---

title: 06 — No-Hardcode Doctrine & Audit date: 2026-05-30

06 — No-Hardcode Doctrine & Audit

Banned patterns

1. Hardcoded counts in Vue/JS/server. 2. Hardcoded species/category/tier/layer arrays in frontend. 3. Handwritten per-entity count(*) in Nuxt/API. 4. Static groups that should come from PG. 5. UNION ALL per species/category/list. 6. Route-specific business logic. 7. Frontend deciding governance/category/status. 8. meta_catalog.record_count as truth without pivot reconciliation. 9. Fake "config" arrays that are really hardcode.

Allowed pattern

PG pivot_definitions define dimensions → PG functions/views assemble the tree → Directus/API exposes → Nuxt renders only. Adding a list = INSERT/UPDATE a data row, not code.

Audit findings

#	Location	Finding	Banned	Severity
H1	registries.vue	Phantom literal rows CAT-SPE/ORP/PHA/UNM/017	#2,#9	HIGH
H2	/api/registry/health	gap = record−actual computed Nuxt-side	#3,#7	HIGH
H3	registries.vue	Cards read record_count as truth (DOT 307 vs 309; species 178,030 vs 42; KG 2,199 vs 2,259)	#8	HIGH
H4	pivot.vue rows 8–9	DOT(307)+Collections(168) from meta_catalog, DOT stale	#8	MED
H5	engine	multiple active pivots over same source_object → must pick canonical total, never sum groups	(engine)	MED

Disguised-hardcode test

Passes only if every displayed number resolves to a pivot_code and every grouped list is a PG row set, proven by: T1 grep (no count literals, no category/species/layer arrays, no phantom rows); T3 API has no aggregation SQL; T-living a list can be added via one pivot_definitions INSERT with no code deploy.

Verdict

/knowledge/registries FAILS (H1–H3). /knowledge/pivot largely passes but H4 must be fixed. Target surface designed to pass all checks by construction (doc 07/08/10).