Prompt Engineering
Rekayasa Prompt
The craft of writing instructions a model follows reliably.
Prompt engineering is writing instructions so a model does the right thing more often. In practice it is less about magic phrasing than about removing ambiguity: state the task, the constraints, the format, and what to do when the instruction is unclear.
It also covers the smaller techniques: giving examples, asking for reasoning before the answer, splitting one hard request into steps, and telling the model what not to do. Each of these works by shaping the distribution the model samples from, not by unlocking a hidden capability.
A useful discipline is to treat a prompt as code. It has inputs, outputs, and failure modes. That makes it versionable, testable, and reviewable, instead of folklore passed between people in a chat window.
Related terms
Prompt
The input you give a model. The word covers more than the question you type.
Few-shot Prompting
Showing the model two or three examples of the task instead of describing it.
Context Engineering
Deciding what goes into the model's context, and what stays out.
Evaluation
Measuring whether the system does the job, on a fixed set of cases, repeatedly.
Guardrails
Checks that stop or constrain an action before it happens.
Want this working inside your team?
It starts with one conversation about your workflow.