MCP & Tool Supply Chain
Securing the Model Context Protocol and agent tools — malicious servers, tool-description poisoning, CVE teardowns and sandboxing.
MCP Security: Tool Poisoning, Rogue Servers, and Supply-Chain Exploits
How attackers exploit the Model Context Protocol: tool poisoning, malicious and rogue MCP servers, rug pulls and supply-chain attacks against agent tooling — the offensive threat model.
Read →Malicious MCP Servers: How Rogue Tool Servers Cross Your Agent's Trust Boundary
A malicious MCP server runs with your agent's privileges and dictates its tool behavior. How rogue servers get in, what they can do, and how to verify, pin, and isolate them.
Read →MCP Server Isolation: Containing Untrusted Tool Servers and the CVEs That Prove Why
Why MCP servers are the new untrusted dependency: a tour of the CVE classes — command injection, tool poisoning, path traversal, SSRF — and how to isolate, scope, and sandbox each server so a compromised tool can't own your agent.
Read →Tool Description Poisoning in MCP: How Rug Pulls and Hidden Instructions Hijack Your Agent
How MCP tool description poisoning and rug pulls hide malicious instructions in tool metadata, plus pinning and audit defenses that actually catch post-install mutation.
Read →