analyzing-time-series

Name: analyzing-time-series
Author: Rami-RK

by @Rami-RK in Data & Analytics

# Install this skill:

npx skills add Rami-RK/skill

Or install specific skill: npx add-skill https://github.com/Rami-RK/skill/tree/main/custom_skills/analyzing-time-series

# Description

Comprehensive diagnostic analysis of time series data. Use when users provide CSV time series data and want to understand its characteristics before forecasting - stationarity, seasonality, trend, forecastability, and transform recommendations.

# SKILL.md

name: analyzing-time-series
description: Comprehensive diagnostic analysis of time series data. Use when users provide CSV time series data and want to understand its characteristics before forecasting - stationarity, seasonality, trend, forecastability, and transform recommendations.

Time Series Diagnostics

Comprehensive diagnostic toolkit to analyze time series data characteristics before forecasting.

Input Format

The input CSV file should have two columns:
- Date column - Timestamps or dates (e.g., date, timestamp, time)
- Value column - Numeric values to analyze (e.g., value, sales, temperature)

Workflow

Step 1: Run diagnostics

python scripts/diagnose.py data.csv --output-dir results/

This runs all statistical tests and analyses. Outputs diagnostics.json with all metrics and summary.txt with human-readable findings. Column names are auto-detected, or can be specified with --date-col and --value-col options.

Step 2: Generate plots (optional)

python scripts/visualize.py data.csv --output-dir results/

Creates diagnostic plots in results/plots/ for visual inspection. Run after diagnose.py to ensure ACF/PACF plots are synchronized with stationarity results. Column names are auto-detected, or can be specified with --date-col and --value-col options.

Step 3: Report to user

Summarize findings from summary.txt and present relevant plots. See references/interpretation.md for guidance on:
- Is the data forecastable?
- Is it stationary? How much differencing is needed?
- Is there seasonality? What period?
- Is there a trend? What direction?
- Is a transform needed?

Script Options

Both scripts accept:
- --date-col NAME - Date column (auto-detected if omitted)
- --value-col NAME - Value column (auto-detected if omitted)
- --output-dir PATH - Output directory (default: diagnostics/)
- --seasonal-period N - Seasonal period (auto-detected if omitted)

Output Files

results/
├── diagnostics.json       # All test results and statistics
├── summary.txt            # Human-readable findings
├── diagnostics_state.json # Internal state for plot synchronization
└── plots/
    ├── timeseries.png
    ├── histogram.png
    ├── rolling_stats.png
    ├── box_by_dayofweek.png  # By day of week (if applicable)
    ├── box_by_month.png      # By month (if applicable)
    ├── box_by_quarter.png    # By quarter (if applicable)
    ├── acf_pacf.png
    ├── decomposition.png
    └── lag_scatter.png

References

See interpretation.md for:
- Statistical test thresholds and interpretation
- Seasonal period guidelines by data frequency
- Transform recommendations

Dependencies

pandas, numpy, matplotlib, statsmodels, scipy

# README.md

Skills with the Claude API

Lesson Files

You can find the lesson's notebook and all the required input files here.

To run the notebook, you need to create a .env file containing an Anthropic API key (no Claude subscription is required):

ANTHROPIC_API_KEY="your-key"

You can get a key from Claude Developer Platform.

About costs: Please note that running through all the notebook cells once will use approximately $0.67 in API credits.

If you'd prefer not to run the notebook, you can:
- view the notebook with pre-run outputs (exactly as shown in the video)
- check out the generated sample outputs

You can also try the same custom skills in Claude.ai.

Notes

Here's the list of pre-installed libraries in the sandboxed environment
Streaming: The lesson's notebook does not implement streaming with the Messages API. So when you run the cells to get the response, you might need to wait for a few minutes. If you'd like to implement streaming, you can check the documentation here.
To see more examples of how to use Agent Skills with the API (like multi-turn conversation), make sure to check this guide.

Additional References

2. Create a Virtual Environment

Go inside the project folder and create a Python virtual environment.

python -m venv venv

Activate the environment:

Windows:
bash venv\Scripts\activate
macOS / Linux:
bash source venv/bin/activate

Install all dependencies:

pip install -r requirements.txt

# Supported AI Coding Agents

This skill is compatible with the SKILL.md standard and works with all major AI coding agents:

⚡ Amp 🚀 Antigravity 🤖 Claude Code 🦀 Clawdbot 📝 Codex ▶️ Cursor 🤖 Droid 💎 Gemini CLI 🐙 GitHub Copilot 🪿 Goose 📊 Kilo Code 🔧 Kiro CLI 💻 OpenCode 🦘 Roo Code 🌲 Trae 🏄 Windsurf

Learn more about the SKILL.md standard and how to use these skills with your preferred AI coding agent.