100% Latest & Updated Cisco 350-901 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!
Question 990
One proposed Python library has a plausible name but no trustworthy verified publisher. Independently, reviewers flag this potentially unsafe practice for assessment: 'Omit all dependency origins from the release manifest'. Which verification addresses that practice without assuming it caused the incident?
- Inspect SBOM or manifest
- Allow unreviewed latest packages into production automation
- Inspect runner trust boundary
- Inspect registry and publisher identity
Explanation
Correct Answer
A:Untracked sources complicate incident response. The intended behavior is: Maintenance requires traceable library selection.
Incorrect Answers
B:New releases may contain breaking or malicious behavior.
C:This checks run isolated tests, not the named failure. A test environment limits potential damage.
D:This checks validate package existence, not the named failure. Fabricated package names can enable dependency confusion.
Question 991
An AI network tool returns a CLI command suggestion in free-form text. The specific implementation objective is to treat model output as untrusted. Which configuration best meets the stated requirement?
- Execute every model suggestion automatically when confidence is high
- Allow arbitrary shell expansion in device command arguments
- Parse and validate proposed commands before any execution
- Assume the model's final answer is an operational state readback
Explanation
Correct Answer
C:Generated output cannot be trusted as executable authorization.
Incorrect Answers
A:Confidence is not authorization.
B:Shell metacharacters can alter execution.
D:Narrative cannot replace telemetry.
Question 992
An AI network tool returns a CLI command suggestion in free-form text. The specific implementation objective is to use structured allowlisted operations. Which design action is most appropriate for this case?
- Assume the model's final answer is an operational state readback
- Execute every model suggestion automatically when confidence is high
- Pass raw LLM text directly to a privileged shell
- Map approved intents to typed device API operations rather than arbitrary shell
Explanation
Correct Answer
D:Constrained operations reduce injection surface.
Incorrect Answers
A:Narrative cannot replace telemetry.
B:Confidence is not authorization.
C:Unvalidated command text can cause arbitrary changes.
Question 993
An AI network tool returns a CLI command suggestion in free-form text. The specific implementation objective is to require approval for side effects. What should the engineer implement for this specific objective?
- Allow arbitrary shell expansion in device command arguments
- Review target and exact config diff before running changes
- Assume the model's final answer is an operational state readback
- Pass raw LLM text directly to a privileged shell
Explanation
Correct Answer
B:Explicit approval controls network mutation.
Incorrect Answers
A:Shell metacharacters can alter execution.
C:Narrative cannot replace telemetry.
D:Unvalidated command text can cause arbitrary changes.
Question 994
An AI network tool returns a CLI command suggestion in free-form text. The specific implementation objective is to validate result. Which adjustment would satisfy this particular constraint?
- Execute every model suggestion automatically when confidence is high
- Compare device state with approved intent and monitor side effects
- Pass raw LLM text directly to a privileged shell
- Allow arbitrary shell expansion in device command arguments
Explanation
Correct Answer
B:Execution response does not prove desired behavior.
Incorrect Answers
A:Confidence is not authorization.
C:Unvalidated command text can cause arbitrary changes.
D:Shell metacharacters can alter execution.
Question 995
An AI network tool returns a CLI command suggestion in free-form text. The deployment plan must both (1) use structured allowlisted operations and (2) validate result. Which TWO actions together satisfy those requirements? Choose TWO.
- Pass raw LLM text directly to a privileged shell
- Compare device state with approved intent and monitor side effects
- Assume the model's final answer is an operational state readback
- Execute every model suggestion automatically when confidence is high
- Map approved intents to typed device API operations rather than arbitrary shell
Explanation
Correct Answers
B:Execution response does not prove desired behavior.
E:Constrained operations reduce injection surface.
Incorrect Answers
A:Unvalidated command text can cause arbitrary changes.
C:Narrative cannot replace telemetry.
D:Confidence is not authorization.
Question 996
An AI network tool returns a CLI command suggestion in free-form text. Three conditions must be met: (1) treat model output as untrusted; (2) use structured allowlisted operations; and (3) validate result. Which THREE actions form the narrowest complete plan? Choose THREE.
- Allow arbitrary shell expansion in device command arguments
- Compare device state with approved intent and monitor side effects
- Parse and validate proposed commands before any execution
- Assume the model's final answer is an operational state readback
- Map approved intents to typed device API operations rather than arbitrary shell
- Pass raw LLM text directly to a privileged shell
Explanation
Correct Answers
B:Execution response does not prove desired behavior.
C:Generated output cannot be trusted as executable authorization.
E:Constrained operations reduce injection surface.
Incorrect Answers
A:Shell metacharacters can alter execution.
D:Narrative cannot replace telemetry.
F:Unvalidated command text can cause arbitrary changes.
Question 997
The suggested command contains additional semicolon-delimited operations not requested by the operator. Independently, reviewers flag this potentially unsafe practice for assessment: 'Pass raw LLM text directly to a privileged shell'. Which inspection most directly evaluates that practice?
- Inspect device confirmation
- Inspect command validation boundary
- Allow arbitrary shell expansion in device command arguments
- Inspect tool audit and approval
Explanation
Correct Answer
B:Unvalidated command text can cause arbitrary changes. The intended behavior is: Generated output cannot be trusted as executable authorization.
Incorrect Answers
A:This checks validate result, not the named failure. Execution response does not prove desired behavior.
C:Shell metacharacters can alter execution.
D:This checks require approval for side effects, not the named failure. Explicit approval controls network mutation.
Question 998
The suggested command contains additional semicolon-delimited operations not requested by the operator. Independently, reviewers flag this potentially unsafe practice for assessment: 'Allow arbitrary shell expansion in device command arguments'. What diagnostic evidence should first be collected to assess that practice?
- Inspect command validation boundary
- Inspect tool audit and approval
- Assume the model's final answer is an operational state readback
- Inspect allowed operation schema
Explanation
Correct Answer
D:Shell metacharacters can alter execution. The intended behavior is: Constrained operations reduce injection surface.
Incorrect Answers
A:This checks treat model output as untrusted, not the named failure. Generated output cannot be trusted as executable authorization.
B:This checks require approval for side effects, not the named failure. Explicit approval controls network mutation.
C:Narrative cannot replace telemetry.
Question 999
The suggested command contains additional semicolon-delimited operations not requested by the operator. Independently, reviewers flag this potentially unsafe practice for assessment: 'Execute every model suggestion automatically when confidence is high'. Which focused check best tests this specific assumption?
- Allow arbitrary shell expansion in device command arguments
- Inspect device confirmation
- Inspect command validation boundary
- Inspect tool audit and approval
Explanation
Correct Answer
D:Confidence is not authorization. The intended behavior is: Explicit approval controls network mutation.
Incorrect Answers
A:Shell metacharacters can alter execution.
B:This checks validate result, not the named failure. Execution response does not prove desired behavior.
C:This checks treat model output as untrusted, not the named failure. Generated output cannot be trusted as executable authorization.
Question 1000
The suggested command contains additional semicolon-delimited operations not requested by the operator. Independently, reviewers flag this potentially unsafe practice for assessment: 'Assume the model's final answer is an operational state readback'. Which verification addresses that practice without assuming it caused the incident?
- Execute every model suggestion automatically when confidence is high
- Inspect device confirmation
- Inspect command validation boundary
- Inspect allowed operation schema
Explanation
Correct Answer
B:Narrative cannot replace telemetry. The intended behavior is: Execution response does not prove desired behavior.
Incorrect Answers
A:Confidence is not authorization.
C:This checks treat model output as untrusted, not the named failure. Generated output cannot be trusted as executable authorization.
D:This checks use structured allowlisted operations, not the named failure. Constrained operations reduce injection surface.
Designing, Deploying, and Managing Network Automation Systems (350-901) Practice Tests