Chain-of-Thought Template

beginner

The zero-shot and few-shot chain-of-thought prompt templates that improve accuracy on multi-step reasoning tasks by making the model show its work.

System Prompt

{{USER_QUESTION}}

Think through this step by step before giving your final answer.

$npx genai-skills add chain-of-thought

What this prompt does

Forces the model to show step-by-step reasoning before giving a final answer. Improves accuracy on math, logic, and complex analysis tasks by making the reasoning process explicit.

When to use it

Multi-step math or logic problems
Complex analysis requiring intermediate conclusions
Tasks where you need to audit the reasoning, not just the answer

When NOT to use it

Simple factual lookups — adds unnecessary tokens
Creative writing — structured reasoning constrains output quality

The Prompt

Zero-shot variant

Few-shot variant

Solve the following problems step by step.

Problem: {{EXAMPLE_PROBLEM_1}}
Step 1: {{EXAMPLE_STEP_1}}
Step 2: {{EXAMPLE_STEP_2}}
Answer: {{EXAMPLE_ANSWER_1}}

Problem: {{EXAMPLE_PROBLEM_2}}
Step 1: {{EXAMPLE_STEP_1_P2}}
Step 2: {{EXAMPLE_STEP_2_P2}}
Answer: {{EXAMPLE_ANSWER_2}}

Problem: {{USER_PROBLEM}}

Why it's structured this way

"Step by step" — This exact phrase is the canonical zero-shot chain-of-thought trigger (Kojima et al., 2022). Other phrasings produce weaker results.

Examples show the format, not just the answer — Few-shot CoT examples must include intermediate steps. Models learn the reasoning format by demonstration.

Answer labeled explicitly — The Answer: label gives a reliable extraction point for programmatic parsing.

Variations

Variant	When to use
Zero-shot	No examples available; works on GPT-4/Claude class models
Few-shot (2–3 examples)	Highest accuracy; use when you have labeled examples
Zero-shot-CoT + self-check	Add: "Now verify your answer by checking each step."

Pattern:Chain-of-Thought Template →