exploratory-data-analysis
Skill8 starsChanged 4 days agoMITView on GitHub
Use when exploring a dataset before formal analysis, modeling, reporting, or dashboarding. Runs a bundled zero-dependency data health check on CSV files that infers field types, detects encoding, pseudo-nulls, numbers stored as text, leading-zero IDs, duplicate keys, partition-vs-event-time drift, missing dates, and orphan join keys, ranks issues P0/P1/P2, and renders an HTML report; then interprets grain, usability, and next analysis directions. 中文触发词:数据探索、EDA、数据体检、数据摸底、拿到一份新数据。
27 other projects use a version of this.
Adopted by
Most recently taken up by
- BrightGold70/skillstheir own version, first seen Nov 8, 2026
- SWEStash/swe-workflow-skillstheir own version, first seen Nov 8, 2026
- MarieLynneBlock/arcanum-artifextheir own version, first seen Nov 1, 2026
- JPeetz/agent-skillstheir own version, first seen Nov 1, 2026
- Big-LAORE/Biblioth-que-de-skillstheir own version, first seen Nov 1, 2026
- PaulDanielYoung/internal-audit-skillstheir own version, first seen Nov 1, 2026
Other versions27 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 27
- wu-yc/LabClaw★ 1.1ksame name
- LeonChaoX/qinyan-academic-skills★ 910same name
- jaechang-hits/SciAgent-Skills★ 386same name
- qinyan-ai/qinyan-academic-skills★ 80same name
- Barrios88/barrios-skills★ 63same name
- ZanderRuss/obsidian-claude★ 19same name
- JPeetz/agent-skills★ 14same name
- fmschulz/omics-skills★ 9same name
- MarieLynneBlock/arcanum-artifex★ 4same name
- JunMA98/Computer-science-claude-skills★ 3same name
17 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository23 other files it gives its agents.
- ab-test-analysisSkill
- business-root-cause-analysisSkill
- daily-report-writerSkill
- dashboard-reviewerSkill
- data-analysis-report-writerSkill
- data-doc-writerSkill
- data-incident-postmortem-writerSkill
- data-lineage-impact-analyzerSkill
- data-presentation-architectSkill
- data-quality-rule-generatorSkill
- data-requirement-clarifierSkill
- data-tool-integration-plannerSkill
- data-viz-kitSkill
- feature-engineering-advisorSkill
- funnel-analysisSkill
- market-research-analystSkill
- metric-definition-reviewerSkill
- model-evaluation-reviewerSkill
- retention-cohort-analysisSkill
- sql-reviewerSkill
- stream-pipeline-designerSkill
- table-design-advisorSkill
- weekly-monthly-report-writerSkill
The file
Files
Exploratory Data Analysis
目标
拿到一份新数据,先回答三个问题:
- 一行代表什么?(粒度、主键、时间范围)
- **能不能直接用?**哪些问题会让结果算错,哪些需要在口径里处理?
- 值得往哪个方向继续分析?
核心原则:数据长什么样由脚本说,数据意味着什么由人判断。scripts/eda.py 负责体检和问题分级;模型负责推断字段含义、确认粒度、区分数据问题和真实业务现象、规划下一步。
使用场景
- 拿到一张新表、一份 CSV 或 Excel 导出,先摸底
- 判断数据能否支撑某个分析、看板或模型
- 接手别人的数据,快速发现坑
- 多张表要关联之前,检查键是否对得上
- 分析结果“看起来不对”,回头排查数据本身
典型输入:
请帮我对这份订单数据做 EDA,看哪些字段有质量问题,能不能用来做渠道分析。不适用场景
- 漏斗、留存、实验等专项分析:体检完成后分别使用
funnel-analysis、retention-cohort-analysis、ab-test-analysis - 为一张表设计长期运行的质量监控规则:使用
data-quality-rule-generator(可以把体检发现的问题作为输入) - SQL 逻辑审查:使用
sql-reviewer - 写完整分析报告:使用
data-analysis-report-writer
只有业务问题、没有任何数据或字段时,先列出需要的数据,不要假装已经完成 EDA。
输入信息
最少输入:
- 数据文件(CSV、TSV 或 xlsx,UTF-8 和 GBK 都可以),或字段列表 + 样例
推荐输入:
- 分析目标:这份数据接下来要用来做什么
- 主键:一行应该由哪些字段唯一确定
- 多表时的关联关系
- 时间字段、分区字段
- 已知问题
上下文建议
推荐使用 数据分析上下文模板;表结构信息可参考 数据表上下文模板。
最有价值的上下文是:分析目标、预期粒度和主键、多表关联关系。没有真实数据时,只能输出检查清单,不能声称发现了具体的缺失率或分布。
工作流程
运行体检(有数据文件且能执行代码时必须运行):
python3 scripts/eda.py profile \ --table orders=orders.csv --table users=users.csv \ --key orders=order_id --key users=user_id \ --join orders.user_id=users.user_id \ --html eda_report.html- 组合主键:
--key events=user_id+event_time+event_name - 指定时间字段:
--time-col orders=pay_time(默认自动选择非分区的时间列) - 大文件:
--max-rows 500000在全表上随机抽样,报告会标注;主键和关联检查建议全量运行 - xlsx 有多个工作表时:
--sheet orders=订单明细 - 业务时区不是北京时间时:
--tz -05:00 - 全部参数见 命令行参数
- 组合主键:
- 确认粒度:根据主键检查结果和字段画像,写出“一行代表 ___”。主键重复时先弄清是重复数据还是粒度理解错了。
- 处理问题清单:脚本已按 P0 / P1 / P2 分级,每条带影响和建议。逐条判断是数据问题还是业务现象(例如负金额可能是退款,突增可能是大促)。检查项说明见 检查项说明。
- 推断字段含义:类型是脚本推断的,含义要结合字段名和取值推测,并明确标注“推测”。
- 规划下一步:这份数据适合做哪些分析、需要先补什么、该用哪个 Skill。
- 交付:Markdown 结论 + HTML 报告路径。
输出格式
## EDA 结果
### 1. 结论
<能不能用:可以使用 / 修正后可用 / 不建议直接使用,以及最关键的 1~3 个原因>
### 2. 数据概览
| 表 | 一行代表 | 行数 | 主键 | 时间范围 | 编码 |
| --- | --- | --- | --- | --- | --- |
### 3. 必须先处理的问题(P0)
| 表.字段 | 问题 | 影响 | 处理方式 |
| --- | --- | --- | --- |
### 4. 需要在口径中处理的问题(P1)
### 5. 字段画像与含义推测
| 字段 | 推断类型 | 含义(推测) | 缺失率 | 备注 |
| --- | --- | --- | --- | --- |
### 6. 值得注意的模式
<分布、趋势、关联中的发现;区分数据问题和业务现象>
### 7. 可继续分析的方向
| 方向 | 需要的前置处理 | 推荐 Skill |
| --- | --- | --- |
### 8. 待确认问题
报告文件:<eda_report.html 路径>质量标准
输出必须:
- 先说明粒度和能否使用,再列细节
- 所有缺失率、重复数、分布数字来自脚本输出,或明确标注为“未实际运行”
- 区分数据质量问题和真实业务异常(负金额、突增、长尾都可能是真的)
- 字段含义不确定时标注为推测
- 不把相关写成因果
- 每个 P0 问题都给出具体处理方式
- 给出下一步分析方向和对应 Skill,而不是只罗列统计量
完成后的下一步
- 专项分析:
funnel-analysis、retention-cohort-analysis、ab-test-analysis - 建模:
feature-engineering-advisor - 把体检发现变成长期校验:
data-quality-rule-generator - 写报告:
data-analysis-report-writer
示例 Prompt
请用 exploratory-data-analysis 体检这两份导出数据,判断能不能做渠道 GMV 分析:
orders.csv:订单导出(Excel 另存,可能是 GBK 编码),主键 order_id
users.csv:用户表,主键 user_id
关联:orders.user_id = users.user_id请用 exploratory-data-analysis 分析下面这张 SaaS 订阅变更事件表:
粒度:一行一条订阅状态变更事件
字段:company_id, event_time, event_type, plan, amount, prev_plan, prev_amount, seats, reason
目标:判断哪些字段适合用于构建客户流失预测模型特征,是否存在数据质量问题Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

