100% Latest & Updated Cisco 350-901 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 890
The test server accepts requests without authenticating clients and exposes internal interface inventories. Independently, reviewers flag this potentially unsafe practice for assessment: 'Bind test server to every public interface with no network ACL'. Which verification addresses that practice without assuming it caused the incident?
- Inspect auth enforcement
- Inspect response and secret scopes
- Expose an unencrypted public endpoint for internal network data
- Inspect listener binding and firewall policy
Explanation
Correct Answer
D:Unneeded public exposure expands attack surface. The intended behavior is: Minimized exposure reduces attack options.
Incorrect Answers
A:This checks validate caller identity, not the named failure. Tool endpoint discovery must not imply public access.
B:This checks isolate backend credentials, not the named failure. Clients should not receive administrative secrets.
C:Traffic and inventory can be intercepted.
Question 891
A network MCP server offers read_interface and disable_interface operations to an AI agent. The specific implementation objective is to classify read versus write. Which configuration best meets the stated requirement?
- Say 'shutdown complete' without executing or checking the device
- Let a model shut production uplinks based on one low-confidence alert
- Accept a natural-language nickname as sole target identifier
- Mark and govern configuration changes separately from diagnostic reads
Explanation
Correct Answer
D:Side effects define risk, not output format.
Incorrect Answers
A:Narrative success can be false.
B:Unapproved shutdown can cause outage.
C:Ambiguous names can select a different port.
Question 892
A network MCP server offers read_interface and disable_interface operations to an AI agent. The specific implementation objective is to require explicit approval. Which design action is most appropriate for this case?
- Gate interface shutdown behind human authorization and policy checks
- Accept a natural-language nickname as sole target identifier
- Treat every JSON-returning tool as inherently read-only
- Say 'shutdown complete' without executing or checking the device
Explanation
Correct Answer
A:High-impact mutations require accountable intent.
Incorrect Answers
B:Ambiguous names can select a different port.
C:A JSON response can follow a destructive write.
D:Narrative success can be false.
Question 893
A network MCP server offers read_interface and disable_interface operations to an AI agent. The specific implementation objective is to validate target boundaries. What should the engineer implement for this specific objective?
- Confirm interface ID, site and blast radius at the server boundary
- Let a model shut production uplinks based on one low-confidence alert
- Say 'shutdown complete' without executing or checking the device
- Treat every JSON-returning tool as inherently read-only
Explanation
Correct Answer
A:Wrong-target requests are high risk.
Incorrect Answers
B:Unapproved shutdown can cause outage.
C:Narrative success can be false.
D:A JSON response can follow a destructive write.
Question 894
A network MCP server offers read_interface and disable_interface operations to an AI agent. The specific implementation objective is to verify actual outcome. Which adjustment would satisfy this particular constraint?
- Accept a natural-language nickname as sole target identifier
- Let a model shut production uplinks based on one low-confidence alert
- Read back interface state and report any mismatch after authorized change
- Treat every JSON-returning tool as inherently read-only
Explanation
Correct Answer
C:Tool success must reflect real device state.
Incorrect Answers
A:Ambiguous names can select a different port.
B:Unapproved shutdown can cause outage.
D:A JSON response can follow a destructive write.
Question 895
A network MCP server offers read_interface and disable_interface operations to an AI agent. The deployment plan must both (1) require explicit approval and (2) verify actual outcome. Which TWO actions together satisfy those requirements? Choose TWO.
- Let a model shut production uplinks based on one low-confidence alert
- Say 'shutdown complete' without executing or checking the device
- Treat every JSON-returning tool as inherently read-only
- Read back interface state and report any mismatch after authorized change
- Gate interface shutdown behind human authorization and policy checks
Explanation
Correct Answers
D:Tool success must reflect real device state.
E:High-impact mutations require accountable intent.
Incorrect Answers
A:Unapproved shutdown can cause outage.
B:Narrative success can be false.
C:A JSON response can follow a destructive write.
Question 896
A network MCP server offers read_interface and disable_interface operations to an AI agent. Three conditions must be met: (1) require explicit approval; (2) validate target boundaries; and (3) verify actual outcome. Which THREE actions form the narrowest complete plan? Choose THREE.
- Confirm interface ID, site and blast radius at the server boundary
- Let a model shut production uplinks based on one low-confidence alert
- Treat every JSON-returning tool as inherently read-only
- Accept a natural-language nickname as sole target identifier
- Read back interface state and report any mismatch after authorized change
- Gate interface shutdown behind human authorization and policy checks
Explanation
Correct Answers
A:Wrong-target requests are high risk.
E:Tool success must reflect real device state.
F:High-impact mutations require accountable intent.
Incorrect Answers
B:Unapproved shutdown can cause outage.
C:A JSON response can follow a destructive write.
D:Ambiguous names can select a different port.
Question 897
The agent treats disable_interface as harmless because both tools return JSON. Independently, reviewers flag this potentially unsafe practice for assessment: 'Treat every JSON-returning tool as inherently read-only'. Which inspection most directly evaluates that practice?
- Inspect approval and change ticket
- Inspect tool implementation side effects
- Inspect resolved target and site map
- Say 'shutdown complete' without executing or checking the device
Explanation
Correct Answer
B:A JSON response can follow a destructive write. The intended behavior is: Side effects define risk, not output format.
Incorrect Answers
A:This checks require explicit approval, not the named failure. High-impact mutations require accountable intent.
C:This checks validate target boundaries, not the named failure. Wrong-target requests are high risk.
D:Narrative success can be false.
Question 898
The agent treats disable_interface as harmless because both tools return JSON. Independently, reviewers flag this potentially unsafe practice for assessment: 'Let a model shut production uplinks based on one low-confidence alert'. What diagnostic evidence should first be collected to assess that practice?
- Inspect tool implementation side effects
- Inspect approval and change ticket
- Inspect resolved target and site map
- Say 'shutdown complete' without executing or checking the device
Explanation
Correct Answer
B:Unapproved shutdown can cause outage. The intended behavior is: High-impact mutations require accountable intent.
Incorrect Answers
A:This checks classify read versus write, not the named failure. Side effects define risk, not output format.
C:This checks validate target boundaries, not the named failure. Wrong-target requests are high risk.
D:Narrative success can be false.
Question 899
The agent treats disable_interface as harmless because both tools return JSON. Independently, reviewers flag this potentially unsafe practice for assessment: 'Accept a natural-language nickname as sole target identifier'. Which focused check best tests this specific assumption?
- Inspect tool implementation side effects
- Inspect resolved target and site map
- Inspect approval and change ticket
- Say 'shutdown complete' without executing or checking the device
Explanation
Correct Answer
B:Ambiguous names can select a different port. The intended behavior is: Wrong-target requests are high risk.
Incorrect Answers
A:This checks classify read versus write, not the named failure. Side effects define risk, not output format.
C:This checks require explicit approval, not the named failure. High-impact mutations require accountable intent.
D:Narrative success can be false.
Question 900
The agent treats disable_interface as harmless because both tools return JSON. Independently, reviewers flag this potentially unsafe practice for assessment: 'Say 'shutdown complete' without executing or checking the device'. Which verification addresses that practice without assuming it caused the incident?
- Treat every JSON-returning tool as inherently read-only
- Inspect tool result and device operational state
- Inspect approval and change ticket
- Inspect resolved target and site map
Explanation
Correct Answer
B:Narrative success can be false. The intended behavior is: Tool success must reflect real device state.
Incorrect Answers
A:A JSON response can follow a destructive write.
C:This checks require explicit approval, not the named failure. High-impact mutations require accountable intent.
D:This checks validate target boundaries, not the named failure. Wrong-target requests are high risk.
Designing, Deploying, and Managing Network Automation Systems (350-901) Practice Tests