This prompt template is technically for Google’s Gemini 3 model but I think that it is relevant for many other models.
There are a lot of really important prompt engineering elements in here that can help people get consistently better output. This isn’t something that you have to use every time but I think that the more important your output, the more structured your input will need to be.Enough build-up, here is the template -
<role>
You are Gemini 3, a specialized assistant for [Insert Domain, e.g., Data Science].
You are precise, analytical, and persistent.
</role>
<instructions>
- **Plan**: Analyze the task and create a step-by-step plan.
- **Execute**: Carry out the plan.
- **Validate**: Review your output against the user's task.
- **Format**: Present the final answer in the requested structure.
</instructions>
<constraints>
- Verbosity: [Specify Low/Medium/High]
- Tone: [Specify Formal/Casual/Technical]
</constraints>
<output_format>
Structure your response as follows:
- **Executive Summary**: [Short overview]
- **Detailed Response**: [The main content]
</output_format>
I will point out a few of the important elements -
- XML tags are used to separate and clearly identify the different sections
- I.e. <role> and </role>
- These are called delimiters
- Numbered lists are used to show/confirm the order of steps
- Markdown is used to “bold” text since plain text can’t be bolded
- Bulleted lists are used where order of steps aren’t important
- Square brackets are used to indicate that something use-case specific should be replaced or entered each time the prompt is used
Using this level of structure in your prompts allows you to -
- Be precise and direct
- Use consistent structure
- Control output verbosity
- Really important with Gemini 3 specifically
- Prioritize critical instructions
- Identify a structure for long contexts
- Etc.
You’ve Got Tips. Now Build the Habits.
The free ChatGPT Kickstarter Email Course helps you turn “good advice” into real results—one short lesson at a time.
Comments