intent-solutions-io/nixtla-uncertainty-quantifier
000-docs/000a-planned-skills/core-forecasting/nixtla-uncertainty-quantifier/SKILL.md
Quantifies prediction uncertainty using conformal prediction. Use when risk assessment, scenario planning, or decision-making under uncertainty is required. Trigger with "quantify uncertainty", "generate prediction intervals", "confidence bands".
$ install --global
skillsauthnpx skillsauth add intent-solutions-io/plugins-nixtla nixtla-uncertainty-quantifierInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: Apr 5, 2026, 11:13 PM12.5s3 files scanned
SKILL.md
- name:
- nixtla-uncertainty-quantifier
- description:
- |
- allowed-tools:
- Read,Write,Bash,Glob,Grep
- version:
- 1.0.0
Uncertainty Quantifier
Generates prediction intervals and confidence bands for time series forecasts using conformal prediction.
Purpose
Provides a measure of the uncertainty associated with time series forecasts, enabling more informed decision-making.
Overview
Calculates prediction intervals and confidence bands around point forecasts, leveraging conformal prediction techniques. Requires a trained forecasting model (e.g., TimeGPT, StatsForecast) and historical data. Use when point forecasts alone are insufficient and an understanding of forecast uncertainty is needed. Outputs CSV files containing the forecast with uncertainty bounds and visualizations.
Prerequisites
Tools: Read, Write, Bash, Glob, Grep
Environment:
- Requires access to a trained forecasting model (e.g., TimeGPT API Key or StatsForecast model).
NIXTLA_TIMEGPT_API_KEY(if using TimeGPT).
Packages:
pip install nixtla pandas statsforecast matplotlib scikit-learn
Instructions
Step 1: Generate forecasts
Load historical data and generate point forecasts using StatsForecast.
Script: {baseDir}/scripts/generate_forecasts.py
Usage:
python {baseDir}/scripts/generate_forecasts.py \
--input data.csv \
--output forecast.csv
The script:
- Loads CSV data with
unique_id,ds,ycolumns - Splits data into training (80%) and validation (20%)
- Trains AutoETS model with weekly seasonality
- Generates forecasts for validation period
- Saves results with columns: unique_id, ds, y, StatsForecast
Step 2: Quantify uncertainty
Execute the uncertainty quantification script with your chosen confidence level and method.
Script: {baseDir}/scripts/quantify_uncertainty.py
Usage:
python {baseDir}/scripts/quantify_uncertainty.py \
--input forecast.csv \
--confidence 0.95 \
--method quantile \
--forecast_col StatsForecast
Parameters:
--input: Path to forecast CSV file--confidence: Confidence level (0.90 for 90%, 0.95 for 95%, etc.)--method: Conformal prediction method (quantileorjackknife+)--forecast_col: Name of the forecast column (default: StatsForecast)
Methods:
- quantile: Uses calibration set residuals to compute prediction intervals
- jackknife+: Uses leave-one-out cross-validation for uncertainty estimation
The script automatically:
- Calculates conformal prediction intervals
- Evaluates calibration (coverage, interval width)
- Generates visualization with confidence bands
- Saves all outputs
Output
- forecast_with_uncertainty.csv: Predictions with lower and upper bounds of the prediction interval.
- uncertainty_plot.png: Visualization of the forecast with confidence bands.
- calibration_metrics.json: Metrics evaluating the calibration of the prediction intervals (coverage, interval width).
Error Handling
-
Error:
Input file not foundSolution: Verify the file path specified in the--inputargument is correct. -
Error:
Invalid confidence levelSolution: Specify a confidence level between 0 and 1 (e.g., 0.95 for 95% confidence). -
Error:
Unsupported conformal prediction methodSolution: Choose a valid conformal prediction method, such as 'quantile' or 'jackknife+'. -
Error:
Forecast file does not contain required columnsSolution: Ensure the input CSV file containsunique_id,ds,y, and forecast columns (e.g., 'StatsForecast'). -
Error:
Data split results in empty calibration setSolution: Ensure you have sufficient data for both calibration and evaluation (minimum 10-20 observations).
Examples
Example 1: Sales forecast with 90% confidence
Input (forecast.csv):
unique_id,ds,y,StatsForecast
store_1,2024-01-01,100,95
store_1,2024-01-02,120,115
store_1,2024-01-03,110,112
Command:
python {baseDir}/scripts/quantify_uncertainty.py \
--input forecast.csv \
--confidence 0.90 \
--method quantile
Output (forecast_with_uncertainty.csv):
unique_id,ds,y,StatsForecast,lower_bound_90,upper_bound_90
store_1,2024-01-01,100,95.0,85.0,105.0
store_1,2024-01-02,120,115.0,105.0,125.0
store_1,2024-01-03,110,112.0,102.0,122.0
Example 2: Energy demand forecast with 95% confidence
Input (data.csv):
unique_id,ds,y
grid_1,2024-01-01 00:00,5000
grid_1,2024-01-01 01:00,5100
...
grid_1,2024-01-31 23:00,5200
Commands:
# Step 1: Generate forecasts
python {baseDir}/scripts/generate_forecasts.py \
--input data.csv \
--output forecast.csv
# Step 2: Quantify uncertainty
python {baseDir}/scripts/quantify_uncertainty.py \
--input forecast.csv \
--confidence 0.95 \
--method jackknife+
Output: Forecast with 95% prediction intervals for risk-aware energy planning.
Example 3: Complete workflow
# Generate forecasts from historical data
python {baseDir}/scripts/generate_forecasts.py \
--input sales_history.csv \
--output sales_forecast.csv
# Quantify uncertainty with 90% confidence using quantile method
python {baseDir}/scripts/quantify_uncertainty.py \
--input sales_forecast.csv \
--confidence 0.90 \
--method quantile \
--forecast_col StatsForecast
# Review outputs:
# - forecast_with_uncertainty.csv (data with intervals)
# - uncertainty_plot.png (visualization)
# - calibration_metrics.json (quality metrics)
Resources
- Conformal prediction overview: https://en.wikipedia.org/wiki/Conformal_prediction
- StatsForecast documentation: https://nixtlaverse.nixtla.io/statsforecast/
- Uncertainty quantification guide: https://otexts.com/fpp3/prediction-intervals.html
- Scripts:
{baseDir}/scripts/directory contains all executable code
Related Skills
intent-solutions-io/managing-test-environments
testing
VerifiedTrustedCommunity
This skill enables Claude to manage isolated test environments using Docker Compose, Testcontainers, and environment variables. It is used to create consistent, reproducible testing environments for software projects. Claude should use this skill when the user needs to set up a test environment with specific configurations, manage Docker Compose files for test infrastructure, set up programmatic container management with Testcontainers, manage environment variables for tests, or ensure cleanup after tests. Trigger terms include "test environment", "docker compose", "testcontainers", "environment variables", "isolated environment", "env-setup", and "test setup".
6SKILL.mdUpdated Jun 5, 2026
intent-solutions-io/generating-test-doubles
tools
VerifiedTrustedCommunity
This skill uses the test-doubles-generator plugin to automatically create mocks, stubs, spies, and fakes for unit testing. It analyzes dependencies in the code and generates appropriate test doubles based on the chosen testing framework, such as Jest, Sinon, or others. Use this skill when you need to generate test doubles, mocks, stubs, spies, or fakes to isolate units of code during testing. Trigger this skill by requesting test double generation or using the `/gen-doubles` or `/gd` command.
6SKILL.mdUpdated Jun 5, 2026
intent-solutions-io/generating-test-data
tools
VerifiedTrustedCommunity
This skill enables Claude to generate realistic test data for software development. It uses the test-data-generator plugin to create users, products, orders, and custom schemas for comprehensive testing. Use this skill when you need to populate databases, simulate user behavior, or create fixtures for automated tests. Trigger phrases include "generate test data", "create fake users", "populate database", "generate product data", "create test orders", or "generate data based on schema". This skill is especially useful for populating testing environments or creating sample data for demonstrations.
6SKILL.mdUpdated Jun 5, 2026
intent-solutions-io/analyzing-test-coverage
development
VerifiedTrustedCommunity
This skill analyzes code coverage metrics to identify untested code and generate comprehensive coverage reports. It is triggered when the user requests analysis of code coverage, identification of coverage gaps, or generation of coverage reports. The skill is best used to improve code quality by ensuring adequate test coverage and identifying areas for improvement. Use trigger terms like "analyze coverage", "code coverage report", "untested code", or the shortcut "cov".
6SKILL.mdUpdated Jun 5, 2026