Opus Post-Discovery Review — P3D Pack 1 Phase 4B Discovery PASS

Date: 2026-05-11 Author: Opus 4.7 For: GPT-5.5 Thinking / Hội đồng AI — thẩm tra + quyết định species/composition/governance Run: p3d-phase4b-discovery-20260511-092233 Report: ...reports/p3d-pack1-phase4b-species-composition-discovery-report.md

---

1. Verdict: Discovery PASS — evidence complete, decisions pending

Agent gathered all required evidence. No mutations. No script aborts. Column detection worked correctly (caught parent_id not parent_ref, caught missing facet_code, caught missing composition_level in species_collection_map). Law crosswalk populated for all 12 compliance matrix items.

---

2. Top 5 discoveries that change the design picture

2.1 — species_collection_map has NO composition_level column

fn_birth_registry_auto JOINs species_collection_map → entity_species to get composition_level. Implication: choosing a species automatically determines composition_level via entity_species.composition_level. They're not independent decisions — species IS the composition decision.

2.2 — SPE-GOV (governance_infra) already exists as gom species

SPE-GOV: species_code=governance_infra, composition_level=atom, management_mode=observed. Already maps 15 governance-plumbing collections (birth_registry, species_collection_map, collection_registry itself, etc.). IU could join this group while observed — immediate Điều 29 compliance with zero new species creation.

2.3 — SPE-LAW already exists for the law domain

SPE-LAW: species_code=law, composition_level=compound, management_mode=governed, prefix=LAW. Maps to normative_registry. This is for the law documents themselves (Điều 35, Điều 28, etc.). IU (text-as-code representation of law content) is related but not identical. Whether IU uses SPE-LAW or a new species depends on whether IU is "law content" or "governance infrastructure."

2.4 — TAC nesting is real structural containment

83/86 TAC logical_units have parent_id (not NULL). Max depth = 2 (3 levels: root→child→grandchild). Children are distinct entities with own UUIDs and section_types (heading→principle, heading→paragraph, etc.). This is NOT just ordering — it's structural containment per Điều 0-B. If migration preserves this → IU is NOT atom.

2.5 — Species tree is flat in practice

All 40 species have parent_id=NULL, depth=1. The 3-level tree design from v1.2 is not materialized. No need to worry about tree positioning for new species — just use depth=1 like everything else.

---

3. Agent compliance

Check	Result
No mutations	✅
Column detection before schema-sensitive queries	✅ — all 17 targets detected or correctly skipped
Graceful degradation (no cascade abort)	✅ — 0 failed sub-queries
detect_col return 0 fix (rev6)	✅ — no script abort from missing optional columns
Counts as snapshots	✅
Candidates labelled candidate_not_approved	✅
Law crosswalk completed	✅ — 12/12 compliance matrix items assessed
fn_birth_registry_auto logic documented	✅ — source captured, JOIN logic verified
No species/composition decision locked	✅

---

4. Decision framework for GPT/User

Three interconnected decisions, with discovery evidence for each:

Decision 1: Governance role (drives everything else)

Option	What it means	Evidence
Keep observed	IU stays pilot. Use gom species (SPE-GOV). Reduced inspection.	67 other collections are observed. IU has 12 pilot rows.
Promote to governed	Full governance. Need dedicated species. QT-005 procedure. Full inspect pipeline.	34 collections are governed. 6 law-domain collections already governed with species.

Decision 2: Species (follows from governance role)

If observed	If governed
Map to SPE-GOV (governance_infra, atom). Immediate compliance. 15 collections already there.	Create new species OR map to SPE-LAW (law, compound). SPE-LAW already governs normative_registry.
No new entity_species row needed. Just INSERT species_collection_map.	If new: decide species_code, display_name, composition_level, prefix. If SPE-LAW: verify IU fits "law" semantics.

Decision 3: Composition level (follows from species choice + TAC nesting decision)

Scenario	Composition	Reasoning
SPE-GOV (observed, gom)	atom (from entity_species)	SPE-GOV is atom. Auto-resolved by fn_birth_registry_auto JOIN.
SPE-LAW (governed)	compound (from entity_species)	SPE-LAW is compound. If IU joins SPE-LAW, composition=compound automatically.
New species	Must decide based on Điều 0-B	If Phase 5 preserves TAC nesting → molecule or compound. If flat migration → atom.

The composition question IS the species question (discovery 2.1). Choose species → composition follows.

---

5. Opus recommendation (staged approach)

Stage 1 (immediate, Phase 4C): Map IU/UV to SPE-GOV (governance_infra) while observed.

• Restores Điều 29 compliance immediately (species mapping exists, birth auto-fills).
• QT-001 backfill for 12 birth_registry rows (species_code + composition_level auto-filled from SPE-GOV → atom).
• No new species creation. No governance promotion. Minimal scope.
• Cost: IU temporarily classified as "governance infrastructure atom" — technically correct (IU tables ARE governance infrastructure during pilot).

Stage 2 (when Phase 5 migration design is locked): Decide whether IU is promoted to governed.

• If promoted: create dedicated species with correct composition_level based on TAC nesting decision.
• Execute QT-005 (promotion: new species → re-map → backfill → update role → verify).
• If not promoted: SPE-GOV stays, IU remains observed governance infrastructure.

Why staged: Blocking Phase 4C/5 until the containment question is fully resolved would delay migration indefinitely. SPE-GOV gom is law-compliant NOW, and promotion can happen cleanly later via QT-005.

---

6. Updated migration gates

vocab_gate                   = OPEN (14 keys, planner plan_ok)
species_mapping_gate         = BLOCKED → OPEN after Phase 4C (SPE-GOV mapping)
composition_decision_gate    = BLOCKED → OPEN after Phase 4C (auto-resolved by SPE-GOV)
governance_promotion_gate    = DEFERRED to Stage 2 (post-Phase 5 design)
birth_backfill_gate          = BLOCKED → OPEN after Phase 4C (QT-001 for 12 rows)
legal_compliance_gate        = BLOCKED → OPEN after Phase 4C (Điều 29 species mapping restored)
label_gate                   = DEFERRED (DOT enrichment, post-migration)
edge_gate                    = DEFERRED (DOT enrichment, post-migration)
migration_allowed            = false → reassess after Phase 4C

---

7. Summary for GPT

Discovery PASS. Evidence complete. Key insight: species determines composition (no independent composition decision needed). SPE-GOV is available as immediate gom species for IU while observed. TAC nesting is real but containment decision can be deferred to Phase 5 migration design. Staged approach (gom now, promote later) is both law-compliant and pragmatic.

phase4b_discovery=PASS
evidence_complete=true
agent_compliance=FULL
species_decision_locked=false
composition_decision_locked=false
opus_recommendation=STAGED_SPE_GOV_GOM_THEN_QT005_LATER
next_action=GPT_USER_DECIDE_SPECIES_GOVERNANCE_THEN_PHASE4C