05 — Supertrack E: RP-04 catalog-family exact-set coverage review

Verdict: RP04_VERIFIED.

Codex delivered (doc 01): reference_contract, operand_column_contract, and structural_literal_class as exact-set families inside the sealed code-catalog root — no new authority surface.

Checklist

Criterion	Result	Basis
reference_contract covered	✅	discovered catalog-typed FK columns vs reference_contract: both EXCEPT, exactly one row each
operand_column_contract covered	✅	ACTIVE operand types vs operand_column_contract: both EXCEPT, exactly one physical column each
structural_literal_class covered	✅	numeric literals in every hash-bound adapter/function source vs structural_literal_class: every literal classified; unclassified → readiness FAIL
all counted / sealed / exact-set	✅	three families inside the single already-counted/sealed code-catalog root; count data-driven by code_catalog_set.expected_family_count (16→18, no hardcoded count)
every catalog-typed FK column has a reference_contract row	✅	both-EXCEPT by (table, column); missing row → SEAL FAIL
every operand type has an operand_column_contract row	✅	both-EXCEPT; union of mapped columns == fixed operand column set; missing/extra → SEAL FAIL
every adapter numeric literal has SA15 classification	✅	SA15 enumerates every literal in every hash-bound source; unclassified → readiness FAIL
missing row fails	✅	seal/readiness fail-closed across all three families

Anti-hardcode note (also feeds Supertrack K)

The structural_literal_class family is the AUTHORITY; the source scan is a fail-closed DETECTOR (every literal must map to a sealed class; unclassified → FAIL). This is detector-not-authority — the same accepted pattern as CP-05/SA15 and the FIX5 structural callgraph — not "source-text as authority." The family count being data-driven (expected_family_count) keeps the 16→18 move from being a hardcoded count.

Conclusion

RP-04 fully resolved: three families in the catalog root, exact-set coverage at seal, fail-closed on any unmapped column / operand type / unclassified literal. No proposal. This makes the anti-hardcode root provably complete (closes the contract-coverage hole in the cross-impact map).