---

title: 07 — Macro 2 Handoff date: 2026-05-31

07 — Macro 2 Handoff

Exact enough that the next agent does not re-investigate from scratch. All items below were rehearsed GREEN this run (BEGIN..ROLLBACK) or are precisely specified.

M2.1 — Threshold registry (display_policy)

• Rehearsed: CREATE TABLE display_policy(species_code text PK, max_ungrouped smallint DEFAULT 50, note text, CHECK(max_ungrouped>0 AND max_ungrouped<=50)). 50 = MAX ceiling (not a trigger-on-50). Seeded __default__=50, dot_tool=30, collection=50.
• Live signal: 28 / 160 leaves exceed their species ceiling (ungrouped) → those get grouped/labelled by default.
• Reuse: taxonomy_facets.max_labels_per_entity exists but is label-count, not a display ceiling — distinct concept; kg_thresholds table exists → reuse for storage.
• Gate: additive table; needs approval to commit.

M2.2 — Label / grouping (PIV-311 + existing machinery)

• entity_labels = 726,850 rows, 23 distinct label_code → PIV-311 (group by label_code) gives 23 groups (proven via real engine, doc 04).
• Reuse: label_rules (38, has skip_wide_warning), taxonomy (58), species_collection_map (164) as discriminator. Do not build new label machinery.

M2.3 — registry_pin

• Rehearsed: CREATE TABLE registry_pin(id, object_ref, object_kind, surface_ref, pinned_by, scope DEFAULT 'user', scope_ref, reason, priority, active, created_at, CHECK(scope IN ('global','user','role','team'))). Inserted 2 pilot rows GREEN.
• No pin/ghim table exists today (REGISTRY_MISSING) → this is NEW, propose-only.
• Once committed, PIV-321 (pins by scope, grouped-count) activates (doc 04).

M2.4 — Phantom law (RG4, LAW_DEFINITION_GAP)

• record > actual is source_model-dependent (doc 06): model-A = write-race, model-B = phantom candidate. There is no authoritative phantom definition and no phantom_count column anywhere.
• Council must (a) define phantom per source_model, (b) ratify a phantom_count, (c) approve PIV-302 promoting from "candidate count" to "phantom count".
• event_type_registry (40) has no count_integrity.* / phantom / pin event types → NEW event types needed; system_issues (≈179k) is the notify substrate (Đ45).

M2.5 — Scan cadence (CAT-023 + 5 unmeasured)

• CAT-023 actual_count is a frozen scan (985,471) while record_count tracks live → net_gap grows. Decide: (a) faster scan cadence, or (b) source actual_count live from PIV-019.
• CAT-1006..1010 (apr_action_types, apr_request_types, apr_approvals, admin_fallback_log, law_version_verification_log): counts NULL, registry_collection NULL → run a count scan + set registry_collection, then pivot.

M2.6 — Directus / API exposure

• Expose the six committed views as Directus read collections / API endpoints so the Nuxt shell reads pivot-backed counts — retiring the legacy /api/registry/health Σ|gap| math (Đ28 violation). REAL no-hardcode violations confirmed on VPS source /opt/incomex/docker/nuxt-repo/web: health.get.ts:123 totalGap=reduce(+Math.abs(gap)), index.vue:312 code:'CAT-017', etc. (full list in memory). pivot-query.get.ts = GOOD pattern to follow.

Macro-2 ordering

RG1+RG2 → six-object commit (Macro 1) → M2.1 display_policy ∥ M2.2 PIV-311 → M2.3 registry_pin → M2.4 phantom-law council (RG4) → M2.5 scan cadence → M2.6 Directus.