This study proposes an instructional and system design for enhancing children’s career planning abilities through community-based experiential learning. Although the Japanese government highlights Career Planning Ability as one of the four Basic and Generic Abilities, prior research—including the authors’ Learning Feedback System (LeaFeS)—showed limited improvement in this domain, despite gains in problem-solving abilities. A key issue was the lack of timely and developmentally appropriate reflection support during short activity cycles. To address this challenge, the present study integrates (1) a four-quadrant job model, (2) explicit behavioral goal setting through Challenge Cards, and (3) instant reflective feedback generated by a fully offline Local Large Language Model (Local LLM). The system was deployed in Children’s City Yotsukaido 2025, where children selected short-term jobs, set goals, and received three-sentence feedback immediately after each activity. A two-day subject experiment (N = 189) compared conditions with and without the system. Statistical analyses—including paired and independent t-tests and mixed ANOVA—showed that the proposed design significantly enhanced goal awareness and execution awareness, core components of career planning abilities. These findings demonstrate both the educational effectiveness of the integrated learning design and the technical feasibility of offline Local LLM–based feedback in a real community event.