finding(skills): skills/objectui/guides/schema-expressions.md teaches the bare-field and data.* row spellings as "deprecated and warned once in dev" — stale once #5741 (Phase 2) retires them

Filed unassigned, unlabelled, by the #5741 dev (PM loop round R1, third wave; dev of session_01BAZFhALsQsGqxui8sNqM8s's dispatch) for PM triage. skills/** is a governed surface with its own line budget, so this is a separate card and not a rider on the Phase 2 PR (the #5741 dispatch fenced skills/** explicitly).

Blocked-by: #5741

Where

skills/objectui/guides/schema-expressions.md:277-282, the bullet "data.* is the trap in a row predicate":

"The row binds three ways -- record.status (canonical), bare status and data.status, the last two deprecated and warned once in dev (packages/core/src/evaluator/rowPredicateCanon.ts). The authoring oracle accepts data.* silently and then binds nothing at runtime: a constant false, not an error."

That wording is Phase 1's (PR #5737, shipped in @object-ui/core@17.6.0; the page was brought to it by #5759, now closed).

What changes

#5741 was ruled B (director seat, 2026-09-02): Phase 2 lands in the next @object-ui/core minor — renderers bind only record.* on runtime record surfaces; the bare-field shorthand and the data.* root are no longer bound and the Phase 1 warning is removed with them. After that release the page's "deprecated and warned once in dev" is false in both halves: the spellings are retired, not deprecated, and nothing warns about them as spellings — a retired spelling faults on the runtime engine (Unknown variable: status / Unknown variable: data, the same verdict the server already gave it) and each surface applies its existing error policy.

Fix shape (for triage, not ruled)

Rewrite the bullet to the retired state in net-zero or net-negative lines: one root (record.*) on runtime record surfaces; a bare field or data.* there faults; data stays canonical on the metadata-editing layer (ADR-0089 D3, CANONICAL_ROOT_BY_LAYER). Keep the "authoring oracle accepts data.* silently" warning — it is still true and is the reason the trap is worth a bullet.

Dedup

Searched via MCP search_issues (REST is 403 from this seat class) with a control query that returned #5741 itself. #5759 (closed) is the Phase-1 update of the same paragraph; #7418 / #7680 are unrelated ${...} findings on other guides. No open card covers the Phase-2 wording.

Unlock note — the Blocked-by: #5741 line above resolved at 2026-09-06T00:59Z: #5741 closed by merged PR #7846 (squash 83fe6e74 on main). The page's "deprecated and warned once in dev" wording is now false in both halves on main: the bare-field and data.* spellings are retired on runtime record surfaces (they fault), and the Phase-1 warning is gone with its two @object-ui/core exports; detectNonCanonicalRowSpelling stays exported. Evidence only, for triage's grading — this card is unlabelled and its landing surface is governed skills/** (draft PR, human merge). Session session_01BAZFhALsQsGqxui8sNqM8s.

Generated by Claude Code

分诊路由 · triage seat

标签：domain:skills · documentation · finding · pm:queue · priority:p2

1. ⭐ `Blocked-by: #5741` 已解除 —— Phase 2 已在树上

origin/main 0558e0f：

packages/react/src/hooks/useExpression.ts:153-157 —— usePredicateRecordContext 返回 { record } 且仅此；
packages/core/src/evaluator/listConditional.ts:281 —— { ...(opts.scope ?? {}), record: rowObj }；
packages/core/src/evaluator/rowPredicateCanon.ts:23-28 —— “Phase 2 (objectui#5741, ruled 2026-09-02 and amended 2026-09-05) retired them: the bare shorthand and data.* are no longer bound on runtime record surfaces, and the Phase-1 warning went with them.”

⇒ 卡片写的是将来时（“stale once #5741 … retires them”），树上已是完成时。扳机已经扣下，本卡现在就可动。 这是本轮第 22 次 branch-facts-as-tree-facts。

2. 锚点复核：逐字命中

skills/objectui/guides/schema-expressions.md:277-282（卡片给的行号，精确）：

“data.* is the trap in a row predicate. The row binds three ways -- record.status (canonical), bare status and data.status, the last two deprecated and warned once in dev (packages/core/src/evaluator/rowPredicateCanon.ts). The authoring oracle accepts data.* silently and then binds nothing at runtime: a constant false, not an error.”

✅ 逐字一致。并且卡片提出的修法所依赖的锚点也存在：CANONICAL_ROOT_BY_LAYER 在 rowPredicateCanon.ts:63 与 listConditional.ts:230 有真实引用（{ runtime: 'record', metadata: 'data' }）⇒ “data 在元数据编辑层仍是 canonical” 这半句有据可依（discipline ⑭：落点已验证存在）。

3. ⭐ 定级 p2，并说明它与同批 #7770 的 p3 差在哪

两张卡改同一个文件，我给了不同的档，判据必须写清楚，否则优先级轴就乱了：

	教的是什么	读者照做的后果
#7770（p3）	一个反例（“别这么写”），只是它的理由因退役而变了	读者的行动仍然正确（他本来就不该那么写）
本卡 #7728（p2）	“这两种拼法是 deprecated，会在 dev 下 warn 一次”	⭐ 读者会合理地推断「它们还能用，只是不推荐」，于是继续写 `data.status` —— 而它现在直接 fault（`Unknown variable: data`）

⇒ 本卡两个半句都已为假：不是 deprecated 而是 retired；不是 warn 而是根本不再有那条 warning（rowPredicateCanon.ts:26-28 明写 warning 随之移除）。⇒ 这属于把一条死通道当作活通道来教，与 #7945（同一份已发布技能仍把 events bag 当作 action system，p2）同类同档。

⛔ 不抬 p1：受众是编写者而非终端用户，且照它写出来的谓词会响亮地 fault（各面套用既有的 fail-closed / fail-soft 策略），⛔ 不会静默给出错误结果。

4. 对实施者的边界

✅ 保留卡片点名要保留的那半句：“the authoring oracle accepts data.* silently” —— ⭐ 它仍然为真，而且正是这条 bullet 值得存在的理由。⛔ 不要把整段删掉。
✅ 改写目标（卡片已给，我复核后认为准确）：运行期记录面只有一个根 record.*；裸字段或 data.* 在那里 fault；data 在元数据编辑层仍是 canonical（ADR-0089 D3 / CANONICAL_ROOT_BY_LAYER）。
⚠️ 行数预算：skills/** 有自己的行数预算（卡片自述），⇒ 请按 net-zero 或 net-negative 改写。⛔ 不要顺手扩写成一节。
⛔ 治理面：skills/** 在 scripts/check-governed-queue-guard.mjs:230 的 GOVERNED_SURFACES 里（id skills-catalog）⇒ PR 停在 draft，需 GOVERNED_APPROVERS（os-zhuang / hotlong，:252）之一人工合并。⛔ 不翻 ready、不入队、⛔ 绝不自己留下那条 approval（AGENTS.md 第五条禁令）。

5. ⭐ 去重 —— 与 #7770 同文件，强烈建议合并为一个治理 draft

卡	同文件的哪几行	触发条件
本卡 #7728	`:277-282`	#5741 Phase 2 —— ✅ 已落地
#7770	`:47` · `:479` · `:485`	#6951 / PR #7761（`TextSchema.value` 退役）—— ✅ 已落地

⛔ 非重复（不同段落、不同触发源），⭐ 但两者的两个扳机都已扣下，且落点是同一个文件、同一道人工合并闸。⇒ 拆成两个 draft 等于让 os-zhuang / hotlong 为同一个文件做两次人工合并。建议并成一个 PR，在描述里分列两处来源。是否合并由实施者与 approver 决定，⛔ 本席不代为决定。

卡片自述的去重（MCP search_issues，带一个返回 #5741 本身的控制查询；#5759 已关闭、#7418 / #7680 无关）带了活控制 ⇒ 我接受为一次有效读数，⛔ 不升级为 exhaustive（discipline ⑪）。⚠️ 卡片声明其 REST 通道 403 —— 原样转录。

6. 车道

domain:skills：落点 skills/objectui/** —— 已发布技能目录，本仓 domain:skills 车道。⛔ 不是 domain:devx：不碰 content/docs、门禁或 CI；⛔ 不是 domain:ui / domain:spec：一行代码不改，packages/core 与 packages/react 只是被引用为证据。

⛔ 本席（triage）的禁令：不认领 · 不派单 · 不写代码 · 不合并 · 不裁决 decision-box 卡（本会话为 claude-opus-5，CONTRACT_REVIEW_TIER 硬门要求 fable）。

Generated by Claude Code

Claim: PM loop round 8 (continued) — flight G, one objectui skills flight: #7728 (the data.* row-predicate bullet in skills/objectui/guides/schema-expressions.md teaches the bare-field and data.* spellings as deprecated and warned once in dev; on main they are retired, fault at runtime, and the warning is gone) folded with the skills/** half of #7770 (the same guide's three text + value counter-examples and the sentence that value is "read but never templated" — TextSchema.value is a tombstone now, refused by name; write content). Same file, same register, one dev, XS. #7770's other half (four code comments under packages/components, packages/react, examples/schema-catalog) is domain:ui's and is NOT taken: the PR carries Fixes #7728 and Refs #7770, and at landing this seat releases #7770 back to pm:queue for its UI half.
Session: session_019RfFHiRCSs3JXLK4cwcfox
Branch: claude/issue-7728-schema-expressions-retired-spellings
Worktree: objectui-issue-7728
Domain: domain:skills
File surface: skills/objectui/guides/schema-expressions.md — net-zero or net-negative lines; the three os:check-marked counter-example fences stay marked so check:skill-examples's marked floor does not shrink. ⛔ No other file.
Container & model: XS, mode:subagent, opus — objectui carries no dispatch-gates.mjs, so the tier is the seat's judgment: prose on a governed surface, no contract accept/reject moves.
Clause-②: no
Serial: no open PR touches this guide. GOVERNED (skills/**) ⇒ draft PR, in-seat review, os-zhuang + hotlong, human merge; objectui has no skip-changeset label — the changeset gate decides; at ACCEPT the seat adds needs-user-decision + the 维护者速读.

Generated by Claude Code

os-dev-report

{
  "issue": 7728,
  "status": "done",
  "branch": "claude/issue-7728-schema-expressions-retired-spellings",
  "pr": "https://github.com/objectstack-ai/objectui/pull/8091",
  "premise_still_valid": true,
  "summary": "Both cards' premises re-verified on origin/main 295804a6 and both hit VERBATIM. #7728: the bullet at :277-282 said the bare-field and `data.*` row spellings were 'deprecated and warned once in dev'; rowPredicateCanon.ts:23-28 says Phase 2 RETIRED them and removed the Phase-1 warning with them. Rewritten to the retired state — one root `record.*`, a bare field or `data.*` faults on the runtime engine (`Unknown variable`) under each surface's existing error policy, `data` canonical only on the metadata-editing layer — and the sentence #7728 asked to KEEP (the authoring oracle still accepts `data.*` silently) is kept. #7770 skills half: the three `text` + `value` counter-examples at :46, :478, :485 now teach the tombstone verdict (refused by name; write `content`), per layout.zod.ts:65-77 and text.tsx:162/:167. Judgment call inside the ruling, flagged in the PR body: the :485 props-envelope counter-example kept its comment (still true) but its demo key moved `value` -> `content`, because leaving a retired spelling directly under a counter-example that has just said `value` is refused invites the reading '`value` is the key, just not in `props`'. NET-ZERO: 569 lines before and after, 9 insertions / 9 deletions, one file, every `os:check` marker kept. GOVERNED: PR #8091 is a DRAFT, not flipped ready, not enqueued, no auto-merge armed, no approval left by this seat, no assignee or label touched on either card. #7770's UI half (four code comments under packages/components, packages/react, examples/schema-catalog) is NOT taken and that card stays open — the PR carries `Refs #7770`, never `Fixes`.",
  "tests": "All at head 3916de68 (`git rev-parse --short HEAD` of the final commit), exit code captured before any pipe (`cmd > file 2>&1; E=$?`). Prerequisite: the scoped build the gate itself prints, `pnpm exec turbo run build $(node scripts/check-skill-examples.mjs --build-filter) --concurrency=2` -> 'Tasks: 29 successful, 29 total'. LOCAL: check:skill-examples exit 0 — 'Marked: 13 ts fence(s) (floor 13), 70 json fence(s) (floor 70)', IDENTICAL before and after the edit (before-run captured on the clean tree), 'JSON phase: 70 fence(s) parsed, 0 failed', 'Every marked skill example holds up against the built types'. check:skill-eval-tokens exit 0 — 'Every must_contain token is taught by its own skill bundle', red 0; no eval names `data.status` or `value` (schema-expressions.json's must_contain is content / ${data. / visible / && / ${ / SUM / DATEFORMAT / IF) and the oracle is bundle-wide, so nothing under skills/objectui/evals/** was touched. check:skills-paths exit 0 — 'OK (88/89 stated path(s) resolve across 20 guide file(s); 1 baselined)'. check:doc-fences exit 0. check:control-bytes exit 0 — 'scanned 6473 tracked text file(s)'; plus a self-scan of the edited file with grep -naP over the control-byte class, no hits. check-governed-queue-guard.mjs --test skills/objectui/guides/schema-expressions.md exit 3 — 'GOVERNED — 1 of 1 path(s) are on a governed surface: skills/** x1 — the published skills catalog'. scripts/pm/check-half-states.mjs --self-test exit 0 — '1574 cases pass'. check-changeset-presence.mjs exit 0 — 'No source or published contract of a released package changed in this range, so no changeset is owed.' Whole-repo `pnpm exec turbo run lint --concurrency=2` exit 0 — 'Tasks: 47 successful, 47 total', 0 errors (the 10 warnings are pre-existing in @object-ui/collaboration, untouched by this diff). Heavy runs went through the shared verify lock (slot flightG-7728); its VERDICT lines were read, never a bare $?. REMOTE CI, polled in the foreground to completion on head 3916de68: 28 check runs, FINAL 25 success / 3 skipped / 0 failure, including Governed Surface Queue Guard, Skill Example Check, Skill Eval Token Check, Skill Guide Path Check, Lint, Type Check, all four Test shards, Build & E2E, Changeset Declaration. One intermediate red is worth recording because it was NOT this diff: README Export Check failed 10s into the job inside `corepack enable` downloading pnpm (undici ERR_ASSERTION, assert(!this.paused)) — the gate never ran. Confirmed by (a) the job log, (b) that same check being green on base 295804a6, and (c) rerun_failed_jobs on the same head turning it green with no code change. Filed as #8099. My own local run of that gate is NOT a reading either: it exits 1 with 'its type entry ./dist/index.d.ts is not on disk -- run `pnpm build` first', because the scoped build covers 30 of 40 packages by design.",
  "mcp_calls": "3 — get_job_logs (the failing job's log; REST log download is HTTP 000, its Azure blob host is egress-denied), actions_run_trigger (rerun_failed_jobs; REST rerun is 403 'Resource not accessible by integration'), search_issues (one targeted dedup query for #8099; REST /search/issues is 403 'sessions are bound to their configured repositories', though repo-scoped REST reads work and carried every card, comment, PR and check-run read in this run)",
  "open_questions": [],
  "out_of_scope_findings": [
    "filed as #8099: a corepack pnpm download failure reds the required `README Export Check` before the gate runs; on the merge_group leg the workflow's own header says that is a 60-minute queue stall, and 15 workflows run `corepack enable`",
    "NOT filed, reported to PM instead — dispatch reading correction, no repo defect: the brief names `node scripts/check-half-states.mjs --self-test`; that path does not exist. The script is at `scripts/pm/check-half-states.mjs` and that is what was run (exit 0, 1574 cases)",
    "NOT filed, reported to PM instead — dispatch asked which gate enforces a `skills/**` line or token budget: NONE does. Searched scripts/ and .github/workflows/ for line/token budget patterns, zero matches. The three skills gates are check-skill-examples (a SHRINK-ONLY floor on the MARKED FENCE population, not on lines), check-skill-eval-tokens and check-skills-paths. The net-zero constraint honoured here is the triage seat's instruction, not a mechanical one — nothing would have gone red had it been missed"
  ]
}

Generated by Claude Code

准入复核 · triage seat（新规回溯）—— 本席站下，交回 skills 席

objectstack@a5eccf92 落地的新规里有一条直接管辖归属的（.claude/skills/pm-dispatch/SKILL.md:380）：

「车道席可附证据/前提重验，⛔ 不定级不改标；skills 车道 finding 由该席自分诊，全仓轮跳过。」

本卡是 domain:skills（落点 skills/objectui/guides/schema-expressions.md）⇒ 它归 skills 席自分诊，全仓分诊轮应当跳过它。我上一轮给它定了级（pm:queue / p2），那是新规落地之前的做法；按现在的条文，那不是我的判断。

⇒ 本席就此站下：⛔ 不再改它的标签、⛔ 不再定级、⛔ 不代它做准入判定。⚠️ 我上一轮已加的标签原样留着，交由 skills 席复核并按其判断调整。

附证据（`SKILL.md:380` 允许车道外席位做的唯一一件事）

供 skills 席直接使用，⛔ 不必重测：

扳机已扣下 —— Carrier for Phase 2 of the row-predicate deprecation: the stored-metadata survey that must size the removal window before any spelling retires #5741 Phase 2 已在 objectui@0558e0f 落地：packages/react/src/hooks/useExpression.ts:153-157 返回 { record } 且仅此；packages/core/src/evaluator/listConditional.ts:281 为 { ...(opts.scope ?? {}), record: rowObj }；rowPredicateCanon.ts:23-28 记 “ruled 2026-09-02 and amended 2026-09-05 — retired them … and the Phase-1 warning went with them”。⇒ 卡片写的将来时已是完成时。
锚点逐字命中：schema-expressions.md:277-282 的那段 bullet 与卡片引文一致。
修法所依赖的锚点存在：CANONICAL_ROOT_BY_LAYER（{ runtime: 'record', metadata: 'data' }）在 rowPredicateCanon.ts:63 与 listConditional.ts:230 有真实引用。
我对准入的看法（仅供参考，⛔ 非判定）：这一段两个半句今天都已为假 —— 不是 deprecated 而是 retired，不是 warn 而是那条 warning 已随之移除。⇒ 读者会合理推断"还能用，只是不推荐"，于是继续写 data.status，而它现在直接 fault。我倾向认为它落在 (c) 类，但这个判断该由 skills 席做。
同文件另一张：finding: prose still describes text.value as read (skills counter-examples + four read-site comments) after objectui#6951 retired the key #7770 也改 schema-expressions.md（:47 / :479 / :485），两者的扳机都已扣下 ⇒ 若都要动，⭐ 建议并成一个治理 draft，省一次人工合并往返（skills/** 在 GOVERNED_SURFACES 内，需 os-zhuang / hotlong 之一人工合并）。

Triage: stood down; rationale: domain:skills findings are self-triaged by that lane's seat and skipped by the whole-repo round (SKILL.md:380). Evidence attached for that seat; no grading or label change made by this seat.

Generated by Claude Code

Navigation Menu

Repository navigation

finding(skills): `skills/objectui/guides/schema-expressions.md` teaches the bare-field and `data.*` row spellings as "deprecated and warned once in dev" — stale once #5741 (Phase 2) retires them #7728

Where

What changes

Fix shape (for triage, not ruled)

Dedup

分诊路由 · triage seat

1. ⭐ `Blocked-by: #5741` 已解除 —— Phase 2 已在树上

2. 锚点复核：逐字命中

3. ⭐ 定级 p2，并说明它与同批 #7770 的 p3 差在哪

4. 对实施者的边界

5. ⭐ 去重 —— 与 #7770 同文件，强烈建议合并为一个治理 draft

6. 车道

准入复核 · triage seat（新规回溯）—— 本席站下，交回 skills 席

附证据（`SKILL.md:380` 允许车道外席位做的唯一一件事）

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(skills): skills/objectui/guides/schema-expressions.md teaches the bare-field and data.* row spellings as "deprecated and warned once in dev" — stale once #5741 (Phase 2) retires them #7728

Description

Where

What changes

Fix shape (for triage, not ruled)

Dedup

Activity

os-justin commented on Sep 6, 2026

os-zhuang commented on Sep 6, 2026

分诊路由 · triage seat

1. ⭐ Blocked-by: #5741 已解除 —— Phase 2 已在树上

2. 锚点复核：逐字命中

3. ⭐ 定级 p2，并说明它与同批 #7770 的 p3 差在哪

4. 对实施者的边界

5. ⭐ 去重 —— 与 #7770 同文件，强烈建议合并为一个治理 draft

6. 车道

claude commented on Sep 6, 2026

claude commented on Sep 6, 2026

os-zhuang commented on Sep 7, 2026

准入复核 · triage seat（新规回溯）—— 本席站下，交回 skills 席

附证据（SKILL.md:380 允许车道外席位做的唯一一件事）

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(skills): `skills/objectui/guides/schema-expressions.md` teaches the bare-field and `data.*` row spellings as "deprecated and warned once in dev" — stale once #5741 (Phase 2) retires them #7728

1. ⭐ `Blocked-by: #5741` 已解除 —— Phase 2 已在树上

附证据（`SKILL.md:380` 允许车道外席位做的唯一一件事）