Skip to main content

Validation Agent

Validate execution results against requirements and update step instructions for better outcomes.

What is the Validation Agent?​

The Validation Agent validates whether execution results meet the defined requirements. When results don't meet expectations, it updates the step instructions to help the executor agent generate better results on subsequent attempts.

How to use it?​

Configure the Validation Agent with these parameters:

  1. Select Language Model:

    • Choose models with strong analytical capabilities like OpenAI GPT-5, GPT-4o, or Claude
    • Ensure sufficient context length for analyzing errors and generating alternatives

  2. Configure Model Parameters:

    • Temperature: Lower values (0.2-0.4) for more consistent and focused analysis
    • Top P: Moderate values (0.8-0.9) for balanced analytical creativity
    • Max Tokens: Sufficient length for detailed analysis and improvement recommendations

  3. Custom System Prompt (Optional):

    • Define specific validation criteria and success requirements
    • Provide domain-specific guidance for result validation
    • Set expectations for how to improve step instructions

  4. Available Tools (Optional):

    • Add tools that may help with validation and testing
    • Tools from the master agent are automatically available unless overridden
    • Use tools for data inspection, quality checks, or validation tests