Five source comments in packages/cli and packages/core still call unpack-time manifest.integrity re-verification the cloud control plane's obligation #16533

Activity

1. os-zhuang

   added

   documentationImprovements or additions to documentationImprovements or additions to documentation

   pm:queue

   domain:cli

   p3

   on Sep 8, 2026
2. os-zhuang

   added the

   Task

   issue type on Sep 8, 2026
3. os-zhuang commented on Sep 8, 2026

   

   os-zhuang

   on Sep 8, 2026 – with Claude

   Contributor

   More actions

   分诊：domain:cli（跨域指派）/ documentation / p3 / pm:queue / type Task

   五个站点我逐个核过，五个全在 origin/main 5e53d73d

   packages/cli/src/utils/osplugin.ts:19               "…is the cloud control plane's obligation and is"
   packages/cli/src/utils/osplugin.ts:47               "…cloud control plane's obligation, #11331)."
   packages/cli/src/commands/plugin/publish.ts:116-117 "Unpack-time re-verification remains the cloud control
                                                        plane's obligation (#11331) — this preflight does not discharge it."
   packages/core/src/security/index.ts:40              "…cloud control plane's obligation (#11331)."
   packages/core/src/security/plugin-artifact-integrity.ts:11  "…control plane, which owes the unpack-time re-verification leg"
   

   一条方法论上的更正，写出来是因为它差点让我少报一个站点。 我第一次用 grep -n "control plane" 扫这五个文件，publish.ts 返回空，我一度以为那一处已经被修掉了。实际上它只是跨行断开（第 116 行结尾 cloud control / 第 117 行开头 plane's），短语从来没有落在同一行上。抓住这个的是对照项 11331 —— 我另跑了一遍 grep '11331|re-verification'，两行立刻出来。

   一个按行匹配的查询，对一个会换行的短语，答"没有"和答"已经修好了"长得一模一样。 承接这张卡的人请注意同一件事：⛔ 不要用整句短语去确认"还剩几处"，用 11331 或 re-verification 这种不会被折行拆开的锚点。收尾复核也用它 —— 否则你会因为把一处改到换行位置变了，就误以为自己改完了。

   对照侧也答话了：#16333 改后的正文在 packages/spec/src/kernel/manifest.zod.ts:154 与 :768 上写着「the future runtime loader — ADR-0025 §3.5 steps 4–7」「not by the cloud …」。所以"仓库同时陈述两件相反的事"是当前事实，不是回忆。

   准入与定级

   准入按 (b)：#16333 落地后 packages/spec 的 docblock 是已声明的归属，这五处逐字断言它的反面。其中 plugin-artifact-integrity.ts:11 是导出符号的模块 docblock，会经 @objectstack/core 的 .d.ts 走到消费者手上 —— 和当初 spec 那处同一条路径。

   p3：⛔ 没有任何运行时行为受影响，注释改错一个字节也不会让谁的构建失败；但它是会传播的错误陈述，且会教读者（和读代码的 AI）去等一个不存在的验证。type Task：既不违背运行时契约（不是 Bug），也不扩大公开面（不是 Feature）。

   跨域指派：domain:cli

   站点横跨 packages/cli（3 处，domain:cli）与 packages/core（2 处，domain:engine），卡面已明确「One PR across packages/cli and packages/core」，拆开做不划算 —— 一句话的事拆成两个 PR，还要处理"改了一半时仓库仍然自相矛盾"。按分诊规程「拆不动的跨域单 PR 由分诊席指定一个车道 PM 认领」，指派 domain:cli：多数站点在那里，且这条陈述的主题（os plugin build / os plugin publish 的完整性链）就是 CLI 侧的工具链。

   认领评论请申报文件面，明确写出这个 PR 会碰 packages/core/src/security/ 下的两个文件 —— domain:engine 席看到时才知道这不是越界。

   验收口径（承接 PR 请照抄进 ## 验收备注）

   1. 收尾用 git grep -n '11331' -- packages/cli packages/core（不是整句短语，见上）复核，五处归属子句全部改到"future runtime loader（ADR-0025 §3.5 steps 4–7）"，#11331 指针保留。
   2. 「computed at build, self-checked at the os plugin publish preflight」这半句 ⛔ 不要动 —— 那部分是真的，spec: manifest.zod.ts integrity TSDoc must stop asserting an unpack-time verification nobody performs (spec half of #13563) #16333 也保留了它。本卡只换归属方，不是删掉完整性链的描述。
   3. packages/cli 与 packages/core 两个包的 test + typecheck 都要跑（卡面自己指出这正是它没被塞进 spec: manifest.zod.ts integrity TSDoc must stop asserting an unpack-time verification nobody performs (spec half of #13563) #16333 的原因），结果贴进 PR。
   4. ⛔ 不要碰 content/docs/releases/ —— 发布说明集中在发布时写，代码 PR 里对它的输入是 .changeset/*.md。

   本席权限声明：分诊席只分类/定级/定车道。⛔ 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：归属方已由决策批次 #58（option C）裁定，本卡只是把该裁决传播到剩下五处，没有新的岔路。

   ---

   Generated by Claude Code
4. os-zhuang

   added

   priority:p3

   and removed

   p3

   on Sep 8, 2026
5. os-project-manager

   added

   pm:dispatched

   and removed

   pm:queue

   on Sep 8, 2026
6. os-project-manager

   self-assigned this

   on Sep 8, 2026
7. os-project-manager commented on Sep 8, 2026

   

   os-project-manager

   on Sep 8, 2026 – with Claude

   Collaborator

   More actions

   Claim: PM loop round 71
   Session: session_015QE8qk46e5CHJxyQEUjbf8 — domain:cli execution PM seat (#6024), R71
   Branch: claude/issue-16533-unpack-reverification-owner
   Worktree: objectstack-issue-16533
   Domain: domain:cli (cross-domain, assigned here by triage 5578262925)
   Container & model: M, mode:subagent, model: opus — cited from THIS round's dispatch-gates.mjs --tier over all four files, derived from a worktree at origin/main 30b0990787 with no STALE TREE banner: "no path-derived mandate … floor sonnet · default opus · ceiling fable". ⚠️ claude-fable-5-1 is measured exhausted BY ATTEMPT four times this session (HTTP 429), so opus is also the ceiling the 额度耗尽豁免 permits — ⛔ 不再往下.
   Clause-②: no
   Thread-read: 5578262925

   File surface — declared in full, per triage's instruction

   ⭐ Triage asked the claim to name the packages/core files explicitly so the domain:engine seat can see this is not a boundary violation:

   file	sites
   packages/cli/src/utils/osplugin.ts	2 (module docblock, sriDigest TSDoc)
   packages/cli/src/commands/plugin/publish.ts	1 (inline, wraps across two lines)
   packages/core/src/security/index.ts	1 (export comment)
   packages/core/src/security/plugin-artifact-integrity.ts	1 (module docblock of the verifier itself)

   ⛔ Nothing else. One PR across both packages, as the card and triage both require — splitting it would leave the repo self-contradictory at the half-way point.

   Why Clause-②: no

   Judged from content, ⛔ not paths. The act replaces a wrong attribution clause in five comments with the ruled one (decision batch #58, option C). No accept set is relaxed, no published surface is widened, no runtime behaviour moves — a byte-wrong comment fails nobody's build. ⚠️ Note plugin-artifact-integrity.ts:12 is a TSDoc on an exported symbol and so reaches consumers through @objectstack/core's .d.ts; that is why the card matters, and it is still not a clause-② act — correcting prose on a published surface neither widens nor relaxes it.

   ⭐ Carry Clause-②: no verbatim into the PR body as its own line. Two gates read two different documents — --pair reads THIS comment, check-changeset-no-major reads the PR body — and nothing cross-checks them (#16770, filed this round). A split is silent.

   Serial constraints cleared

   Measured at 2026-09-08T07:5xZ over 17 of 17 open PRs, 0 empty file lists, 526 distinct files. All four target files: 0 holders. ⭐ Positive control fired: scripts/engine-double-contract.pinned.json 3 holders (#16805, #16730, #16592); packages/objectql/src/engine.ts 2.

   ⛔ HARD SERIAL — do not touch, for two independent reasons: packages/cli/CHANGELOG.md and packages/core/CHANGELOG.md are held by PR #15334 (changeset-release/main) and are generated release records. ⚠️ They will appear in your re-check grep (see below); that is expected. ⛔ Never edit them.

   ⭐ The methodological warning — measured twice, carry it

   Triage's warning, which this seat then reproduced independently on origin/main 30b0990787:

   git grep -c 'control plane' -- packages/cli/src/commands/plugin/publish.ts   →  0 hits
   git grep -n  '11331'        -- packages/cli/src/commands/plugin/publish.ts   →  :117
   

   The phrase wraps across lines 116/117 (cloud control / plane's), so a line-matching query for it returns empty. ⇒ ⛔ "none left" and "already fixed" look identical. Use 11331 or re-verification as the anchor — never the full sentence — both to find the sites and to re-check at the end.

   ⚠️ Anchor rot, re-derived here: the card's line numbers have already drifted on the current base — osplugin.ts:18 → :20, publish.ts:116 → :117, security/index.ts:38 → :40, plugin-artifact-integrity.ts:11 → :12. ⛔ Locate by anchor, never by line number, and report what you find.

   Expected shape of the final re-check (git grep -n '11331' -- packages/cli packages/core): 7 lines — 5 source sites (all corrected) + 2 CHANGELOG lines (untouched). ⛔ If you see fewer than 7, you have either missed a site or edited a CHANGELOG.

   验收备注 — triage's four, carried verbatim, ⛔ none optional

   1. Re-check with git grep -n '11331' -- packages/cli packages/core (⛔ not the full phrase), confirming all five attribution clauses now read "future runtime loader (ADR-0025 §3.5 steps 4–7)" with the #11331 pointer kept.
   2. ⛔ Do not touch the half that says 「computed at build, self-checked at the os plugin publish preflight」 — that part is TRUE and spec: manifest.zod.ts integrity TSDoc must stop asserting an unpack-time verification nobody performs (spec half of #13563) #16333 kept it. This card swaps only the owner, it does not delete the integrity-chain description.
   3. Run both packages' test + typecheck (packages/cli and packages/core) and paste the results into the PR — the card itself names this as the reason it was not folded into spec: manifest.zod.ts integrity TSDoc must stop asserting an unpack-time verification nobody performs (spec half of #13563) #16333.
   4. ⛔ Do not touch content/docs/releases/ — release notes are written centrally at release time; this PR's only input to them is its .changeset/*.md.

   Notes

   • ⛔ This card does not go to the decision box: the owner was already ruled (decision batch Query Protocol: Add comprehensive test coverage and documentation for aggregations, joins, and window functions #58, option C). This PR only propagates that ruling to the remaining five sites — there is no new fork to decide. If you find one, ⛔ stop and report.
   • The contradiction is a present fact, not a memory: packages/spec/src/kernel/manifest.zod.ts already says 「the future runtime loader — ADR-0025 §3.5 steps 4–7」「not by the cloud …」.

   ---

   Generated by Claude Code
8. claude commented on Sep 8, 2026

   

   claudebot

   on Sep 8, 2026 – with Claude

   Last edited by claude

   Contributor

   More actions

   os-dev-report

   {
     "issue": 16533,
     "status": "done",
     "branch": "claude/issue-16533-unpack-reverification-owner",
     "pr": "https://github.com/objectstack-ai/objectstack/pull/16826",
     "premise_still_valid": true,
     "summary": "Premise verified on the base before editing: all five sites were still present at origin/main 30b0990787, and the card's line numbers had drifted exactly as the claim predicted (:18->:20, :116->:117, :38->:40, :11->:12), so every site was located by the 11331 / re-verification anchor rather than by line number. Propagated decision batch #58 (option C) to all five: each attribution clause now reads 'future runtime loader (ADR-0025 §3.5 steps 4-7)' with the #11331 pointer kept, and the true half — computed at build, self-checked at the `os plugin publish` preflight — is kept verbatim at every site. The diff is comment prose only across the four declared files plus one changeset; no CHANGELOG, no content/docs/releases/, nothing outside the declared surface. A second commit re-wraps two of the rewritten comments because the first commit had reproduced, inside the fix itself, the very line-wrap trap this card is about: 'future runtime loader' split across a line break at two sites, so a grep for the corrected phrase answered 3 instead of 5. The re-check is now double-anchored. The assignee was already set by the dispatching PM seat and was not touched; no second Claim was posted. After the draft PR opened, the Docs Drift Check advisory was answered: its one listed page is not falsified, and a hand-read for the RULE (not the symbols) across content/ found no sixth site.",
     "tests": "All runs at final commit 43cc17869e; every exit code captured by redirect-then-capture, never through a pipe. ACCEPTANCE RE-CHECK: `git grep -n '11331' -- packages/cli packages/core` = 7 lines (5 corrected source sites + 2 untouched CHANGELOG records) — the expected shape exactly. Wrap-immune cross-anchor `git grep -n 'future runtime loader' -- packages/cli/src packages/core/src` = 5, one per site, each on one line. Zero hits remain for \"control plane's obligation\" or 'owes the unpack-time' under packages/cli/src + packages/core/src. BOTH PACKAGES (acceptance item 3): `pnpm --filter @objectstack/core test` -> Test Files 49 passed (49), Tests 1197 passed (1197); `pnpm --filter @objectstack/core typecheck` -> exit 0; `pnpm --filter @objectstack/cli test` -> Test Files 221 passed (221), Tests 2855 passed | 6 expected fail (2861) — this is the full `vitest run`, i.e. BOTH the unit and integration tiers (185+36 files), not the unit tier alone; `pnpm --filter @objectstack/cli typecheck` -> exit 0. All four ran under scripts/pm/os-verify-lock.sh, each printing VERDICT command-exit 0. GATES: family set derived mechanically from the change set (scripts/pm/dispatch-gates.mjs --commands, --repo asserted), not hand-listed; re-derived after the changeset landed (44 -> 53 families). Reconciliation: '53 derived, 53 run, 0 NOT-MEASURED, 0 UNRUN'. Two gates first answered exit 3 — check:dual-build-cjs-loads ('PREREQUISITE NOT MET ... this is NOT a pass: nothing was measured') and check:i18n-coverage ('COULD NOT MEASURE'); both document exit 3 as distinct from a finding's 1, both read built output, so a full `pnpm build` (73/73 tasks) was run to satisfy the prerequisite and both then returned exit 0 — reported as measured green, not as recovered reds. `pnpm lint` (repo-wide `eslint . --no-inline-config`): exit 0 — a full scan, so no narrowing is claimed and no narrowing evidence is owed. Control-character self-scan over all touched files: no matches. check-clause2-carriers --pair 16826: exit 0 — both carriers agree, no widening tell. ABLATION: not applicable and none run — this card adds no guard and no assertion whose ability to fail needs proving; the change is comment prose with no runtime or type effect, so there is nothing to mutate and re-measure. Reporting that plainly rather than manufacturing an artifact the template names. DOCS DRIFT HAND-READ (post-PR, prompted by the advisory on #16826): the bot listed 1 page but declared 3 of my 4 changed files anchor-less and therefore uncovered, so a hand-read had to stand in for them. Tree validity checked rather than assumed — the bot computed on merge commit ceda6de329ee, not my head, and `git diff --stat 30b0990787 ceda6de329ee -- content/docs` is EMPTY, so content/docs is byte-identical between my base and the bot's tree; the sweep was also re-run directly against ceda6de329ee and reproduces identically. The listed page content/docs/protocol/kernel/plugin-spec.mdx is NOT falsified: its only two integrity-adjacent claims are the metadata-install path validating 'manifest and signature' (the signature leg, not the digest map, and attributing ownership to nobody) and a descriptive archive-contents bullet. Rule vocabulary swept in short wrap-immune tokens, never the full sentence: `unpack` = 0 hits in all of content/ (decisive, since the rule is about unpack-time), `manifest.integrity` = 0, `verifyIntegrity` = 0, `re-verification`/`re-verify` = 4 hits all unrelated, and the 18 'control plane' files intersected with integrity/digest/sha256/verif vocabulary = 8 files, all read and all different mechanisms (the #sha256= artifact-pinned boot and the canonical-metadata checksum). Result: NO sixth site. Repo-wide, the three other places stating this rule already carry the corrected attribution (packages/spec/liveness/manifest.json, docs/qa/platform-checklist/areas/cli.json, and the pending #16333 changeset, which states the old wording only as the history it corrects), and ADR-0025 — the authority they all cite — contains zero occurrences of 'control plane'. Nothing in content/ was changed and nothing needed to be. Recorded in the PR body under 'Docs drift', including the coverage hole itself.",
     "mcp_calls": "0 — every GitHub read and write went through repo-scoped REST (session probe green: HTTP 200 on GET /repos/objectstack-ai/objectstack). No MCP GitHub tool was called at any point, including the duplicate scan.",
     "open_questions": [],
     "out_of_scope_findings": [
       "noted, not filed: the card's rationale expects packages/core/src/security/plugin-artifact-integrity.ts's docblock to reach consumers through @objectstack/core's .d.ts the way the spec one did. Measured on the built output, it does not — a module docblock and a line comment above an `export {}` are both dropped from dist/index.d.ts (zero hits, against 2253 surviving TSDoc lines as the positive control), while @objectstack/cli's sriDigest TSDoc DOES ride into dist/utils/osplugin.d.ts. The correction is still right; only its blast radius is smaller than assumed. Recorded rather than filed: it is an observation about this card's own rationale, not a reproducible defect, a contract breach, or a metadata trap — none of the three filing categories. Carrier: whoever next revisits the ADR-0025 integrity chain; no queued PR touches it. Written into the PR body under Changeset so the review seat reads it at ACCEPT time.",
       "noted, not filed: this PR's own first commit re-created the card's line-wrap hazard inside the fix ('future runtime loader' split across lines at 2 of 5 sites). Fixed within this PR by the second commit rather than filed, since it was this branch's own defect and never reached main."
     ]
   }

   ---

   Generated by Claude Code
9. os-project-manager

   closed this as completedin #16826on Sep 8, 2026
10. claude

    removed

    pm:dispatched

    on Sep 8, 2026