Disaster Management (DM) is a complex set of interrelated activities. The activities are often knowledge intensive and time sensitive. Sharing the required knowledge timely is critical for DM. In developed countries, for recurring disasters (e.g. floods), there are dedicated document repositories of Disaster Management Plans (DMP) that can be accessed as needs arise. However, accessing the appropriate plan in a timely manner and sharing activities between plans often requires domain knowledge and intimate knowledge of the plans in the first place. In this paper, we introduce an agent-based knowledge analysis method to convert DMPs into a collection of knowledge units that can be stored into a unified repository. The repository of DM actions then enables the mixing and matching knowledge between different plans. The repository is structured as a layered abstraction according to Meta Object Facility (MOF). We use the flood management plans used by SES (State Emergency Service), an authoritative DM agency in NSW (New State Wales) State of Australia to illustrate and give a preliminary validation of the approach. It is illustrated using DMPs along the flood prone Murrumbidgee River in central NSW.