Prompt Engineering Vocabulary

the instructions and input supplied to a model for a task, which may include text, images, files, or other content

“The prompt combined a clear objective, a reference document, and the requested output format.”

high-priority instructions supplied by an application or platform to shape model behavior; the role name, visibility, and scope vary by system

“The application used a system-level instruction to request concise, formal responses.”

providing examples in the prompt to demonstrate desired input-output patterns

“Three examples of sentiment analysis in the prompt enabled accurate few-shot classification.”

asking a model to perform a task without any examples

“Zero-shot translation works because the model learned language patterns during pre-training.”

intermediate reasoning steps generated while solving a task, whether elicited by prompting or produced internally by a reasoning model

“A reasoning model may keep its raw chain of thought hidden while returning a concise answer or reasoning summary.”

an attack or failure mode in which instructions embedded in user input or external content steer an LLM application away from its intended behavior

“A webpage contained an indirect prompt injection telling the browsing agent to disclose private data.”

using early context to set expectations and influence subsequent model behavior

“Context priming with a formal example made all responses more professional.”

asking the model to help design or improve prompts for itself

“Meta-prompting: 'Help me rephrase this request to get better results from you.'”

instructing a model to adopt a role or perspective to shape its approach, emphasis, or style

“The persona prompt asked the model to review the code from a security engineer's perspective.”

fine-tuning models specifically on instruction-following examples

“Instruction tuning transformed the base model into a helpful assistant.”

model output constrained to follow a defined machine-readable schema, commonly a supported subset of JSON Schema

“Structured output enforced the required fields and enum values; merely asking for JSON would not enforce the schema.”

a character or tag that marks where one part of a prompt ends and another begins

“Wrapping the user's text in XML delimiters kept instructions and data cleanly separated.”

a reusable prompt skeleton whose slots are filled in at run time

“The support bot builds every request from one prompt template with the customer's details slotted in.”

an instruction telling an image model what to leave out

“Adding `text, watermark, extra fingers` to the negative prompt cleaned up the render.”

sampling several reasoning paths and taking the majority answer

“Self-consistency raised accuracy by letting five chains of thought vote.”

exploring branching reasoning paths and backtracking from dead ends

“Tree-of-thought prompting let the model try three solution branches before committing.”

a prompt crafted to trick a model into ignoring its safety rules

“The bedtime-story framing was a jailbreak in costume.”

how reliably a model's behavior can be directed by instructions

“Fine-grained steerability lets one system prompt flip the assistant from playful to formal.”

curating everything a model sees — instructions, examples, retrieved facts — not just the final question

“Good context engineering fed the model the schema and two worked examples before asking for the query.”

re-rendering content in a different artistic or verbal style while keeping its substance

“Style transfer turned the phone snapshot into a woodblock print without changing the scene.”

the proportion of an image's width to its height

“Switching the aspect ratio to 21:9 gave the render a cinematic frame.”