Adaptive workflow router for AI coding agents, controlling risk depth
Many AI coding assistants execute risky actions without proper checks, leading to accidental leaks or destructive changes. This tool classifies each request into a depth level (0‑4) and routes it through appropriate verification steps, from simple answers to gated approvals. It ships as a Claude Code and Codex plugin with deterministic JavaScript classifiers and a test suite, making it easy to integrate into existing AI‑assisted development pipelines. Developers gain fine‑grained control over AI‑driven actions without sacrificing speed for low‑risk tasks.
View on GitHub →0xTangerin3/superpowers-a