Free Anthropic CCA-F practice questions

10 free Anthropic CCA-F practice questions with the correct answer and a full explanation for each, taken from the CertStash pack of 85 questions. Work through them, then open each answer to check your reasoning.

Get all 85 questions (US$39) · Download these 10 as a PDF

Question 1

The synthesis agent receives summarized findings from the web search and document analysis agents, then passes a consolidated summary to the report generator. During testing, you discover the generated reports make factual claims without proper citations, the report generator cannot attribute statements to their original sources because that metadata was lost during the summarization steps.

What’s the most effective approach to ensure proper source attribution in the final reports?

  1. Have the report generator query the web search agent to re-locate sources for claims in the final report.
  2. Have each agent output structured data separating content summaries from source metadata (URLs, document names, page numbers).
  3. Skip summarization and pass full raw outputs from web search and document analysis directly to the report generator.
  4. Instruct the synthesis agent to embed source references inline within its summary text using a consistent citation format.
Show answer and explanation

Correct answer: B. Have each agent output structured data separating content summaries from source metadata (URLs, document names, page numbers).

Structured data separation ensures that content summaries and source metadata (URLs, document names, page numbers) flow together through the pipeline without being lost during summarization. When each agent outputs both content and metadata as distinct fields, the synthesis agent and report generator can maintain the link between claims and their origins. Option A creates unnecessary redundant work by re-querying sources that were already found. Option C eliminates the efficiency gains of summarization and overwhelms the report generator with irrelevant detail. Option D embeds citations as text strings, which are easily lost or corrupted when downstream agents process or r-ummarize the content.

Why the other options are wrong

  • A. Re-querying the web search agent wastes computational resources and doesn't solve the root cause of metadata loss during summarization.
  • C. Passing raw 120K token outputs defeats the purpose of the summarization pipeline and exceeds context windows inefficiently.
  • D. Inline text citations are fragile; they degrade when content is re-summarized or reformatted by downstream agents.

Question 2

After the web search agent finds 25 sources (120K tokens of raw content), the document analysis agent extracts key insights (15K tokens), and the synthesis agent produces a coherent narrative draft (3K tokens), the coordinator must pass context to the report generation agent for the final output with proper source citations.

What context-passing strategy provides the best balance of completeness and efficiency?

  1. Pass the full accumulated context from all prior agents.
  2. Pass the synthesis draft along with a structured source index that maps key claims to their source URLs and relevant excerpts.
  3. Pass only the synthesis draft and have a separate post-processing pipeline match claims to sources and insert citations after the report is generated.
  4. Pass a condensed summary of all prior stages that preserves the main findings and attributes them to sources by name only.
Show answer and explanation

Correct answer: B. Pass the synthesis draft along with a structured source index that maps key claims to their source URLs and relevant excerpts.

Passing the synthesis draft (3K tokens) alongside a structured source index maintains both completeness and efficiency. The synthesis draft provides the coherent narrative the report generator needs, while the index, mapping key claims to source URLs and relevant excerpts, enables proper citation without requiring the full 120K tokens of raw content or the 15K tokens of intermediate extracts. This approach preserves fidelity by keeping source context linked to specific claims. Option A wastes tokens by passing all accumulated context when only the final synthesis is needed for narrative flow. Option C risks lost citations because post-processing cannot reliably recover source intent from finished prose. Option D loses methodological detail by attributing findings only by source name, making it impossible to cite specific evidence or methodologies.

Why the other options are wrong

  • A. Passing all 138K prior tokens is inefficient and provides information irrelevant to report generation.
  • C. Post-processing citation matching is unreliable; claims may be paraphrased or combined in ways that obscure their original sources.
  • D. Attribution by name only prevents citation of specific methodologies, data points, or evidence supporting claims.

Question 3

Your multi-agent research pipeline crashed after processin  documents. The web search agent had identified relevant sources, the document analyzer had partially completed extraction, and the synthesizer had begun pattern identification. You need to resume processing without repeating work or losing fidelity of prior findings.

What state management approach best balances information fidelity with context efficiency when restoring agent state?

  1. Have each agent persist a structured export to a known location. On resume, the coordinator loads the manifest and injects relevant state into agent prompts.
  2. Index all agent outputs in a shared vector store. When resuming, each agent queries the store using semantic search to retrieve relevant prior findings.
  3. Have each agent maintain its own persistent state file and reload it independently at the start of each session.
  4. Persist the coordinator’s conversation log containing all task delegations and responses, providing this to agents when resuming.
Show answer and explanation

Correct answer: A. Have each agent persist a structured export to a known location. On resume, the coordinator loads the manifest and injects relevant state into agent prompts.

Having each agent persist structured exports to a known location and injecting relevant state via prompts when resuming provides a balance between fidelity and efficiency. The coordinator loads a manifest of completed work and shares only the pertinent prior findings with each agent through its instructions, avoiding redundant processing while preserving the context needed to continue accurately. This approach maintains research quality without forcing agents to reprocess 12 completed documents. Option B introduces semantic search approximation, which can lose precision in technical research contexts. Option C allows agents to independently reload state without coordinator awareness, risking inconsistent resumption or duplicate work. Option D relies on conversation logs, which are verbose and do not provide structured data easily consumable by agent prompts.

Why the other options are wrong

  • B. Semantic search can conflate similar findings and may retrieve approximate rather than exact prior work, reducing fidelity.
  • C. Independent state reloading bypasses the coordinator's control and risks uncoordinated, inconsistent processing.
  • D. Conversation logs are unstructured, verbose, and difficult for agents to parse into actionable prior findings.

Question 4

You’ve configured the system so that all four subagents have access to the complete set of 18 tools. During testing, agents frequently call tools outside their specialization, the synthesis agent attempts web searches, and the report generator tries to analyze documents.

What is the primary cause of this poor tool selection behavior?

  1. Choosing from 18 tools instead of 4-5 relevant ones increases decision complexity beyond reliable selection thresholds.
  2. The tool definitions consume too much context window space, leaving insufficient room for task content.
  3. The agents’ role descriptions in their system prompts conflict with having access to tools outside that role.
  4. The coordinator cannot track which capabilities each subagent has, leading to misrouted tasks.
Show answer and explanation

Correct answer: A. Choosing from 18 tools instead of 4-5 relevant ones increases decision complexity beyond reliable selection thresholds.

LLM-based agents exhibit degraded tool selection when choosing among many options; the decision space becomes too large and the model's confidence in selecting the appropriate tool decreases. With 18 available tools, agents exceed the reliable selection threshold and resort to broad or off-topic tool use. Constraining each subagent to 4–5 tools aligned with its specialization dramatically improves selection accuracy. The synthesis agent should never have web search tools; the report generator should never have document analysis tools. This is a fundamental limitation of how language models make discrete tool choices under high branching factor. Option B misattributes the issue to context consumption rather than decision complexity. Option C incorrectly suggests that conflicting role descriptions cause poor tool use when the problem is choice overload. Option D incorrectly blames coordination tracking when the issue is the agent's own decision-making under complexity.

Why the other options are wrong

  • B. Tool definitions do not typically consume enough context to starve task content; the problem is decision complexity, not token scarcity.
  • C. Role conflicts do not directly cause poor tool selection; the issue is the number of choices available, not contradictory instructions.
  • D. The coordinator's awareness of agent capabilities is orthogonal to the agent's own poor tool selection from an oversized palette.

Question 5

The coordinator provides detailed step-by-step instructions to the web search subagent, specifying exact search queries, source priorities, and date filters. Production monitoring reveals three issues: (1) the subagent reports “insufficient results” rather than trying alternative approaches when pre-specified searches fail, (2) research quality drops for emerging topics that don’t match expected patterns, and (3) the subagent rarely surfaces valuable tangential sources.

What’s the most effective way to improve subagent adaptability?

  1. Implement a topic classification step where the coordinator categorizes requests as “well-defined” or “exploratory” and uses different instruction styles for each category.
  2. Add explicit fallback directives to the detailed instructions: “If specified searches yield fewer than N results, attempt alternative query formulations before reporting failure.”
  3. Remove procedural details entirely, delegating with simple goals like “research X thoroughly” and relying on the subagent’s general capabilities.
  4. Specify research goals and quality criteria (coverage breadth, source diversity, recency) rather than procedural steps, letting the subagent determine its search strategy.
Show answer and explanation

Correct answer: D. Specify research goals and quality criteria (coverage breadth, source diversity, recency) rather than procedural steps, letting the subagent determine its search strategy.

Specifying research goals and quality criteria (coverage breadth, source diversity, recency) rather than procedural steps allows the subagent to exercise judgment and adapt its strategy based on what it discovers. The subagent becomes empowered to try alternative search formulations, recognize emerging topics outside pre-specified patterns, and identify tangential sources that meet the quality criteria. This approach treats the agent as a capable researcher rather than a script executor. Option A adds complexity without fundamentally changing how the agent responds to unexpected situations. Option B adds fallback directives but still constrains the agent to procedural thinking; it does not encourage strategic adaptability. Option C removes all guidance and risks incoherent unfocused research. Option D balances autonomy with clarity by stating what success looks like, not how to achieve it.

Why the other options are wrong

  • A. Topic classification adds a preliminary step but does not fundamentally change the subagent's procedural adherence or adaptability.
  • B. Fallback directives remain procedural and do not grant the subagent strategic autonomy to adapt its overall approach.
  • C. Removing all guidance risks incoherent, unfocused research that fails to address the actual information need.

Question 6

The synthesis agent completes its initial pass but flags that three key research questions remain unanswered because the web search and document analysis agents didn’t find relevant information on those specific subtopics. The coordinator currently proceeds directly to report generation, producing reports with incomplete coverage.

What change would most effectively improve research completeness?

  1. Have the report generation agent note which research questions couldn’t be answered, so users understand the limitations of the final output.
  2. Increase the initial breadth of queries sent to web search and document analysis to reduce the probability of missing relevant information.
  3. Have the coordinator evaluate synthesis output for gaps, then re-delegate to web search and document analysis with targeted queries before invoking synthesis again.
  4. Give the synthesis agent direct access to web search tools so it can autonomously fill knowledge gaps without returning control to the coordinator.
Show answer and explanation

Correct answer: C. Have the coordinator evaluate synthesis output for gaps, then re-delegate to web search and document analysis with targeted queries before invoking synthesis again.

Having the coordinator evaluate the synthesis output for gaps and re-delegate targeted queries to web search and document analysis agents creates a feedback loop that closes research gaps before report generation. The coordinator acts as quality control, identifying unanswered questions and issuing focused follow-up tasks. This preserves the specialization of agents and ensures comprehensive coverage. Option A documents incompleteness but does nothing to remedy it. Option B increases initial query breadth but is inefficient and does not specifically target the known gaps. Option D gives the synthesis agent tools outside its specialty (web search), undermining the agent specialization architecture and introducing uncontrolled autonomous behavior.

Why the other options are wrong

  • A. Documenting limitations in the final report acknowledges the problem but does not solve it; readers still receive incomplete research.
  • B. Increasing initial breadth is inefficient and speculative; targeted follow-up based on identified gaps is more effective.
  • D. Granting web search tools to the synthesis agent breaks specialization and removes coordinator oversight of research strategy.

Question 7

The web search agent has gathered several relevant sources for a research topic. The document analysis agent now needs to examine these sources.

How does information typically flow between these two specialized subagents?

  1. The agents communicate through an event-driven message queue, with the document analysis agent subscribing to web search completion events.
  2. The web search agent directly invokes the document analysis agent, passing the discovered sources as parameters.
  3. The coordinator agent receives the web search agent’s output and includes relevant findings in the prompt when invoking the document analysis agent.
  4. Both agents access a shared memory store where the web search agent writes findings and the document analysis agent reads them.
Show answer and explanation

Correct answer: C. The coordinator agent receives the web search agent’s output and includes relevant findings in the prompt when invoking the document analysis agent.

In a coordinator-based multi-agent architecture, subagents do not communicate directly. The coordinator receives the web search agent's output (discovered sources) and then includes those findings in the prompt when invoking the document analysis agent. This maintains a clear, auditable flow of information and gives the coordinator visibility and control over task sequencing. Option A assumes an event-driven architecture that is not typical for LLM-based agent systems managing sequential research tasks. Option B bypasses the coordinator, creating hidden dependencies and complicating debugging and state tracking. Option D assumes a shared memory store architecture, which is more complex than the direct coordinator-mediated approach and introduces synchronization concerns.

Why the other options are wrong

  • A. Event-driven message queues are not the typical pattern for LLM-based research coordinators managing sequential tasks.
  • B. Direct agent-to-agent invocation bypasses the coordinator and obscures the flow of control and data.
  • D. A shared memory store adds architectural complexity and synchronization overhead unnecessary for sequential task coordination.

Question 8

Production reviews reveal inconsistent handling of uncertainty in final reports. Sometimes conflicting subagent findings are synthesized into a single confident statement (losing nuance), while other times reports over-hedge with excessive qualifications (becoming unhelpful).

When the web search agent returns “industry analysts estimate $50B market size (methodology varies)” and the document analysis agent returns “peer-reviewed study estimates $35B (±$7B, 95% CI),” the coordinator either picks one arbitrarily or produces vague statements like “the market may be $35B-$50B depending on factors.”

What systematic approach best addresses this?

  1. Configure subagents to only report findings meeting a high-confidence threshold, filtering uncertain information before it reaches the coordinator.
  2. Add a verification subagent that cross-references findings across sources, only passing claims to synthesis that are corroborated by at least two independent sources.
  3. Instruct the synthesis agent to structure reports with explicit sections distinguishing well-established findings from contested ones, preserving original source characterizations and methodological context.
  4. Implement a confidence calibration layer that normalizes subagent uncertainty expressions to standardized probability scores (0.0-1.0), then weight-average findings by their calibrated confidence.
Show answer and explanation

Correct answer: C. Instruct the synthesis agent to structure reports with explicit sections distinguishing well-established findings from contested ones, preserving original source characterizations and methodological context.

Instructing the synthesis agent to structure reports with explicit sections distinguishing well-established findings from contested ones, while preserving original source characterizations and methodological context, maintains nuance and transparency. The report can present both the $50B industry estimate and the $35B±$7B peer-reviewed finding with their respective methodologies intact, allowing readers to understand the basis for differences. This approach honors the input data without losing information to artificial averaging or hedging. Option A filters out uncertain information, reducing completeness and transparency. Option B creates a separate verification step that may arbitrarily discard valid findings that lack corroboration. Option D normalizes uncertainty into probabilistic scores, losing domain-specific methodological context and introducing false precision through numerical conflation of qualitatively different uncertainty sources.

Why the other options are wrong

  • A. Filtering uncertain findings reduces completeness and removes legitimate research results from the final output.
  • B. Requiring corroboration by two sources may exclude valid singleton findings and adds an arbitrary gate.
  • D. Normalizing uncertainty to numerical scores loses methodological context and creates false equivalence between different types of uncertainty.

Question 9

In production, final reports frequently contain claims without proper source attribution. Investigation shows that while the web search and document analysis agents correctly attach citations to their outputs, the synthesis agent loses track of which sources support which conclusions when combining findings.

What’s the most effective architectural change?

  1. Require all subagents to output structured claim-source mappings that the synthesis agent must preserve and merge when combining findings from multiple sources.
  2. Maintain complete transcripts of all subagent interactions and add a citatio-esolution agent to analyze logs and determine attributions before report generation.
  3. Add a verification step where the report generator uses semantic similarity matching against original sources to reconstruct which claims came from which documents.
  4. Have the coordinator inject source identifier prefixes into text before each handoff, then parse these prefixes at report generation to reconstruct citations.
Show answer and explanation

Correct answer: A. Require all subagents to output structured claim-source mappings that the synthesis agent must preserve and merge when combining findings from multiple sources.

The root cause is that the synthesis agent loses the mapping between claims and their sources when merging outputs from multiple subagents. Option A directly solves this by requiring each subagent to output structured claim-source mappings and mandating that the synthesis agent preserve and merge these mappings. This maintains the traceability chain throughout the pipeline. Structured data formats prevent information loss that occurs when agents work with unstructured text, ensuring citations remain attached to their supporting claims.

Why the other options are wrong

  • B. Maintaining transcripts and adding a citation-resolution agent adds complexity pos-acto rather than preventing the loss of citations during synthesis, making it an inefficient and reactive solution.
  • C. Semantic similarity matching is a heuristic that cannot reliably reconstruct precise source attributions and may incorrectly associate claims with sources, introducing new errors.
  • D. Injecting prefixes into text is fragile and loses fidelity during synthesis when agent combines and reformats content, requiring error-prone parsing that doesn't scale to complex multi-source synthesis.

Question 10

After the web search agent and document analysis agent complete their tasks, the coordinator invokes the synthesis agent. However, the synthesis agent responds that it cannot complete the task because no research findings were provided.

What is the most likely cause of this issue?

  1. The subagents need to share a single API connection to enable automatic context sharing between invocations.
  2. The synthesis agent needs tools that can fetch results directly from the other agents’ conversation histories.
  3. The coordinator did not include the outputs from the previous agents in the synthesis agent’s prompt.
  4. The synthesis agent’s context window is not large enough to hold the combined outputs from both previous agents.
Show answer and explanation

Correct answer: C. The coordinator did not include the outputs from the previous agents in the synthesis agent’s prompt.

When the synthesis agent reports receiving no research findings, the coordinator failed to pass the outputs from the web search and document analysis agents into the synthesis agent's prompt context. In agentic architectures, information flow between sequential agents is not automatic; the orchestrator must explicitly include prior results in the next agent's input. Without this explicit handoff, downstream agents have no access to upstream work.

Why the other options are wrong

  • A. Shared API connections enable resource sharing but do not solve context passing; agents still require explicit data transfer to know what work has been completed.
  • B. While tools to fetch conversation history could work, it's an over-engineered solution; the standard pattern is for the coordinator to directly include results in prompts rather than requiring agents to fetch them.
  • D. Context window size becomes a concern only if the data is being passed; the agent didn't even receive the findings, so window size is not the limiting factor here.

That was 10 of 85.

The full Anthropic CCA-F pack has all 85 questions, each with the answer, the explanation and why the other options are wrong, plus a questions-only copy for timed runs. US$39, paid once, with free monthly updates and a pass-or-your-money-back guarantee.

Get the full pack