Anthropic CCA-F (Claude Certified Architect - Foundations (CCAR-F)) Practice Test 8

Question 70

A deterministic XML-to-JSON conversion is tested and stable, but an engineer wants a separate autonomous agent for every field. Which advice is strongest?

  1. Use agents because more model turns necessarily increase accuracy
  2. Ask the coordinator to infer field mappings from unrelated conversation history
  3. Discard the stable converter to make the architecture appear more advanced
  4. Keep the transformation in ordinary code and reserve agents for actual judgment

Explanation

Correct Answer

D:

Known transformations benefit from predictable logic and cheap tests. Agent delegation adds latency, cost and unnecessary uncertainty.

Incorrect Answers

A:

Repeated generations do not guarantee faithful deterministic conversion. They can introduce additional errors.

B:

A stable schema mapping should be explicit, not inferred from noisy context. This would reduce reproducibility.

C:

Complexity is not a goal in itself. A proven deterministic component should be retained when it satisfies the contract.

Question 71

An interrupted coding session must continue from its prior decisions and open tasks. Which session strategy preserves that history?

  1. Start a fresh session without supplying any prior state
  2. Ask the model to reconstruct missing actions from memory alone
  3. Fork an unrelated team member's session by convenience
  4. Resume the saved session, then check current workspace state

Explanation

Correct Answer

D:

Resumption continues the conversation context. The host should still check files and external state that may have changed during the interruption.

Incorrect Answers

A:

A fresh thread omits prior decisions and outstanding work. The agent may repeat steps or contradict previously approved changes.

B:

Model recall is not durable task state. Recovery needs the actual saved session and authoritative artifacts.

C:

A fork branches a specific existing history, not an arbitrary replacement. It would carry irrelevant decisions and potentially private context.

Question 72

A team wants to explore two alternative designs from one approved baseline without one branch changing the other's reasoning history. What should it do?

  1. Fork the baseline session into two independent branches
  2. Use a Stop hook to merge all alternatives before they exist
  3. Resume the same session twice while writing both alternatives into one transcript
  4. Delete the baseline and begin both from unrelated prompts

Explanation

Correct Answer

A:

Session forking preserves a shared starting point while isolating subsequent decisions. The team can compare outcomes before selecting a path.

Incorrect Answers

B:

A termination hook cannot replace deliberate branch creation and comparison of separate design states.

C:

Shared continuation mixes conflicting hypotheses and decisions. Branch isolation is the requirement here.

D:

Throwing away the verified common context increases rework and makes comparison less controlled.

Question 73

A customer account changes owners, and a new case must not see the previous owner's private support dialogue. Which session choice is safest?

  1. Start a new session with only authorized current-case context
  2. Ask the model to forget the old owner but keep the same session
  3. Resume the previous owner's session to save tokens
  4. Fork the previous owner's session and hide its title

Explanation

Correct Answer

A:

A fresh conversation boundary prevents accidental carryover of another person's sensitive history. Required facts should be reloaded from trusted systems.

Incorrect Answers

B:

A text instruction does not guarantee that sensitive prior content is unavailable. Isolation requires a new authorized context.

C:

Continuing that history can expose private content and bias decisions with unrelated facts. Data segregation takes priority over convenience.

D:

Forking still inherits the original conversation context. Renaming does not remove sensitive material.

Question 74

A suspended research run stores a session identifier but loses the associated tool execution ledger. What is the remaining risk after resuming?

  1. The assistant will always receive an end_turn response after resume
  2. The session identifier can authorize refunds without policy checks
  3. None, because every tool call is automatically reversible
  4. Session history alone cannot prevent replayed external writes

Explanation

Correct Answer

D:

Session continuation restores conversational context, but safe side-effect recovery also requires durable records of completed calls and outcomes.

Incorrect Answers

A:

Termination depends on subsequent actions and observations. A stored session alone does not guarantee completion.

B:

A session id is not a transactional permission grant. Authorization remains an application control.

C:

Many external operations are not reversible and may have succeeded before interruption. Recovery logic must check actual state.

Question 75

A platform team is documenting session choices for three situations: continue an interrupted debugging thread, test alternatives from the same baseline, and begin an unrelated customer case. Which THREE mappings are appropriate? Choose THREE.

  1. Fork the baseline session to compare alternative fixes independently
  2. Resume a persisted session to continue the debugging conversation
  3. Resume an unrelated customer's session to save prompt tokens
  4. Use a Stop hook to make the filesystem revert whenever a session resumes
  5. Fork a conversation to renew expired financial permissions automatically
  6. Start a fresh session for an unrelated customer's case

Explanation

Correct Answers

A:

Forking preserves common earlier context while separating subsequent reasoning histories. It supports controlled comparison.

B:

Resumption retains relevant conversational state, while external files and permissions should still be revalidated. It is the continuity path.

F:

A new context boundary avoids accidental disclosure or bias from another customer's history. Authorized facts can be loaded specifically for the new case.

Incorrect Answers

C:

Continuity would carry sensitive and irrelevant context into a new case. Convenience does not justify cross-customer leakage.

D:

A termination hook does not automatically restore external repository state. Session choice and workspace versioning are different concerns.

E:

Branching transcript state does not renew authorization. Permissions are enforced by external policy and expiry checks.

Question 76

A resumed code review references file locations that were renamed after the prior run. Which step should precede edits?

  1. Trust old file paths because the session id matches
  2. Recheck repository state against the saved session context
  3. Create a branch fork solely to make old paths reappear
  4. Treat the rename as proof that all prior findings are invalid

Explanation

Correct Answer

B:

Resumed conversation state may be stale relative to the working tree. The agent must validate paths, diffs and versions before modifying code.

Incorrect Answers

A:

Session identity does not freeze external repository contents. Blind writes can target nonexistent or wrong files.

C:

Forking conversation history does not revert the filesystem. Workspace changes require explicit source-control handling.

D:

Some prior reasoning may remain sound. The correct approach is to verify which dependencies changed.

Question 77

A user asks for a wholly unrelated brainstorming task immediately after a long technical debugging session. Which choice best avoids context contamination?

  1. Fork the debugging session so all old details remain available
  2. Resume the debugging session because its context is already large
  3. Archive the latest tool result as the only prompt for brainstorming
  4. Start a new session for the unrelated task

Explanation

Correct Answer

D:

A fresh session prevents old debugging assumptions from consuming context or biasing the new objective. Relevant stable preferences can be supplied explicitly.

Incorrect Answers

A:

A fork deliberately inherits the history and therefore preserves the contamination risk. It is useful for alternatives on the same baseline.

B:

More historical content is not always helpful. Unrelated context increases noise and can affect the assistant's answers.

C:

A single irrelevant observation cannot establish the new task. The user request should define the fresh context.

Question 78

A session continuation succeeds but the previous run's approval expired while it was paused. How should the agent handle the pending write?

  1. Treat all earlier approvals as permanent because the conversation was resumed
  2. Remove the approval check because the user previously asked for the task
  3. Fork the session to refresh the approval automatically
  4. Revalidate time-sensitive authorization before executing the write

Explanation

Correct Answer

D:

Resumption does not extend an approval's validity. The application must check scope, expiry and external state at the moment of action.

Incorrect Answers

A:

Authorization can be revoked or expire independently of session history. Persisted text is not a live permission check.

B:

A request to work is not necessarily authorization for a sensitive write. The required gate still applies.

C:

Forking duplicates a conversation baseline, not external privileges. It cannot renew authorization.

Question 79

A developer needs to reproduce a failure before experimenting with fixes. Which session arrangement is most useful?

  1. Discard all prior error evidence and start from speculation
  2. Keep a baseline session and fork experimental continuations from it
  3. Apply each experimental fix sequentially inside one irreversible session
  4. Use the same external mutable test database for every branch without resets

Explanation

Correct Answer

B:

A stable checkpoint supports repeatable comparison of hypotheses. Branches can be evaluated against the same initial assumptions and artifacts.

Incorrect Answers

A:

The baseline contains useful observations for reproduction. Losing them prevents meaningful comparison.

C:

Sequential changes contaminate later tests with previous attempts. Isolated branches make attribution clearer.

D:

Even isolated conversation states can interfere through shared external systems. Reproducible trials require controlled test state.

Question 80

A user wants an agent to handle weekly unrelated support cases without leaking old customer data. What is the most suitable session policy?

  1. Rely on a final instruction saying previous cases are irrelevant
  2. Fork every old customer session into a shared weekly master
  3. Create a fresh case-specific session and load only scoped case records
  4. Reuse the previous customer's live session for all cases

Explanation

Correct Answer

C:

New sessions isolate customers and reduce stale context. Long-lived knowledge should reside in authorized systems and be fetched for each case.

Incorrect Answers

A:

Prompts cannot guarantee absence of sensitive information already present in the context. Explicit isolation is preferable.

B:

Forks would carry historical content into a common context. This defeats case-level separation.

D:

That design risks exposing private data and mixing case facts. Session continuity is inappropriate across unrelated customers.

Claude Certified Architect - Foundations (CCAR-F) (CCA-F) Practice Tests