When the traditional "one size fits all" approach is used in designing educational games, the game context is usually arranged in a fixed sequence. However, the designated content may not effectively support the diversity of players. The player's ability and characteristics should be considered and supported with an appropriate learning context embedded in the game to facilitate personalised experiences. Adapting game scenarios to a player's characteristics can boost motivation and ultimately improve learning outcomes. This research applies a context-aware design approach and the Learner-Centered Design approach to establish a personalised adaptation framework for designing educational serious games and enhancing personalised knowledge delivery. The proposed framework decouples the game logic implementation and adaptation mechanism. It dynamically adapts the designed game objects and activities to personal learning objectives, learning levels and learning progress to achieve a non-linear learning sequence. Through synchronous real-time xAPI message exchange mechanisms, system components and learning content adaptation are enabled. The adaptation aims to fit personal learning objectives and provide a non-linear learning sequence in a game environment. The framework provides students with personalised learning experiences. A game named GhostCoder is implemented and used to evaluate the framework. Based on the externalised adaptive mechanism, the game content is adapted to the player's performance by adjusting the difficulty of the learning content within the game. Testing of the game in the lab environment has been performed. At the next stage, an evaluation will be conducted with the target groups of students.