Stack layer / Threat pattern
MCPZoo Study: Scanners Call 96.89% of MCP Servers 'Risky' — But Under Half of Those Alerts Are Real
arXiv
Stack layer
Distributed Backdoors Prove Per-Step Agent Monitors Can Be Right on Every Step and Still Miss the Attack
arXiv
Threat pattern
Malaika Grounds Agentic Malware Analysis in Domain, Program Semantics, and Threat Intel
arXiv
Stack layer / Threat pattern
Mako: A 'Self-Evolving Agentic OS' That Writes New Exploits Against Live Targets and Hot-Loads Them Into Itself
arXiv
Contrast
Shared Selective Persistent Memory Lifts Agentic Task Completion to 96% and Cuts Per-Call Token Cost 97x
arXiv
Stack layer / Contrast
BackendForge: Best Coding Agent (GPT-5.5) Deploys Correct Backends Only 28.6% of the Time Under a Strict Oracle
arXiv
Policy dependency / Stack layer
Cynative Ships an Open-Source Deep-Research Security Agent With Default-Deny Write Permissions
IT Security News
Stack layer / Update thread
SLBench Finds Up to 70% Unsafe Execution When Codex and Claude Code Follow Logical Relations in Skill Files
arXiv