Hypothesis Generation

What it does

This skill enables AI agents to systematically formulate testable hypotheses based on provided context or data. It structures the reasoning process to move from observations to specific, falsifiable predictions suitable for experimentation.

When to use it

• Root cause analysis: When investigating why a system is failing or underperforming without clear error logs.
• Product strategy: During brainstorming sessions to generate potential features or user behavior predictions before building.
• Scientific research: To draft initial experimental designs and expected outcomes for new variables.
• Troubleshooting: When debugging complex codebases where the failure point is not immediately obvious from symptoms.

Key capabilities

• Analyzes provided context to identify patterns or anomalies.
• Generates multiple distinct hypotheses rather than a single assumption.
• Structures outputs to include the core prediction and implied conditions for testing.

Example prompts

• "Analyze these server logs and generate three hypotheses explaining the spike in latency at 2:00 PM."
• "Given our user feedback on the new checkout flow, propose hypotheses about where users are dropping off."
• "I observed that the model performs poorly on images with low contrast. Generate a hypothesis for why this is happening."

Tips & gotchas

Ensure you provide sufficient context or data points; vague inputs will result in generic hypotheses. This skill is designed to spark ideas, not validate them—you must still design experiments to test the generated predictions.