数据与分析
data-cleaning
Use when planning safe data cleaning, deduplication, missing-value handling, anomaly handling, and reproducible cleaned outputs.
BlueSkyXN · 已有静态检查记录
安装与使用
运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。
npx skills add BlueSkyXN/Codex-is-all-you-need --skill data-cleaning技能说明
name: data-cleaning description: Use when planning safe data cleaning, deduplication, missing-value handling, anomaly handling, and reproducible cleaned outputs. metadata: version: "0.1" updated: "2026-05-11"
Data Cleaning Skill
Use this workflow for data cleaning plans or scripts.
Steps
- Identify raw inputs and preserve them unchanged.
- Define the target output format and location.
- Classify issues: missing values, duplicates, type errors, invalid categories, outliers, date problems, and inconsistent identifiers.
- Decide handling rules and document assumptions.
- Implement cleaning in a reproducible script if requested.
- Write cleaned outputs to
outputs/or a task-specific directory, notraw/. - Validate before/after row counts, key counts, missingness, and edge cases.
Do not
- Do not overwrite raw files.
- Do not silently drop rows or impute values without documenting rules.
- Do not treat outliers as errors without evidence.
Output
Return:
- Cleaning issues found
- Cleaning rules
- Files/scripts created
- Before/after validation
- Residual risks