Question 150
A desktop client and a REST backend interpret the same prompt differently. What should the team avoid assuming?
- Cross-check the operation using an idempotent business identifier rather than replaying the update.
- Retain user-provided content in a separate user or quoted data field.
- Employ deterministic code to compute totals and let Claude explain the results.
- Do not assume all Claude interfaces share identical tools, system instructions, and session context.
Explanation
Correct Answer
D:Product surfaces have distinct integrations and permission models, so behavior must be assessed in the actual deployment interface.
Incorrect Answers
A:UI disconnects do not roll back external effects; checking committed state prevents a duplicate shipment; changing reconnect-safe action state does not establish interface-specific Claude behavior in this application.
B:Combining untrusted input with the highest-priority prompt surface risks instruction confusion and makes auditing difficult; it improves separation of user data and policy only, leaving interface-specific Claude behavior unproven.
C:Numerical aggregation is reliably handled by code; the model can interpret verified figures without inventing calculations; it changes separating calculation from explanation; it does not prove interface-specific Claude behavior in this case.
Question 151
A user submits a spreadsheet and asks the model to calculate a large total. What application design reduces arithmetic error?
- Use deterministic code to compute totals and let Claude explain the results.
- Enforce tenant authorization filters before returning retrieval results to the model.
- Insist on tool execution to satisfy trusted user intent and policy, not retrieved text.
- Remove or qualify the unsupported claim and offer source retrieval or escalation.
Explanation
Correct Answer
A:Numerical aggregation is reliably handled by code; the model can interpret verified figures without inventing calculations.
Incorrect Answers
B:Prompt promises cannot secure a shared index; access control must govern which documents enter the context; the consequence is improved retrieval-layer tenant filtering, not verified separating calculation from explanation.
C:External content is lower trust than the task owner; backend guards should reject out-of-scope requests even when the model proposes them; the scenario's controlling requirement is separating calculation from explanation, not the related tool authority based on trusted intent concern.
D:Generation quality alone cannot establish evidence; the application must enforce the required attribution or abstention behavior; the effect concerns evidence-required response policy; the success condition is separating calculation from explanation.
Question 152
A workflow must write a payment after a model recommends it. Where should the final validation run?
- Adopt deterministic code to compute totals and let Claude explain the results.
- Revalidate authorization before loading history or calling privileged tools.
- Insist on reviewed permissions and scoped access for each connector before activation.
- Check business rules and user authority in the payment service before executing the write.
- Associate user requests to validated canonical categories before querying metrics.
Explanation
Correct Answer
D:The model's recommendation is an input, not a permission grant; the execution boundary must independently reject unauthorized actions.
Incorrect Answers
A:Numerical aggregation is reliably handled by code; the model can interpret verified figures without inventing calculations; this may help separating calculation from explanation but cannot by itself deliver model-to-payment boundary.
B:Long-lived chat IDs do not preserve access rights; current identity and permissions must gate each request; when the goal is model-to-payment boundary, optimizing per-turn authorization misses that operational boundary.
C:Plugins may introduce new external data and actions; install-time authorization prevents implicit privilege expansion; where model-to-payment boundary matters, a control for plugin permission review leaves that test unmet.
E:Normalization separates flexible language understanding from deterministic database and UI contracts; a remedy for canonicalization of intent labels is not a substitute for model-to-payment boundary.
Question 153
A document assistant needs to cite a paragraph from a specific policy revision. What should retrieval return?
- Include stable document IDs, versions, and location metadata with retrieved excerpts.
- Enforce defined retention limits, consent and access controls to conversation records.
- Escape or sanitize untrusted output according to its display context.
- Isolate draft generation from an explicit send action controlled by the user.
Explanation
Correct Answer
A:Provenance makes citations and later audits traceable to the exact revision rather than an undifferentiated text blob.
Incorrect Answers
B:More stored context is not inherently better; privacy obligations and data minimization require purposeful retention; for this request, versioned retrieval provenance is the missing piece, not conversation retention governance.
C:Model and tool content can contain scripts or markup; safe rendering is an application responsibility independent of the prompt; the effect concerns multi-criterion integration; the success condition is versioned retrieval provenance.
D:Text generation and delivery are distinct capabilities; the UI and backend should keep drafts inert until authorized; this targets draft-send separation, while the system must demonstrate versioned retrieval provenance.
Question 154
A multi-tenant app shares a retrieval index and must not expose one tenant's private documents to another. Which check is essential?
- Block or qualify the unsupported claim and offer source retrieval or escalation.
- Present the proposed operation and its target before asking for authorization.
- Remove or filter sensitive content before indexing and retrieving it.
- Apply tenant authorization filters before returning retrieval results to the model.
Explanation
Correct Answer
D:Prompt promises cannot secure a shared index; access control must govern which documents enter the context. Attach authorized tenant predicates to retrieval queries before any matching chunk is placed into the model context window.
Incorrect Answers
A:Generation quality alone cannot establish evidence; the application must enforce the required attribution or abstention behavior; where retrieval-layer tenant filtering matters, a control for evidence-required response policy leaves that test unmet.
B:A meaningful approval gate needs specific action details; a generic 'continue' button cannot establish informed consent; it follows actionable human approval UI constraints but leaves retrieval-layer tenant filtering unspecified.
C:Minimizing data at the retrieval layer reduces the material exposed in prompts and accidental output leakage; changing minimal knowledge indexing does not establish retrieval-layer tenant filtering in this application.
Question 155
A support assistant repeatedly asks users for information already in verified account data. What context design is preferable?
- Supply only the necessary authorized account fields for each request.
- Establish a constrained status enum and validate it at the application boundary.
- Look up account IDs through trusted lookup and validate any generated reference against it.
- Filter relevant recent turns and a summary of durable earlier constraints.
Explanation
Correct Answer
A:Reliable context can be loaded from authoritative systems while minimizing sensitive disclosure and token bloat.
Incorrect Answers
B:An explicit schema makes downstream branching predictable; free-form labels require fragile string interpretation; this implements finite machine-readable statuses rather than the demanded minimal authoritative account context.
C:Language models can produce syntactically plausible but nonexistent IDs; authoritative resolution must occur outside generation; this may help verified identifier resolution but cannot by itself deliver minimal authoritative account context.
D:Explicit context assembly controls costs and stale instructions; the API does not autonomously decide which history is authoritative; changing selective conversation assembly does not establish minimal authoritative account context in this application.
Question 156
A chat frontend disconnects after the API has already executed a shipping update. How should reconnection behave?
- Record minimal pseudonymous operational metadata and keep content in approved secure storage.
- Inspect business rules and user authority in the payment service before executing the write.
- Reconcile the operation using an idempotent business identifier rather than replaying the update.
- Tag retrieval as untrusted task data and block tool authority from crossing that boundary.
Explanation
Correct Answer
C:UI disconnects do not roll back external effects; checking committed state prevents a duplicate shipment. A committed shipping update needs a durable operation ID, so retry logic can query actual shipment state before another mutation.
Incorrect Answers
A:Diagnostics and product analytics have different data needs; leaking raw conversation material into telemetry broadens exposure; even if privacy-preserving analytics succeeds, reconnect-safe action state still needs separate verification.
B:The model's recommendation is an input, not a permission grant; the execution boundary must independently reject unauthorized actions; for reconnect-safe action state, this instead manages model-to-payment boundary; the requested decision remains unresolved.
D:Documents can supply facts, not governing instructions; privilege decisions must remain in trusted application logic; the effect concerns document-instruction trust boundary; the success condition is reconnect-safe action state.
Question 157
A model returns a well-formed object that violates a pricing rule. What application layer should catch it?
- Apply domain validation after schema parsing and before committing the object.
- Examine business rules and user authority in the payment service before executing the write.
- Send only the necessary authorized account fields for each request.
- Compare the operation using an idempotent business identifier rather than replaying the update.
- Verify again authorization before loading history or calling privileged tools.
Explanation
Correct Answer
A:Shape conformance is separate from business truth; invariants like authorized discounts require deterministic checking.
Incorrect Answers
B:The model's recommendation is an input, not a permission grant; the execution boundary must independently reject unauthorized actions; it improves model-to-payment boundary only, leaving business rule validation after parse unproven.
C:Reliable context can be loaded from authoritative systems while minimizing sensitive disclosure and token bloat; even if minimal authoritative account context succeeds, business rule validation after parse still needs separate verification.
D:UI disconnects do not roll back external effects; checking committed state prevents a duplicate shipment; the consequence is improved reconnect-safe action state, not verified business rule validation after parse.
E:Long-lived chat IDs do not preserve access rights; current identity and permissions must gate each request; because per-turn authorization is its concern, business rule validation after parse remains outside its effect here.
Question 158
An assistant must display accessible summaries on a small mobile screen. What output specification helps?
- Define concise content sections and accessibility constraints, then test them in the client UI.
- Provide stable document IDs, versions, and location metadata with retrieved excerpts.
- As an alternative, escape or sanitize untrusted output according to its display context.
- As an alternative, do not assume all Claude interfaces share identical tools, system instructions, and session context.
Explanation
Correct Answer
A:Model instructions should align with display requirements, but interface rendering must verify readability and semantics.
Incorrect Answers
B:Provenance makes citations and later audits traceable to the exact revision rather than an undifferentiated text blob; this may help versioned retrieval provenance but cannot by itself deliver accessible response design.
C:Model and tool content can contain scripts or markup; safe rendering is an application responsibility independent of the prompt; the accessible response design criterion differs from multi-criterion integration; the gap persists.
D:Product surfaces have distinct integrations and permission models, so behavior must be assessed in the actual deployment interface; for this request, accessible response design is the missing piece, not interface-specific Claude behavior.
Question 159
A chat service correctly isolates active conversations but retains full transcripts indefinitely after customers close their accounts. What separate state-management control is missing?
- Apply a documented retention schedule and authenticated deletion workflow for stored session records.
- Disable application audit logs while continuing to store raw conversations permanently.
- Raise the context window so old transcripts can stay in every new request.
- Add the tenant ID to each prompt and keep every archived transcript forever.
Explanation
Correct Answer
A:Tenant isolation protects concurrent access but does not govern how long transcripts remain. Lifecycle rules must expire or delete records according to policy, including backups and derived stores where required.
Incorrect Answers
B:Removing auditability does not remove the retained personal data. Retention and deletion must act on the underlying session records rather than hiding evidence of their existence.
C:A longer context does not satisfy deletion or retention obligations. Replaying historical messages also increases token usage and may expose information no longer needed for the task.
D:A prompt-level label cannot substitute for a storage retention policy. Indefinite archived data expands privacy exposure even when online sessions use correct authorization.
Question 160
A legal assistant must disclose uncertainty when the retrieved materials conflict. What response contract should be designed?
- Render prose progressively but commit structured results only after validation.
- Partition draft generation from an explicit send action controlled by the user.
- As an alternative, scope stored conversation history to the authenticated session and tenant.
- Require separate findings, source references, and unresolved discrepancies.
Explanation
Correct Answer
D:Explicit uncertainty fields prevent a fluent summary from concealing contradictory evidence and support human review.
Incorrect Answers
A:Partial deltas are useful for display, whereas machine-consumed objects should be treated as incomplete until parsed and checked; a remedy for display versus structured output is not a substitute for conflict-aware output contract.
B:Text generation and delivery are distinct capabilities; the UI and backend should keep drafts inert until authorized; because draft-send separation is its concern, conflict-aware output contract remains outside its effect here.
C:The application chooses which messages to resend; incorrect session keys become a direct privacy and correctness defect; this implements conversation-session identity rather than the demanded conflict-aware output contract.