record chrome draws the record's picture from the object's `imageField` in the record page header — the renderer half of objectstack#21182 (ruling A on hotcrm#1199)

[objectstack-fleet]

Unblocked: objectstack-ai/objectstack#21182 closed completed through PR objectstack-ai/objectstack#21221 (9969228626), and the fix is installable in @objectstack/spec 17.6.0 (hold released in the comment below).

Filed by the director seat, summon #32, objectstack-ai/objectstack session_016tKoy8NJa35Yih1FdzrVmn, as the renderer half of the maintainer's ruling A on objectstack-ai/hotcrm#1199. The ruling is batch #261 item 3, maintainer 「其他四张同意」; the record is the Ruling: comment on hotcrm#1199. ⛔ Not a claim.

What to build

• The record chrome draws the record's picture beside the title, from the field the object names in its new object-level imageField (spec(data): an object declares which image field is the record's picture (imageField, beside nameField), refused at publish unless it names an image or avatar field — ruling A on hotcrm#1199 objectstack#21182). The chrome lives in page:header's PageHeaderRenderer, packages/components/src/renderers/layout/containers.tsx, the record branch at about :2157–:2194 on main 8001068b9c.
• The value comes from the served row. A field the caller may not read arrives absent, so nothing is drawn: masking is the server's answer, ⛔ never re-derived here.
• No declaration, or an empty value, draws nothing. ⛔ No initials/placeholder avatar (ruled out, zero pull).
• recordChrome: false keeps today's bare header.
• ⛔ No page:header prop and no private fallback key (logo, avatar, image, …) read by name. The object declaration is the only channel.

Unlock

• This card is blocked on objectstack#21182 and on an installable @objectstack/spec release that carries imageField. objectui consumes the spec from npm, so the unlock check is the installed package accepting the key, not the upstream merge.
• Probe at unlock time: the installed @objectstack/spec ObjectSchema parses { …, imageField: 'logo' } (logo being an image field) without an unknown-key refusal.

Acceptance

• A pin over the record chrome: with imageField, the image renders from the row value; without it, the header is unchanged; with an empty value, nothing renders; with recordChrome: false, it is unchanged.
• A changeset for the released package.
• The Seam's acceptance on objectstack#21182 is this read: its liveness-ledger row leaves planned when this lands and the spec pin moves past it.
• The first app reader is hotcrm#1199 (crm_account.logo), which waits on this card.

Lane

domain:ui, level S–M, mode:subagent, Clause-②: no (it reads a key the spec publishes; no objectui export changes).

Dedupe: mcp__github__search_issues repo-scoped, open and closed, 「record header image avatar logo record chrome page header」 → 2 hits (#10499, #7281), both closed and unrelated: title masking and width arbitration.

---

Generated by Claude Code

[objectstack-fleet]

Triage: pm:blocked → pm:on-hold. The upstream half landed, but it is not installable yet

Triage seat (objectstack-wide, seat post objectstack-ai/objectstack#6015) · session_01AavokzJ5DndAwitDXvKy4U · 2026-10-01T19:03Z. ⛔ Not a claim, ⛔ not a dispatch.

Restart-when: an @objectstack/spec release that carries 9969228626 (PR objectstack-ai/objectstack#21221) is published on npm, and objectui resolves it. 17.5.0 does not (read at this write).

• spec(data): an object declares which image field is the record's picture (imageField, beside nameField), refused at publish unless it names an image or avatar field — ruling A on hotcrm#1199 objectstack#21182 closed completed at 2026-10-01T18:56Z, landed by PR feat(spec): ObjectSchema.imageField — an object declares its record picture, refused unless it names an image or avatar field objectstack#21221. The cross-repo unlock keys on the install face, not on the merge, as this card's own Unlock section says.
• 9969228626 is not in the current head of chore: version packages objectstack#20639 (the 17.6.0 release PR), so 17.6.0 cut from that head would not carry it. The line names the commit, not a version number, for that reason.
• At the restart, the card's own probe runs first: the installed ObjectSchema accepts imageField.

---

Generated by Claude Code

[objectstack-fleet]

Hold released: pm:on-hold → pm:queue. The Restart-when: condition holds on the installed @objectstack/spec 17.6.0

domain:ui execution seat 1 @ objectui (seat post #5560) · session_015W8GBu6sBiqus2L2xjMsAL (os-steve) · 2026-10-04T14:04Z. ⛔ Not a claim, ⛔ not a dispatch.

The condition is the Restart-when: line of the latest transition comment, 5938525466.

Readings at this write:

• npm view @objectstack/spec version answers 17.6.0.
• objectui origin/main fd060f0767: pnpm-lock.yaml resolves @objectstack/spec only at 17.6.0 (2 entries, no other version).
• REST compare/9969228626...617f25f8a (617f25f8a is the commit tag @objectstack/spec@17.6.0 points at): ahead 34, behind 0, so 9969228626 is in the 17.6.0 release.
• Install-face probe, with @objectstack/spec installed from npm into a scratch directory at each version. 17.5.0 is the control, and shows the probe can fire:
  • data.ObjectSchema.safeParse of an object with a logo image field and imageField: 'logo': 17.5.0 refuses it (Unrecognized key(s) on this object: \imageField`); **17.6.0 accepts** it and keeps imageField: "logo"`.
  • Control in both versions: the same object with a bogus key imageFieldBogus is refused, so the schema is still closed.
• Release double-check ②: the card's timeline shows no merged PR referencing it after 5938525466.

Correction to two earlier readings. 5938525466 predicted that 17.6.0 would not carry 9969228626, because the commit was not in the release PR's head at that moment. That head moved before the release was cut. The seat post's later lines ("waits on a release newer than 17.6.0") repeated the prediction without re-measuring it. The installed package is the reading, and it carries the key.

The card's own unlock probe ("the installed ObjectSchema accepts imageField") is the one run above. The card returns to the domain:ui queue.

---

Generated by Claude Code

[objectstack-fleet]

Claim: PM loop round 1
Session: session_015W8GBu6sBiqus2L2xjMsAL
Account: os-steve
Branch: claude/issue-11383-record-chrome-image
Worktree: objectui-issue-11383
Domain: domain:ui
Seat: domain:ui#1
File surface: packages/components/src/renderers/layout/containers.tsx (the page:header record branch, which resolves the title from nameField today), the tests beside it, and .changeset/11383-*.md (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: default tier (dispatch-gates --tier --repo objectstack-ai/objectui: no path-derived mandate; floor sonnet · default opus · ceiling fable)
Clause-②: no
Thread-read: 5980846960
Serial constraints cleared: none, read 2026-10-04T14:14Z: the four open objectui PRs (objectui#11614, #11607, #11600 the release PR, #11069) were read in full (29, 202, 101 and 15 files; the release PR touches only changesets, CHANGELOGs, manifests and QUICK_REFERENCE.md) and the three pm:dispatched cards (objectui#11610, #10202, #8347) by their claims; none touches this surface.

Why Clause-②: no: the renderer reads imageField, a key the installed @objectstack/spec 17.6.0 already publishes and judges at parse (liveness row object.imageField: planned, carrier this card). No @object-ui/* export, prop, type member or i18n key is added, and the dispatch fences that: a route that needs one stops and reports.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 11383,
"status": "done",
"branch": "claude/issue-11383-record-chrome-image",
"pr": "#11619",
"session": "session_015W8GBu6sBiqus2L2xjMsAL (dispatched subagent; the parent seat's id)",
"premise_still_valid": true,
"summary": "page:header's record chip now draws the record's picture in the existing RecordTitleChip icon slot, from objectSchema.imageField and the served row's value. It is a Radix Avatar + AvatarImage with no AvatarFallback, so there are no initials and no broken-image icon. Nothing is drawn, and the header markup equals the undeclared header's, for no declaration, an empty value, an absent (masked) field, or recordChrome false. No page:header prop and no conventional key is read. Zone 2 #3 is falsified as a reuse route: the resolver objectui already has (readFileValue in @object-ui/fields) cannot be imported from @object-ui/components because fields depends on components, and moving it to core would add a core export (Clause-2 fence). So containers.tsx carries a small private recordPictureUrl that covers exactly the spec's read forms: expanded {url}, a bare sys_file id via isFileIdToken to the storage path, a URL string, and the first drawable entry of a list. It is labelled in code as a second resolution; consolidating it is open question 1. Landing site as expected: packages/components/src/renderers/layout/containers.tsx.",
"tests": "New pin packages/components/src/tests/page-header-record-picture-11383.test.tsx (13 cases, real SchemaRenderer + registry). At 711adb9: 'Tests 13 passed (13)'; full pnpm exec vitest run packages/components/ gave 'Test Files 358 passed | 1 skipped (359) / Tests 3643 passed | 24 skipped', exit 0. At merged head 4a34ea4 (origin/main 2abec3a merged), the same set ran as --shard=1/2 + --shard=2/2 because a single run nears the 10-minute foreground cap: 'Test Files 180 passed (180) / Tests 1754 passed' and 'Test Files 178 passed | 1 skipped (179) / Tests 1889 passed', exit 0 / 0. That is 359 files and 3643 tests, matching the pre-merge run, and the new file is in that set. A separate single-file rerun at 4a34ea4 was NOT MEASURED: queue-timeout exit 99 twice, with the lock held by another agent's app-shell run. Deps closure pnpm --filter \"@object-ui/components^...\" build then pnpm --filter @object-ui/components type-check at 4a34ea4: exit 0, both tsc projects. --listFiles counts 1 for the new test in tsconfig.test.json and 1 for containers.tsx in tsconfig.json. The first type-check, before deps were built, exited 2 with TS2307 cannot find @object-ui/core: a prerequisite failure, not a red gate. Lint at 4a34ea4: exit 0, '0 errors'; no warning on a changed line (the only nearby hits sit on the pre-existing PageHeaderRenderer declaration). Ablations, all run on committed code via objectstack scripts/ablation-replace.mjs; each mutation was proven on disk (anchor 1 -> 0, blob changed) and each restore by 'blob == HEAD and git diff HEAD is empty'. (1) Deleting 'icon={recordPicture}': '6 failed | 7 passed'; every positive pin went red and the 7 'nothing drawn' controls stayed green. (2) Changing imageField to 'objSchema?.imageField ?? logo': '1 failed'; the conventional-key control went red. (3) 'pictureFieldName ? (' in place of 'pictureUrl ? (': '5 failed'; every empty-value markup-equality control went red. The direction was as expected in all three. Spec probe on the installed 17.6.0 ObjectSchema.safeParse: image and avatar are accepted; an undeclared name is REFUSED ('declares no field'); text and file are REFUSED ('which is not a picture'); the control key imageFieldBogus is REFUSED as unrecognized. So the PM's unverified refusal half holds.",
"mcp_calls": "0",
"api_writes": "3 relayed writes as objectstack-fleet[bot], each carried by one POST /repos/objectstack-ai/objectstack/dispatches: (1) pr_create POST /repos/objectstack-ai/objectui/pulls (draft, read back 9491 bytes sent = 9491 stored); (2) label-write assign POST /repos//issues/11619/assignees (os-steve; read back MATCHES); (3) post-stamped POST /repos//issues/11383/comments (this report). Plus git push of the branch (3 pushes: empty probe, 711adb9, 4a34ea4). Reads only otherwise: REST GET issue 11383 + comments, GET pulls/11619, GET /rate_limit.",
"gates": [
{
"gate": "pnpm exec vitest run packages/components/",
"head": "711adb9",
"exit": 0
},
{
"gate": "pnpm exec vitest run packages/components/ --shard=1/2",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "pnpm exec vitest run packages/components/ --shard=2/2",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "pnpm exec vitest run packages/components/src/tests/page-header-record-picture-11383.test.tsx",
"head": "711adb9",
"exit": 0
},
{
"gate": "same single file at merged head",
"head": "4a34ea4",
"exit": "NOT MEASURED (queue-timeout 99 x2; covered by the shards)"
},
{
"gate": "pnpm --filter "@object-ui/components^..." build && pnpm --filter @object-ui/components type-check",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "pnpm --filter @object-ui/components lint",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "check:new-line-citations",
"head": "711adb9",
"exit": 0
},
{
"gate": "check:control-bytes",
"head": "711adb9",
"exit": 0
},
{
"gate": "check:phantom-deps",
"head": "711adb9",
"exit": 0
},
{
"gate": "check:changeset-claims",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "check:pending-changeset-literals",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "check:unreferenced-sources",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "check:test-path-roots",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "node scripts/check-changeset-no-major.mjs",
"head": "4a34ea4",
"exit": 0
},
{
"gate": "node scripts/check-changeset-presence.mjs",
"head": "711adb9 tree",
"exit": 0
},
{
"gate": "check:spec-floors",
"head": "-",
"exit": "NOT MEASURED: needs every released package's dist; direct reading instead: isFileIdToken is exported by @objectstack/spec/data at 17.0.0 (npm tarball d.ts + mjs), the components floor"
},
{
"gate": "check:eager-closure",
"head": "-",
"exit": "NOT MEASURED: needs the console production build"
}
],
"files_changed": [
".changeset/11383-record-chrome-image.md (+5, @object-ui/components minor)",
"packages/components/src/tests/page-header-record-picture-11383.test.tsx (+233)",
"packages/components/src/renderers/layout/containers.tsx (+87)"
],
"line_budget": "n/a: no skills/** and no governed surface touched (Tier: not governed)",
"deviations": [
"Zone 2 #3 (reuse the existing resolution) falsified by the package graph: @object-ui/fields depends on @object-ui/components, and the core move breaches the Clause-2 fence. So a private, labelled second resolution lives in containers.tsx (recordPictureUrl). It is narrower than readFileValue: an object with no url is not drawn. See open question 1.",
"Zone 3 suggested the record title as the picture's accessible name. Instead the picture is decorative (alt empty, inside the chip's aria-hidden icon slot), because the H1 beside it already names the record and a titled image would read the name twice.",
"The dispatch and the PM heads-up spell the script 'typecheck'; @object-ui/components names it 'type-check'. A 'typecheck' filter matches no script, so 'type-check' was run.",
"Components tests at the merged head ran as two shards rather than one run (the single run took 569s against the ~600s foreground cap).",
"The merge commit message was amended before its first push, to carry the model-free trailer pair. No pushed history was rewritten."
],
"open_questions": [
{
"question": "The header's picture URL resolution is a second copy of readFileValue (in @object-ui/fields), and nothing pins the two together. Should they be consolidated?",
"options": [
"A: keep the private copy (as landed). Cost: silent drift if the storage path or the read forms move.",
"B: move fields' file-value module down into @object-ui/core, have fields re-export it, and have page:header import it. This gives one resolver, but adds public core exports (a Clause-2 yes card).",
"C: keep the copy and add a test-only parity pin in @object-ui/fields that renders page:header over a value corpus and compares the drawn src with readFileValues(value)[0].url (no export; outside this claim's file surface)."
],
"recommendation": "C, by the four axes. Business need: there is one real consumer (hotcrm#1199 crm_account.logo) and no second one below fields, so B's new core surface has no measured pull. Long-term: B is the single-source shape, and C keeps the drift visible until a second consumer justifies B. AI-error: C turns a silent divergence into a red test, while A leaves it silent. Startup focus: C costs one test file and no public surface. Take B only when a further package below fields needs the resolver."
},
{
"question": "content/docs/layout/page-header.mdx (Layout, the record-chip bullet) lists the chip's parts without the picture, and AGENTS.md #2 asks docs to move with the code. The file is outside the claim's declared surface, so it was not edited.",
"options": [
"A: amend the claim's surface and add one clause to the record-chip bullet on this PR as a patch round: '..., and the record's picture when the object names an imageField'.",
"B: a docs-only follow-up after landing."
],
"recommendation": "A: it is a one-line change, docs should land with the behaviour, and the PR stays in one package's docs plus source."
}
],
"out_of_scope_findings": [
"carrier: 承接者:无 · noted, not filed. RecordDetailPanel (@object-ui/plugin-detail) renders DetailView's own heading, not page:header, so the record side panel draws no imageField picture. The spec describe says 'one object-level declaration every record detail page reads'. Not measured in a running app, and the card scopes the change to page:header. Dedupe words: DetailView imageField record panel picture.",
"carrier: the objectstack .objectui-sha pin-bump follow-up (PM-filed, per dispatch Zone 2 #4) · noted, not filed. The @objectstack/spec imageField describe still reads 'Pending renderer: the record chrome does not draw it yet.' and should move with the liveness row object.imageField."
]
}

---

Generated by Claude Code