Question 120
A Git pull request changes the payment tool's schema. What review artifact most directly reduces production risk?
- Employ a bounded worker queue with backpressure and rate-limit awareness.
- Interpret and validate fields, then use parameterized database queries.
- Fix model, prompt, tool schema, and application configuration under change control.
- Require schema diffs and contract tests for the affected tool boundary.
Explanation
Correct Answer
D:Reviewers need to see both the field changes and their behavioral impact on model-proposed arguments and downstream execution.
Incorrect Answers
A:Concurrency improves throughput only until it overwhelms quotas or dependencies; queueing provides controlled resource use; the scenario's controlling requirement is reviewable tool schema changes, not the related scalable request queueing concern.
B:The output is untrusted data even if generated by Claude; parameter binding prevents injection through constructed SQL syntax; when the goal is reviewable tool schema changes, optimizing safe persistence of generated data misses that operational boundary.
C:Reproducibility depends on all behavior-shaping inputs, not only the application binary; it follows reproducible inference configuration constraints but leaves reviewable tool schema changes unspecified.
Question 121
A codebase has the same retry and authentication logic in six Claude call sites. Which refactoring improves reliability?
- Employ asynchronous I/O and enforce concurrency limits on external requests.
- Maintain Claude calls in a server-side service and expose a restricted application API.
- Repeat transient failures with backoff, jitter, and an overall attempt limit.
- Consolidate shared request handling into a tested client adapter.
Explanation
Correct Answer
D:One adapter standardizes error mapping, headers, and retry policy, reducing drift across different application paths.
Incorrect Answers
A:Network calls can overlap without tying up a worker for each wait; bounded parallelism avoids saturating dependencies; it follows nonblocking service I/O constraints but leaves shared Claude transport abstraction unspecified.
B:A backend boundary centralizes secrets, rate limits, authorization, and prompt versions rather than shipping privileged credentials to browsers; changing server-side credential boundary does not establish shared Claude transport abstraction in this application.
C:Temporary service errors often recover; spreading and bounding retries reduces synchronized bursts and runaway latency; where shared Claude transport abstraction matters, a control for transient service retry strategy leaves that test unmet.
Question 122
A development team uses model output as a typed object without parsing failure handling. What should be added?
- Implement defensive parsing with explicit schema-error paths and fallback behavior.
- Decode and validate fields, then use parameterized database queries.
- Validate destinations against an explicit allowlist and network policy.
- Adopt a bounded worker queue with backpressure and rate-limit awareness.
Explanation
Correct Answer
A:Generation and network success do not remove the need to handle malformed or semantically invalid data safely.
Incorrect Answers
B:The output is untrusted data even if generated by Claude; parameter binding prevents injection through constructed SQL syntax; for this request, robust typed output parsing is the missing piece, not safe persistence of generated data.
C:Unchecked endpoints create server-side request forgery risk; the application's network layer must limit reachable addresses; changing safe outbound URL handling does not establish robust typed output parsing in this application.
D:Concurrency improves throughput only until it overwhelms quotas or dependencies; queueing provides controlled resource use; the consequence is improved scalable request queueing, not verified robust typed output parsing.
Question 123
A client app wants to process a stream containing text deltas and terminal events. Which engineering design fits?
- Obtain authoritative data at execution time and supply it as referenced context.
- Collect status, correlation ID, latency, and safe error classification without secrets.
- Store prompts as versioned artifacts linked to tests and releases.
- Build an event-driven stream parser with explicit completion and error states.
Explanation
Correct Answer
D:Streaming protocols deliver more than plain characters; stateful parsing ensures consumers know when content is complete.
Incorrect Answers
A:The model's learned information is not a live database; the application must retrieve current records when freshness matters; for this request, event-stream parser state is the missing piece, not current external data acquisition.
B:Structured metadata supports incident triage and vendor support while avoiding accidental logging of sensitive prompts or keys; a remedy for safe API error telemetry is not a substitute for event-stream parser state.
C:Traceable revisions allow comparing model behavior and restoring a known-good prompt independently of code edits; it improves multi-criterion integration only, leaving event-stream parser state unproven.
Question 124
A test suite compares Claude's free-form response strings byte for byte and fails on paraphrases. What should it evaluate instead?
- Assert task-level invariants and structured fields rather than exact prose.
- Apply domain invariants such as sum reconciliation and nonnegative quantities.
- Inspect status and content type before attempting JSON decoding.
- Verify destinations against an explicit allowlist and network policy.
Explanation
Correct Answer
A:Non-deterministic language variation makes string snapshots brittle, while behavior and schema assertions capture meaningful correctness.
Incorrect Answers
B:A schema can guarantee a field is numeric without guaranteeing that extracted values are financially consistent; where model-response test invariants matters, a control for semantic business validation leaves that test unmet.
C:Assumed JSON parsing can obscure the real transport failure; protocol metadata must be validated first; it changes defensive HTTP content parsing; it does not prove model-response test invariants in this case.
D:Unchecked endpoints create server-side request forgery risk; the application's network layer must limit reachable addresses; this targets safe outbound URL handling, while the system must demonstrate model-response test invariants.
Question 125
A workflow relies on time-dependent external information. What integration pattern limits stale results?
- Insist on schema diffs and contract tests for the affected tool boundary.
- Fetch authoritative data at execution time and supply it as referenced context.
- Link internal failures into safe, stable, actionable client error categories.
- Record emitted content and reconnect semantics to prevent replaying the same deltas.
Explanation
Correct Answer
B:The model's learned information is not a live database; the application must retrieve current records when freshness matters.
Incorrect Answers
A:Reviewers need to see both the field changes and their behavioral impact on model-proposed arguments and downstream execution; a remedy for reviewable tool schema changes is not a substitute for current external data acquisition.
C:Exposing internals leaks implementation details and produces brittle clients; a boundary can preserve detailed logs while presenting safe errors; for current external data acquisition, this instead manages safe error abstraction; the requested decision remains unresolved.
D:Stream resumption and client rendering require explicit ordering and deduplication assumptions; reconnecting is not automatically exactly once; it changes streaming UI deduplication; it does not prove current external data acquisition in this case.
Question 126
A middleware service sees an error with a structured HTTP status and request ID. What should the observability layer record?
- At the application layer, introduce a compatible default or version the contract with a migration path.
- Capture status, correlation ID, latency, and safe error classification without secrets.
- Adopt asynchronous I/O and enforce concurrency limits on external requests.
- Construct an event-driven stream parser with explicit completion and error states.
Explanation
Correct Answer
B:Structured metadata supports incident triage and vendor support while avoiding accidental logging of sensitive prompts or keys.
Incorrect Answers
A:Adding a newly required field can break existing callers; backward compatibility must be managed explicitly; the consequence is improved schema compatibility evolution, not verified safe API error telemetry.
C:Network calls can overlap without tying up a worker for each wait; bounded parallelism avoids saturating dependencies; it improves nonblocking service I/O only, leaving safe API error telemetry unproven.
D:Streaming protocols deliver more than plain characters; stateful parsing ensures consumers know when content is complete; it changes event-stream parser state; it does not prove safe API error telemetry in this case.
Question 127
A service stores every user message in a global list shared by all active sessions. What bug can result?
- A global message list is automatically partitioned by the model into authenticated sessions.
- Cross-user conversation leakage; key state by authenticated session or tenant.
- The principal risk is that messages can no longer be ordered chronologically.
- Global history is harmless because each request contains a separate model identifier.
- Only the latest user message can be visible, regardless of server session storage.
Explanation
Correct Answer
B:Conversation context is caller-owned, so shared mutable history can mix unrelated users and violate data isolation.
Incorrect Answers
A:A shared application data structure provides no tenant isolation unless the developer explicitly partitions access.
C:A globally ordered list can still be sorted by time; the critical issue is that unrelated users may see one another’s text.
D:Model selection does not authenticate conversation ownership or restrict which stored messages are returned. The server must isolate history before it is sent to the model.
E:If an application adds shared history to a prompt, the model can receive those records irrespective of their original owner.
Question 128
A deployer wants the same application behavior in testing and production. What should be versioned together?
- Save prompts as versioned artifacts linked to tests and releases.
- Pin model, prompt, tool schema, and application configuration under change control.
- Build defensive parsing with explicit schema-error paths and fallback behavior.
- Centralize shared request handling into a tested client adapter.
Explanation
Correct Answer
B:Reproducibility depends on all behavior-shaping inputs, not only the application binary. Keep a record of deployment-time overrides and feature flags, because they can change behavior even when the source commit matches.
Incorrect Answers
A:Traceable revisions allow comparing model behavior and restoring a known-good prompt independently of code edits; when the goal is reproducible inference configuration, optimizing multi-criterion integration misses that operational boundary.
C:Generation and network success do not remove the need to handle malformed or semantically invalid data safely; changing robust typed output parsing does not establish reproducible inference configuration in this application.
D:One adapter standardizes error mapping, headers, and retry policy, reducing drift across different application paths; a remedy for shared Claude transport abstraction is not a substitute for reproducible inference configuration.
Question 129
A test API endpoint sometimes yields a non-JSON error page from an upstream proxy. What should the caller do?
- Implement an event-driven stream parser with explicit completion and error states.
- For this workflow, assert task-level invariants and structured fields rather than exact prose.
- Check status and content type before attempting JSON decoding.
- Retain Claude calls in a server-side service and expose a restricted application API.
Explanation
Correct Answer
C:Assumed JSON parsing can obscure the real transport failure; protocol metadata must be validated first. Log a safe correlation identifier and the upstream failure class without attempting to interpret an HTML error page as a model response.
Incorrect Answers
A:Streaming protocols deliver more than plain characters; stateful parsing ensures consumers know when content is complete; for defensive HTTP content parsing, this instead manages event-stream parser state; the requested decision remains unresolved.
B:Non-deterministic language variation makes string snapshots brittle, while behavior and schema assertions capture meaningful correctness; it changes model-response test invariants; it does not prove defensive HTTP content parsing in this case.
D:A backend boundary centralizes secrets, rate limits, authorization, and prompt versions rather than shipping privileged credentials to browsers; the effect concerns server-side credential boundary; the success condition is defensive HTTP content parsing.
Question 130
A web app directly trusts a user-supplied role string such as 'admin' before calling tools. What is the security flaw?
- A role string typed by the user grants database permissions when included in a prompt.
- TLS encryption establishes that the user holds the administrator role.
- Authorization must derive from verified server-side identity, not client-controlled fields.
- The model can authorize sensitive tools if it explains the action clearly.
Explanation
Correct Answer
C:Untrusted request payloads can be altered; tools require authoritative permission checks independently of what the model or client claims.
Incorrect Answers
A:Self-asserted text is not a verified identity claim and cannot justify privileged downstream data access. The downstream tool must make the role decision itself.
B:TLS secures the connection, but role-based authorization still requires identity verification and server policy.
D:A plausible narrative does not replace backend identity, role, target, and action checks. An audit log of prose does not grant any of those permissions.