Two `packages/spec` route response schemas declare a `data` NARROWER than the contract their route relays — `AnalyticsResultResponseSchema` and `TriggerFlowResponseSchema`

[os-litant]

Found while binding the five in-repo return res.json() methods of #12104. Out of that
card's declared scope and deliberately not fixed there — that card binds client return types;
this is a producer-side declaration defect in packages/spec, and it changed what #12104 could
bind.

ⓘ Angle brackets are spaced throughout (Promise< any >) on purpose — GitHub's body sanitizer
strips tag-shaped spans, backticks and fenced code included.

The defect

Both schemas below look like the authoritative response contract for their route. Both declare a
data that is a strict subset of what the route's producer declares and actually serves. A
consumer that binds either — which is exactly what a sweep reaches for — ships a declaration
that refuses reads the wire really carries.

1. AnalyticsResultResponseSchema (packages/spec/src/api/analytics.zod.ts)

POST /analytics/query ends deps.success(await analyticsService.query(body, ctx)), so data
is IAnalyticsService.query's declared return, AnalyticsResult
(packages/spec/src/contracts/analytics-service.ts).

declared on the schema	declared on AnalyticsResult
rows	rows
fields: { name, type }[]	fields: { name, type, label?, format?, currency?, percentScale? }[]
sql?	sql?
—	totals?: { dimensions, rows }[]

fields[].label is not hypothetical: it is served by a real AnalyticsService, measured in
packages/client/src/analytics-automation-json-erasure.test.ts. currency and percentScale
are the ADR-0053 / objectui#3136 renderer chains, and totals is the ADR-0021 marginal-aggregate
channel — every one of them a key a dashboard reads.

2. TriggerFlowResponseSchema (packages/spec/src/api/automation-api.zod.ts)

POST /automation/:name/trigger (and the legacy POST /automation/trigger/:name) end
deps.success(result) where result is IAutomationService.execute's declared return,
AutomationResult.

declared on the schema	declared on AutomationResult
success, output?, error?, durationMs?	those four, plus code?, status?, runId?, screen?, successMessage?, errorMessage?, summary?

The missing members are not decoration. status: 'paused' + runId + screen is the whole
third state of the #9378 / #9510 trigger contract — the payload a caller resumes a screen flow
with — and the SDK's own docblock on automation.trigger tells callers to read exactly those.

Why this is the same shape #6442 already fixed once, in this same file

AnalyticsMetadataResponseSchema used to declare { cubes: Cube[] } for an endpoint that has
always answered a bare CubeMeta[]. The maintainer ruled (2026-08-08) "narrow the declaration",
and the schema's own docblock records the reasoning: the TS contract already agreed with the
runtime and the schema was the lone outlier. These two are the same class, caught the same
way — by reading the producer's declared return next to the schema.

Remedy shape (not prescribed — this is the question the card carries)

Either bring each schema's data up to its producer's contract, or state in each docblock that
the schema is a deliberate PROJECTION and name what a consumer must not bind it to. What must
not stay is the current state: a response schema that reads as the route's contract, is
narrower than it, and has no note saying so. AnalyticsResultResponse additionally has no
exported type at all (protocol.zod.ts keeps a module-local z.infer alias), so a consumer
cannot name it even if it were right.

What #12104 did about it, so this is not re-derived

client.analytics.query and client.automation.trigger were bound to the producer contracts
(BaseResponse & { data: AnalyticsResult } / BaseResponse & { data: AutomationResult })
rather than to these two schemas, and the near-miss is pinned in
packages/client/src/return-type-precision.test.ts so a later sweep cannot quietly retarget
them. Their two siblings, AnalyticsMetadataResponseSchema and AnalyticsSqlResponseSchema,
DO agree with their producer and were used as-is.

---

Generated by Claude Code

[os-trump]

Claim: PM loop round R7 (domain:spec seat)
Session: session_01LpRNHxWZgSUgVnFT9mQQo4
Branch: claude/issue-13078-response-schema-contract-parity
Worktree: objectstack-issue-13078
Domain: domain:spec
File surface: packages/spec/src/api/analytics.zod.ts (AnalyticsResultResponseSchema.data → producer contract AnalyticsResult), packages/spec/src/api/automation-api.zod.ts (TriggerFlowResponseSchema.data → producer contract AutomationResult), the export gap the card names (protocol.zod.ts module-local alias → a nameable exported response type, if the gates and the surface conventions support it), pin tests both directions, regenerated api-surface/export-origins as the gates demand, docblocks recording the parity rule, changeset. Stop on breach; explain in the report.
Container & model: M, mode:subagent, model: claude-fable-5 — Clause-② both legs (PATH: packages/spec/src/api/**; CONTENT: the response accept set widens to the producer contract, and a public type export may be added); dispatch at CONTRACT_REVIEW_TIER.
Clause-②: yes — binds the enqueue gate: needs:contract-review re-attaches the moment this card's PR exists.
PM ruling of route (premise-carried, recorded for the veto window): of the card's two remedy shapes, this seat rules parity — bring each schema's data up to its producer's contract, not the projection-docblock route. The recorded #6442 precedent in this same file governs: when the TS contract and the runtime agree and the schema is the lone outlier, the schema moves. Both missing member sets are served in reality (fields[].label measured in analytics-automation-json-erasure.test.ts; status:'paused'+runId+screen is the whole third state of the #9378/#9510 trigger contract that the SDK docblock tells callers to read), so a "deliberate projection" note would document a refusal of reads the wire carries — the projection route fails the same test #6442 failed. Premise: producer contracts and served reality agree on the wide shape, re-verified on the current ref as the dev's first work item. ⛔ Premise fails, or evidence surfaces that the narrowness is a deliberate projection with a named reader ⇒ report the fork on this card, never force parity and never silently switch routes.
Serial constraints cleared: in-flight #13002 (driver/ui zod prose + check-doc-authoring) and #13135 (kernel/metadata-customization coupling set) touch neither file; shared regenerated artifacts (api-surface) serialize at the landing window; #13136 is docs/adr only. No open PR claims either file (single-writer guard enforces at PR time).

---

Generated by Claude Code

[os-trump]

Dev claim — session session_01LpRNHxWZgSUgVnFT9mQQo4, branch claude/issue-13078-response-schema-contract-parity.

Scope: parity route per the PM claim — widen AnalyticsResultResponseSchema.data and TriggerFlowResponseSchema.data to their producer contracts (AnalyticsResult, AutomationResult), premise re-verified on my ref first. Files to be touched: packages/spec/src/api/analytics.zod.ts, packages/spec/src/api/automation-api.zod.ts, packages/spec/src/api/protocol.zod.ts (export gap, if conventions allow), generated api-surface/export-origins artifacts, plus pin tests and a changeset.

---

Generated by Claude Code

[os-trump]

Dead-claim recovery + tier downgrade re-dispatch (domain:spec seat, session_01LpRNHxWZgSUgVnFT9mQQo4, R7).

What happened — the third death mode, with a host signal: the fable dev terminated mid-work on rate_limit HTTP 429, host message verbatim: "You've reached your Fable 5 limit. Switch to another model, or manage usage credits… (model sent to the API: claude-fable-5)". Per the recorded reading of this death mode: dispatch is dead, the premise is intact, the card is stuck on a resource ⇒ dead-claim recovery + worktree salvage, ⛔ no premise re-verification, ⛔ no escalation.

Salvage (verified, not assumed): the work is NOT lost but was never pushed — the live worktree holds local commit 6c93b2f59 "fix(spec): widen AnalyticsResultResponseSchema and TriggerFlowResponseSchema data to producer-contract parity" ahead of the pushed branch head, with a clean tree. The dev's own last reading, carried forward as the first item for the continuation: that commit message carries a card-relationship line the squash discipline forbids on branch commits — amend before pushing.

Tier: re-dispatched at opus under the quota-exhaustion exemption (maintainer 2026-08-13, verbatim 「fable 如果用完了,可以用 opus」), whose rule states the downgrade applies to a re-dispatch as well ("重派同样可降") and requires the tier and reason be recorded on the claim — done here. Floor is opus, ⛔ no lower.

Container & model (superseding the original claim line): M, mode:subagent, model: claude-opus-5 — fable measured unavailable (429 above, twice across two agents in this lane within minutes). Clause-② is unchanged and both legs still apply; the compensating control is unchanged too: needs:contract-review attaches to the PR and STAYS.

Landing posture: this seat's own tier fuse now reads claude-opus-5 on both fields, so ⛔ this seat does not contract-review or arm this PR. On delivery the PR is left DRAFT with the gate label on both carriers and a parking note — the review belongs to the triage seat's timed round, with the director seat as the summoned fallback.

The ruled route (parity, per the #6442 precedent), its premise-carried stop condition, file surface and serial constraints are unchanged from claim 5460786256 — and per the death-mode rule the premise is NOT re-verified on this re-dispatch.

---

Generated by Claude Code

[os-trump]

Continuation claim — os-dev seat, session session_01LpRNHxWZgSUgVnFT9mQQo4 (domain:spec PM loop R7).

Continuing the quota-terminated dispatch at opus under the recorded quota-exhaustion exemption. The predecessor dev on this card died mid-work on a host-side quota 429; its work survives locally and I am carrying it forward rather than restarting.

• Branch: claude/issue-13078-response-schema-contract-parity
• Worktree: existing, tree clean, one local commit ahead of the (not yet pushed) branch
• Route: parity (bring each schema's data up to its producer's declared contract), per the ruled route in the PM claim comment

Compensating control for the tier: needs:contract-review on the resulting draft PR, which stays draft.

---

Generated by Claude Code