Question 560
A coding agent launched in a repository should use the team's established build and review conventions. Which of the following would work as specified?
- Use CLAUDE.md for repository guidance in Claude Code
- Configure permissions within the relevant client
- Persist conversation history in application storage
- Read project guidance from Claude Code files
Explanation
Correct Answer
D:Claude Code can load contextual project instructions such as CLAUDE.md during its sessions; a crucial qualification is that those instructions do not automatically govern an isolated custom API service.
Incorrect Answers
A:Repository instruction files provide persistent project context to the Claude Code environment; in contrast, claude Code can load contextual project instructions such as CLAUDE.md during its sessions, while they are not automatically loaded into unrelated Messages API clients.
B:Client-specific settings define the environment and its permitted tool behavior; this addresses a separate concern, not the fact that claude Code can load contextual project instructions such as CLAUDE.md during its sessions; configuring a distant backend does not necessarily constrain a local coding session.
C:The application saves and reconstructs relevant turns because Messages requests are stateless; here, the controlling capability is that claude Code can load contextual project instructions such as CLAUDE.md during its sessions, but assuming the API remembers every prior independent call will break continuity.
Question 561
A request worker performs many independent network round trips to the same remote endpoint. What is the appropriate next design choice?
- Use connection reuse for frequent calls
- Use an official SDK for conventional API calls
- Use request timeouts for bounded waits
- Use retry-aware HTTP error handling
Explanation
Correct Answer
A:Persistent connections reduce repeated connection setup overhead; the implementation must still recognize that pool sizing still has to respect concurrency and upstream rate limits.
Incorrect Answers
B:A supported client library wraps the underlying service API and common network details; the missing capability is more specific: persistent connections reduce repeated connection setup overhead; the SDK does not decide business-level retries or permissions automatically.
C:Client networking deadlines convert unbounded waits into controlled failures; by comparison, the intended solution relies on this: persistent connections reduce repeated connection setup overhead; a model token cap does not constrain socket wait times.
D:Status classification drives different responses for transient and permanent errors; the needed capability is different—persistent connections reduce repeated connection setup overhead—and blindly retrying invalid JSON does not improve correctness.
Question 562
Incoming tickets should be placed into one of four known queues before specialized pipelines run. How should this requirement be met?
- Run an explicit fixed-stage workflow
- Route requests with an LLM classifier
- Use a model-directed agent loop
- Use deterministic code for a known calculation
Explanation
Correct Answer
B:A routing call chooses among bounded destinations while downstream code owns the actual execution; in production, classification alone cannot search, repair, or complete an open-ended case.
Incorrect Answers
A:Application code controls the predetermined stage transitions and makes them reproducible in tests; the missing capability is more specific: a routing call chooses among bounded destinations while downstream code owns the actual execution; it cannot decide to introduce an unplanned research step after unexpected evidence appears.
C:Claude selects an action after observing each result, with the application enforcing iteration and permission limits; the needed capability is different—a routing call chooses among bounded destinations while downstream code owns the actual execution—and the autonomy adds variability and demands hard budgets, logs, and a stopping policy.
D:Ordinary code computes the numerical result repeatably rather than trusting text generation for arithmetic, whereas the requirement calls for another behavior: a routing call chooses among bounded destinations while downstream code owns the actual execution; a language model may describe the calculation but should not be its financial source of truth.
Question 563
A production service needs to call Claude but the application repository is shared with many developers. Which design would meet this acceptance criterion?
- Use separate credentials for development and production
- Revoke integration access when offboarding
- Rotate a key after suspected leakage
- Store credentials in a secret manager
Explanation
Correct Answer
D:Protected storage and runtime injection keep billable API keys out of committed files; a crucial qualification is that encryption at rest does not justify exposing secrets in source control.
Incorrect Answers
A:Environment isolation makes accidental or compromised test execution less damaging; in contrast, protected storage and runtime injection keep billable API keys out of committed files, while a single unrestricted key creates unnecessary blast radius across environments.
B:Offboarding revokes credentials and grants to prevent future unauthorized use; this addresses a separate concern, not the fact that protected storage and runtime injection keep billable API keys out of committed files; deleting the employee's saved chat does not revoke external tool credentials.
C:Revoking and replacing the affected secret reduces the time attackers can continue using it; here, the controlling capability is that protected storage and runtime injection keep billable API keys out of committed files, but hiding the log page alone cannot guarantee the secret was not captured.
Question 564
A customer asks a known low-risk question during a temporary model outage. Which approach targets the deciding condition?
- Show uncertainty when evidence is insufficient
- Return a verified FAQ when inference is unavailable
- Disable privileged tools during security incidents
- Reject unsupported operations explicitly
Explanation
Correct Answer
B:A controlled cache of approved responses preserves basic service where exact matches exist; the system must avoid pretending a cached answer is newly generated or universally applicable.
Incorrect Answers
A:The product communicates the gap and offers verification rather than inventing a confident conclusion; by comparison, the intended solution relies on this: a controlled cache of approved responses preserves basic service where exact matches exist; more assertive response wording would make an unsupported answer misleading.
C:A feature flag or policy change can revoke risky execution paths while investigation continues, whereas the requirement calls for another behavior: a controlled cache of approved responses preserves basic service where exact matches exist; a prompt-only warning cannot reliably disable an exposed backend permission.
D:A clear capability boundary returns an accurate unsupported-action response; however, the critical mechanism here is that a controlled cache of approved responses preserves basic service where exact matches exist, and inventing a successful action would violate user trust and corrupt workflow state.
Question 565
A customer support agent should inspect ticket status but not invoke a broad account deletion endpoint. Which production safeguard is most relevant?
- Use idempotency keys for retriable writes
- Use an MCP-provided tool through a configured client
- Allowlist tools per agent role
- Validate caller authorization per invocation
Explanation
Correct Answer
C:Restricting registered capabilities limits possible operations at the dispatch boundary; the important constraint is that a polite instruction cannot fully constrain a tool whose credentials are overprivileged.
Incorrect Answers
A:Stable transaction identifiers allow duplicate attempts to be detected before replaying the business action; the needed capability is different—restricting registered capabilities limits possible operations at the dispatch boundary—and replaying without deduplication can create multiple charges.
B:The MCP client discovers the server's tools and invokes permitted operations through a standardized protocol, whereas the requirement calls for another behavior: restricting registered capabilities limits possible operations at the dispatch boundary; MCP transport does not replace authentication or authorization of the business system.
D:The backend checks owner, resource, action, and policy before accepting the call; the requirement instead turns on this behavior: restricting registered capabilities limits possible operations at the dispatch boundary; model-generated account identifiers are not an authorization proof.
Question 566
A network-incident agent must adapt its next diagnostic tool to surprising observations, yet operations must impose a hard cap on the number of tool actions per request. Which TWO mechanisms provide flexibility without an unbounded loop? Choose TWO.
- Assign action limits in the agent harness
- Use a model-directed agent loop
- Keep model context limited to relevant evidence
- Run an explicit fixed-stage workflow
- Keep a single direct model call
Explanation
Correct Answers
A:Host-controlled counters and deadlines bound execution regardless of what Claude decides next; the important constraint is that prompt instructions alone cannot enforce a hard external quota.
B:Claude selects an action after observing each result, with the application enforcing iteration and permission limits—with this limitation: the autonomy adds variability and demands hard budgets, logs, and a stopping policy.
Incorrect Answers
C:A curated evidence window preserves useful facts while discarding noisy intermediate transcripts; this cannot substitute for the behavior that host-controlled counters and deadlines bound execution regardless of what Claude decides next, because pruning must preserve commitments and results required for downstream steps.
D:Application code controls the predetermined stage transitions and makes them reproducible in tests; in contrast, host-controlled counters and deadlines bound execution regardless of what Claude decides next, while it cannot decide to introduce an unplanned research step after unexpected evidence appears.
E:A direct request minimizes latency and operational state for a bounded answer; this addresses a separate concern, not the fact that host-controlled counters and deadlines bound execution regardless of what Claude decides next; an agent loop would add failure modes without yielding a decision the model must make.
Question 567
A classifier handles ambiguous categories that are difficult to describe by abstract rules alone. How should the implementation handle this need?
- Use multiple diverse examples for edge cases
- Evaluate examples for data leakage
- Use zero-shot instructions for a straightforward task
- Use a fixed regression dataset for stable comparisons
Explanation
Correct Answer
A:Contrasting examples illustrate boundaries and exceptions across several realistic inputs; a crucial qualification is that few-shot prompts add tokens and can create bias if examples are unrepresentative.
Incorrect Answers
B:Approved synthetic or sanitized cases teach the format without disclosing unrelated private content; by comparison, the intended solution relies on this: contrasting examples illustrate boundaries and exceptions across several realistic inputs; good few-shot performance does not justify embedding sensitive identifiers.
C:Direct instructions define the output without consuming tokens on demonstrations; the needed capability is different—contrasting examples illustrate boundaries and exceptions across several realistic inputs—and complex edge cases may still need examples and evaluations.
D:Repeated evaluation over the same representative cases reveals distributional quality differences, whereas the requirement calls for another behavior: contrasting examples illustrate boundaries and exceptions across several realistic inputs; one identical answer does not prove full determinism across deployments.
Question 568
A support answer cannot find an authoritative source for the customer's disputed claim. Which mechanism should the team prioritize?
- Show uncertainty when evidence is insufficient
- Sanitize generated HTML before rendering
- Preserve incremental rendering order
- Map model fields to application fields explicitly
Explanation
Correct Answer
A:The product communicates the gap and offers verification rather than inventing a confident conclusion; a crucial qualification is that more assertive response wording would make an unsupported answer misleading.
Incorrect Answers
B:Escaping or trusted sanitization prevents untrusted content from executing in the browser; by comparison, the intended solution relies on this: the product communicates the gap and offers verification rather than inventing a confident conclusion; system prompts cannot provide a browser-level cross-site scripting defense.
C:Ordered stream assembly avoids scrambled or duplicated text fragments; the needed capability is different—the product communicates the gap and offers verification rather than inventing a confident conclusion—and parallel callbacks that append out of order can corrupt an otherwise valid response.
D:An adapter transforms and validates values at the boundary rather than assuming identical schemas, whereas the requirement calls for another behavior: the product communicates the gap and offers verification rather than inventing a confident conclusion; directly sending raw model content to a write endpoint creates avoidable coupling.
Question 569
An MCP server update changes a tool's input schema consumed by multiple agents. What is the appropriate next design choice?
- Review plugin permissions before activation
- Pin and test MCP server compatibility
- Retain a tested rollback version
- Remove unused integrations
Explanation
Correct Answer
B:Version tracking and integration tests preserve compatibility at the tool boundary; automatically adopting an untested update can break dispatch or authorization.
Incorrect Answers
A:Least-privilege permission review limits what installed integrations can access; by comparison, the intended solution relies on this: version tracking and integration tests preserve compatibility at the tool boundary; a friendly plugin description is not a security audit.
C:Keeping the previously tested artifact enables a measured rollback when safe, whereas the requirement calls for another behavior: version tracking and integration tests preserve compatibility at the tool boundary; a rollback should consider state migrations and outstanding jobs.
D:Decommissioning unneeded capabilities reduces the exposed attack surface and maintenance burden; however, the critical mechanism here is that version tracking and integration tests preserve compatibility at the tool boundary, and merely hiding a tool from a menu may not revoke its underlying credential.
Question 570
Several separate compatible assistants need to find the same data resources and tools dynamically. Which design would meet this acceptance criterion?
- Use an MCP server for a reusable integration
- Prefer MCP for cross-client discovery
- Prefer one application-local function for a one-off feature
- Prefer a prompt template for pure language guidance
Explanation
Correct Answer
B:MCP standardizes discovery and execution interfaces across participating clients; notably, standard interfaces do not automatically secure individual backends.
Incorrect Answers
A:A protocol server exposes shared tools and resources with consistent discovery semantics; the requirement instead turns on this behavior: MCP standardizes discovery and execution interfaces across participating clients; it adds infrastructure and identity concerns compared with a single local function.
C:A local function minimizes protocol, hosting, and dependency overhead; in contrast, MCP standardizes discovery and execution interfaces across participating clients, while reusability infrastructure should be justified by real multiple-client requirements.
D:A maintained prompt template provides reusable text behavior without operating a tool; this addresses a separate concern, not the fact that MCP standardizes discovery and execution interfaces across participating clients; creating an executable integration would add complexity to a text-only requirement.