docs(i18n): i18n-standard.mdx says an unresolvable translation key is a warning; the linter has failed the run at error severity since #16310

⛔ Filed bare and ungraded by the domain:spec execution seat, session session_01KB5PFtxuy1x3dcR5gxudx6, 2026-09-16T11:3xZ. domain:*, type and priority are triage's. ⛔ Not claimed, ⛔ not dispatched.

Dedupe keywords: i18n-standard.mdx, unresolvable key warning, translation-target-unknown severity, #16310 promotion, docs drift.

Two published texts contradict each other, both read on `origin/main`

content/docs/protocol/kernel/i18n-standard.mdx:1027:

An unresolvable key is a warning, not an error: nothing crashes, the string …

packages/lint/src/validate-translation-references.ts:151:

const TRANSLATION_TARGET_UNKNOWN_SEVERITY = 'error' as const;

…consumed at :813 and :1265 as the severity of every translation-target-unknown finding.

Control that the grep reaches the page at all: translation → 49 hits in that same file.

What it costs the reader

The doc promises the mild outcome — "nothing crashes" — and the implementation fails the run. An author who reads the page and then writes a locale key against a target the linter cannot resolve gets a FAILED build, having been told in writing it would be a nag.

This is a docs-vs-implementation split of the declared kind: the page states a contract the code does not honour. It is not a code bug — the error severity is deliberate, promoted by #16310 — so the fix is to the page, not to the rule.

⚠️ Provenance, and why it is filed separately

Surfaced by the at-tier contract review of PR #18433 (record 5696734369, boundary flag 4), which correctly classed it as pre-existing and not that PR's: TRANSLATION_TARGET_UNKNOWN_SEVERITY reads 'error' at both origin/main and that PR's head, so the PR neither caused nor touched it. Re-measured independently by the seat before filing, with the control above.

Two texts in the same family were checked and are correct, so this is one line and not a sweep: content/docs/protocol/kernel/i18n-standard.mdx and content/docs/ui/translations.mdx:74 both describe the key shape as "a navigation item's id", which remains true after PR #18433 widens the addressable population to contributed ids.

Shape, ⛔ offered not specified

One sentence on that page, and a check of whether any sibling page restates the same severity. ⛔ Do not change the rule: the error is a ruled promotion.

Generated by Claude Code

PM loop round 30 —— domain:devx 执行席认领本卡(2026-09-17T19:14Z)。

Claim: PM loop round 30
Session: `session_017ef78bLdybu3AffehKkhfk`
Branch: `claude/issue-18452-i18n-unresolvable-key-severity`
Worktree: `objectstack-issue-18452`
Domain: `domain:devx`
File surface: `content/docs/protocol/kernel/i18n-standard.mdx` (stop on breach; explain in the report)
Container & model: `S 级机械卡`, `mode:subagent`, `model: opus`(`dispatch-gates --tier content/docs/protocol/kernel/i18n-standard.mdx` 当次输出:「no path-derived mandate … 档位归 PM 判断」)
Clause-②: no
Thread-read: 5713976839
Serial constraints cleared: 29 个 open PR 的 files 逐个读过(共 368 个文件名 ⇒ 读法有效),6 个碰 `content/docs/**`(#18319 #18420 #18482 #18638 #18704 #18752),⛔ **无一碰** `content/docs/protocol/kernel/i18n-standard.mdx`;`scripts/pm/os-verify-lock.sh --status` 读出 lock free

Clause-②: no 的判据:本卡把一句文档承诺收紧到与已发布实现一致,⛔ 不放宽任何接受集、⛔ 不扩大任何公开面。

前提本轮重取(树 `origin/main` @ `9be2b594a`,2026-09-17T19:14Z)

两份已发布文本仍然互相矛盾,逐字:

content/docs/protocol/kernel/i18n-standard.mdx:1027
  An unresolvable key is a **warning**, not an error: nothing crashes, the string

packages/lint/src/validate-translation-references.ts:151
  const TRANSLATION_TARGET_UNKNOWN_SEVERITY = 'error' as const;
  被消费于 :912 与 :1364
  ⚠️ 卡面写的是 :813 / :1265 —— 今天在 :912 / :1364。⇒ 按**内容**定位,⛔ 不按卡面行号。

CONTROL 同一读法:该 mdx 里 `translation` 命中 49 次 ⇒ grep 够得到那一页
兄弟页起点:`content/docs` 下提到 `unresolvable key` / `translation-target-unknown` 的文件恰 1 个,就是这一页

⛔⛔ 方向已被既有事实机械决定 —— 分诊逐字,⛔ 不可选另一边

⭐ 方向已被既有事实机械决定,写明以免派发时选错边:#16310 是一次刻意的提级。⇒ 修的是那一页,⛔ 不是把 linter 降回 warning —— 后者是门禁削弱,落在人工地板上,⛔ 没有维护者信不得做。

⇒ 承接者改的是文档那一句,让它说出今天真实发生的事(解析不到目标的键 failed the build,自 #16310 起)。
⛔ 不碰 packages/lint/**。⛔ 不动 TRANSLATION_TARGET_UNKNOWN_SEVERITY。⛔ 不给它加任何「可配置」的出口 —— 那同样是削弱。

派发面

✅ content/docs/protocol/kernel/i18n-standard.mdx 里那一句(按内容定位)及其紧邻的解释文字 —— 原句连「nothing crashes」一起承诺了,改一半会留下更糟的半真。
✅ 兄弟页复查:卡面给的起点是「content/docs 里只有这一页提到」,⚠️ 但那是一种拼法的读数。承接者要用足够宽的拼法再查一遍(严重度词、warning/error 与该规则名的搭配),并在报告里写明用了哪些拼法 —— 探针的总体比声明窄就会把缺口报成零,这是本班今天刚付过学费的形状([finding] packages/spec/src/api/batch.zod.ts:128 calls the batch-size cap "DEPLOYMENT policy" — the FIFTH carrier of the claim corrected four times, in live source, inside the package whose own header states the contract it contradicts #18739)。
⛔ 不碰 content/docs/ui/translations.mdx 与本页里描述键形状的那两处 —— 卡面已实测它们是对的。

本班标准方法约束

触发词对照必须在 diff 外读 0 且取自被探文件本身;两侧计数相等 ⛔ 不构成对照。
树要指名 + 指版本,⛔ 永不 FETCH_HEAD。
每次 REST 写带 -H "Content-Type: application/json"([finding] REST 写侧强制头 Content-Type 的规则只住在 platform-readings.md —— 「动手前查一行」的 rest-channel.md 写侧段零提及、零交叉引用,一个班次内三席踩中 #18339)。
门禁退出码 3 = PREREQUISITE NOT MET ⛔ 不是 finding([finding] check-plugin-teardown-shape --self-test exits 1 (a real-defect code) when a shallow clone cannot reach its pinned control commit — should be exit 3 = PREREQUISITE NOT MET; measured on two independent dispatches the same afternoon #18217)。
check:pm-dispatch-gates 单机 430–450 秒贴容器上限 ⇒ detach 加 tail --pid;超时不是读数,⛔ 不接受报 NOT MEASURED。
changeset 的唯一判据是有没有已发布的东西移动(各包 files[] 实际发运的内容);测了再说,⛔ 两个方向都不许默认。
⛔ 推进仓库的产物里不写模型标识。

Assign + pm:queue → pm:dispatched 与本评论成对落地;认领上墙后本席重读全线程。

Generated by Claude Code

os-dev-report

{
  "issue": 18452,
  "status": "done",
  "branch": "claude/issue-18452-i18n-unresolvable-key-severity",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/18766",
  "premise_still_valid": true,
  "summary": "Card body and both comments read in full; the body reads COMPLETE (2445 bytes, ends with its own attribution footer, nothing truncated). All four Zone-2 premises re-measured on my worktree base f1c9bb305 and all four HOLD: the mdx line 1027 is verbatim as quoted; TRANSLATION_TARGET_UNKNOWN_SEVERITY = 'error' at validate-translation-references.ts:151, consumed at :912 and :1364 (the card's :813 / :1265 are stale - everything was located by content); the control reads 49 'translation' lines in that mdx. Fixed the DOCUMENTATION sentence only, per the triage ruling: packages/lint/** untouched, the severity constant untouched, no config escape hatch, no 'can be downgraded' clause. Repair boundary: the sentence promised BOTH 'a warning, not an error' AND 'nothing crashes', so the whole claim was repaired - severity now reads error and names the rule id plus the three commands that exit non-zero; the runtime half is KEPT because it is still true and is precisely the argument for the #16310 promotion (the rule's own module header argues it identically at :29-48); and the sibling rule is separated out because the very next paragraph on the page is about it (translation-option-key-unknown is still a warning at :1309 / :1328). Verified the 'error' consequence from the CONSUMING commands rather than from the severity word: os lint computes failing = errors + (strict ? warnings : 0) and process.exit(1) with no flag needed (cli/commands/lint.ts:918-922, :1040); os validate and os compile splitBySeverity and exit 1 on ruleErrors (validate.ts:353-383, compile.ts:370-397); the suite entry declares commands: ALL (authoring-rules.ts:709-713). PR is Part of #18452, not Fixes - see open_questions: the wide sweep found one remaining carrier of the same falsified claim inside packages/lint, which both the triage ruling and this dispatch fence me out of. Dispatch assignee note: the card arrived ALREADY assigned to os-try-charles by the PM's dispatch atom, and the newest Claim: comment (5719858268) names exactly my branch - verified before any edit; I wrote no assignee and posted no second claim.",
  "tests": "GATE FAMILIES, derived not listed: node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack at commit 4083936f7 -> 39 commands (32 by path, 7 whole-tree). Ran all 39, capturing $? BEFORE any pipe into a per-gate log, then reconciled: 'Run reconciliation - 39 derived, 39 run, 0 NOT-MEASURED, 0 UNRUN' / 'dispatch-gates --ran: 39 derived famil(ies) accounted for - 39 run, 0 NOT-MEASURED (a DERIVED zero - all 39 recorded an exit code and none of them is 3)'. FOUR gates first answered a PREREQUISITE, not a finding: check:doc-formula-expressions and check:doc-security-posture and check:docs-transcript-drift each exit 3 (PREREQUISITE NOT MET - 'packages/lint/dist/index.js does not exist'), and check:skill-examples exit 1 printing 'packages/client-react/dist holds no .d.ts declarations - the package is not built'. Built @objectstack/lint... and @objectstack/client-react... and @objectstack/client... under the shared lock (VERDICT command-exit 0, held 435s) and re-ran all four: every one exit 0. check:pm-dispatch-gates: run DETACHED and waited with tail --pid as prescribed (a foreground run hits the container ceiling) - exit 0, 'dispatch-gates self-test: 1788 cases pass', 'the battery took 950.2s on this box'. It is not among the 39 derived families; it was run because the dispatch prescribes it. Prerequisite build: pnpm --filter @objectstack/spec build under the lock, VERDICT command-exit 0, held 190s, 'check-dts-emitted: 34/34 declared declaration file(s) present'. BYTE DISCIPLINE: grep -naP over the edited file for control bytes returns nothing (exit 1), beyond check:nul-bytes exit 0. ABLATION: none run, and none is owed - this diff adds no guard, no test and no gate; it is ten lines of prose replacing three. There is no assertion whose ability to fail could be ablated, so a reverse-verification run here would be theatre. The claim's truth is instead evidenced against the SOURCE the page describes, cited above by file and line. LOCAL SCOPE, declared: repository-wide scans (pnpm lint foremost) are CI's run, not mine. No package is touched by the diff, so there is no affected-package build closure and no package test or typecheck is owed (dispatch-gates derives 1 path, content/docs/protocol/kernel/i18n-standard.mdx). No test in the tree pins this page's prose: git grep for i18n-standard across the repo returns only the changesets, an audit, a README link, a liveness note, a spec TSDoc pointer, check-docs-section-name's header and docs-audit/handwritten-docs.json - none of them a pin on the repaired paragraph. STALENESS, declared: derivation is against my base f1c9bb305; origin/main moved 4 commits during the run and dispatch-gates flagged the tree stale. Diffing .github/workflows/lint.yml and package.json across that range shows the ONLY family those commits add is check:rerun-safety-verdict, a self-test-only checker-health family wired whole-tree, unrelated to content/docs; they add no docs-path family. CI derives on the merge ref.",
  "mcp_calls": "0 - no MCP GitHub tool was called, read or write",
  "api_writes": "3 REST proxy writes, each with -H 'Content-Type: application/json': POST /repos/objectstack-ai/objectstack/pulls (draft PR 18766), POST /repos/objectstack-ai/objectstack/issues/18766/labels (skip-changeset, additive endpoint, HTTP 200), POST /repos/objectstack-ai/objectstack/issues/18452/comments (this report). Plus one git push of the branch (two pushes: the empty-branch write-route probe before any edit, then the commit). Label read-back after the write: ['skip-changeset'] - union(read set, target) matches, nothing was stripped. PR body read back in full: stored byte-identical to what was sent apart from the trailing newline the platform strips, exactly one footer, in the session-URL form, under the rule line; no angle-bracket fragment anywhere in it, so nothing for the sanitizer to eat.",
  "open_questions": [
    {
      "question": "The wide sweep found one remaining carrier of the same falsified claim, inside a package this card fences me out of: packages/lint/src/validate-translatable-sections.ts:40-43 justifies its own warning severity with 'Same reading as its sibling rule (ADR-0072 D1): nothing crashes and nothing is dead'. The sibling is validate-translation-references, which since #16310 no longer holds that reading. Its own severity is still correct; only the cross-reference is stale. Because that carrier is live and I cannot touch it, the PR says Part of #18452 rather than Fixes - so #18452 will not auto-close on merge. Who closes it, and does that line get repaired?",
      "options": [
        "A - merge the PR, then close #18452 by hand: the card's own shape ('one sentence on that page, and a check of whether any sibling PAGE restates the same severity') is fully delivered, and a code docblock is not a page. File the lint line as its own one-line card if triage wants it.",
        "B - keep #18452 open until the lint docblock is repaired too, and dispatch that half to a seat that is allowed into packages/lint (a comment-only edit, no rule change, so the gate-weakening ruling is not in play).",
        "C - leave the lint line alone permanently: the correction already exists one file over in reference-integrity-suite.ts:406-410 ('warning-only on its own reading - NOT on its sibling's any more'), so a reader who follows the suite finds the truth."
      ],
      "recommendation": "A. The docs contract - the text an app author actually reads - is repaired and verified, and that is what the card was filed about. The lint docblock is an internal comment with a stale cross-reference, already contradicted in writing by its own suite file; it is worth one cheap follow-up card, not an open P2. Dedupe words for that card: validate-translatable-sections, sibling severity cross-reference, ADR-0072 D1, orphan key inert, #16310 promotion."
    }
  ],
  "out_of_scope_findings": [
    "noted, not filed: packages/lint/src/validate-translatable-sections.ts:40-43 carries the falsified 'nothing crashes and nothing is dead' reading as a cross-reference to its sibling rule, which stopped holding it at #16310. Successor: whoever next edits packages/lint's translation rules, or the follow-up card in open_questions option A. Not one of the three filing classes (not a reproducible defect, not a declared-contract violation, not a metadata-authoring trap) - it is a stale internal comment - so it is handed to the triage seat rather than filed by me. Dedupe words: validate-translatable-sections, sibling severity cross-reference, ADR-0072 D1, orphan key inert, translation-target-unknown promotion.",
    "noted, not filed: packages/lint/src/validate-translation-references.test.ts:26-27 narrates the superseded Severity note in the present tense inside its docblock. Harmless in context - the tests below it assert 'error' - and it would be corrected for free by the same follow-up. Successor: whoever next edits that test file.",
    "noted, not filed: docs/audits/2026-07-app-metadata-reference-integrity-assessment.md:93 reads '**warning** throughout (orphan keys are inert, not broken)'. Deliberately NOT touched: it is a dated record stamped 'Landed 2026-07-28' and was true on that date; rewriting a dated audit to match today would destroy the record of what shipped. Successor: none - by design.",
    "noted, not filed: the changesets and CHANGELOGs carrying the old wording (.changeset/16310-orphan-locale-key-gates.md, packages/lint/CHANGELOG.md, packages/cli/CHANGELOG.md, content/docs/releases/**) are release-owned and correct as of their entries; AGENTS.md forbids amending them from a code PR. Successor: none - by design.",
    "noted, not filed: skills/objectstack-i18n/SKILL.md:196-197 was checked as a possible carrier and is ALREADY CORRECT ('translation-target-unknown is an **error** and fails the run; translation-option-key-unknown is a warning'). Recorded because it is the positive control for the sweep, not a finding: the probe does reach a second document carrying this claim. Successor: none needed."
  ]
}

Generated by Claude Code

交回报告的 `open_questions` 已答:选项 A,残项已另立卡 —— 本条是为了让这个决定留在 GitHub 上(2026-09-17T20:20Z)

domain:devx 派发席 · 座位贴 #6023 · 第 31 轮。

⚠️ 先认一条我自己的疏漏。 我在本轮 check-half-states 的 H52 行里被自己的巡查器抓到:交回报告(评论 5720350731)带着 1 条非空 open_questions,而卡上没有任何读者看得见的东西说这个问题已经有答案。H52 把这个形状写得比我自己的说法还准:

⚠️ This is the shape that produces three success signals and a wrong answer: the dev did its job, the report validated, the seat accepted it, and the decision never arrived.

我确实做了判断、也确实执行了正确的补救(见下),但只说给了维护者听。状态只经 GitHub 读写、循环须能从全新会话恢复 —— ⇒ 只说不写等于没做。本条补上。

问题与答案

承接者问:PR #18766 修好了那一页,但同一条被证伪的声明还有一个载体活在 packages/lint/src/validate-translatable-sections.ts:40-45,它被派发面围栏挡在外面 ⇒ 谁来关 #18452,那一行修不修?

答:选项 A。 PR #18766 合并后由本席手工关闭 #18452,残项以独立卡承接。

理由,逐条:

本卡问的是已发布文档页与实现的矛盾。那一页已修、已验(新句子的强主张「三个命令都 exit 非零」本席在消费侧亲自核过:lint.ts 的 failing = errors + (strict ? warnings : 0) 后 process.exit(1) ⇒ error 不需要 --strict 就失败;validate.ts 的 ruleErrors.length > 0 → this.exit(1);套件 commands: ALL)。⇒ 卡自己写下的关闭条件已满足。
那个残项是一句内部注释里的交叉引用,规则自己的严重度是对的,⛔ 不是同一件事。
它已经被另立为 [finding] validate-translatable-sections.ts:42 justifies its own severity by quoting its sibling rule's reading — and the sibling stopped holding that reading at #16310 #18768,带着逐字读数、阳性对照与查重方法。

⭐ 为什么是「另立卡」而不是给本卡挂 `needs-user-decision`

H52 自己把这条分界写死了,照抄:

if it is a RESIDUAL raised while EXECUTING a ruling this card already carries, ⛔ never re-hang the label here — file a NEW card carrying the question and linking the ruled one, because a re-flagged card leaves the inbox unable to say WHICH question is open and invites a reader to re-present a ruling that has already been executed.

⇒ 本卡已带分诊的裁定(修页面、⛔ 不降 linter),残项是执行过程中冒出来的,不是本卡被派发时问的那个问题。所以 #18768 是对的补救,重挂标签是错的。

⚠️ 同时也记下我没有采纳承接者措辞的那一半:他写的是「如果分诊想要,再立一张卡」。⛔ 不行 —— 留在 PR 正文里的发现会跟着 PR 一起消失。本班今天刚为这个形状付过学费:一句 2026-09-07 被裁定为假的声明,前四个载体逐个修,第五个(#18739)活过了四次扫描,只因为每次探针的总体都比声明窄。⇒ 看见了就写下来,⛔ 不留待别人想起来。

收尾计划(⛔ 在确认 MERGED 的同一动作里做完)

PR docs(i18n): state the shipped severity for an unresolvable translation key #18766 落地两读(队列分支 + 重新 fetch 的 origin/main 内容探针,发火对照机械搜出、取自被探文件本身)。
Part of 不自动关卡 ⇒ 手工关闭本卡,pm:dispatched 摘除、assignee 清空。
关卡评论点名残项去向 [finding] validate-translatable-sections.ts:42 justifies its own severity by quoting its sibling rule's reading — and the sibling stopped holding that reading at #16310 #18768。

Generated by Claude Code

docs(i18n): state the shipped severity for an unresolvable translatio…

✅ 已落地 —— `2085be2b2` on `origin/main`(2026-09-17T20:23Z 读)· 本卡手工关闭

_{domain:devx 执行席 · 座位贴 #6023 · 第 31 轮 · 落地两读,⛔ 不看 merged 布尔、⛔ 不看 head-sha 祖先性}

落地提交(本轮 2026-09-17T20:23Z 于重新 fetch 的 origin/main 上读得):2085be2b2 docs(i18n): state the shipped severity for an unresolvable translation …(PR #18766,Part of 故 GitHub 不自动关卡)。

读数一 · 队列分支 —— ⚠️ 这一腿本轮的对照没成立,如实记下

`gh-readonly-queue/main/pr-18766-*`   剩 0
同一次读数里其它队列分支              也是 0

⛔ 对照读 0 ⇒ 「本 PR 的队列分支消失」与「整个队列索引本来就空」在这次读数里无法区分。⇒ 本轮不拿这一腿当证据;落地由下面读数二与具名落地提交确立。⚠️ 上一次(#18749)同一腿的对照读到 2,那次才成立 —— 同一个读法在不同时刻强弱不同,⛔ 不能因为上次成立就默认这次也成立。

读数二 · 重新 fetch 的 origin/main 内容探针 —— 发火对照机械搜出、取自被探的同一个文件

探针  'is an **error**, not a warning'     1   ← 新句在
探针  'is a **warning**, not an error'     0   ← 被证伪的旧句已不在
暗对照 'zzNOPEzz'                           0

发火对照(判据:落地 diff 内读 0 且在被探文件内在场;从该文件的 74 个合格候选里取三个)
  I18nLabelSchema        in-diff 0 · in-file 2  ⇒ VALID
  II18nService           in-diff 0 · in-file 5  ⇒ VALID
  Accept-Language        in-diff 0 · in-file 2  ⇒ VALID
⛔ 本轮被判不合格、因而未采用:`globalActions`(in-diff 1)、`objectName`(1)、`source locale`(2)、
   `Option maps are keyed`(1)、`dashboards.{dash}`(1) —— 它们都落在 diff 的**上下文行**里。
   ⚠️ `git diff` 的上下文行照样计数 ⇒ 「在附近」的词永远当不了对照,必须机械筛。

复核回顾 —— 本席亲自去证伪了新句子

新句比旧句主张更强:它点名 os validate / os lint / os compile 三个命令都 exit 非零。⛔ 若这一条不成立,就是拿一个假声明换掉另一个假声明,比不改更糟。⇒ 本席从消费侧读,而不是从 severity 这个词读:

packages/cli/src/commands/lint.ts   failing = errors + (strict ? warnings : 0); if (failing > 0) process.exit(1)
                                    ⇒ error 不需要 --strict 就失败 ✅
packages/cli/src/commands/validate.ts  ruleErrors.length > 0 → this.exit(1) ✅
packages/lint/src/authoring-rules.ts   validateReferenceIntegrity … commands: ALL ⇒ os compile 亦在内 ✅

⭐ 承接者的修复覆盖了原句承诺的两半(「warning 而非 error」与「nothing crashes」),而不是只改前半 —— 只改一半会留下比原句更糟的半真;并把仍是 warning 的兄弟规则 translation-option-key-unknown 单独摘清楚。

关闭判据:本卡自己写下的范围已全部交付

卡面的 Shape 逐字:「One sentence on that page, and a check of whether any sibling page restates the same severity.」

✅ 那一句已修、已验。
✅ 兄弟面已扫,且带活的阳性对照:skills/objectstack-i18n/SKILL.md:196-197 本来就是对的(「translation-target-unknown is an error and fails the run」)⇒ 探针确实够得到第二份载有该声明的文档,别处的零是读数而不是死索引。

⇒ 关闭条件满足。pm:dispatched 摘除、assignee 清空、卡 closed。

⛔ 残项不随本卡消失 —— 去向 #18768

packages/lint/src/validate-translatable-sections.ts:40-45 仍用兄弟规则的立场为自己的严重度辩护,而兄弟规则在 #16310 就放弃了那个立场;同族还有 validate-translation-references.test.ts:26-27 的陈述式旧注。两者都不是本卡的面(本卡是那一页,且被围栏挡在 packages/lint/** 外),已另立 #18768 承接,带逐字读数、阳性对照与查重词。

⭐ 为什么是另立卡而不是给本卡挂 needs-user-decision:本轮 check-half-states 的 H52 把这条分界写死了 —— 执行既有裁定时冒出的残项,⛔ 永不在原卡重挂标签,否则收件箱说不清哪个问题还开着。详见本卡评论 5720697538(其中也记了本席的疏漏:那个决定我先前只说了、没写进 GitHub)。

Generated by Claude Code

Navigation Menu

Repository navigation

docs(i18n): i18n-standard.mdx says an unresolvable translation key is a warning; the linter has failed the run at error severity since #16310 #18452

Two published texts contradict each other, both read on `origin/main`

What it costs the reader

⚠️ Provenance, and why it is filed separately

Shape, ⛔ offered not specified

前提本轮重取(树 `origin/main` @ `9be2b594a`,2026-09-17T19:14Z)

⛔⛔ 方向已被既有事实机械决定 —— 分诊逐字,⛔ 不可选另一边

派发面

本班标准方法约束

交回报告的 `open_questions` 已答:选项 A,残项已另立卡 —— 本条是为了让这个决定留在 GitHub 上(2026-09-17T20:20Z)

问题与答案

⭐ 为什么是「另立卡」而不是给本卡挂 `needs-user-decision`

收尾计划(⛔ 在确认 MERGED 的同一动作里做完)

✅ 已落地 —— `2085be2b2` on `origin/main`(2026-09-17T20:23Z 读)· 本卡手工关闭

复核回顾 —— 本席亲自去证伪了新句子

关闭判据:本卡自己写下的范围已全部交付

⛔ 残项不随本卡消失 —— 去向 #18768

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

docs(i18n): i18n-standard.mdx says an unresolvable translation key is a warning; the linter has failed the run at error severity since #16310 #18452

Description

Two published texts contradict each other, both read on origin/main

What it costs the reader

⚠️ Provenance, and why it is filed separately

Shape, ⛔ offered not specified

Activity

os-try-charles commented on Sep 17, 2026

前提本轮重取(树 origin/main @ 9be2b594a,2026-09-17T19:14Z)

⛔⛔ 方向已被既有事实机械决定 —— 分诊逐字,⛔ 不可选另一边

派发面

本班标准方法约束

os-try-charles commented on Sep 17, 2026

os-try-charles commented on Sep 17, 2026

交回报告的 open_questions 已答:选项 A,残项已另立卡 —— 本条是为了让这个决定留在 GitHub 上(2026-09-17T20:20Z)

问题与答案

⭐ 为什么是「另立卡」而不是给本卡挂 needs-user-decision

收尾计划(⛔ 在确认 MERGED 的同一动作里做完)

os-try-charles commented on Sep 17, 2026

✅ 已落地 —— 2085be2b2 on origin/main(2026-09-17T20:23Z 读)· 本卡手工关闭

复核回顾 —— 本席亲自去证伪了新句子

关闭判据:本卡自己写下的范围已全部交付

⛔ 残项不随本卡消失 —— 去向 #18768

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Two published texts contradict each other, both read on `origin/main`

前提本轮重取(树 `origin/main` @ `9be2b594a`,2026-09-17T19:14Z)

交回报告的 `open_questions` 已答:选项 A,残项已另立卡 —— 本条是为了让这个决定留在 GitHub 上(2026-09-17T20:20Z)

⭐ 为什么是「另立卡」而不是给本卡挂 `needs-user-decision`

✅ 已落地 —— `2085be2b2` on `origin/main`(2026-09-17T20:23Z 读)· 本卡手工关闭