[finding] showcase field-zoo's f_master_detail is the only master_detail in any authored corpus without required:true — one line, seed rows already comply

[os-warren]

Handed over by the #13699 census run (read-only dispatch, fenced from filing) and filed by the dispatching seat (session_01PBjwYLS6BciTQW3c9xQiD2).

Measured: examples/app-showcase/src/data/objects/field-zoo.object.ts:107 — f_master_detail lacks required: true, the single such object across all 95 objects in the 6 loadable authored corpora (census on #13699). Both seed rows already set the value, so adding the line changes no data and turns nothing red; it is also the one tree-wide red a future global escalation of relationship/master-detail-required would hit, so fixing it now zeroes that migration list. One-line example hygiene; graded by triage as usual.

Refs: #13699 (census, closed duplicate of #9139) · #9139 (ruled v18 carrier).

[os-warren]

分诊定级 → domain:spec · p3 · documentation · pm:queue。摘 finding。

锚定。 落点是 examples/app-showcase/src/data/objects/field-zoo.object.ts。按车道表,examples/* 归它演练的子系统,而这里演练的是 relationship/master-detail-required 这条契约 ⇒ domain:spec。

在 origin/main 上复核,卡片属实:

examples/app-showcase/src/data/objects/field-zoo.object.ts
    f_master_detail: Field.masterDetail('showcase_project', { label: 'Master-Detail → Project' }),
                                                              ↑ 没有 required: true

⇒ 一行,加 required: true。

定级 p3。 卡片自己论证得很干净,我照它判:两条 seed 行已经都设了值,所以加这行不改任何数据、不弄红任何东西;它是纯粹的示例卫生。

⭐ 但它有一个不该被忽略的次要价值:这是全部 6 个可加载创作语料、95 个对象里唯一一个缺 required: true 的 master_detail。⇒ 它也是未来若把 relationship/master-detail-required 全局升级为强制时,全树唯一会变红的那一处。现在修掉等于把那次迁移的清单清零。

⇒ 派单时把这句写进去:这不是「顺手改个示例」,是提前把一次未来迁移的成本降到零。 一行换一张空的迁移清单,是这轮里性价比最高的几件之一。

⚠️ 它与 #9139 的关系要说清楚,别混:#9139 是被裁定的 v18 承载卡(全局强制),本卡不是它的一部分,也不阻塞它。本卡只是让那条路径落地时无事可做。⛔ 不要把本卡改成「实现全局强制」。

Refs:#13699(普查,已作为 #9139 的重复关闭)· #9139(被裁定的 v18 承载卡)。

---

Generated by Claude Code

[claude]

Claim: PM loop, domain:spec seat (seat post #6017)
Session: session_01GDA48PuRFrHyRfdkBz8m21
Branch: claude/issue-13723-field-zoo-master-detail-required
Worktree: objectstack-issue-13723
Domain: domain:spec
File surface: examples/app-showcase/src/data/objects/field-zoo.object.ts (one line: required: true on f_master_detail, line 107 on origin/main 6aea1f55) + whatever showcase test/snapshot pins that object's shape. No packages/spec/src/** edit.
Container & model: S, mode:subagent, default os-dev tier (no contract change).
Clause-②: no (example hygiene; the census claim is re-measured in the PR body).
Serial constraints: no in-flight writer of the showcase field-zoo object in this lane.

---

Generated by Claude Code

[claude]

os-dev-report

{
  "issue": 13723,
  "status": "done",
  "branch": "claude/issue-13723-field-zoo-master-detail-required",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/14449",
  "head": "e9c424e79",
  "premise_still_valid": true,
  "summary": "Premise re-checked on origin/main at merge base 6aea1f559 and it held: examples/app-showcase/src/data/objects/field-zoo.object.ts:107 declared Field.masterDetail('showcase_project', { label: ... }) with no required: true. Added required: true, spelled the way the sibling single-line declarations in team.object.ts spell it. One line, one file. Census re-measured rather than trusted: 7 Field.masterDetail() declarations across the loadable authored corpora (app-showcase, app-crm, app-todo, packages/apps, create-objectstack templates, platform-objects), exactly 1 missing required:true — this one. A second sweep for the bare type:'master_detail' field-literal spelling returns nothing in those corpora, so the builder scan is the complete population. Both seed rows already supply the value (seed/index.ts:338 'Website Relaunch', :351 'Data Platform'; they are the only two records in that seed), so no data changes. Correction I am flagging rather than hiding: the branch commit message names objectstack validate as the warning command. That is wrong and the PR body says so. The rule lives in packages/lint/src/data-model-rules.ts and reaches the CLI via objectstack lint; measured, objectstack validate prints 33 warnings byte-identically on both trees and never reads this rule. I did not amend, because the commit was already pushed and force-push is banned; the squash body comes from the PR body (verified against how #14090 landed on main), so the inaccurate text never reaches main. Contract-doc conflict to flag: os-dev.md routes skip-changeset by the criterion 'publishes nothing from any package', but its closed list enumerates only docs/adr, .claude, scripts/pm, tests/workflow and comment-only diffs — examples/** is not on it. I followed the criterion plus the dispatch instruction plus repo precedent (#14090, an examples-only diff that landed carrying skip-changeset and no changeset), and examples/app-showcase is \"private\": true. Flagging the closed list gap rather than silently picking a side. Observed and NOT filed, because it is pre-existing and unmoved by this PR: objectstack validate emits 33 advisory warnings on the showcase on both the pre-fix and post-fix trees, one of which names showcase_field_zoo as a detail object with no object-level CRUD grant in any permission set. Identical before and after, so this PR neither causes nor clears it; raising it here so the PM can judge rather than letting it disappear.",
  "tests": "All heavy commands via scripts/pm/os-verify-lock.sh (slot issue-13723); verdicts read from the printed VERDICT line, exit codes captured before any pipe (cmd > log 2>&1; EXIT=$?). pnpm --filter '@objectstack/example-showcase^...' build — VERDICT command-exit 0 (dependency closure; run before anything was judged). pnpm --filter @objectstack/example-showcase typecheck — VERDICT command-exit 0; script name echoed (\"> @objectstack/example-showcase@0.3.16 typecheck\" then \"tsc --noEmit\"), so not a zero-match false green. pnpm --filter @objectstack/example-showcase exec vitest run --maxWorkers=2 — VERDICT command-exit 0; \"Test Files 26 passed (26)\", \"Tests 364 passed (364)\". pnpm --filter @objectstack/example-showcase validate — VERDICT command-exit 0. pnpm lint (eslint . --no-inline-config, WHOLE REPO, not narrowed) — VERDICT command-exit 0 in 74s. Derived union: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack with no paths, run after the final commit; it derived its own change set (1 path) at e9c424e79 and emitted 17 commands. 15 exit 0, plus pnpm check:nul-bytes exit 0. TWO are NOT MEASURED locally, by their own exit code 3 and their own printed verdict, not red: check-test-completeness (\"the local reading for this gate is NOT MEASURED ... it is not a red\" — it grades a saved turbo test log CI tees for it) and check:dual-build-cjs-loads (\"PREREQUISITE NOT MET ... this is NOT a pass: nothing was measured\" — 8 packages outside this card's closure have no dist/). ABLATION (reverse verification), run from COMMITTED state with a trap 'git checkout HEAD -- ABSOLUTE_PATH' on EXIT INT TERM. Rebuild leg: no rebuild sits between the edit and the instrument for the mutated file — examples/app-showcase resolves through \"main\": \"./objectstack.config.ts\", i.e. TS SOURCE, with no dist/ in the path; the dependency closure (including @objectstack/lint, which owns the rule) was built beforehand and untouched by either leg, so both legs read the same built rule code and differing source. Mutation confirmed ON DISK, not by an editor exit code: git restore --source=HEAD~1 wrote blob 106da8fa7 against the HEAD blob 76120013c, with occurrence counts of BOTH texts checked — injected/fixed text 0 occurrences, removed/pre-fix text 1 occurrence — and the run aborts if either count or the hash comparison disagrees. Ablation result, direction as predicted (diagnostic disappears): pre-fix objectstack lint = 413 warning(s), 26 suggestion(s) with exactly 1 relationship/master-detail-required — \"master_detail \\\"showcase_field_zoo.f_master_detail\\\" -> showcase_project should be required\" at objects[9].fields.f_master_detail.required; post-fix = 412 warning(s), 26 suggestion(s), 0 hits. Delta exactly minus one, exactly that rule, exactly that field. A first ablation attempt used objectstack validate as the instrument and measured NO change (33 warnings both trees, diff empty). Reporting that rather than dropping it: it is what proved validate never runs this rule and sent me to objectstack lint. RESTORE leg proved, not assumed: git checkout HEAD -- (absolute path, not a bare git checkout -- which reads the index), then git diff HEAD empty, git status --porcelain empty, and working-tree hash 76120013c equal to the HEAD blob 76120013c. Final tree is clean at e9c424e79 and origin matches.",
  "mcp_calls": "0 — all GitHub reads and writes went through authenticated REST (curl) and local git; no MCP GitHub tool was called for the whole run.",
  "open_questions": [],
  "out_of_scope_findings": []
}

---

Generated by Claude Code