Develop a 500-word article outlining strategies for creating highly effective prompts designed to clean messy datasets using AI agents.  The article should focus on the nuanced aspects of prompt engineering for data cleaning, covering various data types (numerical, categorical, textual) and common data issues like missing values, inconsistencies, outliers, and duplicates.  Include practical examples of prompts for each data type and issue, demonstrating how different prompt structures (e.g., few-shot learning, chain-of-thought prompting) can impact the accuracy and efficiency of the data cleaning process.  Discuss the importance of prompt iteration and refinement, along with techniques for evaluating the quality of the cleaned data. Finally, address the limitations of using AI agents for data cleaning and suggest best practices for human-in-the-loop approaches to ensure data integrity. The article should be targeted towards data scientists, analysts, and anyone working with large datasets who want to leverage AI for data cleaning tasks.