🛡️ CVE-2026-26030

The items that name it 3

• When Prompts Become Shells: RCE Vulnerabilities in AI Agent Frameworks (Semantic Kernel)
  writeup2026-05-07

  Vulnerability in Semantic Kernel's In-Memory Vector Store filter functionality; unsafe string interpolation into an eval()'d Python lambda combined with a bypassable AST-based blocklist allows prompt-injection-driven remote code execution (RCE).

• Agent Threat Rules (ATR) — open detection-rule standard for AI agent threats
  repo

  Semantic Kernel lambda+eval RCE, published by MSRC on 2026-05-07

• AutoJack Attack Lets One Web Page Hijack AI Agent for Host Code Execution
  article