Prompt Optimizer

Transform vague, underperforming prompts into precise, structured prompts that consistently produce high-quality AI outputs.

Description

This skill takes any user prompt — whether vague, ambiguous, or poorly structured — and systematically refines it into a professional-grade prompt following established prompt engineering principles. It applies techniques from chain-of-thought, role-prompting, few-shot learning, and structured output formatting to maximize AI performance.

When to Use

• The user provides a vague prompt like "write something about marketing" and expects better results
• A prompt produces inconsistent or off-topic outputs
• Converting natural language requests into structured prompts
• Building prompt templates for repeated use
• Debugging prompts that fail in edge cases

Instructions

The OPTIMIZE Framework

When refining a prompt, apply these six principles in order:

O — Objective (明确目标)

Problem: Vague verbs like "write about," "explain," "help with" Fix: Specify exact deliverable and success criteria

P — Persona (设定角色)

Assign a specific role to ground the AI's expertise:

• "You are a senior staff engineer at Google with 15 years of distributed systems experience"
• "You are a Nature journal reviewer specializing in immunology"
• "You are a direct-response copywriter trained by Eugene Schwartz's methods"

Include constraints: "Respond only with what you're confident about. If uncertain, say so."

T — Task Structure (任务结构)

Break complex tasks into ordered steps:

1. First, analyze X and identify Y
2. Then, based on Y, generate Z using method A
3. Finally, format the output as...

For multi-step tasks, use numbered steps rather than one compound instruction.

I — Input Specification (输入规范)

Define what the user will provide:

• "I will provide: (1) a product description, (2) target audience, (3) competitor list"
• "Input: A CSV file with columns [date, revenue, expenses]"
• "Here is the code to review: <language>\n...\n"

Explicit input templates reduce ambiguity.

M — Metrics & Constraints (约束条件)

Add specific constraints:

Constraints:
- Maximum 500 words
- Use only peer-reviewed sources
- No jargon; explain all technical terms
- Output in Chinese
- Format as a comparison table
- Must include 3 concrete examples

I — Ideal Output (理想输出)

Show or describe the desired output format: