Normality and distribution checks

what "assumes normality" actually refers to, and the fact that it is usually the residuals and not the raw variable; histogram, Q-Q plot and box plot read together; Shapiro-Wilk and Kolmogorov-Smirnov, and why both become useless at large N; skewness and kurtosis thresholds; the central limit theorem and when it rescues you; transformation as an option and its cost to interpretation; the decision rule for parametric vs non-parametric.