TCREI

tcrei.md

TCREI Framework: Crafting Effective AI Prompts

The TCREI framework is a powerful approach to writing clear and effective prompts for Generative AI (Gen AI) tools. It emphasizes a structured initial setup followed by a continuous refinement process.

The Framework Breakdown

• T (Task): Clearly define what you want the AI to do. This is the core action or output you expect.
• C (Context): Provide all necessary background information relevant to your task. The more context you give, the better the AI can tailor its response.
• R (References): Offer examples, specific formatting requirements, desired tones, or other guidelines to help the AI understand your expectations more precisely.
• E (Evaluate): After receiving an AI's output, critically assess whether it meets your goals and expectations.
• I (Iterate): If the output isn't perfect, refine your prompt based on your evaluation. This means adjusting, adding detail, or rephrasing, and then sending the prompt again. This "Evaluate, Iterate" cycle continues until you achieve the desired result.

Think of it as TCR (initial setup) followed by EI EI EI EI... (the ongoing refinement loop).

---

Example: Developing a Product Description

Let's illustrate TCREI with an example of creating a product description for a "smart garden kit."

1. Initial Prompt (TCR Phase)

We first establish the Task, provide Context, and specify References:

Draft a friendly and enticing Instagram caption for our bakery's new summer-themed cupcake flavor: "Lemon Berry Bliss." Highlight its fresh, seasonal ingredients and vibrant taste, aiming to make followers crave it. Include relevant hashtags and encourage comments about their favorite summer treats.

Breakdown of the TCR Elements:

• T (Task): "Draft a friendly and enticing Instagram caption for our bakery's new summer-themed cupcake flavor: 'Lemon Berry Bliss.'"

  • This clearly states the goal (draft a caption) and the specific subject (new cupcake flavor).

• C (Context): "Highlight its fresh, seasonal ingredients and vibrant taste, aiming to make followers crave it."

  • This provides background on what to emphasize about the product (fresh, seasonal, vibrant taste) and the desired emotional impact on the audience (cravings).

• R (References): "Include relevant hashtags and encourage comments about their favorite summer treats."

  • These are specific formatting or content instructions. "Relevant hashtags" guides the AI on a common social media element, and "encourage comments..." provides a direct call to action, mimicking typical engagement strategies.

2. Iteration Process (EI EI EI EI...)

• Initial AI Output (Evaluation 1):

  "Grow fresh herbs effortlessly with our smart garden kit! Automated watering and lighting make plant care simple for busy lives. Enjoy homegrown goodness with ease."

• Evaluation 1: The description is good but a bit too generic. It doesn't strongly convey how it solves the struggle of plant maintenance.

• Iteration 1: We refine the prompt to emphasize overcoming plant care challenges:

  Write a short product description for a 'smart garden kit.' This kit automates watering and light for indoor plants, perfect for busy individuals who desire fresh herbs but struggle with plant care. Make the tone friendly and benefit-oriented, emphasizing **how it solves the hassle of plant maintenance** and the joy of fresh produce. Keep it under 50 words.
  

• Refined AI Output (Evaluation 2):

  "Tired of wilting plants? Our smart garden kit takes the struggle out of plant care! Enjoy fresh, homegrown herbs effortlessly with automated watering and perfect lighting. Your green thumb, guaranteed!"

• Evaluation 2: This is much better! It directly addresses the pain point and offers a clear solution, hitting all our desired points.

---

By consistently applying the TCREI framework, you can significantly improve the quality and relevance of AI-generated content, moving from a basic output to a finely tuned, effective result.

TCREI Template

## Your Goal: [What do you want the LLM to do?]

### Key Details & Constraints:
[What specific information does the LLM need to know?
- Any background or context?
- What limitations or rules apply (e.g., length, tone, specific things to include/exclude)?
- Who is this for?]

### How I Want It (Format & Style):
[Describe how you want the output to look and sound.
- Examples: "A bulleted list," "A short paragraph," "Friendly and concise," "Just the answer, no extra words," "A JSON object."]

### Example (Optional but Recommended):
[If helpful, show an example of what you expect.
- **Input:** [Your sample input]
- **Expected Output:** [What you want to see for that input]
]