We are building a reactive planning agent (REACT-P) to automate the operadon of a the communications link in NASA’s Deep Space Network (DSN). We desire agent capable of executing plans and reactively replanning when necessary. Since the reactive planning agent must be built to interact with legacy (i.e., unmodifiable) systems, there are a number of significant challenges that must be addressed: (1) what extent and how should the legacy systems be reverse-engineered and modeled in order to support the development of a reactive planning agent? (2) Since the legacy systems were built to support human interactive control, most of the information needed for reactive planning is located in the human computer interface. How do we identify and acquire the knowledgelements needed for reactive planning? To address the fast issue we are using a scenario-based knowledge acquisition methodology and subsequently building models of the legacy systems using a simulation authoring toolkit named RIDES (Munro, 1993). The reactive planner is being built in Soar (Laird et al., 1987), which will be integrated and tested, and refmed with the RIDES-based simulator.