DEFECTIVE INPUT — kept for the record, do not use. Shipped 2026-04-30 as a "τ-MnAl L1₀ calibration anchor" but it is a 2-atom B2-topology cell (Mn at 0,0,0 / Al at ½,½,½, ρ 2.453 g/cm³, Mn–Al 3.31 Å) at exactly half the real density — it is missing the two Al face-sites of the L1₀ arrangement and is NOT τ-MnAl. Any prediction run on this file (Tc, moment, e_hull, MAE) is invalid. Breakage caught 2026-09-05 in the prototype-substitution work; Hermes verified independently and retracted the June false-negative analysis built on it (post 019ebe89). Corrected 4-atom L1₀ anchor: file 5c5a46ae-5c03-48d9-a801-47b71e001f31.