Cisco 810-110 Practice Test 90
Question 890
A team wants fast quality checks before manually reviewing a model patch. Select TWO statements that correctly combine the Static-analysis and lint-assisted review method with evidence or its governing principle.
- Findings are reviewed and policy-breaking defects are blocked.
- Run compiler, lint, type, and security scans on generated code.
- Redact credentials and use structured allowlisted log fields with correlation IDs.
- Compare before-and-after tests, security cases, and performance baselines.
- Assume code formatting alone proves type and security correctness.
Explanation
Correct Answers
A:Findings are reviewed and policy-breaking defects are blocked. This is directly observable validation of the Static-analysis and lint-assisted review result.
B:Run compiler, lint, type, and security scans on generated code. This is the intended action for Static-analysis and lint-assisted review.
Incorrect Answers
C:Redact credentials and use structured allowlisted log fields with correlation IDs. This serves the different Safe structured logging in generated code objective, not the selected Static-analysis and lint-assisted review plan. Static tools identify classes of errors that fluent model output may hide.
D:Compare before-and-after tests, security cases, and performance baselines. This serves the different Regression testing AI-proposed refactors objective, not the selected Static-analysis and lint-assisted review plan. Static tools identify classes of errors that fluent model output may hide.
E:Assume code formatting alone proves type and security correctness. This is not part of a safe or effective Static-analysis and lint-assisted review design. Skipping all automated checks allows obvious defects to survive review.
Question 891
An agent drafts a refund request but has not yet received customer-service approval. If the relevant Human approval before financial side effects safeguard is omitted, which outcome is the MOST directly associated failure?
- A non-interruptible agent can continue harmful side effects after failure is known.
- Replaying all commands from scratch can duplicate orders or updates.
- Automatically treating every reviewer note as new global instructions may destabilize behavior.
- Automatically executing a model-suggested refund can exceed delegated authority.
Explanation
Correct Answer
D:Automatically executing a model-suggested refund can exceed delegated authority. Human-in-the-loop controls reserve consequential decisions for responsible people. The corrective approach is to require explicit approver consent and preserve the audit trail.
Incorrect Answers
A:A non-interruptible agent can continue harmful side effects after failure is known. This failure pattern belongs chiefly to Human override and reversible control; it is not the failure most directly tested in the Human approval before financial side effects situation. Automatically executing a model-suggested refund can exceed delegated authority.
B:Replaying all commands from scratch can duplicate orders or updates. This failure pattern belongs chiefly to Durable agent checkpoints and resume; it is not the failure most directly tested in the Human approval before financial side effects situation. Automatically executing a model-suggested refund can exceed delegated authority.
C:Automatically treating every reviewer note as new global instructions may destabilize behavior. This failure pattern belongs chiefly to Feedback from reviewers without unapproved learning; it is not the failure most directly tested in the Human approval before financial side effects situation. Automatically executing a model-suggested refund can exceed delegated authority.
Question 892
A function should transform a list of prices according to written business rules. After the recommended Test-driven AI code generation change is implemented, which observation MOST directly verifies that it works?
- Requests and responses match the contract including error cases.
- Stakeholders confirm the stories cover their real needs and constraints.
- Generated code meets the test cases including boundaries and invalid inputs.
- Injection payloads are handled as values and unauthorized tables remain inaccessible.
Explanation
Correct Answer
C:Generated code meets the test cases including boundaries and invalid inputs. This directly tests whether the Test-driven AI code generation goal was met. Tests translate requirements into observable behavior for rapid prototyping.
Incorrect Answers
A:Requests and responses match the contract including error cases. This is evidence about API scaffolding with contract-first tests, not direct proof of the required Test-driven AI code generation behavior. The relevant check is that generated code meets the test cases including boundaries and invalid inputs.
B:Stakeholders confirm the stories cover their real needs and constraints. This is evidence about AI-assisted requirements elicitation, not direct proof of the required Test-driven AI code generation behavior. The relevant check is that generated code meets the test cases including boundaries and invalid inputs.
D:Injection payloads are handled as values and unauthorized tables remain inaccessible. This is evidence about Secure generated SQL queries, not direct proof of the required Test-driven AI code generation behavior. The relevant check is that generated code meets the test cases including boundaries and invalid inputs.
Question 893
A coding assistant needs a read-only configuration document exposed by a connected server. If the relevant MCP resource context primitive safeguard is omitted, which outcome is the MOST directly associated failure?
- Treating third-party prompt content as overriding host security rules risks instruction hijacking.
- Treating resource text as developer-level commands can cause prompt-injection failures.
- Using a shared high-privilege token for all users enables cross-user abuse.
- Giving free-form planning unrestricted tool credentials creates avoidable side-effect risk.
Explanation
Correct Answer
B:Treating resource text as developer-level commands can cause prompt-injection failures. MCP resources expose contextual data for clients to retrieve. The corrective approach is to use resource reads for context and maintain trust boundaries.
Incorrect Answers
A:Treating third-party prompt content as overriding host security rules risks instruction hijacking. This failure pattern belongs chiefly to MCP reusable prompt primitive; it is not the failure most directly tested in the MCP resource context primitive situation. Treating resource text as developer-level commands can cause prompt-injection failures.
C:Using a shared high-privilege token for all users enables cross-user abuse. This failure pattern belongs chiefly to MCP authorization and least privilege; it is not the failure most directly tested in the MCP resource context primitive situation. Treating resource text as developer-level commands can cause prompt-injection failures.
D:Giving free-form planning unrestricted tool credentials creates avoidable side-effect risk. This failure pattern belongs chiefly to Planner-executor role separation; it is not the failure most directly tested in the MCP resource context primitive situation. Treating resource text as developer-level commands can cause prompt-injection failures.
Question 894
A research tool returns results with hidden directions to email private conversation data. Testing exposes this failure: Promoting page content into operating instructions can hijack a research assistant. Which remediation should be applied FIRST?
- Isolate retrieved content and constrain side-effect tools to user-authorized tasks.
- Execute scripts or email instructions discovered inside a retrieved webpage.
- Treat user-provided role claims as lower-trust content and enforce application rules.
- Parse tool results as data and validate them against an expected schema and trust boundary.
Explanation
Correct Answer
A:Isolate retrieved content and constrain side-effect tools to user-authorized tasks. Confirmation requires seeing that the assistant answers the user's question without executing page-authored commands.
Incorrect Answers
B:Execute scripts or email instructions discovered inside a retrieved webpage. This retains the Indirect injection in retrieved webpages problem. Indirect prompt injection hides control attempts inside lower-trust material.
C:Treat user-provided role claims as lower-trust content and enforce application rules. It addresses Direct user instruction override attempt, not the diagnosed Indirect injection in retrieved webpages failure. Promoting page content into operating instructions can hijack a research assistant.
D:Parse tool results as data and validate them against an expected schema and trust boundary. It addresses Malicious instructions inside tool output, not the diagnosed Indirect injection in retrieved webpages failure. Promoting page content into operating instructions can hijack a research assistant.
Question 895
A team reduces prompt size by retaining only a short narrative of prior decisions. An engineer must explain the recommendation for Compression and summary quality risks. Which underlying principle is the MOST accurate?
- Compression saves context but can discard operationally critical facts.
- Token limits can truncate syntactically correct-looking partial responses.
- Selective context improves both signal-to-noise ratio and token use.
- AI may help design tests while execution establishes observed behavior.
Explanation
Correct Answer
A:Compression saves context but can discard operationally critical facts. Therefore the implementation decision is to preserve exact interfaces, constraints, and tests in a compact structured state.
Incorrect Answers
B:Overaggressive summaries can silently erase parameters needed for safe changes. This describes the separate Detecting output truncation mechanism; the present case instead requires Compression and summary quality risks.
C:Overaggressive summaries can silently erase parameters needed for safe changes. This describes the separate Repository chunk selection mechanism; the present case instead requires Compression and summary quality risks.
D:Overaggressive summaries can silently erase parameters needed for safe changes. AI may help design tests while execution establishes observed behavior.
Question 896
A research team wants to rerun AI-supported exploratory calculations. Which practical correction or implementation decision is MOST directly appropriate?
- Record data lineage, generated code, and validation steps.
- Publish a graph without retaining the input dataset or aggregation logic.
- Confirm flagged records and evaluate whether anomalies reflect data errors or real events.
- Compute group-specific statistics and verify comparability.
Explanation
Correct Answer
A:Record data lineage, generated code, and validation steps. The intended behavior can be checked because a second analyst reproduces the same summary from the recorded inputs.
Incorrect Answers
B:Publish a graph without retaining the input dataset or aggregation logic. This would not correct the Reproducible EDA with provenance workflow. Untracked edits can make visually convincing plots impossible to audit.
C:Confirm flagged records and evaluate whether anomalies reflect data errors or real events. This could help Exploratory data anomaly identification, but it does not resolve the Reproducible EDA with provenance case. Reproducibility links exploratory claims to data and deterministic preparation steps.
D:Compute group-specific statistics and verify comparability. This could help Segmentation of exploratory datasets, but it does not resolve the Reproducible EDA with provenance case. Reproducibility links exploratory claims to data and deterministic preparation steps.
Question 897
A generator includes accidental extra commentary after a required final delimiter. Which approach BEST meets the requirement while respecting the system constraints?
- Use a shorter stop delimiter and treat every early ending as a finished answer.
- Configure supported max-output and stop conditions and validate response termination.
- Measure combined input context tokens and trim or retrieve content within the model limit.
- Keep a verified compact state summary plus recent turns and retrievable references.
Explanation
Correct Answer
B:Configure supported max-output and stop conditions and validate response termination. Termination controls bound output but do not replace semantic completeness checks. Verification should demonstrate that responses stop at expected boundaries and report explicit truncation status.
Incorrect Answers
A:Use a shorter stop delimiter and treat every early ending as a finished answer. This is an unsafe or ineffective shortcut for Stop conditions and maximum output. Blindly assuming a stop token proves the requested content is complete loses required fields.
C:Measure combined input context tokens and trim or retrieve content within the model limit. This primarily addresses Context window truncation, not the specific Stop conditions and maximum output decision. Here, termination controls bound output but do not replace semantic completeness checks.
D:Keep a verified compact state summary plus recent turns and retrievable references. This primarily addresses Conversation memory summarization, not the specific Stop conditions and maximum output decision. Here, termination controls bound output but do not replace semantic completeness checks.
Question 898
A team compares sentiment tagging with generating a new customer explanation. Which practical correction or implementation decision is MOST directly appropriate?
- Supply the actual image to a vision-capable model and verify visible claims.
- Validate generated source with tests, linting, and human review.
- Treat a binary label score as a complete personalized email.
- Match the model family to whether the task needs creation or classification.
Explanation
Correct Answer
D:Match the model family to whether the task needs creation or classification. The intended behavior can be checked because the chosen system returns the intended free-form draft or bounded label.
Incorrect Answers
A:Supply the actual image to a vision-capable model and verify visible claims. This could help Multimodal visual question answering, but it does not resolve the Conditional generation versus classification case. Generation produces new content while classification selects among predefined categories.
B:Validate generated source with tests, linting, and human review. This could help Code-generation model with execution tests, but it does not resolve the Conditional generation versus classification case. Generation produces new content while classification selects among predefined categories.
C:Treat a binary label score as a complete personalized email. This would not correct the Conditional generation versus classification workflow. Using a category classifier alone cannot satisfy a multi-paragraph drafting task.
Question 899
A support agent needs grounded answers from approved recent knowledge instead of memory alone. Which practical correction or implementation decision is MOST directly appropriate?
- Implement change-triggered re-indexing and deletion propagation.
- Build retrieval with scoped context and citation verification.
- Replace the knowledge index with a model prompt saying it knows all current policies.
- Tune chunking using retrieval evaluation and token budgets.
Explanation
Correct Answer
B:Build retrieval with scoped context and citation verification. The intended behavior can be checked because an answer points to supporting document passages and avoids unsupported policy claims.
Incorrect Answers
A:Implement change-triggered re-indexing and deletion propagation. This could help Retrieval index freshness, but it does not resolve the Retrieval-augmented question answering case. RAG brings external evidence into a model context without requiring weight retraining.
C:Replace the knowledge index with a model prompt saying it knows all current policies. This would not correct the Retrieval-augmented question answering workflow. Using only static model memory for changing policies can hallucinate obsolete guidance.
D:Tune chunking using retrieval evaluation and token budgets. This could help Chunking and overlap for document retrieval, but it does not resolve the Retrieval-augmented question answering case. RAG brings external evidence into a model context without requiring weight retraining.
Question 900
A scheduling workflow must compare local display times with UTC timestamps. An engineer must explain the recommendation for Date and timezone normalization. Which underlying principle is the MOST accurate?
- Schema mapping translates between service contracts and preserves meaning.
- Data mapping should preserve identifiers needed for idempotent processing.
- Timezone conversion requires explicit offsets and daylight-saving handling.
- Persistent state and idempotency preserve progress across interruptions.
Explanation
Correct Answer
C:Timezone conversion requires explicit offsets and daylight-saving handling. Therefore the implementation decision is to convert times to an agreed timezone and preserve original offsets and meaning.
Incorrect Answers
A:Treating a local time string as UTC can schedule the wrong appointment. This describes the separate Mapping agent tool output to a target schema mechanism; the present case instead requires Date and timezone normalization.
B:Treating a local time string as UTC can schedule the wrong appointment. This describes the separate Idempotent event identity across agents mechanism; the present case instead requires Date and timezone normalization.
D:Treating a local time string as UTC can schedule the wrong appointment. This describes the separate Durable agent checkpoints and resume mechanism; the present case instead requires Date and timezone normalization.