Agent, Tool & MCP Security practice test
The lethal trifecta, excessive agency, tool authorization, scoped credentials for agents, human approval that holds, sandboxed code execution, memory poisoning, MCP's trust model and threats, multi-agent trust and cascading failures, and the agent design review. The pool for this topic holds 10 easy, 28 medium and 28 hard items. Each answer comes with its reasoning and a link to the question or concept page it was drawn from.
Sample items with answers
The combination identifies a possible confidentiality failure: untrusted input may influence how private data leaves the system. Verify the actual data path and enforcement points. Removing one capability closes this route only if alternate tools, rendering and agent handoffs do not restore it.
OWASP describes excessive functionality, permissions and autonomy as recurring roots of excessive agency. Compare the effective authority and permitted decisions with the supported task. Automatic work can be appropriate within a valid grant, and tool count alone does not describe reachable harm. Enforce resource, recipient and business constraints outside the model.
After the test
Choosing from options is easier than reasoning it out in front of a panel. These pages make you reason it out: