CClaudeMentor

Architect practice questions — set 2

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
Context Management & Reliability

A team uses stratified random sampling to measure error rates in high-confidence extractions. How should large multi-file reviews be structured to avoid attention dilution?

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

Correct answer: D

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

Question 2
Claude Code Configuration & Workflows

A team integrates Claude Code into CI/CD pipelines for automated code review and test generation. How should a team ensure new members receive project coding standards?

  1. A.Email coding standards to new members during onboarding and hope they remember
  2. B.Document standards in a wiki that Claude Code does not automatically read
  3. C.Store standards in user-level configuration that each member must manually copy
  4. D.Place standards in project-level CLAUDE.md or .claude/CLAUDE.md shared via version control

Correct answer: D

Project-level configuration is version-controlled and automatically available to all team members.

Question 3
Agentic Architecture & Orchestration

An agentic loop control flow continues when stop_reason is tool_use and terminates when stop_reason is end_turn. What is essential for crash recovery in multi-agent systems?

  1. A.Taking snapshots randomly without coordination with ongoing operations
  2. B.Structured state exports to known locations with manifest loading on session resume
  3. C.Relying on agents to manually remember state and report it on restart
  4. D.Storing all state in volatile memory for maximum access speed and simplicity

Correct answer: B

Structured state exports with manifest loading enable reliable crash recovery.

Question 4
Tool Design & MCP Integration

An architect designs observability for tool chains spanning multiple distributed services. How should intermittent network connectivity be handled for in-flight requests?

  1. A.Sending duplicate requests continuously until one randomly succeeds
  2. B.Dropping all requests immediately upon any network error without retry attempts
  3. C.Request queueing with exponential backoff, idempotency keys, and dead-letter channels
  4. D.Blocking the caller indefinitely until connectivity is restored without timeout

Correct answer: C

Queueing with backoff and idempotency ensures reliable delivery despite intermittent connectivity.

Question 5
Agentic Architecture & Orchestration

A developer uses subagent delegation to isolate verbose exploration output while the main agent coordinates. Which approach best mitigates the lost in the middle effect?

  1. A.Randomly shuffling input sections so no single position is consistently overlooked
  2. B.Removing all middle sections and keeping only the beginning and end of the input
  3. C.Placing key findings summaries at the beginning and organizing detailed results with explicit headers
  4. D.Increasing the context window size without any reorganization of content positioning

Correct answer: C

Strategic placement and explicit headers mitigate position-based attention degradation.

Question 6
Agentic Architecture & Orchestration

A coordinator partitions research scope across subagents to minimize duplication by assigning distinct subtopics. What is the primary risk if the delegation mechanism fails without isolation?

  1. A.Improved throughput due to reduced coordination overhead unexpectedly
  2. B.Cascading failure across dependent sub-tasks without boundary containment or recovery
  3. C.Temporary cosmetic formatting issues in the final output presentation layer
  4. D.Slightly increased latency in non-critical response paths only with no data impact

Correct answer: B

Without isolation boundaries, a failure in one delegated task propagates and crashes the entire workflow.

Question 7
Context Management & Reliability

A developer provides concrete input/output examples to communicate expected transformations. What is the purpose of the context:fork frontmatter option in skills?

  1. A.Run skills in an isolated sub-agent context preventing outputs from polluting the main conversation
  2. B.Terminate the main conversation and replace it with the skill execution context
  3. C.Duplicate the main conversation context into the skill for maximum information access
  4. D.Merge skill outputs directly into the main conversation without any isolation boundary

Correct answer: A

context:fork isolates verbose or exploratory skill outputs from the main session context.

Question 8
Context Management & Reliability

A system structures synthesis output with coverage annotations indicating well-supported versus gap areas. Which approach best mitigates the lost in the middle effect?

  1. A.Randomly shuffling input sections so no single position is consistently overlooked
  2. B.Increasing the context window size without any reorganization of content positioning
  3. C.Placing key findings summaries at the beginning and organizing detailed results with explicit headers
  4. D.Removing all middle sections and keeping only the beginning and end of the input

Correct answer: C

Strategic placement and explicit headers mitigate position-based attention degradation.

Question 9
Agentic Architecture & Orchestration

Tool results are appended to conversation history so the model can reason about the next action. What is essential for crash recovery in multi-agent systems?

  1. A.Relying on agents to manually remember state and report it on restart
  2. B.Structured state exports to known locations with manifest loading on session resume
  3. C.Storing all state in volatile memory for maximum access speed and simplicity
  4. D.Taking snapshots randomly without coordination with ongoing operations

Correct answer: B

Structured state exports with manifest loading enable reliable crash recovery.

Question 10
Agentic Architecture & Orchestration

A PostToolUse hook intercepts tool results to normalize heterogeneous data formats before the agent processes them. Which architectural pattern best addresses this requirement?

  1. A.Hierarchical supervisor-worker with state checkpointing and retry logic for subagent delegation
  2. B.Random agent selection with optimistic execution and no rollback or state management
  3. C.Monolithic sequential pipeline with rigid ordering and no failover or recovery capability
  4. D.Flat peer-to-peer mesh with gossip-based consensus and no central coordination mechanism

Correct answer: A

The supervisor-worker pattern with checkpointing directly supports decomposition, delegation, and recovery in multi-agent systems.

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