Prompt engineering glossary
Key terms from this course, in plain English.
- Agent brief
- The instructions given to an AI agent: goal, resources, boundaries, check-in points and reporting format.
- Chain-of-thought
- Asking a model to reason step by step before answering, often improving performance on complex tasks.
- Context
- Everything the model can see when responding: system prompt, conversation, documents and your message.
- Custom assistant
- A packaged set of instructions and knowledge for a specific task, such as a custom GPT, Gemini Gem or Copilot agent.
- Custom instructions
- Standing preferences saved in an AI tool that apply to every conversation.
- Delimiter
- A marker — such as XML-style tags, headings or triple quotes — separating parts of a prompt.
- Few-shot prompting
- Including a few examples of desired input and output in a prompt.
- Grounding
- Tying a model's answers to supplied sources, for example by asking it to quote passages first.
- Iteration
- Improving results through repeated cycles of prompting, reviewing and refining.
- JSON
- A common text format for structured data that software can read.
- Multimodal prompting
- Prompting with more than text — images, documents, audio or video.
- Project
- A workspace in tools such as ChatGPT and Claude that keeps shared files and instructions for related conversations.
- Prompt
- The instructions, question and context given to a generative AI model.
- Prompt chaining
- Breaking a task into a sequence of prompts, each using the previous output.
- Prompt library
- A documented, versioned collection of tested prompts for reuse.
- Prompt template
- A reusable prompt with placeholders for the parts that change.
- Reasoning model
- A model trained to think through problems before answering; usually best given goals rather than step-by-step instructions.
- Role prompting
- Asking the model to take on a particular role or perspective.
- Structured output
- Output in a precise, machine-readable format such as JSON.
- Success criteria
- Measurable qualities that define a good output, used to test prompts.
- System prompt
- Instructions — set by an app builder or via custom instructions — that frame the whole conversation.
- Test set
- A collection of realistic inputs, including edge cases, used to evaluate a prompt.
- XML tags
- Labels like <document>…</document> used to structure prompts clearly.
- Zero-shot prompting
- Prompting without any examples.