For the introduction of robots to a variety of workplaces, accumulation, sharing and reuse of task implementation knowledge including complex motions of robots, tool and device usage for tasks are important. Further this knowledge need to be quickly accessed and utilized through the process of introducing a robot for a task, updating existing systems to more efficient ones, and implementing similar tasks. This paper proposes a framework to store data related to tasks and enable task implementation using search, customization and combination of the stored data. The stored data include motion patterns of robots, various meta-data including texts, images, 3D models and other documents. They are managed based on task models. The meta-data are used by humans to figure out the tasks and reusability of the associated motion patterns. The primary goal of this framework is to assist the process of thinking and implementation by humans. This paper presents an analysis of typical use-cases and system design of the framework, and implementation on Choreonoid. Then an experiment is presented, where a sequence of assembling a complex mechanical unit is described using the proposed framework. We also discuss issues on the current framework specifications and possible cooperation with automation technologies and external data sources.