00 — Overview & Control

Mission: FULL_STACK_GOVERNANCE_ALIGNMENT_AUDIT_FOR_REGISTRIES_PIVOT_GROUPING Class: Read-only full-stack governance audit + independent recommendation. No implementation. Audit run date: 2026-06-01 (package dated 2026-05-31 per mission spec). Host: contabo VPS · DB: directus (Docker container postgres, PG 16.13). Read channel: Agent-Data KB + live PG read-only (context_pack_readonly) + law text-as-code. Apply channel: KB docs only. Zero PG / Directus / Qdrant / Nuxt mutation.

---

Hard Gate 0 — result: PASSED

#	Check	Result
1	KB read access	✅ mcp__agent-data__* + batch_read working
2	Live PG read-only access	✅ role context_pack_readonly, READ ONLY tx, 5s timeout, LIMIT 500
3	Target DB = directus	✅ current_database()='directus'
4	No PG apply channel used	✅ query_pg is AST-validated read-only; ssh/docker exec NOT used
5	No mutation	✅ only SELECT / information_schema / pg_catalog
6	Audit/recommendation only	✅
7	No UI/schema/route/law/approval/queue/event/DOT change	✅ none performed
8	Inspected live PG (not KB only)	✅ 30+ live queries across the governance substrate
9	Separated the six SSOT layers	✅ Law / Authority-ownership / Approval / DOT-execution / Issue-event / UI-display (see doc 01)

---

What we were asked

We are building Registries-Pivot as the independent verification surface for the whole system. The governance risk: grouping / classification / label / pin / phantom / pivot-coverage / cleanup / DOT-grouping must NOT become a local Registries-Pivot governance island. They must reuse the central governance model (Điều 37 / governance_registry / approval spine / DOT governance / pivot law / label law). This audit reconstructs the full-stack governance model, finds where it is centralized vs fragmented, maps the Registries-Pivot artifacts onto it, marks local-island risks, recommends the correct owner + approval path, and proposes law/design patches — without implementing anything.

---

Headline verdict

The central governance spine is REAL and must be reused — but it is INCOMPLETE at the agency-ownership layer, DRIFTED at the schema layer, and ABSENT for the grouping/threshold/pin/phantom concepts. Registries-Pivot grouping is therefore at genuine risk of becoming a local island, NOT because the central model is missing, but because nobody has yet been assigned to own classification/pivot at the agency (TẦNG 3) layer, and because three of the proposed artifacts (display_policy threshold, registry_pin, phantom definition) have no law/domain home at all.

The correct move is NOT to build local governance, and NOT to bolt owner_gov_code columns onto each new policy table (that pattern exists nowhere in the system today). It is to use Điều 37's relational ownership model: assign the already-existing classification and pivot domains to agency owners, route every change through the Điều 32 approval spine (which already has reclassify / rule_change / schema_add / new_dot / retire request types), execute via Điều 35 paired DOTs, and raise findings to system_issues / event_outbox (Điều 45).

Four SSOT verdicts (detail in docs 01–02)

• Law SSOT = normative_registry (47 rows, enacted). Verdict: authoritative but INCOMPLETE — Đ20, Đ23, Đ45 exist as KB law docs but are not registered rows; law_catalog (5 rows) is stale/legacy; governance_docs UI index has drifted (claims "Constitution v3.5 / 25 Điều"; mislabels Đ26 as "Counting", Đ28 as "Matrix").
• Authority / ownership SSOT = governance_registry (9 agencies) + governance_relations (8 edges) + law_jurisdiction (domain ownership). Verdict: real and Đ37-mandated, but SPARSE and DRIFTED — 4 "mother" factories still draft; ownership edges only agency→law (no object/collection edges); classification & pivot domains have no agency owner; live column/CHECK vocabulary diverges from enacted Đ37 v3.3.
• Approval SSOT = approval_requests + apr_approvals (Đ32). Verdict: the real and ONLY approval spine — quorum enforced (high = president + ≥2 ai_council + 0 reject), no-self-approval enacted (§4.3.1). NO grouping/pivot/label/pin/phantom artifact has any approval; os_proposal_approvals=0; the only human approver is president (14 approvals, all S178 DOT-repair). ⇒ NO_APPROVAL_FOUND / COMMIT_FORBIDDEN stands.
• DOT-execution SSOT = dot_tools + law_dot_enforcement + dot_operations/dot_iu_command_catalog under Đ35. Verdict: complete and reusable — classify/audit/health/verify operations already exist; Đ35 B-tier(execute, approval-gated) ↔ A-tier(audit, read-only auto-approve) pairing rule is exactly the scan/propose/apply/audit model grouping needs.

Biggest fragmentation risks

1. Agency-ownership gap — classification (Đ24/Đ29) and pivot (Đ26) domains have no governance_registry agency owner. A law owns them; no body executes them. This vacuum is what tempts a local Registries-Pivot owner.
2. display_policy (50-ceiling) and registry_pin were designed as standalone LOCAL tables with no relational governance binding → highest island risk.
3. Threshold / phantom-definition / pin have no law and no domain → LAW_GAP → tempts local definition.
4. Live direct-pg-Pool API path (server/utils/rpGatewayDb.ts) bypasses the PG→Directus→Nuxt convention → process island needing explicit Đ41/API-exception approval.
5. Schema drift between enacted Đ37 v3.3 and the live governance tables weakens the very anchor the central model relies on.
6. governance_audit_log = 1 stale row → the governance-relation audit loop is not closing.

Recommended owner (detail in doc 12)

Option 4 — split by object type, with named existing owners (reuse-first, no new agency):

• Classification / grouping / threshold / label-dimension / phantom-definition / pin POLICY → GOV-COUNCIL (direct, per Đ37 §4.12(d) cross-cutting tie-break + the tier_registry → GOV-COUNCIL precedent).
• Count-integrity / orphan / phantom-detection / pivot-coverage HEALTH → GOV-SIV (System Integrity, Đ31).
• DOT execution of grouping scan/propose/apply/audit → GOV-DOT (Đ35 paired DOTs).
• Registries-Pivot render templates / route → GOV-MOUT + Đ28 design_templates.

A new GOV agency (e.g. GOV-CLASS) is not recommended unless Council explicitly delegates — reuse-first says assign the un-owned domain to Council/SIV first.

---

Package contents

Doc	Branch	Content
00	—	This control doc
01	A	Full-stack governance SSOT map (six layers)
02	B	Current governance architecture map (live PG inventory)
03	C	Điều 37 ownership interpretation + candidate owners
04	D	Local-governance-island risk review (per artifact)
05	E	Unified Governed Classification & Grouping Framework proposal
06	F	Label / grouping governance alignment (Đ24/Đ29)
07	G	Pivot governance alignment (Đ26)
08	H	DOT governance alignment (Đ35)
09	I	Phantom / orphan governance alignment (Đ31/Đ19) + law-patch outline
10	J	Approval lifecycle model (Đ32)
11	K	Gap matrix
12	L	Decision options + recommendation
13	M	Go / no-go list
14	—	Law / design patch outline
15	—	Next prompts (≥6 paste-ready)
16	—	Self-review vs acceptance criteria

Forbidden-compliance statement

No PG mutation · no Directus mutation · no Qdrant/vector write · no Nuxt/UI implementation · no route change · no law enactment · no approval creation · no self-approval · no new governance owner created · no table created · no schema change · no event/job/notification emit · no hardcode introduced · no local-governance design offered as the final recommendation. All findings are evidence-backed from live PG + enacted law text.