Unverified50% confidenceFactExact time
当前MCP生态缺乏类似App Store审核机制的准入管控，安全研究人员已发现多个MCP投毒攻击案例，包括Typosquatting钓鱼攻击和伪装成Postmark MCP的恶意服务器
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/21/2026
First Seen
Valid until: 10/19/2026
Sources
Related Claims
UnverifiedSecurity auditing of MCP Servers has fallen far behind their rapid growth in numbers, which is currently the biggest shortcoming of the MCP ecosystem.72% similarVerifiedThe agent using only MCP tools did not set security flags and created a view with a data leakage risk, while the MCP+Skill agent correctly identified and implemented the secure configuration.72% similarVerifiedMCP tools are essentially arbitrary code execution, indirect prompt injection is a real-world threat, and attack vectors can be hidden in resource content rather than in direct user input.70% similarUnverifiedMCP currently lacks fine-grained permission control, and third-party servers of unknown origin may contain malicious instructions.70% similarUnverified安全研究人员发现一个恶意MCP服务器伪装成正规邮件转发服务Postmark MCP，实际会在每封邮件密送一份给攻击者69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/577519API
curl https://kongchang.com/api/v1/knowledge/claims/577519MCP
get_claim(id=577519)