Separate Data from Instructions
Independent PiSkill directory guide. The original prompt remains hosted by Anthropic Courses.
What does this prompt do?
Uses structural delimiters to distinguish untrusted or variable input data from the instructions the model should follow.
Primary use case
Safely place documents or user content inside a reusable instruction template.
Expected output
A response based on clearly delimited source data.
Inputs or variables
- instructions
- source data
Related prompts
Few-Shot Pattern Prompt
Uses representative input-and-output examples to teach the model a task pattern and desired response style.
Clear and Direct Instructions
Demonstrates how to turn an ambiguous request into a direct prompt with explicit context, constraints and expected behavior.
Complex Prompt Construction
Combines role, context, examples, constraints and output formatting into one reusable prompt workflow.
Role-Based Expert Response
Shows how assigning a relevant role changes the context and framing used for a specialized response.
Hallucination-Resistant Answer
Adds grounding and uncertainty instructions so the model can distinguish supported answers from missing information.
Implementation Prompt Author
Creates an execution-ready implementation prompt from approved requirements and design evidence, including scope, files, checks, and stopping conditions.