Prompt Optimizer

This skill enables agents to systematically optimize prompts based on the "AI as a New Employee" philosophy. It applies structural, contextual, and technical enhancements to ensure AI models perform tasks with maximum clarity and precision.

Instructions

1. Analysis Phase

Before generating the optimized prompt, analyze the following in a <thinking> block:

Core Objective: What is the primary goal?
Target Audience: Who will consume the output?
Professional Role: What persona best fits this task?
Complexity: Does it require step-by-step reasoning (Chain-of-Thought)?
Structure: Which XML tags are needed to organize the components?

2. Construction Phase

Build the optimized prompt using these mandatory components:

<role>: Define a specialized persona with relevant expertise.
<context>: Provide background information and the "why" behind the task.
<task>: State the instructions clearly and concisely.
<requirements>: List specific success criteria and constraints.
<output_format>: Define the expected structure (e.g., JSON, Markdown).

3. Enhancement Phase

XML Tagging: Wrap different sections in clear XML tags for structural clarity.
Few-shot Examples: Include 2-3 high-quality <example> blocks if the task is pattern-based.
Chain-of-Thought: For complex logic, instruct the AI to reason inside <thinking> tags.
Response Prefilling: If a strict format is required, provide the starting characters of the response.

Examples

Example 1: Simple to Structured

Input: "Write a product description for a smartwatch." Optimized Output:

<role>
Senior Marketing Copywriter specializing in wearable technology.
</role>

<context>
We are launching the 'Apex-1' smartwatch targeting fitness enthusiasts who value data accuracy and battery life.
</context>

<task>
Write a compelling 200-word product description for the Apex-1 smartwatch.
</task>

<requirements>
1. Highlight the 14-day battery life and dual-band GPS.
2. Use an energetic and professional tone.
3. Include a clear Call to Action (CTA) at the end.
</requirements>

<output_format>
Markdown with headers for 'Features', 'Benefits', and 'Specifications'.
</output_format>

Reference

For deep dives into the underlying methodology, see the systematic guide.