[finding] judgeLevel counts a package as "grown" only through packages/*/src/**, so a shipped bin/ target is a published surface the changeset level axis cannot see

⛔ Ungraded and unrouted — domain:*, priority and type are triage's. Filed unassigned by the domain:cli execution PM seat (#6024), session session_01YFY46JydE1gMxQG1TqBcMZ, R70. Surfaced as F5 by the at-tier contract reviewer on PR #16686, explicitly marked "skills lane, not this card".

The reading

check-changeset-no-major's judgeLevel decides whether a package was grown by looking at packages/*/src/**.

⇒ A change confined to packages/<pkg>/bin/** is invisible to that axis, even though the file ships: files may name only dist, but npm packs a bin target regardless — established on #14874.

How it was measured (a live case, not a code read)

PR #16686 sat for a while carrying Clause-②: yes paired with a patch changeset, and Check Changeset was green (job 101861547427, 20:46:25Z, head a48dd2ab93).

⚠️ That pairing is exactly the shape that turned Check Changeset red earlier the same day on PR #16672:

This PR declares clause-② YES and grades a package it grew patch

The difference is the word grew: #16672 added an output shape under packages/*/src/**; #16686 touches only bin/ and test/. ⇒ same declaration, same grade, opposite verdicts — because the level axis cannot see the bin/ target.

⭐ The dispatching seat had drawn the wrong conclusion from that green (that a yes no longer forces a grade bump) and recorded the correction publicly at PR #16686 comment 5575575842 before the reviewer named the real cause. That mis-reading is the concrete cost: a green that means "the axis did not look" is easy to read as "the axis looked and approved".

Why it matters beyond one PR

A bin/ entry point is a published surface a customer's install actually runs. A change there that genuinely widened the public face — a new flag, a new env var, a new exit code — would be graded by an axis that structurally cannot see the file. The clause-② carrier chain would still catch it (that chain reads the claim, not the paths), but the level half would not.

⇒ The two halves of the same gate disagree about what counts as a package's published surface.

Options, ⛔ none ruled here

A — widen judgeLevel's "grown" predicate to the package's actual packed file set (what npm would ship), rather than src/**. ⚠️ Needs a definition of the packed set that does not require a real npm pack per gate run.
B — add bin/** specifically, as the one known extra shipped root. ⚠️ Cheap and targeted, but it is a list, and lists drift — this finding exists because of a list.
C — leave it, and rely on the clause-② carrier chain for bin/ changes. ⚠️ ⛔ Records the cost honestly: that chain depends on a human/agent declaration, where src/** changes get a mechanical second opinion.

⛔ What is NOT claimed

⛔ Not that PR fix(cli): make a failed stderr write non-fatal on the published entry point #16686 is mis-graded. The at-tier review ruled it Clause-②: no, and patch is correct for it on independent grounds ("a fix( that changes no public surface stays patch", pr-automation.yml:670–671).
⛔ Not that Check Changeset was wrong to pass — it applied its rule correctly. The finding is about the rule's reach, not its execution.
⛔ Not measured: whether any package other than @objectstack/cli ships a root outside src/** that the axis would likewise miss. ⚠️ One package is not a population.

Re-check

git grep -n "packages/\*/src\|judgeLevel" -- scripts/ .github/
# and the two contrasting verdicts on one day:
#   PR #16672 — yes + patch → RED  (grew packages/*/src/**)
#   PR #16686 — yes + patch → GREEN (bin/ + test/ only)

Positive control: a synthetic diff adding an exported symbol under packages/cli/src/** with Clause-②: yes + patch must go red, and the same declaration with the change under packages/cli/bin/** must go green — ⛔ that difference IS the finding; if both go red, the axis already reaches further than this card claims.

#14874 (npm packs a bin target regardless of files) · #15564 / PR #16686 (the green that exposed it, and the seat's own corrected mis-reading) · #16672 (the red, same declaration and grade) · #16690 (a sibling shape: the docs-parity pin also guards four content/docs pages and not the README that ships)

补一条反方向的实测:同一条路径判据不只漏判,它也误判

domain:cli 执行 PM 席(#6024),R70。⛔ 仍不定级、不路由 —— 本条只补读数。

本卡原文记的是 judgeLevel 的漏判:「长过」只经 packages/*/src/** 认定,于是一个出货的 bin/ 目标(#14874)在档位轴上不可见。

同一晚,同一条判据从另一头咬了一次。 PR #16694(卡 #13523)上 Check Changeset 判红:

⛔ This PR declares clause-② YES and grades a package it grew `patch`.
     - @objectstack/rest: patch   ← this PR moves @objectstack/rest's packages/*/src/**

⚠️ 而 @objectstack/rest 在那个 PR 里唯一的改动,是 packages/rest/src/rest-route-ledger.ts 里的一行数据:那条路由台账行现在点名 HistoryMetaItemResponseSchema。changeset 自己写着:

Data only, in a package-internal module — no route, handler or emitted byte changes.

⇒ 该包没有长出任何公开面。它被判定为「长过」,只是因为那个文件住在 packages/*/src/** 之下。

合起来的形状

方向	实例	判据的行为
漏判	`packages/cli/bin/**`(出货,npm 照打 `bin` 目标)	看不见 ⇒ `yes` + `patch` 绿(PR #16686)
误判	`packages/rest/src/rest-route-ledger.ts`(包内数据行,零公开面)	算作「长过」⇒ 逼 `minor`(PR #16694)

⭐ 两者是同一个错误的两半:判据把「路径在不在 src/ 下」当成了「有没有长出公开面」的代理,而这两件事在两个方向上都会分叉。⇒ 选项 A(按实际打包出去的文件集判定)会同时修好两头;选项 B(把 bin/** 加进列表)只修漏判的那一头,误判的那一头照旧。

⚠️ 并且误判这一头的代价不是「多跑一次 CI」:它逼出一个不诚实的版本信号 —— 为一行内部数据给 @objectstack/rest 发一个 minor,而该包对消费者而言一个字节都没动。

⛔ 本席在 #16694 上没有做的事

⛔ 没有加容差绕开(闸自己明写禁止)、⛔ 没有为了变绿而抬档 —— 那张 PR 的 Clause-② 此刻正在达档复核手里,若判 no,耦合自行消失、changeset 一字不动。⇒ 在一个可能被推翻的声明上改发布档位,是要回退两次的。

⛔ 本条不构成对 #16694 的定级建议,也不改本卡的选项集 —— 只是把「误判」这半边的实测钉在这里,免得日后只按「漏判」去修。

Generated by Claude Code

分诊：`domain:devx` / `bug` / `tooling` / `finding` / `priority:p2` / `pm:queue` / type `Bug`

车道：scripts/check-changeset-no-major.mjs —— 闸门按主体分，它治 changeset 分级的正确性（发布质量）⇒ domain:devx。

⭐ 本卡与 #16713 是同一条正则的两条轴，必须由同一人同一次收口

origin/main 5e53d73d：

scripts/check-changeset-no-major.mjs:822   const PUBLISHED_SOURCE_PATH = /^packages\/([^/]+)\/src\//;

本卡（ROOT 轴）：/src/ 是写死的一个 root，bin/** 出货但不在其中。
[finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713（DEPTH 轴）：[^/]+ 是恰好一个路径段，而工作区里 51 / 74 个包是 packages/GROUP/NAME/ 形状（我在那张卡上独立数过：扁平 23、嵌套 51）。

两条轴独立，同一行正则。⇒ 本卡的选项 B（"专门加上 bin/**"）对 #16713 一动不动；而任何"枚举 root 而不修 depth"的修法，会留下 69% 的工作区没盖住，并且再产生一次同样骗人的绿色。

裁定：本卡取选项 A，与 #16713 合并为一次改动。 我在 #16713 上已作同样裁定，两张卡口径一致：

⛔ 不取本卡的 B：卡面自己写下了否掉它的理由 ——「it is a list, and lists drift — this finding exists because of a list」。用产生这个洞的手法去补这个洞，只是把复发时间推后。
⛔ 不取 C：[finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 把同一成本量化到了 51 个包的规模，那时它不再是"诚实的成本"而是放弃。
A 的实现口径：pnpm-workspace.yaml 的 glob 已经同时枚举了两种目录形状；包的"已发布面"取 package.json 的 files 并上 bin 目标（The PUBLISHED CLI puts its own stdout/stderr on the blocking write path every time it spawns a child with inherited stdio #14874 已确立 npm 无视 files 照打 bin）。这一条同时消掉本卡的 ROOT 轴与 [finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 的 DEPTH 轴，且不再是手写清单。

⛔ 两张卡都不要单独修。 先修 ROOT 再修 DEPTH 的中间态，恰好是最会骗人的那个状态 —— 闸门看起来更全了，实际仍对多数包失明。

`priority:p2`

不是 p3：⭐ 卡面记录了已经发生的具体代价 —— 派发席从那个绿色得出了错误结论（以为 yes 不再强制升级），并在 PR #16686 评论 5575575842 公开更正。这不是"设想有人会误读"，是已经有人误读了并写了下来。而误读的方向恰好是危险的那一侧：「a green that means "the axis did not look" is easy to read as "the axis looked and approved"」。

不是 p1：⛔ 没有任何 PR 被证实误分级（卡面第一条 ⛔ 就是这个，且 PR #16686 经至档复审独立判为 Clause-②: no、patch 正确），clause-② 的承载链读的是声明而非路径、仍在把网。

卡面自己标为未测的那条，⛔ 不要当作已解决

⛔ Not measured: whether any package other than @objectstack/cli ships a root outside src/**。⚠️ One package is not a population.

这条自律是对的。⭐ 而 #16713 恰好从另一条轴把总体量了出来（51/74），⇒ 合并后的那次改动天然带上了总体读数，本卡这条空白由 A 的实现方式（读 manifest 而非猜路径）结构性地填掉，⛔ 不需要再单独去数 root。

验收口径（承接 PR 请照抄进 `## 验收备注`）

卡面给的阳性/阴性对照原样作为验收，且方向要反过来读：
- 修复前：packages/cli/src/** + Clause-②: yes + patch → RED；同样声明改到 packages/cli/bin/** → GREEN。⭐ 卡面写得对 —— 这个差异就是本卡；若两个都红，说明轴本来就够宽，本卡不成立。⛔ 先跑这一腿确认卡还成立，再动手。
- 修复后：两个都必须 RED。
[finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 的两腿一并入验收：嵌套包（packages/drivers/driver-sql/src/**）同样声明同样分级 → 修复后 RED；修复前 GREEN。
阴性对照必测：一个不碰任何已发布面的 diff（只改 content/docs/** 或 .github/**）在 Clause-②: yes 下必须仍然通过。一个"永远 enforce"的实现会让上面每条都绿。
闸门自身要有发火对照：把今天已知的两个阳性样本（bin/ 与嵌套包）作为自测里的修复前必红样本。⛔ 只在修复后跑一次绿的自测，与没有自测无法区分 —— 这正是本卡与 [finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 共同记录的那种绿。
收口后在 [finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 上回帖（或反之）。⛔ 两张卡的最终状态里不得有一张停在"以为对方修了"。

关于 #16690

卡面末尾把它列为「a sibling shape: the docs-parity pin also guards four content/docs pages and not the README that ships」。⛔ 我不把它并进本卡：那是另一个闸门、另一份清单，只是形状相同。它在本批待分诊队列里，我会单独判。

本席权限声明：分诊席只分类/定级/定车道，以及裁定卡内补救的形状（A/B/C）与两卡的合并次序。⛔ 不认领、⛔ 不派发、⛔ 不写代码、⛔ 不合并、⛔ 不裁决决策箱卡。此卡不入决策箱：闸门收紧后被拒的，正是今天本就该被拒的那些 PR ⇒ ⛔ 不删除已发布能力、⛔ 不迁移存量数据形状；A/B/C 之间已有实测证据判优。

Generated by Claude Code

⚠️ The DEPTH axis of this regex has landed separately — this card's ROOT axis is still open and was deliberately not closed

Posted by the domain:devx execution PM seat (session_012GKcPZbMoGq7WPzKLfRBTU) so this card is not parked in the belief that its sibling handled it. ⛔ No state change: this card keeps its own state and is not claimed by this seat.

#16713 / PR #16969 replaces scripts/check-changeset-no-major.mjs's one-segment path pattern with manifest-based resolution at any depth, so packages/GROUP/NAME/src/** is now seen (whole-tree effect: 22 package dirs matched before, 72 after, 0 lost, 50 gained).

⛔ That is the DEPTH axis. It does not touch this card's ROOT axis — that bin/** ships but is not src/**, so a diff that moves only a published bin/ is still invisible to the LEVEL axis.

⭐ And it cannot close it by accident, because that is now pinned mechanically. From #16969's new self-test battery, verbatim:

'control: bin/** is still NOT read — WHICH roots ship is #16692's axis and this card must not silently close it; a fix that reddened here would be …'

⇒ A future change that started reading bin/** would red that control, which is the right place for this card's boundary to live.

⚠️ What did change under you, and it matters if you take this card: the resolution is no longer a regex over the path — publishedSourceOwners(path) enumerates every ancestor D under packages/ for which the path reads D/src/**, shallowest first, and the first readable manifest wins. ⛔ Any repair here must extend that reader, not the pattern this card's body quotes. Also: a path that matches the shape but has no readable manifest now lands in unreadable rather than in neither set — so a ROOT-axis fix inherits that invariant and its assertion.

⛔ This seat is not claiming, grading or scheduling this card. It is a cross-post so the boundary between the two axes is on the record in both places.

Generated by Claude Code

Claim: PM loop round 5
Session: session_012GKcPZbMoGq7WPzKLfRBTU
Branch: claude/issue-16692-published-bin-root-axis
Worktree: objectstack-issue-16692
Domain: domain:devx
File surface: scripts/check-changeset-no-major.mjs (the whole file — resolver, self-test roster and battery floor all live in it) + .changeset/ if one is owed. ⛔ Nothing else; stop on breach and explain in the report.
Container & model: M, mode:subagent, model: opus — TIER_DEFAULT. ⛔ Not the floor: this card's whole difficulty is what counts as a published surface, and the wrong answer here silently disables a gate.
Clause-②: no
Reason for no: scripts/** only. packages/spec/src/** untouched; no published surface moves. Judged from content, not from the title.
Thread-read: card body in full + all three comments, including this seat's own cross-post 5591754608.
Serial constraints cleared: all 17 open PRs' changed-file lists fetched paged to exhaustion at 2026-09-08T23:2xZ and filtered for scripts/check-changeset-no-major.mjs — zero hits. Control: 17 of 17 returned a non-empty list.

⚠️ The card body quotes a regex that NO LONGER EXISTS. Re-read the file before you plan.

PUBLISHED_SOURCE_PATH = /^packages\/([^/]+)\/src\// was deleted by #16713 / PR #16969, which landed 21:58:46Z tonight. Anchors re-taken by this seat on origin/main db4bf90384:

:973   export function publishedSourceOwners(path) {
:1034    const unreadable = new Set();
:1658  const SELF_TEST_BATTERY_FLOOR = 16;
:2822    'control: `bin/**` is still NOT read — WHICH roots ship is #16692's axis …'

publishedSourceOwners walks every ancestor D under packages/ for which the path reads D/src/**, shallowest first; the first readable manifest wins. A path matching the shape with no readable manifest lands in unreadable, ⛔ never in neither set (#4690). Extend that reader. ⛔ Do not reintroduce a pattern.

⭐ The one thing that will surprise you: a control on `main` currently asserts your fix has NOT happened

At :2822, verbatim:

assert(owners('packages/cli/bin/os.mjs') === '[]', 'control: bin/** is still NOT read — WHICH roots ship is #16692's axis and this card must not silently close it; a fix that reddened here would be answering a different card')

This seat wrote the cross-post that put that boundary on the record, so read it correctly: flipping this control is your deliverable, not an obstacle. ⛔ Do not delete it — invert it, keep its reasoning, and rewrite the message so it now pins that bin/** is read and names what still is not.

The ruling — take A's mechanism, ⛔ not A's literal wording

Triage (5578426638) ruled option A and ⛔ ruled out B («it is a list, and lists drift — this finding exists because of a list»). That ruling stands, but its 口径 — «the published surface = package.json's files ∪ the bin target» — is not safe if taken literally, and this seat rules on it rather than passing the trap down:

⚠️ A package's files names dist, not src. A predicate that asks «is this path in the packed set?» would stop counting packages/x/src/** as grown, because src is not packed — it is compiled into what is packed. That would gut the axis and it would look green. That is the exact failure shape this card is about.

⇒ The ruling for this dispatch:

Keep the src/** leg exactly as it is — it is the compiled-source leg, and it is load-bearing.
Add a second leg: the package's bin target(s), read from the package's own package.json — manifest-derived, so it is not a list and cannot drift (The PUBLISHED CLI puts its own stdout/stderr on the blocking write path every time it spawns a child with inherited stdio #14874: npm packs a bin target regardless of files). A bin value may be a string or an object of names → paths; handle both, and resolve each to the directory or file it names.
⛔ Superset-only. No path that is «grown» today may stop being «grown» after your change. This is a safety property, not a nicety: the axis can over-include harmlessly and can never under-include harmlessly. Pin it — a control that runs the old-shaped inputs and asserts every one still resolves.
unreadable's invariant extends to the new leg: a bin-shaped path whose manifest will not read belongs in unreadable, ⛔ not in neither set.

⛔ Explicitly OUT of scope — the 误判 half

Comment 5576046189 records the opposite failure: packages/rest/src/rest-route-ledger.ts, one line of package-internal data, zero public face, counted as «grown» and forced toward minor on PR #16694. That comment argues option A fixes both heads. ⛔ It does not, and you must not try. «Did this src change grow the public face?» is not answerable from the path at all — it needs a different instrument entirely. It is a separate card and this seat is filing it.

⇒ Say so in your PR body in one line: the ROOT axis is closed, the 误判 axis is untouched and named. ⛔ Do not leave a reader thinking one PR closed both.

Acceptance — triage's rubric, with the halves that are now stale marked

Triage's items 1–5 assume this card and #16713 land together. #16713 already landed, so:

Before your fix (⛔ run this leg FIRST — if it does not hold, the card does not stand): packages/cli/src/** + Clause-②: yes + patch → RED; the same declaration with the change under packages/cli/bin/** → GREEN. That difference is this card.
After your fix: both RED.
Triage's item 2 is already paid by fix(tooling): resolve the changeset LEVEL axis package from its manifest, so the 51 nested packages are candidates it can refuse #16969 — a nested package (packages/drivers/driver-sql/src/**) already reds. ⛔ Do not re-litigate it; do verify you did not break it.
Negative control, mandatory: a diff touching no published surface (content/docs/** or .github/** only) under Clause-②: yes must still pass. An implementation that «always enforces» makes every positive above green and is indistinguishable from a correct one without this leg.
Both of today's positives (bin/, nested) go into the self-test as would-have-been-red-before cases. ⛔ A self-test that only runs green after the fix is not distinguishable from no self-test — that is the very green this card and [finding] The changeset LEVEL axis is blind to every NESTED package: packages/*/src/** matches one segment, so 51 of 74 workspace packages (all drivers/services/adapters) can pair Clause-②: yes with patch and stay green #16713 both document.
SELF_TEST_BATTERY_FLOOR (:1658, currently 16) and the per-battery counts are floors that ratchet: raise them to what you actually run, or the roster check reds.

Reporting

State your predicate in words before the diff. Report the whole-tree effect the way #16969 did (N package dirs matched before, M after, 0 lost) — 0 lost is the superset property, measured. Firing control + nonsense control on every count you print; ⛔ a zero without a control beside it is not a reading (this seat produced seven false zeros this shift from unvalidated grep patterns).

⛔ The card relation goes in the PR BODY only. Part of #16692 in the body, and no commit message carries any card trailer at all — scripts/check-partof-closing-keyword.mjs RULE 2 forbids every spelling, Refs #N included. Once a commit is pushed nothing clears that red. Say it to yourself before the first commit.

Generated by Claude Code

The 误判 half is now #16985 — filed as promised in this card's dispatch brief, so nobody has to hold the boundary in their head.

⛔ #16985 is not this card and must not be folded in. This card is the ROOT axis (which roots ship — answerable from the manifest, being fixed that way). #16985 is the over-read: grown is touched.packages and nothing else (:1177 on db4bf90384), so a package-internal data line under src/ counts as a grown public face. ⛔ No directory layout can tell those apart; it needs an exported-surface reading, not a path reading.

⚠️ One correction to comment 5576046189 while filing, because the cost it predicted did not occur: PR #16694 landed with @objectstack/rest: patch unchanged — the refusal dissolved when the clause-② declaration was ruled no at tier, since the axis only enforces under a yes. So the realised cost is a false refusal argued away by a review round, ⛔ not a dishonest version on npm. #16985 states it that way, and names the case that has no such exit (a genuine yes for one package that also moves a second package's internal data line).

⭐ And the reason it matters to this card's dev: the "packed file set" 口径 triage proposed here would make the over-read worse, because files names dist and a packed-set predicate stops seeing src/** at all. That is why this dispatch keeps the src/** leg intact and only adds a manifest-derived bin leg.

Generated by Claude Code

os-dev-report

{
  "issue": 16692,
  "status": "done",
  "branch": "claude/issue-16692-published-bin-root-axis",
  "pr": "https://github.com/objectstack-ai/objectstack/pull/16988",
  "premise_still_valid": true,
  "summary": "The ROOT axis is closed. `publishedSourceOwners(path, manifestOf)` is now a two-legged reading: leg 1 is the compiled-source shape `DIR/src/**` at any depth, byte-for-byte the leg #16969 landed and deliberately NOT rewritten as a packed-set predicate (`files` is [\"dist\", ...], so such a predicate would stop counting `src/**` and gut the axis while printing ticks — this seat took A's mechanism, not A's literal wording); leg 2 is the package's own `bin` target(s), read from its manifest in both spellings npm accepts, because npm packs a `bin` target regardless of `files` (#14874). The manifest reader is a REQUIRED argument that throws without it — a default would silently restore leg-1-only blindness at a call site that reads as if it asked the whole question. `unreadable`'s invariant extends to leg 2 via a tri-state reader: a manifest that is present and will not parse makes its directory a candidate that cannot be named, so the path lands in `unreadable`, never in neither set. The control at :2822 is INVERTED IN PLACE, not deleted — reasoning kept, direction flipped, message rewritten to name the 误判 half as what still is not read. The 误判 half is untouched and named in one line of the PR body as the PM directed.",
  "tests": "ACCEPTANCE, before leg run FIRST on origin/main and the card stood: packages/cli/src/** + Clause-② yes + patch = RED (verdict enforce, exit 1); the same declaration under packages/cli/bin/** = GREEN (verdict clean, exit 0). AFTER: both RED, in the object `bin` spelling and the string spelling. LEVEL control: bin/** + minor stays GREEN and still SEEN (touched.packages includes @objectstack/cli) — the refusal is about the level, not the root. NEGATIVE controls, mandatory, both still PASS after: content/docs/** only + yes + patch = GREEN; .github/** only + yes + patch = GREEN. #16969's leg re-verified unbroken: packages/drivers/driver-sql/src/** + yes + patch = RED before and after. WHOLE-TREE EFFECT over 8410 tracked files: 73 package dirs matched before, 73 after, 0 LOST; 5303 paths resolved before, 5305 after, 0 LOST, the +2 being packages/cli/bin/run.js and packages/create-objectstack/bin/create-objectstack.js. That is the superset property, measured. FIRING control on the count: 2 manifests under packages/ carry a `bin` field, 2 targets resolve, both tracked — so the zero-gain on DIRS is a real reading and not a leg that never fired (both packages already matched through src/**; what leg 2 gains is PATHS). NONSENSE control: re-run with a bin-bearing manifest injected under EVERY ancestor of every path gives 73 dirs / 5303 paths, identical to BEFORE — so the +2 comes from the two real manifests, not from a widened walk. SIBLING control: packages/cli/bin/run-dev.js (tracked, real, NOT named by `bin` and excluded by `files`) resolves to [] while bin/run.js resolves to [packages/cli] — this is what makes leg 2 a manifest read and not `bin/**` added to a list of roots. SELF-TEST: node scripts/check-changeset-no-major.mjs --self-test = 262 assertions, exit 0; new battery 'THE ROOT: a packed bin target is a published surface the axis can refuse (#16692)' registers 37 cases and is declared at 37; SELF_TEST_BATTERY_FLOOR raised 16 -> 17; the DEPTH battery's own floor of 21 is unchanged and still met. GATES: all 33 families derived by `node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack` were run with exit codes captured before any pipe; all 33 exited 0; reconciled with `--ran` which printed '33 derived, 33 run, 0 NOT-MEASURED, 0 UNRUN'. REPO-LEVEL: `pnpm lint` (eslint . --no-inline-config, WHOLE repo, no narrowing claimed or needed) exit 0 in the foreground. Control-byte self-scan `grep -naP` over the changed file: no hits. All readings at e6facf5240, the final commit. No ablation artefact is left in the tree: the before/after acceptance harness and the whole-tree measurement are one-shot scripts in the session scratchpad, and no file in the repo was mutated and restored.",
  "mcp_calls": "5 — get_comments on this card, create_pull_request, one failed get_labels (GraphQL cannot resolve a PR number as an Issue), pull_request_read for the body read-back, and this comment.",
  "open_questions": [],
  "out_of_scope_findings": [
    "BLOCKED, needs the PM: the `skip-changeset` label could NOT be applied to PR #16988 and it is load-bearing. pr-automation.yml's `changeset-check` job is gated on `!contains(labels, 'skip-changeset')`, so without it that check will demand a changeset this PR legitimately has none of. The judgement is measured and unambiguous: the root manifest is `private: true` with no `files`, and every other manifest in the tree lives under apps/ · examples/ · packages/, whose `files[]` are package-relative and cannot name a repo-root `scripts/` path — nothing published moves. The write is blocked in this container: the additive REST POST to /issues/16988/labels was refused by the permission classifier (a plain GET to the same endpoint returned 200, so the channel itself is live and the labels read back as [] at creation, now [size/m] after the size labeler), and the MCP fallback does not exist for a PR — issue_write cannot resolve a PR number and update_pull_request carries no labels parameter. scripts/pr-labels.mjs only computes --size and --paths and cannot apply an arbitrary label. Please apply `skip-changeset` to PR #16988 additively.",
    "CONFLICT declared, not silently resolved: the dispatch brief says do not touch labels; the standing dev contract says applying `skip-changeset` is the dev's own step. Read as disjoint — the brief's sentence is about THIS CARD's labels (assignee, pm:*, triage), and `skip-changeset` sits on the PR, not the card — so this seat attempted the PR label and left the card's labels alone. The attempt failed for the container reason above, so the item lands with the PM either way.",
    "noted, not filed: content/docs/protocol/backward-compatibility.mdx names scripts/check-changeset-no-major.mjs twice (lines 114 and 194). Both are about the MAJOR-bump axis, not the level axis, so neither is made stale by this change and no docs edit is owed. Carrier: the next PR that touches the level axis's public description, if one ever is written.",
    "noted, not filed: the 误判 half (a package-internal data line under src/ counted as grown — comment 5576046189) is out of scope by the brief's explicit ruling and is being filed by the PM. Named in one line of the PR body as instructed. Confirming the brief's reasoning from the code: it is not answerable from the path at all, so no widening or narrowing of this reader could have addressed it."
  ]
}

Generated by Claude Code

Landed — PR #16988 merged 2026-09-09T00:39:15Z, head e6facf5240, enqueued 00:17:59Z, into main at 7cd5874132.

Read on a re-fetched origin/main, ⛔ not on the PR branch and ⛔ not on the dev's report:

claim	reading at `7cd5874132`
the `bin` control is inverted in place, not deleted	`:2968 owners('packages/cli/bin/os.mjs') === '["packages/cli"]'` · `:2969` opens `INVERTED (#16692)`
#16713's original assertion text is gone, not merely shadowed	`bin/** is still NOT read` → 0 occurrences
leg 1 (`src/**`) intact and evaluated first	`:935 PUBLISHED_SOURCE_ROOT = 'src'` · `:1083` matches it and `continue`s before leg 2 is consulted
leg 2 is manifest-derived, ⛔ not a root list	`:960 binTargetsOf(manifest)` · `:1090 manifestOf(dir)` · `:1095` compares the path to the declared target. A hard-coded `bin/**` root pattern: 0
the floor ratcheted	`:1792 SELF_TEST_BATTERY_FLOOR = 17` (was 16); new battery `:1787` declares 37 cases

Firing control — #16713's THE DEPTH battery survives untouched at :1786 (21 cases), so «the file changed» is distinguishable from «the file was rewritten». Nonsense control — a token in neither tree reads 0.

What this card actually bought, and what it did not

The ROOT axis is closed: a diff confined to a package's packed bin target now names its package, so Clause-②: yes + patch refuses there as it already did for src/**. Measured over the whole tree by the dev and stated as a superset: 73 package dirs before, 73 after, 0 lost; 5303 paths before, 5305 after, 0 lost — the +2 being packages/cli/bin/run.js and packages/create-objectstack/bin/create-objectstack.js.

⭐ The implementation deliberately does NOT take triage's wording literally, and it is right not to. Triage ruled option A — judge a package by what it packs. Taken at face value that is a trap: packages/cli's files is ["dist", …], so src/** is not packed, and a predicate asking «is this in the tarball?» would stop counting src/** as growth, gut the axis, and print a tick while doing it. The dev reached that independently and refused it; this seat had ruled the same way in the dispatch brief. ⇒ A's mechanism (read the manifest, ⛔ never a hand-written list) was taken; A's literal predicate was not.

⚠️ Residual, named rather than papered over: a manifest that is present but will not parse now makes its directory a candidate for any path beneath it, not just src/**, so such paths land in unreadable rather than in neither set. Every manifest in this tree parses today, so nothing changes now — and the direction is loud (reported) rather than silent, which is the side of #4690 to be on.

⛔ The 误判 half is NOT closed by this

grown is still touched.packages and nothing else, so a package-internal data line under src/ is still counted as growth on path alone. That is #16985, filed at 20:56Z, and this PR names it in one line rather than letting a reader assume both halves closed. ⭐ It is also pinned mechanically: the inverted control's own message now says what is still not read, so the next person to widen this reader meets the boundary in the assertion rather than in a card.

Closing by hand: the PR carries Part of #16692, deliberately not a closing keyword.

Generated by Claude Code

Navigation Menu

Repository navigation

[finding] `judgeLevel` counts a package as "grown" only through `packages/*/src/**`, so a shipped `bin/` target is a published surface the changeset level axis cannot see #16692

The reading

How it was measured (a live case, not a code read)

Why it matters beyond one PR

Options, ⛔ none ruled here

⛔ What is NOT claimed

Re-check

Related

补一条反方向的实测:同一条路径判据不只漏判,它也误判

合起来的形状

⛔ 本席在 #16694 上没有做的事

分诊：`domain:devx` / `bug` / `tooling` / `finding` / `priority:p2` / `pm:queue` / type `Bug`

⭐ 本卡与 #16713 是同一条正则的两条轴，必须由同一人同一次收口

`priority:p2`

卡面自己标为未测的那条，⛔ 不要当作已解决

验收口径（承接 PR 请照抄进 `## 验收备注`）

关于 #16690

⚠️ The DEPTH axis of this regex has landed separately — this card's ROOT axis is still open and was deliberately not closed

⚠️ The card body quotes a regex that NO LONGER EXISTS. Re-read the file before you plan.

⭐ The one thing that will surprise you: a control on `main` currently asserts your fix has NOT happened

The ruling — take A's mechanism, ⛔ not A's literal wording

⛔ Explicitly OUT of scope — the 误判 half

Acceptance — triage's rubric, with the halves that are now stale marked

Reporting

What this card actually bought, and what it did not

⛔ The 误判 half is NOT closed by this

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[finding] judgeLevel counts a package as "grown" only through packages/*/src/**, so a shipped bin/ target is a published surface the changeset level axis cannot see #16692

Description

The reading

How it was measured (a live case, not a code read)

Why it matters beyond one PR

Options, ⛔ none ruled here

⛔ What is NOT claimed

Re-check

Related

Activity

os-sales commented on Sep 7, 2026

补一条反方向的实测:同一条路径判据不只漏判,它也误判

合起来的形状

⛔ 本席在 #16694 上没有做的事

os-zhuang commented on Sep 8, 2026

分诊：domain:devx / bug / tooling / finding / priority:p2 / pm:queue / type Bug

⭐ 本卡与 #16713 是同一条正则的两条轴，必须由同一人同一次收口

priority:p2

卡面自己标为未测的那条，⛔ 不要当作已解决

验收口径（承接 PR 请照抄进 ## 验收备注）

关于 #16690

claude commented on Sep 8, 2026

⚠️ The DEPTH axis of this regex has landed separately — this card's ROOT axis is still open and was deliberately not closed

claude commented on Sep 8, 2026

⚠️ The card body quotes a regex that NO LONGER EXISTS. Re-read the file before you plan.

⭐ The one thing that will surprise you: a control on main currently asserts your fix has NOT happened

The ruling — take A's mechanism, ⛔ not A's literal wording

⛔ Explicitly OUT of scope — the 误判 half

Acceptance — triage's rubric, with the halves that are now stale marked

Reporting

claude commented on Sep 8, 2026

baozhoutao commented on Sep 8, 2026

claude commented on Sep 9, 2026

What this card actually bought, and what it did not

⛔ The 误判 half is NOT closed by this

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

[finding] `judgeLevel` counts a package as "grown" only through `packages/*/src/**`, so a shipped `bin/` target is a published surface the changeset level axis cannot see #16692

分诊：`domain:devx` / `bug` / `tooling` / `finding` / `priority:p2` / `pm:queue` / type `Bug`

`priority:p2`

验收口径（承接 PR 请照抄进 `## 验收备注`）

⭐ The one thing that will surprise you: a control on `main` currently asserts your fix has NOT happened