القائمة
Use when analyzing data, creating visualizations, or extracting insights -- covers statistical methodology, pandas patterns, data cleaning, visualization best practices, and insight presentation
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Use after completing a major work batch (plan execution, feature build, multi-task subagent run) to refresh code-graph-derived data in MEMORY.md/CONTEXT.md/SKILL_MEMORY, append dated narrative entries to TODO/CHANGELOG/lessons-learned, mine cross-project structural habits, and ratchet reusable permissions to global settings. Keeps project DNS in sync with reality.
Use when orchestrating multiple ghengis-skills across a single task -- coordinates pre-action validation, execution, post-action verification, and lesson extraction via a shared JSON scratchpad. TRIGGER when dispatching a subagent, finishing a major task, or when the user mentions "run the full chain", "validate this", "fire the pipeline", or explicitly invokes a chain by name.
Use to build and maintain an evolving understanding of user preferences, communication style, and working patterns — learns from interactions and adapts behavior over time
Use when starting a new project or organizing an existing one — creates self-documenting 4-layer project structure with MEMORY.md, CONTEXT.md, modular CLAUDE.md, and workspace routing
Use to accumulate and retrieve domain knowledge from past tasks — builds a grepable plain-text knowledge base that grows over time, enabling agents to learn from experience without vector databases
Use when reorganizing an existing project's file tree based on what actually connects to what — analyzes import graph, classifies layers, proposes an ideal target tree, validates the plan adversarially, then executes file moves with atomic per-language import rewrites. Always plan-first; never moves a file before the user confirms. TRIGGER on "reorganize this project", "clean up the file tree", "the structure is a mess", "make this match a real project layout", or after project-scaffold detects layout drift.
| name | data-analysis |
| description | Use when analyzing data, creating visualizations, or extracting insights -- covers statistical methodology, pandas patterns, data cleaning, visualization best practices, and insight presentation |
| allowed-tools | Read Grep Glob Bash(python *) |
When helping with data analysis, follow this methodology to ensure statistical rigor, clear visualizations, and actionable insights.
Measures of Central Tendency
Measures of Spread
Example
Dataset: [10, 12, 12, 13, 15, 17, 100]
Mean: 25.6 (skewed by outlier 100)
Median: 13 (middle value, better representation)
Mode: 12 (appears twice)
Range: 90 (100 - 10)
Std Dev: 31.5 (high due to outlier)
IQR: 5 (Q3=16, Q1=11, ignores outlier)
Conclusion: Median + IQR better summarize this data (outlier present).
Pearson Correlation (r)
Correlation does NOT equal causation.
Scatterplot Interpretation
Positive correlation: Points trend upward (r > 0)
Negative correlation: Points trend downward (r < 0)
No correlation: Random scatter (r ~ 0)
Non-linear: Curved pattern (Pearson r may not apply)
Simple Linear Regression
R-Squared (R2)
Residuals
Methods
Handling Outliers
Types of Trends
Moving Average: Smooths noisy data to reveal underlying trend. Example: 7-day moving average for daily stock prices.
Year-Over-Year (YoY) Comparison: Compare same period in different years to remove seasonality. Better than month-over-month for seasonal businesses.
import pandas as pd
import numpy as np
# 1. Load data
df = pd.read_csv('data.csv')
# 2. Initial exploration
df.head() # First 5 rows
df.info() # Column types, null counts
df.describe() # Summary stats
# 3. Clean data
df.dropna() # Remove missing values
df['column'] = df['column'].fillna(0) # Fill nulls with 0
df.drop_duplicates() # Remove duplicate rows
# 4. Filter/subset
df[df['age'] > 30] # Rows where age > 30
df[['name', 'age']] # Select specific columns
df.loc[0:5, 'name'] # Rows 0-5, column 'name'
# 5. Transform
df['new_col'] = df['col1'] + df['col2'] # Create column
df['log_price'] = np.log(df['price']) # Log transform
df.groupby('category')['sales'].sum() # Group by and aggregate
# 6. Analyze
df['sales'].mean() # Mean
df['sales'].median() # Median
df['sales'].std() # Standard deviation
df.corr() # Correlation matrix
# 7. Visualize (with matplotlib or seaborn)
df['sales'].hist() # Histogram
df.plot(x='date', y='sales') # Line chart
df.boxplot(column='price', by='category') # Box plot
1. Define question:
"Is there a relationship between ad spend and sales?"
2. Collect data:
- Ensure sufficient sample size (n > 30 for most stats)
- Check data quality (missing values? outliers?)
3. Explore data:
- Summary stats (mean, median, std dev)
- Visualize distributions (histogram, boxplot)
- Check for outliers
4. Run analysis:
- Correlation: df.corr() or scipy.stats.pearsonr()
- Regression: sklearn LinearRegression or statsmodels
- T-test: scipy.stats.ttest_ind() for group comparison
5. Interpret results:
- Check p-value (< 0.05 = statistically significant)
- Check effect size (is difference meaningful?)
- Check R2 (how much variance explained?)
6. Communicate findings:
- Visualization (chart that tells the story)
- One-sentence insight: "Each $1k in ad spend increases sales by $3k (R2=0.72)"
- Caveat: "Correlation, not causation; other factors may contribute"
1. Detect outliers:
- IQR method: x < Q1 - 1.5*IQR or x > Q3 + 1.5*IQR
- Z-score: |z| > 3
- Visual: Boxplot shows dots beyond whiskers
2. Investigate cause:
- Data entry error? (typo: 1000 instead of 100)
- Measurement error? (faulty sensor)
- Real extreme value? (celebrity, rare event)
3. Decision:
- If error: Correct or remove
- If real but rare: Keep (document as outlier)
- If distorts analysis: Transform (log) or use robust stats (median)
4. Document:
- Note which values were outliers
- Explain why you kept or removed them
- Run analysis both ways (with and without outliers)
| Chart Type | Use When | Example |
|---|---|---|
| Line chart | Trends over time | Stock prices, website traffic |
| Bar chart | Comparisons across categories | Sales by region |
| Histogram | Distribution of single variable | Age distribution |
| Scatterplot | Relationship between two variables | Height vs weight |
| Boxplot | Compare distributions across groups | Salary by department |
| Heatmap | Correlation matrix | Multiple variable relationships |
| Pitfall | Problem | Fix |
|---|---|---|
| Mean for skewed data | Mean = $500k, but median = $50k (billionaire skew) | Use median for skewed distributions |
| Correlation = Causation | "Ice cream causes drowning" | Check for confounding variables |
| Small sample size | n=10, conclude "5% difference" | Minimum n=30, use power analysis |
| P-Hacking | Test 20 hypotheses, find 1 with p<0.05 | Pre-register hypothesis, Bonferroni correction |
| Cherry-picking time periods | "Sales up 50% last month!" (down 20% overall) | Show full time series, use YoY |
| Ignoring outliers | Mean of [10, 12, 13, 15, 1000] = 210 | Investigate outliers, use median or remove if errors |
Compare two groups:
- T-test: scipy.stats.ttest_ind(group1, group2)
- Use when: Normal distribution, continuous data
Compare multiple groups:
- ANOVA: scipy.stats.f_oneway(g1, g2, g3)
- Use when: 3+ groups, normal distribution
Correlation:
- Pearson: scipy.stats.pearsonr(x, y)
- Use when: Linear relationship, continuous data
Non-parametric (non-normal data):
- Mann-Whitney U: scipy.stats.mannwhitneyu(g1, g2)
- Spearman correlation: scipy.stats.spearmanr(x, y)
import matplotlib.pyplot as plt
import seaborn as sns
# Matplotlib (basic)
plt.plot(x, y) # Line chart
plt.bar(categories, values) # Bar chart
plt.hist(data) # Histogram
plt.scatter(x, y) # Scatterplot
# Seaborn (advanced)
sns.lineplot(data=df, x='date', y='sales')
sns.barplot(data=df, x='category', y='sales')
sns.histplot(data=df, x='age', bins=20)
sns.scatterplot(data=df, x='x', y='y', hue='category')
sns.boxplot(data=df, x='category', y='price')
sns.heatmap(df.corr(), annot=True) # Correlation heatmap
r = 1.0: Perfect positive (x up -> y up exactly)
r = 0.7 to 0.9: Strong positive
r = 0.4 to 0.7: Moderate positive
r = 0.1 to 0.4: Weak positive
r = 0.0: No correlation
r = -0.1 to -0.4: Weak negative
r = -0.4 to -0.7: Moderate negative
r = -0.7 to -0.9: Strong negative
r = -1.0: Perfect negative (x up -> y down exactly)