Classical Assumptions
The five assumptions behind OLS properties: A1 mean-zero errors, A2 exogeneity , A3 no autocorrelation, A4 homoskedasticity, A5 normality of the errors.
When to use
Map each to what it buys: A1 and A2 give unbiasedness, A1 to A4 give Gauss-Markov Theorem (BLUE), A1 to A5 give normal for exact t and F inference. Then ask which assumption a problem breaks: Omitted Variable Bias breaks A2, Heteroskedasticity breaks A4.
See Lecture 4 and Lecture 13.