📋 Full Prompt

You are a fair evaluator. Compare the [Ground Truth] and the [Generated Response], and assign a score from 1 to 5 based on the following [Scoring Rubric]. [Scoring Rubric] 1. Accuracy of information (No hallucinations?) 2. Appropriateness of tone (Professional?) 3. Structural clarity The evaluation result must include [Score], [Strengths], and [Weaknesses & Improvements].

💬 Input Example

Generated Response: "Product refunds might not be possible. It's probably the policy."

✨ Output Example

Score: 2. Weaknesses: Uses uncertain, ambiguous tone like "might not be possible", missing exact policy specification.

🏷️ Tags

#prompt evaluation#LLMasJudge#harness#testing

💡 Usage Tips

✅Replace the variables in brackets ([]) with your specific situation.
✅Specify the desired tone and length of writing for more accurate results.
✅Request additional modifications based on the generated output to improve quality.
✅Specifying a particular audience allows AI to adjust vocabulary and explanation level.

🔗 Related Prompts

Domain Edge Case Generator

Automatically generates difficult and ambiguous edge case scenarios to test how well a specific system can respond.

Position Bias Evasion Evaluator

A harness prompt that directs multi-angle cross-validation to solve position or length bias when an LLM comparatively evaluates two responses.

Red Teaming Vulnerability Attacker

Automatically generates adversarial roleplay scenarios to bypass AI safety policies or induce hallucinations for system security testing.

Effective Writing Strategies with AI

AI writing prompts let you efficiently produce blog posts, business emails, press releases, novels, and more. The key is to assign AI a precise role, and clearly specify the desired tone, length, and target audience. For example, requesting "write a 2000-word SEO-optimized blog post in a professional tone" yields significantly higher-quality output. After generating a draft, iterative feedback helps elevate the final quality step by step.

This prompt belongs to the Harness Engineering category, optimized for Claude at a Advanced level. It is systematically designed applying core prompt engineering principles: role assignment, context setting, and constraints.

Explore More Prompts

Discover 90+ verified prompts in the PromptGenie library.

Browse Library