Prévia do material em texto
https://www.passcert.com/CCAR-P.html Page 2 18 questions selected from source version V9.02 CLAUDE CERTIFIED ARCHITECT Question 1 HOTSPOT You are evaluating prompting claims in a peer's design document. For each claim, select yes if the claim reflects sound practice. Otherwise, select no. Answer: https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 3 CLAUDE CERTIFIED ARCHITECT Question 2 You are identifying signals that a deployment should re-enter design rather than continue iterating in place. Which signal most directly indicates the need for a new design cycle? A. A runbook step requires a clarification edit to improve on-call guidance accuracy, which can be handled as a documentation update without changes to component responsibilities or core contracts. B. A dashboard alert threshold needs a small numerical adjustment to reduce false-positive noise, which can be handled as an operational configuration change without a new design cycle. C. A minor copy edit is requested in a customer-facing string within the existing UI, which can be handled as a localized content change without altering component responsibilities or contracts. D. The system's current architecture cannot meet the new requirements without changes to component responsibilities or core contracts. Answer: D Explanation A new design cycle is warranted when the required change crosses an architectural boundary. Option D explicitly states that the existing architecture cannot satisfy the new requirements without redistributing component responsibilities or altering core contracts. Such changes may affect service ownership, orchestration, data flows, integration interfaces, security boundaries, failure handling, evaluation strategy, and operational accountability. They therefore require renewed discovery, impact analysis, design review, stakeholder approval, and regression planning. Options A, B, and C are localized maintenance activities. Clarifying a runbook improves operational documentation but does not change the deployed system's structure. Adjusting an alert threshold is a controlled operational tuning activity, provided the change remains within established monitoring policy. Editing customer-facing copy is similarly confined to the presentation or content layer. None of these changes inherently invalidates component contracts or architectural assumptions. An architect should distinguish between iteration within an approved design and evidence that the design itself no longer supports the required outcome. Re-entering design for every minor adjustment creates unnecessary governance overhead; continuing local iteration after architectural assumptions have failed creates unmanaged technical and operational risk. Option D is the only signal that establishes a structural incompatibility. Study Guide references/topics: Lifecycle feedback loops; design re-entry criteria; architectural significance; component responsibilities; interface contracts; controlled operational iteration. CLAUDE CERTIFIED ARCHITECT Question 3 You are integrating Claude Code into a workflow that runs against a production database. Which guardrail design most directly preserves safety on data-modifying operations? A. Allow Claude Code to write directly to the production database without subagent scoping, read-only credential defaults, or human confirmation gates on data-modifying operations. B. Configure the database MCP server with a fully privileged credential that can perform any read or write operation, and allow all operations to proceed without explicit human confirmation. C. Disable all logging and auditing on database operations through the MCP server to reduce alert noise, removing the observability needed to detect unintended data modifications. D. Configure the database MCP server with a read-only credential by default, restrict the subagent's tool list to read-only operations, and require explicit human confirmation on any operation that would modify data. Answer: D Explanation https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 4 Question 3 continued 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. CLAUDE CERTIFIED ARCHITECT Question 4 A Claude architect is auditing configuration scope assignments. Which two statements correctly identify an appropriate use of user-scope configuration versus other scopes? (Select two.) A. Persisting personal editor theme preferences that follow an engineer across projects. B. Saving a preferred Claude response language that applies to all repositories the engineer uses. C. Enforcing a company-wide policy that disables a feature for all engineers. D. Defining MCP server endpoints shared by all contributors to a specific repository. E. Storing API authentication keys so they are not committed to version control. Answer: A, B Explanation 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. https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.htmlPage 5 CLAUDE CERTIFIED ARCHITECT Question 5 During an architectural review, the security team identifies a risk that adversarial content injected into retrieved documents could manipulate the model's behavior. Which mitigation most directly addresses this threat? A. Treat all retrieved content as untrusted input and apply input classifiers with output validation. B. Require citations for each claim and constrain responses to source-supported content. C. Restrict outbound tool calls to an approved destination allow-list. D. Score outputs against a stable adversarial evaluation set on each model-version change. Answer: A Explanation 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. CLAUDE CERTIFIED ARCHITECT Question 6 A Claude architect needs to ensure that a security-hardening flag cannot be disabled by any individual engineer after it is set. Which configuration scope correctly enforces this requirement? A. User scope (~/.claude/settings.json) on each engineer's machine B. Environment variable defined in the CI/CD pipeline only C. Managed configuration applied centrally and marked as non-overridable D. Project scope (.claude/settings.json) committed to the repository Answer: C Explanation Managed configuration is designed for organization-wide security and compliance controls that individual users and repositories must not override. Applying the hardening flag centrally establishes an administrative policy boundary and prevents engineers from disabling the control through user, project, or local settings. Anthropic identifies managed scope as the appropriate location for security policies, non-overridable compliance requirements, and standardized configurations deployed by IT or DevOps. Managed values ordinarily take precedence over command-line arguments and every user-controlled settings scope. Claude Code Settings https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 6 Question 6 continued Option A depends on every engineer maintaining the setting and allows the user to edit or delete it. Option B affects only pipeline execution and does not protect local, interactive, IDE, or other execution paths. Option D standardizes the setting within the repository but does not make it tamper-resistant; contributors with repository write access could alter the file, use higher-precedence local settings where permitted, or operate outside the repository configuration. The implementation should also verify active policy delivery, monitor configuration-change events, and test that startup or execution fails safely if the managed setting is absent or invalid. Central definition without enforcement and verification would not fully satisfy the requirement. Study Guide references/topics: Managed settings; non-overridable controls; enterprise hardening; policy enforcement; configuration governance; defense against local override. CLAUDE CERTIFIED ARCHITECT Question 7 You must present an architectural recommendation to deploy a Claude-based contract review assistant to a steering committee that includes the CFO, the general counsel, and the CIO. Each stakeholder cares about different aspects of the decision. How should you structure the recommendation document? A. Lead with the technical architecture diagram and the full component list before any other section. B. Lead with detailed cost projections across the full multi-year horizon before the rationale section. C. Present the same dense narrative throughout with no stakeholder differentiation in any section. D. Lead with the architectural decision, then address each stakeholder's primary concerns directly. Answer: D Explanation A steering committee needs an answer-first recommendation that states the proposed architectural decision, the business outcome it supports, and the approval or action required. After establishing that common decision frame, the document should address each stakeholder's concerns separately. The CFO requires implementation and operating costs, expected savings, consumption assumptions, financial exposure, and cost-control mechanisms. General counsel needs evidence concerning confidentiality, privilege, data handling, human legal review, explainability, retention, and regulatory risk. The CIO needs integration architecture, security boundaries, availability, scalability, operational ownership, technical dependencies, and lifecycle support. Option A begins at an implementation depth unsuitable for the entire committee. Option B overweights the CFO's concerns and delays the rationale required by the other decision-makers. Option C ignores audience-specific responsibilities and forces every stakeholder to extract relevant information from an undifferentiated narrative. The recommended structure is: decision and executive rationale, expected outcomes, accepted trade-offs, stakeholder-specific impact sections, major risks and controls, implementation implications, success metrics, and the requested governance decision. Technical appendices can preserve component-level detail without overwhelming the principal recommendation. Study Guide references/topics: Executive communication; stakeholder differentiation; decision-first structure; financial, legal, and technology concerns; architectural recommendations. https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 7 CLAUDE CERTIFIED ARCHITECT Question 8 You are responding to an adversarial input pattern in which users include text claiming admin authority and instructing the model to bypass safety restrictions. Which combination of controls most effectively mitigates this attack pattern? A. Trusting that the model will intrinsically recognize and reject all bypass attempts without prompt-level instructions, runtime classifiers, scoped permissions, or audit logging. B. Prompt-level instructions that treat user content as untrusted data, runtime classifiers that detect override attempts, scoped tool permissions that cannot be elevated by user content, and audit logging of attempts. C. Removing all safety restrictions and guardrails to eliminate the attack surface that bypass attempts target, accepting that this makes the assistant unrestricted for all inputs. D. Granting users any privilege level they assert in their message content, on the assumption that cooperative behavior requires honoring self-declared authority without independent verification. Answer: B Explanation Self-declared administrative authority inside a prompt is untrusted data, not authenticatedidentity 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. CLAUDE CERTIFIED ARCHITECT Question 9 A loan pre-qualification assistant shows 94 percent approval recommendations that match the human underwriter decision. The fairness team has reviewed approval rate parity across protected groups and reported no significant difference. A board member has asked whether this evidence is sufficient to declare the assistant fair. Which two Discernment-competency findings should you report? Each correct answer presents part of the solution. (Select two.) A. Approval rate parity does not by itself assess error rate parity across protected groups. B. Match with human underwriters does not establish freedom from underwriter-introduced bias. C. The 94 percent match rate is sufficient evidence of fairness for the assistant's decisions. D. The fairness team's review process likely missed at least some of the protected groups studied. E. A larger sample is needed before any meaningful fairness claim can be made about the model. https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 8 Question 9 continued Answer: A, B Explanation Approval-rate parity measures whether groups receive positive recommendations at similar rates. It does not establish whether false approvals, false denials, sensitivity, specificity, or calibration are comparable across those groups. A system can therefore satisfy aggregate approval parity while imposing materially different error burdens on protected populations. The evaluation must include subgroup confusion matrices, false-positive and false-negative rates, calibration, intersectional analysis, and confidence intervals. The 94 percent agreement rate measures fidelity to human underwriter decisions, not fairness. Human decisions are not automatically unbiased ground truth. If historical underwriting practices contain structural, procedural, or measurement bias, a model that reproduces those decisions accurately can reproduce the same bias. The reference labels must therefore be independently assessed for legitimacy, consistency, and potential discriminatory effects. Nothing in the scenario establishes that protected groups were omitted, so D is unsupported. Similarly, sample size may require examination, but the scenario provides no statistical information proving that sample size is the principal deficiency. The evidence already contains two identifiable conceptual gaps regardless of sample size. Study Guide references/topics: [Defining multidimensional evaluation criteria] (https://docs.anthropic.com/en/docs/build-with-claude/develop-tests); fairness measurement; subgroup error analysis; label and benchmark bias; human-baseline limitations; governance evidence. CLAUDE CERTIFIED ARCHITECT Question 10 You are integrating AI-assisted tooling into the team's documentation workflow. The team wants generated documentation that stays grounded in the actual code. Which integration approach best fits this requirement? A. Generate documentation from the model's training-data recall without reading any of the actual repository code, accepting that the output will not reflect the current implementation. B. Have the subagents publish generated documentation directly to the public-facing site without passing through the team's normal review workflow or any human approval step. C. Configure subagents that read the relevant code files via filesystem and code-search tools, generate the documentation, and emit changes through the team's normal review workflow. D. Disable all filesystem and code-search tools so the subagents cannot read any repository code, accepting that documentation generation will be entirely disconnected from the actual implementation. Answer: C Explanation Documentation remains grounded only when the generating agent can inspect the authoritative implementation. A purpose-built subagent should receive scoped filesystem and code-search capabilities, identify the relevant modules, interfaces, configuration files, tests, and call sites, and generate documentation from that evidence. This approach minimizes reliance on training-data recall, which cannot represent the repository's current private code or recent changes. Subagents are useful because repository exploration occurs in their own context, allowing them to return focused findings or proposed edits without filling the main conversation with every file read. Their tool access should follow least privilege: read access to required paths, narrowly controlled write access, and explicit exclusions for secrets or unrelated repositories. Generated documentation should enter the team's established change-control process as a patch, branch, or pull request. Normal review, automated checks, link validation, example execution, and human approval provide accountability and prevent unverified material from being published. Direct publication bypasses these safeguards, while disabling repository tools makes factual grounding impossible. Study Guide references/topics: [Claude Code custom subagents] (https://docs.anthropic.com/en/docs/claude-code/sub-agents); [codebase exploration workflows] (https://docs.anthropic.com/en/docs/claude-code/common-workflows); source-grounded generation; least-privilege tools; https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 9 Question 10 continued human review and lifecycle controls. CLAUDE CERTIFIED ARCHITECT Question 11 You are rolling out monitoring for a Claude-based deployment and must complete the specification steps before instrumenting the deployment. Which two steps must be completed BEFORE instrumenting the deployment to emit metrics and traces? Each correct answer presents part of the solution. (Select two.) A. Define the metrics and the slices the deployment will monitor across normal and adversarial traffic. B. Tune the alert thresholds based on observed normal-state distributions to reduce false positives. C. Define the service-level objectives (SLOs) and the error budgets the deployment will be held to. D. Build the dashboards that surface metrics across slices at the cadence the team operates on. E. Document the dashboards, alerts, and runbooks for the on-call rotation that will respond. Answer: A, C Explanation Instrumentation must be driven by an explicit measurement specification. The team must first define the metrics and evaluation slices required to detect performance, cost, quality, safety, or reliability failures. Slices should include representative traffic, adversarial inputs, user segments, tools, and relevant content categories.SLOs and error budgets must also be established before instrumentation because they determine required measurements, aggregation windows, and acceptable failure levels. Alert thresholds cannot be tuned until normal-state data has been collected. Dashboards and on-call documentation depend on the final metric definitions and observed operational behavior. Anthropic supports metrics, event, and trace export through Open Telemetry, but telemetry is useful only when tied to measurable success and operating objectives. Claude Code monitoring CLAUDE CERTIFIED ARCHITECT Question 12 You are assessing a Claude-based system whose dominant risk is silent quality drift on safety-relevant outputs after a model-version upgrade. Which assessment activity most directly addresses this risk? A. Disable adversarial evaluation entirely during model-version upgrade cycles to reduce evaluation cost, accepting that safety drift will go undetected until it appears in production. B. Rotate adversarial inputs randomly so no two upgrades are scored on the same set. C. Skip evaluation on each model-version upgrade and rely on user-submitted complaints to surface safety drift after the upgraded model has already served production traffic. D. Maintain an adversarial evaluation set with version-attributed scoring so each upgrade is measured against the same set before promotion. Answer: D Explanation 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 https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 10 Question 12 continued completed and recommends maintaining them continuously to detect backsliding. Demystifying evals for AI agents CLAUDE CERTIFIED ARCHITECT Question 13 You are a platform architect designing an internal Claude-based assistant that serves both finance analysts and external auditors. Each population must access only documents permitted by its role. Where should role-based access control be enforced in the pipeline? A. Inside the system prompt as a natural-language instruction for Claude to ignore unauthorized documents. B. At the retrieval layer, before any role-restricted content reaches the prompt-construction step or the model. C. Nowhere in the pipeline; rely on the model's general refusal behavior to reject unauthorized document access without any enforced access control. D. After the response is generated, by post-filtering content that should not have been retrieved. Answer: B Explanation Authorization must be enforced before restricted information enters model context. The retrieval service should authenticate the user, resolve role and document permissions, and filter candidate records before prompt construction. This prevents unauthorized content from appearing in prompts, model traces, cached context, tool results, or generated responses. A system-prompt instruction is behavioral guidance, not an enforceable access-control boundary. Post-generation filtering is also inadequate because the model has already received and processed the restricted data, creating exposure through output, logging, or indirect inference. Relying on general refusal behavior provides no deterministic authorization. The correct architecture treats RBAC as an application and data-layer control, with the model receiving only information the requesting principal is entitled to access. Study Guide references/topics: Retrieval-layer RBAC; authorization before inference; least privilege; sensitive-context isolation. CLAUDE CERTIFIED ARCHITECT Question 14 A Claude-based research assistant begins producing responses that confidently contradict its retrieved source documents despite no change to the retrieval pipeline. Which two diagnostic actions most directly identify the root cause of this behavior? (Select two.) A. Increase the context-window size to allow more retrieved chunks per query. B. Switch the retrieval index to a denser embedding model to improve chunk-relevance scores. C. Determine whether the failure reproduces on the previous model version to test for a model mismatch. D. Reduce the temperature setting to lower response variance across all query types. E. Inspect the system-prompt grounding instructions to determine whether citation constraints remain intact. Answer: C, E Explanation 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 https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 11 Question 14 continued before applying a fix. Define success criteria and evaluations CLAUDE CERTIFIED ARCHITECT Question 15 You are reviewing an integration specification for security gaps. Which two findings constitute valid security gaps in the specification? Each correct answer presents a complete solution. (Select two.) A. Tool calls execute server-side under a least-privilege service principal scoped to the requested action. B. Service credentials are placed in the prompt context, where they can leak into logs and traces. C. Role-based access control is enforced only at the response-rendering layer after the model accesses restricted data. D. Per-user OAuth tokens are exchanged with scope-restricted permissions and refreshed within the active session. E. Tool inputs and outputs are encrypted in transit using transport-layer security between services. Answer: B, C Explanation Credentials must never be supplied as prompt content. Prompts can be retained in traces, debugging output, caches, or logs, and model output may inadvertently expose included secrets. Credentials should instead be resolved server-side through an approved secret store or delegated authentication mechanism. Option C is also a serious gap because post-response RBAC occurs after unauthorized data has already entered model context. Authorization must be enforced before retrieval and prompt construction. Options A, D, and E describe appropriate controls: least-privilege service identities, per-user scoped OAuth, and encrypted transport. Anthropic's prompt-leak guidance recommends excluding unnecessary sensitive information, while Claude Code security relies on permission-based enforcement rather than natural-language instructions alone. Reduce prompt leak CLAUDE CERTIFIED ARCHITECT Question 16 HOTSPOT You are classifying token-management tactics by where each tactic applies in the request lifecycle: "Input Preparation," "Prompt Construction," or "Output Handling." https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 12 Question 16 continued Answer: CLAUDE CERTIFIED ARCHITECT Question 17 A platformteam operates a self-hosted multi-agent system on Kubernetes that orchestrates seven specialized agents for invoice processing. The team spends approximately 40 percent of engineering capacity on infrastructure maintenance, message bus reliability, and agent state recovery. The CFO has asked you to evaluate moving to managed agent infrastructure to reclaim engineering capacity. The security officer requires that all customer financial data remain within an approved network boundary. Which factor should most heavily influence your recommendation? A. Whether the current seven agents map cleanly to the patterns supported by managed agents. B. Whether managed agents reduce per-invoice token costs across the existing processing volume. C. Whether managed agents support the current bus topology used by the platform team. D. Whether managed agent data handling satisfies the network boundary required by security. Answer: D Explanation The network-boundary requirement is a mandatory architectural constraint; therefore, Option D must be resolved before evaluating operational savings. Managed infrastructure can reduce responsibility for agent loops, sandboxes, session state, recovery, and orchestration, but those advantages are immaterial if financial data leaves the approved boundary. Anthropic describes Claude Managed Agents as a hosted agent harness and provides configurable data-residency controls. The architect must still validate inference location, session persistence, encryption, sub processors, credential handling, auditability, and deletion behavior against the organization's exact security policy. Options A, B, and C affect migration effort or economics, but they cannot override a non-negotiable data-governance requirement. Study Guide references/topics: Architectural constraints; managed versus self-hosted agents; data boundaries; residency; security-gated decision-making. https://www.passcert.com/CCAR-P.html https://www.passcert.com/CCAR-P.html Page 13 CLAUDE CERTIFIED ARCHITECT Question 18 You are preparing an operational runbook for a Claude-based service. Which content is essential to include in the runbook? A. Dashboard and log references only, without alert definitions, triage steps, escalation paths, or rollback procedures for the on-call engineer to act on. B. Common alerts and their triage steps, escalation paths, rollback procedures, and references to the relevant dashboards and logs. C. Alert definitions and triage steps only, without escalation paths, rollback procedures, or references to dashboards and logs for on-call use. D. Escalation paths and rollback procedures only, without alert definitions, triage steps, or dashboard references to guide initial incident response. Answer: B Explanation An actionable runbook must connect detection, diagnosis, containment, recovery, and escalation. Option B supplies that complete operational chain: alerts identify abnormal conditions; triage procedures guide initial investigation; dashboards and logs provide evidence; escalation paths assign authority; and rollback procedures restore a known-safe state. The runbook should cover model errors, latency degradation, rate limits, tool failures, prompt regressions, cost anomalies, data-governance events, and dependency outages. Anthropic's operational materials demonstrate incident workflows in which alerts lead responders to logs and runbooks before corrective action. Options A, C, and D each omit information required for an on-call engineer to progress from notification to controlled resolution. Procedures should be tested through exercises and updated after incidents. Study Guide references/topics: Claude operational and observability patterns; incident response; triage; escalation; rollback; operational readiness. https://www.passcert.com/CCAR-P.html