CI LAW - Prompt Engineering

 

1	Be Explicit with Instructions	Claude 4 responds best to clear, specific instructions rather than implicit expectations	Instead of vague requests, specify exactly what you want
2	Add Context & Motivation	Provide background on why the task matters to help Claude understand your goals	Explain the purpose: "This is for a client presentation, so accuracy is critical"
3	Use Positive Instructions	Tell Claude what TO do instead of what NOT to do	"Write in flowing prose paragraphs" vs "Don't use markdown"
4	Leverage XML Format Tags	Use XML-style tags to control output structure and formatting	<smoothly_flowing_prose_paragraphs>content</smoothly_flowing_prose_paragraphs>
5	Match Prompt Style to Output	Your prompt formatting influences Claude's response style	Remove markdown from prompts if you want plain text output
6	Be Vigilant with Examples	Claude 4 pays close attention to examples - ensure they align with desired behavior	All examples should demonstrate exactly what you want to see
7	Utilize Thinking Capabilities	Guide Claude's reasoning process for complex multi-step tasks	Enable thinking mode for reflection and complex problem-solving
8	Frame with Quality Modifiers	Add descriptors that encourage higher quality, detailed output	"Create a comprehensive analytics dashboard with advanced features"
9	Request Features Explicitly	Don't assume Claude will include features - ask for them directly	"Include animations, hover states, and interactive elements"
10	Avoid Test-Focused Solutions	Prevent Claude from hard-coding solutions just to pass tests	Guide toward robust, generalizable solutions rather than quick fixes

 

 

Master Claude 4 by being explicit, contextual, and positive in your instructions - specificity is your superpower for unlocking optimal AI performance.

 

https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/claude-4-best-practices

Claude 4 prompt engineering best practices - Anthropic