feat: add experiment-designer skill

[xjrmh]

Summary

• Adds an experiment-designer skill that guides users through designing statistically rigorous experiments via a structured 8-step workflow
• Supports 6 experiment types: A/B test, cluster randomized, switchback, causal inference, factorial, and multi-armed bandit
• Includes reference files for statistical formulas, experiment type selection, and a common metrics library with baselines

Files

File	Purpose
skills/experiment-designer/SKILL.md	Main skill — 8-step workflow, inference rules, output template
skills/experiment-designer/experiment-types.md	Detailed reference for 6 experiment types
skills/experiment-designer/statistics.md	Sample size formulas, type-specific adjustments, duration estimation
skills/experiment-designer/metrics-library.md	14 common metrics with baselines and variance values

How it works

Invoke /experiment-designer and describe what you want to test. The skill walks through:

1. Experiment type selection (inferred from context)
2. Metrics definition (PRIMARY + GUARDRAIL required)
3. Statistical parameters & sample size calculation
4. Randomization strategy
5. Variance reduction (CUPED, stratification, etc.)
6. Risk assessment & pre-launch checklist
7. Monitoring & stopping rules
8. Summary — produces a structured experiment design document

Test plan

• Install skill via cp -r skills/experiment-designer ~/.claude/skills/
• Invoke /experiment-designer I want to test a new checkout flow and verify it walks through all 8 steps
• Verify statistical formulas produce correct sample size calculations
• Confirm the final output is a well-structured markdown design document

🤖 Generated with Claude Code