A prompt is the instruction or the input one gives to an AI system to tell it what we want it to do. For example, if I want, let's say Claude to implement a function for me to send emails, then I would input something like "Create a C# method to send emails". That is a prompt. AI takes that instruction along with the context available to it ( in the form of open files, attachments you attached, the code you have opened, etc.) and generates a response with that input.

 

A prompt is not necessarily a complete description of the software you want to build, because it tells AI what to do, but doesn't tell what business rules are applied, what are valid inputs, what are the edge cases and the required system responses, whether any constraints exist, or how best to fit a feature into existing architecture. I am not saying the code will not work or not compile, but the point is that it may not implement what you actually intended and how you actually wanted. My understanding is that, as AI becomes more involved in the software development process, we need to think beyond just the prompts. The next question would be, "What information does AI need to build the right thing?" and that's where the specifications jump into the scene.

 

Enough to cover a prompt in AI, next tip would be around specifications - get ready.