Question 90
A malformed JSON payload prevents an MCP tool handler from parsing the request. What should the server provide?
- Return a structured invalid-arguments failure with an expected-shape hint
- Label the business constraint as nontransient and require a different next step
- Mark partial completion explicitly and identify which data remains unavailable
- Add a safe error category and actionable next-step detail
Explanation
Correct Answer
A:Errors about parsing and schema constraints should identify how the caller can repair the input. Pretending the operation succeeded would poison the workflow with false data.
Incorrect Answers
B:Funds or policy failures generally need user or system-state changes, not immediate identical retries; its benefits for avoiding futile retries on business-rule rejection do not resolve the need for explaining invalid tool arguments.
C:A partial response must not masquerade as complete evidence; supporting communicating incomplete tool output cannot replace the required explaining invalid tool arguments.
D:Structured failure context allows the caller to distinguish bad inputs, authentication problems and outages; this contributes to making tool failure diagnostic rather than opaque without meeting the need for explaining invalid tool arguments.
Question 91
A large project has distinct frontend and database migration conventions. How can instructions be organized without burdening every task?
- Place additional directory-specific CLAUDE.md files near the relevant code
- Keep both frontend and migration conventions in one always-loaded root CLAUDE.md and add explanatory links
- Revise guidance to distinguish tests required by change type and risk
- Record the repository's actual build, lint and test commands with prerequisites
- Record the stable rules once in reviewed project instructions
Explanation
Correct Answer
A:Scoped instruction files can provide rules for particular subtrees. Localizing them reduces clutter and helps agents use conventions relevant to the files in scope.
Incorrect Answers
B:Adding links does not eliminate the irrelevant instructions continually loaded for every task. Localized project guidance should load with the directories where it applies.
C:Persistent rules should guide decisions rather than mandate wasteful work universally; that may serve making standing test instructions proportional, but it is no substitute for localizing instructions to code directories.
D:Executable command knowledge is stable operational context; this may support recording project-specific validation commands but cannot alone meet the requirement for localizing instructions to code directories.
E:Persistent project guidance makes the same expectations available across sessions; the stated problem requires localizing instructions to code directories rather than persisting stable codebase conventions.
Question 92
A rubric classifies security findings but overlabels minor style issues as critical. Which examples would best correct it?
- Choose examples representing distinct layouts and difficult ambiguity cases
- Choose varied fictional entities and emphasize that examples illustrate structure, not actual values
- Provide a missing-field example using the schema's defined null behavior
- Include contrasting high-severity and low-severity labeled cases with rationales
Explanation
Correct Answer
D:Contrastive examples teach the decision boundary between categories. Repeating only severe defects gives little information about false positives.
Incorrect Answers
A:Example diversity matters more than sheer count; the deciding condition is distinguishing severity classes with contrastive examples, whereas this action concerns choosing representative examples rather than redundant ones.
B:Reusing the same names across all demonstrations can invite copying; the option would serve preventing sample values from leaking into outputs, not distinguishing severity classes with contrastive examples as specified in the stem.
C:A deliberate absent-value demonstration discourages invented totals; this addresses demonstrating missing-field semantics, while the question is about distinguishing severity classes with contrastive examples.
Question 93
A long agent run keeps requerying the same unchanged policy document. What context mechanism helps?
- Summarize earlier turns into a verified summary that keeps decisions and open actions
- Keep approval identity, timestamp and supporting policy evidence
- Retain a concise referenced policy finding and its source version
- Provide the minimum relevant schema and task constraints to that specialist
- Document policy version and recheck rules that govern pending decisions
Explanation
Correct Answer
C:Cached validated findings reduce repetitive reads while retaining provenance. A version marker makes it possible to refresh if the source policy changes.
Incorrect Answers
A:When addressing retaining critical state while compacting history, context compaction reduces token load while retaining essential state; the present task instead calls for avoiding repeated retrieval of unchanged evidence.
B:High-impact decisions require an audit trail, not just a terse conclusion; preserving authorization evidence through summarization and avoiding repeated retrieval of unchanged evidence require different controls in this situation.
D:Context selection reduces noise and accidental disclosure; in this scenario, minimizing context passed to specialist workers is secondary to avoiding repeated retrieval of unchanged evidence.
E:When considering detecting stale policy within condensed state, summaries are snapshots rather than proofs of current policy; the requirement for avoiding repeated retrieval of unchanged evidence remains unmet.
Question 94
Frontend files require one lint convention while backend files use another. Where should the conditional rule live?
- Refine the rule's paths glob to target only migration files
- Define path-scoped rules under .claude/rules with file-pattern applicability
- Partition language-specific guidance into separate path-targeted rule files
- Limit the review instruction to schema and migration paths
Explanation
Correct Answer
B:Path-conditioned repository guidance applies specialized instructions only to matching files. This avoids forcing frontend conventions on unrelated backend edits.
Incorrect Answers
A:A narrower glob controls the rule's applicability; this can serve narrowing a conditional rule's file scope, but leaves the requirement for applying code-style rules by path unmet.
C:Distinct rules with distinct globs avoid contradictory language conventions; this concerns separating language-specific rules, leaving the requirement for applying code-style rules by path outstanding.
D:Targeted rules make special approval requirements visible when the affected files warrant them; targeting schema-specific review obligations and applying code-style rules by path are distinct responsibilities under these constraints.
Question 95
A search tool returns a large opaque string containing hundreds of records, overwhelming the agent's context. What output redesign helps?
- Supply each tool a distinct action-focused name and usage description
- Describe the mutation and external communication effect explicitly in the tool description
- Return structured, bounded results with pagination and relevant fields
- Partition the overloaded interface into cohesive operations with narrow schemas
Explanation
Correct Answer
C:Typed and paginated results let the agent request evidence progressively. Huge unstructured dumps increase cost and make source-to-claim mapping difficult.
Incorrect Answers
A:Discriminating names and descriptions specify the business intent and side effects of each operation; supporting differentiating similar tool responsibilities cannot replace the required controlling the volume of tool output.
B:Callers need to distinguish observation from state change before choosing tools; this contributes to signaling irreversible tool side effects without meeting the need for controlling the volume of tool output.
D:Because smaller purpose-built tools make argument requirements and authorization easier to reason about, the measure supports reducing an overloaded tool schema rather than the needed controlling the volume of tool output.
Question 96
A meeting summarizer must omit confidential identifiers, preserve original spelling for proper names, and attribute disputed decisions to their actual speakers. Which THREE prompt instructions address all the requirements? Choose THREE.
- Attribute disputed statements and decisions to the actual meeting speakers
- Preserve proper-name spelling while normalizing only fields specifically designated for normalization
- Spell out which confidential identifiers must be redacted and which summary fields are allowed
- Explain task scope, stopping conditions and prohibited side quests
Explanation
Correct Answers
A:This choice addresses preserving attribution in generated meeting summaries: attribution is part of factual correctness; an explicit decision ledger prevents the output from treating debate as unanimous agreement.
B:The design must include preserving exact source spelling where required: field-level instructions distinguish transcription from interpretation; unnecessary translation can break matching against source systems that store original forms.
C:Data minimization instructions make the output contract sensitive to privacy. The prompt should state which identifiers to omit instead of leaving confidentiality to a generic politeness rule.
Incorrect Answers
D:Bounded instructions tell the agent what enough evidence looks like; a method suited to defining a stopping condition for research does not meet the need for minimizing sensitive data in summaries.
Question 97
A company runs a remote MCP service over an insecure connection on a public network. What deployment correction matters?
- Specify a remote HTTP-based MCP endpoint with appropriate authentication
- Verify a reviewed project .mcp.json definition into version control
- Protect the transport with TLS and authenticate clients before sensitive tool calls
- Choose supported authenticated transport and credential renewal rather than embedding a fixed secret
Explanation
Correct Answer
C:Tool inputs and results may contain confidential operational data. Secure transport and identity checks are necessary controls for network-exposed integrations.
Incorrect Answers
A:Networked MCP transport supports a centrally hosted integration across clients; this addresses connecting to a shared remote MCP service, while the question is about protecting an MCP network connection.
B:Project-scoped MCP configuration can be distributed with the repository; that relates to sharing an MCP server configuration with a project, not the necessary protecting an MCP network connection.
D:Remote MCP services need a secure authentication lifecycle; the deciding condition is protecting an MCP network connection, whereas this action concerns maintaining authenticated MCP transport.
Question 98
Claude Code's long task summary lists implementation ideas but omits which tests actually ran. What should checkpointing preserve?
- Record tested commands, outcomes and affected paths alongside decisions
- Find generation sources and edit the generator or schema instead of generated output
- Examine dependency manifests and lockfiles to confirm the affected version
- Find references and consumer contracts before changing the shared definition
Explanation
Correct Answer
A:Implementation history is not reliable if verification evidence is lost. Explicit test status lets the next session continue without assuming unfinished checks passed.
Incorrect Answers
B:Generated artifacts are often overwritten in the next build; in this scenario, distinguishing generated files from authoritative sources is secondary to preserving code-testing evidence in work summaries.
C:When considering grounding dependency remediation in repository metadata, current local package metadata grounds the remediation plan; the requirement for preserving code-testing evidence in work summaries remains unmet.
D:When addressing mapping consumers of a shared interface, shared APIs can affect distant packages; the present task instead calls for preserving code-testing evidence in work summaries.
Question 99
A developer creates a very complex JSON Schema with unsupported constraints and receives an API error. What is the best correction?
- Employ a finite enumeration for the allowed status values
- Mandate an actual authorized tool invocation and process its result through the dispatcher
- Simplify the schema to supported constructs and validate remaining rules in application code
- Ask for schema-constrained JSON output through the supported structured-output interface
Explanation
Correct Answer
C:Constrained decoding has a documented supported JSON Schema subset. Unsupported keywords should be enforced after parsing rather than assumed to be honored silently.
Incorrect Answers
A:Enumerated values constrain the output vocabulary and simplify downstream branching; this concerns restricting status output to known categories, leaving the requirement for respecting structured-output schema limitations outstanding.
B:A typed JSON response is not a network operation; distinguishing actual tool execution from JSON text and respecting structured-output schema limitations are distinct responsibilities under these constraints.
D:Constrained JSON generation can enforce a declared schema for supported responses, avoiding brittle scraping of prose or fences; this can serve producing machine-parseable schema-conformant JSON, but leaves the requirement for respecting structured-output schema limitations unmet.
Question 100
A human reviewer is inundated with thousands of low-risk extraction exceptions. What routing policy improves scalability?
- Mandate evidence and calibration rather than trusting the self-reported confidence number
- Present policy basis, supporting records, and the requested action before approval
- Hold back from automatic commitment and send the exception to the authorized owner
- Prioritize by material impact, uncertainty and business risk
- Keep individual rationales, evidence and the final adjudication basis
Explanation
Correct Answer
D:Triage uses the cost of errors to allocate scarce review capacity. Not all missing fields need the same urgency or reviewer expertise.
Incorrect Answers
A:A model's expressed confidence may not correspond to measured probability; that serves avoiding uncalibrated confidence as proof, whereas the case requires directing human attention to the highest-risk cases.
B:Because meaningful approval requires adequate context to judge eligibility and consequences, the measure supports giving human reviewers sufficient decision evidence rather than the needed directing human attention to the highest-risk cases.
C:Policy ambiguity is not resolved by fluent reasoning; its benefits for routing uncertain policy exceptions to authorized review do not resolve the need for directing human attention to the highest-risk cases.
E:Disagreement may expose ambiguous policies or missing facts; this contributes to retaining review disagreement and adjudication provenance without meeting the need for directing human attention to the highest-risk cases.