finding(devx): a test that reads SOURCE off disk creates an affected-package edge turbo ls --affected structurally cannot see — every declared narrowing built on it has an unmeasured hole

Filed by the objectui PM seat (session session_01YBWFb5YgMU5dw8p2VKj16S), surfaced by the #7773 dev during PR #8191 and captured here rather than left in a PR body.

The instance that revealed it

packages/types/src/__tests__/object-kanban-group-by-limit-7322.test.ts carries an off-disk ledger, VIEW_LEVEL_ALIAS_SITES, which pins live view-level alias reads by exact source text in other packages.

PR #8191 edits packages/plugin-view/src/ObjectView.tsx and packages/plugin-list/src/ListView.tsx. Those edits are therefore capable of reddening a test in packages/types — a package the diff does not touch and does not import.

turbo ls --affected does not name packages/types for that diff, and structurally cannot: it derives the affected set from the dependency graph, and this dependency is a file read, not an import. There is no edge in the graph to follow.

The #8191 dev found this by reading, ran the ledger explicitly, and it stayed green. Nothing is broken today. The finding is the hole, not this instance.

Why this is worth more than a note

Every seat in this loop uses turbo ls --affected to justify a declared narrowing — "these are the affected packages, I ran all of them, nothing went unrun." That sentence has appeared in most dev reports this session. If the affected set silently omits a package whose tests read the changed files off disk, then:

the narrowing is wrong, not merely incomplete;
it is wrong silently, because the instrument reports success;
and the person relying on it has no way to notice, since the missing package looks unrelated by every graph-based measure.

CI runs the full farm regardless, so this is not a "green PR that should have been red" risk today. It is a method risk: a widely-used justification has an unmeasured hole, and the discipline this repo applies everywhere else — a reading needs a control that could have failed — is not being applied to the affected set itself.

What is NOT claimed here

⛔ I have not enumerated how many such off-disk source-scanning tests exist, or which packages they couple. VIEW_LEVEL_ALIAS_SITES is the one instance measured. The census is the first piece of work this card needs, and it should carry a control that fires (a known off-disk reader found by the same query shape).

⛔ I am also not asserting the fix. Candidate directions, unranked and for triage:

teach the affected-set computation about these ledgers (an explicit declaration file listing the source paths each ledger reads);
or make the ledgers import what they assert on rather than reading text;
or require any PR touching a path named in a ledger to run that ledger's package explicitly, and gate on it;
or accept the hole and require declared narrowings to say they rest on a graph that cannot see file reads.

Related, and why none of them cover this

The nearest family is real but is about a different hole — turbo inputs not being hashed, so a cached task goes stale:

turbo test inputs have the same out-of-package hole #3514 closed for type-check — the root vitest.config.mts every package re-exports is not hashed #4178 — the root vitest.config.mts is not hashed for test
turbo lint inputs have the same out-of-package hole as #3514 / #4178 — the root eslint.config.js that IS the lint program is not hashed; measured frozen at 79872f192ee4828c #4184 — the root eslint.config.js is not hashed for lint
guard: turbo type-check 的手维护 inputs 名单已两次因"包程序越出包目录"失真 —— 评估从 tsconfig include 派生或加钉扎测试 #3514 — the same shape closed earlier for type-check
finding(ci): a claim about a package's BUILT artifact cannot be pinned — turbo's test task builds dependencies only, so a dist-importing test is NOT MEASURED in CI #7183 — a dist-importing test is NOT MEASURED because turbo's test builds dependencies only

Those concern staleness of a task's inputs. This concerns membership of the affected set. All four are closed. Dedup search returned those five results, so the channel was live — the zero for this shape is a reading.

Claim: PM loop round R46 — domain:devx @ objectui execution seat.
Session: session_01FhBNJcLRZLe8M87VcUgpKr
Branch: claude/issue-8195-off-disk-source-readers-census
Worktree: /home/user/objectui-issue-8195
Domain: domain:devx
File surface: MEASUREMENT ONLY — no product or gate code. Read-only over the tree; the census and its control are posted on this card. A PR only if the census is a pinnable test that adds no new file outside scripts/__tests__/.
Dispatch: one os-dev subagent of this session (mode:subagent, model: opus); the dispatch note with the brief follows on this card. Labels pm:queue → pm:dispatched and assignee set in one write, read back at 2026-09-07T07:44Z.

Generated by Claude Code

Dispatched (measurement only) — domain:devx @ objectui execution seat, PM session session_01FhBNJcLRZLe8M87VcUgpKr, R46, 2026-09-07T07:46Z. Devs in flight 3/3 (#7882, #8216, this); PRs #8271 and #8276 are in landing.

Dev. One os-dev subagent of this session on branch claude/issue-8195-off-disk-source-readers-census, worktree /home/user/objectui-issue-8195, base origin/main at b6f821a8b or newer. Race re-read at 07:44Z: the newest Claim: on this card is 5567119075, naming that branch.

Brief, in one paragraph. The card asks for a census with a control before any fix, and lists four unranked directions; this dispatch delivers the census and a four-axis recommendation, ⛔ no gate or product code (a pinnable census test under scripts/__tests__/ is the only code allowed, and only if it carries a control that fires). Method: enumerate every *.test.{ts,tsx} that reads files off disk (readFileSync / readdirSync / globs / repoRoot-relative resolves) under packages/*/src, apps/*/src, scripts/__tests__; classify each read target by workspace package into (a) another package's source — the hole, (b) own package, (c) repo-root files outside any package graph, (d) fixtures; the known instance (VIEW_LEVEL_ALIAS_SITES, object-kanban-group-by-limit-7322.test.ts:100) must classify (a) by the same query, and one own-package reader must classify (b). Cross-check with turbo ls --affected at TWO points (a scratch, unpushed whitespace commit on plugin-view/src/ObjectView.tsx, then one more (a)-class edge from the census), confirming the reader package is absent; reset the worktree to origin/main afterwards. Measure the method risk: how many of the last 30 merged PR bodies cite turbo ls --affected as a narrowing justification. Report table + edge list + recommendation on this card.

Premise check (read 07:43Z on b6f821a8b): the instance is live at lines 100/372, last touched by PRs #8227 and #7774; the rough upper bound of the population is 138 test files (grep for readFileSync|readdirSync under the test globs above), so the classification is the work.

After the report: this seat routes the fork — the four directions touch turbo.json, scripts/**, or every reader test, so the choice is not this lane's to make alone; expect pm:retriage with the census attached, or a decision card if the recommendation crosses a human floor.

Generated by Claude Code

os-dev-report

{
  "issue": 8195,
  "status": "measured",
  "branch": "claude/issue-8195-off-disk-source-readers-census",
  "pr": null,
  "base_sha": "b6f821a8b7feda5636521b603c3e974fa52d5db3",
  "premise_still_valid": true,
  "census": {
    "candidates": 295,
    "in_package_candidates": 190,
    "root_scripts_candidates": 105,
    "class_a": 73,
    "class_a_breakdown": {
      "point_hole": 44,
      "workspace_or_repo_sweep": 17,
      "already_covered_by_dep_graph": 12
    },
    "class_b": 39,
    "class_c": 30,
    "class_d": 48,
    "root_row": {
      "note": "scripts/__tests__ + eslint-rules: the root package is NOT a turbo workspace package (turbo ls lists 46; object-ui-monorepo is absent), so these are outside the instrument entirely",
      "point_hole": 33,
      "sweep": 40,
      "class_c": 31,
      "class_d": 1
    },
    "manual": 78,
    "edges": [
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/components",
        "files": 12
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/example-schema-catalog",
        "files": 9
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/app-shell",
        "files": 7
      },
      {
        "reader": "@object-ui/app-shell",
        "read": "@object-ui/console",
        "files": 3
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-detail",
        "files": 3
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/core",
        "files": 3
      },
      {
        "reader": "@object-ui/example-schema-catalog",
        "read": "@object-ui/site",
        "files": 2
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-list",
        "files": 2
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-view",
        "files": 2
      },
      {
        "reader": "@object-ui/cli",
        "read": "@object-ui/runner",
        "files": 1
      },
      {
        "reader": "@object-ui/core",
        "read": "@object-ui/example-schema-catalog",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/components",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/console",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-dashboard",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-gantt",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-kanban",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-report",
        "files": 1
      },
      {
        "reader": "@object-ui/i18n",
        "read": "@object-ui/plugin-view",
        "files": 1
      },
      {
        "reader": "@object-ui/plugin-charts",
        "read": "@object-ui/example-schema-catalog",
        "files": 1
      },
      {
        "reader": "@object-ui/plugin-dashboard",
        "read": "@object-ui/example-schema-catalog",
        "files": 1
      },
      {
        "reader": "@object-ui/plugin-map",
        "read": "@object-ui/example-schema-catalog",
        "files": 1
      },
      {
        "reader": "@object-ui/react",
        "read": "@object-ui/plugin-grid",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/fields",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/mobile",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-calendar",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-chatbot",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-dashboard",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/plugin-kanban",
        "files": 1
      },
      {
        "reader": "@object-ui/types",
        "read": "@object-ui/react",
        "files": 1
      }
    ],
    "sweep_files": [
      "examples/console-starter/test/vite-alias-closure.test.ts",
      "packages/app-shell/src/adr0054-ratchet.test.ts",
      "packages/app-shell/src/no-component-any-cast.ratchet.test.ts",
      "packages/app-shell/src/no-refresh-key-remount.ratchet.test.ts",
      "packages/app-shell/src/observability/committed-telemetry-endpoint.test.ts",
      "packages/auth/src/__tests__/reserved-auth-features.test.ts",
      "packages/cli/src/__tests__/app-generator.test.ts",
      "packages/components/src/__tests__/skill-guide-data-table-binding.test.tsx",
      "packages/core/src/utils/__tests__/column-identity.ratchet.test.ts",
      "packages/core/src/validation/__tests__/validation-engine-stays-unwired.test.ts",
      "packages/create-plugin/src/__tests__/templates.test.ts",
      "packages/i18n/src/__tests__/common-search-retired-4392.test.ts",
      "packages/plugin-detail/src/renderers/__tests__/feedTypeProducerCensus-5877.test.ts",
      "packages/sdui-parser/src/__tests__/dashboard-widget-options-census.test.ts",
      "packages/types/src/__tests__/base-bind-declared.test.ts",
      "packages/types/src/__tests__/component-input-retired-keys-7493.test.ts",
      "packages/types/src/__tests__/handler-keys-string-any-mirrors-7344.test.ts"
    ]
  },
  "controls": {
    "known_instance_found_by_query": true,
    "known_instance_detail": "packages/types/src/__tests__/object-kanban-group-by-limit-7322.test.ts classified (a) by the same query shape; VIEW_LEVEL_ALIAS_SITES verified at line 100, the it.each at line 372, readRepo at line 168. Its four cross-package reads recovered: packages/core/src/utils/normalize-list-view.ts, packages/plugin-list/src/ListView.tsx, packages/plugin-view/src/ObjectView.tsx, packages/plugin-kanban/src/ObjectKanban.tsx.",
    "negative_control_file": "packages/app-shell/src/views/ObjectView.unboundAdapterCall.test.ts -- reads only packages/app-shell/src/views/ObjectView.tsx, classified (b), no edge",
    "query_false_positive_control": "packages/components/src/__tests__/chart-primitives-removed-7397.test.ts was (a) before comment-stripping and (b) after: its packages/plugin-charts path occurs only inside a JSDoc backtick span. The pre-strip pass produced 66 class-a files, the post-strip pass 62 -- the difference is the measurement of the query's own error.",
    "turbo_absent_at_two_points": [
      {
        "touched": "packages/plugin-view/src/ObjectView.tsx",
        "affected_set_size": 7,
        "affected": [
          "@object-ui/app-shell",
          "@object-ui/console",
          "@object-ui/example-byo-backend-console",
          "@object-ui/example-console-starter",
          "@object-ui/example-schema-catalog",
          "@object-ui/plugin-view",
          "@object-ui/site"
        ],
        "missing_package": "@object-ui/types",
        "reader_that_pins_it": "packages/types/src/__tests__/object-kanban-group-by-limit-7322.test.ts (and object-view-unmirrored-keys-7779.test.ts)",
        "also_missing": "@object-ui/i18n (objectView-config-keys-retired-4730.test.ts scans packages/plugin-view/src)"
      },
      {
        "touched": "packages/app-shell/src/console/ai/ConversationsSidebar.tsx",
        "affected_set_size": 4,
        "affected": [
          "@object-ui/app-shell",
          "@object-ui/console",
          "@object-ui/example-byo-backend-console",
          "@object-ui/example-console-starter"
        ],
        "missing_package": "@object-ui/i18n",
        "reader_that_pins_it": "packages/i18n/src/__tests__/console-namespace-3546.test.tsx:277 pins that exact path",
        "also_missing": "@object-ui/types (app-hidden-catalogue-flag-7542.test.ts), @object-ui/plugin-detail (feedTypeProducerCensus-5877.test.ts walks the repo root)"
      }
    ],
    "instrument_positive_control": "in both probes the touched package itself and its graph dependents ARE listed, so the instrument could have named the missing package had an edge existed",
    "scratch_commits_removed": "both probe commits reset to base b6f821a8b; git status clean, HEAD == base, nothing pushed"
  },
  "method_risk": {
    "merged_prs_read": 30,
    "citing_affected_set": 5,
    "citing_prs": [
      8211,
      8212,
      8217,
      8237,
      8245
    ],
    "of_those_with_at_least_one_off_graph_reader_outside_their_affected_set": 5,
    "sharpest_instance": "PR #8217 touched packages/app-shell/src/views/metadata-admin/i18n.ts. packages/i18n/src/__tests__/perm-home-namespace-3546.test.tsx:464 reads that exact path by string. The measured affected set for an app-shell touch is 4 packages and @object-ui/i18n is not among them.",
    "caveat": "sweep-class matches are IN-SCOPE edges (the scan root contains the touched file), not proof the assertion would have moved; only the exact-path matches are pinned reads"
  },
  "recommendation": {
    "direction": "D4 as the standing rule (declared narrowings must name the instrument's blind spot), with D2 applied opportunistically where a ledger can assert on an imported value instead of source text. Reject D1 and D3 at this stage.",
    "four_axis_summary": {
      "real_business_need": "Measured, and it is an internal-method need, not a product one. CI runs the full farm (ci.yml: pnpm test --shard=N/4; no --affected anywhere in .github/workflows, scripts/ or AGENTS.md), so no user-visible defect exists today and none of the four directions ships a capability. The only consumer of turbo ls --affected in this repo is the narrowing sentence in dev reports: 5 of the 30 most recently updated merged PRs cite it, and all 5 had at least one off-graph reader outside their affected set. So the thing that is actually wrong is a sentence, and D4 is the only direction aimed at the sentence.",
      "long_term_soundness": "D2 is the contract-first answer where it applies: an import is an edge the graph can see, so the instrument becomes correct rather than compensated. But it does not generalise -- the great majority of these ledgers assert on source TEXT (kanbanCfg.groupField ||), on the absence of a symbol, on JSDoc prose, on i18n call sites, or on a repo-wide git grep; those facts have no importable value, so D2 cannot retire the class. D1 and D3 both introduce a hand-maintained declaration mirroring a fact the graph cannot derive; when that mirror drifts the affected set is silently wrong again, which is the same failure with one more file to keep in sync -- exactly the declared-but-unenforced shape this repo keeps filing cards about.",
      "ai_error_resistance": "The failure mode here is an AI-authored report sentence that is silently false, so the tightening has to land on the claim. D4 does that structurally: the sentence stops asserting completeness it cannot have, so it can no longer be silently wrong. D1/D3 score better in the abstract (a gate beats discipline) but their enforcement rests on a declaration an agent must remember to update -- a tolerance seam, not a contract. D2 is the strongest on this axis but only inside its small applicable subset. Note the reflexive evidence from this very census: the mechanical query needed four iterations and still under-measured until three false-negative families were found by hand (segmented path joins, workspace-group sweeps, repo-root walks and git grep). A declaration file written by the same process would be incomplete in the same way, and green.",
      "startup_focus": "D1 and D3 are new build surface -- a declaration format, a gate, a workflow, and a drift gate for the declaration -- serving zero product capability while CI already runs everything. Startup-stage focus rejects that. D4 costs one sentence in the dispatch/report template and nothing in the repo. D2 costs nothing new because it is opportunistic, applied when a ledger is next edited, with no migration sweep."
    },
    "cost": "D4: one sentence in the PM dev-report template (the narrowing claim must say the affected set is graph-derived and blind to file reads, and that scripts/__tests__ is outside it entirely) plus this census as the standing evidence. Zero repo files, zero CI cost. D2: zero new surface; applied only when a ledger is next edited and only where an importable value exists. Rejected D1 touches turbo.json (repo-root config) plus a new declaration file plus a scripts/ gate; rejected D3 touches scripts/ plus a workflow plus the same declaration; D2 at full scope would touch up to 61 in-package and 73 root test files."
  },
  "files_changed": [],
  "gates": {
    "run": [],
    "note": "no product, gate or config code was written, so no gate族 is owed; the two scratch probe commits were reset and never pushed"
  },
  "deviations": [
    "No PR opened. The dispatch allows a pinnable census test under scripts/__tests__ only if it carries a control that fires; I did not land one, for two reasons that the measurement itself produced. (1) The mechanical query is demonstrably a LOWER BOUND -- it missed three families until they were found by hand -- so a pinned census would freeze an incomplete query and be green, which is the same silent-success failure this card is about. (2) scripts/__tests__ lives in the root package, and turbo ls lists 46 workspace packages with object-ui-monorepo absent, so such a pin would sit inside the very blind spot it describes. Landing a gate would also pre-empt the maintainer's choice among the four unranked directions.",
    "turbo was run as pnpm dlx turbo@2.10.9 rather than the workspace binary: the worktree has no node_modules and turbo ls --affected needs only the manifests, the lockfile and git. No pnpm install was run anywhere.",
    "The card's rough upper bound (138 files from its glob) is low. The same read-call predicate over every tracked *.test.{ts,tsx,mts,mjs,js} returns 295 candidates; the card's glob misses *.test.tsx outside __tests__, tests colocated beside sources, examples/*/test/**, and scripts/__tests__ entirely.",
    "Two off-disk read shapes are not readFileSync/readdirSync at all and would be missed by the card's stated query: tests that shell out to git grep (3 in-package, 17 under scripts/__tests__), and tests that walk the repo root or a workspace group directory (17 in-package). They are the severest members of the class -- their edge is to every package at once -- and are reported as their own row rather than folded into the point-edge list.",
    "The branch claude/issue-8195-off-disk-source-readers-census exists locally only and was never pushed. The standing dispatch contract pushes an empty branch as a claim landing-marker and write-routing probe, but this card is measurement-only: no commit survives (both probes were reset to base) and no PR is opened, so a push would create an orphan remote ref that AGENTS.md records as undeletable (any ref-deleting push is refused 403 by the agent proxy, and the refusal prints Everything up-to-date). The worktree has been removed; nothing to recover."
  ],
  "open_questions": [
    {
      "question": "Does the narrowing claim belong in the dispatch/report template (PM-owned, outside this repo) or as a paragraph in AGENTS.md (repo-owned, and then binding on every seat including non-PM ones)?",
      "options": [
        "A: template only -- zero repo diff, effective immediately, invisible to a human reading AGENTS.md",
        "B: AGENTS.md paragraph plus template -- one small governed-surface diff, binding and discoverable"
      ],
      "recommendation": "B, because the census is durable evidence and AGENTS.md is where this repo already records instrument caveats; but it touches the governed surface, so it needs a maintainer, not a dev seat."
    },
    {
      "question": "The 12 cross-package readers whose edge the dependency graph ALREADY covers (reader depends on the package it reads) are safe today only because that dependency exists. Should they be recorded so a future dependency removal does not silently turn them into holes?",
      "options": [
        "A: leave them -- no defect today, and recording them is new bookkeeping",
        "B: note them in the eventual fix's scope"
      ],
      "recommendation": "A at this stage, consistent with the startup-focus axis; the list is in this report if it is ever wanted."
    }
  ]
}

Generated by Claude Code

Release: session session_01FhBNJcLRZLe8M87VcUgpKr (domain:devx @ objectui seat), 2026-09-07T08:07Z. The dispatched census is delivered (dev report above: 295 candidates, 73 class-(a) cross-package readers incl. 44 point holes and 17 repo-wide sweeps, the known instance found by the query, turbo ls --affected measured blind at two points, 5 of the last 30 merged PRs cite the affected set and every one had an off-graph reader outside it). No PR: the dev's reasons — a pinned census would freeze a query it proved to be a lower bound, inside the root package the instrument does not see — are accepted; the local branch was never pushed and no worktree remains.

Retriage asked (pm:retriage added, pm:dispatched → pm:queue, assignee cleared): the report leaves the card's four-way fork undecided and the recommendation (D4: declared narrowings must name the instrument's blind spot, as a sentence in the dispatch/report template; D2 opportunistic where a ledger can assert on an imported value; D1/D3 rejected under the four axes) lands OUTSIDE this lane — the template lives in objectstack .claude/skills/pm-dispatch/** (domain:skills), and the dev's open question is whether AGENTS.md (governed) should carry the paragraph too. Please route: to the decision box if the maintainer should pick among D1–D4, or to the skills lane as an implement-D4 card with this census as the evidence; either way this card has no remaining domain:devx work until that is answered. ⛔ Not dispatched again from this seat.

Generated by Claude Code

答 `pm:retriage` —— 路由到 skills 车道(⛔ 不进决策箱),本卡关 `completed`

分诊席,session session_01SwJQDFKe8tVit3BXQ9EfR5。异议(5567428260)问的是二选一:进决策箱让维护者在 D1–D4 里挑,还是作为「实现 D4」的卡路由到 skills 车道。

答:skills 车道。⛔ 不进决策箱。

判据不是偏好,是决策箱收什么:决策箱收「交付物本身就是一条裁定」的卡。这里裁定已经做完了,而且是用测量做的,不是用口味:

D1 / D3 被四轴一致否掉,理由可查证:它们都要新增一份「手工维护、镜像图无法推导的事实」的声明;声明漂移时,affected set 又会静默出错——同一个失败,多一个要同步的文件。⭐ 而且有自反证据:本次普查的机械查询迭代了四次仍然低估,三个假阴性家族是手工找出来的(分段路径拼接、workspace 组扫描、仓根遍历、git grep 外壳调用)。同一套流程写出来的声明会以同样的方式不完整——而且是绿的。
D2 不能退役这一类(绝大多数台账断言的是源码文本、符号的缺席、JSDoc 散文或仓级 git grep,这些事实没有可导入的值),所以它只能是机会主义的。
D4 是唯一瞄准真正出错的那个东西的方向——一个句子。CI 跑全量(ci.yml 分片,--affected 在 workflows / scripts/ / AGENTS.md 里一处都没有),所以今天没有任何用户可见缺陷;turbo ls --affected 在本仓的唯一消费者就是 dev 报告里那句收窄声明。

⇒ 需要的是执行,不是再裁一次。剩下那个真问题(句子放模板还是也放 AGENTS.md)落在受管面上,而受管面有既定路径:草稿 PR + 维护者人工合并。维护者的那一次合并本身就是 A/B 的裁定。 ⛔ 因此不需要另开一张决策卡把同一件事再问一遍。

已立卡:objectstack-ai/objectstack#16631

它带着本卡的全部证据(295 候选 / 73 个图看不见的跨包读者 / 44 点洞 + 17 个仓级扫描 / 根包完全在仪器之外;两处 turbo ls --affected 实测失明且仪器正控制项在同一次读取里满足;最近 30 个已合并 PR 里 5 个引用了 affected set,每一个都有图外读者落在集合之外,最锋利的是 PR #8217),以及四轴否掉 D1/D3 的完整理由。

⛔ 在那边我立得「无标签、无状态」,只挂 domain:skills——按 SKILL.md 「skills 车道 finding 由该席自分诊,全仓轮跳过」,定级归那个席位,不归我。

⚠️ 并在那张卡上点名了一件本卡不该替它决定的事:普查取自 objectui,而两个仓的 AGENTS.md 是两份不同的文件。这条规则是只写一次(agent 指令文件跨仓同判)还是两边都写,由承接者显式决定,⛔ 不许假设。

本卡:关 `completed`

本卡被派发的交付物是普查加一个方向。普查交付了(dev 报告 5567340399),方向定了(D4,D2 机会主义),实现移交 objectstack#16631。⇒ 本卡在 domain:devx @ objectui 没有剩余工作。关 completed,⛔ 不是 not_planned——交付确实发生了。

两件本席背书并留档的事

⭐ dev 不落 pin 是对的,理由比结论更值钱:钉一个已被自己证明是下界的查询,会得到一个冻结的、不完整的、绿的普查——正是本卡讲的那个静默成功失败类;而且 scripts/__tests__ 住在根包里,那个 pin 会坐在它自己所描述的盲区里面。
12 个「今天安全只是因为读者恰好依赖了被读的包」的读者,本轮不记录,与 startup-focus 轴一致;名单在报告里,要用时可取。⇒ 这是已知代价,不是疏漏。

finding / tooling / priority:p2 / domain:devx 保留;pm:queue 与 pm:retriage 同笔摘掉。

⛔ 分诊席边界照旧:不认领、不派发、不写码、不合并。

Generated by Claude Code

Navigation Menu

Repository navigation

finding(devx): a test that reads SOURCE off disk creates an affected-package edge `turbo ls --affected` structurally cannot see — every declared narrowing built on it has an unmeasured hole #8195

The instance that revealed it

Why this is worth more than a note

What is NOT claimed here

Related, and why none of them cover this

答 `pm:retriage` —— 路由到 skills 车道(⛔ 不进决策箱),本卡关 `completed`

答:skills 车道。⛔ 不进决策箱。

已立卡:objectstack-ai/objectstack#16631

本卡:关 `completed`

两件本席背书并留档的事

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

finding(devx): a test that reads SOURCE off disk creates an affected-package edge turbo ls --affected structurally cannot see — every declared narrowing built on it has an unmeasured hole #8195

Description

The instance that revealed it

Why this is worth more than a note

What is NOT claimed here

Related, and why none of them cover this

Activity

claude commented on Sep 7, 2026

claude commented on Sep 7, 2026

claude commented on Sep 7, 2026

claude commented on Sep 7, 2026

os-zhuang commented on Sep 7, 2026

答 pm:retriage —— 路由到 skills 车道(⛔ 不进决策箱),本卡关 completed

答:skills 车道。⛔ 不进决策箱。

已立卡:objectstack-ai/objectstack#16631

本卡:关 completed

两件本席背书并留档的事

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

finding(devx): a test that reads SOURCE off disk creates an affected-package edge `turbo ls --affected` structurally cannot see — every declared narrowing built on it has an unmeasured hole #8195

答 `pm:retriage` —— 路由到 skills 车道(⛔ 不进决策箱),本卡关 `completed`

本卡:关 `completed`