Components of Prompts

A well-structured prompt consists of several essential elements that influence the quality and accuracy of the response. The most important components are explained below and supplemented with examples.

Basic principles

Not all components need to be present in every prompt. Each component serves a purpose and refines the result in a certain direction, but its importance varies depending on the use case.

Every prompt for more complex tasks could consist of at least the following components:

1. Instruction
2. Context
3. Input
4. Output format
5. Persona

1. Instruction

The instruction describes what the language model should do or answer. It is often the core of a prompt.

Example:
Write a poem about summer.

2. Context

The context provides additional information or background to guide the model’s response in a more targeted manner.

Example:
You are a poetry slam participant and want to wow the audience.

3. Input

Input is specific data, questions, or text to which the response should relate.

Example:
Text: The sun is shining, the sea is sparkling.

4. Output format

The output format describes how the response should look (e.g., as a list, table, continuous text).

Example:
Output the answer as a table with two columns (What happens? Why is it important?).

5. Persona

The persona determines the perspective or role from which the answer should be given (e.g., as an expert, teacher, child).

Example:
You are a computer science professor and explain the term “cloud computing.”

6. Constraints

Constraints are specifications that limit the answer in terms of length, style, or content.

Example:
Answer in a maximum of three sentences and do not use technical terms.

7. Exemplary approach (Examples)

Putting examples at the beginning of the prompt shows the model what the answer might look like.

Example:
Example:
Question: What is AI?
Answer: Artificial intelligence is the ability of machines to solve tasks intelligently.

8. Tools

Sometimes the prompt may contain information about external tools or functions that should be used.

Example:
Use the Wikipedia API to research current data.

Composition of an effective prompt

A typical prompt can consist of a targeted combination of the above elements:

You are an environmental scientist (persona).
Explain the greenhouse effect (instruction), taking into account current CO2 emissions (context).
Answer in a maximum of five sentences (restriction).
Provide the answer as a numbered list (output format).

Each component contributes to creating clarity and precision in prompt engineering. By consciously integrating and combining these elements, interaction with AI models can be made more precise and effective.