Prompt To Eval Rubric

Prompt To Eval Rubric

Purpose

Turn desired model behavior into a reusable evaluation rubric.

Fit

• Use when prompt behavior must be evaluated repeatedly or compared across model/prompt versions.
• Do not use when the task is only rewriting the prompt and no evaluation criteria are needed.

Inputs

• Prompt, system instruction, policy, desired behavior, or bad output.
• User goals, safety requirements, and known failure modes if available.
• Preferred scoring scale if provided.

Workflow

1. Extract expected behaviors and forbidden behaviors.
2. Convert each behavior into a measurable criterion.
3. Tie each criterion to prompt evidence, policy evidence, or a clearly labeled assumption.
4. Add representative test cases, including edge cases.
5. Define pass/fail assertions and scoring guidance.
6. Preserve uncertainty where behavior is underspecified.

Output

Produce Markdown with:

• Eval Goal
• Rubric Table
• Test Cases
• Assertions
• Failure Modes
• Scoring Notes
• Missing Requirements

Validation

• Criteria are observable in outputs.
• Assertions can be judged consistently.
• Safety and correctness are not traded away for style.
• Ambiguous prompt requirements become missing requirements.
• Test cases include at least one likely failure case.