Prompt Library

This is your prompt arsenal. We've collected and organized the most practical, battle-tested prompt templates across the industry -- covering everything from basic text classification to complex code generation.

These templates aren't rigid rules. They're reusable starting points. Copy them, fill in your specific Context and Constraints, and you've got a high-quality prompt in seconds.

Why Use a Template Library?

Core Prompt Structure

A solid general-purpose prompt usually has four core elements (CRDO):

graph LR
    C[Context<br>Background info] --> P[Prompt]
    R[Role<br>Role assignment] --> P
    T[Task<br>Specific task] --> P
    O[Output<br>Output requirements] --> P

Or as a more intuitive text structure:

# Role
You are a [expert role] skilled in [core competency].

# Context
I'm facing [specific situation/problem] and need [specific goal].

# Task
Please complete the following:
1. [Step 1]
2. [Step 2]

# Constraints
- [Constraint 1, e.g., word limit]
- [Constraint 2, e.g., tone of voice]

# Output Format
Please output in [Markdown/JSON/list] format.

Template Library Navigation

We've organized prompts by task type. Click any link to jump to the template page.

📊 Classification & Analysis

For labeling, sentiment analysis, intent detection, and similar tasks.

• Classification: General text classification.
• Sentiment Analysis: Detect sentiment polarity (positive/negative/neutral).
• Few-shot Sentiment: Boost accuracy with Few-shot examples.

🧠 Reasoning & Logic

For tasks that need logical thinking, math, or complex problem-solving.

• Reasoning: General logical reasoning (CoT).
• Mathematics: Math problem solving.
• Truthfulness: Reduce hallucinations, improve factual accuracy.

📝 Generation & Creativity

For writing, brainstorming, image generation prompts, and more.

• Creativity: Creative writing templates.
• Text Summarization: Extract summaries quickly.
• Image Generation: Craft prompts for Midjourney/Stable Diffusion.

💻 Coding & Engineering

For code generation, explanation, debugging, and refactoring.

• Coding: Code generation and optimization.
• Information Extraction: Pull entities or data from unstructured text.
• Question Answering: Precise Q&A based on documents or knowledge bases.

🛡️ Safety & Evaluation

Model security, red teaming, and output quality assessment.

⚠️ Note: Adversarial prompts are for Red Teaming and security research only. Don't use them for malicious purposes.

• Evaluation: Prompts for evaluating model output quality.
• Adversarial: Common attack patterns and defenses.
• Prompt Leaking: Detecting and preventing prompt leakage.
• Jailbreaking: Jailbreak examples (research only).

How to Use These Templates

1. Copy: Find a template that matches your task, copy it.
2. Fill: Replace the [variables] with your actual business content.
3. Test: Run the prompt, observe the output.
4. Iterate: If the result isn't great, tweak the Context or Constraints until you're happy.

🌟 Universal Starter Template

Not sure which template to use? Start here:

# Role

You are an experienced [domain expert] focused on providing high-quality solutions for [target audience].

# Task

Given the following [input content], perform [specific operation (e.g., analyze, summarize, rewrite)].

# Input Data

"""
[Paste your text/data here]
"""

# Constraints

-   Keep the tone [professional/friendly/objective].
-   Focus on [key point 1] and [key point 2].
-   If you encounter [situation A], then [response action].

# Output

Please output the result in [specific format].

Takeaways

• Don't memorize templates: Understanding the logic behind them matters more than rote memorization.
• Context is King: No matter how good the template is, providing accurate, rich context is always the key to better results.
• Keep iterating: Prompt development is an experimental process. There's no perfect prompt -- only prompts that keep getting better.