Topic
Defending MCP tool ecosystems against tool poisoning, typosquatting, hidden instruction injection and description drift.
The vendor reports 10-plus CVEs and up to 200,000 exposed instances, and says Anthropic declined to change the protocol, describing the behaviour as expected.
Reality
A DigiCert CTO argues the human approval step in agentic AI will disappear entirely. If he is right, the control most boards signed off on is going before its replacement is provisioned.
Reality
A Hacker News explainer makes the structural case: plaintext configs, unrotated tokens, broad scopes and prompt injection put agent plumbing outside secrets management.
Reality
Agent Governance Toolkit puts policy checks in the execution path rather than the prompt. The seam: the kernel is middleware inside the agent's process, so containers still do the isolating.
Reality