Do I need to manage my own ML infrastructure?

The skill interacts with a local ML runtime via curl. You may need to run '/llamafarm:start' to ensure the required backend services are active.

Can I use this for real-time monitoring?

Yes, the '/llamafarm:anomaly stream' command allows you to set up streaming detection with configurable window sizes and retraining intervals for live data.

How does the skill handle different data qualities?

It provides a contamination recommendation scale: 0.01-0.05 for clean data, 0.05-0.10 for moderate noise, and up to 0.15 for highly noisy or unknown datasets.

Which machine learning algorithms does this skill support?

The skill supports multiple backends including Isolation Forest (iforest), ECOD, PCA, Autoencoders, HBOS, and LODA, covering a wide range of tabular and high-dimensional data types.

ML Anomaly Detection

Name: ML Anomaly Detection
Author: llama-farm

byllama-farm

0•

数据科学与机器学习

Identifies statistical outliers and patterns in batch or streaming datasets using specialized machine learning backends.

The LlamaFarm Anomaly Detection skill provides a comprehensive machine learning workflow directly within Claude Code for identifying deviations in complex datasets. It guides users through selecting the right algorithms—such as Isolation Forest for tabular data or ECOD for high-dimensional sets—and manages the entire lifecycle from model training (fitting) to real-time streaming detection. Whether you are monitoring system health, detecting fraud, or cleaning data, this skill automates the technical configuration of contamination levels and normalization to ensure high-precision results.

主要功能

01Support for both batch processing and real-time streaming anomaly detection

02Automated parameter recommendations for contamination and normalization

03Model management system to list, save, and reuse trained detection models

04Built-in decision trees to match data types with optimal ML algorithms

05Intelligent backend selection including IForest, ECOD, PCA, and Autoencoders

060 GitHub stars

使用场景

01Detecting hardware sensor malfunctions or manufacturing defects in real-time streams

02Monitoring server logs and system metrics for security breaches or performance spikes

03Identifying fraudulent transactions or unusual user behavior in financial datasets

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add llama-farm/claude-code-marketplace anomaly

For use in Claude.ai and ChatGPT

Download Skill