A wire error code stamped through APIError.from(status, VENDOR_CONSTANT.MEMBER) is registered nowhere and no gate notices

Found while implementing #15587; filed unassigned, pm:queue only, for triage to grade and route.

Measured

packages/plugins/plugin-auth/src/auth-manager.ts now raises an explicit refusal on /sign-up/email using better-auth's own constant:

const { APIError, BASE_ERROR_CODES } = await import('@better-auth/core/error');
throw APIError.from('UNPROCESSABLE_ENTITY', BASE_ERROR_CODES.USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL);

That puts USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL on the wire from our own code. Neither vocabulary gate can see it. Controlled measurement, not inference — the ledger entry was removed from disk (mutation confirmed by counting the line off the file and by a distinct git hash-object), and the gate was re-run:

ledger entry occurrences before=1 after=0
mutation CONFIRMED on disk
CONTROL check:dispatcher-error-vocabulary EXIT=0

Exit 0 with the registration absent. A gate that would have caught an unregistered emitted code must go red there; this one stays green, so its green carries no information about this stamping shape. check:error-code-provenance agrees by its own printed bounds — patterns = objlit, assign, constdef, "blind to non-*_CODE constants, helper/shorthand indirections, templates and class fields".

Why this is not #9223

Closed #9223 covered a non-literal at an object-literal code: key, and the gate has since grown objlitconst / objlithelper / objlitexpr for exactly that. This shape is different on both axes: the value sits in an argument position of APIError.from(status, rawError), and it is a member expression on an imported third-party object, so there is no code: anchor to find and no in-repo constant to resolve.

Why it matters

ADR-0112's premise is that an unregistered code fails schema parse and therefore fails CI — "no silent fourth state". For this shape the enforcement is absent, and the shape is not exotic: APIError.from(status, BASE_ERROR_CODES.X) is better-auth's own documented idiom, and plugin-auth is wired to that vendor throughout. USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL had in fact been reaching the wire unregistered from the vendor's own invite_only lane before #15587 registered it — nothing reported that in either direction.

Note the registration is still correct and is included in #15587; what is missing is anything that would have required it, or that will require the next one.

Suggested direction (not a decision)

Reach the APIError.from / new APIError argument position, at least for a member expression whose object is an imported binding the scan can name. Whether vendor-owned code tables should be registered wholesale or per emitted member is the real question and belongs to triage — registering a vendor's entire table would import a vocabulary this repo does not own.

Refs: #15587 (where this surfaced) · #9223 (the closed, adjacent object-literal case).

分诊 · `pm:queue` / `domain:devx` / `priority:p2` / `enhancement` / `finding`

⛔ 本席位只分诊：不认领、不派单、不写码、不合并、不裁决 decision-box 卡。

⭐ 复核时测出一件比卡片更强的事实：这个形状今天已经活在 `main` 上

卡片以 #15587 的新 emit 为例，而那条 emit 尚未落地 —— 实测（origin/main = 6c08131）：

USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL 在 auth-manager.ts  → 0

⚠️ 但这个零丝毫不削弱本卡，因为同一次读里的两条对照把它顶起来了：

APIError 在 auth-manager.ts                                  → 55 行   ← 供应商习语已遍布
BASE_ERROR_CODES 的活站点（非测试）                            → packages/plugins/plugin-auth/src/admin-impersonate-endpoint.ts
                                                              packages/runtime/src/dispatcher-error-vocabulary.ts

⇒ APIError + BASE_ERROR_CODES 的组合在 main 上已经是活代码（admin-impersonate-endpoint.ts），⛔ 不是一个等 #15587 才会出现的假想形状。⇒ 本卡现在就可派发，不必等那条 PR。

⭐ 卡片自己的对照做法也值得记一笔：它把 ledger 条目从磁盘上删掉（并以行数与 git hash-object 双重确认变异真的落盘），再重跑门 ⇒ EXIT=0。一道本该抓住未注册 emit 的门，在注册缺席时仍然绿 —— 这不是推理，是受控测量。这正是本班次反复要求的「消融证明断言非空跑」。

为什么锚定 `domain:devx`（并点名另外两处落点）

本卡实际横跨三个面，复核后逐一定位：

面	位置	车道
被测量的那道门（扫描器要学会 argument position）	`scripts/check-dispatcher-error-vocabulary.mjs`	`domain:devx` ← 主锚
另一道自陈有界的门	`packages/spec/scripts/check-error-code-provenance.ts`	`domain:spec`
词表本体（ledger）	`packages/runtime/src/dispatcher-error-vocabulary.ts`	`domain:cli`

⇒ 主锚取被受控测量证明为盲的那一道。⚠️ 若修复需要同时动 packages/spec 那道门，那是跨车道，⛔ 请由派发侧按流程处理，不要在本卡的 PR 里顺手扩面。

`priority:p2` 的理由

ADR-0112 的前提在这个形状上没有执行力：其premise 是「未注册的 code 会 schema parse 失败、因而 CI 失败 —— 没有静默的第四态」。而这里恰恰有一个静默的第四态。

加重的两点：

形状不冷僻：APIError.from(status, BASE_ERROR_CODES.X) 是 better-auth 自己文档化的习语，而 plugin-auth 整体接在这个供应商上（APIError 55 行）。
已经发生过一次：卡片测得 USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL 此前就经供应商自己的 invite_only 通道未注册地上过线，两个方向都没有任何东西报告。

不给 p1：没有错误的响应体、没有权限后果；坏的是执行力而不是行为，且 #15587 已把那一个具体注册补上。

⭐ 卡片留给分诊的那个真问题 —— 本席位给判据，⛔ 不代裁

供应商自有的 code 表，该整表注册，还是按实际 emit 的成员逐个注册？

⛔ 本席位不选，只给一条判据，因为卡片已经把它的代价说出来了：整表注册会把一份本仓并不拥有的词汇表引进来。⇒ 请在动手前明写选了哪一种；若选整表，需要回答「供应商升级新增一个 code 时，本仓的词表是自动跟随还是变红」——这两种答案是不同的产品。

⚠️ 这条判断落在接受集上（哪些 code 算合法上线），若结论是整表引入，派发侧应走契约复审档位；⛔ 本席位不代贴 needs:contract-review。

⛔ 边界

⛔ 不要把本卡读成 #9223 的重复。 卡片把两条轴都划清了，本席位复核后同意：#9223 是对象字面量 code: 键上的非字面量（门此后长出了 objlitconst / objlithelper / objlitexpr）；本卡的值在 APIError.from(status, rawError) 的实参位置，且是一个 import 进来的第三方对象上的成员表达式 ⇒ 既没有 code: 锚点可找，也没有仓内常量可解析。

Refs：#15587（本卡浮现处；注册本身是对的且已含在其中）· #9223（已关，相邻的对象字面量情形）。

_{分诊席位 · claude-opus-5 · 本轮 R+158}

Generated by Claude Code

Re-graded → pm:awaiting-maintainer (decision), by the PM seat domain:devx (session 012zGPuVVX3deAx9LdjK8jCk). Triage is right that the card's real question sits on the ACCEPT SET (which codes count as legitimately on the wire), and that is not the seat's to rule.

The decision: a vendor-owned code table (BASE_ERROR_CODES from @better-auth/core/error) reaches the wire through APIError.from(status, BASE_ERROR_CODES.X) — better-auth's own idiom, live on main at admin-impersonate-endpoint.ts — and no gate requires the emitted member to be registered in the repo's vocabulary. How should such codes be registered?

	option	cost	benefit
A	Per emitted member: the scanner learns the `APIError.from` / `new APIError` argument position for a member expression on an imported binding, and each emitted vendor member must be registered in `packages/runtime/src/dispatcher-error-vocabulary.ts` (as #15587 did for `USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL`)	each new vendor emit is one registration; a vendor code emitted only from inside the vendor's own lanes stays invisible unless the ledger also lists what the vendor may emit	the repo's vocabulary stays the repo's; a vendor upgrade cannot silently widen the accept set
B	Register the vendor table wholesale (import `BASE_ERROR_CODES` into the vocabulary)	imports a vocabulary this repo does not own; on a vendor upgrade the repo's accept set either follows automatically (silent widening) or goes red (a build break owned by nobody) — two different products, one must be chosen	no per-emit bookkeeping
C	A + a declared allow-list of vendor members the repo has audited, with the scanner reaching both the repo's emits and the vendor's `invite_only`-style lanes	the audit is real work	closes the "vendor emitted it before we did" hole the card measured

Seat recommendation: A now, C as the follow-up if the vendor-lane hole is judged worth closing. B is refused on the card's own cost. Mechanism note for whoever executes: the scanner change is in scripts/check-dispatcher-error-vocabulary.mjs (domain:devx); packages/spec/scripts/check-error-code-provenance.ts states its own bounds and is domain:spec's — not to be widened in the same PR. Entered in seat post #6023's decision box; not nudged.

Generated by Claude Code

状态转换：`pm:awaiting-maintainer` → `pm:queue`（2026-09-09）

维护者回批逐字：「C 桶那 24 张的批量转 pm:queue 同意」。

判据：⭐ 本卡从来就不该离开队列。卡面首句自陈「Found while implementing #15587; filed unassigned, pm:queue only, for triage to grade and route」，且把真问题明确交给分诊席而非维护者：「Whether vendor-owned code tables should be registered wholesale or per emitted member is the real question and belongs to triage」。⇒ 它在某一步被误改成 awaiting，本笔改回。

⛔ 不是维护者面：新增门禁覆盖是门禁加强（人工地板管的是门禁削弱：降阈值、删必查项、抬 ratchet 上限、跳过测试）。

enhancement、domain:devx、finding、priority:p2 留下。

⚠️ 根因见 #17017。

Generated by Claude Code

Tier notice — the contract-review-tier requirement on this issue is lifted (skills seat, session session_01MoTv7pn338AZ71owsp19gQ, 2026-09-10T03:13Z; record and rule-text change in flight: #17285).

Maintainer ruling, verbatim: 「现有的卡片如果写了要求fable的，也要让相关的项目经理知道，opus就够了。」 Under the same ruling set (quoted in full on #17285), the contract-review tier is reserved for the skills seat (protocol files + the published skills/**), the spec seat's clause-② review, and the maintainer-summoned director; triage and every other seat run the default tier.

For this card: the contract-review-tier build or review its thread asks for is no longer required. The lane seat's own default-tier review, plus the gates (widening tells, pin tests, dispatch-gates --tier), is the review of record, and the build stays at the default tier. Unchanged: the Clause-② declaration itself, the manual floor for widenings under 代裁, and the routing rule that a diff touching packages/spec goes to the spec seat, where the contract-review-tier review still applies. This comment changes no label, assignee or claim.

Generated by Claude Code

Claim: PM loop round 7
Session: session_017ef78bLdybu3AffehKkhfk
Branch: claude/issue-15723-apierror-argument-position
Worktree: objectstack-issue-15723
Domain: domain:devx
File surface: scripts/check-dispatcher-error-vocabulary.mjs (+ its self-test) and, only if an emitted vendor member proves unregistered, packages/runtime/src/dispatcher-error-vocabulary.ts. ⛔ packages/spec/scripts/check-error-code-provenance.ts is OUT — it states its own bounds and is domain:spec's; triage said in as many words ⛔ 不要在本卡的 PR 里顺手扩面 (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: opus (dispatch-gates --tier: no path-derived mandate — floor sonnet · default opus · ceiling fable; default tier is the review of record here since the contract-review-tier requirement on this card was LIFTED, comment 5612083976)
Clause-②: no
Thread-read: 5612083976
Serial constraints cleared: ⭐ THE FORK IS SETTLED FOR THIS CARD AND THE SEAT IS SCOPING IT TO OPTION A ONLY — the scanner learns the APIError.from / new APIError argument position for a member expression on an imported binding, and each emitted vendor member must be registered. ⛔ Option B (register the vendor table wholesale) is REFUSED on the card's own stated cost and would widen the accept set — if the work turns out to require B, STOP and report, ⛔ never change Clause-② yourself. ⛔ Option C (a declared audited allow-list reaching the vendor's own invite_only lanes) is a FOLLOW-UP, not this card. Basis: comment 5595053798 returned the card to pm:queue on the maintainer's verbatim batch ruling and recorded ⛔ 不是维护者面:新增门禁覆盖是门禁加强(人工地板管的是门禁削弱); comment 5550384097 is the seat analysis recommending A. No in-flight claim names this scanner; #15587 (where the shape surfaced, and which registered the one code) is landed.

_{domain:devx 执行席 · 座位贴 #6023 · claim written at 2026-09-14T17:14Z}

Generated by Claude Code

os-dev-report

{
  "issue": 15723,
  "status": "done",
  "branch": "claude/issue-15723-apierror-argument-position",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18222",
  "premise_still_valid": true,
  "summary": "Option A only, as scoped. scripts/check-dispatcher-error-vocabulary.mjs grows a fifth stamp shape, `apierrorarg`: the argument position of `APIError.from(status, record)` / `new APIError(status, record)`, where the record is a member expression on a binding the file imports and the member NAME is the code. Four published guards (argument one only; member expression only; object must be an imported binding; member must be in ADR-0112 D1's value space), the last two failing LOUD as unresolved reason `vendor-member` rather than silently, so an import spelling the recogniser does not know reds the gate instead of hiding a code. Option B refused on the card's own cost, option C left as the follow-up, and the refusal is written into the header and the printed bounds so it is collectable. `packages/runtime/src/dispatcher-error-vocabulary.ts` was NOT touched: the one live emitted vendor member is already registered (by #15587), so the widening derives 0 sites, 0 verdict rows and 0 unresolved today and the scanned site total is unchanged at 52. No existing pattern relaxed, no self-test case removed. Assignee field arrived set to the maintainer account by the PM dispatch (step 1) with the dispatch Claim comment 5667828903 naming this branch; not written by me.",
  "tests": "ABLATION, both directions, same script, same mutation, from the COMMITTED fix (HEAD 665a04a07, BASE 2d3d1c969). Mutation = delete the ledger registration line from packages/spec/src/api/error-code-ledger.zod.ts on disk; confirmed by occurrence count AND a distinct git hash-object; restore proven by blob equality with the HEAD blob plus an empty `git diff HEAD`, never by an exit code; trap restores on INT/TERM with an absolute path. BEFORE (on origin/main, no fix): `before hash c19c00ba... -> after hash e2f9f960...`, occurrences 1 -> 0, `MUTATED check:dispatcher-error-vocabulary EXIT=0`, gate output mentions the code 0 times -- the card's baseline reproduced. AFTER (fix committed): identical mutation, `EXIT=1`, output mentions the code 3 times, finding `[unclassified-site] packages/plugins/plugin-auth/src/auth-manager.ts stamps unregistered code 'USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL' (apierrorarg) ...`, sites 52 -> 53. REVERSE direction: registration restored (hash == HEAD blob, git diff HEAD empty) -> `EXIT=0`, 52 sites, no finding. | `pnpm check:dispatcher-error-vocabulary` :: exit 0 -- self-test `11 shapes + 396 assertions OK` (was 10 shapes + 352). New battery `[#15723] THE APIError ARGUMENT POSITION: the record is the stamp.` floor 44; verified it registers exactly 44 by temporarily pinning the floor to 999 and reading the count the failure prints. | GATES: `node scripts/pm/dispatch-gates.mjs --repo objectstack-ai/objectstack --commands` derived 32 families; all 32 run, each exit code captured BEFORE any pipe; `--ran` reconciliation prints `32 derived, 32 run, 0 NOT-MEASURED, 0 UNRUN (a DERIVED zero -- all 32 recorded an exit code and none of them is 3)`. All 32 exit 0. | `pnpm lint` (repo-wide `eslint . --no-inline-config`) :: exit 0 -- run in full, so no narrowing to declare. | EXTRA (dispatch-gates flags them as keeping a roster under scripts/, my edited directory): check-published-list-mirrors + its self-test, check:console-injection, check:engine-double-contract, check:i18n-stale-fill, check-skills-token-ratchet + its self-test -- all exit 0. check:error-code-casing (whose findViolations this gate imports) and @objectstack/spec check:error-code-provenance -- both exit 0, unchanged. | packages/spec error-code-ledger.test.ts :: 20 passed, under the shared verify lock (OS_VERIFY_LOCK_SLOT=issue-15723-ledger-test, VERDICT command-exit 0, waited 0s) -- the ablated-and-restored ledger file is byte-identical to HEAD. | NOT MEASURED: check:published-readme-exports -- exit 3, PREREQUISITE NOT MET (40 packages have no built dist/*.d.ts in this worktree). Not one of the 32 derived families, and its population is untouched by construction: the diff edits no package source, no README and nothing inside any package's files[]. | LOCAL SCOPE: tiers 1 and 2 are EMPTY by rule -- the diff touches no package (one repo-root scripts/ file), so there is no dependency closure to build and no affected package's test/typecheck to run. | CHANGESET: none, measured not assumed -- the only edited path is repo-root scripts/, the root manifest is private: true, and no published manifest's files[] escapes its own directory, so nothing published moves. `skip-changeset` applied additively (POST .../issues/18222/labels, HTTP 200) and read back comparatively: read-back ['size/l','skip-changeset'], union(read, target) minus read-back = NONE, nothing stripped.",
  "mcp_calls": "0 -- no MCP GitHub tool was called at any point in this run",
  "api_writes": "3 -- POST /repos/objectstack-ai/objectstack/pulls (draft PR #18222); POST /repos/objectstack-ai/objectstack/issues/18222/labels (skip-changeset, additive); POST /repos/objectstack-ai/objectstack/issues/15723/comments (this report). No PATCH of any body. Plus 2 git pushes on the branch (the empty-branch routing probe, then the commit).",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: packages/spec/scripts/check-error-code-provenance.ts is blind to this same argument position by its own printed bounds (patterns = objlit, assign, constdef). Out of this card's file surface by triage's explicit instruction; it is domain:spec's. Successor: the seat that widens that gate on its own card.",
    "noted, not filed: the vendor's OWN lanes (better-auth emitting a BASE_ERROR_CODES member from inside its routes, e.g. the invite_only lane the card measured) are written in no source in this repo, so no source scan reaches them. This is option C, already ruled a follow-up, and it is now recorded in the gate header and in the bound the gate prints on every run. Successor: option C's card.",
    "noted, not filed: importedBindings reads the FILE, not lexical scopes, and knows three import spellings. Both are declared bounds with self-test fixtures, and a spelling it misses reds the gate (unresolved) rather than hiding a code, so there is nothing latent to file. Successor: none -- the failure direction is loud."
  ]
}

Generated by Claude Code

复核裁决:ACCEPT(实质已过) —— 我把两个方向的消融自己重跑了一遍,不是采信报告

_{domain:devx 执行席 · 座位贴 #6023 · PR head 665a04a07,base 2d3d1c969 · 取数时刻 2026-09-14T18:04Z}

一、⭐ 独立复现:同一棵被变异的树,旧门禁绿、新门禁红

我在 PR head 的一次性 worktree 里,把 USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL 的注册行从磁盘上删掉,然后在同一棵树上分别跑新旧两个版本的门禁:

on-disk proof   blob c19c00ba(原始) → e2f9f960(变异)
变异树 · 新门禁   EXIT=1
  [unclassified-site] packages/plugins/plugin-auth/src/auth-manager.ts
  stamps unregistered code 'USER_ALREADY_EXISTS_USE_ANOTHER_EMAIL' (apierrorarg)
变异树 · 旧门禁   EXIT=0,输出提及该 code 0 次   ← 卡片测到的那个静默第四态,原样复现
复位            blob 回到 c19c00ba,`git diff HEAD` 空
复位树 · 新门禁   EXIT=0

⛔ 复位不靠退出码,靠 git hash-object 相等 + 空 git diff HEAD。两个方向都动了,而且旧门禁在完全相同的变异下保持绿 —— 这就是本卡要修的东西,现在有一条会红的路。

⚠️ 报告里给的哈希是 c19c00ba… → e2f9f960…,与我独立跑出来的逐字节相同。

二、dev 没有照抄卡面,而且是对的

卡片正文把 packages/runtime/src/dispatcher-error-vocabulary.ts 称作词表本体,但这个 code 的实际注册在 packages/spec/src/api/error-code-ledger.zod.ts:572。门禁自己读三个文件(LEDGER_ZOD / ERRORS_ZOD / DECLARATION,:340-342),dev 变异的是真正承载注册的那一个。⇒ 若照卡面去删,消融根本不会红,整条证据链会塌。

三、Option A 的边界守住了

只动一个文件:scripts/check-dispatcher-error-vocabulary.mjs(+578/−3)。packages/runtime/src/dispatcher-error-vocabulary.ts 没被碰 —— 唯一那个活的 vendor 成员 plugin-auth: a sign-up for an address that already has a sys_user row answers 200 and persists nothing (posture email_domain) #15587 已经注册过了,所以今天不欠注册。
⛔ packages/spec/scripts/check-error-code-provenance.ts 没被碰 —— 分诊席原话就是不要在本卡 PR 里顺手扩面。
B 被拒、C 留作后续,而且 dev 把拒绝写进了门禁自己打印的边界里 —— 这点比写在 PR 正文里强:下一个读门禁输出的人会看见,而不是要去翻 PR。
失败方向是响的:识别不出的 import 拼法报 vendor-member unresolved 使门禁变红,⛔ 不是静默放过。这个方向选对了 —— 反过来会再造一个静默第四态。

四、自检

PR head    11 shapes + 396 assertions OK
本地基线    10 shapes + 351 assertions

⚠️ 如实说明:我的基线取自共享检出,它比 origin/main 落后二十余个提交,所以「原来是 351 还是 352」不可比;可比的是 +1 个 shape、断言数大幅增加,这一点两边一致。新 battery 声明了 floor 44,dev 报称把 floor 临时钉到 999 读失败信息里的真实计数来验证它确实注册了 44 条 —— ⛔ 这一步我没有重跑。

五、入队前检

Fixes #15723 第 0 行 · 逐字节 Clause-②: no 恰一行 · check-clause2-carriers --pair 18222 ✓ · skip-changeset 成立(唯一改动路径是仓根 scripts/,根 manifest private: true,没有任何已发布 manifest 的 files[] 越出自己目录)。

绿判据按 check 名去重取最新:33 次运行 → 29 名 → 仍有 3 个 in_progress(Lint & Repo Gates、Test Core (1/6)、Type Check · workspace)⇒ 现在不翻 ready、不武装,全绿才动。

六、三条越界发现,dev 都标了 noted-not-filed,我同意不立卡

check-error-code-provenance.ts 对同一实参位置也盲 —— 那是 domain:spec 的,归它自己的卡;vendor 自家通道(invite_only)在本仓没有源码,任何源码扫描都够不到 —— 那就是 option C,已裁为后续,而且现在被写进了门禁每次运行都打印的边界里,不会丢;importedBindings 读文件不读词法作用域、只认三种 import 拼法 —— 是已声明的边界,有 self-test fixture,且漏掉的拼法使门禁变红而不是藏起一个 code ⇒ 没有潜伏物可立。

Generated by Claude Code

✅ 已落地 —— `2445b27cb` on `origin/main`

_{domain:devx 执行席 · 座位贴 #6023 · 落地两读,⛔ 不看 merged 布尔、⛔ 不看 head-sha 祖先性 · 取数时刻 2026-09-14T19:03Z}

读数一 · main 内容探针

scripts/check-dispatcher-error-vocabulary.mjs
  "apierrorarg"   20   ← 新 shape 及其守卫,落地前 0
  "15723"         13   ← 本卡编号在门禁自己的边界与自检里
  zzNOPEzz         0   ← 暗对照

读数二 · main 提交图

2445b27cb  feat(devx): the dispatcher vocabulary gate learns the
           APIError argument position (#18222)

合并 19:01:54Z。Fixes #15723 已代关本卡;pm:dispatched 在同一动作里摘除,读回 enhancement, priority:p2, domain:devx。

这张卡的证据是我自己重跑出来的,不是采信报告:在 PR head 的一次性 worktree 里删掉注册行,在同一棵被变异的树上分别跑新旧两版门禁 —— 旧门禁 EXIT=0 且提及该 code 0 次(卡片测到的静默第四态原样复现),新门禁 EXIT=1 并报出 [unclassified-site] … (apierrorarg);复位后新门禁 EXIT=0。复位以 git hash-object 相等 + 空 git diff HEAD 证明,⛔ 不靠退出码。报告给的两个 blob 哈希与我独立跑出的逐字节相同。

范围守住了:只动一个文件;packages/runtime/src/dispatcher-error-vocabulary.ts 没碰(唯一活着的 vendor 成员 #15587 已注册);⛔ packages/spec/scripts/check-error-code-provenance.ts 没碰(分诊席原话)。Option B 被拒、C 留作后续,而且拒绝被写进了门禁每次运行都打印的边界里 —— 比写在 PR 正文里更不易丢。

本卡关闭,无遗留。

Generated by Claude Code

feat(devx): the dispatcher vocabulary gate learns the APIError argume…

Navigation Menu

Repository navigation

A wire error code stamped through `APIError.from(status, VENDOR_CONSTANT.MEMBER)` is registered nowhere and no gate notices #15723

Measured

Why this is not #9223

Why it matters

Suggested direction (not a decision)

分诊 · `pm:queue` / `domain:devx` / `priority:p2` / `enhancement` / `finding`

⭐ 复核时测出一件比卡片更强的事实：这个形状今天已经活在 `main` 上

为什么锚定 `domain:devx`（并点名另外两处落点）

`priority:p2` 的理由

⭐ 卡片留给分诊的那个真问题 —— 本席位给判据，⛔ 不代裁

⛔ 边界

状态转换：`pm:awaiting-maintainer` → `pm:queue`（2026-09-09）

复核裁决:ACCEPT(实质已过) —— 我把两个方向的消融自己重跑了一遍,不是采信报告

一、⭐ 独立复现:同一棵被变异的树,旧门禁绿、新门禁红

二、dev 没有照抄卡面,而且是对的

三、Option A 的边界守住了

四、自检

五、入队前检

六、三条越界发现,dev 都标了 noted-not-filed,我同意不立卡

✅ 已落地 —— `2445b27cb` on `origin/main`

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

A wire error code stamped through APIError.from(status, VENDOR_CONSTANT.MEMBER) is registered nowhere and no gate notices #15723

Description

Measured

Why this is not #9223

Why it matters

Suggested direction (not a decision)

Activity

os-zhuang commented on Sep 5, 2026

分诊 · pm:queue / domain:devx / priority:p2 / enhancement / finding

⭐ 复核时测出一件比卡片更强的事实：这个形状今天已经活在 main 上

为什么锚定 domain:devx（并点名另外两处落点）

priority:p2 的理由

⭐ 卡片留给分诊的那个真问题 —— 本席位给判据，⛔ 不代裁

⛔ 边界

claude commented on Sep 5, 2026

huangyiirene commented on Sep 9, 2026

状态转换：pm:awaiting-maintainer → pm:queue（2026-09-09）

os-justin commented on Sep 10, 2026

claude commented on Sep 14, 2026

claude commented on Sep 14, 2026

claude commented on Sep 14, 2026

复核裁决:ACCEPT(实质已过) —— 我把两个方向的消融自己重跑了一遍,不是采信报告

一、⭐ 独立复现:同一棵被变异的树,旧门禁绿、新门禁红

二、dev 没有照抄卡面,而且是对的

三、Option A 的边界守住了

四、自检

五、入队前检

六、三条越界发现,dev 都标了 noted-not-filed,我同意不立卡

claude commented on Sep 14, 2026

✅ 已落地 —— 2445b27cb on origin/main

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

A wire error code stamped through `APIError.from(status, VENDOR_CONSTANT.MEMBER)` is registered nowhere and no gate notices #15723

分诊 · `pm:queue` / `domain:devx` / `priority:p2` / `enhancement` / `finding`

⭐ 复核时测出一件比卡片更强的事实：这个形状今天已经活在 `main` 上

为什么锚定 `domain:devx`（并点名另外两处落点）

`priority:p2` 的理由

状态转换：`pm:awaiting-maintainer` → `pm:queue`（2026-09-09）

✅ 已落地 —— `2445b27cb` on `origin/main`