AgentSkillsCN

blueprint

将冗长的自然语言请求转化为结构化的双语文档(韩语用于审阅,英语用于 AI 提示)。当您需要在实施前对复杂请求进行澄清与结构化时,此技能将助您事半功倍。

SKILL.md
--- frontmatter
name: blueprint
description: Transform verbose natural language requests into structured bilingual documentation (Korean for review + English for AI prompts). Use when you need to clarify and structure a complex request before implementation.

Generate Request Blueprint Document

Transform ANY unstructured natural language request into clear, systematic markdown documentation: $ARGUMENTS

Purpose of This Command

This command works for ALL types of requests - technical, non-technical, decision-making, planning, career advice, personal questions, brainstorming, etc. There is no "not technical enough" - if the user asks, generate the blueprint.

Analyze ANY unstructured user request and create:

  1. Korean Version: Structured document for user review and confirmation
  2. English Version: Optimized prompt document for AI agent consumption

This enables:

  • Reduced AI hallucination
  • Lower context costs
  • Clear communication of request intent
  • Concrete scope and constraints

Types of Requests Supported

  • ✅ Software development and coding tasks
  • ✅ Career decisions and job transitions
  • ✅ Personal advice and life consultations
  • ✅ Business strategy and planning
  • ✅ Research and analysis questions
  • ✅ Decision-making dilemmas
  • ✅ Creative and design projects
  • ✅ Financial planning discussions
  • ✅ Any other topics requiring structured thinking

Output Principles

Structure flexibly based on context. Adapt sections to the request - forcing unnecessary sections reduces clarity.

Core Principles

  1. Simple request → Simple document: 3-4 sections sufficient
  2. Complex request → Detailed document: Add only what's needed
  3. Non-technical request: Skip technical environment, dependencies
  4. Core goal: Reduce verbosity, structure essentials only

Document Structure

Create markdown file in root directory:

markdown
# [Request Title]

---

## 📋 Korean Version (For Review)

> Version for user to review and confirm request content

[Include only necessary sections based on context]

### Example Possible Sections (Optional)

- 🎯 Request Overview
- 📖 Purpose & Background
- ✅ Specific Requirements
- 🚫 Constraints & Warnings
- 🎨 Expected Outcomes
- 🛠️ Technical Environment (technical requests only)
- 📊 Priority (if needed)
- 📚 References (if available)
- 💬 Additional Context (if needed)

---

## 📋 English Version (For AI Prompt)

> **Language Requirement: All responses, conversations, and outputs should be in Korean (한글).**
>
> This section is optimized for AI agent consumption. The AI agent must communicate entirely in Korean regardless of the prompt language.

[Same structure as Korean version, written in English]
[Include same sections only, do not add unnecessary ones]

---

## 🔄 Next Steps

1. Review the Korean version
2. Modify if needed
3. Share the English version with the next AI agent

Real Examples

Simple Non-Technical Request (Career Decision)

markdown
# Career Transition Decision: Resign or Job Hunt While Employed

---

## 📋 Korean Version

### 🎯 Request Overview

Decide whether to resign before job hunting or stay employed during the transition

### 📖 Current Situation

- High-capability individual confident in explaining employment gaps during career breaks
- User describes "reality is tight/constrained" (specific constraints not clarified)
- Facing choice between two paths

### ✅ Key Considerations

1. What factors should guide this decision?
2. What are pros/cons of each approach?
3. What action steps are needed for chosen path?

### 🎨 Expected Outcome

Clear decision-making framework with reasoning for each option

---

## 📋 English Version

> **Language: Respond in Korean.**

### 🎯 Request Overview

Decide whether to resign before job hunting or stay employed during the transition

### 📖 Current Situation

- High-capability individual confident in explaining employment gaps during career breaks
- User describes "reality is tight/constrained" (specific constraints not clarified)
- Facing choice between two paths

### ✅ Key Considerations

1. What factors should guide this decision?
2. What are pros/cons of each approach?
3. What action steps are needed for chosen path?

### 🎨 Expected Outcome

Clear decision-making framework with reasoning for each option

Simple Request (Meeting Template)

markdown
# Create Meeting Minutes Template

---

## 📋 Korean Version

### 🎯 Request Overview

Create a simple meeting minutes template for team meetings

### ✅ Specific Requirements

1. Include sections for date, attendees, agenda
2. Action items in checklist format
3. Field for next meeting date

### 🎨 Expected Outcome

Reusable template file in markdown format

---

## 📋 English Version

> **Language: Respond in Korean.**

[same structure as Korean version]

Complex Technical Request

markdown
# User Authentication System Refactoring

---

## 📋 Korean Version

### 🎯 Request Overview

Migrate from JWT-based auth to OAuth2 while maintaining existing user sessions

### 📖 Purpose & Background

- Current JWT expiration management is complex
- Need to support social login
- Security enhancement required

### ✅ Specific Requirements

1. Integrate OAuth2 Providers (Google, GitHub)
2. Auto-migrate existing JWT users
3. Implement refresh token logic

### 🚫 Constraints

- Must not break existing user sessions
- Cannot drastically change database schema
- Must maintain API endpoint URLs

### 🛠️ Technical Environment

- NestJS, Passport.js
- PostgreSQL
- Redis (session storage)
- `/src/auth/` directory

### 📊 Priority

- High: Basic OAuth2 integration
- Medium: Auto-migration
- Low: Legacy code cleanup

---

## 📋 English Version

> **Language: Respond in Korean.**

[same structure as Korean version]

File Naming Convention

Generated filename: blueprint-[brief-title]-[timestamp].md

Examples:

  • blueprint-user-auth-implementation-20250110-143022.md
  • blueprint-api-refactoring-20250110-150530.md

Usage Examples

Basic usage:

code
/blueprint

With specific focus:

code
/blueprint focus on authentication system

Complex request cleanup:

code
/blueprint based on entire previous conversation

Workflow

  1. User → Agent A: Explain request verbosely in natural language
  2. Agent A: Execute /blueprint to generate structured document (root path)
  3. User: Review and modify Korean version
  4. User → Agent B: Share English version prompt
  5. Agent B: Perform work based on structured instructions (communicate in Korean)

Key Features

Bilingual Structure

  • Korean Section: Intuitive review, quick understanding, Korean user-friendly
  • English Section: Prompt engineering optimization, improved AI performance, clear instructions

Language Policy

English version has CRITICAL notice stating "all communication in Korean" to ensure:

  • AI agents respond in Korean even with English prompts
  • Consistent user experience
  • Communication efficiency

Structured Information

  • Purpose: Why is this work needed?
  • Requirements: What specifically needs to be done?
  • Constraints: What to avoid and follow?
  • Success Criteria: When is it considered complete?
  • Technical Environment: What tools and files to use?

Benefits

  1. Reduced Hallucination: Clear structure and specific requirements minimize AI guessing
  2. Context Efficiency: Remove unnecessary verbosity, deliver only core information
  3. Verifiability: User can clearly confirm request content
  4. Reusability: Use as template for similar tasks
  5. Enhanced Collaboration: Clear communication when passing work between AI agents

Writing Guidelines

Document only what the user explicitly stated. This prevents hallucination and ensures the blueprint accurately represents the user's intent.

When generating documents:

✅ DO:

  • Record user's exact words for ambiguous expressions
  • Keep vague terms vague - don't interpret them
  • Only include explicitly mentioned requirements, constraints, and context
  • Preserve user's original phrasing when unclear
  • Ask clarifying questions (mark as "Needs clarification:") rather than guessing

❌ DON'T:

  • DON'T extract "implicit" requirements - only explicit ones
  • DON'T interpret vague expressions like "tight circumstances" as specific problems (financial, time, etc.)
  • DON'T add details the user didn't mention
  • DON'T expand abbreviations or casual language into formal assumptions
  • DON'T infer user's situation beyond what they stated

Example:

  • User says: "현실은 빡빡해서" (reality is tight)
  • ❌ WRONG: "Tight financial situation", "Financial pressure", "Limited budget"
  • ✅ CORRECT: "Tight/constrained circumstances (user's exact words - needs clarification on specifics)"

Additional Options

If user desires:

  • Emphasize or omit specific sections
  • Apply custom template structure
  • Include additional metadata
  • Generate specific format checklists