The article explores the evolution of a no-code programming framework aimed at building logically structured interaction scenarios with large language models (LLMs). The basic framework includes three formally defined primitives: "Condition", "Loop", and "Function". These constructs provide a foundation for implementing simple logical operations through textual prompts. An extended framework is then introduced, enhanced by two new primitives-"Label" and "Jump". These enable the organization of non-trivial execution flows, similar to control statements in classical programming. This approach is particularly important for handling analytical tasks that require a clear structure and precise sequence management. Using specific examples, the article demonstrates how to transition from the basic framework to the extended one while maintaining logical integrity and enhancing process control flexibility. A methodology for creating the final prompt as a result of prompt engineering is also presented: verbal formulation, refinement, structuring, testing, and final adjustment. Keywords: no-code programming, prompt engineering, LLM, formalization, logical structure, procedural programming, artificial intelligence