Lesson 1 / 25
What Context Engineering Means
Prompt wording is one part; deciding what goes in the window is the bigger job.
Wording versus contents
Prompt engineering usually means writing clear instructions: role, task, constraints, format, examples. Context engineering is the wider job of deciding everything the model receives on a call: which instructions, which tool definitions, which examples, how much conversation history, which retrieved documents, which tool results, in what order and structure, and within what token budget. In real applications most of the window is filled by your code, not by a human typing, so the quality of answers depends heavily on that assembly step. A perfect instruction with the wrong documents still gives a wrong answer.
Everything the model sees is the prompt
Context engineering designs the whole input: instructions, tools, examples, history, documents and the question.
Briefing a new colleague
A good brief is not only polite wording: it is choosing which files to hand over, in what order, and what to leave out so they are not buried.
Log the final assembled prompt
Always be able to see the exact text sent to the model; most bugs are visible there and invisible in your template.
Quick check: What does context engineering add beyond prompt wording?
- Training the model weights
- Choosing a nicer font
- Deciding which content enters the window, in what order, structure and budget
- Raising the temperature
Answer
Deciding which content enters the window, in what order, structure and budget — The assembled input decides what the model can know on that call.