Automated testing is well established in software engineering but rarely supported in game design tools. This paper presents an automated testing module for a visual scripting tool used to design narrative-driven point-and-click games, combining model-based testing using Extended Finite State Machines (EFSMs) with LLM-based analysis of playthrough scripts. We evaluate the system on games produced in a game design course, demonstrating that complete EFSM traversal was achieved for 15 out of 23 games within practical computational bounds. LLM-based batch queries, such as automated age rating estimation, show potential for helping teachers rapidly characterise student work. The results indicate that even simple automated testing can reduce the mechanical burden on teachers, freeing supervision time for higher-level design feedback. Ongoing work is focused on evaluating broader integration into educational practice.