name: data-analyst description: 資料分析師自動化工作流。從資料載入、質量審計、資料清洗、探索性分析(EDA)、統計建模到視覺化HTML報告生成，覆蓋完整資料分析管線。支援CSV/Excel/JSON/SQLite多格式輸入，內建4層資料防禦體系。觸發詞：分析資料、資料分析、幫我分析資料、資料包告、EDA、data analysis、analyze data、生成資料包告、資料視覺化、探索性分析。 agent_created: true

資料分析師 (Data Analyst)

AI-powered data analysis workflow. Cover the full pipeline from data ingestion to interactive HTML report generation.

When to Use

Trigger when the user asks to: - Analyze a dataset (CSV / Excel / JSON / SQLite) - Generate a data analysis report - Do exploratory data analysis (EDA) - Clean or preprocess data - Create data visualizations - Understand data distributions and relationships

Workflow Overview

The skill follows a 7-phase CRISP-DM pipeline, executed automatically:

1. Data Loading — Auto-detect format, load into DataFrame
2. Data Audit — 4-layer defense: health check, structure, business rules, model readiness
3. Data Cleaning — Missing values, outliers, type conversion, dedup
4. EDA — Distribution analysis, correlation, group aggregation
5. Statistical Analysis — Descriptive stats, hypothesis tests, trend detection
6. Visualization — Charts for distributions, correlations, category breakdowns
7. Report Generation — Interactive HTML report with scorecards, charts, and insights

Usage

Quick Start

To analyze a data file:

python {baseDir}/scripts/run_analysis.py <data_file> [--output report.html]

The script auto-detects the file format and runs the full pipeline.

Module-Level Usage

Each module can be used independently:

# Load data
from data_loader import load_data
df = load_data("sales.csv")

# Audit data quality
from data_auditor import audit_data
report = audit_data(df)

# Clean data
from data_cleaner import clean_data
df_clean = clean_data(df)

# Run EDA
from eda_runner import run_eda
eda_results = run_eda(df_clean)

# Generate report
from report_builder import build_report
build_report(df_clean, eda_results, "report.html")

Scripts Reference

Templates

小蔥技能站7w4.net發現了升級外掛。

• templates/report.html — Jinja2 template for the final HTML report

Config

• config/business_rules.yaml — Optional business validation rules

Dependencies

Install before first use:

pip install pandas numpy matplotlib seaborn scipy jinja2 pyyaml missingno

Notes

• For files > 100MB, the audit module uses sampling (n=50000) to stay performant
• Business rules in config/business_rules.yaml are optional; skip if no domain-specific rules exist
• All charts are saved to a charts/ subdirectory in the output folder before embedding in HTML

這個 Skill 質量不錯，功能覆蓋完整，從讀取資料、檢查質量、清理資料到生成視覺化報告一氣呵成。支援多種資料格式，內建資料質量自動檢測，生成的 HTML 報告直觀好看，對日常資料分析很有幫助。主要缺點是使用門檻較高，需要安裝 Python 和多個依賴庫，而且需要通過命令列操作，沒有圖形介面。功能雖然豐富但對於不熟悉程式碼的使用者來說不太友好。總體來說適合有技術基礎的使用者使用。