# Be Clear and Specific — Prompt Engineering

Source: https://www.geekswithgeeks.com/en/prompt-engineering/c-clarity

> Replace vague requests with measurable ones.

## Vague in, vague out

"Make this better" or "write about dogs" leaves the model to guess your goal, length, tone and audience. Replace vague words with **specifics you could check**: "Rewrite this email for a customer who is angry about a late delivery; keep it under 120 words; apologise once; offer a refund or reshipment; plain, warm tone." State the **goal** (why), the **audience** (who), the **constraints** (how long, what style, what to avoid), and what **done** looks like. Positive instructions ("use short sentences") usually work better than only negative ones ("do not be wordy"), though both help.

## Be specific, show examples, set boundaries

Clarity, examples and constraints solve most prompt problems.

![Five techniques: clarity, examples, role, format, delimiters.](assets/figures/prompt-engineering/section-2-map.svg) — Figure 2.1 — Clarity, examples, role, format and delimiters.

## Vague versus specific

The same request written two ways.

```text
VAGUE
  Summarise this report.

SPECIFIC
  Summarise the report below for the CFO in 5 bullet points.
  Focus on revenue, costs and risks. Each bullet under 20 words.
  Quote any figure exactly as written. If a figure is missing, write "not stated".
```

## Say what "good" looks like

Add a checkable success criterion, such as "each bullet under 20 words", so you can test the output automatically.

**Quiz:** Which prompt is better?

- [ ] "Summarise this"
- [x] "Summarise for the CFO in 5 bullets, under 20 words each"
- [ ] "Do something useful"
- [ ] "Be smart"

*Answer:* "Summarise for the CFO in 5 bullets, under 20 words each". Specific audience, length and format leave less to guesswork.