Prompt engineering glossary

Short definitions + a copy-paste example per term. Pick one and open it.

← Back to home

Start here

This glossary covers concepts that commonly break prompts: ambiguity, missing delimiters, inconsistent output, or hallucinations. Each term includes a short definition and a copy-ready example.

If you need strict outputs, start with strict JSON. If you want to tailor prompts to each AI, check prompts by model.

Guide: Strict JSON Prompts by model Prompt packs

The instruction layer a model treats as highest priority: its role, rules and boundaries for the whole conversation.

Few-shot prompting

Including one to five worked examples in the prompt so the model copies their format, tone and level of detail.

Markers such as triple quotes, headings or XML-style tags that separate your instructions from the text the model must work on.

Explicit, testable limits on the answer — length, tone, scope, allowed sources, forbidden claims, tech stack — that keep it useful.

A sampling setting that controls how random a model's word choices are: lower values are steadier, higher values more varied.

Fluent, confident output that is false or unsupported — invented facts, figures, quotes or sources — presented as if it were true.

Prompt injection

An attack in which instructions hidden in content the model reads — a web page, an email, a document — try to override your own.

A formal description of the shape of a JSON document — keys, types, required fields, allowed values — used to constrain and validate model output.

Assigning the model a role — editor, tutor, auditor — together with the criteria that role should apply, to set tone and expectations.

Testing a prompt on a fixed set of inputs and scoring the outputs against acceptance criteria, so improvements can be measured instead of guessed.

Zero-shot prompting

Asking a model to do a task from the instruction alone, with no examples, relying on what it already learned in training.

The maximum amount of text, counted in tokens, that a model can take in and produce in a single request.

Chain-of-thought prompting

Asking a model to work through intermediate reasoning steps before giving its final answer, which helps on multi-step problems.

Prompt chaining

Splitting a complex task into a sequence of smaller prompts, where each step's output becomes the next step's input.

Structured output

Getting a model to answer in a fixed, machine-readable format — JSON, a table, labelled fields — instead of free-form prose.

The chunks of text — whole words, word pieces or punctuation — that a model reads and writes; context limits and pricing are counted in them.

Top-p (nucleus sampling)

A sampling setting that limits the model's next-word choice to the smallest set of tokens whose probabilities add up to p.

Retrieval-augmented generation (RAG)

A pattern where relevant passages are retrieved from a knowledge source and put into the prompt so the model answers from them.

Negative prompting

Telling the model what to avoid — a tone, a claim, a format, a topic — alongside what to do, so unwanted output is ruled out explicitly.

Context engineering

Deciding what information goes into the model's context — instructions, documents, examples, history, tool results — and in what order and form.