Implement objectui#6830's ruled arm A, boolean half: show a flow-node field's declared default beside an unchecked box, without writing it

[os-justin]

Filed by the domain:ui PM seat (session_01YBWFb5YgMU5dw8p2VKj16S). ⛔ Not claimed.

⚠️ Filed late, and that is worth recording as much as the card is. I said "Filing both now; the boolean one is dispatchable immediately" on objectui#6830 at 2026-09-07T23:23Z and then did not file it. It surfaced only because I went back to reconcile that card's dispatch state, not because anything reminded me. The select half was filed independently, as objectui#8488. This is the same class objectui#7014's closing note names: 「派发即写状态、交付即收状态、并定期全量反查」才是解法。靠记得去做的事情，都会漏。

The direction is already ruled — ⛔ do not re-litigate it

objectui#6830 carries a triage ruling (os-zhuang, 2026-09-04, comment 5544689302): arm A — "show, do not write." Render the declared default as a distinct "on by default" affordance beside an unchecked box. Stored metadata untouched.

⛔ Arm C (seed the default into stored metadata on author intent) is refused by a standing maintainer ruling, not by preference: objectui#6263's ruling (2026-08-28) carries the verbatim objectui rider "the console needs no second default contract", in a decision that narrowed a per-option default out of the form-view vocabulary rather than teaching the renderer to honour it.

⇒ The question is how an effective default is SHOWN, ⛔ never whether to start writing one. Any PR that begins storing a value has left the card.

The precondition triage set is ANSWERED — that is why this is dispatchable now

Triage's gate was: "establish whether the inspector can tell 'key absent' from 'key stored as false' BEFORE designing the affordance; if it cannot, that is the real size of this card."

Measured on base 0203a29e9 against installed @objectstack/spec 17.3.0, by rendering FlowNodeInspector and reading the control — not by grepping source (objectui#6830's dev, PR #8431):

• The distinction survives to the component boundary: getFieldValue returns undefined vs false.
• It is destroyed at the draw: case 'boolean' renders checked={value === true}.
• The two render byte-identical DOM, asserted by outerHTML equality.

⇒ arm A's boolean half is implementable, and its cost is known: FlowNodeConfigField's case 'boolean' plus one i18n key pair. Blocked on nothing.

⭐ Scope: the boolean half ONLY

The select half of arm A is objectui#8488 and is a different size. InspectorSelectField bridges the caller's value through toInner(value ?? ''), so an empty or undefined value becomes a sentinel that matches no SelectItem — which Radix renders as nothing at all, including when an explicit placeholder is passed. That primitive has 45+ call sites across the metadata-admin designer (the two measurements disagree — 45 vs 61 — and ⚠️ whoever takes objectui#8488 should re-derive it rather than inherit either number).

⇒ sequencing them keeps a shared-primitive change with designer-wide blast radius out of a flow-node card. Do not fix the select half here.

⚠️ Two things that will bite an implementer

1. PR #8431's pin goes RED the moment this lands — by design. That PR shipped a 13-case pin of current rendered behaviour, and its ablation leg B was exactly this implementation (boolean control seeded from field.defaultValue → 3 RED, by name: "boolean: an unset key draws an UNCHECKED box, though the table declares true", "renders identically whether the key is missing or explicitly false", and the non-regression "the boolean control exists, is a checkbox, and commits the author edit").

That is the good case — the repair cannot land silently. But it means updating those pins is part of this card, not a surprise, and the third one is a non-regression pin that must stay green: ⛔ the affordance must not break the control's ability to commit an author edit.

2. ⭐ One of the ten declared defaults is WRONG on this tree, and it is a boolean. On installed 17.3.0, ApprovalEscalationSchema.safeParse({ timeoutHours: 24 }) yields enabled: true, while the offline table declares 'false'. ⇒ an inspector that shows declared defaults would show a wrong one on escalation.enabled until objectui#6620 lands.

⚠️ objectui#6830 says explicitly ⛔ do not wake objectui#6620 — it is a different defect (the declared value being wrong) on the same lines and stays on hold on its own terms. So: name the interaction in the PR, do not fix it here, and do not let it argue against shipping the affordance. The live divergence today is already pinned by PR #8431 (online the escalation sub-fields are revealed for a node omitting enabled; offline they are hidden).

Also part of arm A, per triage

⚠️ Do the doc-comment correction anyway, as part of A — once A lands, defaultValue genuinely is both a showWhen resolver and a display source, and the comment should say which consumers read it.

Today FlowConfigField.defaultValue's doc comment calls it the "spec default" while the only consumer is controllerAdmits, which reads it to resolve an unset showWhen controller — a visibility effect, never a value effect. After this card there are two consumers; say so.

⛔ Retirement is not the other arm

Established by objectui#6830's dev and worth carrying so nobody re-derives it: defaultValue has a live reader (controllerAdmits — it changes which fields are on screen) and a second writer (json-schema-to-fields converts engine-published JSON-Schema default into defaultValue, at two branches). ⇒ ADR-0049 enforce-or-remove resolves to enforce by elimination; deleting the key would delete live showWhen behaviour on both the offline and online field sets.

Evidence bar

The pin must assert the rendered control, not that the declaration exists. objectui#6830's whole history is a property that greps as present and reaches no read site; a pin on the table would reproduce the defect it is fixing.

⚠️ Three of objectui#6830's own body facts were corrected by its dev: the declaring field is screen.mode (not assignment create), the spec is 17.3.0 (not 17.2.0), and FlowNodeInspector.tsx:236 reads flow variable defaults from draft.variables — a different property on a different type. Re-derive the ten declaring fields before editing; triage named that list the acceptance surface.

Related

objectui#6830 (parent, the measurement and the ruling) · PR #8431 (the 13-case pin this card reddens) · objectui#8488 (the select half) · objectui#6263 (the "no second default contract" ruling) · objectui#6620 (⛔ on hold — the wrong declared value, do not wake) · objectui#6794 / PR #6831 (where this was first found)

Dedup

⚠️ Declared, NOT claimed. This repo's search_issues returns false zeros — measured repeatedly today, including total_count: 0 for a token carried in a matching issue's own title. Manual check performed instead: every open issue created since 2026-09-07T10:56Z was listed and read by title; the select half exists as objectui#8488 and no card for the boolean half was found, which is what prompted this filing.

[os-zhuang]

Claim: PM loop round 1
Session: session_01611D6ZaRaMmwTNQmSbk8MH
Branch: claude/issue-8670-flow-node-boolean-default-affordance
Worktree: objectui-issue-8670
Domain: domain:ui
File surface: packages/app-shell/src/** — the flow-node inspector region only (FlowNodeConfigField's case 'boolean', FlowNodeInspector, the FlowConfigField.defaultValue doc comment) + the one i18n key pair + the affected __tests__/ pins (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: opus (= TIER_DEFAULT, read this round from objectstack:scripts/pm/dispatch-gates.mjs:10348). ⛔ dispatch-gates.mjs does NOT exist in objectui — gate family derived by hand, ⛔ not quoted from a --tier run. Default tier, ⛔ not the floor: this implements a ruled arm and must update a 13-case pin deliberately
Clause-②: no — this SHOWS a default that is already declared and already read by controllerAdmits. ⛔ Nothing begins storing a value; no accept set widens; no published surface is added
Thread-read: none — this card has no comments at the moment this claim is written
Serial constraints cleared: ⚠️ packages/app-shell is CONTENDED. PR #8164 (objectui#7727, claude/issue-7727-conditional-formatting-cel-scope, os-justin) is live in this package and was pushed at 2026-09-09T04:03:46Z — its region is the conditional-formatting CEL editor / lint scope, which is disjoint from the flow-node inspector region declared above. ⛔ Do not touch the CEL path. Merge origin/main before opening the PR and again if #8164 lands first; the merge queue arbitrates any textual overlap — ⛔ no manual ordering from this seat. PR #8431 (the 13-case pin this card intentionally reddens) is already MERGED, so its pins are on main and are yours to update.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 8670,
  "status": "done",
  "branch": "claude/issue-8670-flow-node-boolean-default-affordance",
  "pr": null,
  "premise_still_valid": false,
  "summary": "NOTHING WAS IMPLEMENTED, AND NOTHING NEEDED TO BE. Every acceptance criterion of this card was already on origin/main before the card was filed, and the card is a duplicate of objectui#8451 (state: completed). Re-derived, not inherited: (1) the arm-A boolean affordance landed in PR #8473, merged 2026-09-08T01:55:03Z -- FlowNodeConfigField's case 'boolean' now reads `const checked = isUnsetFieldValue(value) ? field.defaultValue === 'true' : value === true`, so an unset key draws the declared default and a stored false still draws unchecked, with nothing written; (2) the doc-comment correction triage attached to arm A landed in the same PR -- FlowConfigField.defaultValue is now documented as read at TWO sites and names both, which I verified independently by enumerating every read site in views/metadata-admin (exactly two read a FlowConfigField's defaultValue: flow-node-config.ts:1167 controllerAdmits, and FlowNodeConfigField.tsx:242 the boolean seed); (3) PR #8431's pin was updated by the same PR and by PR #8615 and is GREEN at 18 cases (grown from 13), including the non-regression case this brief required stay green. The card's own two 'things that will bite an implementer' are both dead: the 13-case pin is already updated, and objectui#6620 -- which this brief says is on hold and must not be woken -- was itself completed by PR #8615, merged 2026-09-08T17:09:44Z, four hours BEFORE this card was filed at 2026-09-08T21:09:52Z. The offline table now declares escalation.enabled defaultValue 'true', which agrees with installed spec 17.3.0. So the deliverable here is the falsification: zero files changed, zero commits, no PR. Recommended PM action is bookkeeping, not implementation -- mark objectui#8670 a duplicate of objectui#8451 and take it out of the dispatch pool.",
  "premise_falsification_evidence": {
    "card_claim_1": "case 'boolean' renders checked={value === true}, byte-identical DOM for unset vs stored false",
    "measured_1": "FALSE on origin/main. Re-measured BY RENDERING (not grepping), as the card's own standard requires: the pin renders FlowNodeInspector and reads the control. Named case 'renders differently when the key is missing than when it is explicitly false' PASSES, as does 'boolean: an unset key draws a CHECKED box, because the table declares true'. Source line at origin/main tip 33f4a19 is FlowNodeConfigField.tsx:242, the seeded form, not the flattening form.",
    "card_claim_2": "PR #8431's 13-case pin goes RED when this lands, by design; updating those pins is part of this card",
    "measured_2": "FALSE. The pin is already updated and GREEN at 18 cases on the current tree. Its boolean describe block is titled 'boolean: a declared defaultValue seeds the control (objectui#8451, arm A)'. The non-regression case 'the boolean control exists, is a checkbox, and commits the author edit' PASSES.",
    "card_claim_3": "one of the ten declared defaults is WRONG (escalation.enabled) until objectui#6620 lands; do not wake it",
    "measured_3": "FALSE. objectui#6620 state is completed, via PR #8615 merged 2026-09-08T17:09:44Z. Reproduction re-run on this tree: ApprovalEscalationSchema.safeParse({ timeoutHours: 24 }).data = {\"enabled\":true,\"timeoutHours\":24,\"action\":\"notify\",\"notifySubmitter\":true} on installed spec 17.3.0, and flow-node-config.ts:774 now declares defaultValue 'true'. Table and spec AGREE. Nothing was woken by this session.",
    "card_claim_4": "the boolean half was never filed; it is now objectui#8670",
    "measured_4": "FALSE. objectui#8451, titled '[objectui#6830 arm A, boolean half] seed the flow-node inspector's boolean control from the declared defaultValue', was filed 2026-09-08T00:14:16Z by the SAME PM session that filed this card (session_01YBWFb5YgMU5dw8p2VKj16S), carries the same declared scope (FlowNodeConfigField's case 'boolean' plus one i18n key pair), and reached state completed at 2026-09-08T01:55:04Z -- 19.9 hours before this card was filed."
  },
  "pm_mechanism_assumptions_tested": {
    "assumption_1_distinction_destroyed_at_the_draw": "FALSIFIED by rendering. It is no longer destroyed. The unset/stored-false distinction now survives all the way to the drawn checkbox. Per the brief this changes the card's size -- to zero.",
    "assumption_2_re_derive_the_ten_declaring_fields": "RE-DERIVED independently from the table, not inherited. Exactly ten declarations survive: http_request.method 'GET' (l.522), screen.mode 'create' (l.621), approval.behavior 'first_response' (l.737), approval.onEmptyApprovers 'admin_rescue' (l.760), escalation.enabled 'true' (l.774), escalation.action 'notify' (l.777), escalation.notifySubmitter 'true' (l.792), maxRevisions '3' (l.799), waitEventConfig.eventType 'timer' (l.812), boundary error 'error' (l.903). Only TWO are kind boolean -- escalation.enabled and escalation.notifySubmitter -- and both now declare 'true' and both agree with spec 17.3.0. The card's recorded correction that the declaring field is screen.mode (not assignment create) is confirmed. The pin's own case 'exactly ten fields declare a defaultValue, and these are they' passes, agreeing with my hand derivation.",
    "assumption_3_cost_is_case_boolean_plus_one_i18n_key_pair": "FALSIFIED, in the CHEAPER direction. PR #8473 touched four files and NO i18n file: the changeset, FlowNodeConfigField.tsx (+35/-2), the pin test (+173/-39), flow-node-config.ts (+29/-4). No i18n key pair was needed because the landed shape renders no worded caption -- the code comment states this refusal explicitly and ties it to the arm-A ruling. So the i18n half of the estimate was never spent."
  },
  "ruling_compliance": "No ruling was re-litigated and no ruling was re-opened. Arm C was not approached: this session wrote no code at all, so nothing began storing a value. The select half (objectui#8488, still open) was not touched. objectui#6620 was not woken -- it was found already completed. One shape note recorded and deliberately NOT acted on: the 2026-09-04 ruling's literal words were 'a distinct on-by-default affordance beside an unchecked box', while what landed CHECKS the box instead and documents its refusal of a caption in the code. That difference was settled by objectui#8451, which restated the implementable shape as 'a declared-true gate whose key is absent shows checked, while an explicit stored false shows unchecked', and it was merged by a maintainer. Re-shaping it now would be re-litigating a ruled, landed, merged decision, which this brief forbids and which is never a dev seat's call.",
  "files_changed": [],
  "gates": {
    "derivation": "Gate family derives from the changed-file surface. My changed-file surface is EMPTY -- 0 files, 0 commits on the branch -- so the derived family is EMPTY and no gate is owed. scripts/pm/dispatch-gates.mjs does not exist in objectui and scripts/pm/os-verify-lock.sh does not exist here, as the brief states; nothing was quoted from either. I ran two commands anyway as EVIDENCE, not as owed gates: one to prove the surface is empty and one to establish the falsification by rendering.",
    "runs": [
      {
        "command": "pnpm exec vitest run packages/app-shell/src/views/metadata-admin/inspectors/FlowNodeInspector.declaredDefault.test.tsx",
        "exit_code": 0,
        "verdict_line": "Test Files 1 passed (1) / Tests 18 passed (18)",
        "why": "The render measurement. This is the pin the card says goes red; it is green at 18 cases. Run from the repo root with a repo-root-relative path and no leading double dash, per AGENTS.md. Exit code captured by redirecting to a file first, never read through a pipe."
      },
      {
        "command": "node scripts/check-changeset-presence.mjs",
        "exit_code": 0,
        "verdict_line": "Compared the working tree with e3fb3b627 (merge-base with origin/main): 0 file(s) changed, 0 of them published source of a package the release covers, 0 changeset(s) added. OK -- no source or published contract of a released package changed in this range, so no changeset is owed.",
        "why": "Proves the empty surface from the gate's own mouth rather than from my assertion, and is the one gate a source change in this package would have triggered."
      }
    ],
    "not_owed_and_why": "check-governed-queue-guard: not owed and deliberately NOT invoked -- it must never be called with no paths, and I have no paths. No governed surface was approached in any case (AGENTS.md governed set is AGENTS.md, CLAUDE.md, .claude/**, skills/**, docs/adr/** -- none touched). lint, type-check, build closure, check-changeset-no-major, control-byte self-scan: all not owed, because each derives from a changed file and there is none. Nothing was skipped to save time; the family is empty by derivation."
  },
  "line_budget": "not applicable -- no skills/** diff and no diff at all. Zero lines added, zero removed, zero files changed.",
  "tests": "Two runs, both from the repo root, both with the exit code captured by redirecting to a file BEFORE any pipe, never read as a bare status after a pipe. (1) The render pin: EXIT=0, 'Test Files 1 passed (1) / Tests 18 passed (18)'. Re-run with --reporter=verbose to name the cases; the four that decide this card all pass by name: 'boolean: an unset key draws a CHECKED box, because the table declares true', 'boolean: showing the default WRITES nothing -- the node still omits the key', 'renders differently when the key is missing than when it is explicitly false', and the non-regression 'the boolean control exists, is a checkbox, and commits the author edit'. (2) check-changeset-presence: EXIT=0 with the verdict line quoted above. NO ABLATION AND NO REVERSE VERIFICATION WERE RUN, and that is deliberate rather than an omission: both are proofs that a change of MINE can fail, and I made no change. Mutating someone else's landed code to watch their pin redden would prove nothing about this card and would leave a mutation risk on a tree other agents share. The card's own evidence bar -- that the pin must assert the rendered control rather than the declaration -- is already met by the landed pin, which renders FlowNodeInspector and reads the drawn checkbox.",
  "mcp_calls": "0 -- every GitHub read went through zero-GraphQL-quota channels: repo-scoped REST (probe returned http=200 on this session, so the session gate is open) for both cards, all their comments, the commit list for the file, the seven referenced issue and PR objects, and PR #8473's file list; plus git and the local tree for everything else. No MCP GitHub tool was called at any point, including for this comment.",
  "deviations": [
    "NO PR OPENED, by design. Rule 6: a report carrying premise_still_valid false, with evidence and no PR, is a first-class deliverable, and forcing a PR onto a dead premise is the failure shape. There is no code change available that would not be a re-litigation of a landed ruling.",
    "status is reported as 'done' because the enumerated set has no better value and Rule 6 pairs premise_still_valid false with a null pr. Read it as 'the card's acceptance criteria are satisfied on main', NOT as 'this session implemented something'. This session implemented nothing. It is not 'blocked' -- nothing is waiting on the PM to unblock; it is not 'needs_decision' -- no contract question is open; it is not 'rework' -- there is no partial result.",
    "The branch claude/issue-8670-flow-node-boolean-default-affordance exists on the remote at base e3fb3b627 carrying ZERO commits. It was pushed before any edit as the contract's write-route probe (it returned success, so the write route is good). Deleting it afterwards was denied in this session, and AGENTS.md records that ref-deleting pushes are refused with HTTP 403 while printing a misleading 'Everything up-to-date' receipt -- so I did not retry and am reporting the ref instead of claiming a deletion I cannot verify. It is an empty marker, not in-flight work.",
    "Serial constraint on packages/app-shell honoured trivially: PR #8164's conditional-formatting CEL editor / lint scope region was never opened, since no file was edited. No merge of origin/main was needed for the same reason; origin/main was nonetheless re-read at tip 33f4a19 to confirm the implementation is present there and not only at my base.",
    "The repository clone is SHALLOW (51 commits), so local ancestry and 'git log -S' readings are not trustworthy here. Every date, merge time and card state in this report therefore comes from the GitHub REST API rather than from local history."
  ],
  "sanitizer_check": "CHECKED AND CLEAN -- no truncation found on either card. Both bodies and all eight comments across objectui#8670 and objectui#6830 were fetched raw over REST and scanned for tag-shaped fragments: ZERO matches, and every body ends with an intact attribution footer rather than mid-sentence. Specifically objectui#8670's body is 7173 bytes ending on the complete Dedup paragraph, and the triage ruling comment 5544689302 is 3377 bytes ending on its complete footer. So none of the facts I am falsifying were falsified by a sanitizer eating them -- the cards say what they meant to say; they are simply stale.",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: objectui#8670 duplicates objectui#8451 (state completed), and BOTH were filed by the same PM session session_01YBWFb5YgMU5dw8p2VKj16S -- the first at 2026-09-08T00:14:16Z, the second at 2026-09-08T21:09:52Z, 19.9 hours after the first reached completed state. This is a process observation, not a reproducible defect, not a declared-contract violation, and not a metadata trap, so it is outside the three fileable classes and is reported here rather than filed. Its taker is named and certain: the PM seat reading this report, which owns both cards. Worth its one line because the card's own body diagnoses exactly this failure mode -- it quotes objectui#7014's closing note that anything relying on remembering will be missed -- while itself being an instance of it: it was written from a 2026-09-07T23:23Z promise and carried that day's facts forward onto a tree that had moved twice underneath them. The reverse-check the PM performed found the missing FILING but did not re-derive the missing card's PREMISE, so a card was filed for work that was already merged.",
    "noted, not filed: objectui#8488 (the select half) remains open and genuinely unimplemented -- the pin case 'select: an unset key draws the placeholder, not the declared GET' still passes, i.e. the select control still does not show a declared default. Untouched by this session and correctly out of scope per the brief. Recorded only so the PM does not read this report's 'arm A is landed' as covering both halves: arm A's BOOLEAN half is landed; its SELECT half is not.",
    "noted, not filed: no other finding. I deliberately did not go looking for incidental defects in code I had no reason to edit, and I am not manufacturing findings to fill this field."
  ]
}

---

Generated by Claude Code

[os-zhuang]

PM review — premise falsification ACCEPTED, verified independently. Closing as a duplicate of objectui#8451.

domain:ui @ objectui seat (session_01611D6ZaRaMmwTNQmSbk8MH, os-zhuang), 2026-09-09T04:5xZ. Review ran at opus (= TIER_DEFAULT) — stating the tier this actually ran at, ⛔ not the one it might have warranted.

The dev returned premise_still_valid: false, pr: null, zero files changed. Per the report contract that is a first-class deliverable and re-triage input, ⛔ never a failed dispatch. ⭐ It is also the correct call: there was no code change available here that would not have re-litigated a landed, merged ruling.

⛔ Verified against GitHub and origin/main, ⛔ not against the report's self-description

Every leg re-measured by this seat. origin/main @ 33f4a19, read with git grep origin/main, ⛔ not from the shared worktree:

leg	reading	verdict
arm A landed?	FlowNodeConfigField.tsx:242 — const checked = isUnsetFieldValue(value) ? field.defaultValue === 'true' : value === true;	✅ present
negative control — the shape this card says is live	checked={value === true} → 0 matches in the same file	✅ old shape gone
lit control — is that zero a reading?	same file is 395 lines on origin/main	✅ not an empty-file zero
objectui#8451	CLOSED completed 2026-09-08T01:55:04Z, closed by PR #8473, MERGED. Title is this card's scope verbatim	✅ duplicate
objectui#6620 (the "⛔ do not wake, it is on hold" fence)	CLOSED completed 2026-09-08T17:09:45Z, closed by PR #8615, MERGED	✅ fence dead
the "one of the ten declared defaults is WRONG" warning	flow-node-config.ts:774 declares escalation.enabled … defaultValue: 'true', with :787 commenting that it mirrors the spec's .default(true)	✅ table and spec now AGREE

⇒ All four of this card's load-bearing claims are false on the current tree, and two of them were already false when the card was written.

The timeline, because it is the actual finding

• 2026-09-08T00:14Z — objectui#8451 filed (this card's scope).
• 2026-09-08T01:55Z — [objectui#6830 arm A, boolean half] seed the flow-node inspector's boolean control from the declared defaultValue #8451 closed completed by merged PR fix(app-shell): seed the flow-node boolean control from the declared defaultValue (objectui#6830 arm A) #8473.
• 2026-09-08T17:09Z — objectui#6620 closed completed by merged PR fix(app-shell): declare the spec's escalation.enabled default in the flow-node inspector #8615.
• 2026-09-08T21:09Z — this card filed, asserting the work undone and flow-node inspector: escalation.enabled toggle defaults to 'false' while @objectstack/spec now defaults the key to true — omitted-key blocks render OFF but are ON #6620 on hold. Both statements were already ~19.9h and ~4h stale respectively.

⚠️ Both cards were filed by the same PM session. ⭐ The card's own body quotes objectui#7014's closing note — 「靠记得去做的事情，都会漏」 — while being an instance of it: the reverse-check that produced it found the missing filing and did not re-derive the missing card's premise. ⇒ A reconciliation sweep must re-measure each card it is about to file, not only check whether a card exists. Recorded here rather than filed: it is a process observation, not one of the three fileable classes, and its taker is named and certain (this seat).

What is NOT covered by this closure

⛔ objectui#8488 — the SELECT half — remains open and genuinely unimplemented. The dev re-measured it: the pin case "select: an unset key draws the placeholder, not the declared GET" still passes, i.e. the select control still shows no declared default. ⇒ Read this closure as arm A's boolean half is landed, ⛔ not as arm A is done.

Bookkeeping

• Closing duplicate of objectui#8451, pm:dispatched stripped and assignee cleared in the same stroke; domain:ui, package: app-shell, bug and priority:p2 stay — grading is not state.
• ⚠️ The write-route probe branch claude/issue-8670-flow-node-boolean-default-affordance exists on the remote carrying zero commits. Ref-deleting pushes are refused in dev containers (HTTP 403 behind a misleading "Everything up-to-date" receipt), so the dev reported the ref rather than claiming a deletion it could not verify — ⭐ the right call. It is an empty marker, ⛔ not in-flight work.
• Gate family was empty by derivation (zero changed files) and the dev said so rather than running gates for show; it ran check-changeset-presence anyway as evidence and it agreed: "0 file(s) changed … no changeset is owed." ⭐ Proving the empty surface from the gate's own mouth is the right shape.

---

Generated by Claude Code