LLM Prompt Engineering Cheat Sheet

Prompt structure, roles, few-shot examples, chain-of-thought, and common techniques for better LLM output.

Reference

prompt-engineering

llm

Prompt engineering is the craft of writing instructions that steer a language model toward accurate, useful output. Clear structure, concrete examples, and explicit output constraints outperform vague requests.

Prompt structure

Table

Element	Purpose	Example
Role	Set persona	"You are a senior reviewer."
Context	Provide background	"This is a Node.js API."
Task	State the goal	"Find bugs in this code."
Format	Specify output shape	"Return a bulleted list."
Constraints	Bound the answer	"Do not invent APIs."

Techniques

Table

Technique	How
Few-shot	Provide 2-3 input/output examples.
Chain-of-thought	Ask for step-by-step reasoning.
Role prompting	Assign an expert persona.
Output schema	Request JSON with specific keys.
Self-consistency	Sample multiple answers and pick the majority.

Example prompt

code

Role: You are an expert TypeScript reviewer.
Context: This function parses user input in a web app.
Task: Identify security and correctness issues.
Format: Return a JSON array of {line, severity, issue, fix}.
Constraints: Only flag real bugs; do not restyle the code.

<code>
function parse(input) { return JSON.parse(input); }
</code>

Common pitfalls

Table

Pitfall	Fix
Vague task	State the exact deliverable.
No format	Specify the output structure.
Overlong context	Trim to relevant facts.
Ambiguous constraints	Add explicit "do not" rules.
No examples	Add one good few-shot pair.

References

Related tools

LLM Token Counter