types: settle the unread designer members on both faces — the 4 node members in UnmirroredDeclared and the 32 record members no renderer reads (split from objectui#10859)

Filing gate: ② enforce-or-remove (ADR-0049) on published authoring members that no renderer reads. Split by triage from #10859, answering that card's pm:retriage (5944582356). That card keeps its REFUSED_AT_TYPE ratchet, the (ii) keys and the (iii) keys. This card carries only the settlement. ⛔ Not a claim.

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01AavokzJ5DndAwitDXvKy4U.

The members (the enumeration is the seat's, cited, not restated)

4 node members recorded in UnmirroredDeclared: DataModelDesignerSchema.autoLayout, ProcessDesignerSchema.version, ProcessDesignerSchema.lanes and ReportDesignerSchema.previewMode. Source: objectui#10859 release 5944563530, item 1.
32 record-type members found unread by a source read and mirrored as declared. They are listed in the ACCEPT 5944406564 on objectui#10859. 3 of them are TS-required: DataModelRelationship.sourceField / targetField and ObjectDefinitionRelationship.relatedObject.

Direction (triage's grade)

Shape: objectui#10874's: one change across both faces (zod and TS), one BREAKING (authoring) banner with a migration line, and a runtime probe per member (the seat note 5861341151 shape).
Per member, the maintainer's criterion: "does the mainstream have it?"
- Yes ⇒ give it a reader.
- No ⇒ retire it on both faces.
- The probe decides which arm applies: no reader at runtime and no producer.
The 3 TS-required members are a real fork. Retiring one breaks a TS consumer that constructs the type. Giving it a reader means a renderer starts using it. The claim measures producers (in-repo and the published examples) first. It records each verdict, with its reading, on this card before the PR.
⛔ No member stays declared-and-unread after this card. ⛔ The ratchet on objectui#10859 does not move here.

Acceptance

Each of the 36 members ends either read (pinned by a render probe) or retired (refused by name on both faces, with the migration line).
zod-mirror-parity stays green.

Generated by Claude Code

Claim: PM loop round 1 (measured first, then built)
Session: session_01YLg8XqWGJ785fwQ5v4pH37
Account: os-elon-musk (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-11434-designer-members
Worktree: objectui-issue-11434
Domain: domain:ui
Seat: domain:ui#1
File surface (triage's direction in this card's body: objectui#10874's shape, the maintainer's criterion per member, the 3 TS-required members a fork measured first):

Phase 1, measurement only, no commit. For each of the 36 members, the dev measures its readers at runtime, its producers (in-repo, examples/**, objectstack), its teaching sites, and a cited mainstream reading. These are the 4 node members in UnmirroredDeclared (objectui#10859 release 5944563530) and the 32 record members (ACCEPT 5944406564 on objectui#10859). The dev recommends read or retire, then returns needs_decision. The seat rules member by member in one comment on this card. The 3 TS-required members (DataModelRelationship.sourceField / targetField, ObjectDefinitionRelationship.relatedObject) get their producers measured before any verdict.
Phase 2, after that ruling. One change across both faces: packages/types/src/zod/designer.zod.ts, the TS face under packages/types/src/, the UnmirroredDeclared / mirror-parity ledger in packages/types/src/__tests__/, and for each member ruled "read", its renderer in packages/plugin-designer/src/** with a render probe. One @object-ui/types changeset with a BREAKING (authoring) banner and a migration line; one per other released package touched.

The dev stops on a breach and explains it in the report.
Container & model: M, mode:subagent, model: opus (default judgment tier; dispatch-gates --tier --repo objectstack-ai/objectui prints "no path-derived mandate" for these paths)
Clause-②: yes
Why yes: phase 2 refuses retired members by name on both faces, or makes a renderer read a member it ignores today. The at-tier contract review is owed before ACCEPT. Phase 1 changes nothing.
Thread-read: none
Serial constraints cleared: read 2026-10-02T04:59Z at objectui origin/main 31987bd50.

Parent objectui#10859 batch 8 (this seat): phase 2a touches the fields loop, plugin-tree / plugin-view, packages/cli's ratchet and the doc-type gate. Phase 2b touches registrations, including packages/plugin-designer/src/index.tsx for the designer four. This card's phase 2 surface is packages/types designer declarations and plugin-designer component files. That is disjoint at file level, except possibly plugin-designer/src/index.tsx, which phase 2 re-reads before editing; whichever lands second merges main. The ratchet on objectui#10859 does not move here (triage).
In-flight claims of this seat: objectui#11373 (comment sweep: plugin-dashboard, plugin-report, data-display.zod.ts, plugin-charts.mdx). data-display.zod.ts is a different file from designer.zod.ts. Landing: objectui#11442, objectui#11443. All disjoint.

Scope: the PR says Fixes #11434.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "phase": "1 of 2 (measurement only)",
 "status": "needs_decision",
 "branch": "claude/issue-11434-designer-members",
 "pr": null,
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id; the newest Claim on objectui#11434, 5945883001, names this branch, checked before any work)",
 "premise_still_valid": true,
 "summary": "Re-derived at objectui origin/main 31987bd50 (worktree base, unchanged). The 4 node members hold exactly as ledgered (UnmirroredDeclared 3 entries / 4 keys). The record population: every member the ACCEPT's open question names is still unread by a non-test grep AND by a runtime probe through the real SchemaRenderer + registry (unset vs distinctive value, initial state and, for the four canvas designers, after selecting the item so its property panel is open): 58 probe rows, 16/16 lit controls. The ACCEPT's '32' does not reproduce from its own named list under one counting rule (29 named members, two of them arrays whose element types add 4 + 3); counted member by member the unread record set is 37, of which one is a NEW finding (DataModelField.label, unread by the same method) and the rest are the named list; ReportDesignerSchema.margins top/right/bottom is a further inline sub-key finding outside the population. Recommendation per the maintainer's criterion: read 2 node + 25 record members (+ margins if admitted), retire 2 node (autoLayout, previewMode) + 4 record (parentId, serviceEndpoint, ObjectDefinition.relationships, DesignerFieldDefinition.validationRules) with their 7 transitive members. The 3 TS-required members split: sourceField / targetField READ (only producer a test fixture; the relationship line should anchor at the field rows), relatedObject RETIRE with its parent (spec and mainstream model relationships as fields; producers are 3 test fixtures + 1 gate pin).",
 "population_reconciliation": {
  "node_members": "4, identical to release 5944563530 item 1: DataModelDesignerSchema.autoLayout, ProcessDesignerSchema.version / .lanes, ReportDesignerSchema.previewMode (UnmirroredDeclared still carries exactly these, 3 entries / 4 keys)",
  "record_members_named_by_the_accept_population": "the dev report 5944010306 open_questions[0] (the list the ACCEPT 5944406564 answers): 27 direct members + ObjectDefinition.relationships (all 4 ObjectDefinitionRelationship members) + DesignerFieldDefinition.validationRules (all 3 DesignerValidationRule members) = 29 named, 36 counted member by member; '32' matches neither rule (it matches 27 + relationships + its 4 members, which drops validationRules, or 27 + 2 arrays + 3 rule members, which drops the relationship members) — the figure is not reproducible from the record and should not be carried forward",
  "re_derived_counting_rule": "every member of a named record type reachable from the six nodes whose value no non-test source reads, plus the members of an unread array member's element type, counted individually",
  "re_derived_result": "37 record members: the 36 named slots + DataModelField.label (new). Card total under this rule: 4 + 37 = 41 rows (+ margins sub-keys flagged separately)",
  "borderline_not_added": "DesignerComponent.props (TS-required): probe IDENTICAL and never drawn, but PageDesigner names it (handleUpdateProperty merges into it, behind a branch the property editor never reaches, and it round-trips through onChange). Recorded for the seat, not counted"
 },
 "tests": "Runtime probe (temporary file packages/plugin-designer/src/__tests__/zz-probe-11434.test.tsx, deleted, git status --porcelain empty after; source kept in the scratchpad issue-11434/ for phase 2): real SchemaRenderer from @object-ui/react + real registry (@object-ui/components, @object-ui/fields, @object-ui/plugin-designer imported for registration), innerHTML diffed after normalising React useId / radix ids, each member unset vs a distinctive value, initial state and (page / data-model / process / report) after clicking the item. Run 3 through os-verify-lock (slot issue-11434-probe): 'Test Files 1 passed (1) / Tests 1 passed (1)', VERDICT command-exit 0. Every member under test: initial=IDENTICAL, selected=IDENTICAL (object-manager / field-designer: initial only). Lit controls, 16/16 changed=true: page component.label, page canvas.width (style), page palette item.label, dm relationship.label, dm field.required, dm readOnly, dm entity.color (a style backgroundColor reaches the markup), dm entity.description (selected), process edge.label, process processName, process node.description (selected), report section.repeat, report element.properties.text, report showPropertyPanel, om object.label, fd field.label. Run 1 VERDICT 1, two causes, both fixed before reading any result: (a) my control 'page canvas.showGrid' read IDENTICAL because PageDesigner draws it as a radial-gradient backgroundImage and happy-dom drops that value from serialized style — a probe blind spot for gradient-valued styles, so canvas.width replaced it as the style control and entity.color was added to prove a colour value is visible; (b) the network-escape guard: the object-manager / field-designer grids call /api/v1/security/explain, so a module-scope fetch double was installed. Run 2 passed but the runner suppresses console output on pass, so run 3 wrote the rows to a scratch file. No ablation applies (nothing built). No commit exists, so no sha to cite; tree = origin/main 31987bd50.",
 "gates": [
  "probe: bash /home/user/objectstack/scripts/pm/os-verify-lock.sh -c 'cd /home/user/objectui-issue-11434 && PROBE_OUT=… pnpm exec vitest run packages/plugin-designer/src/__tests__/zz-probe-11434.test.tsx' → 'os-verify-lock: VERDICT command-exit 0 · held the lock 31s · waited 0s'",
  "node scripts/check-designer-field-key-parity.mjs exit 0 → 'designer-field-key-parity: OK'; its UI-only list includes 'validationRules (DesignerFieldDefinition, vs FieldSchema)' and 'relationships (ObjectDefinition, vs ObjectSchema)' (oracle @objectstack/spec 17.5.0)",
  "pnpm check:component-surface-parity exit 0 (report-only) → 'plugin-designer:data-model-designer.autoLayout … `inputs` publishes `autoLayout`, which is outside (spec ∪ renderer reads)'",
  "spec accept sets read off @objectstack/spec/data 17.5.0 by import: FieldSchema has min, max, minLength, maxLength, deleteBehavior, reference and no validationRules; ObjectSchema has validations and indexes and no relationships",
  "greps: non-test `git grep -w` per member over packages/plugin-designer/src, packages/app-shell/src, apps/console/src (readers) and over the whole objectui tree + objectstack 4e6dc233 (producers); zero-hit claims carry lit controls on the same files (e.g. canvas.zoom / gridSize / showGrid hits in PageDesigner, entity.description / rel.label in DataModelDesigner, node.description / edge.label in ProcessDesigner, section.repeat / properties.text in ReportDesigner, pluralLabel / fieldCount in ObjectManager, placeholder / externalId in FieldDesigner); skills/ (27 files, control: 'designer' matches 4) carries no designer node or member",
  "NOT MEASURED: out-of-repo TS consumers of @object-ui/types (published); reason: no instrument reaches them",
  "NOT MEASURED: the object-manager / field-designer edit-form state in the probe; reason: their form initialValues enumerate named keys (7 / 14), none of them relationships / validationRules, read from source"
 ],
 "phase2_surface": {
  "types": "packages/types/src/designer.ts (TS face; the DataModelDesignerSchema `children` docblock names autoLayout among the registration inputs), packages/types/src/zod/designer.zod.ts (tombstones for retired members; mirror version / lanes and a new BPMNLaneSchema if read; header prose), packages/types/src/zod/index.zod.ts and packages/types/src/index.ts (exports of BPMNLaneSchema, and of ObjectDefinitionRelationship / DesignerValidationRule if removed), packages/types/src/zod/README.md (Designer Components group)",
  "ledgers_and_pins": "packages/types/src/__tests__/zod-mirror-parity.test.ts (the 3 designer UnmirroredDeclared entries / 4 keys leave; the pinned totals line '8 entries / 20 keys' and its split figures move; EXPECTED_MIRROR_PAIRS moves with any pair added or removed; history sentences), packages/types/src/__tests__/designer-zod-arms-10859.test.ts (the four-omission type pins and the full-document fixture that authors every member), scripts/__tests__/check-designer-field-key-parity.test.ts (the 'keeps `group`, `sortOrder` and `relationships` on the UI model' pin), packages/plugin-designer/src/__tests__/ObjectManager.test.tsx and packages/app-shell/src/services/MetadataService.specKeyObjectPayload.test.ts (fixtures authoring relationships); MetadataService.ts prose naming relationships as kept on the UI model",
  "renderers_for_read_members": "packages/plugin-designer/src/PageDesigner.tsx (backgroundColor, children, locked, visible, zIndex, palette icons, preview), DataModelDesigner.tsx (sourceField / targetField line anchors, field label / unique / defaultValue / description, onDelete / onUpdate, backgroundColor; drop DataModelDesignerProps.autoLayout), ProcessDesigner.tsx (version, lanes, properties, assignee, dueDate, script, condition, isDefault, backgroundColor; drop serviceEndpoint use), ReportDesigner.tsx (dataBinding replacing properties.field, format, groupField, pageBreakBefore, margins; drop ReportDesignerProps.previewMode); one render-probe pin per read member",
  "teaching_and_i18n": "packages/plugin-designer/README.md (DataModelDesigner fence `autoLayout`), content/docs/plugins/plugin-form.mdx (DesignerValidationRule near-spelling sentence, if the interface goes), packages/plugin-designer/src/hooks/useDesignerTranslation.ts + packages/i18n/src/locales/*.ts (validationRules / addRule labels, seat's call)",
  "changesets": "@object-ui/types minor with a BREAKING (authoring) banner + migration line (retired members); @object-ui/plugin-designer for the readers and the component-props removals (fixed group, so minor/patch only, never major)",
  "overlap_with_objectui_10859_phase_2b": "YES on packages/plugin-designer/src/index.tsx: retiring autoLayout removes the data-model-designer registration `inputs` row there (and reading `version` may add a process-designer input row). Everything else in this surface is disjoint from that file; whichever lands second merges main and re-reads the registration block"
 },
 "retire_form": "RETIRE_FORM in the rows = the TS face takes a `?: never` tombstone or a plain removal per the retire-vs-remove discriminator complex.ts states, the zod arm a retirementTombstone, and the changeset one migration line per member; the seat picks tombstone vs removal",
 "four_axis_frame": "Applied per row (axis notes in each recommendation) and stated once in the CROSS-CUTTING row: measured business pull is zero for every member; contract-first decides the four retirements that duplicate a spec-modelled concept or another member (relationships, validationRules, parentId; serviceEndpoint is a different shape); declaration-is-enforcement is met by either arm; startup focus favours retire everywhere and is overridden on this card by the maintainer's criterion where the mainstream has the capability. Framework block received verbatim from the dispatch (pm-dispatch SKILL.md at 9dae4752).",
 "line_budget": "n/a (phase 1 changes no file)",
 "files_changed": [],
 "deviations": [
  "No empty-branch push: the role file's rule 1 makes it the first action after creating the branch; the dispatch's phase-1 write budget is exactly one write with 'No commit, no push, no PR'. Phase 1 made no edit, so neither of the push's two purposes (landing marker before an edit, write-route probe before a push) was due; the push moves to phase 2's first action. Conflict declared, not silently resolved.",
  "Install: node v22.22.0 is below jsdom@30's engines floor and objectui's .npmrc sets engine-strict=true; installed with 'pnpm install --config.engine-strict=false' as instructed (exit 0).",
  "Population: re-derived with an explicit counting rule because the ACCEPT's '32' is not reproducible from its named list (see population_reconciliation); one addition, one borderline, one out-of-population sub-key finding.",
  "Mainstream citations: each URL is a page the web search returned for that claim; the claim text rests on the search tool's summary of the page, not a full read of every page. Readings without a returned source are marked 'unverified (from knowledge)'.",
  "One read attempt was refused: 'gh pr list' (GraphQL) answered 403 in this session; no open-PR overlap scan was taken beyond the claim's serial-constraint reading and a read of the objectui#10859 worktree's own diff (no plugin-designer/src/index.tsx change in its phase 2a).",
  "The worktree /home/user/objectui-issue-11434 is KEPT for phase 2 as ordered: HEAD 31987bd50, git status --porcelain empty, node_modules present, nothing started in the background."
 ],
 "mcp_calls": "0",
 "api_writes": "1 — this os-dev-report via node scripts/pm/post-stamped.mjs --repo=objectstack-ai/objectui --comment=11434 (fleet-write relay: POST /repos/objectstack-ai/objectstack/dispatches, executed as POST /repos/objectstack-ai/objectui/issues/11434/comments). Reads only otherwise: gh api GET of the card, its comments, and comments 5944563530, 5944406564, 5944582356, 5944010306, 5861341151; web searches for the mainstream readings.",
 "open_questions": [
  {
   "question": "DataModelDesignerSchema.autoLayout (node) · faces: TS optional; zod UNMIRRORED (UnmirroredDeclared); also DataModelDesignerProps.autoLayout and the data-model-designer registration `inputs` row in plugin-designer index.tsx · readers (grep): 0 (declared on props, never destructured; the toolbar Auto Layout button runs calculateGridAutoLayout on demand) · probe true vs unset and false vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test; teaching sites: plugin-designer README DataModelDesigner fence passes `autoLayout` as a prop; check:component-surface-parity reports the published input as outside (spec ∪ renderer reads) · mainstream: Salesforce Schema Builder offers Auto-Layout as an on-demand button whose result cannot be undone, an action, not a saved model flag (https://micropyramid.com/blog/how-to-access-schema-builder-in-salesforce/, https://www.salesforceben.com/5-ways-to-use-salesforce-schema-builder-for-data-modeling/); the ObjectStack studio ER config models layout as an algorithm enum `layout: force|hierarchy|grid|circular`, not a boolean (objectstack packages/spec/src/studio/object-designer.zod.ts, ERDiagramConfigSchema) · confidence: medium-high",
   "options": [
    "read: on mount, when true, run calculateGridAutoLayout over the authored entities (DataModelDesigner initial state); overrides authored positions, which are TS-required",
    "retire: refuse by name on both faces (RETIRE_FORM); also drop DataModelDesignerProps.autoLayout, the registration input row (index.tsx, overlaps objectui#10859 phase 2b) and the README fence attribute"
   ],
   "recommendation": "retire — the mainstream capability (auto-layout as an action) already exists as the toolbar button; no mainstream model persists an auto-layout flag, and positions are required on every entity. Axes: no business pull (0 producers); contract-first (the spec spells layout as an algorithm, a boolean here would be a second dialect); AI-proofing (a refused key is loud, today it is silent)"
  },
  {
   "question": "ProcessDesignerSchema.version (node) · faces: TS optional; zod UNMIRRORED; ProcessDesignerProps.version · readers (grep): 0 (declared on props, never destructured) · probe 'ZZ_9.9' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test; no teaching site · mainstream: yes, persisted: Camunda `camunda:versionTag` on the process is set in the Modeler properties panel and saved in the BPMN XML (https://docs.camunda.org/manual/latest/user-guide/process-engine/process-versioning/); Salesforce Flow Builder shows and manages versions (https://help.salesforce.com/apex/HTViewHelpDoc?id=sf.flow_ref_properties.htm&language=en_US) · confidence: medium",
   "options": [
    "read: ProcessDesigner header shows the version beside processName (the header span with aria-label \"Process name\"); mirror `version` on the zod arm",
    "retire: refuse by name on both faces (RETIRE_FORM); drop ProcessDesignerProps.version"
   ],
   "recommendation": "read — mainstream yes; the cheapest reader of the set (one header label). Axes: business pull 0 (axis 1/4 favour retire), criterion binds read"
  },
  {
   "question": "ProcessDesignerSchema.lanes (node) · faces: TS optional (BPMNLane[]: id, label, role?, nodeIds); zod UNMIRRORED and BPMNLane has no mirror; ProcessDesignerProps.lanes · readers (grep): 0 (destructured into the unused binding `_lanes`) · probe one lane over both nodes vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test; no teaching site · mainstream: yes, core BPMN: bpmn.io models lanes with add/remove/resize controls (https://bpmn.io/blog/posts/2015-lanes-data-objects-more-bpmn) · confidence: medium",
   "options": [
    "read: ProcessDesigner canvas draws a labelled band per lane behind its nodeIds (the SVG layer that draws edges); mirror `lanes` plus a new BPMNLaneSchema record mirror",
    "retire: refuse by name on both faces (RETIRE_FORM); drop ProcessDesignerProps.lanes and the BPMNLane export if nothing else uses it"
   ],
   "recommendation": "read — mainstream yes (swimlanes are core BPMN). Cost medium: lane geometry over free-positioned nodes. Axes: criterion binds read; axis 4 tension stated in the cross-cutting row"
  },
  {
   "question": "ReportDesignerSchema.previewMode (node) · faces: TS optional; zod UNMIRRORED; ReportDesignerProps.previewMode · readers (grep): 0 (declared on props, never destructured) · probe true vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test; no teaching site · mainstream: preview exists in Report Builder as the Run / Design toggle in the tool, which renders the layout with data (https://learn.microsoft.com/en-us/power-bi/paginated-reports/report-builder-previewing-reports); it is a viewer action, not a report-definition property (that it is not saved in the definition is unverified (from knowledge)) · confidence: medium",
   "options": [
    "read: ReportDesigner renders chrome-off (no dashed outlines, toolbar, property panel); a data preview would also need dataBinding plus a data path this designer does not have",
    "retire: refuse by name on both faces (RETIRE_FORM); drop ReportDesignerProps.previewMode"
   ],
   "recommendation": "retire — mainstream preview is a tool mode, not an authored key; the chrome-off half is already expressible with readOnly + showToolbar:false + showPropertyPanel:false, and a data preview is a new capability (data path) with zero pull. Axes: contract-first (no duplicate switch), startup scope"
  },
  {
   "question": "DesignerCanvasConfig.backgroundColor · faces: TS optional; zod mirrored (DesignerCanvasConfigSchema); record shared by page-, data-model- and process-designer canvases · readers (grep): 0 in all three designers (every `backgroundColor` hit in plugin-designer is another component's style) · probe '#123456' vs unset on all three canvases: initial=IDENTICAL selected=IDENTICAL (style control: entity.color '#123456' DOES reach the markup, so a colour value is visible to this probe) · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes, persisted per page in Figma: the canvas background color is set for everyone with access to the file (https://help.figma.com/hc/en-us/articles/360041064814-Change-the-background-color-of-the-canvas) · confidence: medium",
   "options": [
    "read: the three canvas surfaces paint it through a CSS custom property consumed by a static Tailwind utility (AGENTS.md author-declared colour carve-out), never an inline backgroundColor",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; reader must follow the colour carve-out. Axes: criterion binds read; low cost"
  },
  {
   "question": "DesignerComponent.children · faces: TS optional (recursive); zod mirrored via the TreeNodeSchema construct (DesignerComponentSchema, RECURSION_BREAKING_MIRRORS row) · readers (grep): 0 (PageDesigner draws a flat list; no `children` read) · probe one nested child vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Retool containers group nested components with their own layout (https://docs.retool.com/apps/guides/layout-structure/container); PageDesigner's own default palette offers container / flex / grid / card / tabs · confidence: medium",
   "options": [
    "read: PageDesigner canvas renders a component's children inside its box and the component-tree panel shows the nesting",
    "retire: refuse by name on both faces (RETIRE_FORM); DesignerComponentSchema stops being recursive (its RECURSION_BREAKING_MIRRORS row goes)"
   ],
   "recommendation": "read — mainstream yes, and it is the tree's canonical spelling (keep it over parentId). Cost medium. Axes: criterion binds read"
  },
  {
   "question": "DesignerComponent.parentId · faces: TS optional; zod mirrored · readers (grep): 0 (the only `parentId` hits are NavigationDesigner's own callback parameter) · probe 'ZZ_PARENT' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: nesting is mainstream, but as ONE encoding of the tree; a parent pointer beside a children array is a second spelling of the same fact (that mainstream formats encode the tree once is unverified (from knowledge)) · confidence: medium",
   "options": [
    "read: PageDesigner derives nesting from parentId (then `children` would be the duplicate)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "retire — contract-first: one strict spelling for the tree; `children` already carries it and is the mirror's recursion. If the seat prefers a flat list + parentId instead, invert this row and the `children` row together"
  },
  {
   "question": "DesignerComponent.locked · faces: TS optional; zod mirrored · readers (grep): 0 · probe true vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Figma lock / unlock layers (https://help.figma.com/hc/en-us/articles/360041596573-Lock-and-unlock-layers) · confidence: medium",
   "options": [
    "read: PageDesigner skips drag, resize, property edit and delete for a locked component and shows a lock affordance (the canvas item's mouse/drag handlers and handleUpdateProperty / handleDeleteComponent)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; small reader. Axes: criterion binds read"
  },
  {
   "question": "DesignerComponent.visible · faces: TS optional; zod mirrored · readers (grep): 0 (the `visible` hits are NavigationDesigner's own item.visible) · probe false vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Figma toggle visibility to hide layers (https://help.figma.com/hc/en-us/articles/360041112614-Toggle-visibility-to-hide-layers) · confidence: medium",
   "options": [
    "read: PageDesigner hides (or dims) an invisible component on the canvas and marks it in the component tree",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; small reader"
  },
  {
   "question": "DesignerComponent.zIndex · faces: TS optional; zod mirrored · readers (grep): 0 · probe 999 vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Webflow exposes z-index in the Style panel's Position section (https://help.webflow.com/hc/en-us/articles/33961228002963-CSS-position-properties) · confidence: medium",
   "options": [
    "read: PageDesigner canvas item stacks by zIndex (data-driven, like the left/top/width/height it already places from position)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; small reader"
  },
  {
   "question": "DesignerPaletteCategory.icon · faces: TS optional; zod mirrored · readers (grep): 0 (palette renders category.label only) · probe 'Database' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture; PageDesigner's DEFAULT_PALETTE authors no icon · mainstream: insert/palette panels show an icon per category: unverified (from knowledge) · confidence: low",
   "options": [
    "read: PageDesigner palette header renders the named lucide icon through resolveIcon (exported by @object-ui/components, already a plugin-designer dependency)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read, low confidence — the criterion turns on an unverified reading; if the seat wants a cited reading first, retire is the fallback"
  },
  {
   "question": "DesignerPaletteItem.icon · faces: TS optional; zod mirrored · readers (grep): 0 (palette item renders item.label only) · probe 'Type' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture; DEFAULT_PALETTE authors no icon · mainstream: component palettes show an icon per insertable component: unverified (from knowledge) · confidence: low",
   "options": [
    "read: PageDesigner palette button renders resolveIcon(item.icon) before its label",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read, low confidence — same reading as the category icon"
  },
  {
   "question": "DesignerPaletteItem.preview · faces: TS optional; zod mirrored · readers (grep): 0 · probe URL vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Plasmic code components register a `thumbnailUrl` shown in Studio's insert panel (https://docs.plasmic.app/learn/code-components-ref/) · confidence: low-medium",
   "options": [
    "read: PageDesigner palette shows the thumbnail (hover card or inline image)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes (cited), low cost; low-medium because the cited vendor is one product"
  },
  {
   "question": "DataModelField.label (ADDITION: not in the ACCEPT list) · faces: TS optional; zod mirrored · readers (grep): 0 (DataModelDesigner draws field.name only; every `.label` hit is entity.label or rel.label) · probe 'ZZ_FIELD_LABEL' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Schema Builder's View Options toggle between field labels and API names (https://www.salesforceben.com/5-ways-to-use-salesforce-schema-builder-for-data-modeling/) · confidence: medium",
   "options": [
    "read: DataModelDesigner field row shows the label (name stays the edited identifier), or a labels/names toggle",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes. Found by the same source read the ACCEPT used, confirmed by the probe: a reconciliation difference for the seat"
  },
  {
   "question": "DataModelField.unique · faces: TS optional; zod mirrored · readers (grep): 0 in DataModelDesigner (the 11 `unique` hits are FieldDesigner's own DesignerFieldDefinition.unique) · probe true vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: DBML column settings carry `unique` (https://dbml.dbdiagram.io/docs/) · confidence: medium-high",
   "options": [
    "read: DataModelDesigner field row shows a UQ badge beside the existing PK badge",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; smallest reader in this designer"
  },
  {
   "question": "DataModelField.defaultValue · faces: TS optional (unknown); zod mirrored · readers (grep): 0 in DataModelDesigner · probe 'ZZ_DEFAULT' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes in the model: DBML column setting `default` (https://dbml.dbdiagram.io/docs/); whether ER canvases draw it (vs a tooltip) is unverified (from knowledge) · confidence: low-medium",
   "options": [
    "read: DataModelDesigner field row tooltip (title) or the property panel shows the default",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream has it in the model; reader is a tooltip. Low-medium"
  },
  {
   "question": "DataModelField.description · faces: TS optional; zod mirrored · readers (grep): 0 (the one `description` read is entity.description in the entity property panel) · probe 'ZZ_FIELD_DESC' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: DBML column `note` (https://dbml.dbdiagram.io/docs/) · confidence: medium",
   "options": [
    "read: DataModelDesigner field row tooltip shows the description",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; tooltip reader"
  },
  {
   "question": "DataModelRelationship.sourceField (TS-REQUIRED) · faces: TS REQUIRED; zod mirrored (required) · readers (grep): 0 (the relationship line reads sourceEntity / targetEntity only) · probe 'zz_src_field' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test in objectui and objectstack; the only constructor of the type is the designer-zod-arms-10859.test.ts fixture (README only `declare`s a DataModelRelationship[]); out-of-repo TS consumers NOT MEASURED · mainstream: yes, and it is what ER designers draw: DBML refs join a column to a column (posts.user_id references users.id) (https://dbml.dbdiagram.io/docs/); Schema Builder draws lookup / master-detail relationship lines (https://www.salesforceben.com/5-ways-to-use-salesforce-schema-builder-for-data-modeling/) · confidence: high",
   "options": [
    "read: DataModelDesigner relationship SVG anchors the line at the source field's row instead of the fixed entity-centre offsets it uses today (the `x1={source.position.x + 120}` / `y1={source.position.y + 50}` line)",
    "retire: refuse by name on both faces (RETIRE_FORM); BREAKS every TS consumer that constructs DataModelRelationship (in-repo: one test fixture) since the member is required"
   ],
   "recommendation": "read — answers Zone 2 item 3: the line ignores which fields connect, and the mainstream ER drawing anchors FK row to PK row, so the designer should be drawing it. Retire would also remove the only field-level link in the model"
  },
  {
   "question": "DataModelRelationship.targetField (TS-REQUIRED) · faces: TS REQUIRED; zod mirrored (required) · readers (grep): 0 · probe 'zz_tgt_field' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: as sourceField (one test fixture) · mainstream: yes: as sourceField (https://dbml.dbdiagram.io/docs/) · confidence: high",
   "options": [
    "read: the same reader: the line's target end anchors at targetField's row",
    "retire: refuse by name on both faces (RETIRE_FORM); breaks TS constructors as sourceField"
   ],
   "recommendation": "read — with sourceField; the pair is one reader"
  },
  {
   "question": "DataModelRelationship.onDelete · faces: TS optional (cascade | set-null | restrict | no-action); zod mirrored · readers (grep): 0 (the `onDelete` hits are ObjectGrid callback props in FieldDesigner / ObjectManager) · probe 'cascade' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: DBML relationship settings `[delete: cascade]` (https://dbml.dbdiagram.io/docs/); SQL ON DELETE (https://dev.mysql.com/doc/refman/8.0/en/create-table-foreign-keys.html) · confidence: medium",
   "options": [
    "read: DataModelDesigner relationship label / tooltip shows the delete action",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes. AI-proofing caveat for the seat: on ObjectStack field documents the spec refuses `onDelete` as an alias of `deleteBehavior` (field.zod.ts alias table) and spells `set_null`, not `set-null`; this designer's vocabulary is a second dialect of that concept"
  },
  {
   "question": "DataModelRelationship.onUpdate · faces: TS optional; zod mirrored · readers (grep): 0 · probe 'restrict' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes in ER tools: DBML `[update: no action]` (https://dbml.dbdiagram.io/docs/), SQL ON UPDATE (https://dev.mysql.com/doc/refman/8.0/en/create-table-foreign-keys.html); the ObjectStack spec has no update behaviour · confidence: low-medium",
   "options": [
    "read: shown with onDelete on the relationship label / tooltip",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — ER mainstream yes; lower confidence because the platform's own field contract has no counterpart"
  },
  {
   "question": "BPMNNode.properties · faces: TS optional (Record); zod mirrored · readers (grep): 0 (property panel fields are label / type / description) · probe {zz:'ZZ_PROP'} vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Camunda `zeebe:properties` key/value extension properties on any element (https://github.com/camunda/camunda/issues/9868) · confidence: medium",
   "options": [
    "read: ProcessDesigner property panel lists the key/value pairs of the selected node (propertyFields)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes"
  },
  {
   "question": "BPMNNode.assignee · faces: TS optional; zod mirrored · readers (grep): 0 · probe 'ZZ_ASSIGNEE' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Camunda user tasks `assignee` (https://docs.camunda.io/docs/components/modeler/bpmn/user-tasks/) · confidence: high",
   "options": [
    "read: ProcessDesigner property panel field for user-task nodes (and an optional subtitle on the node)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes, core user-task attribute"
  },
  {
   "question": "BPMNNode.dueDate · faces: TS optional; zod mirrored · readers (grep): 0 · probe 'ZZ_DUE' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Camunda user task schedule `dueDate` (https://docs.camunda.io/docs/components/modeler/bpmn/user-tasks/) · confidence: medium-high",
   "options": [
    "read: ProcessDesigner property panel field for user-task nodes",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes"
  },
  {
   "question": "BPMNNode.script · faces: TS optional; zod mirrored · readers (grep): 0 (the `script` hits are the 'script-task' type literal) · probe 'ZZ_SCRIPT' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Camunda script tasks carry an inline expression (https://docs.camunda.io/docs/components/modeler/bpmn/script-tasks/) · confidence: medium-high",
   "options": [
    "read: ProcessDesigner property panel textarea for script-task nodes",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes"
  },
  {
   "question": "BPMNNode.serviceEndpoint · faces: TS optional; zod mirrored · readers (grep): 0 · probe URL vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: the mainstream service task references an implementation (Camunda: a job type the worker subscribes to), not an endpoint URL (https://docs.camunda.io/docs/components/modeler/bpmn/service-tasks/) · confidence: low-medium",
   "options": [
    "read: ProcessDesigner property panel field for service-task nodes, as-is",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "retire — the mainstream member is a different shape (implementation / job-type reference); reading a URL would teach a model the mainstream does not use. A spec-shaped implementation reference is a separate decision. Low-medium: the seat may read 'service tasks are mainstream' as yes"
  },
  {
   "question": "BPMNEdge.condition · faces: TS optional; zod mirrored · readers (grep): 0 · probe 'ZZ_COND' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: BPMN conditional sequence flows (https://docs.cibseven.org/manual/2.0/reference/bpmn20/gateways/sequence-flow/) · confidence: high",
   "options": [
    "read: ProcessDesigner edge renders the condition (label / tooltip) and an edge property entry",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes, core BPMN"
  },
  {
   "question": "BPMNEdge.isDefault · faces: TS optional; zod mirrored · readers (grep): 0 · probe true vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: BPMN default sequence flow, whose condition is ignored (https://docs.cibseven.org/manual/2.0/reference/bpmn20/gateways/sequence-flow/) · confidence: high",
   "options": [
    "read: ProcessDesigner edge draws the default-flow slash marker",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes, core BPMN"
  },
  {
   "question": "ReportDesignerElement.dataBinding · faces: TS optional; zod mirrored · readers (grep): 0 (field elements draw the undeclared `properties.field` instead: the field element's interpolated span and the property panel's Field entry) · probe 'ZZ_BIND' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture (that fixture authors BOTH properties.field and dataBinding) · mainstream: yes: Report Builder text boxes bind through expressions (https://learn.microsoft.com/en-us/power-bi/paginated-reports/expressions/expression-uses-reports-report-builder) · confidence: medium",
   "options": [
    "read: ReportDesigner field element and property panel read dataBinding, and stop reading the undeclared properties.field (spec declaration over implementation)",
    "retire: refuse by name on both faces (RETIRE_FORM); the renderer keeps its undeclared properties.field convention"
   ],
   "recommendation": "read — mainstream yes. Finding for the seat: the renderer already reads the same concept from an undeclared bag key; contract-first says the declared member wins, so read = move the renderer onto dataBinding"
  },
  {
   "question": "ReportDesignerElement.format · faces: TS optional (12 sub-keys); zod mirrored · readers (grep): 0 (no sub-key read either) · probe all 12 sub-keys set vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Report Builder text box formatting (https://learn.microsoft.com/en-us/power-bi/paginated-reports/report-design/textbox/format-text-in-a-text-box-report-builder-and-service) · confidence: medium",
   "options": [
    "read: ReportDesigner element applies weight / style / alignment through utility classes, colours through the CSS-custom-property carve-out; numberFormat / dateFormat only matter once values are drawn",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes; the largest reader of the record set"
  },
  {
   "question": "ReportDesignerSection.groupField · faces: TS optional; zod mirrored · readers (grep): 0 · probe 'ZZ_GROUP' vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Report Builder row groups (Group Properties) (https://learn.microsoft.com/en-us/power-bi/paginated-reports/report-design/add-page-break-report-builder) · confidence: medium-high",
   "options": [
    "read: ReportDesigner section header shows the group field on group-header / group-footer sections (where it already draws the repeat marker)",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes, small reader"
  },
  {
   "question": "ReportDesignerSection.pageBreakBefore · faces: TS optional; zod mirrored · readers (grep): 0 · probe true vs unset: initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: yes: Report Builder page break at the start of a group (https://learn.microsoft.com/en-us/power-bi/paginated-reports/report-design/add-page-break-report-builder) · confidence: medium-high",
   "options": [
    "read: ReportDesigner section header shows a page-break marker",
    "retire: refuse by name on both faces (RETIRE_FORM)"
   ],
   "recommendation": "read — mainstream yes, small reader"
  },
  {
   "question": "ReportDesignerSchema.margins.top / .right / .bottom (ADDITION: inline sub-keys of a mirrored, read node member; outside the ACCEPT population) · faces: TS optional inline object; zod mirrored · readers (grep): only margins.left is read (it places a new element); top / right / bottom: 0 · probe margins {top:97,…,left:40} vs unset (default all 40): initial=IDENTICAL selected=IDENTICAL · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: page margins are standard page setup in paginated-report tools: unverified (from knowledge) · confidence: medium",
   "options": [
    "read: ReportDesigner page surface pads by all four margins",
    "retire: narrow the member to the read sub-key (BREAKING for authors of the other three)"
   ],
   "recommendation": "read — for the seat to admit or leave out of this card; not counted in the 36"
  },
  {
   "question": "ObjectDefinition.relationships · faces: TS optional (ObjectDefinitionRelationship[]); zod mirrored · readers (grep): 0 in ObjectManager / MetadataObjectsPage; MetadataService.toObjectPayload deliberately keeps it off the wire (objectui#6223) · probe initial=IDENTICAL (no selection state for this designer; the edit form's initialValues enumerate named keys, none of them this one) · producers: 0 non-test (toObjectDefinition never sets it); test-only: designer-zod-arms fixture, ObjectManager.test.tsx MOCK_OBJECTS, MetadataService.specKeyObjectPayload.test.ts MANAGED · mainstream: the mainstream models a relationship as a FIELD (Salesforce Lookup / Master-Detail created under Fields & Relationships) (https://help.salesforce.com/s/articleView?id=platform.overview_of_custom_object_relationships.htm&language=en_US&type=5); the ObjectStack spec agrees: ObjectSchema refuses object-level `relationships` by name (check-designer-field-key-parity lists it uiOnly) and the field form is already read as DesignerFieldDefinition.referenceTo · confidence: high",
   "options": [
    "read: ObjectManager shows a relationships column / panel from this array",
    "retire: refuse by name on both faces (RETIRE_FORM); the parity gate's self-test pin 'keeps `group`, `sortOrder` and `relationships` on the UI model' and the two app-shell / plugin-designer fixtures change"
   ],
   "recommendation": "retire — mainstream and spec both say 'relationship = field'; reading this array would fossilize an object-level dialect the platform refuses. Axes: all four agree"
  },
  {
   "question": "ObjectDefinitionRelationship.relatedObject (TS-REQUIRED; transitive) · faces: TS REQUIRED; zod mirrored (required) · readers (grep): 0 (only reachable through ObjectDefinition.relationships) · probe: covered by the relationships row (the array authored with relatedObject 'zz_related': IDENTICAL) · producers: the same three test fixtures + the parity gate pin; out-of-repo TS consumers NOT MEASURED · mainstream: as ObjectDefinition.relationships · confidence: high",
   "options": [
    "read: only as part of a relationships reader",
    "retire: goes with its parent; ObjectDefinitionRelationship becomes orphaned (seat: remove the export, BREAKING for TS importers, or keep it unused)"
   ],
   "recommendation": "retire with relationships — Zone 2 item 3: producers exist only in tests; under retire those 3 fixtures + 1 pin change and any external constructor of ObjectDefinition.relationships stops compiling"
  },
  {
   "question": "ObjectDefinitionRelationship.type (REQUIRED) / .label / .foreignKey (transitive, one row for three members) · faces: TS; zod mirrored · readers (grep): 0 each · probe: covered by the relationships row · producers: the same three test fixtures · mainstream: as ObjectDefinition.relationships · confidence: high",
   "options": [
    "read: only as part of a relationships reader",
    "retire: goes with its parent"
   ],
   "recommendation": "retire with relationships"
  },
  {
   "question": "DesignerFieldDefinition.validationRules · faces: TS optional (DesignerValidationRule[]); zod mirrored · readers (grep): 0 in FieldDesigner / MetadataFieldsPage / MetadataService (toDesignerField, fromDesignerField and toFieldPayload all skip it); i18n carries labels for an editor never built ('appDesigner.fieldDesigner.validationRules', 'addRule' in 10 locales + the designer fallback table) · probe initial=IDENTICAL (no selection state for this designer; the edit form's initialValues enumerate named keys, none of them this one) · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: Salesforce validation rules are OBJECT-level formula rules with an error location (https://help.salesforce.com/s/articleView?language=en_US&id=platform.fields_validation_rules_fields.htm&type=5); the ObjectStack spec refuses `validationRules` on FieldSchema (uiOnly in the parity gate) and models the same constraints as first-class field keys min / max / minLength / maxLength plus object-level `validations` (objectstack lint names `validationRules` a refused alias of `validations`) · confidence: high",
   "options": [
    "read: FieldDesigner drawer edits the rule list; it would still never reach the wire (no converter carries it)",
    "retire: refuse by name on both faces (RETIRE_FORM); the i18n labels go too or stay (seat)"
   ],
   "recommendation": "retire — not the mainstream shape and a spec-refused spelling; the honest 'read' is a different, spec-shaped feature (min / max controls). Axes: all four agree"
  },
  {
   "question": "DesignerValidationRule.type (REQUIRED) / .value (REQUIRED) / .message (transitive, one row for three members) · faces: TS; zod mirrored · readers (grep): 0 each · probe: covered by the validationRules row (one pattern rule authored: IDENTICAL) · producers: 0 non-test (objectui, objectstack 4e6dc233); test-only: designer-zod-arms-10859.test.ts full-document fixture · mainstream: as DesignerFieldDefinition.validationRules · confidence: high",
   "options": [
    "read: only as part of a validationRules reader",
    "retire: goes with its parent; DesignerValidationRule stays exported unless removed, and is named as a near-spelling in content/docs/plugins/plugin-form.mdx and in a comment of the objectui#10719 pin (not one of its firing controls, measured)"
   ],
   "recommendation": "retire with validationRules"
  },
  {
   "question": "CROSS-CUTTING · the maintainer's criterion (binding here) vs the four axes. Measured business need: zero producers of any of the six designer node types or of these members outside test fixtures, in objectui and objectstack; the four canvas designers (page / data-model / process / report) are not mounted anywhere in the product (no import outside plugin-designer, only registered and exported); ObjectManager / FieldDesigner are mounted only through MetadataObjectsPage / MetadataFieldsPage, whose converters never produce the two retire-candidates. The recommendations above apply the criterion and yield about 27 new readers across four components in one change.",
   "options": [
    "A: apply the criterion as ruled, member by member (rows above): about 27 readers + 6 retirements (+7 transitive) in one change",
    "B: apply it, but stage phase 2 per designer (one PR per component plus one retirement PR) while keeping 'one change across both faces' per member",
    "C: ask the maintainer whether 'does the mainstream have it?' should outrank zero pull for designers the product does not mount (axis 4 would retire nearly all 36)"
   ],
   "recommendation": "A or B, seat's call; the card binds the criterion, so I do not recommend C, but I state the tension: axis 1 (no measured user) and axis 4 (startup focus, no pull ⇒ implementation-first) point at retire for every row, axis 2 (contract-first) and axis 3 (declaration is enforcement) are satisfied by either arm as long as no member stays declared-and-unread. B lowers review risk on a large reader surface without leaving any member unsettled at the card's close."
  }
 ],
 "out_of_scope_findings": [
  "class: none (dead tolerance, not filed) · reach: none — the spec refuses `validationRules` on an object at save, so no stored object document can carry it · evidence: @object-ui/components metadata-viewer StateMachineView reads `obj?.validations ?? obj?.validationRules`, a consumer-side alias of a spec-refused spelling (AGENTS.md #0.1) · carrier: 承接者:无 · noted, not filed · dedupe words: metadata-viewer validationRules alias, state machine validations fallback",
  "class: none (probe-instrument limit, recorded for phase 2) · evidence: happy-dom drops a radial-gradient backgroundImage from serialized style, so a style-carried read of that shape is invisible to an innerHTML diff (PageDesigner canvas.showGrid drew IDENTICAL while its source reads it) · carrier: this card's phase 2 render probes (they must not use a gradient-valued style as their only signal) · noted, not filed",
  "class: none (observation) · evidence: ReportDesignerElement.properties.field is an undeclared bag key the report renderer reads for the same concept as the declared, unread dataBinding · carrier: this card (dataBinding row) · noted, not filed"
 ]
}

Generated by Claude Code

Seat ruling: objectui#11434, member by member; phase 2 ships staged (option B)

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T05:31Z. This answers the phase-1 report on this card (claim 5945883001). It rules under this card's direction: objectui#10874's shape, the maintainer's criterion per member ("does the mainstream have it? Yes ⇒ give it a reader. No ⇒ retire it on both faces"), and "⛔ No member stays declared-and-unread after this card". Readings, probes and citations are the dev's measurement at objectui 31987bd50, unless marked as the seat's.

Population. 4 node members plus 37 record members. The ACCEPT's "32" does not reproduce from its own list (the dev's reconciliation), so 37 is the count carried forward, including the new DataModelField.label. ReportDesignerSchema.margins.top / .right / .bottom (declared, unread, in the same designer) are admitted, so nothing in this family stays declared-and-unread.

The cross-cutting question (option C) is not escalated. The four canvas designers are not mounted in the console, and no producer writes these members, so business pull is zero and axis 4 alone would retire nearly everything. But the criterion is the maintainer's standing ruling for exactly this family (triage 5944767525 (b) on objectui#10859: "The criterion is already the maintainer's standing ruling"), and a maintainer ruling outranks the four-axis frame in this skill's order. The designers are also authorable SDUI node types since objectui#10859 batch 7 armed them, so they are product surface. The criterion is applied as ruled.

Retire on both faces

In the zod arm, a retirement tombstone refused by name. On the TS face, a ?: never tombstone. Each gets one migration line in the changeset. A removed element type (ObjectDefinitionRelationship, DesignerValidationRule) also leaves the exports.

DataModelDesignerSchema.autoLayout: auto-layout is an action, and the toolbar already runs it on demand.
ReportDesignerSchema.previewMode: preview is a tool mode, not authored state.
DesignerComponent.parentId: a second spelling of the tree; children is the canonical one.
BPMNNode.serviceEndpoint: the mainstream service task references an implementation, not a URL.
ObjectDefinition.relationships, with its element type's relatedObject (TS-required), type, label and foreignKey: the spec and the mainstream both model a relationship as a field. Its only producers are three test fixtures and one gate pin, which move with it.
DesignerFieldDefinition.validationRules, with DesignerValidationRule.type / .value / .message: a spec-refused spelling, never read, and its i18n strings belong to an editor that was never built (the dev's measurement; the strings may stay or go as the census says).
DataModelRelationship.onUpdate: retire (seat override of the dev's low-medium "read"). The platform's own relationship contract has no update behaviour, so a designer that drew it would declare a capability the runtime never delivers. The baseline rule is that the protocol wins, and "declaration is enforcement" decides it.

Read: give each a reader, pinned by a render probe that diffs against unset

Data-model designer: DataModelField.label / .unique / .defaultValue / .description; DataModelRelationship.sourceField / .targetField (TS-required): read, the relationship line anchored at the field rows. Their only constructor is a test fixture, so retiring them would gain nothing and lose the ER semantics. DataModelRelationship.onDelete: read, in the platform's own dialect (seat override). The spec refuses onDelete as an alias of deleteBehavior on field documents and spells set_null, not set-null, and code aligns to the protocol. So the member is spelled deleteBehavior, with the spec's value vocabulary, and onDelete is refused by name with a migration line (0 non-test producers).
Process designer: ProcessDesignerSchema.version and .lanes (with a new BPMNLaneSchema mirror); BPMNNode.properties / .assignee / .dueDate / .script; BPMNEdge.condition / .isDefault.
Page designer and the shared canvases: DesignerCanvasConfig.backgroundColor (all three canvases, following the colour carve-out the dev names); DesignerComponent.children / .locked / .visible / .zIndex; DesignerPaletteItem.preview; DesignerPaletteCategory.icon and DesignerPaletteItem.icon. The seat settles the icon reading: per-item icons in an insert or palette panel are standard in mainstream builders (Retool's component panel, Power Apps' Insert pane). That is the seat's reading, from knowledge; it is not fetched.
Report designer: ReportDesignerElement.dataBinding, where read means the renderer moves onto dataBinding and stops reading the undeclared properties.field, because a declared member beats an undeclared bag key; ReportDesignerElement.format; ReportDesignerSection.groupField / .pageBreakBefore; ReportDesignerSchema.margins.top / .right / .bottom.

Shape of phase 2 (option B)

At most three Fixes-or-Part of PRs, in order: (1) every retirement above (with the onDelete → deleteBehavior respelling kept for PR 2 or 3); (2) and (3) the readers, split by designer as review size suggests. Each member is settled on both faces in the PR that carries it, and the mirror-parity ledger (UnmirroredDeclared, the totals line, EXPECTED_MIRROR_PAIRS) moves in the same PR as its members. The final PR says Fixes #11434; earlier ones say Part of #11434. Clause-②: yes throughout, with the at-tier contract review before each ACCEPT. Render probes must not rely only on a gradient-valued style (the dev's happy-dom note). packages/plugin-designer/src/index.tsx is shared with objectui#10859 phase 2b: whichever lands second merges main.

Resumption follows when a batch slot frees.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "phase": "2, PR 1 of at most 3 (every retirement), under seat ruling 5946170549",
 "status": "done",
 "branch": "claude/issue-11434-designer-members",
 "pr": "https://github.com/objectstack-ai/objectui/pull/11451",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id; the card's newest Claim 5945883001 names this branch)",
 "premise_still_valid": true,
 "summary": "PR 1 of the card retires on both faces every member that seat ruling 5946170549 rules RETIRE. The members are DataModelDesignerSchema.autoLayout, ReportDesignerSchema.previewMode, DesignerComponent.parentId, DataModelRelationship.onUpdate (the seat override), BPMNNode.serviceEndpoint, ObjectDefinition.relationships and DesignerFieldDefinition.validationRules. Each is a `?: never` tombstone on the TypeScript face and a retirement tombstone on the zod face, refused by name with the reason and what to write instead. The two element types left the root barrel and the zod barrel with their mirrors: ObjectDefinitionRelationship (relatedObject / type / label / foreignKey) and DesignerValidationRule (type / value / message). The component face agrees: DataModelDesignerProps.autoLayout and ReportDesignerProps.previewMode are gone, and so is the data-model-designer registration's `autoLayout` input row (it measured unread). The mirror-parity ledger moves in the same PR: UnmirroredDeclared goes from 8 entries / 20 keys to 6 / 18, its LOCAL half from 5 / 6 to 3 / 4, the union from 13 to 11, and EXPECTED_MIRROR_PAIRS from 192 to 190, each with its history sentence. i18n decision: retire both dead editor strings. appDesigner.fieldDesigner.validationRules and .addRule are NEEDS-REVIEW in the reverse sweep, with no call site and only a defaults-map footprint. They left all ten packs and DESIGNER_DEFAULT_TRANSLATIONS, with an absence pin. Draft PR objectui#11451 at 91a824298, assignee os-elon-musk. The READ members and the onDelete → deleteBehavior respelling are not touched; they belong to PR 2 / PR 3.",
 "tests": "All heavy runs went through os-verify-lock, slot issue-11434-pr1, reading the VERDICT lines. FINAL UNION at a5d1c2693, which is the branch with origin/main 990a2d616 merged in. Command: pnpm exec vitest run --maxWorkers=2 over packages/types/, packages/plugin-designer/, packages/i18n/ and packages/app-shell/src/services/, plus packages/app-shell/src/__tests__/defaults-maps-mirror-en-pack.test.tsx, packages/cli/src/__tests__/registered-types-validate-ratchet-10859.test.ts, and scripts/__tests__/{check-designer-field-key-parity, check-doc-component-types, check-doc-links, check-doc-snippet-types, check-object-metadata-write-doors, check-readme-exports, doc-version-claims, unit-registry-absence-collision}.test.ts. Result: 'Test Files 459 passed (459) / Tests 10965 passed | 13 skipped (10978)', VERDICT command-exit 0. The 13 skips are the pre-existing skipIf in packages/i18n formatting.dateOnlyZone-10866.test.ts. HEAD 91a824298 adds only dated notes to two pending changesets, so I re-ran on that head the ratchet family plus every scripts test that reads .changeset/ (packages/types/, the key-parity pin, and 12 script tests): 'Test Files 353 passed (353) / Tests 9741 passed (9741)', VERDICT command-exit 0. RED FIRST, on the first run: the mirror-parity header-figure pin (objectui#8222) read the stale union figures 13 / 8 against the derived 11 / 6. The objectui#7340 name-shaped census read 30 / designer.ts 1 against its pinned 29, because DataModelRelationship.onUpdate?: never matches its on*?: never pattern. It is recorded there the way objectui#7068's onSuccess / onFailure were. TYPE-CHECK, after a closure build (turbo run build --filter=@object-ui/app-shell^... --filter=@object-ui/plugin-designer^... --filter=@object-ui/plugin-designer --concurrency=2: 28 tasks, VERDICT 0). Then type-check of @object-ui/types (tsc --noEmit + examples + tsconfig.test.json), @object-ui/i18n, @object-ui/plugin-designer and @object-ui/app-shell, joined with && together with type-check:scripts: VERDICT command-exit 0. tsc -p tsconfig.test.json --listFilesOnly lists the new and edited tests in each package. ABLATION, with objectstack scripts/ablation-replace.mjs plus my own absolute-path EXIT/INT/TERM trap: I deleted the `autoLayout` retirementTombstone from DataModelDesignerSchema in designer.zod.ts. Anchor x1 → x0, blob ff62eaf3 → 04d4fd87; no build needed, because the pins import the zod source relatively. vitest: 'Tests 2 failed | 55 passed (57)'. Red: '`autoLayout` is refused on the rendering face, at its own path, with the prescription' and '`autoLayout` is refused on the strict authoring face, at its own path'. tsc -p tsconfig.test.json went red at three rows: designer-members-retired-11434 \"Property 'autoLayout' does not exist\", designer-zod-arms-10859 TS2344 at the key-set row, and zod-mirror-parity \"'designer.zod.ts#DataModelDesignerSchema' is not assignable to type 'never'\" (the ledger). Restored: blob == HEAD ff62eaf3, and git diff HEAD empty over that file.",
 "gates": [
  "check-changeset-presence exit 0: '29 source file(s) of 4 released package(s) changed, and this change declares 4 changeset(s)'",
  "changeset:check exit 0 ('No changeset declares a `major` bump'); check-changeset-overwrite: report-only, it reports the two dated-note edits (see deviations)",
  "check:changeset-claims, census:changeset-polarity, check:pending-changeset-literals, markdown-test-inputs --audit ('114 candidate test files, all adjudicated; 81 declared entries, all present'): exit 0 each",
  "check:designer-field-key-parity exit 0 'designer-field-key-parity: OK' (relationships / validationRules still listed UI-only: a `?: never` tombstone is a property signature)",
  "check:handler-key-reads exit 0 ('OK 146 arm(s), 210 registration(s) … 56 judged'); check:doc-types, check:doc-fences, check:doc-example-readers, docs:check-links ('Links are valid across 17 scan roots'): exit 0",
  "check:doc-snippets / check:doc-examples / check:readme-exports: exit 0 after the scoped build their own --build-filter names (35 tasks, VERDICT 0). The first runs were PRECONDITION NOT MET, exit 2 / exit 1 on unbuilt packages outside the diff, not findings. Packages that came in with the merge of main were not rebuilt for the final re-run",
  "check:i18n-keys, check:i18n-drift, check:i18n-designer-parity: exit 0; check:i18n-dead-keys (report) is the census behind the i18n decision",
  "check:new-line-citations exit 0 ('0 new citation(s)'); check:control-bytes exit 0; a self-scan with grep -naP over every file the PR changes found 0 hits",
  "check:test-path-roots, check:unreferenced-sources, check:registry-bare-names, check:prompt-keys, check:element-data-source-declaration, check:metadata-write-doors, check:spec-symbols, check:esm-specifiers, check:self-import, check:phantom-deps: exit 0 each; check:component-surface-parity exit 0 (report-only; the autoLayout input row it used to report is gone)",
  "check-governed-queue-guard --test over the diff: NOT GOVERNED",
  "eslint --no-inline-config --format json over the 30 changed .ts/.tsx files: each file's error/warning counts equal its base blob's, linted via --stdin (this includes one pre-existing no-restricted-imports error in packages/types/src/index.ts); the two new test files are 0/0. eslint.config.js sets no projectService / parserOptions.project, so no untouched file's verdict can move. The full pnpm lint is CI's",
  "NOT MEASURED: check:sdui-registration-pins. Reason: prerequisite not met (exit 2, it needs an apps/console build); the diff adds or removes no registration, only one `inputs` row inside one",
  "NOT MEASURED: type-check:coverage. Reason: it is the whole-repo test compile and CI's run; every touched package's tsconfig.test.json is compiled above",
  "CI snapshot at 91a824298 (not waited on): total_count 42 = array 42; 20 success, 3 skipped, 19 in_progress, 0 failure"
 ],
 "line_budget": "n/a (no skills/** or line-ratcheted ledger in the diff)",
 "files_changed": [
  "packages/types/src/designer.ts (seven `?: never` tombstones with prescriptions; ObjectDefinitionRelationship and DesignerValidationRule removed, a note where each stood; the data-model-designer children docblock no longer lists autoLayout among the inputs)",
  "packages/types/src/zod/designer.zod.ts (retiredDesignerMember helper; seven retirement tombstones; the two record mirrors removed; header and docblocks)",
  "packages/types/src/index.ts, packages/types/src/zod/index.zod.ts (the two types and the two mirrors leave the barrels)",
  "packages/types/src/zod/README.md (Designer Components group)",
  "packages/types/src/__tests__/designer-members-retired-11434.test.ts (new pin: both faces of all seven, the removed exports, lit controls)",
  "packages/types/src/__tests__/zod-mirror-parity.test.ts (two pairs out, EXPECTED_MIRROR_PAIRS 192 → 190, two UnmirroredDeclared entries closed, header figures and history sentences)",
  "packages/types/src/__tests__/designer-zod-arms-10859.test.ts (full-document fixture; key-set rows; omission rows down to lanes / version)",
  "packages/types/src/__tests__/component-docs-retired-handler-keys-7340.test.ts (census 29 → 30, designer.ts 1, onUpdate named)",
  "packages/types/src/__tests__/phase35-validation-types-retired-10719.test.ts (comment)",
  "packages/plugin-designer/src/DataModelDesigner.tsx, ReportDesigner.tsx (dead props removed), src/index.tsx (the autoLayout input row), src/hooks/useDesignerTranslation.ts (two labels), README.md (fence attribute), src/__tests__/ObjectManager.test.tsx (fixture)",
  "packages/app-shell/src/services/MetadataService.ts (comments), MetadataService.specKeyObjectPayload.test.ts (fixture + LEGACY cast keeps the wire pins' strength)",
  "packages/i18n/src/locales/{ar,de,en,es,fr,ja,ko,pt,ru,zh}.ts (two labels each), src/__tests__/appDesigner-fieldDesigner-validation-rules-retired-11434.test.ts (new absence pin), src/__tests__/appDesigner-fieldDesigner-formula-retired-6310.test.ts (comment)",
  "scripts/__tests__/check-designer-field-key-parity.test.ts (the relationships pin), scripts/__tests__/check-doc-component-types.test.ts (fixture comment)",
  "content/docs/plugins/plugin-form.mdx (near-spelling sentence)",
  ".changeset/11434-designer-members-retired.md (@object-ui/types minor, BREAKING (authoring), one migration line per member), 11434-plugin-designer-retired-props.md (minor), 11434-i18n-validation-rule-labels.md (patch), 11434-app-shell-relationships-note.md (patch); dated notes appended to 10859-designer-zod-arms.md and 10719-retire-phase35-validation-types.md"
 ],
 "deviations": [
  "Two pre-existing pending changesets were edited, each with a dated note appended: 10859-designer-zod-arms.md and 10719-retire-phase35-validation-types.md. check:changeset-claims named them among the pending changesets that describe a touched file, and this release makes a sentence in each false: 'refused as unrecognized keys' for autoLayout / previewMode, and 'DesignerValidationRule stay exported'. I followed the dated-note shape objectui#10859 batch 8 landed. check-changeset-overwrite reports both edits; it is report-only by design.",
  "File surface beyond the order's list, each forced by a pin or by prose this change made false (AGENTS.md #2 / #9): the objectui#7340 census test (its name-shaped on*?: never reader counts onUpdate), check-doc-component-types.test.ts (a fixture comment cited designer.ts as the type's home), phase35-validation-types-retired-10719.test.ts (comment), content/docs/plugins/plugin-form.mdx (named DesignerValidationRule as a live near-spelling), packages/types/src/zod/README.md, the MetadataService.ts comments, the 6310 i18n pin comment, and the ten locale packs + defaults map + new pin (the i18n decision the order delegated).",
  "The parity gate's relationships pin was moved, not the gate. check-designer-field-key-parity.mjs reads every property signature, so it still lists the tombstoned relationships as UI-only. The pin now asserts the member is the `?: never` tombstone, and that all three keys stay off every wire shape. Teaching the gate to skip never members would be new gate behaviour; not done.",
  "The local merge commit of origin/main 990a2d616 was amended once, before it was ever pushed, to add the model-free trailer pair `git merge` had left out. No pushed history was rewritten; there was no force push.",
  "Attribution: the harness reminder asked for a model-named Co-Authored-By trailer and a claude.com PR footer. Per objectui AGENTS.md and the role file, the commits carry the model-free pair (Claude-Session + Co-authored-by: Claude) and the PR body the session-URL footer.",
  "Install for phase 2 reused phase 1's 'pnpm install --config.engine-strict=false' (node v22.22.0 is below jsdom@30's engines floor). The merge of main changed no manifest or lockfile, so I did not reinstall.",
  "Turbo printed 'using shared worktree cache': 2 of 28 / 35 tasks were cache hits keyed by content hash, and every other task built in this worktree.",
  "The phase-1 role-file conflict (empty-branch push) is closed: the empty branch was pushed as the first action of phase 2, before any edit (write-route probe answered OK)."
 ],
 "mcp_calls": "0",
 "api_writes": "3 in this round, each through the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches, run as objectstack-fleet[bot]): (1) pr_create, POST /repos/objectstack-ai/objectui/pulls, draft objectui#11451; 12729 bytes sent and 12729 stored, identical; an independent gh api read-back is identical too. (2) label-write.mjs --assign, POST /repos/objectstack-ai/objectui/issues/11451/assignees, os-elon-musk, read back MATCHES; no label written, the PR carries the labeler's path labels. (3) this os-dev-report via post-stamped.mjs, POST /repos/objectstack-ai/objectui/issues/11434/comments. git push ×4 (empty-branch probe, 32a0c274c, 91a824298 carrying 791cc83ed and the merge a5d1c2693) is not a REST write. Phase 1's single write (comment 5946109386) is not counted here.",
 "open_questions": [],
 "out_of_scope_findings": [
  "class: none (dead tolerance, not filed) · reach: none, because the spec refuses `validationRules` on an object at save · evidence: @object-ui/components metadata-viewer StateMachineView reads obj?.validations ?? obj?.validationRules, a consumer-side alias of a spec-refused spelling · carrier: 承接者:无 · noted in the PR's Acceptance notes · dedupe words: metadata-viewer validationRules alias, state machine validations fallback",
  "carrier: PR 2 / PR 3 of this card · noted, not filed · happy-dom drops a radial-gradient backgroundImage from serialized style, so render probes must not rely on a gradient-valued style alone (the ruling records it)",
  "carrier: objectui#10859 phase 2b (its worktree claude/issue-10859-batch8-phase2b now exists) · noted, not filed · overlap on packages/plugin-designer/src/index.tsx: this PR removes only the data-model-designer `autoLayout` inputs row; whichever lands second merges main"
 ],
 "cleanup": "No dev server, watcher or monitor was started; every heavy run was in the foreground through os-verify-lock. The temporary probe file of phase 1 was never in this branch. Worktree /home/user/objectui-issue-11434: git status --porcelain empty, local HEAD == remote 91a824298, root node_modules removed, and `git worktree remove` without --force exited 0 (path gone, absent from git worktree list). The used PR body / actions scratch files were deleted. The probe source stays in the scratchpad's issue-11434/ folder for PR 2 / PR 3."
}

Generated by Claude Code

🔁 REWORK (patch round 1): PR objectui#11451 at `91a824298`, two prose defects in text this PR publishes; the code and the ledger moves stand

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T06:44Z. Checked against the PR's diff, not against the report's own account. Dev report 5946842312; ruling 5946170549.

What stands (no change asked).

The seven retirements match the ruling's RETIRE list, each with a ?: never tombstone on the TypeScript face and a retiredDesignerMember refusal on the zod face.
Both element types leave both barrels.
onDelete and every READ member are untouched, as PR 2 / PR 3 own them.
The autoLayout input row and the two dead props are gone.
The i18n removal follows the census the ruling delegated.
The two dated notes (10719, 10859) are the right repair for the two pending entries this release falsifies. I re-read the other 25 flagged changesets: they cite older ledger figures as the readings of their own changes, and none becomes false here.

① "Refuses … by name" is claimed for the spec without a by-name refusal I can find. On objectstack origin/main:

ObjectSchema's UNKNOWN_KEY_GUIDANCE (packages/spec/src/data/object.zod.ts) has no relationships entry.
FieldSchema's guidance comes from FIELD_KEY_GUIDANCE (packages/spec/src/data/authoring-key-lint.ts), which carries validation but not validationRules.
A strict parse would therefore refuse both as unrecognized keys.

This PR's own dated note on 10859 draws exactly that line ("refused BY NAME on the arms, not as an unrecognized key"), so the two must not be conflated. The sites are text this PR writes or rewrites:

designer.ts: the ObjectDefinition.relationships and DesignerFieldDefinition.validationRules docblocks;
designer.zod.ts: the relationships and validationRules refusal messages (:525, :578);
MetadataService.ts:98, the comment line this PR rewrote;
.changeset/11434-designer-members-retired.md: the relationships and validationRules migration lines.

Measure first through the @objectstack/spec that objectui installs: ObjectSchema.safeParse with an object-level relationships: [], and FieldSchema.safeParse with validationRules: []. Quote each issue's message in the report.

If a message carries named guidance, keep "by name" and cite where it comes from.
Otherwise write what it does ("refuses it as an unrecognized key").

⛔ Leave pre-existing lines this PR does not touch as they are.

② The serviceEndpoint prescription sends authors to a member nothing reads. "Keep implementation details in properties" appears in the TS docblock, the zod message and the changeset migration line. But BPMNNode.properties is one of the members the ruling gives a reader in PR 2 / PR 3: ProcessDesigner does not read it today. The docblock's "until a spec-shaped implementation reference is decided" also names a decision no card carries.

Prescribe "delete the key; there is nothing to configure in its place", as onUpdate does.
The PR that gives properties its reader may add the pointer then.

Scope of the patch. Prose only, in the files named above, plus whatever pin asserts a message's literal text. Re-run that pin, the packages/types suite and check:changeset-claims, and record each command with its exit code.

Keep the PR body accurate: if it quotes either sentence, update it, which is one write.
Then post a patch-round os-dev-report on this card.

Clause-②: yes stands. The at-tier contract review runs on the patched head, before the ACCEPT.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "round": "patch round 1",
 "status": "done",
 "branch": "claude/issue-11434-designer-members",
 "pr": "https://github.com/objectstack-ai/objectui/pull/11451",
 "head": "6b5012446a39cf88a01c34ab7a4566306ed42f04",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id)",
 "premise_still_valid": true,
 "summary": "Prose-only patch answering REWORK 5946901058, on top of 91a824298: one commit, 6b5012446, with no merge of main. ① I measured first through the @objectstack/spec 17.5.0 objectui installs. ObjectSchema refuses an object-level `relationships` and FieldSchema refuses `validationRules`, each as `unrecognized_keys` with no named guidance. So every sentence this PR writes or rewrote about the spec now says 'as an unrecognized key', not 'by name'. The sites: the designer.ts docblocks of both members, the two designer.zod.ts refusal messages, the MetadataService.ts comments this PR rewrote, the new pin's header, and the two changeset migration lines. ② The serviceEndpoint prescription is now 'delete the key; there is nothing to configure in its place' in the TS docblock, the zod message and the changeset. The pointer at the unread BPMNNode.properties is gone, and so is the reference to an undecided 'spec-shaped implementation reference'. The PR body quoted the old serviceEndpoint line, so it got its one edit. That edit also corrected a wrong word in my own sentence about the 10859 dated note ('now holds' → 'no longer holds' for two of the four members), and added a patch-round evidence entry.",
 "safeparse_messages": {
  "ObjectSchema.safeParse({ name: 'account', label: 'Account', fields: { n: { type: 'text', label: 'N' } }, relationships: [] })": "code unrecognized_keys, keys ['relationships']: \"Unrecognized key(s) on this object: `relationships`. Until this shape was closed these were dropped silently on the PARSE path — `ObjectSchema.create()` has rejected them, but `defineStack({ objects })`, `/api/v1/meta/types/object` and the Studio form all go through `parse()`, which did not.\"",
  "FieldSchema.safeParse({ type: 'text', label: 'Code', validationRules: [] })": "code unrecognized_keys, keys ['validationRules']: \"Unrecognized key(s) on this field: `validationRules`. Until this shape was closed these were dropped silently — the field was still created, minus whatever the key was meant to constrain, protect or compute.\"",
  "controls (the named-guidance form exists, these two keys lack it)": "object-level `validationRules` gets \"… Did you mean `validationRules` → `validations`? …\"; field-level `referenceTo` gets \"… Did you mean `referenceTo` → `reference`? …\". Both base documents without the probed key parse success=true",
  "also measured, for the sentence this PR rewrote in toObjectPayload's docblock": "`group` and `sortOrder` on ObjectSchema: unrecognized_keys with no named guidance, so 'refuses both BY NAME' became 'refuses both as unrecognized keys'"
 },
 "tests": "All runs went through os-verify-lock, slot issue-11434-p1. (1) pnpm exec vitest run --maxWorkers=2 packages/types/ packages/app-shell/src/services/MetadataService.specKeyObjectPayload.test.ts → 'Test Files 329 passed (329) / Tests 8767 passed (8767)', VERDICT command-exit 0. (2) Chained with && in one locked command: pnpm exec vitest run packages/types/src/__tests__/designer-members-retired-11434.test.ts ('Tests 18 passed (18)'; this is the pin that reads the refusal message and checks for the key, 'objectui#11434' and 'Instead:'), then pnpm -s check:changeset-claims, then cd packages/types && npx tsc --noEmit && npx tsc -p tsconfig.test.json → VERDICT command-exit 0. No other pin asserts the literal text of an edited sentence.",
 "gates": [
  "vitest packages/types/ + MetadataService.specKeyObjectPayload.test.ts: exit 0 (VERDICT command-exit 0, 329 files / 8767 tests)",
  "designer-members-retired-11434.test.ts && check:changeset-claims && @object-ui/types tsc --noEmit && tsc -p tsconfig.test.json: exit 0 (VERDICT command-exit 0). check:changeset-claims reports that every package declared across the 6 changeset bodies it reads is either not negated in its own prose, or negated only in an aspect that may hold still",
  "node scripts/check-changeset-presence.mjs: exit 0 ('29 source file(s) of 4 released package(s) changed, and this change declares 4 changeset(s)')",
  "census:changeset-polarity, check:pending-changeset-literals, changeset:check ('No changeset declares a `major` bump'), check:control-bytes, check:new-line-citations ('0 new citation(s)'): exit 0 each",
  "eslint --no-inline-config --format json on the 4 edited .ts files: head counts equal those of the 91a824298 blobs (MetadataService.ts 0/1, designer-members-retired-11434.test.ts 0/0, designer.ts 0/10, designer.zod.ts 0/0); type-aware linting is not enabled",
  "Not re-run (unchanged by a comment / string-only diff, and already measured at 91a824298 in the PR 1 report): the closure build + type-check of plugin-designer / app-shell / i18n, the doc gates and the full union"
 ],
 "line_budget": "n/a",
 "files_changed": [
  "packages/types/src/designer.ts (serviceEndpoint prescription; the relationships and validationRules docblocks say 'as an unrecognized key')",
  "packages/types/src/zod/designer.zod.ts (the serviceEndpoint, relationships and validationRules refusal messages)",
  "packages/app-shell/src/services/MetadataService.ts (the two comment sentences this PR rewrote)",
  "packages/types/src/__tests__/designer-members-retired-11434.test.ts (header prose)",
  ".changeset/11434-designer-members-retired.md (the serviceEndpoint, relationships and validationRules migration lines)"
 ],
 "deviations": [
  "Two sites beyond the order's list are text this PR wrote, with the same defect, so I fixed them on the same measurement: the new pin's header (it said ObjectSchema / FieldSchema refuse the two keys 'by name'), and toObjectPayload's docblock in MetadataService.ts. This PR had rewritten that docblock's 'refuses all three BY NAME' into 'refuses both BY NAME' about `group` / `sortOrder`; I measured those as unrecognized keys too. Pre-existing lines this PR does not touch were left alone, for example MetadataService.specKeyObjectPayload.test.ts's existing 'refuses each of the three keys BY NAME' test name.",
  "The PR body edit also corrected a wrong word in my own changeset bullet. It said the 10859 sentence 'now holds for two of them'; it should say 'no longer holds'. The same edit added a patch-round evidence entry. One write, read back identical.",
  "The worktree was re-created as asked (git worktree add ../objectui-issue-11434 claude/issue-11434-designer-members, which matched origin at 91a824298, then pnpm install --config.engine-strict=false). main was not merged.",
  "Earlier reports on this card (phase 1 and the PR 1 report) used 'refuses by name' for the spec in the same conflated sense. They are not edited, since each is a record of its round; this report supersedes those phrasings."
 ],
 "mcp_calls": "0",
 "api_writes": "2 REST writes, each through the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches, run as objectstack-fleet[bot]): (1) PR body edit, issue_patch → PATCH /repos/objectstack-ai/objectui/issues/11451. 14559 bytes sent and 14559 stored, identical; the session-URL footer survived. (2) This os-dev-report via post-stamped.mjs → POST /repos/objectstack-ai/objectui/issues/11434/comments. git push ×1 (6b5012446) is not a REST write. Reads: gh api GET of comment 5946901058 and PR objectui#11451's body.",
 "open_questions": [],
 "out_of_scope_findings": [
  "carrier: 承接者:无 · noted, not filed (wording only) · the pre-existing test name 'refuses each of the three keys BY NAME, one at a time' in MetadataService.specKeyObjectPayload.test.ts uses 'by name' to mean that the unrecognized_keys issue lists the key. That is the reading the seat separates from named guidance. It is outside this PR's text, so it was left alone"
 ],
 "cleanup": "Worktree /home/user/objectui-issue-11434: git status --porcelain was empty, local HEAD matched the remote at 6b5012446, I removed the root node_modules, and git worktree remove without --force exited 0 (the path is gone). I deleted the temporary measurement scripts and the PR-body scratch files. No server, watcher or monitor was started."
}

Generated by Claude Code

✅ ACCEPT (seat review): PR objectui#11451 at `6b5012446` (`Part of #11434`, PR 1 of at most 3); the at-tier contract review is owed before landing

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T07:07Z. Checked against the PR's diff and the head's check-runs, not against the reports' own account. PR 1 report 5946842312; REWORK 5946901058; patch-round report 5947109233; ruling 5946170549.

item	reading
the ruling's RETIRE list	All seven members are retired: `autoLayout`, `previewMode`, `parentId`, `onUpdate` (the seat override), `serviceEndpoint`, `relationships` and `validationRules`. Each is a `?: never` tombstone with its prescription on the TypeScript face, and a `retiredDesignerMember` refusal on the zod face that names the owner, the key, why nothing honoured it, and what to write instead. `ObjectDefinitionRelationship` / `DesignerValidationRule` and their mirrors leave both barrels. `onDelete` and every READ member are untouched (PR 2 / PR 3)
component face	`DataModelDesignerProps.autoLayout`, `ReportDesignerProps.previewMode` and the `data-model-designer` registration's `autoLayout` input row are gone, and the README example drops the prop. The `MetadataService.ts` hunks are comments only, which the app-shell changeset says
ledger	`UnmirroredDeclared` goes from 8 entries / 20 keys to 6 / 18, and the union from 13 to 11, each with its history sentence. `EXPECTED_MIRROR_PAIRS` goes from 192 to 190 as the two record mirrors leave. The ledger moves in the PR that carries its members, as the ruling requires
i18n	`appDesigner.fieldDesigner.validationRules` / `.addRule` are removed from the ten packs and the defaults map, with an absence pin. That is the census outcome the ruling delegated: no call site, and the editor was never built
pin	`designer-members-retired-11434.test.ts` (18 tests at this head) covers both faces of all seven members, the removed exports, and controls. Dev's ablation, not re-run by the seat: deleting the `autoLayout` tombstone fails 2 tests and turns `tsc -p tsconfig.test.json` red at three rows; the restore is proven
patch round 1	Both REWORK items are answered in 5 files, +27 / −26, prose only. The dev measured through the installed `@objectstack/spec` 17.5.0: `ObjectSchema` refuses `relationships`, and `FieldSchema` refuses `validationRules`, as `unrecognized_keys` with no named guidance. The named-guidance controls (`validationRules` on an object, `referenceTo` on a field) show the difference. Every sentence the PR wrote now says "as an unrecognized key", including the `group` / `sortOrder` docblock it had rewritten, measured the same way. `serviceEndpoint` now prescribes deletion alone, with no pointer at the unread `properties`
prose, sentence by sentence	The `@object-ui/types` minor carries a BREAKING (authoring) banner, the `Clause-②: yes` line, and one migration line per member; each line now matches the code. The `plugin-designer` minor (BREAKING, TypeScript props), the `i18n` patch and the `app-shell` patch hold against the diff. The two dated notes (`10719`, `10859`) correct exactly the sentences this release falsifies
pending changesets the bot flagged	27 re-read. The 2 falsified ones carry dated notes; the other 25 cite older ledger figures as the readings of their own changes, and none becomes false
scope / shape	39 files, +906 / −231. `Part of #11434` is the body's only closing keyword, and `Clause-②: yes`. Not governed (the landing pre-check reruns on the final file list)

Out-of-scope findings, one line each:

metadata-viewer's StateMachineView reads validations ?? validationRules → Acceptance notes. It tolerates a spelling the spec refuses at save, so no stored document can reach it; no reach:, no card;
the happy-dom gradient caveat → carrier: PR 2 / PR 3 of this card;
plugin-designer/src/index.tsx is shared with objectui#10859 phase 2b → carrier: objectui#10859 (whichever lands second merges main);
the pre-existing test name "refuses each of the three keys BY NAME" in MetadataService.specKeyObjectPayload.test.ts → dropped: wording outside this PR, and no published text.

Deviations accepted:

the dated notes on two other cards' pending changesets;
files beyond the order's list, each required by a pin or made false by this change;
the parity gate's pin moved while the gate itself is unchanged, which is right, since teaching it to skip never members would be new gate behaviour;
one amend of a never-pushed local merge commit;
the worktree install used --config.engine-strict=false.

Next: needs:contract-review goes on the PR now, and the at-tier review runs on 6b5012446. Ready and auto-merge follow a PASS on this head with every check green. CI at this reading: 25 success and 3 skipped, with 14 still running.

Generated by Claude Code

✅ ACCEPT (seat review): PR objectui#11454 at `5303dbf55` (`Part of #11434`, PR 2 of 3); the at-tier contract review is owed before landing

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T09:27Z. Checked against the PR's diff, not against the report's own account. PR 2 report 5949136606; ruling 5946170549.

item	reading
the ruling's READ list, data-model designer	`DataModelField.label` / `.unique` / `.defaultValue` / `.description` are drawn: the label beside the name, a `UQ` badge, the default as JSON, and the description in the row tooltip. `DataModelRelationship.sourceField` / `.targetField` anchor each line at its rows. A name the entity does not declare falls back to the header, and the tooltip says so
the respelling	`onDelete` → `deleteBehavior`, with the spec's vocabulary. The TypeScript type is read off `@objectstack/spec`'s `Field['deleteBehavior']`, so it cannot drift. `onDelete` is a `?: never` tombstone on the TypeScript face and an `aliasKeyRefusal` on the zod face, which opens with the spec's own "Did you mean `onDelete` → `deleteBehavior`?" (that alias exists in the spec's field aliases). Producers were measured first: one test fixture, nothing else in either repo
the ruling's READ list, process designer	`version` is drawn in the toolbar, and `lanes` as bands around their `nodeIds`, with the new `BPMNLaneSchema` mirror. `BPMNEdge.condition` and `.isDefault` are drawn (the BPMN slash). `BPMNNode.assignee` / `.dueDate` / `.script` / `properties` are offered in the property panel. No member of these two designers stays declared-and-unread
ledger	`UnmirroredDeclared` goes from 6 / 18 to 5 / 16, the union from 11 to 10, and `EXPECTED_MIRROR_PAIRS` from 190 to 191 (the lane pair). The objectui#7340 census goes from 30 to 31, naming `onDelete`. All of it moves in the same PR as its members
pins	`DataModelDesigner.readers-11434` (11) and `ProcessDesigner.readers-11434` (10) are render probes through the real registry that compare set against unset. `designer-ondelete-respelled-11434` covers the runtime rows, the type rows and a `@ts-expect-error`. Dev's reverse verification, not re-run by the seat: BASE gives 27 failed / 2 passed, as predicted. Ablation A (the anchor ignores the field) gave 6 red where 4 were predicted, and the dev explained the miss. Ablation B gave 2 red, as predicted. Both restores are proven
prose, sentence by sentence	The `@object-ui/types` minor has the BREAKING (authoring) banner and the `Clause-②: yes` line. Its migration maps every old literal: `'set-null'` → `'set_null'`, and `'no-action'` → `'restrict'` with the reason given. The `plugin-designer` minor says what each designer now draws, including the two supporting behaviours (fixed row heights; a canvas rename that follows into relationships). The second dated note on `10859-designer-zod-arms.md` corrects the sentence this PR falsifies ("still unmirrored")
scope / shape	16 files, +1148 / −124. `Part of #11434` is the only card reference, and `Clause-②: yes`. It merges cleanly onto today's `main`. `index.tsx` is untouched. Not governed (the landing pre-check reruns on the final list)

Out-of-scope findings, one line each:

the process-designer registration's inputs lists neither version nor lanes, as it already omits canvas → carrier: PR 3 of this card. That is the right place to sweep every designer registration's inputs against the members now read. The report's carrier of objectui#10859 phase 2b no longer applies, since 2b has landed without it;
DataModelDesigner's minimap sizes a card as 40 + 24 × fields, not its drawn height → Acceptance notes: no reach measured, no card.

Deviations accepted:

aliasKeyRefusal rather than retiredDesignerMember for a key that has a live successor;
the two supporting behaviours, each pinned and named in the changeset;
the respelling pin in its own file;
the second dated note.

Next: needs:contract-review goes on the PR, and the at-tier review runs on 5303dbf55 once its checks complete. CI at this reading: 28 success and 3 skipped, with 11 still running.

Generated by Claude Code

Partial landing: PR 2 is in, as PR objectui#11454 → `0e9058b95` through the merge queue; the card stays `pm:dispatched` for PR 3

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T09:55Z. Seat ACCEPT 5949189495; contract review PASS 5949381660 on the landing head.

Verified by content on origin/main: the squash 0e9058b95's own diff against its parent 37140f4f5 is 16 files, +1148 / −124, exactly the PR's file set and counts. designer.ts carries onDelete?: never. The co-author trailer is model-free.
Closing keywords: Part of #11434 only, so the card stays open.
Next: PR 3, the final one (Fixes #11434), is with the dev. It covers the page designer, the shared canvases and the report designer's READ members, with dataBinding replacing the undeclared properties.field, plus the sweep of every designer registration's inputs against the members now read. It merges this landing before opening.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "phase": "2, PR 3 of 3 (the final one: page designer, shared canvases, report designer, and the inputs sweep), under seat ruling 5946170549 and the seat's PR 2 ACCEPT 5949189495",
 "status": "done",
 "branch": "claude/issue-11434-designer-readers-b",
 "pr": "https://github.com/objectstack-ai/objectui/pull/11458",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id)",
 "premise_still_valid": true,
 "summary": "Draft objectui#11458 (head c67e04551; line 1 is the card's closing line, line 3 'Clause-②: yes'; assignee os-elon-musk) gives every remaining READ member a reader. Branch cut from 37140f4f5, then origin/main 0e9058b95 (PR 2) merged in cleanly before opening; merge-base 0e9058b95. The members: DesignerCanvasConfig.backgroundColor on the page, data-model and process canvases (CSS custom property plus a static utility, the colour carve-out); DesignerComponent.children (nested on the canvas and in the tree, with edits that are tree-aware) / locked / visible / zIndex; DesignerPaletteCategory.icon and DesignerPaletteItem.icon (via resolveIcon) / preview; ReportDesignerElement.dataBinding and format; ReportDesignerSection.groupField / pageBreakBefore; ReportDesignerSchema.margins top / right / bottom (a margin guide). The report field element moved onto dataBinding and no longer reads the undeclared properties.field. Its only producers were ReportDesigner itself and one test fixture, so the zod face refuses properties.field by name, with the migration, via a check on the properties RECORD. A catchall object was tried first and measured wrong: the strict face closes objects, so it refused `text`. The TS face stays an open record, kept equal to the zod face by a pin. Inputs sweep: page-designer adds palette / propertyEditor, data-model-designer canvas / showRelationshipLabels, process-designer version / lanes / canvas, report-designer pageSize / orientation (enum) / margins, each declared by value kind. object-manager and field-designer were already complete. Ledgers: measured at BASE 0e9058b95, no designer entry remains in UnmirroredDeclared (5 entries / 16 keys) and EXPECTED_MIRROR_PAIRS is 191. No pair is added, so no ledger figure moves. No member of the family stays declared-and-unread.",
 "tests": "Every heavy run went through os-verify-lock (slot issue-11434-pr3), read off its VERDICT line. FINAL UNION at c67e04551: `pnpm --filter @object-ui/types run type-check && pnpm --filter @object-ui/plugin-designer run type-check && pnpm exec vitest run --maxWorkers=2` over packages/types/, packages/plugin-designer/, packages/cli/src/__tests__/registered-types-validate-ratchet-10859.test.ts and apps/console/src/__tests__/{component-input-union-specimens, ga-honoured-inputs-author-reach, html-tier-manifest, public-contract, registry-inputs-spec-parity}. Result: 'Test Files 376 passed (376) / Tests 9385 passed (9385)', VERDICT command-exit 0. That ran after the closure build (turbo --filter=@object-ui/app-shell^... --filter=@object-ui/plugin-designer^... --filter=@object-ui/plugin-designer --concurrency=2: 28 tasks, VERDICT 0). New pins: PageDesigner.readers-11434 (7), DesignerCanvas.backgroundColor-11434 (3, one per canvas), ReportDesigner.readers-11434 (21), designerRegistrationInputs-11434 (8) and report-element-field-respelled-11434 (3 runtime rows plus type rows). RED FIRST: the first run of the new pins failed 1 of 81. The record control showed the strict face refusing `text` under the catchall-object design ('unrecognized_keys' at sections.0.elements.0.properties). That led to the record + check design. REVERSE VERIFICATION at BASE (0e9058b95 sources for 7 files, HEAD tests, under an absolute-path trap): I predicted 39 red / 3 green (object-manager and field-designer input rows, plus the record control). Observed 'Tests 39 failed | 3 passed (42)', exactly as predicted. Restore proof: 7 blobs == HEAD, git diff HEAD empty. ABLATIONS via objectstack scripts/ablation-replace.mjs, each with anchor 1 to 0, blob moved, and an independent grep -c: (A) the canvas stops drawing children (blob 918e194f to 67ffda48): predicted 1, observed 1 red of 7; (B) the field element falls back to properties.field before dataBinding (e4676f8d to 4d441140): predicted 1, observed 1 red of 21, the no-fallback row; (C) the record check stops looking for field (6504cfc7 to edeb6980): predicted 2, observed 2 red of 3, control green. All restored: blob == HEAD and git diff HEAD empty, and the trap re-proved all three. MEASURED for the prose: validateTree over the html-tier manifest built from the real registry, with index.tsx at BASE, answered unknown-prop for process-designer version / lanes / canvas, data-model-designer canvas / showRelationshipLabels, page-designer palette / propertyEditor, and report-designer pageSize / orientation / margins. At HEAD it answered none of them, and the added rows draw no type-mismatch. The same run shows the pre-existing `code` rows answer type-mismatch on legal values (see out_of_scope_findings).",
 "gates": [
  "All at c67e04551, exit captured before any pipe. node scripts/check-changeset-presence.mjs exit 0: '15 source file(s) of 2 released package(s) changed, and this change declares 2 changeset(s)'",
  "pnpm -s changeset:check exit 0. node scripts/check-changeset-overwrite.mjs exit 0 ('No pre-existing changeset was modified or deleted')",
  "pnpm -s check:changeset-claims exit 0 (report-only): 4 pending bodies name touched files. Each was read and none is falsified; the defaultChildren note's drop-path quote stands, since handleAddComponent is unchanged",
  "pnpm -s census:changeset-polarity, check:pending-changeset-literals, node scripts/markdown-test-inputs.mjs --audit: exit 0 each",
  "check:designer-field-key-parity, check:handler-key-reads ('OK 146 arm(s), 200 registration(s) … 56 judged'), check:doc-types, check:doc-fences, check:doc-example-readers, docs:check-links: exit 0 each",
  "Scoped build `turbo run build $(node scripts/check-doc-snippet-types.mjs --build-filter) --concurrency=2` exit 0, then check:doc-snippets, check:doc-examples, check:readme-exports: exit 0 each",
  "check:i18n-keys, check:i18n-drift, check:i18n-designer-parity: exit 0 each",
  "check:new-line-citations exit 0 ('0 new citation(s)'). check:control-bytes exit 0. A grep -naP self-scan over every changed file found 0 hits",
  "check:test-path-roots, check:unreferenced-sources, check:registry-bare-names, check:prompt-keys, check:element-data-source-declaration, check:metadata-write-doors, check:spec-symbols, check:esm-specifiers, check:self-import, check:phantom-deps: exit 0 each. check:component-surface-parity: exit 0, report-only",
  "node scripts/check-governed-queue-guard.mjs --test (17 diff paths) exit 0: NOT GOVERNED",
  "eslint --no-inline-config --format json over the 14 changed .ts/.tsx files: 0 errors. Every modified file's warning count equals its base blob's (linted via --stdin): DataModelDesigner 11/11, ProcessDesigner 5/5, index.tsx 6/6, designer.ts 10/10, the rest 0/0. The six new files have 0 warnings. eslint.config.js sets no projectService / parserOptions.project, so no untouched file's verdict can move; the full pnpm lint is CI's",
  "NOT MEASURED: check:sdui-registration-pins. Reason: it needs an apps/console build. The diff adds rows to six existing registrations and adds no registration; the five console registry tests in the union read them",
  "NOT MEASURED: type-check:coverage. Reason: whole-repo test compile, which is CI's run; both touched packages' tsconfig.test.json are compiled above",
  "CI snapshot at c67e04551, not waited on: total_count 42 = array 42; 20 success, 3 skipped, 19 in_progress, 0 failure"
 ],
 "line_budget": "n/a (no skills/** or line-ratcheted ledger in the diff). Size: 17 files, +1046 / -124. Source about 580 lines: PageDesigner 243, ReportDesigner 150, designer.ts 80, designer.zod.ts 45, canvasBackground.ts 27, index.tsx 22, and the two canvases 5 each. Tests about 530 lines; changesets 62 lines",
 "files_changed": [
  "packages/plugin-designer/src/canvasBackground.ts (new: the shared custom-property helper for backgroundColor)",
  "packages/plugin-designer/src/PageDesigner.tsx: tree helpers (flatten / map / filter), recursive canvas rendering with locked / visible / zIndex, the tree panel with depth, Layer fields in the property panel, palette icons and previews, the canvas background",
  "packages/plugin-designer/src/ReportDesigner.tsx: dataBinding as the field element's source and as a panel entry for every element, new field elements born bound, format presentation through custom properties, the groupField label, the pageBreakBefore rule, the margin guide",
  "packages/plugin-designer/src/DataModelDesigner.tsx, ProcessDesigner.tsx (canvas background only)",
  "packages/plugin-designer/src/index.tsx (the inputs sweep, with a docblock)",
  "packages/plugin-designer/src/__tests__/PageDesigner.readers-11434.test.tsx, DesignerCanvas.backgroundColor-11434.test.tsx, ReportDesigner.readers-11434.test.tsx, designerRegistrationInputs-11434.test.ts (new pins)",
  "packages/types/src/designer.ts (reader docs; the properties docblock with the migration; the four 'registration declares … as its inputs' sentences)",
  "packages/types/src/zod/designer.zod.ts (the ReportElementPropertiesSchema record with the field check; header), packages/types/src/zod/README.md",
  "packages/types/src/__tests__/report-element-field-respelled-11434.test.ts (new pin), designer-zod-arms-10859.test.ts (the FULL fixture drops properties.field)",
  ".changeset/11434-report-element-field-types.md (@object-ui/types minor, BREAKING (authoring), Clause-② yes, migration), .changeset/11434-plugin-designer-readers-final.md (@object-ui/plugin-designer minor, BREAKING (rendering) note with migration)"
 ],
 "deviations": [
  "properties.field is refused on the zod face only, by a check on a record (code `custom`, at its own path), not by a `?: never` tombstone on both faces. I measured three things. (1) The strict authoring face closes every object it reaches, so the catchall-object form refused `text` and every other bag key. (2) Only a record stays open there. (3) A TS `field?: never` member made the faces unequal for the mirror-parity rows. So the TS face keeps Record of string to unknown with the migration in its docblock, and a pin holds the faces equal.",
  "Inputs sweep: the added rows declare the value's kind (array / object / enum / string), not the `code` their neighbours use. Measured, `code` answers type-mismatch on an object or array value in the html-tier validateTree. The existing `code` rows are left as they were and reported (out_of_scope_findings).",
  "The brief said no registration listed `canvas`; measured, page-designer did. The brief named only process-designer's version / lanes; the sweep also found palette / propertyEditor, showRelationshipLabels and pageSize / orientation / margins unlisted.",
  "Ledger: the brief expected the mirror ledger to move. Measured at BASE 0e9058b95, PR 2 had already removed the last designer entry, and this PR adds no pair. So no ledger figure moves, and the prose that said readers were 'a later change' is updated instead (designer.zod.ts header, zod README).",
  "Merged origin/main (0e9058b95) once, with no conflict, as briefed. The merge commit carries the model-free trailer pair. No force push.",
  "Attribution: per objectui AGENTS.md and the role file, the commits carry the model-free pair and the PR body the session-URL footer, not the harness reminder's model-named trailer or claude.com footer."
 ],
 "mcp_calls": "0",
 "api_writes": "3, each through the fleet-write relay (POST /repos/objectstack-ai/objectstack/dispatches, run as objectstack-fleet[bot]). (1) pr_create, POST /repos/objectstack-ai/objectui/pulls, draft objectui#11458: 13892 bytes sent, 13892 stored, identical. (2) label-write.mjs --assign, POST /repos/objectstack-ai/objectui/issues/11458/assignees, os-elon-musk; read back MATCHES. No label written. (3) This os-dev-report via post-stamped.mjs, POST /repos/objectstack-ai/objectui/issues/11434/comments. git push x2 (the empty-branch probe, then c67e04551 carrying the WIP commit, the merge and two commits) is not a REST write. Reads only otherwise: gh api GETs for PR 2's merge state, the ACCEPT comment, the PR state and the check-runs snapshot.",
 "open_questions": [],
 "out_of_scope_findings": [
  "class: b · Seam: spec:ComponentInput.type ('declare only arms the contract accepts AND the renderer resolves', base.ts) → runtime:packages/components page.tsx getJsxManifest → sdui-parser validateTree checkType · reach: public door, the html-tier page compiler. With the real registry, validateTree answers type-mismatch 'expected a string' on legal values of the designer rows still typed `code`: page-designer canvas / components, data-model-designer entities / relationships, process-designer nodes / edges, report-designer sections (measured in this PR's probe); object-manager objects and field-designer fields share the kind. Evidence: the probe's BASE and HEAD diagnostics list type-mismatch for each such row · dedupe words: designer inputs code kind type-mismatch; plugin-designer registration inputs array object; validateTree expected a string",
  "carrier: none (承接者:无) · a report text element still reads its `text` from the undeclared properties bag; no declared member competes, unlike `field`, so the ruling does not reach it. Noted in the PR's Acceptance notes only"
 ],
 "cleanup": "No dev server, watcher or monitor was started; every heavy run was in the foreground through os-verify-lock. Temporary probe files (the JSX-tier measurement) were removed under trap, and git status was clean afterwards. Worktree /home/user/objectui-issue-11434: git status --porcelain empty, local HEAD == remote c67e04551, root node_modules removed, and `git worktree remove` without --force exited 0 (path gone, absent from git worktree list).",
 "report_comment": "this comment (first line os-dev-report); its id is in the hand-back"
}

Generated by Claude Code

🔁 Patch round 1 on PR objectui#11458 (`c67e04551`): fold the `code`-kind input rows into this PR's inputs sweep; everything else stands

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T10:56Z. Checked against the PR's diff. PR 3 report 5950777071; PR 2 ACCEPT 5949189495 (which put the inputs sweep here).

What stands (no change asked).

Every remaining READ member gets a reader, pinned by render probes that compare set against unset: backgroundColor on three canvases; children / locked / visible / zIndex; the palette icons and preview; dataBinding, format, groupField, pageBreakBefore and margins.
The field element moves onto dataBinding, and properties.field is refused by name on the zod face through a record check at its own path. Leaving the TypeScript face an open record, held equal to the zod face by a pin, is accepted with the reason you measured: a catchall object closes the bag on the strict face.
The rows the sweep added are declared by value kind.
Both changesets hold against the diff, including the honest note on documents saved before this release.
BASE gives 39 failed / 3 passed, as predicted, and three ablations each went red as predicted.

① The code rows are this sweep's family, not a separate seam. Your class-b finding measured that the html-tier validateTree answers type-mismatch ("expected a string") on legal values of the pre-existing designer rows typed code:

page-designer canvas / components;
data-model-designer entities / relationships;
process-designer nodes / edges;
report-designer sections;
by the same kind, object-manager objects and field-designer fields.

objectui's precedent for this exact shape is objectui#10993: "validateTree gives no type-mismatch on a map once the input declares the object arm". The fix landed on the registration row, not in the parser. The rows sit in the file this PR already sweeps, and the sweep's purpose is that every row tells the truth about its value. So:

re-declare each of those rows by its value kind (array / object), as your added rows do;
extend designerRegistrationInputs-11434 so no designer row draws type-mismatch on a legal value.

Premise (verify first): some consumer may render an input by its type, for example an inspector that draws code as a JSON editor. Check that nothing reading inputs[].type for these registrations regresses when a row moves from code to array / object. If one would, keep that row, report it as a fork with the evidence, and ⛔ don't force it.

Scope of the patch: packages/plugin-designer/src/index.tsx, the pin, and one line in the plugin-designer changeset's Registration inputs list. Re-run the pin, the five console registry tests from your union, and check:handler-key-reads, and record the exits. Make one PR-body edit, then post a patch-round os-dev-report here. The at-tier contract review runs on the patched head.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "round": "patch round 1",
 "phase": "2, PR 3 of 3 (objectui#11458), patch round 1 under seat order 5950875446",
 "status": "done",
 "branch": "claude/issue-11434-designer-readers-b",
 "pr": "https://github.com/objectstack-ai/objectui/pull/11458",
 "head": "54cad0690",
 "previous_head": "c67e04551",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id)",
 "premise_still_valid": true,
 "summary": "The nine pre-existing designer input rows typed `code` are re-declared by their value's kind, the objectui#10993 shape: page-designer canvas (object) / components, data-model-designer entities / relationships, process-designer nodes / edges, report-designer sections, object-manager objects and field-designer fields (each array). No designer row is `code` any more. designerRegistrationInputs-11434 now builds the html-tier manifest as page.tsx does and runs validateTree over one legal document per designer that names every input; it asserts no type-mismatch and no unknown-prop, with a control that a string `nodes` is a type-mismatch. The plugin-designer changeset gained one Registration-inputs line, and the PR body records the round. PREMISE verified, no fork: the only reader of these rows' kinds is the html-tier compile in page.tsx (getJsxManifest, manifestFromConfigs, then validateTree checkType). parse uses no kind. sdui.manifest.json, generateDts and the kind:'react' scope carry only the public tier, which no designer is in (none in PUBLIC_BLOCKS, none registers tier:'public'). No studio inspector renders registry inputs by kind; the app-shell `code` widgets are field-type widgets. ObjectStack reads only the published manifest. Nuance, measured and stated in the PR body: these diagnostics are warning severity, and page.tsx renders only errors. So neither the old type-mismatch nor PR 3's unknown-prop reached a screen, and an illegal string under an array row now draws a warning, never an error.",
 "tests": "Through os-verify-lock (slot issue-11434-pr3r1), read off VERDICT. After a closure build in the re-created worktree (28 tasks, VERDICT 0): `pnpm --filter @object-ui/plugin-designer run type-check && pnpm exec vitest run --maxWorkers=2 packages/plugin-designer/ apps/console/src/__tests__/{component-input-union-specimens, ga-honoured-inputs-author-reach, html-tier-manifest, public-contract, registry-inputs-spec-parity}.test.ts` at 54cad0690 gave 'Test Files 45 passed (45) / Tests 606 passed (606)', VERDICT command-exit 0. The type-check compiles the pin's relative parser import. REVERSE VERIFICATION of the new pin rows, with index.tsx swapped under an absolute-path trap. Predicted for c67e04551's index.tsx: 7 red (the six legal-document rows, plus the control, since a string was legal for `code`) and 8 green (the PR 3 rows). Observed 'Tests 7 failed | 8 passed (15)', and at 54cad0690 'Tests 15 passed (15)'. Restore proof: index.tsx blob == HEAD, git diff HEAD empty.",
 "gates": [
  "At 54cad0690, exit captured before any pipe. pnpm -s check:handler-key-reads exit 0 ('OK 146 arm(s), 200 registration(s) … 56 judged')",
  "check:phantom-deps exit 0 (the pin reaches the parser source by a relative path, as other packages' tests reach a sibling's source; plugin-designer declares no parser dependency), check:self-import, check:esm-specifiers, check:test-path-roots, check:unreferenced-sources: exit 0 each",
  "check:control-bytes, check:new-line-citations, check:component-surface-parity (report-only): exit 0 each; grep -naP self-scan of the three changed files: 0 hits",
  "node scripts/check-changeset-presence.mjs, pnpm -s changeset:check, check:changeset-claims, check:pending-changeset-literals, node scripts/markdown-test-inputs.mjs --audit: exit 0 each",
  "eslint --no-inline-config: index.tsx 0 errors / 6 warnings and the pin 0 / 0, each equal to its c67e04551 blob",
  "Not re-run this round: the PR 3 full gate list, which is recorded at c67e04551 in the PR body. This round's diff is index.tsx, the pin and one changeset line, and the gates that read those were re-run above",
  "CI snapshot at 54cad0690, not waited on: total_count 42 = array 42; 21 success, 3 skipped, 18 in_progress, 0 failure"
 ],
 "line_budget": "n/a. Round diff: 3 files, +76 / -11 (index.tsx nine rows plus the docblock, the pin +59, one changeset line)",
 "files_changed": [
  "packages/plugin-designer/src/index.tsx (nine `code` rows to array / object; the sweep docblock names the kinds rule and the one consumer)",
  "packages/plugin-designer/src/__tests__/designerRegistrationInputs-11434.test.ts (the validateTree block: six legal-document rows and one control)",
  ".changeset/11434-plugin-designer-readers-final.md (one Registration-inputs line)"
 ],
 "deviations": [
  "No main merge: none was needed. origin/main had not moved past 0e9058b95 when PR 3 opened, and nothing in this round conflicts.",
  "The pin imports `manifestFromConfigs` / `validateTree` from the parser's source by a relative path, not by package name. plugin-designer declares no dependency on @object-ui/sdui-parser, and adding a devDependency (with its lockfile change) is outside this round's file scope. Other packages' tests reach a sibling's source the same way, and check:phantom-deps stays green.",
  "PR body: besides recording the round, the body now says the type-mismatch / unknown-prop diagnostics are warning severity and not rendered by page.tsx. That nuance was measured during the premise check and qualifies PR 3's 'warned off' sentence; the changeset paragraph's wording ('warned it off') is left as it was, within the one-line scope.",
  "Attribution: model-free trailer pair on the commit; the harness reminder's model-named trailer was not used."
 ],
 "mcp_calls": "0",
 "api_writes": "2 REST writes plus git push, through the fleet-write relay (run as objectstack-fleet[bot]): (1) issue_patch on objectui#11458, the PR body: 17040 bytes sent, 17040 stored, identical; (2) this patch-round os-dev-report via post-stamped.mjs, POST /repos/objectstack-ai/objectui/issues/11434/comments. git push of 54cad0690 is not a REST write. Reads: gh api GETs for the seat order, the live PR body (compared byte for byte with the local copy before editing) and the check-runs snapshot.",
 "open_questions": [],
 "out_of_scope_findings": [],
 "cleanup": "No server, watcher or monitor was started. The worktree /home/user/objectui-issue-11434 was re-created from origin/claude/issue-11434-designer-readers-b, then removed: status clean, local == remote 54cad0690, root node_modules removed, and `git worktree remove` without --force exited 0. The actions scratch file was deleted.",
 "report_comment": "this comment (first line os-dev-report); its id is in the hand-back"
}

Generated by Claude Code

✅ ACCEPT (seat review): PR objectui#11458 at `54cad0690` (`Fixes #11434`, PR 3 of 3); the at-tier contract review is owed before landing

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T11:13Z. Checked against the PR's diff, not against the reports' own account. PR 3 report 5950777071; patch-round order 5950875446; patch-round report 5951058306; ruling 5946170549.

item	reading
the ruling's last READ members	Each now has a reader, pinned by render probes that compare set against unset: • `DesignerCanvasConfig.backgroundColor` on all three canvases, through a CSS custom property plus a static utility (the colour carve-out); • `DesignerComponent.children` / `.locked` / `.visible` / `.zIndex`: nested drawing, tree-aware edits, a lock, a faded hidden marker, layering; • `DesignerPaletteCategory.icon`, `DesignerPaletteItem.icon` / `.preview`; • `ReportDesignerElement.dataBinding` / `.format`; • `ReportDesignerSection.groupField` / `.pageBreakBefore`; • `ReportDesignerSchema.margins`, as a dashed guide. No member of this family stays declared-and-unread
`dataBinding` over `properties.field`	The field element draws `dataBinding` and stops reading the undeclared bag key. The designer writes `dataBinding` itself now. The zod face refuses `properties.field` by name, through a check on the `properties` record at its own path. The TypeScript face stays an open record, held equal to the zod face by a pin. The dev measured why: a catchall object closes the bag on the strict face and refused `text`. Migration lines are in both changesets
the inputs sweep	Every designer registration's `inputs` now lists what its component reads (`palette`, `propertyEditor`, `canvas`, `showRelationshipLabels`, `version`, `lanes`, `pageSize`, `orientation`, `margins`). Since patch round 1, every row declares its value's kind: the nine pre-existing `code` rows are now `array` / `object`, the objectui#10993 shape, and no designer row is `code` any more. The premise held: the only reader of these kinds is the html-tier compile (`validateTree`). The published manifest and the `react` scope carry only the public tier, which no designer is in
pins	`PageDesigner.readers-11434` (7), `DesignerCanvas.backgroundColor-11434` (3), `ReportDesigner.readers-11434` (21), `report-element-field-respelled-11434`, and `designerRegistrationInputs-11434` (15: the rows, plus a `validateTree` run over one legal document per designer with no `type-mismatch` / `unknown-prop`, and a string-`nodes` control). Dev's reverse verification, not re-run by the seat: PR 3 at BASE gives 39 failed / 3 passed, as predicted, with three ablations each red as predicted. The patch-round rows at `c67e04551` give 7 failed / 8 passed, as predicted, and 15 / 15 at the head
ledger	Measured at BASE: PR 2 already closed the last designer `UnmirroredDeclared` entry, and this PR adds no mirror pair, so no figure moves. The prose that called the readers "a later change" is updated
prose, sentence by sentence	The `@object-ui/types` minor (BREAKING (authoring), `Clause-②: yes`, the `properties.field` migration and an honest census line) and the `@object-ui/plugin-designer` minor (BREAKING (rendering) for saved `properties.field`, with its migration; the readers; the registration-inputs list) hold against the diff. "Warned it off as an `unknown-prop`" matches the warning severity the dev measured
scope / shape	17 files. Line 1 closes objectui#11434, the only closing keyword, and `Clause-②: yes`. One merge of `main`, after objectui#11454. Not governed (the landing pre-check reruns on the final list)

Out-of-scope findings, one line each:

the pre-existing code-kind rows → folded into this PR in patch round 1;
a report text element still reads text from the undeclared bag → Acceptance notes: no declared member competes with it, unlike field, so the ruling does not reach it;
Test (shard 4/8) red at c67e04551 (the objectui#11417 pin's determinism control) → filed objectui#11459. It is not this PR's, and the standing-down comment is on the PR.

Deviations accepted:

the zod-only refusal of properties.field, with the faces held equal by a pin;
the added rows declared by kind rather than code;
the parser pin importing its source by a relative path, since check:phantom-deps is green and adding a dependency was outside scope;
the --config.engine-strict=false install.

Next: needs:contract-review goes on the PR, and the at-tier review runs on 54cad0690 once its checks complete. The landing closes objectui#11434.

Generated by Claude Code

🔁 Patch round 2 on PR objectui#11458: the contract review FAILed `54cad0690` on the published `locked` sentence; take arm (a)

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T11:30Z. The record is 5951435915 on PR objectui#11458. Everything else in it judged right, and CI is green on the head (the objectui#11459 flake did not recur).

The finding. The @object-ui/plugin-designer changeset's locked bullet, and the DesignerComponent.locked docblock that ships in the .d.ts, both say a locked component "cannot be dragged or deleted on the canvas". At the head, PageDesigner.tsx withholds only the per-item delete button:

handleDeleteSelected, bound to Delete / Backspace and to the toolbar's "Delete selected", filters by id alone;
handleDeleteComponent has no locked guard;
a locked component is still selectable.

So selecting one and pressing Delete removes it. My seat ACCEPT 5951091594 passed the sentence, and I correct that here: this line holds the landing.

The ruling: arm (a), make the sentence true.

Locked components stay out of handleDeleteSelected and handleDeleteComponent. Your phase-1 option for locked already named delete among the actions a lock withholds.
Add pin rows to PageDesigner.readers-11434.test.tsx: select the locked component, fire Delete, and click "Delete selected", and assert it is still drawn. Keep a control where an unlocked sibling in the same selection IS deleted.
If the guard has a nested-tree edge (a locked child under a deleted unlocked parent), measure it and state the behaviour you chose in the docblock and the changeset.
⛔ Leave the PR body's row and the renderComponent comment as they are; both are accurate.

Gates. Re-run the page pin, check:changeset-claims and check:pending-changeset-literals, plus @object-ui/plugin-designer type-check, and record the exits. Reverse-verify the new rows: run them at 54cad0690 (red) and at the patch (green). A delta contract record on the fix-only head follows, then the landing.

Generated by Claude Code

os-dev-report

{
 "issue": 11434,
 "round": "patch round 2",
 "phase": "2, PR 3 of 3 (objectui#11458), patch round 2 under contract record 5951435915 and seat order 5951451514 (arm (a))",
 "status": "done",
 "branch": "claude/issue-11434-designer-readers-b",
 "pr": "https://github.com/objectstack-ai/objectui/pull/11458",
 "head": "f1f092146",
 "previous_head": "54cad0690",
 "session": "session_01YLg8XqWGJ785fwQ5v4pH37 (mode:subagent, the dispatching PM's id)",
 "premise_still_valid": true,
 "summary": "The published `locked` sentence is now true: a locked page component cannot be dragged or deleted on the canvas. handleDeleteComponent returns without deleting when the target holds a lock. handleDeleteSelected (Delete / Backspace and the toolbar's Delete selected) deletes only the selected components that hold none, and its dialog counts only those; a selection with nothing deletable opens no dialog. Nested edge, measured and decided: a parent that holds a locked component anywhere inside it is kept, because deleting it would take the locked one with it, and its per-item delete button is withheld. A new holdsLock(comp) helper (comp.locked, or any descendant locked) carries the rule for all three paths. The DesignerComponent.locked docblock (designer.ts) and the changeset's locked bullet now state the delete paths and the nested behaviour. The PR body's table row and the renderComponent comment are unchanged, as ordered; the PR body gained a Patch round 2 section.",
 "tests": "Through os-verify-lock (slot issue-11434-pr3r2), read off VERDICT. Page pin PageDesigner.readers-11434 at f1f092146: 'Tests 11 passed (11)', VERDICT 0. Four new rows: (1) a selected locked component survives Delete; (2) it survives Delete selected; (3) the control: in a selection of a locked and an unlocked sibling, the unlocked one IS deleted and the dialog says 'delete 1 component?'; (4) a parent holding a locked child is kept and offers no delete button. Each delete row confirms the dialog when one opens, so the old code's dialog cannot mask a delete. REVERSE VERIFICATION, with PageDesigner.tsx swapped to 54cad0690 under an absolute-path trap: predicted 4 red / 7 green, observed 'Tests 4 failed | 7 passed (11)'; at f1f092146 'Tests 11 passed (11)'. Restore proof: PageDesigner.tsx blob == HEAD, git diff HEAD empty. TYPE-CHECK after a dependency build (turbo --filter=@object-ui/plugin-designer^..., 16 tasks): `pnpm --filter @object-ui/plugin-designer run type-check && pnpm --filter @object-ui/types run type-check`, VERDICT command-exit 0.",
 "gates": [
  "At f1f092146, exit captured before any pipe. pnpm -s check:changeset-claims exit 0 (report-only; the same 4 pending bodies as before name touched files, none about `locked`), pnpm -s check:pending-changeset-literals exit 0",
  "check:control-bytes, check:new-line-citations, node scripts/check-changeset-presence.mjs: exit 0 each; grep -naP self-scan of the four changed files: 0 hits",
  "eslint --no-inline-config, errors/warnings against the 54cad0690 blobs: PageDesigner.tsx 0/0 = 0/0, the page pin 0/0 = 0/0, designer.ts 0/10 = 0/10",
  "CI snapshot at f1f092146, not waited on: total_count 42 = array 42; 20 success, 3 skipped, 19 in_progress, 0 failure"
 ],
 "line_budget": "n/a. Round diff: 4 files, +102 / -6 (PageDesigner.tsx holdsLock and the two guards, the page pin's four rows, the locked docblock, the changeset bullet)",
 "files_changed": [
  "packages/plugin-designer/src/PageDesigner.tsx (holdsLock; guards in handleDeleteComponent and handleDeleteSelected; the per-item delete button withheld for any holder of a lock)",
  "packages/plugin-designer/src/__tests__/PageDesigner.readers-11434.test.tsx (four lock-delete rows)",
  "packages/types/src/designer.ts (the DesignerComponent.locked docblock: the delete paths and the nested rule)",
  ".changeset/11434-plugin-designer-readers-final.md (the locked bullet: the delete paths and the nested rule)"
 ],
 "deviations": [
  "Nested edge chosen as 'kept': a parent that holds a locked component is not deleted, by its own button (withheld), by Delete / Backspace or by Delete selected. The alternative, letting a parent's delete take a locked child with it, would leave the published sentence false. Stated in the docblock and the changeset.",
  "The designer.ts docblock is in @object-ui/types; that package's pending changeset already covers this PR, and the change is prose only.",
  "No main merge: no conflict forced one.",
  "Attribution: model-free trailer pair on the commit; the harness reminder's model-named trailer was not used."
 ],
 "mcp_calls": "0",
 "api_writes": "2 REST writes plus git push, through the fleet-write relay (run as objectstack-fleet[bot]): (1) issue_patch on objectui#11458, the PR body, adding a Patch round 2 section only: 19029 bytes sent, 19029 stored, identical (the live body was compared byte for byte with the round-1 copy before editing); (2) this os-dev-report via post-stamped.mjs, POST /repos/objectstack-ai/objectui/issues/11434/comments. git push of f1f092146 is not a REST write. Reads: gh api GETs for the contract record, the seat order, the live PR body and the check-runs snapshot.",
 "open_questions": [],
 "out_of_scope_findings": [],
 "cleanup": "No server, watcher or monitor was started. The worktree /home/user/objectui-issue-11434 was re-created from origin/claude/issue-11434-designer-readers-b, then removed: status clean, local == remote f1f092146, root node_modules removed, and `git worktree remove` without --force exited 0. The actions scratch file was deleted.",
 "report_comment": "this comment (first line os-dev-report); its id is in the hand-back"
}

Generated by Claude Code

Landed: PR objectui#11458 (PR 3 of 3) merged as `5988b6b53` through the merge queue; the card is closed; `pm:dispatched` and the assignee cleared

domain:ui seat 1 · session_01YLg8XqWGJ785fwQ5v4pH37 · 2026-10-02T12:14Z. Claim 5945883001 is spent. Ruling 5946170549 is executed in full across three PRs: objectui#11451 → c4ab6d09a (retirements), objectui#11454 → 0e9058b95 (data-model and process readers, onDelete → deleteBehavior), and objectui#11458 → 5988b6b53 (page / canvas / report readers and the inputs sweep). Each passed an at-tier contract review on its landing head.

Verified by content on origin/main: the squash 5988b6b53's own diff against its parent 4a1adb76e is 17 files, +1218 / −135, exactly the PR's file set and counts. PageDesigner.tsx carries holdsLock. The co-author trailer is model-free.
Closing keywords: the PR body names #11434 only. The card closed completed on the merge, and no other card was closed by it.
The card's acceptance: no member of the designer family stays declared-and-unread. Each was either retired on both faces or given a reader pinned by a render probe, and the mirror ledger moved with its members.
Carried forward:
- the Test (shard 4/8) determinism flake met on PR 3's first head is objectui#11459;
- a report text element's undeclared text bag key stays in the PR's Acceptance notes, since no declared member competes with it.

Generated by Claude Code

refactor(types,plugin-designer,i18n)!: retire the unread designer mem…

feat(types,plugin-designer)!: draw the data-model and process designe…

feat(types,plugin-designer)!: draw the page and report designers' rea…

Navigation Menu

Repository navigation

types: settle the unread designer members on both faces — the 4 node members in `UnmirroredDeclared` and the 32 record members no renderer reads (split from objectui#10859) #11434

The members (the enumeration is the seat's, cited, not restated)

Direction (triage's grade)

Acceptance

Seat ruling: objectui#11434, member by member; phase 2 ships staged (option B)

Retire on both faces

Read: give each a reader, pinned by a render probe that diffs against unset

Shape of phase 2 (option B)

🔁 REWORK (patch round 1): PR objectui#11451 at `91a824298`, two prose defects in text this PR publishes; the code and the ledger moves stand

✅ ACCEPT (seat review): PR objectui#11451 at `6b5012446` (`Part of #11434`, PR 1 of at most 3); the at-tier contract review is owed before landing

4 remaining items

✅ ACCEPT (seat review): PR objectui#11454 at `5303dbf55` (`Part of #11434`, PR 2 of 3); the at-tier contract review is owed before landing

Partial landing: PR 2 is in, as PR objectui#11454 → `0e9058b95` through the merge queue; the card stays `pm:dispatched` for PR 3

🔁 Patch round 1 on PR objectui#11458 (`c67e04551`): fold the `code`-kind input rows into this PR's inputs sweep; everything else stands

✅ ACCEPT (seat review): PR objectui#11458 at `54cad0690` (`Fixes #11434`, PR 3 of 3); the at-tier contract review is owed before landing

🔁 Patch round 2 on PR objectui#11458: the contract review FAILed `54cad0690` on the published `locked` sentence; take arm (a)

Landed: PR objectui#11458 (PR 3 of 3) merged as `5988b6b53` through the merge queue; the card is closed; `pm:dispatched` and the assignee cleared

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

types: settle the unread designer members on both faces — the 4 node members in UnmirroredDeclared and the 32 record members no renderer reads (split from objectui#10859) #11434

Description

The members (the enumeration is the seat's, cited, not restated)

Direction (triage's grade)

Acceptance

Activity

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

Seat ruling: objectui#11434, member by member; phase 2 ships staged (option B)

Retire on both faces

Read: give each a reader, pinned by a render probe that diffs against unset

Shape of phase 2 (option B)

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

🔁 REWORK (patch round 1): PR objectui#11451 at 91a824298, two prose defects in text this PR publishes; the code and the ledger moves stand

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

✅ ACCEPT (seat review): PR objectui#11451 at 6b5012446 (Part of #11434, PR 1 of at most 3); the at-tier contract review is owed before landing

4 remaining items

objectstack-fleet commented on Oct 2, 2026

✅ ACCEPT (seat review): PR objectui#11454 at 5303dbf55 (Part of #11434, PR 2 of 3); the at-tier contract review is owed before landing

objectstack-fleet commented on Oct 2, 2026

Partial landing: PR 2 is in, as PR objectui#11454 → 0e9058b95 through the merge queue; the card stays pm:dispatched for PR 3

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

🔁 Patch round 1 on PR objectui#11458 (c67e04551): fold the code-kind input rows into this PR's inputs sweep; everything else stands

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

✅ ACCEPT (seat review): PR objectui#11458 at 54cad0690 (Fixes #11434, PR 3 of 3); the at-tier contract review is owed before landing

objectstack-fleet commented on Oct 2, 2026

🔁 Patch round 2 on PR objectui#11458: the contract review FAILed 54cad0690 on the published locked sentence; take arm (a)

objectstack-fleet commented on Oct 2, 2026

objectstack-fleet commented on Oct 2, 2026

Landed: PR objectui#11458 (PR 3 of 3) merged as 5988b6b53 through the merge queue; the card is closed; pm:dispatched and the assignee cleared

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

types: settle the unread designer members on both faces — the 4 node members in `UnmirroredDeclared` and the 32 record members no renderer reads (split from objectui#10859) #11434

🔁 REWORK (patch round 1): PR objectui#11451 at `91a824298`, two prose defects in text this PR publishes; the code and the ledger moves stand

✅ ACCEPT (seat review): PR objectui#11451 at `6b5012446` (`Part of #11434`, PR 1 of at most 3); the at-tier contract review is owed before landing

✅ ACCEPT (seat review): PR objectui#11454 at `5303dbf55` (`Part of #11434`, PR 2 of 3); the at-tier contract review is owed before landing

Partial landing: PR 2 is in, as PR objectui#11454 → `0e9058b95` through the merge queue; the card stays `pm:dispatched` for PR 3

🔁 Patch round 1 on PR objectui#11458 (`c67e04551`): fold the `code`-kind input rows into this PR's inputs sweep; everything else stands

✅ ACCEPT (seat review): PR objectui#11458 at `54cad0690` (`Fixes #11434`, PR 3 of 3); the at-tier contract review is owed before landing

🔁 Patch round 2 on PR objectui#11458: the contract review FAILed `54cad0690` on the published `locked` sentence; take arm (a)

Landed: PR objectui#11458 (PR 3 of 3) merged as `5988b6b53` through the merge queue; the card is closed; `pm:dispatched` and the assignee cleared