Dev Sciomc (Scientific Method)

Derived from oh-my-claudecode (MIT, Yeachan Heo). Adapted for the EvoNexus Engineering Layer.

Scientific method discipline applied to engineering investigations. Forces explicit hypothesis statement, experimental design, evidence collection, and provisional conclusions.

Use When

• Investigation requires rigor beyond "let me try X"
• Performance optimization (you need controls and measurements, not guesses)
• A/B comparison of two implementations
• Anything where the cost of being wrong is high

Do Not Use When

• Trivial bug → use @hawk-debugger
• Pure exploration → use @scout-explorer

Workflow

Phase 1 — Hypothesis

• State the hypothesis as a falsifiable claim
• "X is faster than Y" not "X feels faster"
• Identify the dependent variable, independent variables, controls

Phase 2 — Experiment Design

• What measurement will prove/disprove the hypothesis?
• What's the minimum sample size for statistical significance?
• What confounders need to be controlled?

Phase 3 — Evidence Collection

• Run the experiment
• Collect raw data
• Note environmental factors that could affect results

Phase 4 — Analysis

• Apply statistical tests (delegate to @prism-scientist)
• Calculate effect size, CI, p-value
• Compare against the hypothesis

Phase 5 — Conclusion

• Provisional, never absolute
• State limitations
• Identify follow-up experiments

Output

Saved to workspace/development/research/[C]sciomc-{topic}-{date}.md:

markdown
## Scientific Investigation — {topic}

### Hypothesis
{Falsifiable claim}

### Experimental Design
- Dependent variable: {what we measure}
- Independent variables: {what we vary}
- Controls: {what we hold constant}
- Sample size: {N}

### Method
{Step-by-step protocol}

### Results
{Raw data summary}

### Statistical Analysis
[delegated to @prism-scientist]

### Conclusion
{Provisional conclusion + limitations}

### Follow-ups
- {next experiment}

Pairs With

• @prism-scientist (for statistical analysis)
• @trail-tracer (when investigation is causal)
• @apex-architect (when conclusion implies architecture change)