Question 230
An unsafe call must be blocked even if Claude confidently explains why it is acceptable. What safety architecture meets that requirement?
- Introduce a PreToolUse hook that denies matching unsafe invocations
- Enforce the boundary in executable hook or application policy code
- Limit the hook matcher to the intended tool name or invocation pattern
- Examine result payload semantics in a post-tool validation hook
Explanation
Correct Answer
B:Model reasoning can propose an operation but should not override deterministic authorization. Executable policy separates permissions from probabilistic language generation.
Incorrect Answers
A:PreToolUse runs before a tool call and can participate in the decision to block it; its benefits for blocking disallowed commands before tool execution do not resolve the need for keeping permission enforcement independent of model judgment.
C:Matchers determine which hook executions apply; this contributes to targeting a hook to relevant operations without meeting the need for keeping permission enforcement independent of model judgment.
D:Transport success does not necessarily mean business success; that serves detecting business failure despite transport success, whereas the case requires keeping permission enforcement independent of model judgment.
Question 231
A task runner resumes after a crash with the export step marked complete. Which behavior avoids duplicate exports?
- Add a concise case summary, relevant evidence and the escalation reason
- Sequence or lock the conflicting state-changing section
- Formalize a post-action reconciliation stage before marking the case resolved
- Persist workflow checkpoints and skip already-confirmed idempotent stages
- Introduce a human approval gate between proposal and execution
Explanation
Correct Answer
D:Stage-level state permits resumable orchestration. Replaying every step after a restart can reissue notifications or modify external records twice.
Incorrect Answers
A:For transferring sufficient escalation context, a meaningful handoff preserves the context the reviewer needs to act, avoiding repeated questioning and incorrect policy decisions; however, this requirement concerns restarting from a durable workflow checkpoint.
B:Concurrency control protects shared mutable state; that may serve preventing conflicting concurrent mutations, but it is no substitute for restarting from a durable workflow checkpoint.
C:Workflow success should depend on confirmed downstream state, not just a tool invocation; this may support verifying all required stages of case resolution but cannot alone meet the requirement for restarting from a durable workflow checkpoint.
E:Human-in-the-loop approval must be enforced by the controller for side effects; a method suited to inserting human approval before side effects does not meet the need for restarting from a durable workflow checkpoint.
Question 232
A project-wide instruction says to run a slow integration suite on every tiny documentation edit. What improvement reduces needless runs?
- Spell out the repository's actual build, lint and test commands with prerequisites
- Rewrite guidance to distinguish tests required by change type and risk
- Retain mandatory security policy in enforced controls rather than editable personal preferences
- Verify the intended CLAUDE.md is discovered and loaded for that working context
Explanation
Correct Answer
B:Persistent rules should guide decisions rather than mandate wasteful work universally. Task-sensitive criteria preserve thoroughness where code behavior changes.
Incorrect Answers
A:Executable command knowledge is stable operational context; the option would serve recording project-specific validation commands, not making standing test instructions proportional as specified in the stem.
C:Instruction hierarchy should not be the sole defense for high-risk operations; that relates to separating organization policy from user preferences, not the necessary making standing test instructions proportional.
D:Although rules are useful only when the assistant encounters the relevant file and scope, the choice addresses verifying instruction-file loading and not the requested making standing test instructions proportional.
Question 233
An agent cannot reproduce a reported production problem and starts modifying unrelated modules. What is a better response?
- Split reproducible regressions from known nondeterministic failures
- Employ failing integration output to refine the patch and rerun relevant tests
- Gather reproduction inputs and compare observed environment differences first
- Examine the actual dependency version and local interface before rewriting calls
Explanation
Correct Answer
C:Without a reproducible failure or causal evidence, broad edits are speculative. Environment-aware diagnosis narrows the problem before mutation.
Incorrect Answers
A:A single intermittent result is weak evidence; interpreting flaky test signals correctly and establishing a reliable bug reproduction require different controls in this situation.
B:Iterative development uses tool feedback as evidence; the unsatisfied requirement is establishing a reliable bug reproduction, not refining code based on integration-test feedback.
D:When considering aligning generated code to installed dependency APIs, version drift can make a plausible call invalid; the requirement for establishing a reliable bug reproduction remains unmet.
Question 234
An automated test-generation job writes snapshots directly to the protected branch. What workflow correction is needed?
- Launch Claude Code non-interactively with claude -p in CI
- Narrow allowed tools and repository credentials to read-only review operations
- Produce proposed changes in an isolated workspace or pull request for review
- Employ least-privilege credentials and sandbox the untrusted review workload
- Solicit machine-readable JSON output and validate its schema
Explanation
Correct Answer
C:Generated tests may be incorrect or brittle. Isolating changes preserves branch protections and allows maintainers to inspect before merging.
Incorrect Answers
A:The print-style command supports scripting without a chat session; that addresses invoking Claude Code in a noninteractive pipeline, whereas this case calls for isolating generated CI code changes.
B:CI runners often have elevated tokens; this concerns enforcing read-only CI review permissions, leaving the requirement for isolating generated CI code changes outstanding.
D:PR diffs can contain prompt injection or malicious scripts; the action supports limiting trust for untrusted pull-request code but falls short of isolating generated CI code changes in this case.
E:Structured output is easier for CI to parse and gate deterministically; this can serve parsing review output safely in CI, but leaves the requirement for isolating generated CI code changes unmet.
Question 235
A team uses one giant CLAUDE.md with rules for dozens of unrelated packages. What organization makes project guidance leaner?
- Store globally applicable policy in an always-loaded or enforced layer
- Move conditional package conventions into dedicated rules files with path selectors
- Correct the rule's paths glob to target only migration files
- Shift writing-style guidance into a documentation-path rule
Explanation
Correct Answer
B:Splitting the corpus by applicability reduces irrelevant context. Shared baseline principles can stay in CLAUDE.md while specialized rules live beside their scope definitions.
Incorrect Answers
A:A conditional file rule may not activate in every session; this contributes to keeping organization-wide policy outside conditional globs without meeting the need for avoiding an oversized global project rule set.
C:Because a narrower glob controls the rule's applicability, the measure supports narrowing a conditional rule's file scope rather than the needed avoiding an oversized global project rule set.
D:Documentation conventions are meaningful when documents are the target; its benefits for limiting editorial guidance to documentation edits do not resolve the need for avoiding an oversized global project rule set.
Question 236
A model produces well-formed invoice JSON but includes a nonexistent line item. What test catches this?
- Merge repeated records from extracted records using source identifiers and document context
- Cross-check extracted items against their source spans or document evidence
- Check only that referenced product identifiers exist in an inventory database, without inspecting the invoiced source document
- Implement domain-specific calculation and policy rules after schema parsing
Explanation
Correct Answer
B:Per-field provenance validation detects additions not supported by the original content. Structural checks alone will accept a fabricated yet correctly typed item.
Incorrect Answers
A:Overlapping page captures can lead to double counting; this may support removing duplicate extraction artifacts but cannot alone meet the requirement for matching extracted entities to source evidence.
C:A real product can still be hallucinated as a line item on a particular invoice. Document-level provenance is needed to verify each extracted entry actually appears in the source.
D:For validating business invariants after schema checks, schema guarantees shape, not compliance with business calculations; however, this requirement concerns matching extracted entities to source evidence.
Question 237
A customer-facing chatbot needs an answer in seconds during a live support call. Why should it generally avoid batch processing?
- Exclude streaming from batch requests and process completed result artifacts
- Use a synchronous interactive request for the latency-sensitive turn
- Give a unique custom_id to every request and match returned results by that identifier
- Carry forward each request's custom_id and associated error classification
Explanation
Correct Answer
B:A batch job runs asynchronously and completion is not immediate. The user-facing session needs direct response latency rather than deferred bulk throughput.
Incorrect Answers
A:Batch execution yields collected request outcomes rather than an interactive token stream; that relates to using supported response delivery for batch processing, not the necessary keeping interactive requests out of deferred batches.
C:Batch completion order is not guaranteed; the deciding condition is keeping interactive requests out of deferred batches, whereas this action concerns correlating out-of-order batch results.
D:Stable item identifiers connect failures to inputs and retry history; the option would serve linking each batch error to its source record, not keeping interactive requests out of deferred batches as specified in the stem.
Question 238
A data retrieval tool returns a structured error instead of a shipment record. How should the model loop receive it?
- Resume the paused assistant turn with its response content preserved
- Hold the tool implementation in the authorized execution environment
- Enable automatic tool selection based on the available tool definitions
- Return an explicit failed tool result with diagnostic information rather than fabricate data
Explanation
Correct Answer
D:The tool-error signal should flow into the conversation so the agent can retry, escalate, or communicate limits. Pretending the result is a successful record masks operational failure.
Incorrect Answers
A:A pause_turn can mark a server-tool iteration boundary; in this scenario, continuing a paused server-tool loop is secondary to surfacing client-tool execution failure.
B:When considering distinguishing model intent from tool execution, client tools are interfaces advertised to the model; the hosting application runs their operations; the requirement for surfacing client-tool execution failure remains unmet.
C:When addressing choosing optional rather than mandatory tool use, automatic selection lets Claude choose between answering and requesting a tool when appropriate; the present task instead calls for surfacing client-tool execution failure.
Question 239
A customer lookup tool sometimes receives email where its code expects account_id. What should be improved?
- Define typed required parameters with separate identifier semantics
- Choose an unambiguous timestamp representation with timezone requirements
- Tell apart the read-only lookup from the authorized state-changing operation
- Divide the overloaded interface into cohesive operations with narrow schemas
Explanation
Correct Answer
A:A precise input schema can distinguish account identifiers from email addresses and reject unsupported shapes. A free-form identifier field invites misbound customer records.
Incorrect Answers
B:Machine-oriented time fields and explicit timezone semantics prevent interpretation drift; the action supports normalizing temporal tool parameters but falls short of distinguishing identity fields in a tool interface in this case.
C:Tool metadata should expose whether a call merely inspects or mutates an order; this can serve clarifying read versus write tool intent, but leaves the requirement for distinguishing identity fields in a tool interface unmet.
D:Smaller purpose-built tools make argument requirements and authorization easier to reason about; that addresses reducing an overloaded tool schema, whereas this case calls for distinguishing identity fields in a tool interface.
Question 240
A long-lived support session stores all retrieved customer records forever. Which retention practice reduces risk?
- Retain a session identifier and resume its authorized conversation state
- Cross-check the fork against the current codebase and rerun relevant checks
- Persist only necessary state and apply an explicit retention or deletion policy
- Verify the user's session privileges again before protected work continues
Explanation
Correct Answer
C:Resumability does not justify indefinite storage of sensitive context. State minimization and lifecycle controls balance continuity with privacy obligations.
Incorrect Answers
A:Because session continuity reconnects the agent to relevant history and progress, the measure supports resuming a saved agent session rather than the needed limiting persisted session data.
B:A successful result on an older branch does not prove current compatibility; its benefits for validating branch changes before merge do not resolve the need for limiting persisted session data.
D:A stored conversation is not a perpetual permission grant; this contributes to revalidating authorization on session resume without meeting the need for limiting persisted session data.