Multi-hop cross-prompt injection benchmark for multi-agent AI systems. 250 attack cases, 8 taxonomy categories, 4 defenses evaluated. Watch: https://www.youtube.com/watch?v=fGOlMij4HPQ
-
Updated
Jun 16, 2026 - Python
Multi-hop cross-prompt injection benchmark for multi-agent AI systems. 250 attack cases, 8 taxonomy categories, 4 defenses evaluated. Watch: https://www.youtube.com/watch?v=fGOlMij4HPQ
Defensive 7-layer runtime guardrails gateway for LLM agents: a reverse proxy that blocks prompt injection (incl. indirect/XPIA), tool misuse, and data exfiltration on every turn. Polyglot — Python control plane + Rust security core + React HITL. Metrics reported split (ASR vs FPR), never one blended F1.
Add a description, image, and links to the xpia topic page so that developers can more easily learn about it.
To associate your repository with the xpia topic, visit your repo's landing page and select "manage topics."