5.1.1. Prompt Structure and Context
💡 First Principle: A strong prompt pairs intent (what you want) with context (what the model needs to know to deliver it). Structure is just the discipline of making both explicit instead of assuming the model can infer them.
An effective prompt generally makes three things clear: the goal, the relevant context, and any constraints or format. In Copilot, that context comes from two places — what you type (a comment or Chat message) and what you expose (open files, the active selection, neighboring tabs). Good structure aligns both.
Compare:
| ❌ Weak prompt | ✅ Strong prompt |
|---|---|
// sort the list | // sort the list of User objects by lastLogin descending, nulls last, returning a new array |
| Vague intent, no constraints | Clear goal, type, ordering, edge case, and output shape |
The strong version isn't longer for its own sake — every added word removes an ambiguity the model would otherwise guess at. Notice it specifies the type (User objects), the key and order (lastLogin descending), an edge case (nulls last), and the output shape (new array).
✅ Best Practice: Start general to set direction, then get specific. A short context-setting line followed by a precise request usually beats one long, unstructured paragraph.
⚠️ Exam Trap: More words is not the goal — less ambiguity is. A prompt padded with irrelevant detail can perform worse than a short, precise one because it dilutes the signal.
Reflection Question: What four ambiguities does the "strong prompt" above resolve that the weak one leaves to chance, and why does resolving them improve the suggestion?
