This article introduces the implementation principles of Prompt reverse engineering through multiple examples, allowing AI to return its built-in prompts. The methods include: 1) using reverse prompts to make AI ignore preset settings and return its own prompts; 2) using different reverse prompts to confuse AI and obtain the original prompts; 3) obtaining key prompt information through partial reverse engineering. Reverse engineering faces difficulties such as the influence of prompt structure and AI reinforcement. In the future, there may be prompt defense positions to ensure the safety of AI applications. Overall, Prompt reverse engineering is a new application direction in the AI field, which helps to understand the working principles of AI, but also carries the risk of being misused.