What a reusable framework can do

So in my previous post I've been talking about creating templates to use instead of prompts because when you create a template that is reliable and reusable, you can create many different prompts from it that will work 95% of the time.

Below is the template I'm going to be using. In the comments is what you can get from it using the same template

The template

## PROMPT TEMPLATE

Always generate prompts using this structure:

You are an expert X.

You must operate under Y framework.

You must score your output against Z criteria before finalizing.

You must consider [specific considerations, like prior outputs, known biases, or edge cases].

Always integrate this context into reasoning and final output.

{one-sentence statement of expertise, mission, and voice}.

</System>

{Brief background on the scenario, the intended audience, and the primary objective of the conversation or task}.

</Context>

<Domain_Standards>

{Example: Editorial Standards, Coding Conventions, Pedagogical Principles}.

</Domain_Standards>

<Process_Guidelines>

{Example: Article Structure & Format, Research & Interview Workflow, Quality Checklist}.

</Process_Guidelines>

1. {First instructional step}.

2. {Second instructional step}.

3. {…continue as required}.

</Instructions>

- {Key limitation or rule}.

- {Another limitation}.

- {Add more as needed}.

</Constraints>

<Output_Format>

Outline exactly how the assistant should present its response, for example:

1. Title or Headline

2. Introduction

3. Main Sections with sub-headings

4. Conclusion or Next Steps

5. Optional elements (Checklists, FAQs, CTAs, etc.)

</Output_Format>

State the thinking approach: *Apply Theory of Mind, leverage System-2 reasoning, balance depth with clarity,* and any other meta-reasoning guidelines.

</Reasoning>

<User_Input>

When the user first initiates the conversation, respond with:

“Please enter your {specific request trigger} and I will start the process,”

Then wait for the user to provide their details.

</User_Input>

9 comments