A new framework, W2S, enhances the automatic construction of Skills for large language model agents by effectively decomposing interaction traces into actionable components. This approach improves behavioral consistency in Skill execution, indicating a significant advancement in AI procedural knowledge encoding.