What is the general workflow for processing a dataset during evaluation?	Test Case is merged with the Prompt, fed into Claude, and the result is passed through a Grader.
What is the primary function of the `run prompt` function?	To merge a test case with the prompt, generate text using Claude, and return the resulting output.
What is currently lacking in the initial prompt design?	Formatting instructions (e.g., requiring Python, JSON, or regular expressions) to ensure structured output.
What is the purpose of the `run test case` function?	To take a single test case, call `run prompt`, grade the resulting output, and return a summary dictionary.
How does the `run eval` function operate?	It loads the entire dataset and loops through every test case, calling `run test case` for each one.
What is the final output structure returned by the `run eval` function?	A large JSON object (an array of objects), where each object summarizes the output of one individual test case.
What is the final, yet-to-be-implemented step in the evaluation pipeline?	Feeding the input and Claude's generated result into a Grader.