HYPE Skills

HYPE Skills is a small open-source framework that converts real-world security incidents into machine-readable detection skills for AI coding agents — specifically Claude Code. When a new exploit happens in the wild, the useful artifact is not a blog post; it's a structured pattern that an agent can reuse to scan other codebases for the same class of bug.

The repo contains an attack taxonomy and an incident dataset covering eight real-world DeFi exploits. For each incident, I produced a YAML record with the attack primitives, the vulnerable pattern, and a narrative case study for human readers. The goal is that a Claude Code session handed this skill can spot the same class of bug in an unfamiliar contract without the user having to describe the bug first.

It's early and deliberately narrow — Hyperliquid ecosystem incidents are the starting point. The format is designed to be extended, so that other researchers can contribute their own incident records without changing the framework.