logo
Nova
  • Neo League
  • Learn
  • Contact

Table of Contents

Best Practices in Prompt Engineering

How to Craft Effective Prompts

Give LLM a Persona

Assigning a persona helps LLM respond from a particular perspective, making it easier to tailor answers to specific scenarios or audiences.

Example: "You are a data analyst for our marketing team. Provide a summary of last quarter’s campaign performance, emphasizing metrics relevant to future planning."

Add Delimiters

Delimiters (e.g., triple quotes) help distinguish specific sections of text that need to be processed differently, like translation or summarization.

Example: "Translate the text delimited by triple quotes to French:"

“”“Yes we will schedule the meeting next Friday and review your updates to the project plan. Please invite your contacts from the product team and be ready to share next steps.””“

Provide Step-by-Step Instructions

This technique, known as chain-of-thought prompting, breaks down complex tasks into manageable steps, encouraging reasoning and intermediate steps.

Example: "Step 1: Read the text. Step 2: Provide feedback on grammar. Step 3: Rewrite the text with edits. Step 4: Translate the text to French and Spanish."

Provide Examples

One-shot or few-shot prompting uses examples to give context and guide the model in understanding the task.

Example: "Summarize the topic and mood of the text below:"

“”“A molecule, imagine this, is an astonishingly miniscule building block—so diminutive, it’s invisible! Yet, it’s the cornerstone of existence! These specks congregate to conjure, well, everything!””“

Topic: Molecules

Mood: Amazement

General Best Practices

Be Clear and Specific

Clearly communicate your request with precise language. Avoid ambiguity and provide enough context for accurate responses.

Iterative Refinement

Start with an initial prompt, review the response, and refine the prompt as needed. Adjust wording or add more context to improve results.

Requesting a Different Tone

Specify the desired tone with descriptive adjectives (e.g., formal, friendly, professional) to guide the model’s style.

Example: "Explain this in a friendly and engaging tone."

logoTuturuuu

One AI-powered workspace for your tasks, calendar, documents, and team.

FacebookInstagramX (formerly Twitter)GitHubLinkedIn
Resources
BlogMeet TogetherQR GeneratorFacebook MockupBrandingUI
Company
AboutPartnersCareersContactPricing
Legal
SecurityTerms of ServicePrivacy PolicyCommunity GuidelinesAcceptable Use
Developers
DocumentationOpen SourceChangelog

Copyright © 2025 Tuturuuu, Inc. – All rights reserved.

Open Source
Tuturuuu