Cisco 350-901 Exam Dumps, Practice Test Questions

100% Latest & Updated Cisco 350-901 Practice Test Questions, Exam Dumps & Verified Answers!
30 Days Free Updates, Instant Download!

Cisco 350-901 Premium Bundle
$79.97
$59.98

350-901 Premium Bundle

  • Premium File: 604 Questions & Answers. Last update: Nov 7, 2026
  • Training Course: 106 Video Lectures
  • Study Guide: 1255 Pages
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates

350-901 Premium Bundle

Cisco 350-901 Premium Bundle
  • Premium File: 604 Questions & Answers. Last update: Nov 7, 2026
  • Training Course: 106 Video Lectures
  • Study Guide: 1255 Pages
  • Latest Questions
  • 100% Accurate Answers
  • Fast Exam Updates
$79.97
$59.98

Cisco 350-901 Practice Test 87

Question 860

A prototype stores one full-admin token in prompt context for every user conversation. Independently, reviewers flag this potentially unsafe practice for assessment: 'Store only a model summary saying 'done' after changes'. Which verification addresses that practice without assuming it caused the incident?

  1. Inspect user-to-tool policy
  2. Use one unrestricted administrator token for all agent activities
  3. Inspect prompt and runtime secret design
  4. Inspect tool and device audit records

Explanation

Correct Answer

D:

Narration cannot prove execution. The intended behavior is: Action traceability supports accountability.

Incorrect Answers

A:

This checks check user authority, not the named failure. Conversation intent is not sufficient authorization.

B:

Token compromise exposes entire network.

C:

This checks keep secrets outside prompts, not the named failure. Prompt context can be disclosed or manipulated.

Question 861

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. The specific implementation objective is to register a named Python tool. Which configuration best meets the stated requirement?

  1. Run show-interface tools through a global write-admin identity by default
  2. Return unbounded full configs including secrets for a status request
  3. Use a FastMCP-decorated function with a narrow purpose and typed parameters
  4. Allow the agent to submit any reachable management IP

Explanation

Correct Answer

C:

Explicit tool schemas help agents select allowed operations.

Incorrect Answers

A:

Needless privilege increases compromise impact.

B:

Excess output risks leakage and confusion.

D:

Unbounded targets risk cross-scope access.

Question 862

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. The specific implementation objective is to validate device selector. Which design action is most appropriate for this case?

  1. Accept only authorized device IDs from the server-side inventory
  2. Expose unrestricted arbitrary shell execution as the interface lookup tool
  3. Return unbounded full configs including secrets for a status request
  4. Run show-interface tools through a global write-admin identity by default

Explanation

Correct Answer

A:

User input must not become arbitrary target selection.

Incorrect Answers

B:

General shell access expands attack surface.

C:

Excess output risks leakage and confusion.

D:

Needless privilege increases compromise impact.

Question 863

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. The specific implementation objective is to return structured bounded results. What should the engineer implement for this specific objective?

  1. Allow the agent to submit any reachable management IP
  2. Provide interface status with identifiers and a bounded result size
  3. Expose unrestricted arbitrary shell execution as the interface lookup tool
  4. Run show-interface tools through a global write-admin identity by default

Explanation

Correct Answer

B:

Predictable outputs make downstream reasoning checkable.

Incorrect Answers

A:

Unbounded targets risk cross-scope access.

C:

General shell access expands attack surface.

D:

Needless privilege increases compromise impact.

Question 864

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. The specific implementation objective is to keep tool read-only. Which adjustment would satisfy this particular constraint?

  1. Allow the agent to submit any reachable management IP
  2. Expose unrestricted arbitrary shell execution as the interface lookup tool
  3. Return unbounded full configs including secrets for a status request
  4. Use a read-scoped controller identity and avoid configuration mutations

Explanation

Correct Answer

D:

Read-only design matches diagnostic need.

Incorrect Answers

A:

Unbounded targets risk cross-scope access.

B:

General shell access expands attack surface.

C:

Excess output risks leakage and confusion.

Question 865

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. The deployment plan must both (1) register a named Python tool and (2) return structured bounded results. Which TWO actions together satisfy those requirements? Choose TWO.

  1. Provide interface status with identifiers and a bounded result size
  2. Run show-interface tools through a global write-admin identity by default
  3. Expose unrestricted arbitrary shell execution as the interface lookup tool
  4. Return unbounded full configs including secrets for a status request
  5. Use a FastMCP-decorated function with a narrow purpose and typed parameters

Explanation

Correct Answers

A:

Predictable outputs make downstream reasoning checkable.

E:

Explicit tool schemas help agents select allowed operations.

Incorrect Answers

B:

Needless privilege increases compromise impact.

C:

General shell access expands attack surface.

D:

Excess output risks leakage and confusion.

Question 866

A network AI assistant needs an MCP server that lists interface operational states from approved lab routers. Three conditions must be met: (1) validate device selector; (2) return structured bounded results; and (3) keep tool read-only. Which THREE actions form the narrowest complete plan? Choose THREE.

  1. Run show-interface tools through a global write-admin identity by default
  2. Return unbounded full configs including secrets for a status request
  3. Use a read-scoped controller identity and avoid configuration mutations
  4. Expose unrestricted arbitrary shell execution as the interface lookup tool
  5. Provide interface status with identifiers and a bounded result size
  6. Accept only authorized device IDs from the server-side inventory

Explanation

Correct Answers

C:

Read-only design matches diagnostic need.

E:

Predictable outputs make downstream reasoning checkable.

F:

User input must not become arbitrary target selection.

Incorrect Answers

A:

Needless privilege increases compromise impact.

B:

Excess output risks leakage and confusion.

D:

General shell access expands attack surface.

Question 867

The MCP prototype exposes a general shell command instead of a constrained diagnostic tool. Independently, reviewers flag this potentially unsafe practice for assessment: 'Expose unrestricted arbitrary shell execution as the interface lookup tool'. Which inspection most directly evaluates that practice?

  1. Inspect execution identity and API methods
  2. Allow the agent to submit any reachable management IP
  3. Inspect response schema and result size
  4. Inspect registered tool schema

Explanation

Correct Answer

D:

General shell access expands attack surface. The intended behavior is: Explicit tool schemas help agents select allowed operations.

Incorrect Answers

A:

This checks keep tool read-only, not the named failure. Read-only design matches diagnostic need.

B:

Unbounded targets risk cross-scope access.

C:

This checks return structured bounded results, not the named failure. Predictable outputs make downstream reasoning checkable.

Question 868

The MCP prototype exposes a general shell command instead of a constrained diagnostic tool. Independently, reviewers flag this potentially unsafe practice for assessment: 'Allow the agent to submit any reachable management IP'. What diagnostic evidence should first be collected to assess that practice?

  1. Inspect execution identity and API methods
  2. Inspect registered tool schema
  3. Inspect parameter validation and inventory lookup
  4. Return unbounded full configs including secrets for a status request

Explanation

Correct Answer

C:

Unbounded targets risk cross-scope access. The intended behavior is: User input must not become arbitrary target selection.

Incorrect Answers

A:

This checks keep tool read-only, not the named failure. Read-only design matches diagnostic need.

B:

This checks register a named Python tool, not the named failure. Explicit tool schemas help agents select allowed operations.

D:

Excess output risks leakage and confusion.

Question 869

The MCP prototype exposes a general shell command instead of a constrained diagnostic tool. Independently, reviewers flag this potentially unsafe practice for assessment: 'Return unbounded full configs including secrets for a status request'. Which focused check best tests this specific assumption?

  1. Inspect parameter validation and inventory lookup
  2. Inspect registered tool schema
  3. Inspect response schema and result size
  4. Run show-interface tools through a global write-admin identity by default

Explanation

Correct Answer

C:

Excess output risks leakage and confusion. The intended behavior is: Predictable outputs make downstream reasoning checkable.

Incorrect Answers

A:

This checks validate device selector, not the named failure. User input must not become arbitrary target selection.

B:

This checks register a named Python tool, not the named failure. Explicit tool schemas help agents select allowed operations.

D:

Needless privilege increases compromise impact.

Question 870

The MCP prototype exposes a general shell command instead of a constrained diagnostic tool. Independently, reviewers flag this potentially unsafe practice for assessment: 'Run show-interface tools through a global write-admin identity by default'. Which verification addresses that practice without assuming it caused the incident?

  1. Allow the agent to submit any reachable management IP
  2. Inspect response schema and result size
  3. Inspect registered tool schema
  4. Inspect execution identity and API methods

Explanation

Correct Answer

D:

Needless privilege increases compromise impact. The intended behavior is: Read-only design matches diagnostic need.

Incorrect Answers

A:

Unbounded targets risk cross-scope access.

B:

This checks return structured bounded results, not the named failure. Predictable outputs make downstream reasoning checkable.

C:

This checks register a named Python tool, not the named failure. Explicit tool schemas help agents select allowed operations.

Designing, Deploying, and Managing Network Automation Systems (350-901) Practice Tests

UP

SPECIAL OFFER: GET 10% OFF

This is ONE TIME OFFER

ExamSnap Discount Offer
Enter Your Email Address to Receive Your 10% Off Discount Code

A confirmation link will be sent to this email address to verify your login. *We value your privacy. We will not rent or sell your email address.

Download Free Demo of VCE Exam Simulator

Experience Avanset VCE Exam Simulator for yourself.

Simply submit your e-mail address below to get started with our interactive software demo of your free trial.

Free Demo Limits: In the demo version you will be able to access only first 5 questions from exam.