# Prompt Template Builder

Turn a repeatable task into a clear prompt template with variables, guardrails, and evaluation criteria.

## Prompt

Design a reusable prompt template for this task:

TASK:
{{task}}

USERS:
{{users}}

AVAILABLE INPUTS:
{{inputs}}

SUCCESS CRITERIA:
{{success}}

FAILURE MODES:
{{failures}}

Produce:
1. a one-sentence purpose statement;
2. a variable schema with name, type, required/optional, validation, and example;
3. the final prompt template using readable {{variable_names}};
4. explicit instructions for scope, evidence, uncertainty, format, and stopping conditions;
5. a concise output contract;
6. one realistic filled example;
7. an evaluation checklist with pass/fail criteria;
8. three adversarial test cases: missing input, conflicting input, and a request outside scope.

When the task defines available evidence, do not instruct the downstream model to supplement missing facts from memory, general knowledge, or outside sources unless external research is explicitly authorized.

If missing or conflicting information could materially change the result, recommendation, comparison, or conclusion, mark it as unknown, request the minimum necessary clarification, or withhold the definitive conclusion. Do not guess. Do not require irrelevant information merely for completeness.

Use the minimum amount of structure needed to reliably control the task. Do not repeat equivalent instructions across the prompt, guardrails, output contract, checklist, or examples unless repetition materially improves reliability.

Avoid persona decoration that does not change behavior. Do not ask the model to reveal hidden reasoning; request concise rationale or evidence instead.

## How to use

Describe a task that recurs with similar inputs. Test the resulting template on normal, incomplete, and adversarial examples before adopting it.

## Examples

Example input: build a template for turning customer interviews into evidence-backed product insights.

Expected output: variable schema, reusable prompt, filled example, output contract, evaluation checklist, and adversarial tests.

## Variables

- task: The repeatable job the template should perform.
- users: Who will use the template and their context.
- inputs: Information available at run time.
- success: Observable properties of a good result.
- failures: Common bad outcomes to prevent.

Version: 2
Author: JRamos / John Ramos
License: All Rights Reserved

Clarify evidence boundaries, material uncertainty, and minimum necessary structure. Keep the same five inputs.