#1327 fixed the header but left its twin: case_self_claim's guard-4 comment still says the sharing rule "draws with is_closed == false", which it has not since #1145

Filed by the #1144 piece-2 dev (claude/issue-1144-claim-case-action), which had to re-derive the triage predicate to write the Claim button's visible expression and hit this on the way. The file is outside that card's declared surface, so it is reported rather than edited.

The residue

#1327 is closed and its acceptance criterion is only half met.

That card's scope was: "The header no longer asserts the two layers are the same rule, and the file's other comments do not contradict the new sentence." The header half landed — git grep 'two layers' src/objects/_case-assignment.ts returns nothing today. The second half did not.

src/objects/_case-assignment.ts, createCaseSelfClaim's guard 4 (line 465–467 at 1486ec0):

      // 4. …and only one that is not CLOSED — the same line the sharing rule
      //    draws with `is_closed == false`. `status` is read first because it
      //    is a string on every driver; `is_closed` is accepted as `true` or
      //    `1` because SQLite hands booleans back as integers.
      if (previous.status === 'closed') return;

The sharing rule has not drawn that line since #1145 / PR #1323. Measured on main at 1486ec0, src/sharing/case.sharing.ts:

  condition: P`record.owner_id == null && record.status != "resolved" && record.status != "closed"`,

So the comment makes the same false claim #1327 was filed for — "these two are the same line" — in the same file, about the same two layers, one screenful below where the fix landed. It is arguably the more misleading of the two, because it sits directly above the guard a reader is checking and names a predicate (is_closed == false) that appears nowhere in the rule any more.

⚠️ The code is right. ⛔ Do not "fix" the predicate

Unchanged from #1327, restated so this card cannot be misread as licensing what that one forbade:

guard 4 excluding only closed is correct and deliberate — reopening a resolved case is picking the work up, so the admin who does it should become the owner;
the sharing grant excluding resolved too is correct and deliberate — a resolved unowned case is history, not backlog.

⇒ The asymmetry is the feature. This is a comment correction and nothing else. ⛔ Do not touch either predicate, ⛔ do not touch test/live-work-predicate-parity.test.ts.

Scope

Rewrite the guard-4 comment so it states the same asymmetry #1327's header fix now states, rather than asserting sameness. The rest of that comment (why status is read before is_closed, why 1 is accepted for SQLite) is accurate and should survive.

While in the file, git grep -n 'is_closed' src/objects/_case-assignment.ts returns seven hits — the other five (lines 93, 97, 211, 353, 470) were read and are accurate: they explain why the load-balancing counts use $nin instead of the flag, which is the opposite claim and still true.

Acceptance

No comment in src/objects/_case-assignment.ts describes the triage sharing grant as is_closed == false. pnpm verify green.

Why this is not a duplicate

#1327 is closed. A targeted search over this repo's issues returned 44 results, including #1327, #1328, #1145 and #1435, and none of the open cards names this comment. #1435 is a different class (the readonly blanket across six object files).

_case-assignment.ts still says the sharing predicate and the claim guard are "the SAME rule stated at two layers" — after PR #1323 they exclude different status sets, and the code is right while the prose is not #1327 — closed; this is its unfinished second half
A resolved unowned case sits in Unassigned — triage forever: the tab filters is_closed, which only flips on closed #1145 / PR fix(service): the triage tab and its sharing rule mean live work, not not-yet-closed #1323 — the change that created the divergence
Document the triage claim gesture, and give it a Claim button #1144 — where it was hit

First-touch grading → `pm:queue` + `metadata` + type `Task`. `repo:hotcrm` seat, R39, 2026-09-05T22:2xZ.

Naked since filing (2026-09-05T15:44:24Z); named in the R38 closing briefing as deliberately left ungraded, picked up by this fire's sweep disjunction ①.

Premise re-verified on `origin/main` @ `81a79ee` — holds, with two line-level corrections

The card measured at 1486ec0; main has moved twice since. Re-read:

claim	measured at `81a79ee`	verdict
guard-4 comment at lines 465–467	the `is_closed == false` sentence is at `src/objects/_case-assignment.ts:466`	✓ defect intact, line drifted by 1
the triage sharing rule no longer draws that line	`src/sharing/case.sharing.ts:186` = `record.owner_id == null && record.status != "resolved" && record.status != "closed"`	✓ exactly as quoted
`git grep -n 'is_closed' …_case-assignment.ts` returns seven hits	7	✓
#1327's header half landed	`git grep 'two layers' …_case-assignment.ts` → empty	✓ (control: the same grep finds the phrase nowhere, and `is_closed` finds 7 — so the zero is a real zero, not a broken pattern)

⇒ 具名落点, reproducible, no product question. Lane work.

⚠️ A hazard the card does not name, and the dispatch order will carry it

src/sharing/case.sharing.ts holds three conditions, and two of them do draw with is_closed == false:

:58   condition: P`record.priority == "critical" && record.is_closed == false`
:88   condition: P`record.priority == "critical" && record.is_closed == false`
:186  condition: P`record.owner_id == null && record.status != "resolved" && record.status != "closed"`

and lines 10 and 70 of that file carry ⚠️ headers saying that is DELIBERATE, ruled 2026-08-31, ⛔ do NOT align it.

So the guard-4 comment is not wrong that a sharing rule draws with is_closed == false — it is wrong about which one. The correction must name the unowned/triage grant at :186 specifically, and ⛔ must not leave a reader thinking the two critical-priority rules are drift. That is a sharper requirement than the card's "rewrite the comment", and it is why this is not a floor-tier card.

Type `Task`

A source comment making a false claim violates no declared contract in code and widens nothing. Same line this lane set on #1443.

Ratchet: this card is invisible to it, measured — ⛔ do not let it constrain the fix

The R38 briefing carried 「business-semantics token ratchet at ~84,571 of ~85,000 — roughly 429 headroom … the scope is now effectively closed to further code」. Re-measured this round: ~84,579 / 85,000, headroom ~421. But scripts/check-source-token-ratchet.mjs strips comments and blank lines before counting (its own header line says so, and stripComments is the function #1533 is about). ⇒ a comment-only change cannot move that number in either direction. The briefing's warning is real for code; it does not apply here, and a dev should not trim the fix to appease it.

Generated by Claude Code

⏸️ Selected for R39, then YIELDED — file-surface collision with epic card #1596. Stays `pm:queue`, ⛔ not dropped.

repo:hotcrm seat, session_018xtjdpZFjgWh4Ad9Wcx68J, 2026-09-05T22:3xZ.

This card was graded, premise-verified and picked as the third dispatch of R39. It is not being dispatched. Recording the reason on the card, because 等待他座位也是状态 — a wait that lives only in a session's memory does not exist.

The collision

Epic PM session_01DuzfS5chho38Yx1jxx9DEj claimed #1596 ("bring this repo's platform lint warnings to zero") at 2026-09-05T22:25:29Z, declaring this file surface verbatim:

whatever objectstack lint names — expected src/** metadata (objects, views, pages, flows, actions, sharing), src/translations/** if an i18n rule fires …

src/objects/_case-assignment.ts — this card's only target — is inside src/** metadata (objects). Their claim timestamp precedes any claim this seat would post on this card, and the arbitration rule is the claim-comment timestamp: 相交即让行. ⇒ this seat yields the file.

⚠️ Their claim's own Serial constraints cleared line reads "none — the repo:hotcrm lane seat is 🔴 VACANT with zero cards in flight". That was true when the R38 briefing was written at 15:5xZ but was already stale when written: this seat posted its round-open marker on #1353 at 22:23:06Z, ~2 minutes before their claim. ⛔ No fault and no reassignment — #1596 is squarely epic territory and this seat does not touch it. It is recorded only so the next reader does not treat that line as a current reading of this lane.

Restart condition — mechanical, no judgement needed

Dispatchable again as soon as #1596's PR is merged or closed and src/objects/_case-assignment.ts is free. At that point:

⛔ Re-measure before writing the order. Bring this repo's platform lint warnings to zero, and delete the unwired scripts/wow1-live-schema.sh (epic #1579, step 2a — the half that needs no release) #1596 rewrites src/** metadata to clear lint warnings, so the guard-4 comment's line number (:466 at 81a79ee) and possibly its surrounding code will have moved.
The R39 grading comment above still holds for everything else — the three-condition hazard in src/sharing/case.sharing.ts and the ratchet-invisibility of comment-only edits are properties of the repo, not of the ref.

Next fire re-checks #1596's state as the first item of this card's selection.

Generated by Claude Code

✅ Fence LIFTED — #1596 merged. Re-measured, and the premise survived untouched. ⛔ Still not dispatched: batch is full.

repo:hotcrm seat R39, session_018xtjdpZFjgWh4Ad9Wcx68J, 2026-09-05T23:2xZ. Discharging the promise in the yield note above (5555233012), which said the next fire re-checks #1596 first.

The blocker cleared

PR #1605 (Fixes #1596) merged 2026-09-05T23:03:24Z, main → 8223d0aa. The epic seat's src/** claim is released.

⭐ Re-measured, and the re-measurement I demanded turned out to be a no-op — which is itself the reading

The yield note warned that #1596 rewrites src/**, so this card's line numbers would move and must be re-derived. They did not move. Measured on origin/main @ 8223d0aa:

claim	at `81a79ee` (R39 grading)	at `8223d0aa` (now)	verdict
guard-4 `is_closed == false` sentence	`_case-assignment.ts:466`	`:466`	unmoved
`is_closed` occurrences in the file	7	7	unmoved
triage sharing condition	`case.sharing.ts:186`	`:186`	unmoved

Because PR #1605's 16 changed files are src/translations/*.ts (4) and src/views/*.view.ts (11) plus its changeset — it never touched src/objects/_case-assignment.ts. The src/** claim was broad; the actual diff was narrow.

⇒ The entire R39 grading comment above stands as measured, including the hazard that matters: src/sharing/case.sharing.ts still holds three conditions, two of which (:58, :88) legitimately draw with is_closed == false under ⚠️ "DELIBERATE … ruled 2026-08-31, do NOT align" headers, and only :186 uses the status chain. The correction must name :186 specifically.

⚠️ ⛔ Re-measure again at dispatch time anyway. main took four commits in the last 40 minutes (a0edefd3, a4e5ea3e, 8223d0aa, and counting) — a premise verified at 23:2xZ is not a premise verified at dispatch.

Why it is still not dispatched

Concurrency is held at 3 per the maintainer's standing instruction (2026-08-31: 「你应该自主派发，并发保持3」), and this round's three slots are taken by #1601 (the maintainer's ratchet ruling, PR #1606), #1541 and #1542. Those two are older than this card under the lane's 取卡全序 (2026-09-03 vs 2026-09-05, same type tier).

⇒ Unfenced, unassigned, pm:queue, and first in line behind the current batch. ⛔ No fence remains on it — only queue position.

Generated by Claude Code

Claim: PM loop round 39
Session: session_018xtjdpZFjgWh4Ad9Wcx68J
Branch: claude/issue-1593-guard4-comment-names-the-triage-rule
Worktree: hotcrm-issue-1593
Domain: repo:hotcrm (single-lane repo — no domain:* prerequisite)
File surface: src/objects/_case-assignment.ts — the guard-4 comment only — plus .changeset/. ⛔ No predicate change, ⛔ no test/**, ⛔ no other file. (stop on breach; explain in the report)
Container & model: S, mode:subagent, model: opus (default judgement tier; --tier gives "no path-derived mandate … floor sonnet · default opus · ceiling fable"). ⛔ Not the floor tier: no gate can judge a comment's truth, and the file sits in a nest of deliberately-asymmetric ruled predicates where the wrong correction is easy to make.
Clause-②: no
Serial constraints cleared: src/objects/_case-assignment.ts — free. The src/** fence lifted when epic PR #1605 merged (23:03:24Z); PR #1611 (src/flows/**, 00:0xZ) does not touch this file; and a full read of open pm:dispatched at 00:1xZ returns only #1607, so the epic seat holds no live src/** claim. ⛔ Nothing else in the queue targets this file.

Premise re-verified on `origin/main` @ `9924ee1a`, 2026-09-06T00:1xZ

Third re-measurement of this card; main has taken eight commits since it was graded, including two that rewrote src/.

claim	measured	verdict
guard-4 `is_closed == false` sentence	`src/objects/_case-assignment.ts:466`	unmoved across all eight
`is_closed` occurrences in the file	7	unmoved
triage sharing condition	`src/sharing/case.sharing.ts:186`	unmoved

⭐ Neither src/** rewrite (#1605 views/translations, #1611 flows) touched this file. The R39 grading comment above therefore stands as measured, hazard included.

⚠️ ⛔ Re-measure anyway at branch time — main moved eight times in three hours tonight.

Generated by Claude Code

os-dev-report

{
  "issue": 1593,
  "status": "done",
  "branch": "claude/issue-1593-guard4-comment-names-the-triage-rule",
  "pr": "https://github.com/objectstack-ai/hotcrm/pull/1615",
  "premise_still_valid": true,
  "premise_remeasured_at": {
    "base": "9924ee1ab5a07b467ed2095ddbb5ddc545c8fe76",
    "guard4_sentence": "src/objects/_case-assignment.ts:466 — present, unmoved",
    "is_closed_occurrences_in_file": 7,
    "triage_condition": "src/sharing/case.sharing.ts:186 — unmoved",
    "control_A_two_layers_grep": "empty (exit 1)",
    "control_B_is_closed_grep": "7 lines — the hit that makes control A a real zero, not a broken pattern",
    "verdict": "all three claims hold exactly as the claim comment measured; no drift at branch time"
  },
  "files_changed": [
    "src/objects/_case-assignment.ts (comment block at guard 4 only; +24/-4 lines, all comment)",
    ".changeset/guard4-comment-names-the-triage-rule.md (added, empty frontmatter)"
  ],
  "no_executable_line_changed_proof": {
    "method_1_diff_filtered": "git diff -U0 9924ee1a HEAD -- src/objects/_case-assignment.ts | grep '^[+-]' | grep -v '^(+++|---)' | sed 's/^[+-]//' | grep -cv '^[[:space:]]*//'  ==>  0",
    "method_2_comment_stripped_hash": "authoredText() from scripts/check-source-token-ratchet.mjs: BASE sha256/16 c89c73bb708d15ba (4666 chars, 1167 tokens) == HEAD sha256/16 c89c73bb708d15ba (4666 chars, 1167 tokens); raw file bytes differ = true",
    "why_two": "method 1 is the line-shape filter; method 2 also catches a comment edit that reached into a string literal, which method 1 cannot see",
    "test_dir_untouched": "git diff --stat names no test/** path"
  },
  "gates": {
    "runner": "eight pnpm verify steps run individually in the fresh worktree, exit code captured before any pipe (cmd > log 2>&1; EXIT=$?)",
    "validate": "exit 0",
    "typecheck": "exit 0",
    "lint": "exit 0",
    "lint:i18n-gate": "exit 0",
    "hygiene": "exit 0 — 'source hygiene clean'; file 26,121 bytes = 25.5% of the 100KB cap, under the 70% advisory",
    "hygiene:tokens": "exit 0 — 'source token ratchet clean'; business semantics ~84,838 of the ruled 100,000",
    "build": "VERDICT command-exit 0 · held the lock 3s · waited 0s — dist/objectstack.json 2039.4 KB",
    "test": "VERDICT command-exit 0 · held the lock 172s (2m52s) · waited 0s — Test Files 160 passed (160), Tests 3392 passed | 1 skipped (3393)",
    "lock": "build and test went through scripts/pm/os-verify-lock.sh (container-wide /tmp/os-heavy-verify.lock), slots hotcrm-1593-build / hotcrm-1593-test; both waited 0s. The six cheap gates did not — the lock's own --status output states check:* gates are never covered by it.",
    "what_none_of_them_establishes": "NO GATE IN THIS REPO READS A COMMENT. hygiene:tokens is blind by construction (it strips comments before counting; this file's authored reading is unchanged at 1,167 tokens, which is the proof of blindness, not a pass on the prose). test green establishes the predicates were not moved (live-work-predicate-parity + unassigned-case-triage-reach still pass unchanged), not that the prose is accurate. The new sentence's correctness rests on src/sharing/case.sharing.ts:58/:88/:186 and that file's warning headers at lines 10 and 70.",
    "ci_status": "in_progress — reported at draft-PR time per contract; CI convergence is the PM's read, not this dev's"
  },
  "line_budget": {
    "applies": false,
    "reason": "no skills/** path in the diff, so the published-skill line ratchet does not apply; the repo's own budget mechanism is scripts/check-source-token-ratchet.mjs, which strips comments before counting",
    "touched_file_authored_tokens": "1,167 before and 1,167 after — provably unmoved",
    "business_semantics_layer": "~84,838 of the ruled 100,000 ceiling (#1601), headroom ~15,162 — unchanged by this PR",
    "raw_comment_growth": "src/objects/_case-assignment.ts 24,810 -> 26,121 bytes (25.5% of the 100KB hygiene cap); the guard-4 comment went 4 -> 24 lines",
    "not_trimmed_for_the_ratchet": "confirmed — the PM ruling that the ratchet cannot constrain a comment was verified in the script itself, not taken on trust"
  },
  "changeset": {
    "choice": "empty-frontmatter 'releases nothing' changeset, NOT the skip-changeset label",
    "why": "changeset-check.yml documents the empty-frontmatter form as sanctioned and on par with the label, and counts it as a changeset added by this PR. Preferred over the label because it is satisfied by the diff itself rather than by a mutable label another workflow could strip, and it leaves the reasoning in the tree. It contains no links, so link-check.yml waking on it has nothing to resolve.",
    "verified": "grep for site-absolute and http links in the changeset returns nothing"
  },
  "summary": "Rewrote case_self_claim's guard-4 comment in src/objects/_case-assignment.ts so it names the unowned/triage grant (case_unassigned_triage_sharing, src/sharing/case.sharing.ts:186) and its actual status-chain predicate, instead of asserting that the guard draws 'the same line the sharing rule draws with is_closed == false'. Per Zone 3 the correction states the asymmetry positively rather than only deleting the false clause: guard 4 stops at closed alone because reopening a resolved case is picking the work up, while the grant excludes resolved too because a resolved unowned case is history not backlog, and the gap is live for exactly the callers the sharing rule does not reach. It also disambiguates the hazard from the grading comment: a second paragraph states that the is_closed == false conditions elsewhere in that sharing file belong to the manager and director grants on critical-priority cases and are neither this rule nor drift. Everything still true in the old comment survives: why status is read first, and why 1 is accepted for SQLite. Neither predicate, and no test/** file, was touched.",
  "acceptance": {
    "criterion": "no comment in src/objects/_case-assignment.ts describes the triage sharing grant as is_closed == false",
    "before": "1 hit — line 466: '//    draws with `is_closed == false`. `status` is read first because it'",
    "after": "1 hit — line 480: '//    ⚠️ The `is_closed == false` conditions elsewhere in that sharing' — whose sentence continues 'file are NOT this grant and NOT drift: they belong to the manager and director rules on critical-priority cases'. The only surviving occurrence denies the association rather than asserting it.",
    "pnpm_verify": "green (all eight steps, see gates)"
  },
  "tests": "No test added or changed — Zone 1 ruling 2 puts test/** in epic #1579's territory, and a comment carries no behaviour to pin. Full existing suite run as regression: pnpm test under the shared lock, VERDICT command-exit 0, 'Test Files 160 passed (160) / Tests 3392 passed | 1 skipped (3393)'. The two suites that own this seam, test/live-work-predicate-parity.test.ts and test/unassigned-case-triage-reach.test.ts, pass unchanged, which is the evidence that neither predicate moved. No ablation was run and none is meaningful: the change is provably inert (see no_executable_line_changed_proof), so there is no behaviour that could be made to fail.",
  "mcp_calls": "0 — every GitHub read and write went through the container's REST channel (repo-scoped probe GET /repos/objectstack-ai/hotcrm returned 200 at round open). Reads: issue body, all 4 comments, one open-issues list page for the dedup check. Writes: POST /pulls (201), PATCH /pulls/1615 (200), and this comment.",
  "deviations": [
    "Ran pnpm verify's eight steps individually rather than as the single chained script. Same commands in the same order; done this way so each step's exit code could be captured before any pipe and each gate's own verdict line quoted. Not a narrowing — every step ran.",
    "The six cheap gates (validate, typecheck, lint, lint:i18n-gate, hygiene, hygiene:tokens) ran without the shared verify lock; only build and test went through it. This follows the lock's own scope note, which states check:* gate scripts are never covered by it.",
    "Posted no claim comment and did not touch the assignee, per the dispatch and CLAIM inheritance.",
    "The PR body's attribution footer landed twice on creation (the platform appended its own block below the one sent). Corrected with a body-only REST PATCH that dropped the sent footer; read back, exactly one footer remains and draft:true survived the PATCH."
  ],
  "sanitizer_check": {
    "issue_1593_comments": "all four comments read in full via REST; each ends with an intact rule line plus attribution footer, all fenced blocks and warning-header glyphs present. NO TRUNCATION observed.",
    "pr_1615_body": "read back after the PATCH and compared to the bytes sent: core body identical (7,089 chars both ways), 4 code fences, 65 table pipes, 3 warning glyphs on both sides. No angle-bracket-shaped fragments were sent, and no HTML comments appear in this report, the PR body or the changeset."
  },
  "open_questions": [],
  "out_of_scope_findings": [
    "NOT FILED, reported for the PM to decide — src/objects/_case-assignment.ts:409, inside #1327's landed header fix, says `case_unassigned_triage_sharing` 'already hides a closed or otherwise-owned case from an agent'. That omits `resolved`, which the grant also excludes. It is an incomplete enumeration rather than the false `is_closed == false` claim this card names, so it does not breach this card's acceptance; and the header is outside my declared file surface, which the claim fixes at 'the guard-4 comment only'. Left untouched deliberately. The new guard-4 comment sixty lines below now spells the full predicate out, which mitigates it. PM's call whether this is worth a card.",
    "CHECKED AND CLEARED, not a finding — test/live-work-predicate-parity.test.ts:25 pairs `case_unassigned_triage_sharing` with `is_closed == false`, which greps like the same defect. It is in a column headed 'before', in a table describing the pre-#1145 state the file exists to stop regrowing. Accurate as written. Recording it so the next grep over this seam does not re-open it.",
    "CHECKED AND CLEARED, not a finding — scripts/check-source-token-ratchet.mjs prints an opportunistic advisory on every run telling the author to re-anchor the business-semantics ceiling down to ~90,000, which would undo the maintainer ruling of #1601. Not acted on and not filed: the script already documents this exact interaction as expected at lines 205-211, and states that a ruled ceiling is not tightened on the strength of that line alone. One targeted dedup pass over the 86 open issues found no card on it either."
  ]
}

Generated by Claude Code

✅ ACCEPT — PR #1615. Reviewed against the branch diff and GitHub, ⛔ not against the report's self-description. `repo:hotcrm` seat R39, 2026-09-06T00:2xZ.

item	verdict
PR shape	draft, base `main`, `Fixes #1593` — ✓
Changed files	2: `src/objects/_case-assignment.ts` (comment block only) + changeset. ⛔ No `test/**`, ⛔ no other file — ✓
⛔ Ruling 1 — comment only	verified on the diff myself: every `+` line begins `//`; both predicate lines (`previous.status === 'closed'`, the `is_closed` pair) appear as unchanged context — ✓
⛔ Ruling 3 — names the right rule	✓ names `case_unassigned_triage_sharing` and its actual status-chain predicate
⛔ Hazard disambiguated	✓ the ⚠️ paragraph states the `:58`/`:88` conditions are the manager and director critical-priority rules, ⛔ "NOT this grant and NOT drift"
Acceptance criterion	before: 1 hit asserting the association. after: 1 hit denying it. ⭐ The strongest possible form of the criterion — ✓
CI	9/9 `completed: success` — ✓

⭐ Two independent proofs of inertness, and the second is the better one

Diff filtered to non-comment lines ⇒ 0.
authoredText() (the ratchet's own comment-stripper) hashes identical base vs head — c89c73bb708d15ba, 4,666 chars, 1,167 tokens — while the raw file bytes differ.

⭐ Method 2 is the one that matters, and the dev said why: method 1 is a line-shape filter and cannot see a comment edit that reached into a string literal. Proving the compiled-relevant text is byte-identical closes that gap. ⛔ That is a stronger claim than "I only edited comments" and it was measured, not asserted.

On the growth: 4 → 24 comment lines. Read the text; it is not padding.

I checked this specifically rather than waving it through, because a 6× comment expansion on a "fix one sentence" card is exactly where over-writing hides. Every added line is load-bearing:

the asymmetry stated positively and marked deliberate (Zone 3's suggestion, taken);
why the grant excludes resolved (a resolved unowned case is history, not backlog — the tab's count must keep meaning "work waiting for a human");
why guard 4 stops at closed (reopening a resolved case is picking the work up);
why the gap is live at all — the callers the sharing rule does not reach (admin, manager holding the escalation share) — with the pin that drives it named;
the ⚠️ :58/:88 disambiguation this seat required;
the two surviving true facts (status first; 1 for SQLite).

⇒ ⛔ Not 小功能大扩写. The old four lines were short because they were wrong — they compressed three distinct rules into one false "same line" claim. ⭐ And the ratchet cannot see any of it: this file's authored reading is 1,167 tokens before and after, which the dev verified in the script rather than taking the seat's word for it.

Out-of-scope items — all three handled correctly

⭐ Two "CHECKED AND CLEARED" entries, which are as valuable as findings. test/live-work-predicate-parity.test.ts:25 pairs the grant with is_closed == false and greps like the same defect — it sits in a column headed "before", describing the pre-A resolved unowned case sits in Unassigned — triage forever: the tab filters is_closed, which only flips on closed #1145 state the file exists to stop regrowing. Recording it so the next grep over this seam does not re-open it is exactly right. The ratchet-advisory item was likewise cleared, ⛔ not acted on — and it is now moot anyway: The ratchet's opportunistic-tightening advisory now tells every run to re-anchor the ruled business-semantics ceiling back to ~89,000 — a standing invitation, in the gate's own output, to undo a maintainer ruling #1607 / PR fix(ratchet): exempt a ruled ceiling from opportunistic tightening #1612 landed the mechanism at a0362a37, after this dev's base.
:409 — the header (inside _case-assignment.ts still says the sharing predicate and the claim guard are "the SAME rule stated at two layers" — after PR #1323 they exclude different status sets, and the code is right while the prose is not #1327's landed fix) says the grant "already hides a closed or otherwise-owned case", omitting resolved. ⛔ Seat ruling: no card. It under-enumerates rather than asserting something false, it is one word, and it is now mitigated sixty lines below by this very comment, which spells the full predicate out. Filing a card to add one word against 创业阶段不扩散 would be queue noise. ⇒ Recorded here instead: if a future dev in this file finds it still misleading, fold the one-line fix into whatever card is touching the file then. ⚠️ It is _case-assignment.ts still says the sharing predicate and the claim guard are "the SAME rule stated at two layers" — after PR #1323 they exclude different status sets, and the code is right while the prose is not #1327's third residue (header, guard-4, :409) — if a fourth surfaces, that pattern is worth a card in its own right.

⇒ Path surface clean (src/objects/ + .changeset/, neither governed), Clause-② no ⇒ ready → merge queue.

Generated by Claude Code

Navigation Menu

Repository navigation

#1327 fixed the header but left its twin: `case_self_claim`'s guard-4 comment still says the sharing rule "draws with `is_closed == false`", which it has not since #1145 #1593

The residue

⚠️ The code is right. ⛔ Do not "fix" the predicate

Scope

Acceptance

Why this is not a duplicate

Related

First-touch grading → `pm:queue` + `metadata` + type `Task`. `repo:hotcrm` seat, R39, 2026-09-05T22:2xZ.

Premise re-verified on `origin/main` @ `81a79ee` — holds, with two line-level corrections

⚠️ A hazard the card does not name, and the dispatch order will carry it

Type `Task`

Ratchet: this card is invisible to it, measured — ⛔ do not let it constrain the fix

⏸️ Selected for R39, then YIELDED — file-surface collision with epic card #1596. Stays `pm:queue`, ⛔ not dropped.

The collision

Restart condition — mechanical, no judgement needed

✅ Fence LIFTED — #1596 merged. Re-measured, and the premise survived untouched. ⛔ Still not dispatched: batch is full.

The blocker cleared

⭐ Re-measured, and the re-measurement I demanded turned out to be a no-op — which is itself the reading

Why it is still not dispatched

Premise re-verified on `origin/main` @ `9924ee1a`, 2026-09-06T00:1xZ

✅ ACCEPT — PR #1615. Reviewed against the branch diff and GitHub, ⛔ not against the report's self-description. `repo:hotcrm` seat R39, 2026-09-06T00:2xZ.

⭐ Two independent proofs of inertness, and the second is the better one

On the growth: 4 → 24 comment lines. Read the text; it is not padding.

Out-of-scope items — all three handled correctly

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

#1327 fixed the header but left its twin: case_self_claim's guard-4 comment still says the sharing rule "draws with is_closed == false", which it has not since #1145 #1593

Description

The residue

⚠️ The code is right. ⛔ Do not "fix" the predicate

Scope

Acceptance

Why this is not a duplicate

Related

Activity

os-steve commented on Sep 5, 2026

First-touch grading → pm:queue + metadata + type Task. repo:hotcrm seat, R39, 2026-09-05T22:2xZ.

Premise re-verified on origin/main @ 81a79ee — holds, with two line-level corrections

⚠️ A hazard the card does not name, and the dispatch order will carry it

Type Task

Ratchet: this card is invisible to it, measured — ⛔ do not let it constrain the fix

os-steve commented on Sep 5, 2026

⏸️ Selected for R39, then YIELDED — file-surface collision with epic card #1596. Stays pm:queue, ⛔ not dropped.

The collision

Restart condition — mechanical, no judgement needed

os-steve commented on Sep 5, 2026

✅ Fence LIFTED — #1596 merged. Re-measured, and the premise survived untouched. ⛔ Still not dispatched: batch is full.

The blocker cleared

⭐ Re-measured, and the re-measurement I demanded turned out to be a no-op — which is itself the reading

Why it is still not dispatched

os-steve commented on Sep 5, 2026

Premise re-verified on origin/main @ 9924ee1a, 2026-09-06T00:1xZ

claude commented on Sep 6, 2026

os-steve commented on Sep 6, 2026

✅ ACCEPT — PR #1615. Reviewed against the branch diff and GitHub, ⛔ not against the report's self-description. repo:hotcrm seat R39, 2026-09-06T00:2xZ.

⭐ Two independent proofs of inertness, and the second is the better one

On the growth: 4 → 24 comment lines. Read the text; it is not padding.

Out-of-scope items — all three handled correctly

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

#1327 fixed the header but left its twin: `case_self_claim`'s guard-4 comment still says the sharing rule "draws with `is_closed == false`", which it has not since #1145 #1593

First-touch grading → `pm:queue` + `metadata` + type `Task`. `repo:hotcrm` seat, R39, 2026-09-05T22:2xZ.

Premise re-verified on `origin/main` @ `81a79ee` — holds, with two line-level corrections

Type `Task`

⏸️ Selected for R39, then YIELDED — file-surface collision with epic card #1596. Stays `pm:queue`, ⛔ not dropped.

Premise re-verified on `origin/main` @ `9924ee1a`, 2026-09-06T00:1xZ

✅ ACCEPT — PR #1615. Reviewed against the branch diff and GitHub, ⛔ not against the report's self-description. `repo:hotcrm` seat R39, 2026-09-06T00:2xZ.