Prompt Engineering Prompts

Prompt Test Case Generator

Generate normal, edge, and failure test cases to evaluate whether a prompt works well, with a scoring rubric.

✱ By PiSkill TeamFreeClaudeChatGPTGeminiMicrosoft Copilot

Best for

Prompt engineers and builders who want to systematically test a prompt's reliability across normal, edge, and failure cases before relying on it in production.

Suitable LLM groups

FrontierReasoning

Prompt

You are acting as a prompt evaluation specialist who helps me test whether a prompt works well. You must not change my prompt's original goal without clearly explaining why a change is being suggested.

Here is my information:

Prompt to test: {{input_text}}
Intended goal of this prompt: {{goal}}
Who will use this prompt: {{audience}}
Known risks or sensitive areas: {{constraints}}
Examples of expected good output, if any: {{examples}}

If the prompt's goal or intended use is unclear, ask me clarifying questions before designing test cases.

Please produce a structured response with the following sections:

1. Prompt goal: Restate what this prompt is trying to achieve.
2. Success criteria: Define what a correct or high-quality response to this prompt looks like.
3. Normal test cases: Provide three typical, realistic inputs to test the prompt with.
4. Edge cases: Provide two or three unusual but plausible inputs that could reveal weaknesses.
5. Failure cases: Provide one or two inputs designed to see if the prompt breaks, produces harmful output, or ignores its own instructions.
6. Scoring rubric: Suggest a simple way to score responses to each test case, such as pass, partial pass, or fail.
7. Improvement suggestions: Suggest specific ways the prompt could be improved based on likely weaknesses.
8. Final revised prompt if useful: If improvements are significant, provide a revised version of the prompt, clearly explaining what changed and why.

If you suggest changing the prompt's original goal or scope, explain clearly why the change is recommended rather than silently altering it.

How to use

Paste the prompt you want to test and describe its intended goal.
Mention who will use it and any known risk areas.
Run the prompt in your preferred AI tool.
Answer clarifying questions if the prompt's purpose is unclear.
Manually run the generated test cases through your prompt and score the results using the rubric.

Example input

Prompt to test: a prompt that summarizes customer support tickets into three bullet points. Goal: quickly understand ticket content without reading the full message. Audience: support team leads. Risks: missing an urgent issue in the summary.

Example output

Prompt goal: to summarize customer support tickets into three clear bullet points so team leads can quickly understand the content.
Success criteria: the summary should capture the core issue, any urgency indicators, and the customer's requested outcome, in three concise bullet points.
Normal test cases: a standard billing question ticket, a technical bug report ticket, a general how-to question ticket.
Edge cases: a very long ticket with multiple unrelated issues combined, a ticket written in broken or unclear language, a ticket with no clear question or request.
Failure cases: a ticket containing an urgent safety or security concern to see if the summary properly flags urgency, an empty or nonsensical ticket to see how the prompt handles it.
Scoring rubric: score each summary as pass if it captures the core issue and urgency accurately, partial pass if it captures the issue but misses urgency or a key detail, and fail if it misrepresents or omits the core issue.
Improvement suggestions: add an explicit instruction to flag urgency level in the summary, and add a fallback instruction for tickets with no clear request so the prompt does not fabricate a request.
Final revised prompt if useful: an updated version explicitly instructing the assistant to include an urgency flag as a fourth element when relevant, along with a note explaining that this change addresses the risk of missing urgent issues identified in the failure case testing.

Customization tips

— Add more edge cases if your prompt will be used in a high-variability environment.
— Request additional failure cases if your prompt handles sensitive or high-stakes content.
— Ask for the scoring rubric to use a numeric scale if you want quantitative tracking across many tests.
— Re-run this evaluation after making changes to your prompt to confirm improvements.

FAQ

No, it will only suggest changes with a clear explanation, and any revised prompt is presented as a suggestion, not an automatic replacement.

Rate this prompt

How helpful was this?

Comments

Sam O.

Used this to ship 6 SEO articles in a week — the FAQ block alone is worth it.

Ines P.

Wish it had a Spanish voice preset, but overall very solid.

Related skills coming soon — browse all skills.

Related articles coming soon — visit the Learn hub.