---

title: 13 — Self-Review date: 2026-05-31 final_status: PASS

13 — Self-Review

A. Acceptance criteria (§15) — self-scored

criterion	met?	evidence
master design sources read	✅	all 14 design docs + readiness report + GPT review read (docs 00–13 of design pack)
leaf-set rule explicit	✅	composition_level<>'meta' AND entity_type NOT LIKE '%_total' AND entity_type<>'all' → 160 (doc 02)
count-integrity rehearsal ran read-only	✅	E1–E3 read-only; leaf invariant 1,954,686/1,917,903; FAILED status (doc 03)
BEGIN..ROLLBACK view/fn rehearsal completed	✅	9 objects compiled, rollback clean, 0 leftover, idle_tx 0 (doc 04/12)
dynamic drill-down tested against live	✅	pivot_query('PIV-104') child layer + fn_substrate leaf; tree flat blocker (doc 05)
label/grouping threshold clarification applied	✅	50=MAX ceiling; 27/160 exceed; THRESHOLD_REGISTRY_GAP (doc 06)
pin/ghim checked/rehearsed/proposed	✅	REGISTRY_MISSING; registry_pin rehearsed valid (doc 07)
orphan/phantom/notification explicit	✅	orphan REUSE; phantom LAW_DEFINITION_GAP + unreliable def; issue/event map (doc 08)
no-hardcode tests defined + current violations inspected	✅	live gap-math violation captured; test catalogue (doc 09)
readiness verdict strict	✅	NOT_READY; staged unlocks (doc 10)
no persistent mutation	✅	exit==entry; 0 leftover; idle_tx 0 (below)

Result: PASS (all criteria met; not PARTIAL/BLOCKED).

B. Mutation proof (entry vs exit)

table	entry	exit	Δ
meta_catalog	169	169	0
pivot_definitions	37	37	0
pivot_results	126	126	0
collection_registry	168	168	0
table_registry	21	21	0
entity_species	42	42	0
species_collection_map	164	164	0
taxonomy_facets	10	10	0
taxonomy	58	58	0
label_rules	38	38	0
dot_tools	309	309	0
dot_iu_command_catalog	54	54	0
information_unit	219	219	0
iu_relation	60	60	0
v_registry_counts	169	169	0
rehearsal views/table/fn leftover	—	0/0/0	clean
idle in transaction	—	0	clean

Append-only tables (birth_registry, system_issues, event_outbox, entity_labels) were written to by none of this session's actions; any movement is from other system processes. No COMMIT was issued; the workflow_admin channel was used only inside BEGIN..ROLLBACK.

C. Forbidden compliance (§16)

No persistent PG mutation ✅ · no COMMIT ✅ · no Directus mutation ✅ · no Qdrant/vector write ✅ · no production Nuxt change ✅ · no committed schema ✅ · no route implementation ✅ · no fake approval ✅ · no frontend business logic introduced ✅ · no hardcoded counts/categories/species/layers/labels/pins ✅ (the one 50 literal lives only inside a rolled-back rehearsal view and is explicitly flagged THRESHOLD_REGISTRY_GAP) · no event/job execution ✅ · no notification emit ✅ · no self-approval ✅.

D. What this session changed about the design (honest deltas)

1. Leaf invariant is far cleaner than doc-05 implied once scoped: 3 drift rows, not 10; orphans are a rollup-row property (leaf orphan = 0).
2. The phantom operational definition (record>actual) is unreliable — direction depends on source_model. This is a sharper finding than the design's LAW_DEFINITION_GAP and must shape the law patch (doc 08, P6).
3. Hidden drift exists that no record-vs-actual check sees (entity_labels +28,403; system_issues +7,135) → pivot-vs-stored is the only valid check, and most leaves lack a pivot (PIVOT_MISSING) → the real failure is pivot coverage, not arithmetic.
4. A naive pivot JOIN re-creates the double-count — the canonical view must use scalar EXISTS. Caught only because we rehearsed, not just designed.
5. meta↔pivot needs a mapping column, not just parent_code — source_location (File:…) doesn't join cleanly.
6. The live gap-math anti-pattern is still serving and actively mislabels birth_registry as 985k orphans — a concrete, dated implementation blocker.

E. Weaknesses / open items

• Nuxt source not server-readable → static no-hardcode scan deferred to CI over the repo (P10); used built-bundle + live API instead (sufficient to prove the violation).
• LEAF_SET is not provably MECE → PIV-500 over the universal substrate is the needed cross-check (P5).
• Ghost (cross-store) detection still DEFERRED (no Qdrant read).
• Governance owner per node is heuristic (primary_collection NULL — carried from prior sessions).

F. Recommended next macro

RG1 + RG2 human ratification, then P1 REGISTRIES_PIVOT_VIEW_COMMIT_AFTER_RATIFY (commit the corrected, fan-out-free v_count_integrity/v_count_drift/v_registry_leaf_set/fn_registries_pivot_node_substrate + phantom_count/pivot_code EXTEND), in parallel with P5 (PIV-500/30x/31x to close the four-disagreeing-totals gap and the hidden drift).