The growing number of online accessible services call for effective techniques to support users in discovering, selecting, and aggregating services. We present WS-Advisor, a framework for enabling users to capture and share task memories. A task memory represents knowledge (e.g., context and user rating) about services selection history for a given task. WS-Advisor provides a declarative language that allows users to share task definitions and task memories with other users and communities. The service selection component of this framework enables a user agent to improve its service selection recommendations by leveraging task memories of other user agents with which the user share tasks in addition to the local task memories.