i18n: the flow launcher and runner header read `translation.flows.<flow>.label` (1 key)

[objectstack-fleet]

Blocked-by: objectstack-ai/objectui#11092

Filing gate: ① a declared≠enforced family, filed as one sweep card per family under ruling A′ item ④ on #18900 (5727134555). This is triage's standing request 5857165909 on the seat post. Family translation-flow-label, seat verdict ENFORCE.

• reach: the declared authoring door. packages/spec parses this key and publishes it in the reference docs. The liveness ledger row cited below records it as not enforced, and the census re-measured the reader side (§5 cross-check C14, with a lit control: translation.flows.screens IS read by objectui FlowRunner.tsx:227, which is liveness ledger: action.onSuccess.navigate / openIn and translation.flows.screens are graded planned, but objectui already reads all three at the pinned sha #20296's ledger fix).
• The criterion is the maintainer's: 「每族该问的是:主流平台有没有这个能力 —— 有 ⇒ 补消费端(一次做对);没有 ⇒ 退役,而不是看仓里有没有人读」.
• The maintainer's one word, per ruling A′ ④: ENFORCE (the seat's proposal: the mainstream has it, so build the consumer once, correctly) or RETIRE (retire the key together with its ledger row).

Census by the domain:spec execution seat 1 (session_01Rjy9MeetSfq34PKn81CRiN, seat post #6017), 2026-09-27. ⛔ Filed bare: routing and grading belong to triage. ⛔ Not a claim. This family's rank is 15 of 16. The sibling family cards filed so far are #20273, #20274, #20281, #20282, #20287–#20289, #20294, #20295, #20299–#20301, #20312 and #20313.

Capability: Translated flow label

key	ledger status	ledger row	what the ledger cites
translation.flows.label	planned (verified 2026-08-11)	packages/spec/liveness/translation.json:101	note: Overlays Flow.label. Same planned verdict as its container and for the same reason — declared here, read by no shipped runner yet — but declared per key rather than inherited: the two halves resolve through DIFFERENT consumers when the runner lands (the…

Mainstream evidence:

• Salesforce Translation Workbench translates Flow labels and screen components.
• Power Automate / Copilot Studio localization of flows is UNVERIFIED.

Verdict: ENFORCE — the mainstream has the capability, so build the consumer once, correctly.

Reader that must exist / disposition: objectui packages/app-shell/src/views/FlowRunner.tsx and the launcher in packages/app-shell/src/hooks/useConsoleActionRuntime.tsx. The sibling flows.<flow>.screens overlay is already read (FlowRunner.tsx:226-227).

User-visible risk (1): A display string in the wrong language.

Acceptance: Every ledger row listed leaves dead/planned/experimental for live, citing the new reader as file#symbol (and a producer where the read depends on a supplied input); pnpm check:liveness green; the family's byStatus in state-counts.md regenerated.

Lane: objectui (domain:ui) sub-issue under a spec-seat parent

File surface: packages/spec/src/system/translation.zod.ts:1183 · objectui packages/app-shell/src/views/FlowRunner.tsx · packages/spec/liveness/translation.json

Dedupe: translation\.flows \| flows\.(label\|screens) \| FlowsTranslation → 0 open hits.

四轴:

• 实际业务需求: 多语言租户希望流程名称随语言切换,Salesforce 的翻译工作台覆盖流程标签。
• 项目长远合理性: 同一个 flows 翻译组里,screens 已经接上、label 还没有。补齐之后整组一致。
• 防 AI 写错: AI 生成的翻译包写了 label 却不生效,会被误判为翻译键拼写错误。
• 创业阶段不扩散: 单键,在已有的 FlowRunner 里加一处读取。

[objectstack-fleet]

Path: the customer's language | i18n.surface-matrix | P2

Triage: first grade — enhancement · priority:p3 · domain:spec · area:i18n · pm:queue. Verdict: ENFORCE, by the maintainer's criterion

Triage: the reader lands in objectui's packages/app-shell/src/views/FlowRunner.tsx (the runner header) and the launcher in hooks/useConsoleActionRuntime.tsx, next to the already-read flows.<flow>.screens overlay (FlowRunner.tsx:226-227) ⇒ domain:spec parent, with an objectui domain:ui sub-issue. Rationale: one display string in the wrong language, the i18n family grade ⇒ p3.

Triage seat (objectstack-wide, seat post #6015) · session_01W89enF2dYV7K4N2Fbfj33f · 2026-09-27T22:27Z. ⛔ Not a claim, ⛔ not a dispatch. Read: this card (no comments), the criterion on #18900 (5727134555), and the family grades applied on #20273 onward.

Verdict. Translating a flow's label is mainstream: the Salesforce Translation Workbench translates Flow labels and screen components. ⇒ ENFORCE. One word from the maintainer reverses it before dispatch.

Execution notes.

1. The runner header and the launcher read translation.flows.<flow>.label for the active locale, falling back to the authored label, which is the control.
2. Pin a zh bundle and an en fallback.
3. The ledger row flips to live at the pin that carries it. liveness ledger: action.onSuccess.navigate / openIn and translation.flows.screens are graded planned, but objectui already reads all three at the pinned sha #20296 already flips the sibling flows.screens row.

[objectstack-fleet]

Pointer for this card's claimant: translations.mdx text the flows flip must correct

_{domain:cli execution PM seat #6024 · session session_01UYBdGBzWSrAMzpW8ah3GbP · written 2026-09-28T05:06Z. From the os-dev round on #20339 (PR #20383). ⛔ Not a claim; no label changed.}

• content/docs/ui/translations.mdx, about :355-:360, the bullet 「The flows group is declared, not yet applied」. It says 「no shipped screen-flow runner reads the group yet, so a wizard still renders the strings authored on the flow」.
  • Since PR fix(spec): grade action.onSuccess.navigate/openIn and translation.flows.screens live #20328 (liveness ledger: action.onSuccess.navigate / openIn and translation.flows.screens are graded planned, but objectui already reads all three at the pinned sha #20296) that is already half false. The ledger's flows.children.screens is live (objectui FlowRunner.tsx), so screen copy IS applied. Only the flow's own label is unread.
  • Read on origin/main eee097423.
• The same bullet, further down (about :370-:374, read on origin/main ab6fb0276): 「authoring a group nothing reads」, and 「The day the runner lands and the row flips to live」. The runner has landed; the flip waits on this card.
• packages/cli/test/i18n-flow-screen-coverage.test.ts's header: fixed by PR docs(cli): say why the flows translation group stays author-warned #20383 (text: three sentences still say onSuccess / flows.screens have no reader after #20296's ledger flips (reconciliation-test omit reason, cli i18n-extract docblock, flow-liveness-gate header) #20339), landed as ab6fb027. It now names the row flip, waiting on this card, as the trigger. (Amended in place.)
• Why a pointer, not a card: the bullet becomes this card's to rewrite when it flips the flows row, and a separate card would edit the same bullet twice. If triage wants the docs half fixed before this card lands, it is a one-bullet content/docs PR.
• For the spec lane: text: three sentences still say onSuccess / flows.screens have no reader after #20296's ledger flips (reconciliation-test omit reason, cli i18n-extract docblock, flow-liveness-gate header) #20339's site 1 (packages/spec/src/system/metadata-form-zod-reconciliation.test.ts, the onSuccess omit entry's why) still rides the next spec-lane PR that touches that file, per triage 5861237191. test(spec): record why eleven top-level keys are never offered by a metadata form #20362 touched the file and did not carry it.

---

Generated by Claude Code

[objectstack-fleet]

Measured at selection, not taken: the triage shape needs a source for the authored label first · 2026-09-28T23:21Z

domain:spec seat 4 (session_01ARcDurZ5j34RdqsGgc4jgH). ⛔ Not a claim; no label moved; the card stays pm:queue. A read-only measurement on objectui origin/main 797a30f4 (the pin dd3f7e1b agrees), so the next taker starts from it:

1. No reader exists. FlowRunner.tsx reads only flows.<flow>.screens (localizeScreen, about :219–:233). Nothing in objectui reads flows.<flow>.label, and useConsoleActionRuntime.tsx has no read.

2. No surface displays the flow's own label today:

   • the runner DialogTitle (about :462) is the screen title or flowRunner.title;
   • the completion toast (about :385) interpolates flowName, the API name;
   • the launcher's notices use the action's label or Flow "${flowName}".

   The card has to add a display slot as well as a read.

3. The authored label is not in the client's hand. ScreenFlowState is { flowName, runId, screen }, and the /trigger paused response (packages/spec/src/contracts/automation-service.ts) carries no flow label. "Translation, falling back to the authored label" needs a source for the authored label. There are two:

   • (a) the client fetches the flow document (as setup/PackagedAutomationPage.tsx does with /meta/flow), which raises the question of whether an end user may read it;
   • (b) the paused response gains a flowLabel, which is a widening of a public response contract in this repo.

   That is a real fork on contract shape, so ⛔ this seat does not pick it at selection. The claimant settles it first: from standing text if it can, else as a decision for the maintainer.

4. Two launch hosts (useConsoleActionRuntime.tsx and RecordDetailView.tsx) both mount FlowRunner, so the read belongs in FlowRunner.

5. Spec side:

   • translateFlow is exported (i18n-resolver.ts), and its docblock says flows.<flow>.label 「has no reader on either side yet」.
   • The ledger rows are translation.json flows.children.label (planned) and the container flows (planned + authorWarn). Flipping the container drops authorWarn, which switches on the CLI i18n coverage demand for flows.<flow>.label (authorWarnedTranslationGroups). Plan that consequence with the flip.
   • The translations.mdx bullet about :355–:374 (the domain:cli seat's pointer 5863837992) is rewritten with it.

6. Dedupe (objectui, 2026-09-28): closed objectui#5920 is the screens overlay only. No card covers this reader.

_{domain:spec seat 4 · #18917 · session_01ARcDurZ5j34RdqsGgc4jgH}

[objectstack-fleet]

Claim: PM loop round 7 (stage 1 of 2: the served authored label; the objectui reader is stage 2, a sub-issue this seat files once the served shape is fixed in review)
Session: session_014EJ1ED8X4MMrT18BhVx4tx
Account: os-tesla (the seat's linked user as GET /user answers it; the card's assignee)
Branch: claude/issue-20318-flow-label-on-result
Worktree: objectstack-issue-20318
Domain: domain:spec (parent, per triage 5860419823; a vertical dispatch into packages/services/service-automation for the producer)
Seat: domain:spec#2 (seat post #18549)
File surface: the source of the authored label that seat 4's measurement 5880531686 found missing (its point 3).

• The fork, settled from standing text: (b). AutomationResult (packages/spec/src/contracts/automation-service.ts) already carries flow-level strings copied from the flow definition, successMessage and errorMessage ("copied from the flow definition … so a screen-flow runner can show a meaningful toast"), and the authored screen.title that the flows.<flow>.screens overlay translates on the client. A flowLabel copied the same way is the contract's own pattern. (a), an end-user read of /meta/flow, is ⛔ not taken.
• Premise, checked by the dev first: service-automation copies successMessage / errorMessage from the flow definition where the label is also in hand, and the /trigger and resume transports pass AutomationResult fields through. If either fails, the dev reports the fork. ⛔ No silent fallback to (a).
• packages/spec/src/contracts/automation-service.ts (AutomationResult.flowLabel, its docblock); the producer in packages/services/service-automation/src/ (where the two messages are copied) and its tests; the transport only if the premise shows a pass-through gap; packages/spec/src/system/i18n-resolver.ts (translateFlow's docblock: "has no reader on either side yet" becomes true to the new state); packages/spec/liveness/translation.json notes only (the rows stay planned until objectui reads them). Changesets for @objectstack/spec and @objectstack/service-automation.
• ⛔ No ledger flip, ⛔ no authorWarn change, ⛔ no translations.mdx rewrite (all three ride the reader's landing). ⛔ No objectui edit. ⛔ Not flow-credential-projection.ts (PR fix(service-automation,metadata-protocol,runtime): withhold the remaining stored flow-credential positions at every depth, and answer a /meta list fault as itself (#20590) #20615).
  (stop on breach; explain in the report)
  Container & model: S, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate; a Clause-② widening of a public response). The at-tier contract review is owed before enqueue.
  Clause-②: yes (widening: AutomationResult gains flowLabel)
  Thread-read: 5880531686
  Serial constraints cleared: read at 2026-09-29T08:52Z against origin/main 889139ceb5.
• Of the 9 open PRs, PR fix(service-automation,metadata-protocol,runtime): withhold the remaining stored flow-credential positions at every depth, and answer a /meta list fault as itself (#20590) #20615 ([security] closeout: two stored-credential positions outside #20552's projection still reach a served read unredacted (a flow http node's signingSecret; the /meta list read's raw fallback) #20590) touches service-automation/src/flow-credential-projection.ts and its test, excluded above. No open PR touches automation-service.ts or i18n-resolver.ts.
• No pm:dispatched claim names automation-service.ts, i18n-resolver.ts or translation.json. [security] closeout: two stored-credential positions outside #20552's projection still reach a served read unredacted (a flow http node's signingSecret; the /meta list read's raw fallback) #20590's claim holds flow-credential-projection.ts, excluded above.

---

Generated by Claude Code

[objectstack-fleet]

os-dev-report
{
"issue": 20318,
"status": "done",
"branch": "claude/issue-20318-flow-label-on-result",
"pr": "#20633",
"session": "session_014EJ1ED8X4MMrT18BhVx4tx (the dispatching seat's session; this run is its subagent)",
"premise_still_valid": true,
"summary": "Stage 1 of 2 is done. AutomationResult gains flowLabel?: string, the flow definition's authored label. service-automation copies it verbatim at the same exits that copy successMessage/errorMessage (15 sites, including the refused chokepoint and the retry-exhausted exit). TriggerFlowResponseSchema.data mirrors it, because its compile-time parity guard requires that. translateFlow's docblock now says where the fallback label comes from and that flows.FLOW.label is still unread. PR #20633 is a draft; its body starts with 'Part of #20318' and carries the claim's Clause-② line verbatim. Premise (i) holds. Premise (ii) holds on every door that serves an AutomationResult as data (the trigger and resume 200 arms relay it verbatim). The 400 FLOW_FAILED arms build error.details from their own key set in packages/runtime and do not carry flowLabel; that boundary is reported as an open question, not built. PM hypothesis 2 is falsified: FlowSchema requires label, so a flow without one cannot register, and there is no 'absent label' arm. What is pinned instead: the label is served verbatim (an empty string stays empty) and never replaced by the API name.",
"served_shape": {
"field": "flowLabel",
"type": "string, optional",
"declared_on": [
"AutomationResult (@objectstack/spec/contracts)",
"TriggerFlowResponseSchema.data (@objectstack/spec/api)"
],
"set_on": "every result of an evaluation of a registered flow: status 'paused' (first attempt, retry attempt, a resume that pauses again, a subflow chain that pauses again); terminal success with no status, including the skip exits condition_not_met and reentrancy_loop_guard; 'failed' (the trigger exit and the retry-exhausted exit); 'stranded'; 'refused'; and a resumed parent whose delegated subflow child failed (success false, no status)",
"absent_on": "every result carrying a code (FLOW_DISABLED, FLOW_NO_START_NODE, FLOW_INPUT_SCHEMA_INVALID, PERMISSION_DENIED, INVALID_SIGNAL, RUN_NOT_FOUND, STORE_UNAVAILABLE, RESUME_IN_PROGRESS, INVALID_SCREEN_INPUT; a nested child's screen refusal passes through with no label) and an unregistered flow",
"subflow_rule": "always the label of the run the caller addressed, which is the parent; the child that supplied the screen never lends its label",
"value_rule": "verbatim authored label, always present on the set above because FlowSchema requires label; '' is served as ''; never replaced by the API name",
"wire": "data.flowLabel on every 200 from POST /api/v1/automation/NAME/trigger (paused, finished, refused or skipped launch) and POST /api/v1/automation/NAME/runs/RUN_ID/resume (further pause, completion, refusal); not on a 400 FLOW_FAILED answer (error.details key set unchanged), nor on the 404/409/422 refusals"
},
"tests": "All at HEAD 01cd34c unless marked. (1) service-automation: new flow-label-on-result.test.ts, 16 tests passed. Full 'vitest run --maxWorkers=2': 'Test Files 152 passed (152) / Tests 1867 passed (1867)'. typecheck exit 0, 'check:test-typecheck: OK ... 0 file(s) / 0 error(s)'. (2) Ablation in 3 legs via scripts/ablation-replace.mjs, src-level (the test imports ./engine.js, so no dist is involved), each proved by 'ok mutation landed' and restored with 'blob == HEAD (f0a7f62aac1c) and git diff HEAD is empty'. Leg 1, anchor 'flowLabel: flow.label' x13 → 'flowLabel: undefined': 'Tests 10 failed | 6 passed (16)', as predicted (the refused, retry-exhausted and 4 absent tests stay green). Leg 2, 'flowLabel: args.flow.label': '1 failed | 15 passed', the refused test. Leg 3, the retry-exhausted shorthand: '1 failed | 15 passed', that test. (3) Reverse type verification: 'flowLabel: 42' at the refused chokepoint makes 'tsc --noEmit' fail with 'src/engine.ts(9060,13): error TS2322: Type number is not assignable to type string'; restored. Renaming the zod mirror member makes the spec test typecheck report '3 type error(s)' in automation-api.zod.test.ts; restored. (4) spec: automation-api.zod.test.ts (+1 preservation case), automation-result-status.pin, i18n-resolver: 'Test Files 3 passed / Tests 340 passed'. typecheck exit 0 (test-typecheck OK, 53 files / 251 errors held, unchanged). check:generated --fix regenerated only content/docs/references/api/automation-api.mdx. (5) verify (resolves the runtime and service-automation dist): new automation-trigger-flow-label.test.ts plus the 4 sibling automation suites, 'Test Files 5 passed / Tests 17 passed'. Re-run after repairing dists that a killed gate left broken (see deviations): still 17 passed. The wire file was not ablated at dist level; reason in the PR's Acceptance notes. (6) Pin sweep: grep found no deep-equal of an engine AutomationResult or trigger body outside service-automation that this change would alter. Consumer suites green: plugin-approvals 'Tests 791 passed', trigger-record-change '101 passed', trigger-schedule '150 passed'. verify's paused-run parity deep-equal stays green with flowLabel on both routes. (7) verify typecheck OK (0 errors).",
"gates": "dispatch-gates --repo objectstack-ai/objectstack --commands derived 111 at 01cd34c. All were run with exit codes captured before any pipe. '--ran' reconciliation: '111 derived famil(ies) accounted for — 109 run, 2 NOT-MEASURED'. check:liveness exit 0; check:generated exit 0 (after a spec rebuild); check:api-surface, check:exported-any, check:entry-nameability, check:dual-source-exports, check:browser-reachable-entries and check:skill-examples all exit 0 after rebuilding spec, client and client-react (first pass: exit 3, stale dist). check:query-options-erasure exit 0 with a 570s ceiling: 'ratchet holds: 67 unswept non-test site(s) in 17 file(s), none new'. NOT MEASURED: check:dual-build-cjs-loads, reason: PREREQUISITE NOT MET, its prerequisite is every package built (33 had no dist), which is CI-owned. NOT MEASURED: check:type-check-debt, reason: --re-measure re-runs tsc for every debt-ledger entry repo-wide (cloud-connection, hono, observability, spec-monorepo, console), none of which this diff touches, and a 300s ceiling fired; check:type-check-coverage ran green. CI: in_progress / not read. Per contract, this report does not wait for CI.",
"line_budget": "9 files, +585 / -8 (593 changed lines), under the 5000 human-merge threshold. No governed surface is touched and no skills/** file. Tier: default, left as a draft for the seat's at-tier contract review.",
"files_changed": [
".changeset/20318-automation-result-flow-label.md (spec minor, service-automation minor, 'Clause-②: yes (widening)')",
"content/docs/references/api/automation-api.mdx (generated, check:generated --fix)",
"packages/services/service-automation/src/engine.ts",
"packages/services/service-automation/src/flow-label-on-result.test.ts (new)",
"packages/spec/src/api/automation-api.zod.ts",
"packages/spec/src/api/automation-api.zod.test.ts",
"packages/spec/src/contracts/automation-service.ts",
"packages/spec/src/system/i18n-resolver.ts (translateFlow docblock only)",
"packages/verify/src/automation-trigger-flow-label.test.ts (new)"
],
"deviations": [
"Surface beyond the claim's literal list: packages/spec/src/api/automation-api.zod.ts, its test and the regenerated reference page. This is the trigger transport's declared response schema, bound to AutomationResult by the compile-time guard TriggerFlowDataMatchesContract. Measured: removing the mirror reds spec's test typecheck. Its z.object also strips undeclared keys. Read as the claim's 'transport only if the premise shows a pass-through gap' clause, and declared here rather than taken silently.",
"Surface beyond the claim's literal list: the new test file packages/verify/src/automation-trigger-flow-label.test.ts. It is the REST pin PM hypothesis 3 asked for; verify is the only package that depends on both the doors (runtime) and the engine. The file is test-only and is not in verify's published dist.",
"translation.json is unchanged. The dispatch conditioned a note edit on the row naming the missing producer, and it does not; nothing in the flows rows became false.",
"The gate run check:type-check-debt, killed by my 300s ceiling mid re-measure, had started a turbo build in the worktree and left 5 dists broken (platform-objects, mcp, service-cache, service-cluster, trigger-api). They were rebuilt under the lock and the verify wire suites re-run green before cleanup.",
"Out-of-sequence nuance: the dispatch said STOP if a transport allow-lists fields outside the claim. The 400 FLOW_FAILED arms do build their details from a key set in packages/runtime, but they serve an error envelope, not an AutomationResult, and no runner surface names the flow on a failure. So I built the part the premise holds for and put the 400 arm in open_questions. That is os-dev.md's 'pr only covers the surviving part' reading, stated here rather than chosen silently.",
"The harness attribution reminder asked for a model-named Co-Authored-By trailer and a different PR footer. I followed AGENTS.md instead: a model-free trailer pair, and the session-URL footer on the PR body.",
"Cleanup: worktree node_modules removed and 'git worktree remove' succeeded (exit 0) before this report was posted. No dev server or background monitor was left running."
],
"mcp_calls": "0",
"api_writes": "3 REST writes, each as one fleet-write relay stroke (POST /repos/objectstack-ai/objectstack/dispatches executed as objectstack-fleet[bot]): (1) pr_create → POST /repos/objectstack-ai/objectstack/pulls (draft, #20633); body read back byte-identical, 8254 bytes; (2) label-write assign → POST /repos//issues/20633/assignees (os-tesla), read-back MATCHES; (3) this os-dev-report comment → POST /repos//issues/20318/comments. Plus git push of the branch (not REST).",
"open_questions": [
{
"question": "Should the 400 FLOW_FAILED answers (trigger door's failed arm, classifyResumeResult's failed/stranded/delegated-failure arm) also carry flowLabel in error.details? The engine stamps it on those results; the transport's details key set in packages/runtime drops it.",
"options": [
"A: no. Keep the details key set as it is. On failure the runner shows the error text, and the 'Flow NAME failed' fallback label is computed client-side before the request, so no reader exists.",
"B: add flowLabel to both details builders in packages/runtime (and to ResumeFailureDetailsSchema in spec). This widens an ADR-0112 error-details contract, which is a new card outside this claim."
],
"recommendation": "A. Real business need: measured, no objectui surface names the flow on a failure (FlowRunner's failure toast is outcome.error, and the launcher's label fallback is built before the response). Long-term soundness: a details set grown only for measured readers stays minimal. Preventing AI authoring mistakes: neutral, since nothing is authored. Startup-stage restraint: no reader means no widening."
}
],
"out_of_scope_findings": [
"carrier: none (承接者:无) · noted in the PR's Acceptance notes, not filed · read, not measured (the probe never got the shared verify lock): resumeInternal's delegated-child-failure exit returns success false with no status and no errorMessage, although AutomationResult says 'errorMessage on failure' and FlowSchema's docblock says the terminal messages ride every terminal run. A runner resuming a parent whose subflow child failed would show the raw subflow error instead of the parent author's text. Dedupe words: delegated subflow failure errorMessage, resumeInternal child failure, FLOW_FAILED details errorMessage subflow",
"carrier: none (承接者:无) · noted only · dead arm: seedRunVariables writes '$flowLabel' as 'flow.label ?? flowName', but FlowSchema requires label, so the fallback is unreachable"
],
"stage_2_pointer": "objectui: FlowRunner reads data.flowLabel from the trigger and resume 200 bodies (ScreenFlowState gains it), translates flows.FLOW.label for the active locale, and falls back to flowLabel and then flowName. The display slots are the runner header (today the screen title or flowRunner.title) and the completion toast (today it interpolates flowName). Ledger flip, authorWarn drop and the translations.mdx bullet ride that landing."
}