The intrinsic flexibility and dynamism of service-centric applications preclude their pre-release validation and demand for suitable probes to monitor their behavior at run-time. Probes must be suitably activated and deactivated according to the context in which the application is executed, but also according to the confidence we get on its quality. The paper supports the idea that significant data may come from very different sources and probes must be able to accommodate all of them. The paper presents: (1) an approach to specify monitoring directives, called monitoring rules, and weave them dynamically into the process they belong to; (2) a proxy-based solution to support the dynamic selection and execution of monitoring rules at run-time; (3) a user-oriented language to integrate data acquisition and analysis into monitoring rules.