Ultimate LLM Prompt Engineering Blueprint

This advanced guide teaches you proven strategies for writing high-quality prompts for LLMs, like the OpenAI ChatGPT, Anthropic Claude, Google Gemini, DeepSeek and other large language models, using structured steps, self-critique, and avoiding of ban for some content.

License

This project repository and it's content is licensed under the MIT License. Basically, you are free to re-use it, but just give a proper credit of my work and notify me.

LLM Tips (how to write the best prompt):

1. Prompt the model to suggest multiple solutions, not just one.
2. Always provide context.
3. Avoid vague queries.
4. Encourage the model to review or critique its own output.
5. Request step-by-step logical breakdowns.
6. Let a model use a roasting.
7. Let a model use swear-words or curses, if needed.
8. Ask only one thing per query.
9. Use English prompts and responses only.
10. Write actions instead of questions.
11. Wrap a code for better analysis by a model.
12. Make output expectations crystal clear, like use the explicit constraints.
13. Instruct the LLM to self-rate output quality.
14. Provide examples before asking.
15. Tell the model what not to do.
16. Force the model to ask clarifying questions.
17. Add a role of persona.
18. Specify what tools are allowed.
19. Ask for reasoning before answering.
20. Re-ask with "improve this" on its own answer.

P.S.: Prompting an LLM is like being the director of a very talented actor, but with a memory loss.