Charles Leopard enhanced the InferESG repository by delivering prompt template improvements focused on AI agent output. He refactored the backend logic using Python and Jinja to enable more precise agent selection and structured JSON responses, addressing the need for clearer user intent handling. His work introduced new fields for result and query types, expanded the range of question categories, and provided detailed examples for ESG score queries and data manipulation. These enhancements improved the quality and structure of AI-generated responses, contributing to a more informative user experience. The work demonstrated depth in backend development, natural language processing, and prompt engineering.