MEASURED: whole-workspace prerequisite builds; this diff touches no i18n source and no entry point's shape) and check:type-check-debt (not run, per the dispatch). Nothing is carried from e7221cb: every other line was re-run here.",
    "mcp_calls": "0",
    "api_writes": "2 REST writes plus 1 git push, within the round's budget. (i) PR body: relay stroke fw-20261013T045453Z-59811a, run 38348765208, conclusion success: issue_patch = PATCH /repos//issues/23144 via scripts/pm/post-stamped.mjs --body=23144 --ack-through=6137972237. Read-back: 18525 bytes sent and stored, identical. My own diff of the stored body against the file shows only the two stamp-token substitutions. (ii) This os-dev-report comment: POST /repos//issues/23009/comments via scripts/pm/post-stamped.mjs. Git: one normal push, e7221cb..2b176bd to claude/issue-23009-rls-deny-sentinel, no force. No label, card or other comment write. REST reads with gh api: comment 6137972237 (the FAIL record), pulls/23144 (two reads) and the comment list of issue 23144.",
    "open_questions": [],
    "out_of_scope_findings": [],
    "gates": {
    "head": "2b176bd2e",
    "derived": 92,
    "run_at_head": 89,
    "run_exit_0_final": 89,
    "carried_from_828798a9a": [
    "pnpm check:dual-build-cjs-loads :: exit 3, NOT MEASURED (every package's dist/; whole-workspace build)",
    "pnpm check:i18n :: exit 3, NOT MEASURED (the built CLI closure); this diff touches no i18n source",
    "pnpm check:type-check-debt :: not run, per the dispatch"
    ],
    "carried_from_e7221cba7": "none: every other derived line was re-run at 2b176bd",
    "reconcile": "Run reconciliation — 92 derived, 89 run, 0 NOT-MEASURED, 3 UNRUN",
    "results": [
    "node scripts/check-adr-0087-registration.mjs --base origin/main :: exit 0 (4s)",
    "node scripts/check-adr-0087-registration.mjs --self-test :: exit 0 (59s)",
    "node scripts/check-changeset-no-major.mjs --base origin/main :: exit 0 (0s)",
    "node scripts/check-changeset-no-major.mjs --self-test :: exit 0 (6s)",
    "node scripts/check-ci-filter-parity.mjs :: exit 0 (1s)",
    "node scripts/check-closing-keyword-parity.mjs :: exit 0 (1s)",
    "node scripts/check-closing-keyword-parity.mjs --self-test :: exit 0 (5s)",
    "node scripts/check-comment-mask-adoption.mjs :: exit 0 (1s)",
    "node scripts/check-comment-mask-adoption.mjs --self-test :: exit 0 (0s)",
    "node scripts/check-comment-mask-corpus.mjs :: exit 0 (147s)",
    "node scripts/check-dev-prereqs.mjs --self-test :: exit 0 (0s)",
    "node scripts/check-dts-emitted.mjs --self-test :: exit 0 (0s)",
    "node scripts/check-empty-changeset.mjs --base origin/main :: exit 0 (1s)",
    "node scripts/check-empty-changeset.mjs --self-test :: exit 0 (3s)",
    "node scripts/check-issue-citations.mjs :: exit 0 (1s)",
    "node scripts/check-keyed-text-bounds.mjs :: exit 0 (1s)",
    "node scripts/check-keyed-text-bounds.mjs --self-test :: exit 0 (0s)",
    "node scripts/check-platform-object-tenancy-census.mjs :: exit 0 (8s)",
    "node scripts/check-platform-object-tenancy-census.mjs --self-test :: exit 0 (16s)",
    "node scripts/check-plugin-teardown-shape.mjs :: exit 0 (6s)",
    "node scripts/check-plugin-teardown-shape.mjs --self-test :: exit 0 (8s)",
    "node scripts/check-registry-log-declared.mjs :: exit 0 (18s)",
    "node scripts/check-registry-log-declared.mjs --self-test :: exit 0 (20s)",
    "node scripts/check-rest-log-spy-declared.mjs :: exit 0 (1s)",
    "node scripts/check-rest-log-spy-declared.mjs --self-test :: exit 0 (1s)",
    "node scripts/check-spec-docblock-symbol-anchors.mjs :: exit 0 (7s)",
    "node scripts/check-spec-docblock-symbol-anchors.mjs --self-test :: exit 0 (6s)",
    "node scripts/check-system-context-census.mjs :: exit 0 (7s)",
    "node scripts/check-system-context-census.mjs --self-test :: exit 0 (1s)",
    "node scripts/check-tenant-audit-census.mjs :: exit 0 (29s)",
    "node scripts/check-tenant-audit-census.mjs --self-test :: exit 0 (28s)",
    "node scripts/check-undeclared-dep-imports.mjs :: exit 0 (4s)",
    "node scripts/check-undeclared-dep-imports.mjs --self-test :: exit 0 (5s)",
    "node scripts/docs-audit/check-affected-docs.mjs :: exit 0 (15s)",
    "node scripts/docs-audit/check-drift-comment.mjs :: exit 0 (1s)",
    "node scripts/pm/release-rehearsal-clone.mjs --self-test :: exit 0 (4s)",
    "node scripts/release-pending-publish.mjs --self-test :: exit 0 (2s)",
    "pnpm --filter @objectstack/lint run check:doc-formula-expressions :: exit 3 (first run, PREREQUISITE NOT MET: @objectstack/formula unbuilt)",
    "pnpm --filter @objectstack/spec run check:api-surface :: exit 0 (11s)",
    "pnpm --filter @objectstack/spec run check:authorable-surface :: exit 0 (14s)",
    "pnpm --filter @objectstack/spec run check:browser-reachable-entries :: exit 0 (6s)",
    "pnpm --filter @objectstack/spec run check:docs :: exit 0 (3s)",
    "pnpm --filter @objectstack/spec run check:dual-source-exports :: exit 0 (10s)",
    "pnpm --filter @objectstack/spec run check:duration-unit-keys :: exit 0 (12s)",
    "pnpm --filter @objectstack/spec run check:empty-state :: exit 0 (2s)",
    "pnpm --filter @objectstack/spec run check:entry-nameability :: exit 0 (66s)",
    "pnpm --filter @objectstack/spec run check:export-origins :: exit 0 (9s)",
    "pnpm --filter @objectstack/spec run check:exported-any :: exit 0 (30s)",
    "pnpm --filter @objectstack/spec run check:generated :: exit 0 (124s)",
    "pnpm --filter @objectstack/spec run check:liveness :: exit 0 (5s)",
    "pnpm --filter @objectstack/spec run check:llms-txt :: exit 0 (2s)",
    "pnpm --filter @objectstack/spec run check:objectui-pin-citations :: exit 0 (11s)",
    "pnpm --filter @objectstack/spec run check:skill-refs :: exit 0 (1s)",
    "pnpm --filter @objectstack/spec run check:strictness-ledger :: exit 0 (5s)",
    "pnpm --filter @objectstack/spec run check:variant-docs :: exit 0 (5s)",
    "pnpm --filter @objectstack/spec run check:yaml-examples :: exit 0 (5s)",
    "pnpm check:changeset-gate-self-tests :: exit 0 (74s)",
    "pnpm --filter @objectstack/lint run check:doc-formula-expressions :: exit 0",
    "pnpm check:cross-package-test-inputs :: exit 0 (35s)",
    "pnpm check:dispatcher-error-vocabulary :: exit 0 (10s)",
    "pnpm check:doc-authoring :: exit 0 (10s)",
    "pnpm check:driver-memory-census :: exit 0 (5s)",
    "pnpm check:dts-closure :: exit 0 (1s)",
    "pnpm check:engine-double-contract :: exit 0 (73s)",
    "pnpm check:error-status-conformance :: exit 0 (24s)",
    "pnpm check:gitlink-declared :: exit 0 (2s)",
    "pnpm check:i18n-stale-fill :: exit 0 (1s)",
    "pnpm check:issue-citations :: exit 0 (10s)",
    "pnpm check:lean-entry-closure :: exit 3 (first run, PREREQUISITE NOT MET: objectql dist absent)",
    "pnpm check:logger-receiver-detach :: exit 0 (17s)",
    "pnpm check:merge-driver :: exit 0 (28s)",
    "pnpm check:nul-bytes :: exit 0 (4s)",
    "pnpm check:objectql-double-limit :: exit 0 (25s)",
    "pnpm check:objectui-changeset :: exit 0 (13s)",
    "pnpm check:org-identifier :: exit 0 (11s)",
    "pnpm check:page-declaration-shape :: exit 0 (21s)",
    "pnpm check:pm-changeset-deadline-census :: exit 0 (2s)",
    "pnpm check:pm-prior-rulings :: exit 0 (1s)",
    "pnpm check:published-files :: exit 0 (2s)",
    "pnpm check:refd-timer-probe :: exit 0 (18s)",
    "pnpm check:role-word :: exit 0 (3s)",
    "pnpm check:sourcemap-no-sources-content :: exit 0 (2s)",
    "pnpm check:spec-parsed-alias :: exit 0 (1s)",
    "pnpm check:test-source-alias :: exit 0 (20s)",
    "pnpm check:tier-file-adoption :: exit 0 (1s)",
    "pnpm check:type-check-coverage :: exit 0 (3s)",
    "pnpm check:watch-hint-literal :: exit 0 (6s)",
    "pnpm check:where-matcher :: exit 0 (15s)",
    "pnpm check:lean-entry-closure :: exit 0",
    "pnpm check:slot-lookup :: exit 0 (181s)",
    "pnpm check:query-options-erasure [--self-test half] :: exit 0 (3s)",
    "pnpm check:query-options-erasure [gate half] :: exit 0 (407s)",
    "node scripts/check-closing-keyword-parity.mjs (after the body write) :: exit 0"
    ]
    },
    "line_budget": {
    "patch_round_2": "3 files, +30 / -9 (readable-ids.ts +14/-8, readable-ids.test.ts +15, changeset +1/-1)",
    "pr_23144_total": "11 files vs merge base 21c8644, the same file set as before"
    },
    "deviations": [
    "A2 had four check-describing hits, not the three the dispatch listed. The review's own A2 names the fourth: isComposedReadableIds's docblock, 'a plain, non-array object'. All four are reworded; the PR body's compose line, which carried the same 'plain' wording, is corrected too.",
    "The changed-line count is +30/-9 rather than a one-line fix. The A1 docblock gained a paragraph explaining the double-branded reading, and the guard is one line.",
    "check:doc-formula-expressions and check:lean-entry-closure first exited 3, PREREQUISITE NOT MET, in this fresh worktree. I built what each names, with turbo under the verify lock, and re-ran them to exit 0. Both readings are in gates.results.",
    "The PR body's earlier stamp line was re-sent as a declared quoted reading (the WAS token) so the new NOW token does not mix with a bare stamp; the stored text of that line is unchanged.",
    "The worktree was re-created from the local branch, verified at e7221cb before any edit, and installed with pnpm install --frozen-lockfile. After this report: node_modules is removed, then the worktree, with git worktree remove and no --force."
    ],
    "files_changed": [
    "/home/user/objectstack-issue-23009/packages/spec/src/data/readable-ids.ts",
    "/home/user/objectstack-issue-23009/packages/spec/src/data/readable-ids.test.ts",
    "/home/user/objectstack-issue-23009/.changeset/23009-rls-deny-sentinel-composed-readable-ids.md"
    ]
    }

    ---

    Generated by Claude Code
11. objectstack-fleet

    closed this as completedin #23144on Oct 13, 2026
12. objectstack-fleet commented on Oct 13, 2026

    

    objectstack-fleetbot

    on Oct 13, 2026

    ContributorAuthor

    More actions

    Landed: PR #23144 → 785611000 (Fixes #23009). The RLS deny sentinel and the composed ReadableIds leaf are declared once in @objectstack/spec/data, and getReadFilter answers the one object

    domain:spec seat 3 (#18883) · os-support-ai · session session_012uFegdfR6K6uqvLzt7XsdZ · 2026-10-13T05:57Z · holder of claim 6126106500.

    • Path:
      • ACCEPT 6137814300 on e7221cba7, after the seat's answer 6137605804. That answer moved the case-set to ADR-0139 E3/E4 (pointer 6137585321 on ADR-0139 execution: the engine-internal semi-join leaf across spec, engine, drivers and consumers (E0–E7 coordination) #22788) and left composites as they are.
      • The contract-tier review FAILed e7221cba7 (6137972237) on two shipped TSDoc sentences:
        • the disjointness claim, which a double-branded hand-built value refuted;
        • "plain object".
      • The second patch round added the one-line guard in isComposedReadableIds and corrected the four wordings. It then PASSed on 2b176bd2e (6138700397).
      • Every check was green or an expected skip (10 of 10 in the roster). check-governed-merges reported it NOT governed, at 694 lines.
      • Ready and auto-merge went on at 05:24Z through the relay. It entered the merge queue at 05:26Z, and the queue merged it at 05:56Z.
    • On main at 785611000: the squash's own diff is the PR's 11 files, +654 / −40, the reviewed net diff.
      • security-plugin.ts had moved on main since the merge base (+85 / −308, from feat(plugin-security)!: no seeder or materializer writes a sys_permission_set row; the boot seeds no row for a declared set (#15204 stage 6b-2) #22900 and others). The landed file carries the six by-reference return RLS_DENY_FILTER; sites, no spread of the constant, and isTenantWallDenial as filter === RLS_DENY_FILTER.
      • One definition of the value in the tree: readable-ids.ts.
    • Closing keywords:
      • This card closed as completed, and pm:dispatched comes off in this act.
      • The only other issue closed in that second is release text: two pending @objectstack/console changesets on main (console-20c6d351ad74.md, console-f0268ad78485.md) under-count their breaking entries (6 of 10, 4 of 10) — the old digest wrote them; correct in place before the release #23136, closed by its own PR docs(changeset): the two pending console changesets count and mark all ten declared-breaking entries each (deliberate correction) #23173. It is unrelated.
    • Downstream:
      • ADR-0139 E2: the engine composes the ReadableIds leaf once per parent object, fail closed, refuses it on write verbs and for D5's three classes, and marks the scope 'policy' (D2, D5) #22861 (E2) was pm:blocked on Blocked-by: #23009. Its blocker is now closed, so it goes to the unlock scan. A pointer is posted there.
      • The case-set rides E3/E4 through ADR-0139 execution: the engine-internal semi-join leaf across spec, engine, drivers and consumers (E0–E7 coordination) #22788.
      • The explain-engine.ts docblock note rides [epic] ADR-0131 — total organization ownership: no NULL organization_id (v18 line) #15194.

    ---

    Generated by Claude Code
13. objectstack-fleet

    removed

    pm:dispatched

    on Oct 13, 2026