How do you write a concise prompt that still carries enough context?

Question

Accepted Answer

A good prompt is concise but **complete**: it states the goal, the constraints, the expected output format, and the minimal relevant context up front — so the model can answer correctly on the first try instead of forcing a back-and-forth.

## What a sharp prompt includes

- **Goal** — what you actually want produced.
- **Constraints** — language, version, libraries, style, what to avoid.
- **Output format** — code only, a diff, JSON, a bullet list, etc.
- **Minimal context** — just the relevant snippet or fact, not the whole file.

## Vague vs sharp

```text
Vague:
  "Fix my date function."
  → The model has to guess the language, the bug, and what "fixed" means.
```

```text
Sharp:
  "In this TypeScript function, parseDate returns NaN for 'DD/MM/YYYY' input.
   Fix it to parse day-first dates. Return only the corrected function.
   function parseDate(s: string) { return Date.parse(s); }"
  → goal (fix day-first parsing), constraint (TypeScript), context (the function),
    output format (only the function) — all in four lines.
```

The sharp version removes every guess. The model knows the language, the exact failure, the desired behavior, and how to hand back the result — so you get a usable answer immediately.

## Why it matters

Concise, well-specified prompts cut the slow back-and-forth where the model guesses wrong and you correct it. Stating the goal, constraints, output format, and just-enough context up front saves tokens, time, and frustration — and makes the model's answer far more likely to be right the first time.