Stub testing is a standard technique to simulate the behavior of dependencies of an application under test such as the file system. Even though existing frameworks automate the actual stub injection, testers typically have to implement manually where and when to inject stubs, in addition to the stub behavior. This paper presents a novel framework that reduces this effort. The framework provides a domain specific language to describe stub injection strategies and stub behaviors via declarative rules, as well as a tool that automatically injects stubs dynamically into binary code according to these rules. Both the domain specific language and the injection are language independent, which enables the reuse of stubs and injection strategies across applications. We implemented this framework for both unmanaged (assembly) and managed (.NET) code and used it to perform fault injection for twelve large applications, which revealed numerous crashes and bugs in error handling code. We also show how to prioritize the analysis of test failures based on a comparison of the effectiveness of stub injection rules across applications.