What is the primary goal of the custom prompt in this workflow?	To help users write code specific to AWS use cases.
What are the three specific types of output the prompt must be able to generate?	Python code, JSON configuration, or a raw regular expression.
What constraint must the prompt adhere to regarding its output format?	It must provide the output without any explanation, header, or footer.
How is a data set defined in the context of prompt evaluation?	A collection of inputs that are fed into the prompt to test its performance.
What is the required structure of the inputs within the data set?	An array of JSON objects, where each object contains a 'task' property.
What are the two methods for assembling a data set?	Manually or automatically using Claude.
Which model is recommended for automatically generating the data set using Claude?	Haiku.
In the code implementation, what is the role of the stop sequence?	To ensure the model outputs only the desired JSON format and stops generating text prematurely.
What is the final step after the evaluation data set has been generated?	Writing the data set to a file (e.g., `dataset.json`) for later use.