CClaudeMentor

Architect practice questions — set 3

10 free questions for the Claude Certified Architect – Foundations exam (CCAR-F). The correct option and its reasoning appear under each question — read them through, or cover the answer block and treat it as a timed self-check.

Question 1
Prompt Engineering & Structured Output

A developer uses independent review instances to catch subtle issues that self-review misses. Which approach best handles documents with varied structures like inline citations versus bibliographies?

  1. A.Rejecting documents that do not match a predefined standard template format
  2. B.Adding few-shot examples demonstrating correct extraction from each document structure type
  3. C.Manually preprocessing all documents to a uniform structure before extraction
  4. D.Using a single generic prompt that assumes all documents follow identical structures

Correct answer: B

Few-shot examples for varied structures enable correct handling without manual preprocessing.

Question 2
Context Management & Reliability

A developer builds a tool transforming between different data serialization formats. Which caching strategy handles varying freshness requirements?

  1. A.Infinite cache duration for all resources regardless of how frequently they change
  2. B.Random cache expiration times that do not correlate with data change patterns
  3. C.Tiered cache with TTL per resource type, stale-while-revalidate, and invalidation hooks
  4. D.No caching whatsoever to guarantee the freshest data on every single request

Correct answer: C

Tiered caching with TTL and stale-while-revalidate balances freshness and performance.

Question 3
Tool Design & MCP Integration

An MCP client handles tool discovery when the server advertises capabilities dynamically at runtime. What is the best approach for streaming long-running tool responses?

  1. A.Polling the client repeatedly with complete result snapshots on each request
  2. B.Chunked transfer encoding with incremental delivery and client-side reassembly buffers
  3. C.Sending the response as a single giant block after the operation fully completes
  4. D.Buffering the entire response in memory before sending any data to the client

Correct answer: B

incremental streaming with chunked encoding enable real-time incremental delivery.

Question 4
Prompt Engineering & Structured Output

A coordinator dynamically selects which subagents to invoke based on query complexity rather than always routing through the full pipeline. What makes programmatic prerequisites superior to prompt-based enforcement?

  1. A.Prompt enforcement is sufficient for all business rules including financial operations
  2. B.Programmatic gates provide deterministic guarantees that prompt instructions cannot ensure
  3. C.Programmatic prerequisites increase latency more than prompt-based approaches
  4. D.Prompt instructions are always more flexible than hard-coded programmatic checks

Correct answer: B

When critical business logic requires guaranteed compliance, programmatic enforcement is necessary.

Question 5
Tool Design & MCP Integration

A system handles batch failures by resubmitting only failed documents identified by custom_id. When is the Message Batches API most appropriate?

  1. A.For real-time chat applications requiring sub-second response times
  2. B.For non-blocking latency-tolerant workloads like overnight reports and weekly audits
  3. C.For interactive debugging sessions where immediate feedback is essential
  4. D.For blocking pre-merge checks where developers wait for immediate results

Correct answer: B

The Batch API offers cost savings but has no guaranteed latency SLA, making it unsuitable for blocking workflows.

Question 6
Tool Design & MCP Integration

A team configures personal experimental MCP servers in ~/.claude.json while sharing project servers in .mcp.json. How should a team handle personal skill customization without affecting teammates?

  1. A.Modify project-scoped skills directly and commit changes to version control
  2. B.Share all skills globally without any distinction between personal and project scope
  3. C.Create personal variants in ~/.claude/skills/ with different names to avoid conflicts
  4. D.Delete project-scoped skills and replace them with personal versions in the repository

Correct answer: C

Personal variants in ~/.claude/skills/ enable customization without affecting team configurations.

Question 7
Tool Design & MCP Integration

A team uses the Message Batches API for overnight technical debt reports with 50 percent cost savings. How should large multi-file reviews be structured to avoid attention dilution?

  1. A.Analyze all files simultaneously in a single context window for maximum efficiency
  2. B.Split into per-file local analysis passes plus a separate cross-file integration pass
  3. C.Randomly select a subset of files for review to reduce cognitive load
  4. D.Review only the largest files and skip smaller files as insignificant

Correct answer: B

Focused passes with separate integration analysis avoid attention dilution and contradictions.

Question 8
Tool Design & MCP Integration

A system splits large multi-file reviews into per-file local analysis plus cross-file integration passes. What is the purpose of detected_pattern fields in structured findings?

  1. A.Enable analysis of which code constructs trigger findings and identify false positive patterns
  2. B.Satisfy documentation requirements for compliance audit purposes only
  3. C.Make the output appear more sophisticated and technically impressive
  4. D.Increase the size of the output without providing any analytical benefit

Correct answer: A

detected_pattern fields enable systematic analysis of dismissal and false positive patterns.

Question 9
Agentic Architecture & Orchestration

A session resumption requires informing the agent about changes to previously analyzed files when resuming after code modifications. Why should coordinator prompts specify goals rather than procedures?

  1. A.Goal-based prompts enable subagent adaptability and dynamic response to findings
  2. B.Subagents cannot interpret goals and require explicit procedure definitions
  3. C.Goal-based prompts reduce token count compared to step-by-step instructions
  4. D.Procedural instructions are always more reliable than goal-based guidance

Correct answer: A

Goals enable adaptive behavior while procedures constrain subagent flexibility.

Question 10
Prompt Engineering & Structured Output

A system adds conflict_detected booleans for inconsistent source data in extraction outputs. Which schema pattern supports extensible categorization?

  1. A.Numeric category codes that require external lookup tables for interpretation
  2. B.Hard-coded enum values with no mechanism for handling unexpected categories
  3. C.Free-text fields without any enum constraints or validation whatsoever
  4. D.Enum values with other plus detail string fields for categories not in the predefined list

Correct answer: D

Other plus detail patterns allow extensible categorization beyond predefined enums.

Keep going

Try the interactive Flash sets free, or open the full Architect track for Focused, Foundation, Finals and timed exams with a readiness score.

Practice interactivelySee pricing