# Argument selection The named quantity is the primitive uncompleted Hasse–Weil central value L(E,1), in arithmetic normalization with center 1, for rational elliptic curves of algebraic rank zero. The ordered keys N,c4,c6 are exact integers. They jointly identify a reduced integral global minimal Weierstrass model, not three independently varied arguments. No continuous coupling, decimal mesh, branch choice or statistical threshold is present. N is the positive conductor. c4 and c6 are the integral minimal-model invariants, with c4^3-c6^2=1728*Delta and Delta nonzero. They are dependent discrete curve identifiers; neither c4 nor c6 has an independent numerical cutoff. The differential is dx/(2*y+a1*x+a3), with the five exact coefficients retained in roster.json and the curve crosswalk. All source model changes and algebraic rank bounds will be checked before any values are uploaded. The natural domain is all such rank-zero curves over Q. The finite default selection is curve number 1 from every rank-zero isogeny class with N<=100 in the frozen ecdata commit 25cec5ecfec8b9f016eb1631ac633194c2bed39f, allgens/allgens.00000-09999. Its source roster has 284 individual curves and 76 class representatives. The separate alllabels file at the same commit supplies an explicit Cremona/LMFDB crosswalk. L-functions are isogeny invariant; the representative is a source convention, not a mathematical distinction in the central value. Every individual curve is retained in the attachment for joining to the separate primitive-period family. This is broad bounded reference coverage by an arithmetic complexity measure. The noncanonical practical cutoff 100 matches the existing Faltings-height comparison population and includes every source class within that bound, with ordinary controls and both signs of discriminant. No selection is made by predicted Sha size or numerical attractiveness. The cutoff is independent of numeral spelling and is not a completeness claim about the infinite family. Taking all classes within the bound provides useful breadth before extending conductor depth. Repeating the same central value for all 284 source models is unnecessary here; the crosswalk preserves that relation without merging periods into the table. The three parameter selections should be assessed as genuine dependent discrete indices with a practical finite reference cutoff. If a critic instead classifies the finite population as reference_sample, its policy must explicitly acknowledge that noncanonical conductor budget and joint arithmetic selection. A Cartesian product of selected N,c4,c6 values is never intended. The generator, Programs explanation and complete-note must retain exactly the 76 representative triples. complete is no. Checked source locations: Cremona's ecdata page, Tables One and Two (model, rank, generator and torsion semantics); frozen allgens and alllabels source bytes; Sage L-series documentation, at1 (central normalization and split Mellin series); PARI elliptic-curve documentation, ellan, ellap, ellminimalmodel, ellrank and ellrootno. The source's rank is an input claim until separately checked by descent. A numerical central enclosure excluding zero is analytic nonvanishing, not a computation of Sha. Skill evidence: ../20261120T044104Z-build-skills/manifest.json, status ready; identical local/live bundle SHA-256 b5c3c7985b7c9e178b4899d083a9e3bc4284989dff76f05e5b0e52835d47eed2. No independent argument-review verdict is issued by this builder.