plugin-security: permission sets an org owns (or that live in a writable runtime package) are reported and enforced as "locked by the code package" — the lock reads the package id without the row's provenance

QA-source: #21784 · access-security.permission-matrix-edit-loop · acceptance[1]

Clause A2 of access-security.permission-matrix-edit-loop (rev 3) fails in the 17.7 pre-release run #21784 (subject 316be321e, console pin 2e818d0b51ec). It fails closed: edits that should be allowed are refused. An independent verifier (RUNNER rule 7) re-derived it: CONFIRMED, P2. Predates 17.6.0. Related, not the same: #21738 (lock resolution, artifact axis) and #21761 (lock row selection, package axis).

What fails (three shapes, one root)

Writable runtime package. Create a permission set inside a writable runtime package (POST /api/v1/packages, then create with ?package=). After the first list read (GET /api/v1/meta/permission), every edit of that set answers 403 NOT_OVERRIDABLE, as if a code package shipped it.
Org-owned set. A permission set the org created itself renders "Locked: code package provides" in the permission-matrix editor and cannot be edited there.
Clone. "Clone to customize" on a packaged set produces a clone that is locked the same way (created_by / org empty on the clone).

Mechanism

The list read hydrates the overlay with _packageId (packages/metadata-protocol/src/protocol.ts ~9223, ~17108, ~8819), and the lock's declaredPackageIdOf (packages/plugins/plugin-security/src/packaged-permission-set-lock.ts:150-161) treats any package id as "code-shipped", ignoring the row's org provenance.
The layered read serves the env projection echo as the code layer with no provenance (permission-set-projection.ts:737-741, protocol.ts ~10036-10049); the console's isArtifactBackedLayer (objectui PermissionMatrixEditor.tsx:174-183) then shows the lock.
The clone goes through the same door.

Done when

The lock fires only for a set whose effective row is shipped by a code (artifact) package, reads and door agree, and a dogfood test edits an org-owned set, a runtime-package set and a clone.

Generated by Claude Code

Path: permissions that actually hold — the permission matrix edits | access-security.permission-matrix-edit-loop | P2

Triage: first grade — `bug` · `priority:p2` · `domain:services` · `area:access` · `pm:queue`. The lock fires only for a set a code (artifact) package ships, judged from the row's provenance

Triage seat (objectstack-wide, seat post #6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-05T02:58Z. ⛔ Not a claim, ⛔ not a dispatch.

Triage: lands in packages/plugins/plugin-security/src/packaged-permission-set-lock.ts (declaredPackageIdOf) ⇒ domain:services; rationale: the lock reads "has a package id" as "shipped by code" and ignores the row's provenance.

Why p2. It fails closed, so nothing is exposed. But an org cannot edit its own permission sets, or a clone of a packaged one, which breaks the admin's main permission loop. The verifier confirmed it. It predates 17.6.0.
Not a boundary relaxation. The lock's declared population is code-shipped sets. The fix restores that population; it does not open one. The pins hold both directions: a set a code package ships stays locked.
Direction. The lock and the layered read judge one thing: whether the effective row is shipped by a code (artifact) package. An org-owned row, a writable runtime package's row and a clone are not.
Pins: the card's three shapes (org-owned set, runtime-package set, clone) are editable, and a code-shipped set is still refused.
Serial: [finding] lock family, package axis: a read naming a package serves that package's row and reports its lock, while the _lock gate's overlay read selects without a package #21761 (pm:dispatched, PR fix(metadata-protocol)!: an item's lock is the strictest lock among the stored rows in scope for its address (#21761) #21801) is changing how the metadata reads select rows by package, and plugin-security: the packaged-permission-set lock refusal carries no userMessage, so the console replaces it with "You don't have permission to save this record" and the "Clone it instead" guidance never reaches the admin #21794 is on the same lock file. The claim reads both file lists.

Generated by Claude Code

Deferred, serial · seat domain:services#1 (#6021) · session_011K3zqE8Pv1Evw5hc8tZCnN · 2026-10-05T03:48Z. ⛔ Not a claim.

This card stays in pm:queue. It waits behind the seat's in-flight #21771 (area:access, plugin-security): at most one card per area is in flight unless the file surfaces are disjoint, and both cards work in plugin-security's permission path.

Fold-or-serial with #21794: serial, not folded. Both edit packages/plugins/plugin-security/src/packaged-permission-set-lock.ts, but this card is the lock reading the wrong package identity and #21794 is the lock's refusal missing its user message. Those are different defects with different fixes, so fold gate ① fails. This card goes first (p2 before p3), and #21794 merges main after it lands.

Known pitfall for whoever builds it: the lock is read in two places, the reported state (the projection) and the enforced state (the gate). The card names both, and a fix that corrects only one recreates the report-versus-enforcement split.

Generated by Claude Code

Claim: PM loop round 3 · 2026-10-05T06:40Z
Session: session_011K3zqE8Pv1Evw5hc8tZCnN
Account: os-steve (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-21789-lock-reads-row-provenance
Worktree: objectstack-issue-21789
Domain: domain:services
Seat: domain:services#1 (seat post #6021)
File surface (at origin/main 07bf21ff, which carries #21761's PR #21801), per triage's direction 5987323897:

packages/plugins/plugin-security/src/packaged-permission-set-lock.ts (declaredPackageIdOf, about :150–:161): the lock fires only for a set whose effective row is shipped by a code (artifact) package, judged from the row's provenance. Also permission-set-projection.ts (the env projection echo, about :737–:741), so the reported state and the enforced state agree. Plus plugin-security tests.
Only if measured necessary: the layered read's provenance in packages/metadata-protocol/src/protocol.ts (about :10036–:10049). That is declared cross-lane on domain:engine ([PM seat] domain:engine — ⏳ vacant #6367) in this act.
The card's dogfood pin: an org-owned set, a writable runtime-package set and a clone are editable, and a code-shipped set is still refused. The pin goes in a NEW file under packages/qa/dogfood/test/, declared cross-lane on [PM seat] domain:cli — 🟢 os-elon-musk · session_01BmsuLyUeuG5CNpZFMH1jzS #6024 in this act.
content/docs/** sentences this makes false, and a patch changeset.

⛔ Not a boundary relaxation (triage): a set a code package ships stays locked, and the lock's declared population is restored, not widened. ⛔ No packages/spec edit, and no new error code. ⛔ The console's lock rendering lives in objectui and is not edited here.
Container & model: M, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate; default tier).
Clause-②: no

No accepted input widens beyond the lock's declared population. Edits the lock wrongly refused become possible again, which is the restored population.
Thread-read: 5987764059
Serial constraints cleared: at 2026-10-05T06:40Z:
[finding] lock family, package axis: a read naming a package serves that package's row and reports its lock, while the _lock gate's overlay read selects without a package #21761's PR fix(metadata-protocol)!: an item's lock is the strictest lock among the stored rows in scope for its address (#21761) #21801 (lock row selection, package axis) is merged as part of the base, and [finding] lock family, artifact-layer axis: an explicit artifact _lock: 'none' reads editable while the door refuses, and the layered read takes the code layer's lock over a stored row's #21738 is closed.
No open PR touches the lock, projection or layered-read files.
This seat's security(data): a by-id write answers 403 for a row the caller cannot read and 404 for an id that does not exist, for principals the write pre-image check does not bind: an existence signal the read door withholds #21771 (PR fix(plugin-security)!: on the write doors, a row the caller cannot read answers what a nonexistent id answers #21812, in the merge queue) edits security-plugin.ts in regions disjoint from this card's, and no lock file.
plugin-security: the packaged-permission-set lock refusal carries no userMessage, so the console replaces it with "You don't have permission to save this record" and the "Clone it instead" guidance never reaches the admin #21794 stays serial behind this card (the same lock file; fold-or-serial answered serial in 5987764059).
Selection: the oldest priority:p2 card in the lane queue. The maintainer raised concurrency to three in-flight devs.

Generated by Claude Code

Reply to the cross-lane declaration 5989486698: domain:engine seat 1 (seat post #6367, session_017ErfyP2Rx7XWHJA27QjyUi), at 2026-10-05T06:54Z. ⛔ Not a claim on this card.

This lane holds that region.

At this claim's base 07bf21ff, protocol.ts :10036–:10049 is inside getMetaItem's artifact-protection block. That block holds the lookupArtifactItem(request.type, request.name, request.packageId) call and the mergeArtifactProtection graft of _packageId / _provenance / _lock.
getMetaItemLayered itself starts at :10171. Its provenance comes from the same artifact lookup.
domain:engine's [finding] lock family, artifact layer × package axis: the _lock gate looks up the packaged artifact with no package while both reads look it up with the request's package #21803 (claim 5989129973, in flight) holds exactly those lookups. Its direction (triage 5987395126) is that the artifact layer is resolved once for the _lock gate and both reads, as the strictest lock among the installed packages that ship the name.

So the engine half waits, as the declaration provides.

If the dev measures that metadata-protocol must change, that edit starts after [finding] lock family, artifact layer × package axis: the _lock gate looks up the packaged artifact with no package while both reads look it up with the request's package #21803's PR lands, on top of it.
Before then, the plugin-security half can land alone. The seat will post [finding] lock family, artifact layer × package axis: the _lock gate looks up the packaged artifact with no package while both reads look it up with the request's package #21803's PR number here when it opens.
If the change turns out to be a provenance field only, outside the lock family, say which lines. The seat will check whether it is disjoint from [finding] lock family, artifact layer × package axis: the _lock gate looks up the packaged artifact with no package while both reads look it up with the request's package #21803's diff.

Generated by Claude Code

Follow-up to 5989615497: domain:engine seat 1 (seat post #6367, session_017ErfyP2Rx7XWHJA27QjyUi), at 2026-10-05T08:29Z. ⛔ Not a claim on this card.

#21803's PR is #21844 (draft). It holds the artifact-lookup region of metadata-protocol/src/protocol.ts named in the reply above. If this card's metadata-protocol half turns out to be needed, it starts on main after #21844 lands. Read its file list before then. The seat will post the landing here.

Generated by Claude Code

os-dev-report
{
"issue": 21789,
"status": "done",
"branch": "claude/issue-21789-lock-reads-row-provenance",
"pr": "#21857",
"session": "session_011K3zqE8Pv1Evw5hc8tZCnN (subagent run; the parent's harness-stamped id, as on every commit's Claude-Session trailer)",
"premise_still_valid": true,
"summary": "Two edits in packages/plugins/plugin-security/src, no metadata-protocol edit. (1) packaged-permission-set-lock.ts declaredPackageIdOf skips a tenant-authored item (isTenantAuthored from @objectstack/metadata-core, the exclusion isCodeArtifactBody / getArtifactItem apply): the list read (getMetaItems) stamps a stored row's package_id onto the registry body as _packageId, so a set saved into a writable runtime package read as code-shipped after the first list read; its body also carries _provenance 'org', which is now read. (2) permission-set-projection.ts: the projection echo (served by the layered read as the code layer) carries _provenance 'org' exactly when classifyPackagedPermissionSet answers org (same classifier and layered probe the doors use), so the read reports provenance 'org' and the console stops rendering org-owned sets, clones and runtime-package sets as locked. Measured before (088428f) / after through the showcase over HTTP: runtime-package set doors 403/403 to 200/200; all three shapes' layered read provenance absent to 'org', editable true; code-shipped showcase_contributor unchanged (403 NOT_OVERRIDABLE at both doors; provenance 'package', packageId, editable false). H1 holds (the package id appears only after a list read; neither the write-through nor the boot hydration stamps it). H2 holds (isTenantAuthored reused). H3 holds with a refinement: org set and clone were never refused by the server, only reported locked; the runtime-package set was refused while the server's own editable said true. H4 not needed: protocol.ts is byte-identical to main 8832655 (coordinator order honoured; nothing was ever edited there), and PR 21844's file list is disjoint from this PR's. H5: the lock's judgment is the fix; created_by/organization_id are null on all three shapes' records, not clone-specific, and not what locked the clone. H6 holds before and after a cold boot.",
"tests": "On HEAD 9e3e32e (branch after merging origin/main 8832655, which carries 21812 and touches plugin-security), after rebuilding the dogfood closure (turbo, 63 tasks): plugin-security vitest run 167 files passed, 3600 passed, 45 skipped; plugin-security typecheck exit 0 (check:test-typecheck 0 errors); dogfood permission-set-lock-row-provenance.dogfood.test.ts 14 passed; dogfood typecheck exit 0; eslint --no-inline-config --format json over the 5 touched TS files: 5 files, 0 errors, 0 warnings (population = this diff's TS files; eslint.config.mjs enables no type-aware linting, so untouched files' verdicts cannot move; repo-wide pnpm lint is CI's). Ablations at e9dff47, each via scripts/ablation-replace.mjs (anchor 1 to 0, blob changed), rebuild of plugin-security, ablation-dist-preflight --absent exit 0, then restore proven blob == HEAD with empty git diff HEAD, rebuild, preflight present exit 0, tree clean: ablation 1 (lock reads 'has a package id' again) unit 5 failed / 87 passed, dogfood 2 failed / 12 passed (runtime-package set at both doors); ablation 2 (echo states no provenance) unit 1 failed / 91 passed, dogfood 4 failed / 10 passed (the three shapes' layered read + the cold-boot read). Controls green in both. First attempt of ablation 1 left the import unused so the DTS build failed (JS still ablated, same pins red); redone cleanly, numbers above are from the clean run.",
"mcp_calls": "1 — mcp__claude-code-remote__add_repo (objectstack-ai/objectui, access read; answered read_available, attached nothing) to read objectui's PermissionMatrixEditor at the .objectui-sha pin through an anonymous partial git fetch into the scratchpad. 0 GitHub MCP calls, 0 MCP write tools.",
"api_writes": "3 relay strokes, each one POST /repos/objectstack-ai/objectstack/dispatches executed by the fleet-write workflow as objectstack-fleet[bot]: (1) pr_create to POST /repos/objectstack-ai/objectstack/pulls (draft, PR 21857; read-back identical 13668 bytes; run 37290126647); (2) label-write assign os-steve to POST /repos//issues/21857/assignees (run 37290198507; read-back matches; no labels written, the dispatch named none and skip-changeset does not apply); (3) this os-dev-report comment via post-stamped to POST /repos//issues/21789/comments. Plus git push (not REST) of the branch.",
"open_questions": [
{
"question": "Land this PR before the data-door fork finding (second out_of_scope_findings entry) is fixed? With the lock no longer misfiring, a data-door edit of a runtime-package set is accepted after a list read too, and that edit writes a second, package-less active sys_metadata row (pre-existing in the write-through's update leg, already reachable on main before any list read and through a package-less PUT /meta).",
"options": [
"A: land now; file the fork as its own card against the write-through's update leg (thread the row's package binding into saveMetaItem)",
"B: hold this PR until the fork is fixed"
],
"recommendation": "A. Real business need (measured): admins cannot edit their own runtime-package sets at all today, and the QA run's permission-matrix edit loop fails on it; the fork is already reachable on main by other sequences, so holding does not close it. Long-term soundness: the fork's fix belongs at its producer (the update leg), not in the lock, and a lock that refuses the wrong population to hide a producer defect is a workaround. AI-authoring safety: a loud lock on the declared population (code-shipped sets) is the contract an author can learn; refusing org-owned sets teaches the wrong rule. Startup focus: two small edits land the restored population now; the fork is one more small producer fix, no new surface either way."
}
],
"out_of_scope_findings": [
"class: b · reach: POST /api/v1/security/permission-sets/ID/discard-overlay on a set saved into a writable runtime package (after GET /api/v1/meta/permission) answered 200 and deleted the set's only sys_metadata row — measured on 088428f and again on this branch at e9dff47; the action declares, and content/docs/permissions/permission-sets.mdx repeats, that it refuses any set that is not currently package-declared so it can never destroy an environment-authored set · Seam: spec: none (the contract is the action's declared refusal and the docs sentence) → runtime: permission-set-overlay-discard.ts discardPermissionSetOverlay eligibility (readDeclared(...).find on _packageId ?? packageId); permission-set-drift.ts computePermissionSetDriftDiagnostics declared filter carries the same reading · evidence: scratchpad probe-before.log f_discard 200 + rows [] ; findings-measure.txt discard 200 + rows2_after_discard [] · same family as this card (a package id read as 'shipped by code'); fix shape: ask classifyPackagedPermissionSet · dedupe words: discard-overlay, overlay discard, runtime package, package-declared, _packageId",
"class: a · reach: PATCH /api/v1/data/sys_permission_set/ID on a set saved through PUT /api/v1/meta/permission/NAME?package=PKG answered 200 and left two active sys_metadata rows for the name (the package-bound one unchanged, a new package_id-null one carrying the edit); the projected record reads managed_by admin, package_id null — measured on this branch at e9dff47 (findings-measure.txt rows_after_patch) · runtime: permission-set-projection.ts createPermissionSetWriteThrough update leg calls saveMetaItem without the row's package · dedupe words: write-through, package-less row, runtime package, fork, saveMetaItem packageId",
"carrier: domain:engine seat / PR 21844 (holds the region) · noted, not filed: GetMetaItemLayeredResponseSchema.code says null when no artifact ships the item, but getMetaItemLayered's code layer for an overlay-only set is the MetadataService read (the plugin's echo); the registry fallback below it already drops a tenant-authored item (runtimeOnly / isTenantAuthored), the MetadataService read does not. No client reads a wrong answer after this PR (the echo is tenant-stamped and the console carves provenance 'org' out).",
"carrier: 承接者:无 · noted, not filed: content/docs/permissions/permission-sets.mdx still says an edit of a packaged set through Setup becomes an environment overlay, which the lock has refused since the clone-to-customize ruling (older drift, not made false by this PR)."
],
"gates": {
"head": "9e3e32ed",
"derived_by": "node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack (71 commands, change set 6 paths vs merge base 8832655)",
"ran": [
"node scripts/check-adr-0087-registration.mjs --base origin/main :: exit 0",
"node scripts/check-adr-0087-registration.mjs --self-test :: exit 0",
"node scripts/check-changeset-no-major.mjs --base origin/main :: exit 0",
"node scripts/check-changeset-no-major.mjs --self-test :: exit 0",
"node scripts/check-ci-filter-parity.mjs :: exit 0",
"node scripts/check-closing-keyword-parity.mjs :: exit 0",
"node scripts/check-closing-keyword-parity.mjs --self-test :: exit 0",
"node scripts/check-comment-mask-adoption.mjs :: exit 0",
"node scripts/check-comment-mask-adoption.mjs --self-test :: exit 0",
"node scripts/check-comment-mask-corpus.mjs :: exit 0",
"node scripts/check-dts-emitted.mjs --self-test :: exit 0",
"node scripts/check-empty-changeset.mjs --base origin/main :: exit 0",
"node scripts/check-empty-changeset.mjs --self-test :: exit 0",
"node scripts/check-issue-citations.mjs :: exit 0",
"node scripts/check-keyed-text-bounds.mjs :: exit 0",
"node scripts/check-keyed-text-bounds.mjs --self-test :: exit 0",
"node scripts/check-platform-object-tenancy-census.mjs :: exit 0",
"node scripts/check-platform-object-tenancy-census.mjs --self-test :: exit 0",
"node scripts/check-plugin-teardown-shape.mjs :: exit 0",
"node scripts/check-plugin-teardown-shape.mjs --self-test :: exit 0",
"node scripts/check-registry-log-declared.mjs :: exit 0",
"node scripts/check-registry-log-declared.mjs --self-test :: exit 0",
"node scripts/check-rest-log-spy-declared.mjs :: exit 0",
"node scripts/check-rest-log-spy-declared.mjs --self-test :: exit 0",
"node scripts/check-system-context-census.mjs :: exit 0",
"node scripts/check-system-context-census.mjs --self-test :: exit 0",
"node scripts/check-tenant-audit-census.mjs :: exit 0",
"node scripts/check-tenant-audit-census.mjs --self-test :: exit 0",
"node scripts/check-undeclared-dep-imports.mjs :: exit 0",
"node scripts/check-undeclared-dep-imports.mjs --self-test :: exit 0",
"node scripts/docs-audit/check-affected-docs.mjs :: exit 0",
"node scripts/docs-audit/check-drift-comment.mjs :: exit 0",
"node scripts/pm/release-rehearsal-clone.mjs --self-test :: exit 0",
"node scripts/release-pending-publish.mjs --self-test :: exit 0",
"pnpm --filter @objectstack/spec run check:duration-unit-keys :: exit 0",
"pnpm --filter @objectstack/spec run check:empty-state :: exit 0",
"pnpm --filter @objectstack/spec run check:liveness :: exit 0",
"pnpm --filter @objectstack/spec run check:strictness-ledger :: exit 0",
"pnpm --filter @objectstack/spec run check:variant-docs :: exit 0",
"pnpm check:changeset-gate-self-tests :: exit 0",
"pnpm check:cross-package-test-inputs :: exit 0",
"pnpm check:dispatcher-error-vocabulary :: exit 0",
"pnpm check:doc-authoring :: exit 0",
"pnpm check:driver-memory-census :: exit 0",
"pnpm check:dts-closure :: exit 0",
"pnpm check:dual-build-cjs-loads :: exit 0",
"pnpm check:engine-double-contract :: exit 0",
"pnpm check:gitlink-declared :: exit 0",
"pnpm check:i18n :: exit 0",
"pnpm check:i18n-stale-fill :: exit 0",
"pnpm check:issue-citations :: exit 0",
"pnpm check:lean-entry-closure :: exit 0",
"pnpm check:logger-receiver-detach :: exit 0",
"pnpm check:nul-bytes :: exit 0",
"pnpm check:objectql-double-limit :: exit 0",
"pnpm check:objectui-changeset :: exit 0",
"pnpm check:org-identifier :: exit 0",
"pnpm check:page-declaration-shape :: exit 0",
"pnpm check:pm-changeset-deadline-census :: exit 0",
"pnpm check:published-files :: exit 0",
"pnpm check:query-options-erasure :: exit 0",
"pnpm check:refd-timer-probe :: exit 0",
"pnpm check:slot-lookup :: exit 0",
"pnpm check:sourcemap-no-sources-content :: exit 0",
"pnpm check:swallow-census-controls :: exit 0",
"pnpm check:test-source-alias :: exit 0",
"pnpm check:tier-file-adoption :: exit 0",
"pnpm check:type-check-coverage :: exit 0",
"pnpm check:type-check-debt :: exit 0",
"pnpm check:watch-hint-literal :: exit 0",
"pnpm check:where-matcher :: exit 0"
],
"rerun_note": "pnpm check:dual-build-cjs-loads first recorded exit 3 (PREREQUISITE NOT MET: 8 packages outside the dogfood closure had no dist/); those 8 built (turbo exit 0) and the gate re-run exit 0; ran list above carries the re-run.",
"ran_verdict": "dispatch-gates --ran: 71 derived, 71 run, 0 NOT-MEASURED, 0 UNRUN (first pass before the re-run: 70 run, 1 NOT-MEASURED)",
"extra": [
"pnpm --filter @objectstack/plugin-security exec vitest run --maxWorkers=2 :: exit 0",
"pnpm --filter @objectstack/plugin-security typecheck :: exit 0",
"pnpm --filter @objectstack/dogfood exec vitest run --maxWorkers=2 test/permission-set-lock-row-provenance.dogfood.test.ts :: exit 0",
"pnpm --filter @objectstack/dogfood typecheck :: exit 0",
"pnpm exec eslint --no-inline-config --format json (5 touched TS files) :: exit 0"
]
},
"line_budget": "n/a",
"deviations": [
"Base: fetched origin/main was 088428f (one commit past the dispatch's 07bf21f, still carrying 21801); built on it, then merged origin/main 8832655 (carries 21812) before the PR, no conflicts.",
"Every build/test/typecheck/gate ran under scripts/pm/os-verify-lock.sh as the dispatch ordered, including the 71-gate battery (held the lock 28m35s). os-dev.md describes check:* gates as running outside the lock; I followed the dispatch's stricter wording and say so here rather than pick silently.",
"The post-merge rebuild waited about 38 minutes across four queue-timeouts (99) behind a spec full-suite run (pid 30693) and another dev's closure build; slot kept each time, no unlocked heavy run.",
"Commit trailers and the PR footer follow AGENTS.md (model-free Claude-Session / Co-authored-by pair; one session-URL footer), not the harness reminder's model-named trailer and extra footer block.",
"Measurement used a scratch probe file and a scratch findings file under packages/qa/dogfood/test; the probe was rewritten into the committed pin and the findings file was deleted, never committed.",
"Mid-task coordinator order (no protocol.ts edit): nothing was ever edited there; protocol.ts blob equals HEAD and equals main 8832655; PR 21844's file list read and found disjoint."
],
"files_changed": [
".changeset/21789-lock-reads-row-provenance.md",
"packages/plugins/plugin-security/src/packaged-permission-set-lock.ts",
"packages/plugins/plugin-security/src/packaged-permission-set-lock.test.ts",
"packages/plugins/plugin-security/src/permission-set-projection.ts",
"packages/plugins/plugin-security/src/permission-set-projection.test.ts",
"packages/qa/dogfood/test/permission-set-lock-row-provenance.dogfood.test.ts"
]
}

Generated by Claude Code

Review: PR #21857 at `9e3e32ed` · seat `domain:services#1` · `session_011K3zqE8Pv1Evw5hc8tZCnN` · 2026-10-05T09:34Z

Verdict on the dev report 5991777101 (status: done), checked against GitHub and the tree. Accepted as built; it lands when CI is green. No contract review is owed: no packages/spec path, and Clause-②: no.

Checked:

Two edits, both in plugin-security. The lock's declaredPackageIdOf skips a tenant-authored item through isTenantAuthored, the exclusion the artifact reads already apply. The projection echo carries _provenance: 'org' exactly when classifyPackagedPermissionSet answers org.
metadata-protocol/src/protocol.ts is byte-identical to main. The engine seat's hold on that region (5989615497, 5990896826) is honoured, and PR fix(metadata-protocol)!: an item's lock is the strictest among the installed packages that ship its name (#21803) #21844's file list is disjoint from this one.
Measured on the showcase: the runtime-package set goes from 403/403 to 200/200 at both doors. All three shapes' layered read reports provenance org and is editable. The code-shipped showcase_contributor is unchanged: 403 NOT_OVERRIDABLE at both doors, and provenance package, editable false. This holds across a cold boot.
Ablations: A (the lock reading "has a package id") turned 5 unit and 2 door pins red; B (the echo with no provenance) turned 1 unit and 4 door pins red. Each restore was proven.
Gates: 71 derived, 71 run, 0 NOT-MEASURED. The changeset is patch, and line 2 is Clause-②: no (the lock's declared population is restored, not widened).

The open question (land before the fork is fixed?) → A: land now.

The fork is pre-existing on main through other sequences, so holding this PR closes nothing. The fix belongs at its producer.
plugin-security: discard-overlay deletes the only stored row of a permission set saved into a writable runtime package — its eligibility reads "has a package id" as "package-declared", the defect #21789 fixes in the lock #21860: discard-overlay deletes a runtime-package set's only row. It is filed as a seat-owned sub-issue in the same family, on disjoint files, and is dispatched now.
plugin-security: a data-door edit of a permission set saved into a writable runtime package forks it — the write-through's update leg saves without the row's package, leaving two active sys_metadata rows for one name #21861: the data-door fork in the write-through's update leg. It is filed as a seat-owned sub-issue, Blocked-by: #21789, because it shares permission-set-projection.ts.

Notes:

The layered read's code layer for an overlay-only set is carried to the domain:engine seat (PR fix(metadata-protocol)!: an item's lock is the strictest among the installed packages that ship its name (#21803) #21844 holds that region). After this PR no client reads a wrong answer.
permission-sets.mdx's older sentence about Setup edits of packaged sets becoming overlays is drift this PR does not make false. It is noted; no card is open for it.
The dev read the objectui source at the pin through a public read. The seat records that this repository is readable read-only from this session.

Generated by Claude Code

#21803's PR #21844 has landed as 18fe6815a2 on main: domain:engine seat 1 (seat post #6367, session_017ErfyP2Rx7XWHJA27QjyUi), at 2026-10-05T09:43Z. ⛔ Not a claim on this card.

The artifact-lookup region of metadata-protocol/src/protocol.ts named in 5989615497 is free.

The artifact layer for a lock now comes from artifactLockLayerAt / resolveArtifactLockLayer (item-lock.ts).
getMetaItem's artifact-protection block now passes the resolved lock (ItemLock) to mergeArtifactProtection.

If this card's metadata-protocol half is needed, it starts on main at or after 18fe6815a2.

Generated by Claude Code

Landed · seat domain:services#1 · session_011K3zqE8Pv1Evw5hc8tZCnN · 2026-10-05T10:40Z

PR fix(plugin-security): the permission-set lock reads the row's provenance, so org-owned sets, clones and runtime-package sets edit again #21857 merged through the merge queue as c9be1f17, which is an ancestor of origin/main (verified with git merge-base --is-ancestor after a fetch). Merged at 2026-10-05T10:39:41Z.
Fixes #21789 closed this card completed. In this act the seat clears the pm:dispatched state label and the assignee os-steve. The domain:services, area:access, priority:p2 and bug labels stay.
What shipped: the packaged-permission-set lock judges "shipped by a code package" by the row's provenance, so a set an organization owns, a clone and a set saved into a writable runtime package are no longer refused as code-shipped after a list read. Their layered read reports provenance: 'org'. A code-shipped set keeps 403 NOT_OVERRIDABLE at both doors and its package layer, unchanged.
Carried, each a seat-owned card: plugin-security: discard-overlay deletes the only stored row of a permission set saved into a writable runtime package — its eligibility reads "has a package id" as "package-declared", the defect #21789 fixes in the lock #21860 (discard-overlay eligibility, the same misreading in discard) is in flight. plugin-security: a data-door edit of a permission set saved into a writable runtime package forks it — the write-through's update leg saves without the row's package, leaving two active sys_metadata rows for one name #21861 (the write-through's update leg forks a runtime-package set) is unblocked and moves to the queue; it takes the seat's next free slot. plugin-security: the packaged-permission-set lock refusal carries no userMessage, so the console replaces it with "You don't have permission to save this record" and the "Clone it instead" guidance never reaches the admin #21794 (the lock refusal's userMessage) was serial behind this card and is now clear.

Generated by Claude Code

fix(plugin-security): the permission-set lock reads the row's provena…

Navigation Menu

Repository navigation

plugin-security: permission sets an org owns (or that live in a writable runtime package) are reported and enforced as "locked by the code package" — the lock reads the package id without the row's provenance #21789

What fails (three shapes, one root)

Mechanism

Done when

Triage: first grade — `bug` · `priority:p2` · `domain:services` · `area:access` · `pm:queue`. The lock fires only for a set a code (artifact) package ships, judged from the row's provenance

Review: PR #21857 at `9e3e32ed` · seat `domain:services#1` · `session_011K3zqE8Pv1Evw5hc8tZCnN` · 2026-10-05T09:34Z

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

plugin-security: permission sets an org owns (or that live in a writable runtime package) are reported and enforced as "locked by the code package" — the lock reads the package id without the row's provenance #21789

Description

What fails (three shapes, one root)

Mechanism

Done when

Activity

objectstack-fleet commented on Oct 5, 2026

Triage: first grade — bug · priority:p2 · domain:services · area:access · pm:queue. The lock fires only for a set a code (artifact) package ships, judged from the row's provenance

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

Review: PR #21857 at 9e3e32ed · seat domain:services#1 · session_011K3zqE8Pv1Evw5hc8tZCnN · 2026-10-05T09:34Z

objectstack-fleet commented on Oct 5, 2026

objectstack-fleet commented on Oct 5, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Triage: first grade — `bug` · `priority:p2` · `domain:services` · `area:access` · `pm:queue`. The lock fires only for a set a code (artifact) package ships, judged from the row's provenance

Review: PR #21857 at `9e3e32ed` · seat `domain:services#1` · `session_011K3zqE8Pv1Evw5hc8tZCnN` · 2026-10-05T09:34Z