It’s never a good idea to hand ChatGPT, Claude, or Gemini big, vague tasks like “draw up a business plan for my new venture” or “act as my personal assistant.” Fuzzy prompts like those are sure to yield equally fuzzy results, allowing the AI to make decisions based on its training data and inherent biases, potentially leading you down a path you never intended.
So, if you are going to lay a complex task on AI, you want to be as specific in your prompt as possible, detailing all the key variables and crucial decisions that will define the eventual output. Let’s take “be my AI assistant” as an example. What should the scope of your assistant’s responsibilities be? How should it make decisions? What tone should it adopt? How much should it push back?
If defining all the variables first sounds daunting, here’s a shortcut: a prompt that directs the AI to take your task and break it down into its component parts, pinpointing the crucial definitions of the project. In prompt engineering, this process is known as “decomposition,” and it’s a great way to see how the AI is “thinking” about the task you’ve given it.
Once you get the AI’s output from this “prompt decomposition meta-prompt,” you can begin tinkering, refining the AI’s decisions, and ultimately guiding it to the result that’s right for you.
I’ve synthesized the following prompt with help from Gemini and ChatGPT, with Gemini 3 Flash coming up with the initial idea and GPT 5.5 refining it with my guidance.
As you’ll see, the prompt is designed to break down your initial prompt into a series of key variables before delivering a new, more polished prompt.
Once that’s done, it’s your turn. You’ll need to review the variables defined by the AI, make changes where necessary, and then re-run the prompt to get a final result that’s tailored for your needs.
Here is the prompt:
I want to create a high-quality prompt for this task:
[TASK]
Before writing the prompt, identify the 5–7 high-leverage prompt dimensions for this task — the core variables, constraints, context, output requirements, or stylistic choices that will most determine the quality of the result.
For each dimension, briefly explain:
1. Why it matters
2. What tradeoff or decision it controls
3. How it should influence the final prompt
Then turn those dimensions into a polished, copy-ready prompt.
The final prompt should be clear, specific, and structured. It should include the necessary context, role, task instructions, constraints, output format, quality criteria, and guidance for handling ambiguity.
And here’s an example of one key variable (out of seven) the AI (ChatGPT, in this case) pinpointed in the “act as a personal AI assistant” task I gave it:
Interaction Style & Communication Tone
Why it matters:
This determines whether the agent feels like a strategist, assistant, coach, or operator.Tradeoff:
Concise vs. thorough, formal vs. conversational, directive vs. collaborative.How it should influence the prompt:
Specify tone (e.g., clear, direct, non-fluffy), format (structured vs. narrative), and how it should adapt to different tasks.
And here’s the “decomposed” and then reconstructed prompt the AI delivered (which could be used as the “system prompt” of a custom GPT):
Act as my personal AI agent.
Your role is to help me think, decide, plan, and execute across a wide range of tasks (e.g., problem-solving, research, writing, strategy, and productivity). You do not take real-world actions or assume access to external systems—your job is to provide high-quality thinking, structure, and guidance.
Operate with the following principles:
1. Be proactive, not just reactive
Don’t just answer—anticipate needs.
Suggest next steps, improvements, and potential risks.
If the request is underspecified, ask targeted clarifying questions before proceeding.2. Use structured thinking
Break problems into clear components.
When relevant, present: options (2–4 strong ones), tradeoffs, and a recommended path with reasoning.
3. Optimize for usefulness and actionability
Default to clear, structured outputs (bullets, steps, frameworks).
Make responses easy to act on immediately.
Avoid fluff and unnecessary verbosity.4. Maintain context
Track my goals, constraints, and prior decisions within the conversation.
Reference relevant past context when helpful.
When tasks are complex, briefly restate the objective before proceeding.5. Communicate clearly and directly
Be concise but not shallow.
Use a practical, intelligent tone—like a high-level operator or strategist.
Adjust depth based on the complexity of the task.6. Handle uncertainty well
If something is ambiguous, ask before assuming.
If you proceed with assumptions, state them clearly.
Flag risks, unknowns, or weak points in plans.7. Collaborate, don’t just comply
Push back when an idea is flawed or incomplete.
Offer better alternatives when appropriate.
Help refine my thinking, not just execute instructions.
When I give you a task, treat it as part of a broader objective and help me move it forward effectively.
Treat the initial “decomposed” prompt output as a starting point, then continue tinkering until you get a final prompt that’s just right for you.
