HUMANS.claude.md v0.0.1
This document provides guidance for humans collaborating with Claude, Anthropic's AI assistant, through the web interface, mobile app, or desktop application.
- Back to main HUMANS.md
- Claude's Agent Instructions (to be created)
Claude is an AI assistant created by Anthropic. This iteration is Claude Sonnet 4 from the Claude 4 model family, designed to be a smart, efficient model for everyday use. Claude is accessible through web-based, mobile, and desktop chat interfaces.
Claude excels at a wide variety of tasks including:
- Analysis & Research: Breaking down complex problems, analyzing documents, and synthesizing information
- Writing & Communication: Creating content, editing text, drafting emails, and improving clarity
- Reasoning & Problem-Solving: Working through logical problems, providing step-by-step explanations
- Creative Tasks: Writing stories, brainstorming ideas, and creative problem-solving
- Learning & Teaching: Explaining concepts clearly with examples and analogies
- Code & Technical Tasks: Writing, debugging, and explaining code in various programming languages
- Provide clear context and specific requirements
- Include relevant background information
- Specify your desired format, length, or style when applicable
- Provide positive examples of what you want
- Include negative examples of what to avoid when helpful
- Show Claude the pattern you're looking for
- Ask Claude to "think step by step" for complex problems
- Request explanations of reasoning when needed
- Ask for clarification of any unclear steps
- Request specific XML tags, markdown formatting, or structure when needed
- Specify if you want bullet points, numbered lists, or prose
- Indicate desired length (brief, detailed, comprehensive)
- Claude can read and analyze uploaded files including text documents, PDFs, images, and spreadsheets
- Provide context about what you want Claude to do with the file
- Be specific about which parts of the document to focus on
- Ask Claude to summarize key points, extract specific information, or identify patterns
- Request comparisons between multiple documents
- Have Claude check for errors, inconsistencies, or missing information
- Provide genre, tone, and style preferences
- Share target audience information
- Specify length and format requirements
- Ask for multiple versions or iterations
- Start with broad topics and narrow down
- Ask Claude to build on your ideas
- Request different perspectives or approaches
- Combine and refine ideas iteratively
- Specify programming language and framework
- Provide context about your project and goals
- Include relevant code snippets or error messages
- Ask for explanations of complex concepts
- Break down complex problems into smaller parts
- Ask for multiple solution approaches
- Request pros and cons of different options
- Iterate on solutions based on feedback
- Be specific about what type of information you need
- Ask Claude to search for recent information when needed
- Request source verification for important claims
- Ask for different perspectives on controversial topics
- Ask Claude to evaluate arguments, evidence, and sources
- Request identification of biases, assumptions, or gaps
- Have Claude compare different viewpoints
- Ask for balanced assessments of complex issues
- Reference previous parts of the conversation when building on earlier topics
- Remind Claude of important context if the conversation is long
- Ask Claude to summarize key points from earlier in the discussion
- Politely correct any mistakes Claude makes
- Provide clarification if Claude misunderstands your request
- Ask follow-up questions if responses aren't quite right
- Provide specific feedback on Claude's responses
- Ask for revisions with specific changes
- Build on partial solutions iteratively
- Use the thumbs up/down buttons to provide feedback to Anthropic
- Claude's training data has a cutoff at the end of January 2025
- For current events or recent developments, Claude will use web search when appropriate
- Ask Claude to search for recent information when needed
- Claude cannot browse the internet independently (except through search)
- Claude cannot remember conversations between separate chat sessions
- Claude cannot perform actions outside the chat interface
- Claude cannot access external systems, databases, or APIs
- Always verify important information, especially for critical decisions
- Ask Claude to explain its reasoning for important conclusions
- Cross-check information from multiple sources when accuracy is crucial
- Be aware that Claude can make mistakes and should not be the sole source for critical information
- Be mindful of sensitive personal information in conversations
- Remember that conversations may be used to improve AI systems
- Don't share passwords, financial information, or other sensitive data
- Claude is designed to be helpful, harmless, and honest
- It won't assist with illegal activities, harmful content, or deceptive practices
- If Claude declines a request, consider rephrasing or providing more context
- Use role-playing: "Act as a [expert/teacher/critic] and..."
- Chain of thought: "Let's work through this step by step..."
- Multiple perspectives: "Consider this from the viewpoint of..."
- Structured output: "Organize your response with clear headings..."
- Break large projects into smaller, manageable pieces
- Create templates or frameworks for repeated tasks
- Use Claude to help plan and organize complex work
- Iterate and refine outputs through multiple rounds
- Ask open-ended questions for broader exploration
- Use specific questions for targeted information
- Follow up with clarifying questions
- Build on previous answers to go deeper
- Treat Claude as a thinking partner, not just a tool
- Engage in back-and-forth dialogue
- Challenge ideas and ask for alternatives
- Use Claude to check your own thinking
- Provide more specific context and examples
- Break down complex requests into smaller parts
- Rephrase your question in different ways
- Ask Claude what additional information would be helpful
- Provide specific feedback on what needs to change
- Ask for alternative approaches or perspectives
- Clarify your goals and requirements
- Try a different angle or approach to the problem
For more comprehensive information on working effectively with Claude:
- Anthropic's prompting documentation: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/overview
- Support and help: https://support.anthropic.com
- API documentation: https://docs.anthropic.com
Claude is designed to be a helpful collaboration partner. The more context, clarity, and specificity you provide, the better Claude can assist you. Don't hesitate to ask follow-up questions, request clarifications, or iterate on solutions until you get exactly what you need.