Trust inversion and its exploitation: A threat model of the Model Context Protocol for agentic AI. Journal of the Nigerian Society of Physical Sciences, v. 8, n. 4, p. 3748, 2026.