[KongchangAI]
Unverified50% confidenceSolutionExact time

在MCP层可以注入恶意工具返回值、拦截审计Agent的工具调用、模拟工具失败等边界情况,使MCP层成为对抗测试的注入点与观测窗

1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen

Sources

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/535985
API
curl https://kongchang.com/api/v1/knowledge/claims/535985
MCP
get_claim(id=535985)