Claude Models Prompt Guide
1. Be Explicit and Detailed in Instructions
Claude responds more effectively when instructions leave little ambiguity. General requests often produce minimal results, while explicit ones yield richer, higher-quality outputs.
Example: Creating an analytics dashboard
Less effective:
Create an analytics dashboard
More effective:
Create an analytics dashboard. Include as many relevant features and interactions as possible. Go beyond the basics to create a fully-featured implementation.
Best Practice: Add modifiers such as “fully featured,” “detailed,” “professional tone,” or “interactive with animations” when those qualities are desired.
2. Add Context to Improve Performance
Claude uses contextual information to align its output with your goals. The more Claude knows about the use case, the better the results.
Example: Formatting Preferences
Less effective:
NEVER use ellipses
More effective:
Your response will be read aloud by a text-to-speech engine, so never use ellipses since the text-to-speech engine will not know how to pronounce them.
Best Practice: Provide Claude with:
The purpose of the task (e.g., “quarterly report for executives”).
The audience (e.g., “non-technical customers”).
The workflow where the output will be used (e.g., “feeds into a slide deck”).
The success criteria (e.g., “clear, jargon-free summary under 300 words”).
3. Provide Instructions as Sequential Steps
Breaking tasks into steps improves reliability and reduces missed requirements.
Example: Anonymizing Customer Feedback
Unclear prompt:
Please remove all personally identifiable information from these customer feedback messages: {DATA}
Clear prompt with steps:
Your task is to anonymize customer feedback for our quarterly review.
Instructions:
Replace all customer names with CUSTOMER_[ID]
Replace email addresses with EMAIL_[ID]@example.com
Redact phone numbers as PHONE_[ID]
If a message mentions a specific product, leave it intact
If no PII is found, copy the message verbatim
Output only the processed messages, separated by "---"
Data to process: {DATA}
4. Leverage Claude’s Thinking Capabilities
Claude 4 supports thinking and interleaved thinking, which is especially useful for multi-step reasoning or when reflecting on tool outputs.
Best Practice Prompt:
After receiving tool results, carefully reflect on their quality and determine optimal next steps before proceeding. Use your thinking to plan and iterate based on this new information, and then take the best next action.
5. General Prompting Checklist
Before sending a prompt, ask:
Have I provided purpose, audience, workflow, and success criteria?
Did I state exactly what Claude should and should not output?
Are instructions broken into clear, sequential steps?
Did I add quality modifiers where needed (e.g., detailed, concise, interactive)?
Do I want Claude to reflect and plan next actions?
If coding, did I tell Claude whether to clean up temporary files?
Quick Prompt Templates for Claude
1. Task-Oriented Prompt
Use this when you need Claude to carry out a specific task with a clear outcome.
Template:
Context: [Explain where this will be used and who the audience is]
Task: [Say exactly what you want Claude to do, add modifiers like “detailed,” “concise,” “fully featured”]
Steps:
[Step one of the task]
[Step two of the task]
[Step three of the task, etc.]
Constraints: [Set limits—tone, format, length, what to avoid]
Optional: [Ask Claude to reflect on the result and improve if needed]
Sample:
Context: This will be part of a board presentation for executives with no technical background.
Task: Summarize the attached technical report into 3 clear bullet points.
Steps:
Find the 3 most important insights.
Remove all technical jargon and keep it simple.
Keep each bullet under 20 words.
Constraints: Output only the 3 bullets, no intro or conclusion.
2. Instruction + Reason Prompt
Use this when formatting or rules are important and you want Claude to understand why.
Template:
Task: [Describe the task or rule you want followed]
Reason: [Explain why the rule matters—Claude will follow it more reliably]
Example:
Task: Do not use ellipses (…) in your response.
Reason: The output will be read aloud by a text-to-speech system, which cannot pronounce ellipses correctly.
3. Step-by-Step Workflow Prompt
Use this when the task involves multiple operations or transformations (like cleaning data).
Template:
Task: [State the overall goal]
Instructions:
[Step one instruction]
[Step two instruction]
[Step three instruction]
…
Final Output: [Specify exactly how the results should be presented]
Example:
Task: Anonymize customer feedback for our quarterly review.
Instructions:
Replace all customer names with CUSTOMER_[ID].
Replace email addresses with EMAIL_[ID]@example.com.
Replace phone numbers with PHONE_[ID].
Leave product names unchanged.
If no personal info is found, copy the text as-is.
Final Output: Show only the processed messages, separated by "---".
Last updated