A Practical Guide to Analyzing ClawHub Security Signals and Predicting ClawScan Verdicts with Machine Learning

Summary

This tutorial demonstrates how to use the ClawHub Security Signals dataset to analyze how different security scanners assess AI skills and related files. It covers loading the dataset from Hugging Face Parquet conversion, inspecting columns, verdict distribution, scanner outputs, and severity labels. The article explores scanner disagreement and overlap patterns, then builds a machine learning pipeline combining SKILL.md text with numerical scanner signals to predict ClawScan verdicts using logistic regression.