What it does
A prompt injection scanner for AI coding tool hooks, written in Rust. It checks tool inputs and outputs for injection attacks, leaked secrets and data exfiltration, using models that run locally. It is built for Claude Code hooks. The author describes it as early development, so false positives are possible.
Use cases
- 01Flag injected instructions in fetched web content
- 02Catch secrets leaking through tool output
- 03Add a scanning step to Claude Code hooks