What is prompt context? Definition for LLMs
Prompt context is the text supplied alongside an instruction to a large language model — the background, examples, documents, or transcripts the model reads before producing its answer.
Prompt context is the text supplied alongside an instruction to a large language model — the background, examples, documents, or transcripts the model reads before producing its answer.
A model handed only "fix the bug" with no further information can do nothing useful. Handed the same instruction plus the relevant source file, the error message, and a screenshot of the broken state, it can produce a specific change. The difference is the context.
Curating context
The instinct to dump everything into the prompt is usually wrong. Models pay attention better to shorter, more relevant context than to longer context full of noise. The right material is whatever a knowledgeable teammate would need to answer the same question — and nothing else.
For visual product work, the most useful context format is markdown with the relevant screenshot inline, the source URL on a line of its own, and the question stated last. The model reads it the same way a person would: image first, then where it came from, then what to do about it.
Frequently asked questions
How is prompt context different from a prompt?
The prompt is the instruction. The context is the supporting material — documents, examples, history — provided so the model can answer the instruction well. Most production prompts are mostly context with a short instruction at the end.
Does more context always produce better answers?
No. Beyond a point, extra context introduces noise the model has to filter through and increases cost. The useful version is curated and relevant, not exhaustive.
Capture your first review.
About a minute from open tab to a shareable URL your agent can ingest.
Start capturing