[objectstack-fleet]

Requested by Elon · project thread
Fixes #30479
Clause-②: no
Metadata-protocol: none

维护者速读

• 改了什么:新增「请假与假期余额」中档场景卡:7 个对象(部门、员工、假期类型、假期余额、请假单、请假明细、审批步骤),3 天及以内只需经理审批、超过 3 天再加 HR 审批,批准后自动扣减余额;3 个角色其中 2 个按行限定;1 个请假天数报表;120 张请假单跨 7–10 月;10 道问数、6 个写操作、4 个 agent 动作。另配一份只属于这张卡的真值测试。
• 为什么改:落实 cloud#3600 九个场景中的 HR 场景,让 AI 建应用、问数、代办写操作都有一张能用代码判分的请假与余额卡。
• 风险与回滚:只加测试数据和测试,所在包私有、不发布,运行时零改动;回滚即删两个新文件。
• 先看:① docs/qa/scenarios/leave_balances.json 的 run.operations(六个写操作及预期结果);② 同文件的 run.questions 与 run.personas;③ packages/qa/scenarios/test/leave-balances.test.ts 的 EXPECTED_TRUTHS。
• 两分钟验证:pnpm --filter @objectstack/scenarios test,预期 5 个测试文件、73 个用例全绿。
• 需要你做的:无,席位落地。

What changes

• docs/qa/scenarios/leave_balances.json (new): the 请假与假期余额 leave-and-balances mid-tier card.
  • Seven objects: department, employee, leave_type (annual, sick, personal), leave_balance (one per employee, tracked type and leave year), leave_request, leave_day (one line per date, a full day or a half day) and approval_step.
  • The route follows the days: up to 3 days goes to the department manager only, more than 3 days to the manager and then HR (e13), and a manager's own request goes to the HR manager (e14) at step 1. On the last approval the balance's used_days (a roll-up of the approved requests) rises and remaining_days (entitlement + carry-over - used) falls.
  • Two validation rules: a request cannot be submitted for more days than its balance has left, and a sick request over 2 days needs a certificate.
  • Three personas: employee (own rows) and dept_manager (own department) carry a row scope; hr_officer sees everything except the medical note, which only the requester sees.
  • Seed: 120 requests starting July to October 2026, with 307 day lines, 141 approval steps and 28 balances. The rule that generated it is stated in the card's $comment, so every truth can be worked by hand.
  • Run half: 10 questions, 6 operations and 4 agent actions (one confirm-gated); iterate adds time off in lieu.
• packages/qa/scenarios/test/leave-balances.test.ts (new): this card's own test file.
  • It pins all 10 computed truths and the 9 per-persona truths against hand-worked values, with the derivations beside them.
  • It asserts the mid-tier floor and checks the seed against the rules the card states: a request's days equal its weekday lines, a balance equals its approved requests and the tenure rule, each request's steps match its days and status, and both validation rules hold on every submitted request. It also checks that LR-0091 really exceeds its balance, so the refused submission is a real refusal.
  • It checks that every where in an applied post-state matches no seeded row, so it can only name the row the write creates.
  • It shows that a wrong seed cell and a wrong truth spec in a fixture copy go red against those pins.

Why

#30479 is one of the six mid-tier cards of the open scenario library. This one covers HR: leave requests whose approval route depends on their length, and balances that move when a request is approved.

Tests

pnpm --filter @objectstack/scenarios test: 5 files, 73 tests passed. pnpm --filter @objectstack/scenarios typecheck: exit 0, after building the closure with pnpm --filter '@objectstack/scenarios^...' build. eslint --no-inline-config on the new test file: exit 0.

Acceptance notes

• File names. The card is leave_balances.json, not leave-balances.json as the issue spells it: a card's id is a snake_case machine name that must equal the file's basename. The test file follows the model's spelling, leave-balances.test.ts.
• No shared-file edit. Since travel_expense landed, library.test.ts checks only that the three control cards are present, and the README's "Adding a card" step points at a per-card test file. So this PR touches neither, and the sibling cards no longer conflict on a shared line.
• No schema field for some floor items. As with travel_expense, the approval flow, "declarative preferred" and a report as distinct from a dashboard have no field in the schema. The card states them in prompt.detailed (zh and en) and in capabilities, seeds approval_step as an expected object, and declares the report under expect.dashboards. Nothing in the loader grades them.
• Capability tags. Where a tag in the closed tag table of the open scorer PR feat(scenarios): the scenario scorer moves into @objectstack/scenarios and judges capability tags, agent-action exposure, write outcomes and per-persona answers #30594 fits, the card uses that spelling (approvals.approval-flow, access-security.row-scope, dashboards.widget-measures, records-forms.lookup), so those tags are judged once the scorer lands. Tags outside that table read unscored there, not failed.
• Seven operation kinds, six operations. file_and_submit_from_text covers two of them: entry from free text with lines, and the multi-step chain (file the request, add a full day and a morning, submit, route to the manager).
• Routed approval step not asserted. The post-state of file_and_submit_from_text does not assert the routed step-1 row. A where cannot name the new request's id, and approver e05 · pending also matches seeded rows.
• Dates. asOf is 2026-09-30, so October's requests are the upcoming ones and every draft or pending request is for future leave; "last month" is August. Balances belong to leave year 2026/27, which starts on 1 July, so the seed is the whole leave year so far. Days count Monday to Friday; public holidays are out of scope and the $comment says so.
• Reference. objectstack-ai/templates is not attached to this session, so the domain is designed from the card text alone.

Verification

scenarios test + typecheck

 Test Files  5 passed (5)
      Tests  73 passed (73)
os-verify-lock: VERDICT command-exit 0
> tsc --noEmit
os-verify-lock: VERDICT command-exit 0

one-time check: a wrong cell in the real card goes red

scripts/ablation-replace.mjs set LR-0098's days from 5 to 6 in docs/qa/scenarios/leave_balances.json and ran this card's test file: 4 of 9 tests failed (the pinned truths, the lines check, the balance check, and the fixture-copy pin, which reads the real card). It then restored the file: blob equal to HEAD, git diff HEAD empty.

dispatch-gates --commands (50 commands, local, head 42b3e57)

All 50 derived commands exit 0 on head 42b3e57a (the commit before the main merge). The --ran reconciliation reads: 50 derived, 50 run, 0 NOT-MEASURED, 0 UNRUN, a derived zero because every line recorded its exit code.

Four of them first exited 3 (PREREQUISITE NOT MET). Each exited 0 after the step it names:

• check-plugin-teardown-shape.mjs --self-test, after git fetch --unshallow origin main (the clone was shallow);
• check:doc-formula-expressions, check:lean-entry-closure and check:dual-build-cjs-loads, after a full turbo run build (119 of 119 tasks).

The merge of main (9 commits) touches neither docs/qa/scenarios/ nor packages/qa/scenarios/, and the derivation on the merged head 9bd3d3c9 names the same 50 commands. On that head the scenarios tests (5 files, 73 tests) and typecheck pass again.

Changed lines: 1342 (+1342 / -0, 2 files), under the 3000 threshold.

Human Minimum

• 1. Schemas start as Zod; packages/spec holds no runtime logic. (Not touched.)
• 2. Every name I added uses the casing for its kind.
• 3. No formatter ran; pnpm lint passes. (eslint on the new test file locally; the repo-wide run is CI's.)
• 4. One issue, one branch off an up-to-date main, one pull request.
• 5. A change that publishes carries a changeset, and both declaration lines are filled in here and in the changeset. (@objectstack/scenarios is private: true and publishes nothing, so no changeset, skip-changeset, as travel_expense landed.)
• 6. packages/spec touched: its generated artifacts are regenerated and committed. (Not touched.)
• 7. No hand edits to content/docs/references/, packages/*/CHANGELOG.md or content/docs/releases/.
• 8. Title, body and comments are in English.
• The Before Submitting list in CONTRIBUTING.md is done.

---

Generated by Claude Code

[github-actions]

📓 Docs Drift Check

Nothing in this diff resolved to a documentable surface (no symbol, route or SDK anchor derived from 0 changed package(s)), so this run has no opinion about the docs.

What this run could not see

• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.
• a key NAME is not a key, so the hand re-read the line above prescribes can land on the wrong schema. The same spelling is authorable on one governed type and a [REMOVED] tombstone on another for each of active, aria, joins, objects, template, tools and version (censused on [finding] tools is a key on BOTH AgentSchema (tombstoned, dead) and SkillSchema (live, cloud-attested), so a name-based search attributes skill examples to the agent key — it produced a false stop-the-line alarm on PR #19059 #19093 over the liveness ledger's governed types, top-level keys); nothing in a search result distinguishes the two, so a grep hit on a LIVE example reads as evidence about the DEAD key. Measured on fix(spec): the agent.tools liveness row says dead — it claimed live on a key the schema tombstoned #19059: content/docs/ai/agents.mdx was reported as contradicting the agent.tools tombstone over its tools: example at :161, which is inside the defineSkill({ block opened at :155 — the page was already correct. Settle ownership by PARSING the value against both schemas, never by the name: that literal PASSES SkillSchema, and as an AgentSchema it FAILS at tools with the tombstone prescription. ⛔ These names are not the whole class — a key retired through a .strict() guidance map leaves no tombstone in the walked shape and none of them here (tool.category, live as AIToolDefinition.category).

Coarse fallback — 0 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json a2437b560bd8872feece6ec8b57c6fdf21cff1f4 → packageMentionDocs.

[objectstack-fleet]

domain:devx#3 · os-elon-musk · session_016WV4cz3RTQjLh7hZBLzgtZ

席位核对 · ACCEPT · 9bd3d3c9 · 2026-11-05T23:01Z

Head-sha: 9bd3d3c99f989426a61ee20cd1463888e6716d02

• Shape: 2 files, +1,342/−0 (< 3,000), both on the amended claim surface (docs/qa/scenarios/leave_balances.json, packages/qa/scenarios/test/leave-balances.test.ts). Not governed. Fixes #30479, Clause-②: no, Metadata-protocol: none at line start.
• Premise measured: schema from feat(scenarios): scenario-card schema + loader (private packages/qa package) and nine open seed cards #30340 on main; test(qa-scenarios): add the travel_expense mid-tier scenario card #30600 left library.test.ts and the README card-agnostic, so no shared line is touched.
• Gates: check-runs on this head read by the seat: 35 success, 18 skipped (path filter), 0 failed or pending. Dev's local dispatch-gates 50/50 exit 0, --ran 0 NOT-MEASURED.
• Spot checks (seat's own arithmetic against the pinned truths): remaining_annual_by_department Engineering (15+3−4) + (10+5−13.5) + (5−3) + (10+2−9) = 20.5 ✓, Operations 9 + 2.5 + 3.5 + 3 = 18 ✓, Sales 17 + 1 + 8 + 5.5 = 31.5 ✓, HR 12 + 20 = 32 ✓; lowest_annual_remaining e02 1, e06 1.5, e07 2 ✓. The acceptance "a deliberately wrong truth goes red" is two cases (wrong seed cell, wrong truth spec).
• Accepted deviations: snake_case basename (id = basename, as test(qa-scenarios): add the travel_expense mid-tier scenario card #30600); the routed step-1 row of file_and_submit_from_text is not asserted in the post-state (the routing rule itself is pinned by the "routes each request by its days" case).
• Out of scope: none. Skills-docs: n/a.
• Changeset: skip-changeset, the package is private (as test(qa-scenarios): add the travel_expense mid-tier scenario card #30600).
• Landing: non-governed; flips when the maintainer's landing word for this term is on record.