Data Analysis - Statistical Computing & Insights

When to use this skill

• User mentions "数据分析", "统计", "计算指标", "数据洞察"
• Need to analyze structured data (CSV, JSON, database)
• Calculate statistics, trends, patterns
• Financial analysis (returns, volatility, technical indicators)
• Business analytics (sales, user behavior, KPIs)
• Scientific data processing and hypothesis testing

Workflow

1. Get data

• File names MUST NOT contain Chinese characters or non-ASCII characters
• Use only English letters, numbers, underscores, and hyphens
• Examples:

  data.csv

  ,

  sales_report_2025.xlsx

  ,

  analysis_results.json

• ❌ Invalid:

  销售数据.csv

  ,

  数据文件.xlsx

  ,

  報表.json

• This ensures compatibility across different systems and prevents encoding issues

• Read from file (CSV, JSON, Excel)
• Query database if available

• Ask the user to rename the file to English/ASCII characters
• Or rename the file when saving it to the agent directory

• Automatically activate

  data-base

  skill
• Scrape/collect required data
• Save to structured format

2. Understand requirements

• What questions do you want to answer?
• What metrics are important?
• What format for results? (summary, chart, report)
• Any specific statistical methods?

3. Analyze

• Descriptive statistics (mean, median, std, percentiles)
• Distribution analysis (histograms, box plots)
• Correlation analysis
• Group comparisons

• Return calculation (simple, log, cumulative)
• Risk metrics (volatility, VaR, Sharpe ratio)
• Technical indicators (MA, RSI, MACD)
• Portfolio analysis

• Trend analysis (growth rates, YoY, MoM)
• Cohort analysis
• Funnel analysis
• A/B testing

• Hypothesis testing (t-test, chi-square, ANOVA)
• Regression analysis
• Time series analysis
• Statistical significance

4. Output

• Summary statistics: Tables with key metrics
• Charts: Save as PNG files
• Report: Markdown with findings
• Data: Processed CSV/JSON for further use

Python Environment

cd skills/data-analysis

if [ ! -f ".venv/bin/python" ]; then
    echo "Creating Python environment..."
    ./setup.sh
fi

.venv/bin/python your_script.py

Analysis scenarios

General data

import pandas as pd

# Load and summarize
df = pd.read_csv('data.csv')
summary = df.describe()
correlations = df.corr()

Financial data

# Calculate returns
df['return'] = df['price'].pct_change()

# Risk metrics
volatility = df['return'].std() * (252 ** 0.5)
sharpe = df['return'].mean() / df['return'].std() * (252 ** 0.5)

Business data

# Group by category
grouped = df.groupby('category').agg({
    'revenue': ['sum', 'mean', 'count']
})

# Growth rate
df['growth'] = df['revenue'].pct_change()

Scientific data

from scipy import stats

# T-test
t_stat, p_value = stats.ttest_ind(group_a, group_b)

# Regression
from sklearn.linear_model import LinearRegression
model = LinearRegression()
model.fit(X, y)

File path conventions

Temporary output (session-scoped)

import time
from datetime import datetime

# Use timestamp for unique filenames (avoid conflicts)
timestamp = datetime.now().strftime('%Y%m%d_%H%M%S')

# Charts and temporary files
plt.savefig(f'analysis_{timestamp}.png')      # → $KODE_AGENT_DIR/analysis_20250115_143022.png
df.to_csv(f'results_{timestamp}.csv')        # → $KODE_AGENT_DIR/results_20250115_143022.csv

• Use timestamps:

  analysis_20250115_143022.png

• Use descriptive names + timestamps:

  sales_report_q1_2025.csv

• Use random suffix for scripts:

  script_{random.randint(1000,9999)}.py

User data (persistent)

$KODE_USER_DIR

import os
user_dir = os.getenv('KODE_USER_DIR')

# Save to user memory
memory_file = f"{user_dir}/.memory/facts/preferences.jsonl"

# Read from knowledge base
knowledge_dir = f"{user_dir}/.knowledge/docs"

Environment variables

• KODE_AGENT_DIR

  : Session directory for temporary output (charts, analysis results)

• KODE_USER_DIR

  : User data directory for persistent storage (memory, knowledge, config)

Best practices

• File names MUST be ASCII-only: No Chinese or non-ASCII characters in filenames
• Always inspect data first:

  df.head()

  ,

  df.info()

  ,

  df.describe()

• Handle missing values: Drop or impute based on context
• Check assumptions: Normality, independence, etc.
• Visualize: Charts reveal patterns tables hide
• Document findings: Explain metrics and their implications
• Use correct paths: Temporary outputs to current dir, persistent data to

  $KODE_USER_DIR

Quick reference

• REFERENCE.md - pandas/numpy API reference
• references/financial.md - Financial analysis recipes
• references/business.md - Business analytics recipes
• references/scientific.md - Statistical testing methods
• references/templates.md - Code templates

Environment setup

# Navigate to the skill directory
cd apps/assistant/skills/data-analysis

# Run the setup script (creates venv and installs dependencies)
./setup.sh

# Activate the environment
source .venv/bin/activate

• Create a Python virtual environment in

  .venv/

• Install required packages (pandas, numpy, scipy, scikit-learn, statsmodels)

# Use the virtual environment's Python
.venv/bin/python script.py

# Or activate first, then run normally
source .venv/bin/activate
python script.py