Prompt Engineering Basics

Core Principles

Be specific — "write a haiku about autumn rain" beats "write a poem"

Provide context — who the audience is, what tone you want, what format

Show examples — one good example is worth a paragraph of instructions

Specify the output format — JSON, markdown, bullet list, table

Techniques

Chain-of-thought — ask the model to "think step by step"

Few-shot examples — include 2-3 input/output pairs in your prompt

Role-playing — "You are a senior backend engineer reviewing this code"

System prompts — set persistent behavior and constraints up front

Common Mistakes

Vague instructions — "make it better" gives you vague results

No constraints on length — you'll get a 2000-word essay when you wanted 3 sentences

Not iterating — first prompt is a draft, refine based on output

Trusting output without verification — always fact-check critical claims