DeepYardDeepYard
C

ClawTrap

MITM red-teaming framework for testing autonomous web agent security in real-world scenarios

Open SourceFree

About

ClawTrap is a man-in-the-middle (MITM) based red-teaming framework designed to evaluate the security of autonomous web agents like OpenClaw in production-like environments. Unlike static sandbox benchmarks, it provides network-layer security testing that simulates real-world attack vectors and adversarial conditions. Built for security researchers and AI safety teams working with autonomous agents that interact with live web services.

Details

Type
Integrations
Language

Tags

autonomousevaluationopen-sourceobservabilitytool-use