Cairn
agentic-offensive-framework
Overview
Open-source AI harness that loops coding agents (Claude Code, DeepSeek) toward intrusion goals such as unauthenticated RCE and uploader shells. Panels ship with no authentication and bind to 0.0.0.0:8000, exposing task text, tooling, and secrets. Observed achieving roughly 10% success rate with ~23 hours to goal against e-commerce and payment sites.
- Type
- agentic-offensive-framework