The new generation of Army command-post digital systems, e.g., Command Post of the Future (CPOF), increasingly utilizes non-linear interfaces. Non-linear interfaces use a customizable workspace that is based on the user's needs rather than a static data format, and interaction with the interface is not based on prescribed or hierarchical sequences of steps. Successful application of non-linear interfaces requires the user to decide which functions will best address a problem or need. One major difficulty in learning non-linear interfaces such as CPOF is that the interfaces do not support the organization of knowledge necessary for the individual to successfully interact with the system. This paper describes a two-part research effort to define the structured knowledge of CPOF that can serve as a basis for CPOF training. In the first part, a cognitive task analysis based on functional use of CPOF was conducted to produce a framework of CPOF knowledge. In the second part, data were collected on the retention of CPOF skills. Thirty-six participants from CPOF training at two battle command training centers completed a skills test immediately following training and again five weeks after training. The retention data were then integrated with the CPOF knowledge framework by analyzing the patterns of skill retention within the major categories of the framework. For example, the number of skills retained differed between the "Construct" category of the framework and the "Collaborate" category. The pattern of retention was then used to identify the specific skills and the progression of skills that are critical in developing CPOF expertise. The results can be used to indicate that some training techniques are more appropriate to CPOF skills than other techniques. For example, training techniques that leverage the execution of sub-goals and that illustrate overlapping CPOF procedures should most efficiently train CPOF skills.