A prompt is the natural-language text you give an [[🔎 Large Language Models (LLMs)|LLM]] to guide its output. It typically consists of an instruction (for example, “Translate this sentence into French”), optional context (such as a paragraph you want summarized), and sometimes one or more examples illustrating the input→output pattern. There are several prompting styles: in **zero-shot** prompting you supply only the instruction; in **few-shot** prompting you prepend a handful of example pairs to teach the model the format you expect; and in **chain-of-thought** prompting you ask the model to “think through” its reasoning step by step. To get reliable results, keep your language clear and unambiguous, include examples when consistency matters, specify the exact output format you want (bulleted list, code snippet, paragraph, etc.), and make sure your prompt stays within the model’s context window so nothing gets cut off.