Question 100
A user inserts 'ignore all rules' into a quoted customer review. Which architecture change should be approved first?
- For this workflow, parameterize tenant context and keep global rules separate from per-request variables.
- For the affected service, define a narrow output contract and validate structured responses before consumption.
- In the project scope, treat the review as untrusted data and enforce higher-priority task boundaries.
- Within the rollout design, state explicit scope and refusal/escalation conditions for unauthorized commitments.
Explanation
Correct Answer
C:The embedded instruction has no authority to override the system's allowed summarization task; one operational check should confirm that attack prompts embedded in reviews do not change tool permissions or task objective.
Incorrect Answers
A:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; this applies when a reusable prompt contains extensive tenant-specific details hard-coded into global instructions, but here a user inserts 'ignore all rules' into a quoted customer review.
B:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the alternative helps if a report generator repeatedly produces prose where downstream systems require JSON, instead of when a user inserts 'ignore all rules' into a quoted customer review.
D:Behavioral boundaries should make the assistant defer approval rather than invent authority; its scope includes cases where a support bot makes undocumented promises about customer reimbursements, whereas the problem presented is that a user inserts 'ignore all rules' into a quoted customer review.
Question 101
A monthly compliance reporting pipeline emits a JSON object consumed by software and includes cited legal statements. Testing reveals malformed fields and plausible claims unsupported by the source packet. Which TWO design controls must operate before the response reaches the consumer? Choose TWO.
- At the system boundary, require source-attributed claims and explicit uncertainty when evidence is missing.
- For the primary workstream, define a narrow output contract and validate structured responses before consumption.
- In the proposed design, parameterize tenant context and keep global rules separate from per-request variables.
- For the affected service, treat the review as untrusted data and enforce higher-priority task boundaries.
- In the project scope, separate stable policy instructions from user content and supply a version-controlled system template.
Explanation
Correct Answers
A:Clear evidence instructions help distinguish supported facts from interpretation in the final response; production testing should demonstrate that unsupported-claim probes trigger an uncertainty statement rather than fabricated citations; another obligation arises when a report generator repeatedly produces prose where downstream systems require JSON.
B:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the practical acceptance condition is that fuzzed inputs yield schema-valid objects or controlled validation failures; another obligation arises when a model must cite provided materials yet sometimes blends source statements with inference.
Incorrect Answers
C:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; this technique fits where a reusable prompt contains extensive tenant-specific details hard-coded into global instructions, and not where a model must cite provided materials yet sometimes blends source statements with inference; another obligation arises when a report generator repeatedly produces prose where downstream systems require JSON.
D:The embedded instruction has no authority to override the system's allowed summarization task; the trigger would be that a user inserts 'ignore all rules' into a quoted customer review, rather than the risk that a report generator repeatedly produces prose where downstream systems require JSON; another obligation arises when a model must cite provided materials yet sometimes blends source statements with inference.
E:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; the design helps when a billing assistant gives different refund criteria depending on the user's writing style, rather than the condition that a model must cite provided materials yet sometimes blends source statements with inference; another obligation arises when a report generator repeatedly produces prose where downstream systems require JSON.
Question 102
A support bot makes undocumented promises about customer reimbursements. Which approach should the design team choose?
- For the intended users, separate stable policy instructions from user content and supply a version-controlled system template.
- Within the delivery plan, define a narrow output contract and validate structured responses before consumption.
- At the system boundary, state explicit scope and refusal/escalation conditions for unauthorized commitments.
- For the primary workstream, put prompt versions through review, evaluation and controlled release.
- For this workflow, require source-attributed claims and explicit uncertainty when evidence is missing.
Explanation
Correct Answer
C:Behavioral boundaries should make the assistant defer approval rather than invent authority; the solution is testable when refund-limit probes consistently produce escalation instead of unauthorized commitments.
Incorrect Answers
A:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; that is defensible if a billing assistant gives different refund criteria depending on the user's writing style, whereas the current issue is that a support bot makes undocumented promises about customer reimbursements.
B:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the other workload is one where a report generator repeatedly produces prose where downstream systems require JSON, whereas the release criterion addresses a support bot makes undocumented promises about customer reimbursements.
D:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; the alternative helps if a team edits production prompts directly and cannot explain changed answers, not the risk where a support bot makes undocumented promises about customer reimbursements.
E:Clear evidence instructions help distinguish supported facts from interpretation in the final response; it is justified if a model must cite provided materials yet sometimes blends source statements with inference, not this situation where a support bot makes undocumented promises about customer reimbursements.
Question 103
A reusable prompt contains extensive tenant-specific details hard-coded into global instructions. Where should the solution architect focus the implementation?
- Under the reviewed approach, require source-attributed claims and explicit uncertainty when evidence is missing.
- For this production path, put prompt versions through review, evaluation and controlled release.
- For the intended users, define a narrow output contract and validate structured responses before consumption.
- Within the delivery plan, parameterize tenant context and keep global rules separate from per-request variables.
Explanation
Correct Answer
D:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; the audit trail can establish that tenant-switch tests show only authorized tenant context influences a response.
Incorrect Answers
A:Clear evidence instructions help distinguish supported facts from interpretation in the final response; it is relevant if a model must cite provided materials yet sometimes blends source statements with inference, instead of when a reusable prompt contains extensive tenant-specific details hard-coded into global instructions.
B:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; it serves conditions where a team edits production prompts directly and cannot explain changed answers, not the case in which a reusable prompt contains extensive tenant-specific details hard-coded into global instructions.
C:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the other case is when a report generator repeatedly produces prose where downstream systems require JSON, whereas the problem presented is that a reusable prompt contains extensive tenant-specific details hard-coded into global instructions.
Question 104
A model must cite provided materials yet sometimes blends source statements with inference. What revision would make the proposal more defensible?
- In the initial solution, require source-attributed claims and explicit uncertainty when evidence is missing.
- In the proposed design, define a narrow output contract and validate structured responses before consumption.
- Under the reviewed approach, treat the review as untrusted data and enforce higher-priority task boundaries.
- For this production path, state explicit scope and refusal/escalation conditions for unauthorized commitments.
Explanation
Correct Answer
A:Clear evidence instructions help distinguish supported facts from interpretation in the final response; the practical acceptance condition is that unsupported-claim probes trigger an uncertainty statement rather than fabricated citations.
Incorrect Answers
B:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; it serves a setting where a report generator repeatedly produces prose where downstream systems require JSON, and not where a model must cite provided materials yet sometimes blends source statements with inference.
C:The embedded instruction has no authority to override the system's allowed summarization task; that is defensible if a user inserts 'ignore all rules' into a quoted customer review, rather than the risk that a model must cite provided materials yet sometimes blends source statements with inference.
D:Behavioral boundaries should make the assistant defer approval rather than invent authority; the other workload is one where a support bot makes undocumented promises about customer reimbursements, rather than the condition that a model must cite provided materials yet sometimes blends source statements with inference.
Question 105
A team edits production prompts directly and cannot explain changed answers. How should the architect revise the proposed solution?
- At this decision point, define a narrow output contract and validate structured responses before consumption.
- As a candidate control, parameterize tenant context and keep global rules separate from per-request variables.
- In the initial solution, separate stable policy instructions from user content and supply a version-controlled system template.
- In the proposed design, put prompt versions through review, evaluation and controlled release.
Explanation
Correct Answer
D:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; a meaningful test result is that each output trace identifies the approved prompt revision and linked evaluation run.
Incorrect Answers
A:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the approach helps when a report generator repeatedly produces prose where downstream systems require JSON, whereas the release criterion addresses a team edits production prompts directly and cannot explain changed answers.
B:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; that belongs to situations where a reusable prompt contains extensive tenant-specific details hard-coded into global instructions, while the required safeguard concerns a team edits production prompts directly and cannot explain changed answers.
C:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; it is relevant if a billing assistant gives different refund criteria depending on the user's writing style, not the risk where a team edits production prompts directly and cannot explain changed answers.
Question 106
Two equivalent customer requests reach different refund outcomes after a wording-only prompt change. Which result should the pilot review prioritize?
- The release report verifies that paraphrased customer requests produce the same refund-policy decision.
- The controlled replay verifies that fuzzed inputs yield schema-valid objects or controlled validation failures.
- A reproducible run shows that unsupported-claim probes trigger an uncertainty statement rather than fabricated citations.
- The signed test record confirms that each output trace identifies the approved prompt revision and linked evaluation run.
Explanation
Correct Answer
A:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; the required safeguard is exercised when a billing assistant gives different refund criteria depending on the user's writing style.
Incorrect Answers
B:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; this control is needed when a report generator repeatedly produces prose where downstream systems require JSON, whereas the problem presented is that two equivalent customer requests reach different refund outcomes after a wording-only prompt change.
C:Clear evidence instructions help distinguish supported facts from interpretation in the final response; that would be useful if a model must cite provided materials yet sometimes blends source statements with inference, but this decision concerns two equivalent customer requests reach different refund outcomes after a wording-only prompt change.
D:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; it serves a setting where a team edits production prompts directly and cannot explain changed answers, but the actual risk is that two equivalent customer requests reach different refund outcomes after a wording-only prompt change.
Question 107
Imported customer text contains a forged instruction claiming to supersede the system policy. What result would directly validate the necessary safeguard?
- The release check confirms that unsupported-claim probes trigger an uncertainty statement rather than fabricated citations.
- The reviewer observes that paraphrased customer requests produce the same refund-policy decision.
- The release report verifies that fuzzed inputs yield schema-valid objects or controlled validation failures.
- The controlled replay verifies that attack prompts embedded in reviews do not change tool permissions or task objective.
Explanation
Correct Answer
D:The embedded instruction has no authority to override the system's allowed summarization task; it tests the operating case in which a user inserts 'ignore all rules' into a quoted customer review.
Incorrect Answers
A:Clear evidence instructions help distinguish supported facts from interpretation in the final response; its scope includes cases where a model must cite provided materials yet sometimes blends source statements with inference, and not where imported customer text contains a forged instruction claiming to supersede the system policy.
B:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; it targets cases where a billing assistant gives different refund criteria depending on the user's writing style, rather than the risk that imported customer text contains a forged instruction claiming to supersede the system policy.
C:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; the approach helps when a report generator repeatedly produces prose where downstream systems require JSON, rather than the condition that imported customer text contains a forged instruction claiming to supersede the system policy.
Question 108
A model-produced order record parses as JSON but omits a mandatory downstream field. What evidence would reveal whether the control truly works?
- The regression suite verifies that each output trace identifies the approved prompt revision and linked evaluation run.
- A production-like test shows that attack prompts embedded in reviews do not change tool permissions or task objective.
- The release check confirms that tenant-switch tests show only authorized tenant context influences a response.
- The reviewer observes that fuzzed inputs yield schema-valid objects or controlled validation failures.
Explanation
Correct Answer
D:Schema enforcement at the application boundary catches malformed fields instead of trusting prompt wording; it demonstrates the required control when a report generator repeatedly produces prose where downstream systems require JSON.
Incorrect Answers
A:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; the trigger would be that a team edits production prompts directly and cannot explain changed answers, while the required safeguard concerns a model-produced order record parses as JSON but omits a mandatory downstream field.
B:The embedded instruction has no authority to override the system's allowed summarization task; the design helps when a user inserts 'ignore all rules' into a quoted customer review, not the risk where a model-produced order record parses as JSON but omits a mandatory downstream field.
C:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; that fits cases where a reusable prompt contains extensive tenant-specific details hard-coded into global instructions, not this situation where a model-produced order record parses as JSON but omits a mandatory downstream field.
Question 109
A support reply follows a generic helpfulness instruction and promises a concession staff cannot grant. Which result proves the requested control, not just configuration?
- The end-to-end check confirms that refund-limit probes consistently produce escalation instead of unauthorized commitments.
- The observed outcome is that attack prompts embedded in reviews do not change tool permissions or task objective.
- The regression suite verifies that unsupported-claim probes trigger an uncertainty statement rather than fabricated citations.
- A production-like test shows that each output trace identifies the approved prompt revision and linked evaluation run.
- A controlled trial finds that tenant-switch tests show only authorized tenant context influences a response.
Explanation
Correct Answer
A:Behavioral boundaries should make the assistant defer approval rather than invent authority; this checks the situation in which a support bot makes undocumented promises about customer reimbursements.
Incorrect Answers
B:The embedded instruction has no authority to override the system's allowed summarization task; it is justified if a user inserts 'ignore all rules' into a quoted customer review, but this decision concerns a support reply follows a generic helpfulness instruction and promises a concession staff cannot grant.
C:Clear evidence instructions help distinguish supported facts from interpretation in the final response; its scope includes cases where a model must cite provided materials yet sometimes blends source statements with inference, but the actual risk is that a support reply follows a generic helpfulness instruction and promises a concession staff cannot grant.
D:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; it targets cases where a team edits production prompts directly and cannot explain changed answers, but the requested boundary concerns a support reply follows a generic helpfulness instruction and promises a concession staff cannot grant.
E:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; the approach helps when a reusable prompt contains extensive tenant-specific details hard-coded into global instructions, not the present case where a support reply follows a generic helpfulness instruction and promises a concession staff cannot grant.
Question 110
Global prompt maintenance accidentally alters a tenant's policy-specific behavior across unrelated services. Which acceptance result provides the most specific evidence?
- The measured evidence confirms that attack prompts embedded in reviews do not change tool permissions or task objective.
- The scenario test demonstrates that paraphrased customer requests produce the same refund-policy decision.
- The end-to-end check confirms that each output trace identifies the approved prompt revision and linked evaluation run.
- The observed outcome is that tenant-switch tests show only authorized tenant context influences a response.
Explanation
Correct Answer
D:Modularity reduces cross-tenant leakage and makes policy changes easier to audit; the safeguard must work when a reusable prompt contains extensive tenant-specific details hard-coded into global instructions.
Incorrect Answers
A:The embedded instruction has no authority to override the system's allowed summarization task; this makes sense when a user inserts 'ignore all rules' into a quoted customer review, rather than the risk that global prompt maintenance accidentally alters a tenant's policy-specific behavior across unrelated services.
B:Instruction hierarchy and fixed templates keep governing business rules consistent across requests; this technique fits where a billing assistant gives different refund criteria depending on the user's writing style, rather than the condition that global prompt maintenance accidentally alters a tenant's policy-specific behavior across unrelated services.
C:Prompt text is application logic whose untracked changes can cause regression and unsafe behavior; the trigger would be that a team edits production prompts directly and cannot explain changed answers, while the immediate question involves global prompt maintenance accidentally alters a tenant's policy-specific behavior across unrelated services.