Null and research hypothesis

How I can write null and research hypotheses? How I can know one-tailed and two-tailed? when to reject the null based on the probability (p-value or significance level) of doing so mistakenly;?