antecedent-validate

Effect refuters and validation diagnostics: placebo treatment, random common cause, dummy outcome, bootstrap and data-subset refuters, unobserved-common-cause and E-value sensitivity (linear through nonparametric, including Riesz-based checks for the binary ATE), overlap diagnostics and rules, discovery stability validators, Bayesian predictive checks, and composable refutation suites.

Refuters report applicability explicitly; a validator that does not apply to an estimator returns NotApplicable rather than a vacuous pass.