experiment-design · yeachan-heo — Claude Code skill: what it does, how to install
experiment-design is a Claude Code skill that collects practices for reproducible experiments, from seeding to pre-registering an analysis plan. It supports runs and A/B tests another person must rerun.
no license 1 installs ★ 243 updated 2026-02-21
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a claude-codeWhat experiment-design does
The skill sets out a routine for experiments you can repeat and defend statistically. Templates are Python: seeding randomness, recording library versions, a stratified train/test split, cross-validation, and the frame of an A/B test with a control and a treatment group.
Sample size gets its own section. It calls the statsmodels power classes for a two-group t-test, one-way ANOVA and a proportions test, and adds a post-hoc check of achieved power so a thin sample isn't read as a null result. Those snippets need statsmodels and numpy installed.
Then comes pre-registration: write the analysis plan before touching the data, tag results as confirmatory or exploratory, keep a log of where the run drifted from the plan. That matters once the number lands in a report someone else will act on.
What it helps with
- seed a run and log library versions so it repeats
- size an A/B test before collecting data
- split data into stratified train and test sets
- label a result exploratory when it wasn't planned
- log where the analysis drifted from the pre-registered plan
Example prompts for the agent
- Plan the conversion A/B test: work out the sample size and write the analysis plan before any data comes in.
- Add seed fixing and version logging for Python, NumPy and Pandas to
train.py. - Tag the subgroup result as exploratory since it wasn't pre-registered.
How to install experiment-design
1Skills CLI (Vercel)
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a claude-code2Manually: put the files in place
~/.claude/skills/experiment-design/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/experiment-design/
1Skills CLI (Vercel)
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a codex2Manually: put the files in place
.agents/skills/experiment-design/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a cursor2Manually: put the files in place
.cursor/skills/experiment-design/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install Yeachan-Heo/My-Jogyo skills/experiment-design2Manually: put the files in place
.github/skills/experiment-design/SKILL.md1Skills CLI (Vercel)
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a gemini-cli2Platform CLI
gemini skills install https://github.com/Yeachan-Heo/My-Jogyo --path skills/experiment-designManually: put the files in place
.devin/skills/experiment-design/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about experiment-design
How do I install experiment-design in Claude Code?
Run npx skills add Yeachan-Heo/My-Jogyo --skill experiment-design -a claude-code. The skill file lands at ~/.claude/skills/experiment-design/SKILL.md.
What does experiment-design do?
It lays out how to run an experiment: seed and version logging, a controlled data split, power analysis, a pre-registered plan, and confirmatory versus exploratory labels on findings.
What do I need for the power calculation?
The examples rely on statsmodels and numpy in a Python environment. Sizing happens before data collection, and a separate check measures achieved power against the sample you actually got.
Similar skills
All skills for Science and research →experiment-design by other authors comes first.