Page 105 - CyberSecurityEssentialsEnglish

Output: A prompt with an explicit language specification, and a five-case test showing it
was followed.
Assessment criterion: The test includes at least one case that mixes the two languages
and one case containing numbers or dates.

↻ Exercise 9: A small library
Goal: Turn a working prompt into an asset another person can use.
Required: Two prompts you have already tested. Allow about 40 minutes.
Steps: Give each one a name and a version. Record the task it serves. Attach at least
three evaluation cases with their expected outputs. Write the known limits and the cases
where it should not be used. Then note the date and the tool it was tested on.
Output: Two template records, each with a version, evaluation cases, and stated limits.

Assessment criterion: A colleague could run the evaluation cases and tell whether the
template still works after a tool update.

References and Selected Resources
I. Scientific and Professional References

[1] U.S. Copyright Office Review Board. (2023). Second Request for Reconsideration for
Refusal to Register Théâtre D’opéra Spatial
[2] OpenAI. (2026). Prompt engineering best practices for ChatGPT. Accessed June 6, 2026

[3] OpenAI. (2026). Best practices for prompt engineering with the OpenAI API. Accessed
June 6, 2026
[4] Anthropic. (2026). Prompt engineering overview and prompting best practices. Accessed

June 6, 2026
[5] Microsoft. (2026). Prompt engineering techniques: Microsoft Foundry. Accessed June 6,

2026
[6] Google Cloud. (2026). Overview of prompting strategies. Accessed June 6, 2026
[7] Brown, T. B., et al. (2020). Language Models are Few-Shot Learners. Advances in Neural

Information Processing Systems, 33
[8] Sharma, M., et al. (2024). Towards Understanding Sycophancy in Language Models. ICLR
2024

[9] OWASP Gen AI Security Project. (2025). LLM01: Prompt Injection
[10] OpenAI. (2026). Reasoning model best practices and prompting guide. Accessed June 6,
2026

[11] OpenAI. (2026). Structured model outputs. Accessed June 6, 2026
[12] Huang, J., et al. (2024). Large Language Models Cannot Self-Correct Reasoning Yet. ICLR
2024

[13] Liu, N. F., et al. (2024). Lost in the Middle: How Language Models Use Long Contexts.
Transactions of the Association for Computational Linguistics
[14] Sahoo, P., et al. (2024). A Systematic Survey of Prompt Engineering in Large Language

Models: Techniques and Applications

104