Instructions

The Instructions block breaks down a task into clear, sequential steps. It provides a roadmap for the AI to follow, ensuring complex tasks are handled systematically and nothing is missed.

When to Use

Use Instructions blocks when:

Complex multi-step tasks: Anything with more than 2-3 logical steps
Processes that require order: Steps that must happen in sequence
Quality-sensitive work: When you need thorough, methodical output
Reproducible workflows: Tasks you want handled consistently every time
Teaching or documentation: When the AI needs to explain its reasoning

How to Use

Add an Instructions block to your prompt canvas
Break the task into logical, sequential steps
Number the steps for clarity
Include decision points and conditions when needed
Place after Role, Context, and Goal blocks

Examples

Code Review Instructions

Follow these steps to review the code:

1. First, read through the entire code to understand its purpose
2. Check for security issues:
   - SQL injection vulnerabilities
   - XSS attack vectors
   - Exposed secrets or credentials
3. Evaluate code quality:
   - Naming conventions
   - Function length and complexity
   - DRY principle violations
4. Review error handling:
   - Are all errors caught?
   - Are error messages helpful?
5. Assess test coverage needs
6. Summarize findings with severity levels

Content Creation Instructions

Create the blog post following these steps:

1. Start with a hook that addresses the reader's pain point directly
2. Introduce the problem with a brief real-world example
3. Present your solution in 3-4 main sections:
   - Each section should have a clear subheading
   - Include one code example or screenshot per section
   - End each section with a key takeaway
4. Address common objections or questions
5. Conclude with actionable next steps
6. Add a compelling CTA related to our product

Data Analysis Instructions

Analyze the dataset as follows:

1. Data validation:
   - Check for missing values in each column
   - Identify outliers using IQR method
   - Report any data quality issues

2. Exploratory analysis:
   - Calculate summary statistics
   - Identify correlations between variables
   - Note any unexpected patterns

3. If issues are found in step 1:
   - Propose cleaning strategies
   - Document assumptions made

4. Generate visualizations:
   - Distribution plots for numeric columns
   - Correlation heatmap
   - Time series if date columns exist

5. Summarize key insights in bullet points

Tips

Use numbered lists: Easier to follow than paragraphs or bullets
One action per step: "Check X and validate Y and update Z" should be three steps
Include decision points: "If X, then do Y. Otherwise, do Z."
Specify output for each step when needed: "List all issues found" vs just "Check for issues"
Consider error cases: What should happen if a step fails or finds nothing?
Keep steps atomic: Each step should be completable before moving to the next
Test with edge cases: Do your instructions handle unusual inputs?

Instructions

On this page