A complete Claude Skill for prompt engineering expertise, located at:
~/Documents/prompt-engineering-expert/
- Go to Claude.com and start a conversation
- Click the "+" button next to the message input
- Select "Upload a Skill"
- Choose the skill folder:
~/Documents/prompt-engineering-expert/ - Claude will load the skill and you can start using it
- Open Claude Code in Claude.com
- Create a new project
- Upload the skill folder to your project
- Reference the skill in your prompts
from anthropic import Anthropic
client = Anthropic()
# Load the skill
skill_path = "~/Documents/prompt-engineering-expert"
# Use in your agent
response = client.messages.create(
model="claude-opus-4-1",
max_tokens=2048,
system=f"You have access to the prompt engineering expert skill at {skill_path}",
messages=[
{
"role": "user",
"content": "Review this prompt and suggest improvements: [PROMPT]"
}
]
)Once uploaded, you can ask Claude:
"Review this prompt and suggest improvements:
[YOUR PROMPT]"
"Using your prompt engineering expertise:
1. Analyze this prompt: [PROMPT]
2. Identify issues
3. Suggest specific improvements
4. Provide a refined version
5. Explain what changed and why"
"Design custom instructions for an agent that:
- Analyzes customer feedback
- Extracts key themes
- Generates recommendations
- Outputs structured JSON"
"This prompt isn't working:
[PROMPT]
Issues I'm seeing:
- [ISSUE 1]
- [ISSUE 2]
How can I fix it?"
- SKILL.md - Metadata and overview
- CLAUDE.md - Main instructions
- README.md - User guide
- docs/BEST_PRACTICES.md - Best practices guide
- docs/TECHNIQUES.md - Advanced techniques
- docs/TROUBLESHOOTING.md - Common issues
- examples/EXAMPLES.md - Real-world examples
- INDEX.md - Complete index and navigation
- SUMMARY.md - What was created
"Analyze this prompt for clarity and effectiveness:
'Summarize this article'
What could be improved?"
"Create a prompt for analyzing customer support tickets.
The prompt should:
- Classify tickets by category
- Extract key issues
- Suggest responses
- Output as JSON"
"I'm creating a custom instruction for an AI agent.
Here's my draft:
[YOUR DRAFT]
Please improve it using prompt engineering best practices."
"My prompt keeps producing inconsistent results.
Here's the prompt:
[YOUR PROMPT]
What's wrong and how do I fix it?"
- Start with README.md
- Read docs/BEST_PRACTICES.md (Core Principles)
- Review examples/EXAMPLES.md (Examples 1-3)
- Read docs/TECHNIQUES.md (Sections 1-4)
- Review examples/EXAMPLES.md (Examples 4-7)
- Use docs/TROUBLESHOOTING.md as needed
- Study docs/TECHNIQUES.md (All sections)
- Review examples/EXAMPLES.md (All examples)
- Combine multiple techniques
- Prompt writing best practices
- Advanced techniques (CoT, few-shot, XML, etc.)
- Custom instructions design
- Prompt optimization
- Anti-pattern recognition
- Evaluation frameworks
- Multimodal prompting
- Analyze existing prompts
- Generate new prompts
- Refine and optimize
- Design custom instructions
- Teach best practices
- Identify issues
- Develop test cases
- Create documentation
Edit examples/EXAMPLES.md to add examples for your domain.
Add domain-specific best practices to docs/BEST_PRACTICES.md.
Add common issues to docs/TROUBLESHOOTING.md.
prompt-engineering-expert/
├── INDEX.md # Navigation guide
├── SUMMARY.md # What was created
├── README.md # User guide
├── SKILL.md # Metadata
├── CLAUDE.md # Main instructions
├── docs/
│ ├── BEST_PRACTICES.md # Best practices
│ ├── TECHNIQUES.md # Advanced techniques
│ └── TROUBLESHOOTING.md # Troubleshooting
└── examples/
└── EXAMPLES.md # Examples
- Comprehensive documentation
- Real-world examples
- Best practice checklists
- Troubleshooting guides
- Quick reference tables
- Claude Docs: https://docs.claude.com
- Anthropic Blog: https://www.anthropic.com/blog
- Claude Cookbooks: https://github.com/anthropics/claude-cookbooks
- Start Simple - Begin with basic prompts before advanced techniques
- Use Examples - Provide examples to guide Claude's responses
- Be Specific - The more specific your request, the better the results
- Test Thoroughly - Always test refined prompts with real data
- Iterate - Use feedback to continuously improve
- Document - Keep notes on what works for your use case
- Upload the skill using one of the methods above
- Try a simple example to get familiar with it
- Review the documentation for deeper learning
- Apply to your prompts and iterate
- Share with your team for collaborative improvement
- Check INDEX.md for navigation
- Review TROUBLESHOOTING.md for common issues
- Look at EXAMPLES.md for similar cases
- Read BEST_PRACTICES.md for guidance
Q: How do I upload the skill? A: See "Option 1: Upload via Claude.com" above
Q: Can I customize the skill? A: Yes! Edit the markdown files to add domain-specific content
Q: What if my prompt still doesn't work? A: Check TROUBLESHOOTING.md or ask Claude to debug it
Q: Can I use this with the API? A: Yes! See "Option 3: Use with Agent SDK" above
Your Prompt Engineering Expert Skill is ready to use. Start by uploading it and asking Claude to review one of your prompts!
Questions? Check the documentation files or ask Claude directly!