Prompt engineering questions assess a candidate’s ability to write structured instructions that guide AI models to generate accurate and reliable outputs.

You can add Prompt Engineering questions from the HackerRank Library when you create a test or interview.

How prompt engineering questions work

When a candidate writes a prompt and clicks Run prompt for the first time, the system sends the prompt, the test case input, and the surrounding context to the underlying AI model, which generates a response.

The system then performs the following actions:

• Caches the response against that exact prompt for that test case, so that it can be reused for a later run or for the submission.

• Compares the response with the sample test case’s expected output, using the same exact match logic that scoring uses.

• Displays a Pass or Fail result for each sample test case.

Running a prompt helps candidates iterate on their prompts. It does not record a final score for the question.

Scoring prompt engineering questions

Prompt engineering questions are scored automatically using preconfigured test cases. Each case compares the model's output to the expected output using exact text matching.

Scoring logic

1. When a candidate submits the test, the system sends the prompt, the test case input, and the context to the underlying AI model.

2. The system checks whether an identical prompt was already run on that question during the attempt. If a cached response exists, the system reuses it. If not, the model generates a new response.

3. The system compares the model’s output with the expected output. The comparison requires an exact match, including characters, casing, punctuation, accents, spacing, and word order. Any deviation, such as extra words, explanations, code blocks, or formatting, results in a Fail.

4. The result for each case is either Pass or Fail.

The system calculates the final score as:

Final score = Weight per test case × Test case status

Response caching

AI models can generate different responses for the same input. To provide consistent and repeatable scoring, HackerRank caches the model's response when a prompt runs against a test case.

When a cached response exists, HackerRank reuses it instead of sending the same prompt to the model again.

Response caching works as follows:

• The same prompt for the question returns the same cached response and produces the same Pass or Fail result.

• Running the same prompt again does not generate a new response from the AI model.

• Changing the prompt creates a new prompt. HackerRank sends the revised prompt to the model and caches the new response.

• Changes to wording, structure, or punctuation cause HackerRank to treat the prompt as a new prompt.

• Caching applies to the specific question.

This behavior prevents candidates from repeatedly running an unchanged prompt to receive different responses from the model.

Examples

Example 1: Consistent pass

The prompt asks the model to translate Thank you into French and to return only the translation.

• Expected output: Merci

• Model responds: Merci

• Result: Pass

If the candidate reruns the same prompt, the system returns the cached response, and the result remains a Pass.

Example 2: Consistent fail, followed by a fix

The prompt asks for the same translation, but the model returns additional text.

• Expected output: Merci

• Model responds: Sure! The translation is: Merci

• Result: Fail (extra text)

Rerunning the same prompt returns the same cached response, so the result remains a Fail. If the candidate edits the prompt to state "Return only the word, no explanation," then the system treats the prompt as a new prompt, and the model generates a new response. If the model returns Merci, the result is a Pass, and the system caches the new response for the revised prompt.

Frequently Asked Questions (FAQs)