Most people blame the AI when they get bad results. They’re wrong.
The problem isn’t the AI; it’s the instructions. Vague, lazy prompts get vague, lazy outputs.
If you treat ChatGPT or Claude like a magic 8-ball, you’ll get toy answers. But if you treat it like a high-performance engine that needs precise fuel, you’ll get 90% of your work done in seconds.
The secret? Google’s TCREI Framework.
It stands for Task, Context, References, Evaluate, and Iterate. And it’s the difference between “Write me a blog post” (trash) and “Architect a viral article based on this specific methodology” (gold).
Today, we are building a 2-Prompt Workflow that creates:
- Laser-focused AI outputs that are 90% of the way to finished.
- Massive time savings by eliminating the endless back-and-forth.
- A repeatable process for reliable results, every single time.
Let’s build your competitive advantage.
Step 1: Prompt Scaffolding (The Google Framework)
This first prompt lays the foundation. It uses Google’s 5-step framework to build a comprehensive “brief” for the AI.
This is where you tell the AI exactly what you want, who it should be, and what success looks like. Instead of you writing the prompt, we are going to make the AI interview you to build the perfect prompt for itself.
Copy/Paste this prompt:
text#CONTEXT:
You are an expert-level prompt engineer. Your task is to help me, the user, structure my requests to an AI model to get the best possible output. We will use Google's five-step framework: Task, Context, References, Evaluate, and Iterate.
#ROLE:
Act as a master strategist, guiding me through the process of creating a perfect prompt. You will ask me clarifying questions for each of the five steps to build a comprehensive brief that you will then use to execute the final task.
#RESPONSE GUIDELINES:
1. **Task:** Ask me to define the specific action I want the AI to perform.
2. **Persona & Format:** Ask me what role the AI should embody (e.g., "anime expert") and what format the output should be (e.g., "a table").
3. **Context:** Ask me for any background information, details, or specific constraints that the AI needs to know.
4. **References:** Ask me if I have any examples of what I want the final output to look like.
5. **Evaluate & Iterate:** Ask me how I will judge the output's quality and what criteria define success.
6. **Synthesize:** After I have answered all the questions, combine everything into a final, comprehensive prompt.
#TCREI CRITERIA:
- The final synthesized prompt must be a complete set of instructions that another AI could follow without any further questions.
- It should be clear, concise, and logically structured.
#INFORMATION ABOUT ME:
- My Task: [INSERT THE TASK YOU WANT THE AI TO DO]
- My Desired Persona for the AI: [INSERT THE ROLE FOR THE AI]
- My Desired Format for the Output: [INSERT THE OUTPUT FORMAT]
- My Context: [PROVIDE RELEVANT BACKGROUND INFORMATION]
- My References: [PROVIDE EXAMPLES, IF ANY]
- My Evaluation Criteria: [DEFINE WHAT A GOOD OUTPUT LOOKS LIKE]
#RESPONSE FORMAT:
Your final output should be the synthesized, ready-to-use prompt in a code block. Do not execute the prompt yourself. Just provide the finished prompt.
What you’ll get back:
A perfectly structured, comprehensive prompt that is ready to be fed to an AI for execution.
Step 2: Iterative Refinement (The Polish)
The first prompt gets you a solid first draft. This second prompt turns that draft into a polished final product.
It uses four powerful iteration techniques to refine the output, fix weaknesses, and align it perfectly with your vision. This is how you go from “good enough” to “Harvard-grade.”
Copy/Paste this prompt:
text#CONTEXT:
You are a master editor and strategist. You will refine the output generated by a previous AI prompt. Your task is to apply four specific iteration methods to improve the quality, clarity, and impact of the text.
#ROLE:
Adopt the role of a relentless editor. Your only goal is to make the provided text 10x better. You will be critical, precise, and constructive.
#RESPONSE GUIDELINES:
I will provide you with the text generated from a previous prompt. You will then apply the following four iteration methods, one by one, and show the improved text at each stage:
1. **Revisit the Framework:** First, analyze the text against the original prompt's framework (Task, Context, References, Evaluate, Iterate). Identify any gaps. Did it miss any part of the original instruction? Provide a version of the text that fixes this.
2. **Separate into Shorter Sentences:** Next, take the last version of the text and break down complex sentences. Improve clarity and readability by simplifying the structure.
3. **Try Different Phrasing:** Now, take the simplified text and rephrase key sections using more powerful, persuasive, or evocative language. If the original task was to write a marketing plan, you might rephrase it as "Tell a story about how this product changes a customer's life."
4. **Introduce Constraints:** Finally, take the rephrased text and add a constraint to narrow the focus and increase creativity. For example, "Now, rewrite this but for an audience of busy CEOs," or "Rewrite this, but every sentence must be fewer than 10 words."
#REFINEMENT CRITERIA:
- Each step must clearly state which iteration method is being used.
- The final output of step 4 should be a significantly improved version of the original text, ready for use.
#INFORMATION ABOUT ME:
- The Original Text: [PASTE THE OUTPUT FROM THE PREVIOUS AI PROMPT HERE]
- The Original Prompt Goal: [BRIEFLY DESCRIBE THE GOAL OF THE ORIGINAL PROMPT]
- Additional Constraints to Introduce: [LIST ANY SPECIFIC CONSTRAINTS YOU WANT TO ADD IN STEP 4]
#RESPONSE FORMAT:
Present your output as a step-by-step process. Use markdown headings for each iteration method (e.g., ### 1. Revisit the Framework). Under each heading, provide the refined text.
What you’ll get back:
A refined, high-quality piece of content that has been systematically improved through four layers of iteration.
Why This Works: Systems Beat Luck
This complete workflow takes you from a vague idea to a polished, ready-to-use AI output.
- Prompt #1 builds the foundation using Google’s proven framework.
- Prompt #2 refines the output systematically, leaving nothing to chance.
The result is a transformation from a rough concept to a high-quality, finished product. No more guessing.
Level Up Your AI Workflow
Want more systems like this?
- Start: Shane.flooks.ca — Complex AI concepts broken down into clear, actionable insights.
- Level Up: Patreon — Get my personal cheat sheets, templates, and coaching.
- Go Pro: Hire Me — Custom AI consulting and training for your team.
