# Judgment recipe, job 1138 Read488's next-step requirement/failure condition and the report's total-expectation, covariance and total-variation derivations. Check that the sensitivity inequality is an upper bound and identifies no sign or nonzero lower effect. Inspect the selected Lakens publisher sections in prior-art1138.md to distinguish a justified effect of interest from a true-effect prediction. Reuse469's finite pilot and488's design/analogue numbers; do not run their programs or regenerate the census. Check the conditional/constructed-instance scope qualifications explicitly. Scientific compute0; judgment ten minutes. No checker/target/stdout or numerical guarantee is submitted. This establishes the precise missing design premise, not impossibility of sampling or arithmetic transfer.