Complete the #757 deferrals: the exclusion pass, the measured U8 lever, the release-cut certification, and the re-armed thresholds

changelog.d/757-certification-contract.added.md

@@ -0,0 +1,8 @@
+The publication contract verifies the multi-part release certification
+(`release_certification.json`): exact field set, the three mirrored part
+scopes and their committed scoped-manifest digests, full-manifest spec
+pins, union/no-gap/no-overlap over the declared entry set, per-part
+fully-passed status censuses (shippability recomputed, never read off the
+flag), the diagnostics digest join, and the release-key signature over the
+whole document. The mirrored constants are held in lockstep with the build
+shard by the contract-pins sync tests.

changelog.d/757-compile-parity-registers-current.fixed.md

@@ -0,0 +1,11 @@
+The two ledger compile-parity signed-difference registers are regenerated
+against the pinned chronicle feed before their runner ships: 13 stale
+entries pruned (scotgov council-tax stock and SCP spending, three SLC
+recipient rows - the live compilation now matches the fixture) and 13 SLC
+entries re-kinded `fixture_only` -> `calibration_drift` with measured
+values (the SLC chronicle waves completed after the June fixtures froze).
+Zero entries added: the live diff carried no unsigned differences, so the
+regeneration is exactly the correction the gate's own anti-rot refusals
+demanded. Both release-cut preflight gates now pass against the pinned
+feed - verified live, so the producer's first real invocation does not
+open on a known-stale register.

changelog.d/757-national-release-id.added.md

@@ -0,0 +1,10 @@
+The certified UK national line gets its release id (ruling 2026-08-27):
+`microcosm-uk-2024-25-national`, one constant name across cuts - run
+identity and ordering live in the Logbook chain and artifact versioning,
+not in the id. The `2024-25` segment follows the uk-data survey-vintage
+convention and the `-national` segment keeps the id disjoint from the
+exact-k ladder's `-k<N>` shape. The seam refuses it (a shippable name
+belongs to the release-cut producer), and `required_release_files()` now
+demands `release_certification.json` for it - completing the review
+round's finding 1 at the id-keyed layer, with the content-keyed refusal
+staying as the belt for part-carrying directories under other ids.

changelog.d/757-phase3-acceptance.changed.md

@@ -0,0 +1,12 @@
+The committed spine acceptance receipt moves to spine-i, the candidate
+rebuilt at the follow-up tip: payload-identical to spine-g (the
+certification machinery, exclusion pass, QRF re-arm, and lever revert are
+all payload-inert), 14/14 battery at release-candidate strictness with a
+signed report, fit-weight records in the sidecar, identity ladder e4-e8
+green, strict parity `signed_parity` with 0 unsigned. The first certified
+cut itself is blocked at `uk_target_fit` on 13 characterized cells - 8 UC
+caseload/two-child cells (the U8 lever is measured and exhausted; the
+binding constraint is capital-test support, microcosm#750), 4
+exclusion-set-dependence artifacts (microcosm#792), and 1 sparse-band
+sibling - with the signed blocked-run receipts in the 757-swap acceptance
+evidence and the adjudication queue in the PR.

changelog.d/757-qrf-tail-rearm.changed.md

@@ -0,0 +1,12 @@
+`uk_qrf_tail_concentration` is re-armed from the #686 L3 baselines (#757
+B4): top_k 100 stays the measurement grid anchor, max_top_share moves to
+the exact measured maximum over the checked surface (0.9994670564654868,
+hmrc_spi_other_social_security_income at 104 carriers on spine-a), and
+min_nonzero_records to the thinnest measured column above the grid anchor
+(104). The three saturated sub-anchor columns (taxable termination pay,
+charitable investment gifts, SDA; 12-24 carriers, top-100 share 1.0) go
+thin visibly on every run. The gate notes record the measuring run - the
+baselines file digest, the measured artifact, and the defining column per
+threshold. The 12 household-surface grids are measured but not yet armed
+(declared follow-up). Gate policy/manifest/fingerprint digests re-cut from
+the live producer payload.

changelog.d/757-release-cut-certification.added.md

@@ -0,0 +1,16 @@
+The release-cut certification producer (#757 B5): the 16 declared national
+preflight/terminal gates get their executable home back
+(`tools/certify_uk_release_cut.py` over
+`uk_runtime/release_certification.py`), running as a scoped release-candidate
+battery against the calibrated candidate with evidence reconstructed from
+the persisted artifacts - the spine sidecar (which now carries each fitting
+stage's `FitWeightRecord`s across the run boundary), the seam's diagnostics
+and build record, and the per-run licensed input-mass reference. The
+multi-part certification the 5413502559 audit specified composes over the
+spine, seam, and release-cut reports: union to the full declared entry set,
+no gap, no overlap beyond `uk_aggregate_admin`, per-part signatures and
+committed-spec scoped digests, full phase coverage, a closed identity join
+(spine report -> sidecar -> build record -> diagnostics -> candidate
+bytes), the doctrine and its receipted overrides recorded verbatim, and the
+rule-1 score receipt cross-pinned. A candidate's shippability verdict comes
+only from the certification.

changelog.d/757-review-dispositions-793.fixed.md

@@ -0,0 +1,18 @@
+The #793 review round (four findings, all the reports-success-without-
+testing class): a release shipping national-line gate artifacts without
+`release_certification.json` now refuses instead of validating clean by
+omission (the id-keyed required-files rule waits on the canonical national
+release-id); the rule-1 cross-pin reads `artifacts.candidate.sha256` - the
+field that names what the scorer measured - instead of a substring scan; a
+malformed sidecar fit-weight block raises as corruption instead of
+degrading to the empty tuple (a genuinely empty fitting stage still fails
+the audit, never vacuously passes); and the parity evidence's shared
+name@period grain is now a loud refusal rather than an implicit
+convention. Alongside, the duplication the round pointed at is
+consolidated: one scope-filtering helper (`uk_scoped_gate_manifest`, with
+a source parameter serving the spine driver's stub point) replaces three
+copies and a dead zero-caller variant, one `finalize_uk_scoped_gate_report`
+grafts and re-signs every scoped report, the certification's private
+cross-module imports become public names, and `GateBatteryRun` exposes its
+attested digests as properties so derivation stops routing through a
+signed payload.

changelog.d/757-seam-refuses-release-candidate.changed.md

@@ -0,0 +1,7 @@
+The calibration seam refuses `--release-candidate` outright and refuses
+canonical UK release ids (the #757 release-cut audit, issue comment
+5413502559): its scoped battery covers 6 of the declared gate entries and
+must never sign a shippability claim. The hand-written build-record
+`shippable` literal retires with it, replaced by a pointer to the
+release-cut certification artifact
+(`<staging>.release_certification.json`) whose verdict is authoritative.

changelog.d/757-target-fit-exclusion-pass.changed.md

@@ -0,0 +1,12 @@
+The calibration measure-exclusion register carries the adjudicated
+`uk_target_fit` disposition pass (microcosm#757, issue comment 5427936411):
+42 windowed exclusions covering the eleven 1m+ top-income channel cells,
+the three SLC zero/structural-support channels, the sixteen UC
+payment-distribution cells, the OBR welfare-cap pair, the three ONS
+household-composition cells pending a relationship-to-head frame column
+(microcosm#791), the six sparse HMRC band cells, and `obr.fuel_duties`
+(universe scope; ledger retarget pending chronicle-side adjudication).
+The six UC caseload/two-child-limit cells are deliberately not excluded —
+they ride the `would_claim_uc` lever run — and neither are
+`couple_no_children` nor `state_pension_income_band_40_000_to_50_000`,
+which the exclusion re-run is expected to pull inside the band.

changelog.d/757-would-claim-uc-lever.changed.md

@@ -0,0 +1,11 @@
+The pre-registered `would_claim_uc` U8 lever (uk-data#452) was run and
+measured per the #757 `uk_target_fit` dispositions (issue comment
+5427936411), and reverted on the receipts: a raise to 0.85 leaves every
+failing UC caseload cell unchanged (`dwp.uc.households` -44.9% at both
+rates) while destroying the legacy housing-benefit surface
+(`obr.housing_benefit` -0.0% at 0.55 vs -52.4% at 0.85 - the raise moves
+claimants off legacy benefits) and perturbing the QRF predictor surface
+through engine-computed `household_net_income`. The contract entry stays
+frozen at 0.55 and now records the run; the receipts live in the 757-swap
+acceptance evidence. The binding constraint on UC caseload is capital-test
+support (microcosm#750), not take-up support.

packages/microcosm-build/src/microcosm/build/gate_battery.py

@@ -835,6 +835,24 @@ def _next_phase(self) -> str | None:
                 return phase
         return None
 
+    @property
+    def gates_manifest_sha256(self) -> str:
+        """Canonical digest of the scoped manifest this run attests."""
+
+        return self._gates_manifest_sha256
+
+    @property
+    def spec_fingerprint(self) -> str:
+        """Composition fingerprint derived from the manifest digest."""
+
+        return self._spec_fingerprint
+
+    @property
+    def policy_sha256(self) -> str:
+        """Canonical digest of the declared gate policy entries."""
+
+        return self._policy_sha256()
+
     def run_phase(self, phase: str, context: EvidenceContext) -> GatePhaseReport:
         """Evaluate one phase and persist the full report before returning.