Option C provides the minimum coherent package required to move from architectural intent to implementable and supportable delivery. The architecture overview establishes system boundaries and component relationships. Architecture Decision Records preserve the context and trade-offs behind key choices. Component contracts define interfaces, schemas, responsibilities, failure behavior, and integration assumptions.

The evaluation framework and reference set allow the implementation team to confirm that prompts, models, retrieval components, and tool integrations continue to meet the defined quality, safety, latency, and cost thresholds. Runbooks translate operational knowledge into repeatable procedures, while the on-call playbook defines detection, escalation, containment, and recovery actions. The known-limitations register prevents unsupported behavior from being mistaken for an implementation defect and helps stakeholders manage residual risk.

Options A and B depend excessively on slides or undocumented verbal knowledge. Option D provides implementation material without explaining the system-level decisions, operating model, acceptance criteria, or limitations. Source code cannot substitute for architectural and operational documentation.

A clean transition does not mean the architect can never be consulted. It means routine implementation and support do not depend on undocumented knowledge held by a single individual.

Study Guide references/topics: Design-to-implementation handoff; ADRs; component contracts; reference evaluations; runbooks; operational playbooks; limitations management.

===============

Yes, Yes, Yes, No, No

Dense vector retrieval represents semantic similarity, making it suitable for matching paraphrases and conceptually related language even when the query and source do not share identical words. Sparse lexical retrieval, including BM25-style matching, retains strong sensitivity to exact tokens and is therefore valuable for identifiers, product codes, technical names, and uncommon terminology.

Structured query retrieval is appropriate when the required operation depends on explicit fields, predicates, joins, counts, grouping, or aggregation over a relational schema. In that situation, generating or invoking a constrained database query is more precise than approximating the operation through semantic similarity.

Hybrid retrieval is not identical to dense retrieval. It combines semantic and lexical candidate sets, typically followed by rank fusion or reranking. Anthropic’s contextual-retrieval guidance explains that semantic search captures meaning and paraphrases, while BM25 captures exact terminology; combining them improves coverage. Contextual Retrieval, Contextual Retrieval Cookbook

Random sampling is not a relevance strategy. It provides no systematic relationship between the query and selected evidence, producing unstable coverage and preventable hallucination risk. Production Q & A requires deterministic or evaluated relevance mechanisms, access filters, suitable indexes, and measurable retrieval metrics such as recall at k.

Study Guide references/topics: Dense retrieval; sparse retrieval; structured queries; hybrid search; rank fusion; retrieval evaluation; production RAG design.

===============

An ethics review must examine unsupported generalizations, stereotyping, disparate treatment, and other conclusions about people that are not justified by the evidence. It must also preserve identifiable human accountability for consequential decisions; responsibility cannot be transferred to a probabilistic system. Licensing is a legal and commercial-control question, while latency and throughput are operational requirements. Both remain important, but neither is the central ethical test presented here. Restricting review to low-confidence or threshold-triggering outputs is also inadequate because models can express harmful or biased conclusions with high confidence. Ethical review should cover representative populations, protected and intersectional groups, high-impact outcomes, override mechanisms, transparency, and channels for challenge or appeal.

Study Guide references/topics: Bias and unsupported generalization; human accountability; consequential decisions; ethical oversight.

User scope applies to one engineer across every project and is therefore appropriate for portable personal preferences. Option A fits this definition because an interface theme belongs to the individual rather than to a repository or organization. Option B also fits because Claude Code provides a language setting for the preferred response language, and placing that preference in user scope makes it apply across repositories.

Option C requires managed configuration because an organization-wide security control must be centrally deployed and resistant to individual override. Option D belongs in project scope because repository-specific MCP endpoints must be shared consistently with that repository’s contributors.

Option E describes credential handling rather than merely a preference assignment. Credentials must not be placed in a repository, but avoiding version control alone is insufficient. They must be stored through a supported secure authentication mechanism, protected environment configuration, operating-system credential facility, or approved secrets manager—not inserted as plaintext into an ordinary settings file.

Anthropic defines user scope as affecting one user across all projects, project scope as shared repository configuration, and managed scope as the location for enforceable organizational controls. It also explicitly lists themes and other personal settings as user-scope use cases. Claude Code Configuration Scopes

Study Guide references/topics: User scope; project scope; managed configuration; personal preferences; repository-shared MCP configuration; credential protection.

===============

Over-refusal and hallucination are characteristic model-behavior risks. An overly broad safety response can reject a legitimate request even when the surrounding application operates correctly. A plausible but unsupported answer is a groundedness failure arising from probabilistic generation and inadequate source adherence. These risks require model-specific evaluations, grounding instructions, citations, verification, and calibrated fallback behavior. Expired certificates, database timeouts, and load-balancer misconfiguration are conventional distributed-system failures that should be addressed through infrastructure monitoring, retries, health checks, and configuration management. Distinguishing these categories matters because ordinary software remediation will not resolve behavioral failures, while prompt changes will not repair infrastructure. Anthropic’s hallucination guidance recommends auditable citations and supporting evidence for generated claims. Reducing hallucinations

Authorization and compliance logging must remain enforceable application-layer controls. Option B requires the application to authenticate the requester, resolve the user’s role and entitlements, and deny the operation when the user lacks authority. A Skill can instruct Claude how to behave, but it is not a reliable security boundary and cannot replace identity-based access control enforced by the surrounding system.

Option A similarly belongs at the application layer. The application must create an authoritative audit event containing the requester, timestamp, Skill and model version, relevant input or content identifiers, generated output, review status, and retention classification. Audit evidence cannot depend on whether the model remembers or chooses to invoke a logging instruction.

Options C, D, and E describe reusable generation behavior that appropriately belongs within the approved Skill: house style, the standardized disclaimer, and the prohibited-language policy. The application may still validate their presence after generation, but the Skill remains the primary reusable content package.

The correct architecture separates behavioral guidance from deterministic enforcement. Skills supply domain instructions and templates; the application enforces identity, authorization, audit retention, data access, output validation, and consequential-action controls.

Study Guide references/topics: Skills versus application controls; authentication; authorization; compliance logging; deterministic guardrails; policy enforcement boundaries.

===============

Role-based access control directly supports HIPAA’s access-control and minimum-necessary principles. The retrieval and orchestration layers must verify the authenticated user’s role before protected health information enters the model context. Authorization should be applied to individual records, data sources, tools, and actions, with audit events recording the requesting identity, accessed resource, purpose, and result. Model instructions alone are not a security boundary.

Patient information must also remain within services and processing activities covered by the organization’s Business Associate Agreement. Anthropic states that PHI processing requires a HIPAA-ready organization and an accepted BAA, and that only specified eligible services and configurations are covered. Patient data must not be submitted through training, feedback, beta, third-party, or integration pathways that fall outside the applicable agreement. A BAA is not blanket authorization; architects must verify feature eligibility and disable inappropriate data-sharing mechanisms.

Reducing max_tokens, selecting a more capable model, and enabling streaming affect output length, quality, or latency. They do not establish authorization, contractual coverage, minimum-necessary access, or compliant PHI handling.

Study Guide references/topics: Anthropic Business Associate Agreement requirements; HIPAA-ready services; PHI access control; minimum necessary; eligible-service boundaries; auditability.

===============

Responsible deployment requires users to receive an accurate representation of the system with which they are interacting. Option C provides that transparency while also establishing a practical escalation path. Disclosure should be proportionate to the material role AI plays, expressed in language appropriate to the audience, and aligned with organizational policy and applicable regulatory obligations.

A documented human-contact route is important when the assistant cannot resolve an issue, produces a disputed result, encounters a high-impact exception, or handles a matter requiring human authority. Anthropic’s Responsible Scaling Policy recognizes escalation to human reviewers as an appropriate safeguard for edge cases and situations requiring human judgment. Responsible Scaling Policy

Options A and D intentionally conceal material AI involvement from affected users, weakening informed decision-making, auditability, and trust. Option B is also unsuitable because transparency does not require revealing proprietary prompts, confidential controls, or internal security mechanisms; however, it does require an honest explanation of AI involvement and the system’s operational role.

The correct design combines disclosure, understandable limitations, human escalation, and records demonstrating that the organization’s transparency requirements are consistently applied.

Study Guide references/topics: AI transparency; informed user interaction; human escalation; responsible deployment; disclosure policy; operational accountability.

===============

A stable adversarial regression set provides a controlled basis for detecting behavioral change between model versions. Version-attributed results show whether the proposed upgrade improves, preserves, or degrades performance on the same safety-relevant cases. The evaluation must run before promotion so a regression can block deployment rather than becoming a production incident. Randomly replacing the test inputs removes comparability and makes score changes difficult to attribute. Disabling evaluation or relying on complaints converts a preventable release-control failure into reactive incident handling. Anthropic characterizes regression evaluations as tests of whether an agent still handles tasks it previously completed and recommends maintaining them continuously to detect backsliding. Demystifying evals for AI agents.

Robust guardrails require both preventive enforcement and continuous empirical validation. Per-role tool allow-lists satisfy the preventive requirement by restricting each authenticated role to explicitly authorized tools. Enforcement must occur in the orchestration or permission layer before execution; relying on the model to decide whether a tool call is authorized is not an adequate security boundary. Anthropic recommends allow-list-based validation and provides permission rules and pre-tool-use controls capable of denying calls before they execute.

Adversarial evaluation with regression tracking supplies the validation component. The evaluation set should include prompt injections, indirect injections, encoded instructions, multi-turn escalation, role-confusion attempts, and requests designed to trigger unauthorized tool use. Guardrail performance must then be measured across releases so that changes to prompts, models, tools, or orchestration logic do not silently reduce protection.

Centralized logging and user feedback are useful detective and lifecycle controls, but neither prevents an unsafe action nor proves guardrail effectiveness. Periodically refreshing refusal wording is also insufficient unless the revised behavior is evaluated against defined safety criteria. The strongest design therefore combines deterministic authorization with adversarial regression testing.

Study Guide references/topics: Claude Code permissions and pre-tool enforcement; tool allow-list guidance; adversarial evaluations; regression testing; least-privilege orchestration.

===============

Architecture X fails a stated production gate: its 4.2-second average exceeds the required three-second response SLA. It should therefore not remain a qualifying architecture unless it is redesigned and retested before selection. A promise to optimize after deployment is not evidence of compliance.

Architecture Y meets the latency requirement at 2.1 seconds. Its annual inference cost is $180,000 at two million sessions, compared with $360,000 for Architecture X—a difference of $180,000 annually. However, cost alone does not establish suitability. The organization must determine whether Y’s lower quality score causes material omissions, correction effort, clinician dissatisfaction, documentation risk, or reduced workflow productivity.

Option C correctly treats the SLA as a threshold and then analyzes the remaining architecture across cost, efficiency, quality, and productivity. Option A ignores clinical quality. Option B disregards an explicit SLA failure. Option D may eventually be viable, but it introduces routing complexity and leaves unanswered how complex cases are detected, whether X’s latency becomes acceptable for them, and how both paths are evaluated.

The final choice should be gated by clinical-quality evaluation, human review requirements, safety metrics, operational latency, and total annual cost.

Study Guide references/topics: Business-value pillars; performance gates; solution cost; workflow efficiency; quality trade-offs; healthcare architecture evaluation.

===============

Category 1 tools are necessary for the agent to complete its approved responsibilities and must remain. Category 2 tools are also justified when they are regularly used and eliminate predictable hand-offs without expanding the agent beyond its defined operating role.

Categories 3 and 4 create capability bloat. A tool that is occasionally useful only for unrelated work does not belong to this agent’s responsibility boundary. A speculative tool has no validated requirement and adds attack surface, context consumption, authorization complexity, evaluation burden, and operational dependencies without demonstrated value.

Tool inclusion should be based on task traceability: every configured capability should map to a documented user journey, responsibility, permission scope, and evaluation case. Anthropic recommends least privilege so that a successful injection or model error can cause minimal damage. Mitigate Jailbreaks and Prompt Injections

The architect should periodically review tool-call telemetry and remove unused capabilities. New tools can be added when a validated workflow requires them, after security review and evaluation. The objective is not the smallest possible catalog regardless of usefulness; it is the smallest catalog that reliably completes the agent’s approved tasks.

Study Guide references/topics: Capability-bloat analysis; tool necessity; hand-off reduction; least privilege; attack-surface management; tool lifecycle review.

===============