English | 简体中文

Prompt Optimizer v2.1

A professional Claude Code Skill that transforms simple instructions into production-ready prompts.

✨ Features

🎯 Core Capabilities

• CLARITY Framework: A systematic prompt optimization method covering 7 key elements (Context, Logic, Action, Role, Input/Output, Tone, Yardstick).
• 61 Prompt Frameworks: Categorized by complexity and domain, with detailed definitions and examples.
• Quality Validation: Automatic validation using a 7-point checklist after generation to ensure high-quality output.
• Multi-Version Output: Provides Basic, Enhanced, and Expert versions for different scenarios.
• Boundary Handling: Intelligently handles vague inputs and situations where users refuse to clarify.
• Multiple Installation Methods: Supports Claude Marketplace, npx, and local installation.

🆕 New in v2.1

| Feature | Description | | :--- | :--- | | Structured Framework Index | frameworks/index.json with id, category, elements, domains, use_cases for all 61 frameworks. | | CLARITY Rubric | Binary pass/fail checklist with explicit criteria per letter — reproducible scores across runs. | | CLI template Command | Renamed from optimize to make it explicit that CLI prints static scaffolds; optimize/o remain as aliases. | | CLI Framework Queries | frameworks --json, --filter <domain>, --category <cat> for structured lookups. | | Workflow Tracking | Step 1 now uses TaskCreate for the 7 steps instead of plain-text checklists. | | Deduplicated Tests | 29 unique test cases (was 34 with duplicates) with bash sanity script. |

📦 v2.0 Features

| Feature | Description | | :--- | :--- | | Framework Library | 61 detailed framework definitions with structure, examples, and best practices. | | Quality Validation | CLARITY 7-point checklist to ensure output quality. | | Multi-Version | Different optimization levels based on user needs. | | Boundary Handling | Graceful handling of vague inputs and refusal to clarify. | | Test Suite | 29 test cases covering various usage scenarios. | | Claude Plugin | Support for installation via Claude Code Marketplace. | | npx Support | Use directly via npx without installation. |

📦 Installation

Method 1: Claude Code Marketplace (Recommended)

Install directly from GitHub using the marketplace command:

/plugin marketplace add dreamor/prompt-optimizer-skill
/plugin install prompt-optimizer@prompt-optimizer-skill
/reload-plugins

Method 2: npm Install

Install globally or locally via npm:

# Install globally
npm install -g prompt-optimizer-skill

# Or install as a project dependency
npm install prompt-optimizer-skill

After global installation, use the CLI directly:

prompt-optimizer-skill template "Write an email to a customer"
prompt-optimizer-skill frameworks
prompt-optimizer-skill version

Method 3: npx Usage (No Installation Required)

Use directly via npx without local installation:

# Generate a prompt template (static scaffold)
npx prompt-optimizer-skill template "Write an email to a customer"

# List all frameworks
npx prompt-optimizer-skill frameworks

# Query frameworks by domain or category
npx prompt-optimizer-skill frameworks --filter marketing
npx prompt-optimizer-skill frameworks --category medium

# Output frameworks as JSON
npx prompt-optimizer-skill frameworks --json

# Run tests
npx prompt-optimizer-skill test

# View help
npx prompt-optimizer-skill help

Quick Commands:

# Use short aliases
npx pos "Write an email to a customer"
npx prompt-optimizer-skill optimize "Write code"   # alias for template

# Specify output version
npx prompt-optimizer-skill template "Write code" --basic      # Basic version
npx prompt-optimizer-skill template "Write code" --enhanced   # Enhanced version (default)
npx prompt-optimizer-skill template "Write code" --expert     # Expert version

Method 4: Symbolic Link (Development/Customization)

Suitable for users who need to modify or customize:

# Clone the repository
git clone https://github.com/dreamor/prompt-optimizer-skill.git

# Create symbolic link
cd ~/.claude/skills/
ln -s /path/to/prompt-optimizer-skill prompt-optimizer

Verify Installation

# Claude Plugin method
claude plugin list | grep prompt-optimizer

# npm global install method
prompt-optimizer-skill version

# npx method
npx prompt-optimizer-skill version

# Symbolic link method
ls -la ~/.claude/skills/prompt-optimizer/
# Should see: SKILL.md, frameworks/, tests/, CHANGELOG.md, etc.

🚀 Usage

Method A: Use in Claude Code

In Claude Code conversation:

optimize this prompt: Write an email to a customer

Or:

/prompt-optimizer help me optimize this prompt

Method B: Use via npx Command Line

No installation needed, use directly:

# Basic usage
npx prompt-optimizer-skill "Write an email to a customer"

# Specify output version
npx prompt-optimizer-skill "Write code" -b    # Basic version
npx prompt-optimizer-skill "Write code" -e    # Enhanced version (default)
npx prompt-optimizer-skill "Write code" -x    # Expert version

# List all frameworks
npx prompt-optimizer-skill frameworks

# View help
npx prompt-optimizer-skill help

Workflow

User Input → Analysis → Framework Selection → Clarification → Optimization → Quality Validation → Output

Multi-Version Output

Based on your needs, you can get optimization results with different levels of detail:

| Version | Use Case | Characteristics | | :--- | :--- | :--- | | Basic | Quick use, simple tasks | Core elements, concise and clear | | Enhanced | Regular work, team collaboration | Complete structure with examples | | Expert | Complex projects, high-quality requirements | Full elements + constraints + validation criteria |

Examples

Example 1: Code Generation

Before:

write a function to sort a list

After (Expert Version):

# Role
You are a senior Python developer specializing in algorithms and data structures.

# Task
Implement a robust sorting function with comprehensive error handling.

# Instructions
1. Implement the function with type hints following Python 3.10+ standards
2. Include detailed docstrings with examples in Google style
3. Handle edge cases: empty list, None values, mixed types
4. Provide both ascending and descending sort options
5. Include time and space complexity analysis

# Constraints
- Use only Python standard library
- Follow PEP 8 style guidelines
- Include at least 5 test cases covering edge cases
- Keep function under 50 lines (excluding tests and docs)

# Output Format
1. Function implementation with type hints
2. Comprehensive docstring with examples
3. Complexity analysis table
4. Test suite with pytest-style assertions
5. Usage examples

# Quality Criteria
- Function passes all test cases
- Code is idiomatic Python
- Documentation is clear and complete

Example 2: Content Writing

Before:

write a blog post about AI

After (Enhanced Version):

# Role
You are a technology journalist and AI researcher with 10+ years of experience.

# Context
The blog targets software developers and tech professionals.

# Task
Write an informative blog post about practical AI tools for developers.

# Instructions
1. Start with a compelling hook about AI's impact
2. Cover 3-5 specific AI tools developers can use today
3. Include real-world examples and code snippets
4. Address common concerns and misconceptions
5. End with actionable next steps

# Constraints
- Target audience: intermediate to senior developers
- Tone: informative, practical, conversational
- Length: 1,500-2,000 words
- Avoid overly technical jargon

# Output Format
1. Catchy headline (under 60 characters)
2. Introduction with hook and thesis
3. Body sections with clear subheadings
4. Conclusion with call-to-action
5. Author bio (2-3 sentences)

📚 Framework Library

Simple Frameworks (≤3 Elements)

| Framework | Full Name | Use Case | | :--- | :--- | :--- | | APE | Action-Purpose-Expectation | Quick prompt building | | ERA | Expectation-Role-Action | Simple task instructions | | TAG | Task-Action-Goal | Quick task definition | | RTF | Role-Task-Format | Specific format output required | | BAB | Before-After-Bridge | Marketing promotion | | PEE | Point-Evidence-Explanation | Academic writing | | ELI5 | Explain Like I'm 5 | Complex concept explanation |

Medium Frameworks (4-5 Elements)

| Framework | Full Name | Use Case | | :--- | :--- | :--- | | RACE | Role-Action-Context-Expectation | Role-playing dialogue | | COAST | Context-Objective-Actions-Scenario-Task | AI dialogue system design | | ROSES | Role-Objective-Scenario-Expected Solution-Steps | Role-playing scenarios | | SMART | Specific-Measurable-Achievable-Relevant-Time-bound | Goal setting | | FOCUS | Features-Objective-Constraints-User-Setup | Product analysis and evaluation |

Complex Frameworks (6+ Elements)

| Framework | Full Name | Use Case | | :--- | :--- | :--- | | CRISPE | Capacity-Role-Insight-Statement-Personality-Experiment | Marketing campaign planning | | RACEF | Role-Action-Context-Expectation-Format | Complex analysis tasks | | RISEN | Role-Input-Steps-Expectation-Narrowing | Detailed plan formulation |

Detailed definitions can be found in the frameworks/ directory.

🧪 Testing

Run Tests

# Run full test suite
node tests/run-tests.js

# Or manually check
ls frameworks/simple/   # Should see 16 frameworks
ls frameworks/medium/   # Should see 33 frameworks
ls frameworks/complex/  # Should see 9 frameworks

Test Cases

View tests/test-cases.md for complete test coverage:

• Boundary condition tests (completely vague, ultra-short, meaningless inputs)
• Framework selection tests
• Clarification handling tests
• Quality validation tests
• Multi-version output tests
• Iterative optimization tests

🏗️ Project Structure

prompt-optimizer-skill/
├── SKILL.md                 # Main skill file
├── README.md                # This file
├── README_zh.md             # Chinese README
├── CHANGELOG.md             # Changelog
├── LICENSE                  # MIT License
├── VERSION                  # Current version
├── index.js                 # Module entry point
├── claude.json              # Claude Plugin configuration
├── package.json             # npm package manifest
├── bin/                     # CLI tools
│   └── prompt-optimizer.js  # CLI (template scaffolds + framework lookup)
├── frameworks/              # Framework library (61 total)
│   ├── index.json           # Structured metadata (id, category, elements, domains)
│   ├── simple/              # Simple frameworks (16)
│   ├── medium/              # Medium frameworks (33)
│   ├── complex/             # Complex frameworks (9)
│   └── patterns/            # Reusable patterns (3)
├── tests/                   # Test cases
│   └── test-cases.md
├── references/              # Reference materials
│   ├── Frameworks_Summary.md
│   ├── Quick_Reference.md
│   └── Decision_Tables.md
├── scripts/                 # Build & release scripts
│   ├── postversion.js
│   └── extract-changelog.js
└── .claude-plugin/          # Claude Plugin marketplace metadata
    └── marketplace.json

🤝 Contributing

Contributions are welcome! Please check CHANGELOG.md to understand the project's development direction.

Ways to Contribute

1. Add New Frameworks: Add new framework definitions under frameworks/
2. Improve Documentation: Enhance examples and descriptions for existing frameworks
3. Add Tests: Supplement with more test cases
4. Report Issues: Submit issues describing problems encountered

📄 License

MIT License - See LICENSE file for details.