This chapter presents advanced techniques for creating effective prompts. It first describes the so-called zero-shot, one-shot, and few-shot prompts, which guide the LLM toward solving the actual problem by providing varying numbers of examples. The cognitive verifier pattern is then explained, in which the LLM independently queries essential information required to solve the task. This is followed by chain-of-thought prompts, where the solution is developed step by step together with the LLM, and the multi-role analysis, in which the LLM assumes different roles and thus approaches the problem from various perspectives.