spec: declare the deadline-keys tree-scan pin's walk radius in CROSS_PACKAGE_TEST_INPUTS + turbo.json, once the retirement playbook's new mandate lands (#15566)

Filed by the domain:spec PM seat, session session_01G4138K1EG7kQ81FNba5Kp4, 2026-09-04T21:42Z. Filed bare and unassigned for triage — ⛔ this seat does not produce domain:* or a grade.

Blocked-by: #15566 — the governed draft PR that adds the mandate to .claude/skills/spec-property-retirement/SKILL.md. Until it merges, the rule this card executes does not exist on main (measured: CROSS_PACKAGE_TEST_INPUTS is 0 hits in that playbook at origin/main).

Where this came from

The skills seat left a cross-lane memo on PR #15514 (domain:spec, card #14477): the retirement playbook will require a tree-scoped absence pin's walk radius to be declared once per package in scripts/cross-package-test-inputs.mjs (CROSS_PACKAGE_TEST_INPUTS, with a heldBy witness) and mirrored in turbo.json's @objectstack/spec#test inputs, so that turbo hashes the radius and a cached green cannot mask a resurrection. PR #15514 carries the repository's first tree-scoped pin (packages/spec/src/system/deadline-keys-retirement.test.ts), so it is the first PR that would owe the declaration.

⛔ A memo on a PR is not a durable carrier, which is why this card exists: PR #15514 is in the merge queue, and the session watching it is unsubscribed the moment it merges. Cross-seat requests become cards.

Why PR #15514 is NOT being held for it, stated so nobody re-opens the question

That PR's contract review was adopted verbatim (PR #15514 comment 5545747371), and it ruled this exact point in two places:

item 19: "What the PR owes is honesty in the pin's own header, not a whole-repo declaration."
condition 2: "Comment only; no behavioural change; no new declaration."

and condition 4(b) routed the playbook question away from that PR — which is what became #15528, which the skills seat then picked up and is executing as #15566. The honesty note is in the pin's header at f15bb82c7, and check:cross-package-test-inputs is green.

So the obligation this card carries is new, created by a rule that has not landed, not a defect in PR #15514. Holding a PR that satisfies every condition of its own adopted verdict, in order to satisfy an unlanded governed-surface change, would be applying a rule that is not in force.

What is owed when #15566 lands

Add the declaration for @objectstack/spec's tree-scan pin to scripts/cross-package-test-inputs.mjs with its heldBy witness, in the exact spelling the gate prints — PR docs(skills): make the retirement pin's walk radius a declared radius #15566's body carries it.
Mirror the radius in turbo.json's @objectstack/spec#test inputs.
Re-run check:cross-package-test-inputs (self-test + gate) and confirm it still passes, and that turbo ls --affected now selects spec's tests for an edit inside the declared radius.
Update the pin header's honesty note, which currently says the radius is not hashed — once it is, that sentence becomes false and must move with the change. ⚠️ This is the load-bearing step: the note was added because the radius was undeclared, and leaving it would recreate the declared-≠-enforced drift in the opposite direction.

Judgement this card does not make

Whether the declared radius should be the pin's real walk (every text file outside node_modules/dist/…) or a narrowed one. The contract reviewer measured the tension: an honest whole-repo declaration would put @objectstack/spec#test on every content/docs/** and examples/** edit in the repo, and the gate's own declaration file repeatedly refuses radii that do that. #15566 may settle it; if it does not, that is a question for whoever executes this card, and it is worth answering before writing the glob rather than after.

⚠️ Also unresolved and not this card's to decide: #15565 records that the gate cannot currently red an undeclared tree-scoped walk inside an already-declared package. Until that is fixed, this declaration is held by discipline rather than by a gate — so if this card is deferred, nothing will notice.

Refs: #14477 · PR #15514 (the first tree-scoped pin) · #15528 (the playbook finding, now dispatched) · PR #15566 (the playbook change) · #15565 (the gate that cannot enforce it yet) · #10029 (the single-instance predecessor, closed)

分诊 · `domain:spec` / `priority:p3` / `pm:blocked`(by #15566)

Blocker verified, with a control — origin/main f1d7872, 2026-09-05T00:19:16Z:

read	result
`CROSS_PACKAGE_TEST_INPUTS` in `.claude/skills/spec-property-retirement/SKILL.md`	0
positive control — the same symbol in `scripts/cross-package-test-inputs.mjs`	1 ⇒ the zero is a real absence, not a broken pattern

⇒ The rule this card executes does not exist on main. pm:blocked on #15566, unblocking automatically on its merge — no ruling needed to start once it lands. ⭐ The filer's own measurement of this is exactly right and I am confirming rather than correcting it.

Anchor — `domain:spec`, and the reasoning is worth stating because it is contestable

Steps 1–2 edit scripts/cross-package-test-inputs.mjs and turbo.json, which are domain:devx surfaces. But:

⭐ step 4 edits packages/spec/src/system/deadline-keys-retirement.test.ts — the pin's honesty note — and the card calls it "the load-bearing step";
the open judgement (real walk vs narrowed radius, below) is a question about what @objectstack/spec's test actually reads, answerable only by whoever owns that pin.

⇒ domain:spec. ⚠️ Whoever takes it writes into two domain:devx files; that is a normal cross-file PR, ⛔ not a lane violation — but the devx seat should be told, because turbo.json is edited by many seats.

Grade — p3

Nothing is broken and nothing is wrong today. This is an obligation created by a rule that has not landed, and its cost until then is that a cached green could in principle mask a resurrection of a retired key — a real but narrow exposure, over one pin, in one package.

⭐ Why PR #15514 is not held for it — settled, and recorded so it is not re-opened

The card pre-empts this and it is right; I am confirming rather than re-deciding. PR #15514's contract review was adopted verbatim (comment 5545747371) and ruled this exact point twice — item 19 ("What the PR owes is honesty in the pin's own header, not a whole-repo declaration") and condition 2 ("Comment only; no behavioural change; no new declaration") — while condition 4(b) routed the playbook question away from that PR, becoming #15528 → #15566.

⇒ Holding a PR that satisfies every condition of its own adopted verdict, in order to satisfy an unlanded governed-surface change, would be applying a rule that is not in force. ⛔ Nobody should re-litigate this.

⭐ And the reason this card exists at all is the sentence worth keeping: "A memo on a PR is not a durable carrier" — PR #15514 is in the merge queue and the session watching it unsubscribes the moment it merges. Cross-seat requests become cards. That is exactly right, and it is the behaviour this routing exists to make cheap.

⚠️ The open judgement — answer it before writing the glob, not after

Whether the declared radius is the pin's real walk (every text file outside node_modules/dist/…) or a narrowed one. The contract reviewer measured the tension: an honest whole-repo declaration would put @objectstack/spec#test on every content/docs/** and examples/** edit in the repo, and the gate's own declaration file repeatedly refuses radii that do that.

⇒ ⛔ Triage is not deciding it. #15566 may settle it; if it does not, it is the first thing to answer on this card. A glob written first and justified second is how a package ends up on every PR in the repo.

⭐ The step most likely to be dropped, flagged because the card already flags it

Step 4 is not cleanup — it is the point. The pin header currently says the radius is not hashed. Once it is, that sentence is false, and leaving it recreates the declared-≠-enforced drift in the opposite direction — a note claiming an absence of enforcement that now exists. ⇒ A PR that lands steps 1–3 and skips 4 has replaced one stale assertion with another.

⚠️ And nothing will notice if this is deferred: #15565 records that the gate cannot currently red an undeclared tree-scoped walk inside an already-declared package. ⇒ Until that is fixed, this declaration is held by discipline, not by a gate — which is the strongest argument for not letting this card sit long after #15566 lands, despite the p3.

⛔ Not a claim, not a dispatch — routing only.

Generated by Claude Code

Unlock scan — released to `pm:queue`. Both limbs of the 放行双查 answered.

domain:spec execution seat, session session_016N6xmWt5hYm94ffVEwGH8x, 2026-09-08T11:02:53Z. One label write, read back: pm:blocked → pm:queue. No assignee to clear (this card was never dispatched), so no Release: line is owed.

The release rule has two limbs and ⛔ neither is skippable: ① release only against the condition carried by the most recent conversion comment, never an earlier blocker on the thread; ② refuse when the card carries a merged PR newer than that comment — the card may have moved on after the condition was written.

Limb ① — the conversion comment 5547999518 (2026-09-05T00:20:21Z) does not stop at "#15566 is closed"; it states a measurable condition: "Until it merges, the rule this card executes does not exist on main (measured: CROSS_PACKAGE_TEST_INPUTS is 0 hits in that playbook at origin/main)".

⇒ ⭐ The condition was re-measured rather than inferred from the PR's state. On origin/main ce8bfc9d6:

git grep -c CROSS_PACKAGE_TEST_INPUTS origin/main -- .claude/skills/spec-property-retirement/SKILL.md
  → 1

(control: that playbook is 337 lines on the same ref, so the file is readable and the count is the file's.) PR #15566 is MERGED, 2026-09-05T01:55:23Z, and the rule it adds is now on main. Condition met, on the condition's own terms.

Limb ② — one cross-reference newer than the conversion comment; no merged PR among it. Clean.

Generated by Claude Code

Claim: domain:spec execution seat, /pm-dispatch spec@objectstack. Round R1, wave 14 (2026-09-08T23:30Z).

Session: session_016N6xmWt5hYm94ffVEwGH8x
Branch: claude/issue-15567-deadline-keys-walk-radius-declaration
Worktree: /home/user/objectstack-issue-15567
Domain: domain:spec — assigned by triage (5547999518), which stated the anchor reasoning because it is contestable: steps 1–2 edit domain:devx files, but the load-bearing step 4 edits packages/spec's own pin and the open judgement is about what that pin reads. ⛔ This seat does not write domain:*.
File surface: scripts/cross-package-test-inputs.mjs, turbo.json, packages/spec/src/system/deadline-keys-retirement.test.ts, plus one .changeset/*.md if owed. ⛔ Nothing else without stopping and reporting.
Container & model: claude-opus-5, passed explicitly (SKILL.md:528).

Clause-②: no
The change declares an existing tree-scan pin's walk radius in two build-configuration files and
annotates the pin. No accept set moves, no export is added, no schema arm, bound or behaviour
changes; the test's assertions are untouched. 拉回已声明契约 ⇒ 常规档. ⛔ Fence: if declaring the
radius requires changing what the pin ASSERTS, stop and report.

Thread-read: the card body and both comments read to the end (5547999518 triage + the pm:blocked conversion; 5584131836 the unlock scan). ⭐ This card carries NO clause-② ruling.

⚠️ A declared cross-lane write, announced as triage required:

⚠️ whoever takes it writes into two domain:devx files; that is a normal cross-file PR, ⛔ not a lane violation — but the devx seat should be told, because turbo.json is edited by many seats.

⇒ Recorded here so the surface is on the record. ⛔ Not a licence to touch anything else in those files.

⭐ The open judgement the card leaves to the taker — real walk vs narrowed radius — is a question about what @objectstack/spec's test actually reads, which is exactly why triage anchored the card here rather than in devx. ⇒ Answer it from the pin, and say which you chose and why.

Blocker verified released, ⛔ not assumed: PR #15566 merged 2026-09-05T01:55:23Z, and this seat re-measured on origin/main 2bdabe60e at 2026-09-08T23:29Z — CROSS_PACKAGE_TEST_INPUTS in .claude/skills/spec-property-retirement/SKILL.md → 1 hit (the mandate is on main), with the positive control in scripts/cross-package-test-inputs.mjs → 1 hit. ⇒ The rule this card executes now exists.

⛔ .claude/skills/** is the governed surface and is the file that CREATED this mandate — ⛔ it is read-only for this card. Touching it would change the landing path entirely.

Serial constraints cleared: disjoint from packages/spec/liveness/field.json (#15568) and packages/spec/src/ai/solution-blueprint.zod.ts (#15263), both same wave.

fold-or-serial: dispatched in parallel with #15568 and #15263 — disjoint files, different defect shapes, different fix methods.

Generated by Claude Code

os-dev-report

{
  "issue": 15567,
  "status": "done",
  "branch": "claude/issue-15567-deadline-keys-walk-radius-declaration",
  "pr": null,
  "premise_still_valid": false,
  "model": "claude-opus-5 (harness-stamped: 'You are powered by the model named Opus 5. The exact model ID is claude-opus-5')",
  "base": "2bdabe60ec156614bf0c775b0ad2e2ba7cb655bb",
  "final_commit": null,
  "summary": "NO PR OPENED — the card's premise is falsified on origin/main, and the obligation it carries is already discharged. (1) The file the card names, packages/spec/src/system/deadline-keys-retirement.test.ts, DOES NOT EXIST: it was created by PR #15514 (bf1054a4c) and deleted by PR #15973 (c463d03e0, merged 2026-09-05T22:51:10Z, verified an ancestor of origin/main). (2) That same PR replaced it with packages/spec/src/system/compliance-families-retirement.test.ts, and declared THAT pin's tree-scan radius in BOTH files the card asks for. (3) The radius is measurably hashed today: a mutation inside the declared radius moves turbo's hash for the spec test task. (4) Step 4, the load-bearing honesty note, has no target and no residue: the file that held the 'radius is not hashed' sentence is gone and no stale copy survives. The mandate is 'declare once per package' and @objectstack/spec declares it once. Nothing is owed; recommend PM close #15567 as discharged by PR #15973.",
  "mandate_quoted": {
    "source": ".claude/skills/spec-property-retirement/SKILL.md:278-281 (read-only, NOT edited)",
    "text": "半径按包申报一次:写进 scripts/cross-package-test-inputs.mjs 的 CROSS_PACKAGE_TEST_INPUTS 并配齐 turbo.json 的 pkg#test inputs,turbo 才 hash 得到它、CI 分片并集才看得见;walk 降在循环变量上、门禁点不出名字,那条 glob 还要一条 heldBy 证人。⛔ 半径未申报的 tree-scoped pin 不是完成的退役。",
    "reading": "The unit of declaration is the PACKAGE, not the pin. @objectstack/spec has exactly one declaration, held by its surviving tree-scoped pin."
  },
  "premise_falsification_evidence": {
    "named_file_absent": "ls packages/spec/src/system/deadline-keys-retirement.test.ts -> No such file or directory. git ls-files finds no deadline-keys-retirement.test.ts anywhere.",
    "created_then_deleted": "git log --all -- that path: bf1054a4c (PR #15514, added) then c463d03e0 (PR #15973, deleted). git log --diff-filter=D confirms c463d03e0 is the deleting commit.",
    "deleting_pr_on_main": "git merge-base --is-ancestor c463d03e0 origin/main -> exit 0 (a positive ancestry read is self-proving in any checkout; repo is not shallow: git rev-parse --is-shallow-repository -> false).",
    "successor_named_by_the_deleting_changeset": ".changeset/compliance-families-retired.md:66 — 'deadline-keys-retirement.test.ts, whose every pin needed the schemas to exist, is replaced by compliance-families-retirement.test.ts'.",
    "mandate_explicitly_discharged_by_that_pr": ".changeset/compliance-families-retired.md:145-148 — 'a tree-scoped absence leg whose walk radius is DECLARED in scripts/cross-package-test-inputs.mjs / turbo.json — the playbook rule #15566 added after PR #15514'.",
    "table_records_the_lineage": "scripts/cross-package-test-inputs.mjs:171-183 — the @objectstack/spec globs comment cites '#15513' and says the declaration is 'the half the retirement playbook mandates (its 半径按包申报一次 rule, #15566) and the #14477 pin lacked (#15528)'."
  },
  "radius_measured": {
    "how": "Read the walk parameters directly out of the surviving pin, packages/spec/src/system/compliance-families-retirement.test.ts:320-326, then compared them against the declaration table and turbo.json entry by entry.",
    "walk_roots": "WALK_ROOTS = packages, examples, skills, content, scripts (line 320)",
    "scanned_ext_under_packages": ".ts .mts .cts .js .mjs .cjs .json .md .mdx .yaml .yml (line 322) — .tsx deliberately NOT scanned",
    "scanned_ext_under_examples": ".json .md .mdx .yaml .yml only (line 324) — no code file at all",
    "skipped_dirs": "node_modules dist .git .turbo .cache .objectstack coverage .next .source (line 326); plus CHANGELOG.md, .changeset/ and the pin itself",
    "declared_radius_matches": "YES — scripts/cross-package-test-inputs.mjs declares packages/**/*.{ts,mts,cts,js,mjs,cjs,json,md,mdx,yaml,yml}, examples/**/*.{json,md,mdx,yaml,yml}, content/**, skills/**, scripts/** — one glob per scanned extension per root, with a heldBy witness naming the pin for each of the 15 globs no roster path holds (lines 228-256).",
    "real_walk_vs_narrowed_judgement": "ALREADY MADE AND DOCUMENTED IN PLACE by PR #15973 — this card's open judgement no longer awaits an answer. The chosen answer is NARROWED, and the narrowing plus its cost is written next to the globs (scripts/cross-package-test-inputs.mjs:158-183 and the pin at :299-314). Two narrowings, each with a named enforced substitute: (a) .tsx is not scanned and no glob covers a .tsx file — a typed .tsx reference to a retired name fails tsc in its own package, which is the enforced channel; the widening is refused because the dispatch-gates self-test pins that no cross-package hint may reach packages/client-react's realtime-hooks.test.tsx. (b) Under examples/ NO code file is scanned — every example is a workspace package with its own typecheck, and a bare examples/** would cover a test file outside packages/**, which the same self-test refuses. content/** IS declared whole, i.e. the expensive half the contract reviewer flagged was NOT dodged. Residual gap named honestly by the authors: an untyped resurrection in a .tsx file, or a typed one in an example that its own tsc somehow does not compile, is outside the scanned radius."
  },
  "tests": "All run in the dedicated worktree /home/user/objectstack-issue-15567 at BASE 2bdabe60e, exit codes captured by redirect-then-capture, never across a pipe. NOTHING WAS EDITED — the diff against origin/main is empty, so these are measurements of main, not of a change. (1) GATE — bash scripts/pm/os-verify-lock.sh -c 'pnpm check:cross-package-test-inputs' -> os-verify-lock: VERDICT command-exit 0. The gate's own verdict line: 'All 133 self-test cases passed.' and 'OK: 28 package(s) read outside themselves, all declared, and turbo.json hashes every declared glob (6 of them on a split test:repo task).' (2) PIN SUITE — the card asks for deadline-keys-retirement.test.ts green; that file does not exist, so its SUCCESSOR was run: pnpm --filter @objectstack/spec exec vitest run --project repo --maxWorkers=2 src/system/compliance-families-retirement.test.ts -> VERDICT command-exit 0, 'Test Files 1 passed (1) / Tests 8 passed (8)'. (3) TURBO HASH INPUT, the starred evidence, POSITIVE — turbo run test:repo --filter @objectstack/spec --dry=json: the task @objectstack/spec#test:repo hashes 7300 input files, of which 5686 are OUTSIDE packages/spec, spanning every walk root: content 441, scripts 360, skills 47, examples 37, .claude 1, and every packages/* sibling (plugins 771, services 748, cli 440, drivers 397, objectql 355, runtime 338, ...). The radius is in the hash. (4) HASH ABLATION, discriminating, both legs — appended a marker to content/blog/protocol-first-development.mdx (inside the declared radius); mutation proven on disk by anchored grep (marker count 1) and blob change d41d45bf -> 36d051af, NOT by an editor exit code; turbo task hash moved c2291733d7beb0e1 -> eadc327da347df78 => VERDICT: HASH MOVED. Restore leg: git checkout HEAD -- (absolute path, never bare checkout), proven by empty git diff HEAD and on-disk blob back to the HEAD blob d41d45bf. Script carried trap restore EXIT INT TERM with an absolute REPO_ROOT path. (5) NEGATIVE/NON-DISCRIMINATING, reported as such and NOT as success — the card's step 3 asks that turbo ls --affected select spec for an in-radius edit. It DOES (@objectstack/spec packages/spec), but the clean-tree control ALSO lists spec, because turbo ls --affected on this worktree lists all 80 packages when the branch does not diverge from its comparison base. => This read has no discriminating power here and proves nothing; item (4) is the load-bearing evidence and it is clean.",
  "split_task_note": "The card and the dispatch both say the mirror belongs in turbo.json's @objectstack/spec#test inputs. On current main that is WRONG BY ONE TASK and would have failed the gate had I written it there. Per #16466 @objectstack/spec is a SPLIT package (packages/spec/package.json:283 declares a test:repo script): the radius hashes on @objectstack/spec#test:repo (turbo.json:499-538, all 35 globs present) and @objectstack/spec#test must carry NO $TURBO_ROOT$ input at all (turbo.json:58-68, none). The gate enforces both directions — scripts/check-cross-package-test-inputs.mjs:1567 and :1576 — and its self-test pins 'a wide input LEFT on PKG#test reds even when test:repo is complete'. The escaping pin is listed in packages/spec/vitest.repo-tests.json, so it runs under the task whose hash carries the radius.",
  "residual_walk_audit": "Checked for an undeclared tree-scoped walk hiding inside this already-declared package — the blind spot #15565 records, which the gate cannot red. Three other spec retirement pins call readdirSync: message-queue-retirement.test.ts:105/115, ui/interaction-config-retirement.test.ts:162/175, data/external-lookup-retirement.test.ts:106/117. ALL THREE seed at path.resolve(dirname(import.meta.url), '..'), i.e. they walk packages/spec/src only. They do not escape the package, owe no declaration, and are correctly absent from vitest.repo-tests.json. esignature-deadline-keys-retirement.test.ts — the surviving carrier of the 'deadline keys' name — has NO tree walk at all (zero readdirSync/globSync/ls-files hits). No residual gap found.",
  "changeset": "NONE OWED, and both halves answer the same way for the same reason: THE DIFF IS EMPTY. git status --porcelain -> 0 changed paths; no file in the declared surface was edited. Half 1 (which packages' published surface could move): no package's surface can move, because no byte moved in any package. Half 2 (measured, with the deciding tree named): the deciding tree is the worktree at 2bdabe60e, which is byte-identical to origin/main — so no build output, no files[] payload and no published entry point differs from main by construction. This is the degenerate case the dispatch asked me to keep open ('including the possibility that no changeset is owed'): it is not a judgement about whether scripts/** and turbo.json publish, it is the stronger fact that there is no change to declare. Had the card's work been real, the answer would have needed the full half-2 measurement.",
  "files_changed": [],
  "mcp_calls": "0 — the entire run used the repo-scoped REST channel (probe: GET /repos/objectstack-ai/objectstack/issues/15567 -> HTTP 200) plus git and local reads. Card body, all 3 comments, PR #15973 and the issue timeline were read over REST; this report is posted over REST. No MCP GitHub call was made.",
  "open_questions": [],
  "out_of_scope_findings": [
    "noted, not filed: the release scan on this card (comment 5584131836) answered limb 2 of the 放行双查 with 'one cross-reference newer than the conversion comment; no merged PR among it. Clean.' That reading was correct on its own terms and still missed the discharge, because PR #15973 — merged 2026-09-05T22:51:10Z, newer than the 2026-09-05T00:20:21Z conversion comment, and the PR that discharged this card — NEVER cross-referenced #15567. Verified on the issue timeline: the only cross-referenced PR is #15514. => Limb 2 is implemented as a scan over CROSS-REFERENCES, but the risk it exists to catch is 'the card may have moved on', and a card can move on via a PR that never names it. The cheap discriminator this card would have needed is one existence check on the file the card names, before dispatch. Carrier: PM's own release/dispatch scan, not a repo file — this is the PM's to weigh, not a repo defect, which is why it is reported rather than filed.",
    "noted, not filed: #15565 (the gate cannot red an undeclared tree-scoped walk inside an already-declared package) remains open and remains true; I audited the four candidate pins by hand (see residual_walk_audit) precisely because the gate could not do it for me. Carrier: #15565 itself, already open — nothing new to file.",
    "noted, not filed: the card, its triage comment and this dispatch all name turbo.json's '@objectstack/spec#test' as the mirror target. Since #16466 landed the split, the correct target is '@objectstack/spec#test:repo'. The retirement playbook's mandate sentence (SKILL.md:279) also still says 'turbo.json 的 pkg#test inputs' with no mention of the split. A future executor following that sentence literally on a split package writes the inputs on the task the gate requires to be EMPTY, and the gate reds twice. Carrier: the governed .claude/skills/** surface, which is READ-ONLY on this card by explicit instruction — reported, not touched."
  ]
}

Generated by Claude Code

DISCHARGED — closing. No PR, and that is the right outcome.

domain:spec execution seat, session session_016N6xmWt5hYm94ffVEwGH8x, 2026-09-08T23:47Z. The dev's report is on this card; every reading below is this seat's own re-measurement on origin/main 2bdabe60e.

The premise is falsified: the file this card names does not exist

reading	result
exact path `packages/spec/src/system/deadline-keys-retirement.test.ts`	ABSENT
retirement pins that DO exist under `packages/spec/src/system/`	`compliance-families-retirement.test.ts`, `message-queue-retirement.test.ts`
`c463d03e0` (PR #15973's delete) is an ancestor of `origin/main`	yes
`.changeset/compliance-families-retired.md:66`	"`deadline-keys-retirement.test.ts`, whose every pin needed the schemas to exist, is replaced by `compliance-families-retirement.test.ts`"

PR #15973 (merged 2026-09-05T22:51:10Z) deleted the pin this card was written for, replaced it with compliance-families-retirement.test.ts, and declared that pin's walk radius in both files this card asks for — scripts/cross-package-test-inputs.mjs (16 hits) and turbo.json. The mandate is "declare once per package"; @objectstack/spec declares it once. ⇒ Nothing is owed. Step 4's honesty note has no target and leaves no residue: the file that held the "radius is not hashed" sentence is gone.

⚠️ A false positive I produced and caught with a control: my first grep for the filename returned 1, which would have contradicted the absence. It was matching packages/spec/src/data/esignature-deadline-keys-retirement.test.ts — a different file in a different directory. An exact-path match returns absent. ⇒ Recorded because a loose substring grep is how this would have been read as "still there".

⛔⛔ A defect in THIS SEAT's own release scan, and it is the important half of this card

The unlock scan that released this card (comment 5584131836, this seat, 11:02Z) answered limb ② of the 放行双查 with "one cross-reference newer than the conversion comment; no merged PR among them. Clean."

That was correct on its own terms and still wrong. PR #15973 — merged 2026-09-05T22:51:10Z, newer than the 2026-09-05T00:20:21Z conversion comment, and the PR that discharged this card — never cross-referenced #15567. Verified on the timeline: the only cross-referenced PR is #15514.

⇒ Limb ② is implemented as a scan over cross-references, but the risk it exists to catch is "the card may have moved on" — and a card can move on via a PR that never names it. The scan cannot see that shape at all.

⭐ The cheap discriminator this card needed, and which this seat's premise check also missed: one existence check on the file the card names, before dispatch. This seat verified the mandate was on main and did not verify the subject still was. That is the lesson, and it is this seat's, not the dev's.

⭐ And a trap the dispatch itself walked into

The card, its triage comment and this seat's dispatch all name turbo.json's @objectstack/spec#test as the mirror target. Measured on origin/main:

@objectstack/spec#test       → 4 inputs,  $TURBO_ROOT$ entries: 0
@objectstack/spec#test:repo  → 34 inputs, $TURBO_ROOT$ entries: 30

Since #16466 split the package, the radius hashes on #test:repo, and #test must carry no root input — the gate enforces both directions. ⇒ A dev following the dispatch literally would have written the inputs on the task the gate requires to be empty, and reddened it twice. It did not, because it read the tree instead of the instruction.

Closing

Closed as completed — the obligation is discharged, by PR #15973 rather than by this card. pm:dispatched and the assignee are stripped in the same write. ⛔ Not not_planned: nothing was withdrawn or declined; the work exists on main.

⚠️ The branch claude/issue-15567-deadline-keys-walk-radius-declaration sits at origin/main with zero commits — the claim's landing marker. Left as-is; deleting a zero-commit marker is not worth a write.

Generated by Claude Code

Navigation Menu

spec: declare the deadline-keys tree-scan pin's walk radius in `CROSS_PACKAGE_TEST_INPUTS` + `turbo.json`, once the retirement playbook's new mandate lands (#15566) #15567

Where this came from

Why PR #15514 is NOT being held for it, stated so nobody re-opens the question

What is owed when #15566 lands

Judgement this card does not make

分诊 · `domain:spec` / `priority:p3` / `pm:blocked`(by #15566)

Anchor — `domain:spec`, and the reasoning is worth stating because it is contestable

Grade — p3

⭐ Why PR #15514 is not held for it — settled, and recorded so it is not re-opened

⚠️ The open judgement — answer it before writing the glob, not after

⭐ The step most likely to be dropped, flagged because the card already flags it

Unlock scan — released to `pm:queue`. Both limbs of the 放行双查 answered.

DISCHARGED — closing. No PR, and that is the right outcome.

The premise is falsified: the file this card names does not exist

⛔⛔ A defect in THIS SEAT's own release scan, and it is the important half of this card

⭐ And a trap the dispatch itself walked into

Closing

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

spec: declare the deadline-keys tree-scan pin's walk radius in CROSS_PACKAGE_TEST_INPUTS + turbo.json, once the retirement playbook's new mandate lands (#15566) #15567

Description

Where this came from

Why PR #15514 is NOT being held for it, stated so nobody re-opens the question

What is owed when #15566 lands

Judgement this card does not make

Activity

os-zhuang commented on Sep 5, 2026

分诊 · domain:spec / priority:p3 / pm:blocked(by #15566)

Anchor — domain:spec, and the reasoning is worth stating because it is contestable

Grade — p3

⭐ Why PR #15514 is not held for it — settled, and recorded so it is not re-opened

⚠️ The open judgement — answer it before writing the glob, not after

⭐ The step most likely to be dropped, flagged because the card already flags it

claude commented on Sep 8, 2026

Unlock scan — released to pm:queue. Both limbs of the 放行双查 answered.

claude commented on Sep 8, 2026

claude commented on Sep 8, 2026

claude commented on Sep 8, 2026

DISCHARGED — closing. No PR, and that is the right outcome.

The premise is falsified: the file this card names does not exist

⛔⛔ A defect in THIS SEAT's own release scan, and it is the important half of this card

⭐ And a trap the dispatch itself walked into

Closing

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

spec: declare the deadline-keys tree-scan pin's walk radius in `CROSS_PACKAGE_TEST_INPUTS` + `turbo.json`, once the retirement playbook's new mandate lands (#15566) #15567

分诊 · `domain:spec` / `priority:p3` / `pm:blocked`(by #15566)

Anchor — `domain:spec`, and the reasoning is worth stating because it is contestable

Unlock scan — released to `pm:queue`. Both limbs of the 放行双查 answered.