[[ChatGPT Prompt Engineering for Developers|Fulford&Ng-2023]] propose two principles and six tactics for prompting ### Principle 1: Write clear and specific instructions Be aware that clear does not mean short. Longer prompts provide more clarity and a larger context for the model. - Tactic 1: Use [[use delimiters|delimiters]] - Tactic 2: Ask for [[ask for structured output]] - Tactic 3: Be sure to [[check conditions]] before executing a task - Tactic 4: Use [[few-shot prompting]] ### Principle 2: Give the model time to think - Tactic 5: Specify [[use numbered steps]] to complete the task - Tactic 6: Define the [[chain of thought prompting]] before rushing to a conclusion [[prompt engineering]] < [[Hands-on LLMs]]/[[5 Prompting]] > [[use delimiters]]