100% Latest & Updated Cisco 350-901 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 970
The suggestion cites a generic tutorial instead of device-specific switch output. Independently, reviewers flag this potentially unsafe practice for assessment: 'Reward the model for filling every missing fact with a confident guess'. Which verification addresses that practice without assuming it caused the incident?
- Apply the tutorial's default CLI to every switch blindly
- Inspect evaluation evidence matrix
- Inspect device output and cited object
- Inspect source labels and timestamps
Explanation
Correct Answer
B:Fabricated details inflate apparent usefulness. The intended behavior is: Unsupported recommendations weaken reliability.
Incorrect Answers
A:Default examples may be unsafe for current topology.
C:This checks link claim to live source, not the named failure. Incident claims require instance-specific evidence.
D:This checks separate norms and observations, not the named failure. Documentation and telemetry play different roles.
Question 971
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. The specific implementation objective is to include negative authorization cases. Which configuration best meets the stated requirement?
- Approve new agent version because its average answer length improved
- Test only clean prompts that contain no adversarial text
- Treat all VLAN changes as correct regardless of site
- Test that unapproved users cannot trigger privileged mutations
Explanation
Correct Answer
D:Security requirements need explicit failing examples.
Incorrect Answers
A:Verbose answers are not a safety metric.
B:Clean examples cannot expose tool-redirection failures.
C:Correct syntax on wrong devices is failure.
Question 972
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. The specific implementation objective is to score target precision. Which design action is most appropriate for this case?
- Evaluate only scenarios where every command is approved
- Approve new agent version because its average answer length improved
- Measure whether device, interface and site IDs match the request
- Test only clean prompts that contain no adversarial text
Explanation
Correct Answer
C:Wrong-target actions are dangerous even when commands are valid.
Incorrect Answers
A:Happy-path coverage misses access violations.
B:Verbose answers are not a safety metric.
D:Clean examples cannot expose tool-redirection failures.
Question 973
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. The specific implementation objective is to measure prompt injection resistance. What should the engineer implement for this specific objective?
- Treat all VLAN changes as correct regardless of site
- Approve new agent version because its average answer length improved
- Use controlled hostile documents in tests to see whether tool policy holds
- Evaluate only scenarios where every command is approved
Explanation
Correct Answer
C:Lower-trust content must not gain authority.
Incorrect Answers
A:Correct syntax on wrong devices is failure.
B:Verbose answers are not a safety metric.
D:Happy-path coverage misses access violations.
Question 974
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. The specific implementation objective is to block unsafe regressions. Which adjustment would satisfy this particular constraint?
- Evaluate only scenarios where every command is approved
- Treat all VLAN changes as correct regardless of site
- Test only clean prompts that contain no adversarial text
- Set release thresholds on critical unauthorized actions and evaluate before promotion
Explanation
Correct Answer
D:Severe failures should halt deployment.
Incorrect Answers
A:Happy-path coverage misses access violations.
B:Correct syntax on wrong devices is failure.
C:Clean examples cannot expose tool-redirection failures.
Question 975
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. The deployment plan must both (1) score target precision and (2) block unsafe regressions. Which TWO actions together satisfy those requirements? Choose TWO.
- Evaluate only scenarios where every command is approved
- Measure whether device, interface and site IDs match the request
- Treat all VLAN changes as correct regardless of site
- Set release thresholds on critical unauthorized actions and evaluate before promotion
- Approve new agent version because its average answer length improved
Explanation
Correct Answers
B:Wrong-target actions are dangerous even when commands are valid.
D:Severe failures should halt deployment.
Incorrect Answers
A:Happy-path coverage misses access violations.
C:Correct syntax on wrong devices is failure.
E:Verbose answers are not a safety metric.
Question 976
An internal agent can suggest and, with approval, implement VLAN configuration through network tools. Three conditions must be met: (1) include negative authorization cases; (2) measure prompt injection resistance; and (3) block unsafe regressions. Which THREE actions form the narrowest complete plan? Choose THREE.
- Test that unapproved users cannot trigger privileged mutations
- Test only clean prompts that contain no adversarial text
- Approve new agent version because its average answer length improved
- Evaluate only scenarios where every command is approved
- Set release thresholds on critical unauthorized actions and evaluate before promotion
- Use controlled hostile documents in tests to see whether tool policy holds
Explanation
Correct Answers
A:Security requirements need explicit failing examples.
E:Severe failures should halt deployment.
F:Lower-trust content must not gain authority.
Incorrect Answers
B:Clean examples cannot expose tool-redirection failures.
C:Verbose answers are not a safety metric.
D:Happy-path coverage misses access violations.
Question 977
After an LLM upgrade, the agent starts selecting broad interface targets more often than before. Independently, reviewers flag this potentially unsafe practice for assessment: 'Evaluate only scenarios where every command is approved'. Which inspection most directly evaluates that practice?
- Approve new agent version because its average answer length improved
- Inspect injection-test tool traces
- Inspect test authorization matrix
- Inspect selected target IDs
Explanation
Correct Answer
C:Happy-path coverage misses access violations. The intended behavior is: Security requirements need explicit failing examples.
Incorrect Answers
A:Verbose answers are not a safety metric.
B:This checks measure prompt injection resistance, not the named failure. Lower-trust content must not gain authority.
D:This checks score target precision, not the named failure. Wrong-target actions are dangerous even when commands are valid.
Question 978
After an LLM upgrade, the agent starts selecting broad interface targets more often than before. Independently, reviewers flag this potentially unsafe practice for assessment: 'Treat all VLAN changes as correct regardless of site'. What diagnostic evidence should first be collected to assess that practice?
- Inspect selected target IDs
- Inspect injection-test tool traces
- Inspect critical-failure gates
- Evaluate only scenarios where every command is approved
Explanation
Correct Answer
A:Correct syntax on wrong devices is failure. The intended behavior is: Wrong-target actions are dangerous even when commands are valid.
Incorrect Answers
B:This checks measure prompt injection resistance, not the named failure. Lower-trust content must not gain authority.
C:This checks block unsafe regressions, not the named failure. Severe failures should halt deployment.
D:Happy-path coverage misses access violations.
Question 979
After an LLM upgrade, the agent starts selecting broad interface targets more often than before. Independently, reviewers flag this potentially unsafe practice for assessment: 'Test only clean prompts that contain no adversarial text'. Which focused check best tests this specific assumption?
- Inspect test authorization matrix
- Inspect injection-test tool traces
- Approve new agent version because its average answer length improved
- Inspect selected target IDs
Explanation
Correct Answer
B:Clean examples cannot expose tool-redirection failures. The intended behavior is: Lower-trust content must not gain authority.
Incorrect Answers
A:This checks include negative authorization cases, not the named failure. Security requirements need explicit failing examples.
C:Verbose answers are not a safety metric.
D:This checks score target precision, not the named failure. Wrong-target actions are dangerous even when commands are valid.
Question 980
After an LLM upgrade, the agent starts selecting broad interface targets more often than before. Independently, reviewers flag this potentially unsafe practice for assessment: 'Approve new agent version because its average answer length improved'. Which verification addresses that practice without assuming it caused the incident?
- Inspect critical-failure gates
- Inspect injection-test tool traces
- Treat all VLAN changes as correct regardless of site
- Inspect test authorization matrix
Explanation
Correct Answer
A:Verbose answers are not a safety metric. The intended behavior is: Severe failures should halt deployment.
Incorrect Answers
B:This checks measure prompt injection resistance, not the named failure. Lower-trust content must not gain authority.
C:Correct syntax on wrong devices is failure.
D:This checks include negative authorization cases, not the named failure. Security requirements need explicit failing examples.
Designing, Deploying, and Managing Network Automation Systems (350-901) Practice Tests