Anthropic CCDV-F (Claude Certified Developer - Foundations) Practice Test 1

Question 1

An agent proposes terminating an account based on uncertain evidence and the decision carries legal consequences. Which mechanism should the team prioritize?

  1. Inspect outputs before taking business actions
  2. Enforce least-privilege infrastructure permissions
  3. Apply input validation at ingress
  4. Keep a human review gate for high-impact operations

Explanation

Correct Answer

D:

Mandatory human approval separates recommendations from irreversible execution—with this limitation: model confidence scores cannot replace the required authority of a reviewer.

Incorrect Answers

A:

Output checks validate schema, scope, and safety before any backend mutation; the needed capability is different—mandatory human approval separates recommendations from irreversible execution—and post-generation checks cannot repair unauthorized access already performed upstream.

B:

Backend credentials and IAM policies restrict it to the minimum necessary operations, whereas the requirement calls for another behavior: mandatory human approval separates recommendations from irreversible execution; an unrestricted credential remains dangerous regardless of the model's stated intent.

C:

Input controls reject invalid types, excessive sizes, and disallowed patterns before expensive processing; however, the critical mechanism here is that mandatory human approval separates recommendations from irreversible execution, and content screening alone cannot establish the caller's resource permissions.

Question 2

A model answer contains markup copied from a malicious support ticket and will appear in a web interface. What is the most suitable pattern?

  1. Return an explicit unknown for missing evidence
  2. Preserve partial-stream status
  3. Show structured refusals or errors faithfully
  4. Escape text before inserting into HTML

Explanation

Correct Answer

D:

Output encoding ensures untrusted characters display as text rather than executable DOM content, although JSON validation is not a browser security measure.

Incorrect Answers

A:

A nullable or unknown state avoids inventing a plausible but unsupported date; the missing capability is more specific: output encoding ensures untrusted characters display as text rather than executable DOM content; guessing a value would convert absence of evidence into false data.

B:

The UI marks the response incomplete and avoids committing partial structured output; by comparison, the intended solution relies on this: output encoding ensures untrusted characters display as text rather than executable DOM content; a few visible sentences do not establish a complete API completion.

C:

The client handles refusal or error states without fabricating a completed action; the needed capability is different—output encoding ensures untrusted characters display as text rather than executable DOM content—and hiding these states may incorrectly imply the backend executed something.

Question 3

An assistant must summarize a user-provided contract containing quoted instructions addressed to another party. Which engineering practice is the strongest match?

  1. Separate requirements from source data
  2. Specify a strict response schema
  3. Avoid unnecessary verbosity constraints
  4. Use explicit delimiters for source passages

Explanation

Correct Answer

D:

Marking the passage's bounds clarifies which content is data rather than a user command, subject to this requirement: delimiters help interpretation but do not replace tool-level security.

Incorrect Answers

A:

Structural boundaries distinguish data to analyze from commands the assistant should follow; the requirement instead turns on this behavior: marking the passage's bounds clarifies which content is data rather than a user command; external content should not gain authority merely because it appears near instructions.

B:

Schema constraints and subsequent validation improve consistent downstream parsing; this cannot substitute for the behavior that marking the passage's bounds clarifies which content is data rather than a user command, because they cannot guarantee that the values are factually correct.

C:

Output guidance should match the amount of reasoning and explanation the task actually needs; in contrast, marking the passage's bounds clarifies which content is data rather than a user command, while longer responses can add unsupported details rather than increasing precision.

Question 4

An employee needs to work with Claude in an installed personal desktop client. Which production safeguard is most relevant?

  1. Use CLAUDE.md for repository guidance in Claude Code
  2. Use Claude Code for repository-scoped coding assistance
  3. Use a direct API integration for a branded customer feature
  4. Use Claude Desktop for an interactive desktop workspace

Explanation

Correct Answer

D:

A desktop application manages interactive conversations and integrations under that client's configuration, although a hosted API request is not the same user-interface session.

Incorrect Answers

A:

Repository instruction files provide persistent project context to the Claude Code environment; the missing capability is more specific: a desktop application manages interactive conversations and integrations under that client's configuration; they are not automatically loaded into unrelated Messages API clients.

B:

Claude Code operates in a coding environment with its own tools and repository context; by comparison, the intended solution relies on this: a desktop application manages interactive conversations and integrations under that client's configuration; a generic text completion endpoint does not automatically manipulate local files.

C:

The Messages API allows application code to own product behavior and its user experience; the needed capability is different—a desktop application manages interactive conversations and integrations under that client's configuration—and consumer chat sessions are not an app's programmable backend contract.

Question 5

A claims-review chatbot must accept a bulky evidence packet yet still preserve enough context for the current claim. Historic chat turns fill most of its prompt window. Which TWO measures address preflight size estimation and retention of relevant context? Choose TWO.

  1. Estimate token usage with supported token counting
  2. Prune irrelevant historical turns when near capacity
  3. Reserve enough output tokens for the desired result
  4. Use low-temperature settings when supported for consistency
  5. Count the complete request against context limits

Explanation

Correct Answers

A:

Token counting estimates model tokens more accurately than character length or word count; its operational boundary is that different formats and models may tokenize the same text differently; in claims intake, the preflight check precedes both history pruning and generation.

B:

Context selection removes stale content while preserving state and key constraints, subject to this requirement: blind truncation from the end can delete vital recent instructions.

Incorrect Answers

C:

The configured generation maximum must be consistent with the required output length; here, the controlling capability is that token counting estimates model tokens more accurately than character length or word count, but raising it arbitrarily can increase possible cost without improving content accuracy.

D:

Lower randomness can reduce sampling variability in supported configurations; the missing capability is more specific: token counting estimates model tokens more accurately than character length or word count; it does not guarantee deterministic output or eliminate model errors.

E:

The context budget encompasses the relevant input contents and model-specific output allowances; by comparison, the intended solution relies on this: token counting estimates model tokens more accurately than character length or word count; counting only the latest user text understates true token pressure.

Question 6

An orchestrator plans infrastructure changes but wants a separate guarded component to perform approved modifications. Which solution matches the actual constraint?

  1. Use idempotent side-effect handling
  2. Use a summarizing worker for context compression
  3. Delegate quality checking to a critic worker
  4. Assign tool side effects to an execution worker

Explanation

Correct Answer

D:

A tightly permissioned executor maps authorized operations into deterministic tool calls—with this limitation: broad access in the planning component defeats the intended separation of duties.

Incorrect Answers

A:

Stable operation identifiers and deduplication at the action service protect against duplicate writes; the needed capability is different—a tightly permissioned executor maps authorized operations into deterministic tool calls—and retrying a model-generated tool call blindly can repeat the business operation.

B:

A dedicated summarizer condenses retained facts with provenance and open issues for the parent context, whereas the requirement calls for another behavior: a tightly permissioned executor maps authorized operations into deterministic tool calls; compression may omit nuance unless the system checks critical commitments and references.

C:

An evaluator examines the artifact against explicit tests instead of composing the primary answer; however, the critical mechanism here is that a tightly permissioned executor maps authorized operations into deterministic tool calls, and critique must have actionable criteria and should not be equated with guaranteed correctness.

Question 7

A research agent encounters a page that tells it to reveal a deployment secret before answering. Which response is most defensible under these conditions?

  1. Filter outgoing sensitive data at the egress layer
  2. Treat a retrieved webpage as untrusted data
  3. Authorize each requested operation at the resource level
  4. Isolate risky browsing from privileged tools

Explanation

Correct Answer

B:

External content is evidence, not an instruction source allowed to override the application's policies; the important constraint is that prompt labeling helps but tool-side access restrictions must remain enforced.

Incorrect Answers

A:

Deterministic output and action checks block unauthorized exfiltration before it leaves the trusted environment, whereas the requirement calls for another behavior: external content is evidence, not an instruction source allowed to override the application's policies; model self-policing alone cannot guarantee detection of every malicious instruction.

C:

Access policies check whether that principal may perform the exact requested change on the target object; the requirement instead turns on this behavior: external content is evidence, not an instruction source allowed to override the application's policies; successful login alone does not grant permission across all records.

D:

Separate trust zones and scoped workers prevent public content from directly wielding private capabilities; this cannot substitute for the behavior that external content is evidence, not an instruction source allowed to override the application's policies, because isolation fails if credentials or tool results leak back across the boundary.

Question 8

A customer support product is moving to a newly released Claude model but regressions must be contained. Which action most directly closes the gap?

  1. Update cost and context assumptions
  2. Refresh prompt evaluations after upgrade
  3. Run a canary rollout for a new model version
  4. Recheck behavior after version migration

Explanation

Correct Answer

C:

Limited exposure allows observation of quality, refusal, latency, and tool usage before full migration—with this limitation: a model upgrade can affect the entire user experience even when code is unchanged.

Incorrect Answers

A:

Recalculating budgets avoids relying on stale token or feature assumptions, whereas the requirement calls for another behavior: limited exposure allows observation of quality, refusal, latency, and tool usage before full migration; product names alone do not prove pricing or capacity parity.

B:

Re-running the fixed evaluation set detects behavior shifts requiring prompt adjustment; however, the critical mechanism here is that limited exposure allows observation of quality, refusal, latency, and tool usage before full migration, and copying old pass rates forward would misstate new performance.

D:

Regression tests evaluate prompts, tool arguments, latency, and output shape on the new version; this cannot substitute for the behavior that limited exposure allows observation of quality, refusal, latency, and tool usage before full migration, because version changes can affect behavior even when the API request remains syntactically valid.

Question 9

Customer questions include non-ASCII names and multilingual text that should survive transport. Which capability is necessary for this outcome?

  1. Preserve UTF-8 content safely
  2. Escape output for its destination context
  3. Deserialize JSON with expected schema checks
  4. Negotiate supported content types properly

Explanation

Correct Answer

A:

Proper encoding preserves characters consistently between client, API, and storage; the important constraint is that re-encoding bytes as the wrong character set can corrupt meaning.

Incorrect Answers

B:

Context-aware escaping prevents text from becoming active script or HTML; the requirement instead turns on this behavior: proper encoding preserves characters consistently between client, API, and storage; JSON escaping is not interchangeable with HTML or SQL escaping.

C:

Typed validation enforces the documented contract after parsing; this cannot substitute for the behavior that proper encoding preserves characters consistently between client, API, and storage, because JSON syntax validity does not prove field meaning or user authorization.

D:

Content-type checks help the parser distinguish valid API data from an unexpected gateway body; in contrast, proper encoding preserves characters consistently between client, API, and storage, while blind JSON parsing obscures transport failures.

Question 10

A conversation continues after a published service policy has changed since earlier messages. Which approach best meets this requirement?

  1. Reassert the current task objective concisely
  2. Label external text as untrusted input
  3. Use an external source refresh for changing facts
  4. Segment work into explicit milestones

Explanation

Correct Answer

C:

Refetching authoritative information avoids treating outdated context as eternally current; a crucial qualification is that the model should not invent a change simply because its history is old.

Incorrect Answers

A:

Reintroducing the active goal and constraints guides continuation back to the approved scope; this addresses a separate concern, not the fact that refetching authoritative information avoids treating outdated context as eternally current; repetition cannot override the need to preserve factual state accurately.

B:

Trust-boundary marking prevents source content from being promoted to application authority; here, the controlling capability is that refetching authoritative information avoids treating outdated context as eternally current, but labels should be backed by tool restrictions because prompts can be attacked.

D:

Milestones with acceptance criteria limit aimless elaboration and provide reliable checkpoints; by comparison, the intended solution relies on this: refetching authoritative information avoids treating outdated context as eternally current; a plan without tested completion criteria will still drift.

Claude Certified Developer - Foundations (CCDV-F) Practice Tests