How it works
A prompt is the input you give to a generative AI model. It can be a simple question, like "summarize this text", or a detailed instruction with context, desired format and constraints. The model interprets the prompt and generates a response based on what it learned during training.
Prompts can be short or long. A short prompt works for simple tasks. A long prompt, with examples and clear rules, improves quality on complex tasks. The practice of writing good prompts is called prompt engineering.
Beyond user text, there is the system prompt, which defines the model's behavior before the conversation starts. Developers use system prompts to create specialized assistants.
Prompts can also be chained: the output of one prompt feeds the next, creating more sophisticated workflows. In multimodal systems, prompts include images, audio and other data in addition to text.
Practical example
Instead of asking "write an email", a better prompt would be: "Write a short email to a client who is late on payment, with a professional and cordial tone, offering 5 extra days. Maximum 80 words." The second prompt makes the goal, tone, deadline and size limit clear.
What it's for
Prompts are the entry point for any use of generative AI. Knowing how to write good prompts increases response quality and reduces rework. In AI agents, the prompt defines the task, the available tools and the safety rules.
In bridgecode, each agent receives a prompt that describes its role, such as plan, build or test. Adjusting that prompt is the main way to control the agent's behavior.