CCAR-P Claude Certified Architect - Professional Questions and Answers

The retrieval pipeline is stated to be unchanged, so the first investigation should isolate changes in generation behavior and grounding instructions. Replaying the failing cases on the previous model version determines whether the regression follows the model upgrade or current model-prompt combination. Inspecting the system prompt establishes whether source-use rules, uncertainty behavior, citation requirements, or prohibitions against unsupported claims were removed or weakened. Increasing context or replacing the embedding model changes a component for which no failure evidence exists and may introduce additional noise. Lowering temperature may reduce variation but does not explain systematic contradiction of retrieved evidence. The investigation should compare identical requests, retrieved passages, prompts, and model versions before applying a fix. Define success criteria and evaluations

The evaluation specification must exist before inference begins. Option D defines what the evaluation will measure and how results will be segmented. Representative cases measure normal workload performance, edge cases test boundary behavior, and adversarial cases examine safety, security, or robustness under hostile inputs.

Option B then supplies the correctly curated and labeled dataset corresponding to those slices. Labels, expected outcomes, grading rubrics, provenance, and dataset versions must be established before the deployment generates outputs. Otherwise, evaluators risk modifying criteria after seeing results and introducing confirmation bias.

Anthropic defines an evaluation as an input combined with grading logic used to measure success. Its evaluation guidance distinguishes tasks, trials, graders, traces, outcomes, and the harness that runs and aggregates them. Demystifying Evals for AI Agents

Option E occurs after model execution because outputs must exist before they can be scored. Option A follows scoring and aggregation. Option C is an iterative improvement step performed after failure cases become available, although rubric calibration may also be conducted on separate development data.

Study Guide references/topics: Evaluation specification; metrics and slices; representative and adversarial datasets; labeling; scoring; aggregation; release gates.

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

A Skill is the appropriate packaging mechanism for reusable procedural knowledge, instructions, templates, examples, and supporting resources. The filing capability follows a defined authoring method and requires no live external-system interaction, so an MCP server would introduce unnecessary transport, authentication, hosting, and operational complexity. MCP becomes appropriate when Claude must invoke live databases, APIs, or enterprise services. Command-line availability does not independently determine the design because Skills can be used across supported Claude environments. Direct identity-provider authentication would favor an integration or MCP service rather than a self-contained Skill. Anthropic describes Agent Skills as portable packages that provide agents with specialized procedural and organizational expertise. Agent Skills

Option D applies three complementary controls. First, a read-only database credential creates an authorization boundary outside the model; prompt instructions alone cannot convert that credential into write access. Second, restricting the subagent’s tool list reduces capability exposure by preventing the agent from selecting unrelated or unnecessarily privileged operations. Third, explicit human confirmation creates a deliberate approval gate before any exceptional data-changing action is executed.

This is defense in depth. If Claude misinterprets a request or processes malicious instructions from untrusted content, the restricted credential and tool configuration limit the available action surface. Human review then protects operations with potentially irreversible production consequences. Logging and audit trails should remain enabled to record the actor, request, tool call, approval, affected records, and outcome.

Anthropic documents that Claude Code begins with read-only permissions in Manual mode and requests approval for actions that modify the environment. Its permission system also supports granular allow, ask, and deny controls for MCP tools and subagents. Claude Code Security, Configure Permissions

Options A and B eliminate least privilege and approval boundaries. Option C removes essential detection and forensic evidence.

Study Guide references/topics: Production database safety; least privilege; read-only defaults; MCP permissions; subagent scoping; human confirmation; auditability.

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

Federated authorship allows each domain team to maintain the procedural knowledge it understands while preserving organization-wide trust through a central publication gate. Team authors can update coding standards, review practices, and postmortem procedures without waiting for a single platform group to write every change. Central reviewers can validate ownership, security, tool permissions, duplication, versioning, testing evidence, and compliance before publication. Per-developer or fully decentralized publication creates inconsistent and potentially unsafe Skills. Fully centralized authorship becomes a scalability bottleneck and separates content maintenance from domain expertise. Anthropic describes Skills as packages of instructions, scripts, and resources and warns that repository Skills may grant broad tool access, making formal review essential. Claude Code Skills

Before a pilot begins, the architect must understand the workflows being enabled, the associated security boundaries, and the correct ownership of each configuration decision. Option D establishes this foundation. Organization-wide, non-overridable controls belong in managed configuration, whereas repository-specific MCP servers, shared permissions, and team subagents generally belong in project scope.

Option A then converts those decisions into an initial project baseline that the pilot group can use. The baseline should identify approved MCP endpoints, minimum permission rules, denied operations, subagent responsibilities, configuration ownership, and observable success criteria. Anthropic distinguishes managed, user, project, and local scopes according to who they affect and whether they are shared with the team. Claude Code Configuration Scopes

Options B and C are broad-rollout activities and should not occur before pilot evidence is available. Option E is explicitly post-pilot: findings must first be collected before the configuration can be iterated and stabilized.

The correct lifecycle is discovery, boundary definition, baseline design, limited pilot, evidence review, baseline refinement, documentation and training, followed by controlled expansion with a support and rollback process.

Study Guide references/topics: Claude Code rollout planning; scope ownership; MCP baselines; permission design; pilot strategy; operational enablement.

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

Tool Search applies progressive discovery to a large tool catalog. Instead of inserting every MCP tool’s complete JSON schema and description into the initial context, Claude receives limited discovery information and loads the full definitions only when a task requires them. This preserves the available context for instructions, repository content, conversation history, and tool results without removing the underlying capabilities.

Anthropic’s Claude Code documentation states that Tool Search keeps MCP context usage low by deferring tool definitions until Claude needs them. Only the tools actually selected enter the working context. Claude Code MCP Tool Search

Anthropic’s advanced-tool-use guidance also describes conventional multi-server configurations consuming tens of thousands of tokens before work begins, while on-demand discovery substantially reduces that overhead. Advanced Tool Use

Option B solves context pressure by destroying required functionality. Option C worsens the constraint by adding more prompt tokens. Option D expands the exact source of context bloat.

The team should also improve server and tool descriptions so discovery returns the correct tools, monitor tool-selection quality, and keep frequently required foundational tools non-deferred only when evaluations justify the additional context cost.

Study Guide references/topics: MCP scaling; Tool Search; progressive discovery; deferred schemas; context-budget management; tool-selection accuracy.

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

Control A directly constrains the outbound action surface. An allow-list limits which tools and destinations may receive data, while per-call review introduces an approval boundary before information leaves the controlled environment. These measures therefore address data exfiltration through outbound tool calls, making Option C correct. Anthropic’s security guidance recommends least privilege, narrowly scoped permissions, sandboxing, and limiting access to sensitive information and actions so that a compromised agent can cause minimal damage. Mitigate Jailbreaks and Prompt Injections

Control B addresses model-version drift. Running a stable adversarial evaluation set after each version change creates a consistent comparison baseline and detects regressions that might otherwise remain invisible during routine testing. This correctly maps Control B to silent quality drift, making Option E correct. Anthropic’s migration guidance repeatedly calls for workload reevaluation because instruction following, style, reasoning, and agent behavior can differ between versions. Model Migration Guide

Control A may limit the consequences of successful prompt injection, but it does not prevent or directly detect adversarial instructions in retrieved content. Control B measures behavioral performance but does not impose outbound authorization controls. Each control must therefore be mapped to the risk it most directly mitigates.

Study Guide references/topics: Tool governance; least privilege; exfiltration controls; adversarial evaluations; model-change regression testing; defense in depth.

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

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.

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

The prompt contains mutually incompatible behavioral requirements. The architect should resolve the contradiction explicitly so the model receives one coherent grounding policy. If the application requires source-supported answers, the prompt should prohibit unsupported completion, permit an insufficient-evidence response, and require citations or supporting passages. Adding another priority instruction leaves the underlying conflict in place and creates ambiguity about when speculation is permitted. Increasing temperature makes behavior more variable rather than more correct. Retaining both instructions produces inconsistent outputs across requests and model versions. Anthropic recommends clear, direct, specific prompting and notes that conflicting project instructions can cause Claude to select among them unpredictably. Prompting best practices; project-instruction troubleshooting

Empirical model testing requires a defined decision frame. Option D must occur first because the quality threshold, acceptable latency, expected request volume, and associated cost envelope determine what constitutes a successful result. Without these criteria, a representative-sample test can produce measurements but cannot support a defensible selection.

Option C is the next upstream step. Known model capabilities and the workload profile are used to identify a plausible candidate rather than testing models arbitrarily. Anthropic’s model-selection guidance recommends evaluating capability, speed, cost, and effort requirements before narrowing the candidate set. Choosing the Right Model

Option A occurs after empirical testing because the architect cannot know which lightest model consistently meets the quality bar until sample results have been evaluated. Option B belongs to ongoing lifecycle management after a new version becomes available. Option E is a deployment-governance activity that follows successful model selection, security review, and readiness validation.

The correct sequence is: define workload criteria, shortlist candidate models, run a representative evaluation, compare quality and operational metrics, select the least costly model that reliably satisfies all gates, and periodically retest the decision.

Study Guide references/topics: Requirements-driven model selection; representative sampling; quality thresholds; latency and volume requirements; capability–cost trade-offs; model lifecycle reassessment.

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

The strongest investigation combines configuration comparison with controlled replay. Comparing the current system prompt and output-schema settings against the last known good version identifies maintenance-related configuration drift. Relevant differences include removed formatting instructions, altered schema requirements, incompatible parser expectations, changed stop sequences, or a disabled structured-output mechanism.

Replaying known premaintenance requests against the current configuration creates a controlled test set. Inspecting the raw response before parsing then establishes whether malformed JSON originates in model generation or is introduced by streaming assembly, transport handling, post-processing, or the downstream parser. If the raw response is valid but the pipeline rejects it, the fault lies after generation. If the raw response is malformed, the prompt, schema enforcement, stop reason, and request configuration become the primary investigation targets.

Clearing the prompt cache assumes cache corruption without evidence. Changing model tiers introduces another variable and weakens causal attribution. Raising `max_tokens` is appropriate only when raw responses show truncation and the returned stop reason indicates the token limit was reached; it is not the most direct initial test.

Study Guide references/topics: [Increasing output consistency](https://docs.anthropic.com/en/docs/test-and-evaluate/strengthen-guardrails/increase-consistency); [agent evaluation methodology] (https://www.anthropic.com/engineering/demystifying-evals-for-ai-agents); regression replay; last-known-good comparison; raw-output inspection.

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

Self-declared administrative authority inside a prompt is untrusted data, not authenticated identity or authorization evidence. Option B correctly combines independent controls across the model, runtime, authorization, and monitoring layers.

Prompt instructions establish that user content cannot override system policy. Runtime classifiers detect known and generalized attempts to bypass controls. Tool permissions are enforced outside the model and must derive from authenticated identity, role, and approved scope—not statements contained in the conversation. Audit logging records the actor, attempted override, classifier result, tool requests, and final disposition for investigation and control improvement.

Anthropic recommends input screening, hardened system prompts, safe handling of untrusted content, narrowly scoped permissions, red-team testing, and continuous monitoring. Mitigate Jailbreaks and Prompt Injections

Option A depends entirely on model behavior and provides no containment if the model fails. Option C removes the protections the attacker is attempting to defeat. Option D commits a fundamental authorization error by accepting an unverified claim as privilege elevation.

The strongest design also rate-limits repeated attempts, escalates suspicious activity, validates outputs, and requires human confirmation for consequential actions.

Study Guide references/topics: Direct prompt injection; untrusted user content; runtime classifiers; non-model authorization; scoped tools; audit logging; defense in depth.

Sound human-in-the-loop design places review before a high-impact or irreversible action occurs. The reviewer must receive sufficient context and explicit approval criteria so that the checkpoint functions as a meaningful control rather than a ceremonial step. This is especially important when Claude outputs could initiate financial transactions, publish sensitive material, alter access rights, affect employment decisions, or trigger other consequential actions. Option A provides only retrospective auditing and cannot prevent harm. Option C may support monitoring or quality measurement, but random sampling does not protect every case that requires mandatory approval, particularly when review criteria are undefined. Option D removes the accountability and intervention mechanism entirely. Therefore, B provides the strongest design by combining risk-based placement, human authority, and clearly defined verification requirements before execution.

Regression testing determines whether a change has broken behavior that previously satisfied the defined acceptance criteria. A stable reference set preserves known-good cases and enables the updated prompt to be compared with the prior baseline using consistent scoring rules. This makes Option B directly aligned with the described failure. Adversarial testing focuses on malicious or boundary inputs, such as prompt injection. Smoke testing confirms that the system remains broadly available, but it does not measure whether previously correct model behavior has degraded. Integration testing verifies contracts and interactions among components and may not detect output-quality regressions caused specifically by a prompt change. Anthropic recommends task-specific evaluations that mirror real workloads, include edge cases, and measure performance against explicit success criteria throughout prompt iteration. Anthropic: Define success criteria and build evaluations

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

A structured feedback session must enable informed review, traceable decisions, and accountable follow-through. Option B allows stakeholders to examine architecture diagrams, evaluation results, risk registers, or prototypes before the meeting. The focused agenda keeps discussion aligned with decisions that must be made rather than allowing the session to become an unstructured status conversation. Written decision records prevent different participants from leaving with incompatible interpretations. Assigning an owner and due date to every action converts feedback into executable delivery work. Options A, C, and D depend heavily on memory, provide inadequate preparation time, and create weak decision traceability. In-flight deployments particularly require documented feedback because accepted changes may affect scope, cost, risk, testing, and release timing.

Study Guide references/topics: Stakeholder feedback; decision records; action ownership; lifecycle alignment; communication planning.

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.

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

A workflow is appropriate when processing stages and transitions are known in advance. Metadata extraction, classification, summarization, and persistence can be represented as deterministic nodes with explicit schemas, validation rules, retry policies, and failure handling. This provides predictable execution, cost, observability, and testability without paying for repeated model-driven planning. Anthropic’s Building Effective AI Agents distinguishes workflows, where predefined code paths control execution, from agents, where the model dynamically determines its process. Option A reaches the correct pattern for an incorrect reason: workflows and agents can both invoke tools. Options B and C introduce unnecessary autonomy and make unsupported claims about consistency or universal accuracy.

Study Guide references/topics: Workflow versus agentic patterns; deterministic orchestration; batch processing; structured outputs; predictable cost; error handling.

A minimal reproduction and MCP status inspection are foundational diagnostic actions because they produce direct, interpretable evidence. A minimal reproduction removes unrelated prompts, tools, settings, hooks, and repository state while changing one variable at a time. This determines whether the failure follows a particular configuration, server, command, model request, or project artifact. It also creates a reproducible evidence package if escalation later becomes necessary.

MCP inspection addresses a common Claude Code failure domain. `claude mcp list` displays configured servers and their health states, including connected, authentication-required, pending-approval, and failed-to-connect conditions. `claude mcp get < name > ` provides server-specific details, while `/mcp` exposes status within an interactive session. These checks distinguish model behavior from transport, authentication, configuration, and workspace-trust failures.

Increasing temperature reduces repeatability and introduces additional variation, making root-cause isolation more difficult. Filing a support ticket before collecting local evidence reverses the appropriate diagnostic sequence. An immediate rollback may eventually be used as a controlled version comparison, but it changes the environment before a baseline has been captured and is therefore not a primary toolkit action.

Study Guide references/topics: [Managing and diagnosing MCP servers](https://docs.anthropic.com/en/docs/claude-code/mcp); minimal reproducible cases; controlled-variable troubleshooting; authentication and connection-state diagnostics.

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

Model Stage, Post-Processing, Post-Processing, Model Stage, Pre-Processing, Pre-Processing

Pre-processing prepares source material before Claude receives it. Optical character recognition converts scanned invoice images into machine-readable text. Because PII redaction is explicitly performed before model invocation, it also belongs in pre-processing. Additional activities at this layer can include file validation, malware scanning, normalization, page separation, and metadata extraction.

The model stage contains tasks requiring Claude’s language-understanding capability. Classifying the invoice type involves interpreting textual and contextual features, while extracting structured invoice fields requires mapping unstructured content into defined business attributes.

Post-processing verifies and commits the model’s result. Schema validation confirms that required fields exist, data types are correct, enumerated values are permitted, and structural constraints are satisfied. Persistence must occur only after validation and any required human review because the system of record should not receive malformed or unapproved model output.

This decomposition separates probabilistic inference from deterministic processing. OCR, redaction, validation, and persistence do not need to be delegated to Claude when conventional components can execute them more predictably. The resulting architecture reduces model workload, strengthens privacy controls, improves auditability, and prevents unvalidated output from directly changing authoritative records.

Study Guide references/topics: Task decomposition; pre-processing; model inference; post-processing; deterministic validation; privacy-by-design; system-of-record protection.

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

The 80,000-token filing is a large, stable prefix reused across fourteen requests, making prompt caching the strongest optimization. The first request creates the cache; subsequent identical-prefix requests use lower-priced cache reads while retaining the complete filing, preserving the evidence available to Claude. Anthropic documents cache reads at a fraction of standard input cost and confirms that caches are isolated by organization and, on supported platforms, by workspace. Cached representations remain in memory rather than persistent storage. Options A and C may omit context through summarization or retrieval and therefore require additional quality evaluation. Option D lowers unit cost but may reduce reasoning quality. The design must preserve prefix identity, choose an appropriate TTL, and verify the tenancy configuration.

Study Guide references/topics: Prompt caching; cache isolation; stable prefixes; long-document analysis; cost optimization.

Effective documentation begins with audience analysis and structural planning. Identifying the audience determines the required technical depth, terminology, decision context, and operational emphasis. Defining the questions the guide must answer prevents sections from becoming generic descriptions. The outline then establishes complete coverage across architecture overview, component responsibilities, interface contracts, dominant flows, operational runbooks, and known limitations. Translation occurs only after stable source content exists. Implementation-team validation also follows an initial draft because there must be material to review. Version control, reviewers, and cadence are important governance controls, but they do not replace the two content-planning activities required before section drafting. The guide should be planned around the decisions and tasks its intended readers must perform.

Option C preserves both transparency and stakeholder agency. The architect acknowledges the underlying business need while making the delivery consequences explicit. Presenting controlled alternatives converts an unbounded request into a governed decision: defer the capability, exchange it for existing scope, or accept a later release date. The selected option should be documented with revised success criteria, dependencies, risk, cost, ownership, and approval. Concealing the request prevents accurate planning and creates unmanaged delivery risk. Rejecting it without exploring alternatives ignores legitimate stakeholder value. Quietly absorbing additional work produces scope creep and makes later delays appear unexpected. Effective lifecycle communication does not promise every requested capability; it presents the trade-offs necessary for an informed decision.

Study Guide references/topics: Scope governance; expectation alignment; transparent trade-offs; change control; stakeholder decision rights.

Haiku is the appropriate starting point because the workload is high-volume, latency-sensitive, and based on a stable closed set of moderation labels. These characteristics favor a fast, cost-efficient model capable of consistent classification without incurring the additional inference time and expense associated with deeper reasoning.

Anthropic’s model-selection guidance requires architects to balance capability, speed, and cost rather than automatically selecting the most capable model. Its content-moderation guidance specifically identifies Haiku as a cost-effective option for processing moderation workloads at substantial scale. Choosing the Right Model, Content Moderation

Opus is disproportionate to a routine closed-set classification problem. Sonnet may become justified if evaluation demonstrates that Haiku fails materially on complex policy distinctions, multilingual ambiguity, or adversarial edge cases, but it should not be selected merely because it is larger. Enabling extended thinking on every request would further increase latency and token consumption without evidence that the additional reasoning improves the defined success metrics. The correct architectural practice is to establish a representative moderation evaluation set, validate Haiku against accuracy and safety thresholds, and escalate only the cases that genuinely need deeper reasoning.

Study Guide references/topics: Model selection; capability–latency–cost trade-offs; classification workloads; evaluation-driven routing; moderation architecture.

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

The evaluation metrics must map directly to the two stated requirements. Role-based access-control enforcement rate, tested against a red-team dataset, measures whether requests from unauthorized identities, roles, or access paths are consistently denied. It converts the privacy and authorization requirement into an observable security metric and exposes bypass conditions that ordinary functional tests may miss. Because the requirement says patient data must never reach unauthorized roles, the acceptance threshold should be effectively 100% enforcement across representative and adversarial cases.

Per-query token cost directly measures compliance with the financial constraint. The metric should include uncached input, cache reads and writes where applicable, output tokens, thinking tokens, and any other billable components attributed to the request. Results can then be compared with the defined ceiling across normal, boundary, and worst-case inputs.

Anthropic’s evaluation guidance identifies privacy preservation and price as distinct success criteria and emphasizes that criteria must be specific, measurable, and relevant to the application. Define Success Criteria and Build Evaluations

BLEU may assess textual similarity but does not test authorization or cost. Throughput and percentile latency are valuable operational metrics, yet neither directly answers the two requirements in the scenario.

Study Guide references/topics: Multidimensional evaluation; privacy preservation; authorization testing; red-team datasets; token-cost measurement; requirement-to-metric traceability.

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

Least privilege means granting the exact capabilities required for the defined responsibility—no more and no less. Option B permits the assistant to read the approved knowledge base and write only to the draft queue. Enforcing the allow-list at the orchestration layer creates a deterministic authorization boundary outside the model’s prompt and reasoning.

The rule should also be role-specific so different users or operational roles receive only the tools and data scopes they require. The knowledge-base tool should remain read-only, while the draft-queue tool should be restricted to draft creation rather than external sending, approval, or production publication.

Option A exposes every globally readable tool, creating unnecessary access to unrelated information. Option C relies on voluntary model behavior instead of enforceable authorization. A mistaken tool selection or successful prompt injection could therefore invoke an inappropriate capability. Option D minimizes permissions by eliminating the assistant’s required functionality and is not a viable least-privilege implementation.

Anthropic’s permission guidance distinguishes allow, ask, and deny rules and confirms that these controls are enforced by Claude Code rather than by model instructions. Configure Permissions

Study Guide references/topics: Least privilege; per-role authorization; orchestration-layer enforcement; capability scoping; read/write separation; tool allow-lists.

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

Authentication and authorization run before retrieval so retrieval can filter by identity — Yes

An asynchronous queue absorbs bursty traffic between intake and the model-invocation layer — Yes

Tax computation is encoded directly in the system prompt rather than in code — No

Conversation logs include unredacted government identifiers to maximize tuning signal — No

Authentication and authorization must precede retrieval so unauthorized information never enters model context. An asynchronous queue is appropriate for burst absorption, backpressure, controlled concurrency, and retry management at scale. Deterministic tax computation should be implemented in validated code or a controlled calculation tool, not delegated to probabilistic prompt interpretation. Including unredacted government identifiers in conversation logs violates data-minimization principles and creates unnecessary privacy, security, and regulatory exposure. Sensitive fields should be removed, tokenized, or redacted before logging or model use unless specifically required and authorized. The resulting design separates deterministic computation, access control, scalable orchestration, and language reasoning into appropriate architectural layers.

The scenario describes indirect prompt injection: malicious instructions are embedded in third-party content that Claude processes as retrieved evidence. Option A addresses the threat at the appropriate boundary by classifying retrieved material as untrusted, screening it before use, and validating the model’s output before it can influence downstream actions.

Anthropic recommends placing untrusted third-party content in clearly identified tool-result structures, explicitly instructing Claude that retrieved content cannot override system or user instructions, screening tool output for injection attempts, and applying output validation. Mitigate Jailbreaks and Prompt Injections

Citations in Option B improve factual traceability but do not prevent embedded instructions from attempting to redirect model behavior. An outbound destination allow-list in Option C limits the damage that a compromised agent can cause through tool calls, but it does not directly detect or neutralize the injected content. Option D is valuable for detecting behavioral regression after a model change, not for controlling a specific malicious document during runtime.

A production implementation should combine content provenance, structural separation, injection screening, least-privilege tools, output-schema validation, and human confirmation for consequential actions.

Study Guide references/topics: Indirect prompt injection; untrusted retrieved content; input classification; output validation; layered runtime guardrails.

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

A complete handoff package must help different stakeholders understand, operate, troubleshoot, and evolve the Claude-based solution. An architecture overview explains the system’s structure and responsibilities, while Architecture Decision Records document important decisions, alternatives, and trade-offs. Component-level diagrams provide implementation context without forcing maintainers to reconstruct dependencies from source code. Operational runbooks and an on-call playbook enable reliable incident detection, diagnosis, escalation, and recovery. Documenting known limitations is particularly important for Claude applications because model behavior can vary with prompts, context, input distribution, and external dependencies. Options B, C, and D each provide only a partial or temporary understanding of the system. They would create knowledge gaps, increase operational risk, and make future maintenance dependent on individuals rather than durable organizational documentation.

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

Configuration Y is the appropriate recommendation because architectural optimization must follow the deployment’s explicitly defined priorities and acceptance criteria. Accuracy is the primary requirement for this compliance assistant, and Configuration X introduces a measured two-percent accuracy regression. That reduction cannot be dismissed as ordinary benchmark variation unless repeated evaluation establishes that it is statistically insignificant. In a compliance workload, even a small accuracy decline can increase incorrect guidance, missed policy requirements, human-review effort, and regulatory exposure. Configuration Y preserves the established accuracy level while maintaining the current latency profile. Although its latency does not improve, latency is explicitly the secondary objective. Configuration X could be reconsidered only if further testing eliminates the accuracy regression or stakeholders formally approve a revised trade-off. Options C and D incorrectly impose a general preference for latency over the documented deployment priorities.

The request is not yet design-ready because it lacks a defined failure, affected user journey, supported decision, target population, and measurable outcome. The architect should therefore facilitate structured discovery rather than prematurely selecting an AI pattern or model.

Discovery should identify where the current customer journey breaks down, who experiences the problem, what decision or task requires assistance, what data is available, and which metric would demonstrate improvement. The team can then establish baseline performance, target thresholds, operational constraints, risks, and explicit non-goals. Anthropic’s evaluation guidance states that an LLM application should begin with clearly defined success criteria and a method for measuring them. Define Success Criteria and Build Evaluations

Option A abandons the sponsor instead of helping translate a business concern into an actionable problem. Option B risks producing a generic demonstration unrelated to a validated customer need. Option C substitutes the architect’s assumptions for stakeholder evidence and may commit the organization to the wrong scope.

Competent delegation begins by defining the decision, intended outcome, authority boundaries, and target metric before determining which responsibilities should be assigned to Claude, conventional software, or human reviewers.

Study Guide references/topics: Structured discovery; delegation competency; problem definition; target metrics; decision support; measurable business outcomes.

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

An operational SLA must be objectively measurable and enforceable. Option B defines the metric, percentile, threshold, and observation window, allowing both parties to determine whether service performance complied with the agreement. A complete SLA also defines what follows a breach, such as escalation, remediation, service credits, capacity changes, or architectural review. Percentile latency is preferable to an average because it exposes degraded tail performance experienced by a meaningful portion of users. Anthropic provides usage, cost, and operational data that can support monitoring and alerting through the Usage and Cost API. Options A and C are subjective. Option D remains incomplete because neither assessment periods nor breach handling can be applied consistently.

Study Guide references/topics: SLA construction; p95 latency; thresholds; evaluation windows; breach consequences; operational measurement.

Project-scope configuration is appropriate for shared MCP definitions, permissions, hooks, and other repository-specific behavior that every team member should receive consistently. Managed configuration complements this by enforcing organization-wide security and compliance requirements that individual developers cannot override. Together, these mechanisms separate collaborative project configuration from centrally controlled policy. Editor themes are personal preferences and do not materially establish a secure environment. Allowing each engineer to select an arbitrary model can undermine consistency when model choice affects behavior, cost, or approved-use requirements. Storing a shared API key in a committed project file is a serious credential-management failure. Claude Code documentation identifies .claude/settings.json as the shared project file and managed settings as the highest-precedence organizational policy layer. Claude Code: Settings files and precedence

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

The required sequence must reduce repeated computation without changing the information or model capability used to generate projections. Prompt caching the static 6,000-token policy preamble directly addresses the largest cost contributor while preserving the complete instruction set. Anthropic states that cache reads cost substantially less than uncached input tokens, making repeated, stable prompt prefixes an ideal caching target. Prompt Caching

Caching frequently reused historical-sales chunks applies the same principle to the retrieval layer. When identical, version-controlled chunks are repeatedly fetched and supplied to the model, caching their retrieval or reusable representation eliminates redundant work while maintaining the same evidence available to the projection process. The cache must use appropriate invalidation or source-version keys so updated sales data cannot be replaced by stale content.

Options A and B modify the information supplied to Claude. Fewer sales chunks could remove relevant historical evidence, while truncating policy instructions could alter constraints or projection behavior. Option D introduces a model-capability change and therefore cannot guarantee unchanged accuracy without a comparative evaluation. Those interventions may be considered later, but only after representative regression testing establishes equivalence.

Study Guide references/topics: Cost profiling; prompt caching; retrieval caching; cache invalidation; accuracy-preserving optimization; model and context trade-offs.

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