Rewrite the family-4 record in filter-logic-conformance.ts: the 2026-08-10 exclude ruling was withdrawn — include re-affirmed

[claude]

Follow-up ordered by the 2026-08-10 maintainer ruling on #5299 (recorded there in full). Filed by the domain:engine-core seat; dispatchable immediately.

The fact

PR #7391 (5f7669e0c) recorded the 07:33Z exclude ruling as a "RULED but not yet enrolled" family-4 note in packages/spec/src/data/filter-logic-conformance.ts's header, with matching "ruled target" annotations in packages/formula/src/matches-filter-not-null-safe.test.ts and the two driver-memory pin files. Minutes after it merged, the maintainer withdrew cells 1/3 of that ruling and re-affirmed #5298's include direction ($ne/$nin/$notContains match no-value rows) as the platform semantics.

So the header currently proclaims a ruled target that is no longer ruled — a standing declared≠enforced contradiction, in the exact file whose job is to be the single source of truth.

The work (XS, annotation-only, ⛔ zero behaviour)

1. Rewrite the family-4 header note: the exclude direction was ruled 2026-08-10 07:33Z, measured (the eleven-surface table), and withdrawn the same day; include is re-affirmed platform semantics. Keep the measurement table reference — it is the record of why.
2. Flip the matching "ruled target / superseded direction" annotations in the formula and driver-memory pin files to say include is the affirmed direction. ⛔ The pins' assertions stay byte-identical — they already assert the include row sets.
3. The four enrolled d-column cases' pointers (added by docs(spec,formula,driver-memory): record the ruled EXCLUDE direction for negative operators over no-value rows (#5299) #7391) likewise updated: they state the affirmed direction, not a superseded one.

⛔ No case rows added or removed; check:driver-conformance must be byte-identical before/after. Tests-only+comments ⇒ read the changeset gates for the skip-changeset vs patch decision. packages/spec touched ⇒ all 9 spec gates + the PM files the #6017 declaration.

Refs #5299, #5298, #5146, PR #7391 (5f7669e0c).

---

Generated by Claude Code

[claude]

CLAIM — domain:engine-core seat (#6019). Session session_01MwoubC3jL271FYt9rGXwxb · branch claude/issue-7406-family4-record-rewrite.

Race re-read done; assignees measured EMPTY via REST immediately before assigning.

---

Generated by Claude Code

[claude]

Claim-format amendment — the 10:1xZ claim above predates this seat loading today's SKILL.md update (#7357/#7379); re-stating in the step-4 fixed shape so the half-state sweeper's H2 predicate can read it:

Claim: PM loop wave 7 (seat #6019)
Session: session_01MwoubC3jL271FYt9rGXwxb
Branch: claude/issue-7406-family4-record-rewrite
Worktree: objectstack-7406
Domain: domain:engine-core

Labels repaired in the same write: pm:queue → pm:dispatched (H3 discipline).

---

Generated by Claude Code

[claude]

ACCEPT — PR #7419. Ready + queued; #6017 declaration filed (comment-only spec touch).

Verified against all criteria:

1. ✅ Family-4 header record rewritten: ruled 07:33Z → measured → withdrawn same day; include re-affirmed — the eleven-surface measurement stays referenced as the record of why.
2. ✅ Assertions byte-identical — verified at the diff level: the single expected-containing changed line is header prose, not an assertion row. check:driver-conformance unchanged.
3. ✅ The four enrolled d-column pointers and the formula/driver-memory pin annotations now state the affirmed direction; driver-memory changes are comment-only (sanctioned [裁决] driver-memory / driver-mongodb 投入冻结 —— 维护者 2026-08-05 口径（跨单锚点） #5499 exception).
4. ✅ skip-changeset label per the gate — four files, comments only, nothing ships.
5. ✅ Full spec gate suite green.

The tree no longer proclaims a withdrawn target in its single source of truth. Auto-merge armed; issue closes on merge.

---

Generated by Claude Code