Question 480
A regulated archival platform must require two separate authorized custodians to approve permanent bulk erasure after the requesting customer has confirmed intent. Which choice should be used here?
- Use transaction controls for critical writes
- Audit privileged changes with actor attribution
- Record immutable provenance for sensitive decisions
- Require two independent approvers for regulated bulk deletion
Explanation
Correct Answer
D:A dual-control workflow with separate identities prevents one actor or agent from approving the entire irreversible removal; a second confirmation button by the same person does not meet the two-person policy.
Incorrect Answers
A:Atomic transactions or compensating workflows preserve business consistency across approved changes; here, the controlling capability is that a dual-control workflow with separate identities prevents one actor or agent from approving the entire irreversible removal, but model reasoning does not guarantee database rollback.
B:An audit trail connects the principal, approved operation, target, and result; the missing capability is more specific: a dual-control workflow with separate identities prevents one actor or agent from approving the entire irreversible removal; a text summary by the model is not sufficient evidence of a committed security change.
C:Audit metadata preserves input references, applicable policy, actor, and versioned output; by comparison, the intended solution relies on this: a dual-control workflow with separate identities prevents one actor or agent from approving the entire irreversible removal; a final answer without source history is not adequate for reconstructing the decision.
Question 481
A new agent must search a supplier inventory API whose schema and rate limits are controlled by another team. What is the appropriate next design choice?
- Document data-handling and retention constraints
- Specify audit and observability needs
- Identify external integration contracts
- Identify the upstream source of truth
Explanation
Correct Answer
C:Interface requirements document endpoints, authentication, payloads, limits, and failure handling; the implementation must still recognize that a prose-level prompt is not a substitute for a validated service contract.
Incorrect Answers
A:Privacy requirements identify permitted collection, transmission, storage, retention, and deletion controls; in contrast, interface requirements document endpoints, authentication, payloads, limits, and failure handling, while developer convenience cannot override legal or organizational handling obligations.
B:Trace requirements define correlation IDs, redaction rules, event retention, and source attribution; this addresses a separate concern, not the fact that interface requirements document endpoints, authentication, payloads, limits, and failure handling; a final natural-language answer is insufficient forensic evidence by itself.
D:Architecture requirements name the maintained policy repository and its freshness constraints; the missing capability is more specific: interface requirements document endpoints, authentication, payloads, limits, and failure handling; instructing Claude to remember policy text does not keep that source current.
Question 482
A worker that previously worked now fails authorization before any model output is produced. Which design would meet this acceptance criterion?
- Escalate a persistent authorization denial
- Treat 5xx as potential upstream service faults
- Treat 401/403 responses as credential or access problems
- Treat HTTP 429 as throttling pressure
Explanation
Correct Answer
C:Authentication and entitlement checks distinguish invalid credentials from forbidden resource access—with this limitation: changing prompt wording cannot grant missing API permission.
Incorrect Answers
A:Operators investigate credentials and permissions through the proper access workflow, whereas the requirement calls for another behavior: authentication and entitlement checks distinguish invalid credentials from forbidden resource access; retries or prompt pressure must not be used to bypass access controls.
B:The client applies guarded retries and monitoring while preserving safe degraded behavior; however, the critical mechanism here is that authentication and entitlement checks distinguish invalid credentials from forbidden resource access, and treating an outage as a prompt defect misdirects investigation.
D:The caller reduces concurrency and uses bounded backoff while considering available quota; this cannot substitute for the behavior that authentication and entitlement checks distinguish invalid credentials from forbidden resource access, because blind parallel retries amplify rather than relieve throttle pressure.
Question 483
A parent agent needs the result of a lengthy research subtask without all intermediate searches. What is the appropriate next design choice?
- Use retrieval from a source repository
- Detect and resolve conflicting worker findings
- Use a compact task summary for handoff
- Use ephemeral scratch context for one run
Explanation
Correct Answer
C:A narrow summary communicates findings, evidence, and unresolved questions within a smaller token budget; the implementation must still recognize that it must not silently claim unverified findings are authoritative.
Incorrect Answers
A:Fetching current documentation from a maintained source avoids depending on recollected assertions; in contrast, a narrow summary communicates findings, evidence, and unresolved questions within a smaller token budget, while retrieved snippets still need source and version checks before action.
B:The coordinator compares sources, dates, and evidence before emitting a final claim; this addresses a separate concern, not the fact that a narrow summary communicates findings, evidence, and unresolved questions within a smaller token budget; voting by worker count is not a substitute for checking authoritative documentation.
D:Transient context supports local reasoning without creating long-term customer state; the missing capability is more specific: a narrow summary communicates findings, evidence, and unresolved questions within a smaller token budget; it should not be mistaken for durable memory across new sessions.
Question 484
After a risky assistant release, unexpected outbound calls increase. The same system serves harmful-content prompts and includes a records lookup that must not write. Which THREE controls address monitoring, response policy, and infrastructure privilege? Choose THREE.
- Monitor and investigate unusual tool activity
- Enforce least-privilege infrastructure permissions
- Keep a human review gate for high-impact operations
- Perform a least-privilege permissions review
- Use a policy-based refusal and safe redirect
- Sandbox uncertain code execution
Explanation
Correct Answers
A:Telemetry and alerts identify suspicious action patterns and allow prompt, tool, or policy remediation; a crucial qualification is that monitoring should complement, not replace, preventive execution controls.
B:Backend credentials and IAM policies restrict it to the minimum necessary operations; in production, an unrestricted credential remains dangerous regardless of the model's stated intent.
E:A policy layer handles the request appropriately and communicates a safe boundary; the important constraint is that policy adherence should be evaluated; it is not guaranteed by a single warning prompt.
Incorrect Answers
C:Mandatory human approval separates recommendations from irreversible execution; the needed capability is different—telemetry and alerts identify suspicious action patterns and allow prompt, tool, or policy remediation—and model confidence scores cannot replace the required authority of a reviewer.
D:Reviewing actual credentials and granted scopes shrinks the impact of a compromised prompt or tool, whereas the requirement calls for another behavior: telemetry and alerts identify suspicious action patterns and allow prompt, tool, or policy remediation; the prompt's intended task does not limit the external credential automatically.
F:Runtime isolation limits privileges, file access, and network effects of untrusted programs; the requirement instead turns on this behavior: telemetry and alerts identify suspicious action patterns and allow prompt, tool, or policy remediation; static text review cannot guarantee the code will behave safely.
Question 485
A customer requests account-specific shipment tracking through a privileged backend. Which solution should the engineer adopt?
- Validate the caller's identity before exposing tools
- Apply authorization before retrieval
- Authorize each requested resource operation
- Validate generated tool parameters
Explanation
Correct Answer
A:Authentication establishes which principal is making the request before access decisions; the implementation must still recognize that a conversational claim of ownership is not reliable proof of identity.
Incorrect Answers
B:The retrieval layer filters sources by verified caller access before any text reaches Claude; the missing capability is more specific: authentication establishes which principal is making the request before access decisions; filtering only the final answer cannot undo earlier disclosure to the model.
C:Resource-level policy checks verify that the caller can execute that exact action on that object; by comparison, the intended solution relies on this: authentication establishes which principal is making the request before access decisions; authentication alone does not confer access to all records.
D:Server-side validation constrains types, ranges, ownership, and supported operations; the needed capability is different—authentication establishes which principal is making the request before access decisions—and a well-described function schema does not replace authorization checks.
Question 486
The API sometimes returns a temporary gateway failure and sometimes rejects malformed requests. Which engineering practice is the strongest match?
- Use HTTPS for remote model requests
- Send a typed fallback error to clients
- Use retry-aware HTTP error handling
- Use connection reuse for frequent calls
Explanation
Correct Answer
C:Status classification drives different responses for transient and permanent errors; the implementation must still recognize that blindly retrying invalid JSON does not improve correctness.
Incorrect Answers
A:TLS protects data in transit against passive interception and many tampering threats; in contrast, status classification drives different responses for transient and permanent errors, while encryption does not substitute for caller authentication or payload minimization.
B:A well-defined error object lets clients distinguish outage, rate limit, and invalid input; this addresses a separate concern, not the fact that status classification drives different responses for transient and permanent errors; returning fabricated success data makes incident diagnosis harder.
D:Persistent connections reduce repeated connection setup overhead; the missing capability is more specific: status classification drives different responses for transient and permanent errors; pool sizing still has to respect concurrency and upstream rate limits.
Question 487
A response can contain text and tool-use blocks in one ordered content array. Which mechanism should the team prioritize?
- Parse typed content blocks rather than raw concatenation
- Inspect stop_reason before returning output
- Pass an image as a typed image content block
- Read usage counts from the response
Explanation
Correct Answer
A:Typed block handling preserves whether an element is displayable text or a tool request; the important constraint is that blindly concatenating heterogeneous blocks can lose tool metadata and semantics.
Incorrect Answers
B:The completion reason tells the client whether to deliver, dispatch, or handle truncation; the requirement instead turns on this behavior: typed block handling preserves whether an element is displayable text or a tool request; treating all returned content as a complete final answer can hide partial results.
C:Image blocks identify the media source while text instructions explain the intended analysis; this cannot substitute for the behavior that typed block handling preserves whether an element is displayable text or a tool request, because putting a local path in ordinary text does not upload the image bytes to the API.
D:Response usage fields provide the token measures used for billing and diagnostics; in contrast, typed block handling preserves whether an element is displayable text or a tool request, while character count is not an exact substitute for model tokenization.
Question 488
A generated policy summary should be iteratively checked against a quality rubric and revised until it passes. Which response is most defensible under these conditions?
- Apply orchestrator-worker decomposition
- Keep a single direct model call
- Use a model-directed agent loop
- Use an evaluator-optimizer loop
Explanation
Correct Answer
D:A grader produces targeted feedback that the writer consumes on the next bounded iteration; a crucial qualification is that without a maximum attempts policy the system can burn tokens without improving the draft.
Incorrect Answers
A:A coordinator decomposes the task and merges workers' results after their bounded subtasks complete; in contrast, a grader produces targeted feedback that the writer consumes on the next bounded iteration, while workers need nonoverlapping responsibilities and explicit result contracts to avoid duplication.
B:A direct request minimizes latency and operational state for a bounded answer; this addresses a separate concern, not the fact that a grader produces targeted feedback that the writer consumes on the next bounded iteration; an agent loop would add failure modes without yielding a decision the model must make.
C:Claude selects an action after observing each result, with the application enforcing iteration and permission limits; here, the controlling capability is that a grader produces targeted feedback that the writer consumes on the next bounded iteration, but the autonomy adds variability and demands hard budgets, logs, and a stopping policy.
Question 489
Multiple client sessions need a standard checklist to guide security code reviews. What would best address the stated failure mode?
- Expose an MCP resource for read-oriented context
- Expose an MCP prompt for a reusable template
- Expose a custom API adapter outside MCP when local-only
- Use stdio for a local subprocess integration
Explanation
Correct Answer
B:Prompt templates provide reusable structured instructions rather than executing business operations; a crucial qualification is that templates do not automatically grant access to any tool or private resource.
Incorrect Answers
A:Resources represent accessible data that a client can retrieve and supply as context; here, the controlling capability is that prompt templates provide reusable structured instructions rather than executing business operations, but a resource's text cannot authorize additional tool actions by itself.
C:A direct adapter may be simpler than running and maintaining a separate protocol server; by comparison, the intended solution relies on this: prompt templates provide reusable structured instructions rather than executing business operations; choosing MCP solely for naming consistency introduces unnecessary lifecycle work.
D:Standard input and output suit tightly coupled local client-server interactions; the needed capability is different—prompt templates provide reusable structured instructions rather than executing business operations—and the server's standard output must remain protocol-clean instead of mixing human debug prints.
Question 490
A document parser should inspect confidential attachments without leaking their raw contents into the main coordinator's thread. How should this requirement be met?
- Load task-specific memory at the start
- Start a fresh isolated context for a subtask
- Run independent read-only tool calls concurrently
- Summarize old steps before continuing
Explanation
Correct Answer
B:A separate worker receives only the necessary files and reports a minimal result; the implementation must still recognize that isolation only works when tool rights and returned text follow the same boundary.
Incorrect Answers
A:Relevant stored facts are retrieved and inserted into the allowed context for the current task; this addresses a separate concern, not the fact that a separate worker receives only the necessary files and reports a minimal result; the context selection must be scoped to the authenticated customer.
C:Concurrent dispatch shortens latency when tool requests are independent and safely isolated; the missing capability is more specific: a separate worker receives only the necessary files and reports a minimal result; shared write operations require separate coordination and consistency protections.
D:A compact summary can replace bulky intermediate transcripts for the next model call; by comparison, the intended solution relies on this: a separate worker receives only the necessary files and reports a minimal result; summaries can lose details unless critical facts and provenance are retained.