An outlier is an observation very far from the others, either through an entry or measurement error, or through a rare real phenomenon. You detect it with business rules, the interquartile range method or charts, then decide whether to fix, exclude or keep it. No automatic removal is justified without analysis.
Example
q1, q3 = s.quantile([.25, .75])
limite = q3 + 1.5 * (q3 - q1)