🧰 MCPTox

The items that name it 2

• Threat-Preserving Representation Sensitivity in Agent-Security Benchmarks
  paper2026-10-02

  Security benchmark used to test effect of tool-name wording on ASR for LLM agents

• MCPTox: A Benchmark for Tool Poisoning Attack on Real-World MCP Servers
  paper

  First benchmark to systematically evaluate agent robustness against Tool Poisoning attacks in realistic MCP settings; constructed on 45 live MCP servers and 353 tools, comprising 1312 malicious test cases across 10 risk categories, generated via three attack templates and few-shot learning; released