Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/danielrosehill/claude-data-analyst-plugin/sample-sizenpx skills add danielrosehill/Claude-Data-Analyst-plugin --skill sample-sizegit clone --depth 1 https://github.com/danielrosehill/Claude-Data-Analyst-pluginWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/danielrosehill/claude-data-analyst-plugin/sample-size)<a href="https://agentmods.dev/skills/danielrosehill/claude-data-analyst-plugin/sample-size"><img src="https://agentmods.dev/badge/skills/danielrosehill/claude-data-analyst-plugin/sample-size.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00062 | $0.01330 |
| Opus 5 | $0.00031 | $0.00665 |
| Sonnet 5 | $0.00012 | $0.00266 |
| Haiku 4.5 | $0.00006 | $0.00133 |
Grade A, and why
sample-size scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 92 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Sample Size
Row count is not sample size. This skill characterises the effective sample the user has for the specific analysis they want to run, and tells them whether that's enough.
Inputs
- Path to a dataset (CSV / Parquet / Excel / DuckDB table).
- Optional: the analytical question (e.g. "compare conversion rate between A and B", "estimate mean spend per country", "train a classifier on 12 features"). If not provided, produce a general description only.
- Optional: target column, predictor columns, grouping columns, effect size / precision the user cares about.
Recommended CLI tooling
duckdb— counts, group-by cardinality, null rates.uv run --with statsmodels python -c '...'— power calculations (statsmodels.stats.power), sample-size-for-proportions, sample-size-for-regression.
Procedure
Step 1 — Raw size description
Report:
- Total rows and total columns.
- Non-null rows per column (effective n per variable).
- Unique rows (if duplicates matter — e.g. a "customers" table with repeated IDs isn't really that many customers).
- Time span (if a date column exists): first timestamp, last timestamp, duration, record density (rows/day).
- Memory footprint — helpful for deciding between in-memory vs. chunked processing.
Step 2 — Unit of analysis
Ask (or infer): what is one row? Customer? Transaction? Country-year? Sensor reading?
The honest sample size for most claims is the number of independent units, not the number of rows. If the dataset has 1M transactions from 50 customers and the question is about customer behaviour, n = 50, not 1,000,000. Surface this distinction — it's the single biggest source of misstated power.
Step 3 — Per-group and per-class cells
For each grouping column or categorical target the user mentions:
- Group sizes: the count of rows in each level.
- Smallest cell: the minimum count across groups. This is the binding constraint on any comparison.
- Imbalance ratio: largest / smallest. Flag if > 10:1 — statistical tests and ML models both degrade.
- For cross-tabs (two categorical vars): the smallest cell in the contingency table. Chi-square starts being unreliable when any expected cell is < 5.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 3d ago First seen · 92 lines · 62 tokens per session scan A 55c440ea7c4b
sample-size is a skill published in the GitHub repository danielrosehill/Claude-Data-Analyst-plugin (11 stars, last pushed 4mo ago), licensed MIT. It adds 62 tokens to every session and 1,330 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
ppt-analysis
PPT (.pptx/.ppt) 全量解析。覆盖：所有 slide 文本/表格/图表提取、嵌入图片 caption、纯图片 slide 渲染识别、数据标签提取。.
large-file-conditional-formatting
根据Excel总行数自动切换Parquet加速读取，计算特定维度的时间序列平均值，并使用openpyxl输出带有条件格式（如低于均值标绿）和自定义样式的分析报告。.
trend-analysis
基于多维度数据进行分级评估与趋势预测，通过设定差异化增长率计算预测值，并生成对比可视化图表，适用于绩效评估、目标设定等场景。.
outlier-detection-and-quality-assessment
执行全面的异常值检测与数据质量评估，利用 IQR 方法识别异常值并结合偏度、峰度分析数据分布特征，适用于非正态分布数据的预处理阶段。.
pie-chart-data-analysis
Step1 读取文件并统计所有 Sheet 的行数，确认数据规模以决定处理策略。.
multi-sheet-reading-and-analysis
用于读取多工作表Excel文件，动态评估数据量以启用Parquet大文件优化，并执行正则清洗、分类汇总、线性拟合及生成带格式的图表与结果文件。.