Imagine a world where writing unit tests is no longer a tedious chore, but an automated process. CasModaTest, a cutting-edge framework, is making this a reality through a cascaded and model-agnostic approach to unit test generation. Traditional methods of crafting unit tests can be time-consuming and often produce tests that are difficult to interpret. Existing ML-based methods have shown promise but still suffer from limitations, often generating only parts of a test or producing code that mismatches intended functionality. Enter CasModaTest, which tackles these challenges head-on. This novel framework breaks down the test generation process into two key stages: creating the test prefix, which sets up the inputs and conditions, and generating the test oracle, which defines the expected outcome. By separating these phases, CasModaTest ensures a more coherent and accurate test generation process. To further enhance its performance, CasModaTest utilizes meticulously crafted demo pools, containing high-quality test prefixes and oracles as examples. This innovative use of few-shot learning allows CasModaTest to produce highly effective tests. Finally, CasModaTest intelligently assembles the generated test prefixes and oracles, and then compiles and executes them to validate functionality. It even includes a self-debugging process to fix compilation or execution errors. This comprehensive approach leads to a dramatic improvement in accuracy and focal method coverage compared to state-of-the-art methods, demonstrating the potential of CasModaTest to reshape the world of software testing. CasModaTest's impressive performance is not limited to a single language model. It achieves significant improvements with various LLMs, including open-source models, making it a practical and adaptable solution for a wider range of developers and projects. By automating the creation of robust and comprehensible unit tests, CasModaTest frees up developers to focus on what truly matters: crafting high-quality code.