This document provides an adapted set of heuristic strategies for mathematical problem solving based on George Pólya’s How to Solve It (1949). Carefully reformulated for clarity and instructional use, the strategies are structured to support static context injection in large language models (LLMs). The goal is to enhance the quality of AI-generated solutions in mathematics education by embedding domain-specific reasoning patterns. The resource is suitable for prompt design and use in retrieval-augmented generation (RAG) scenarios.