context-engineering

Using XML tags to structure complex inputs helps Claude distinguish between different data types, instructions, and examples within a single context window. This Skill provides patterns for applying XML tagging to improve Claude's comprehension and output accuracy.

Implementing chain-of-thought reasoning in Claude prompts involves structuring instructions to encourage step-by-step problem solving. This Skill offers detailed patterns and best practices for applying chain-of-thought and extended thinking to enhance complex problem-solving accuracy.

Writing an effective system prompt for an LLM agent requires clear instructions, role prompting, and prefilling responses to guide behavior. This Skill provides comprehensive guidance on designing system prompts and optimizing context windows for maximum consistency.

Multishot examples can be used to improve Claude's accuracy by providing clear demonstrations of the desired input-output mapping. This Skill provides guidance on integrating multishot examples into your prompt engineering workflow.

Claude may ignore parts of a long prompt when complex inputs are poorly structured within the context window. This Skill addresses challenges in structuring complex inputs and optimizing context for accuracy using XML tagging and clear instruction design.

Prefilling responses is a supported prompt engineering technique for Claude API calls that constrains the model's output format and initial reasoning. This Skill details how to implement prefilling alongside role prompting and multishot examples.