jev-superpowers

jev-superpowers is a framework that adds disciplined engineering safeguards to AI coding agents. It replaces loose conversational logic with six structured skills—routing, ideation, dependency vetting, command guarding, log triage, and test gates—each backed by TypeSafe Jev System One typed decisions that evaluate in 70–120ms. The framework intercepts agent actions before execution, catching hallucinated packages, blocking dangerous commands like rm -rf, and rejecting incomplete work.

Designed for developers using Claude Code, Cursor, Codex, and other AI agents, jev-superpowers runs with zero framework fees and includes a 22-test offline suite. It supports both cloud and local inference via the open-weight Laya model, giving teams control over where decisions are made.