scientific-method

djgrant's avatarfrom djgrant

When a task requires observation and verification, use this skill to extend work packages to deliver hypothesis-driven iteration.

⭐2stars🔀0forks📁View on GitHub🕐Updated Jan 8, 2026⬇️1downloads

When & Why to Use This Skill

The Scientific Method skill implements a rigorous, hypothesis-driven framework for Claude agents to solve complex problems. By transforming vague tasks into a structured cycle of observation, prediction, and verification, this skill ensures that every action is intentional and every result is documented. It is an essential tool for high-stakes troubleshooting, optimization, and analytical tasks where trial-and-error is insufficient.

Use Cases

  • Complex Debugging: Systematically identifying the root cause of software defects by forming hypotheses about code behavior and verifying them through controlled execution.
  • Prompt Optimization: Refining AI model performance by treating prompt adjustments as variables in an experiment to achieve specific output goals.
  • Root Cause Analysis: Investigating system failures or data anomalies by predicting potential failure points and observing actual system states to validate theories.
  • Performance Tuning: Iteratively testing infrastructure or algorithm configurations to optimize for speed, cost, or accuracy based on empirical evidence.
namescientific-method
descriptionWhen a task would benefit from hypothesis-driven iteration, load this skill to observe, predict, and verify.

::workflow::

DO

  1. evaluate current state
  2. hypothesize what will fix it
  3. predict the expected result
  4. act on your hypothesis
  5. observe the actual result
  6. repeat until goal achieved END

Tracking Template

Record each iteration in a work package:

### Iteration {n}
Hypothesis: {theory}
Action: {what you did}
Prediction: {expected result}
Observed: {actual result}
Conclusion: {validated or falsified}