Anthropic CCA-F Exam Dumps

Anthropic CCA-F Sample Question Answers

A developer has built their complete agent system and needs to do a final review before productiondeployment. According to the exam guide's reliability checklist, which of the following is NOT arecommended pre-deployment check?

A. Verify that all safety-critical rules are enforced via hooks/code rather than only via prompts 
B. Confirm that rate limiting and circuit breakers are configured for all external tool calls 
C. Ensure 100% test coverage of all possible user input combinations 
D. Validate that compaction preserves critical state information through custom instructions

Your organization is building a document review agent that processes hundreds of contracts daily. Eachcontract review generates a structured report. They want to measure the quality of reviews over time todetect drift or degradation. What evaluation architecture supports continuous quality monitoring?

A. Randomly sample 5% of reviews for manual human evaluation weekly 
B. Run every review through a second Claude evaluation pass that scores quality on predefined dimensions
C. Compare each review's structure to a template and flag deviations 
D. Use a combination: automated structural checks on 100% of reviews plus LLM-based evaluation on 10% sample plus human review of flagged outliers

A developer needs to understand how Claude handles system prompt instructions vs. user instructionswhen they conflict. A system prompt says 'Always respond in formal English' but a user says 'talk tome casually bro.' What is the expected behavior according to Claude's instruction hierarchy?

A. User instructions always override system prompt instructions
B. Claude will average the two styles and respond in semi-formal language 
C. System prompt instructions take precedence over user instructions — Claude should maintain formal English
D. The most recent instruction takes precedence regardless of source 

A team implements an agent with the Agent SDK. They want to add observability: logging each toolcall with timing, inputs, outputs, and model decisions. The Agent SDK uses setting_sources forconfiguration. Where should they implement the logging?

A. In the system prompt: 'Log every tool call you make with its parameters and timing' 
B. Enable debug mode in the Agent SDK to get automatic logging 
C. Wrap each tool handler with a logging decorator that captures inputs, outputs, execution time, and errors before returning results to the Agent SDK
D. Add a 'log' tool that Claude calls after every operation 

# Simple example of the pattern
def logging_decorator(tool_func):
    def wrapper(*args, **kwargs):
        start = time.time()
        result = tool_func(*args, **kwargs)
        log(tool=tool_func.__name__, inputs=args, 
            output=result, time=time.time()-start)
        return result
    return wrapper

An architect wants their agent to handle a multi-language customer base (English, Spanish, Japanese).The system prompt is in English. When a customer writes in Japanese, the agent should respond inJapanese. System prompt instructions and tool results are in English. How should the translation behandled?

A. Keep the system prompt and tools in English — Claude naturally responds in the customer'slanguage. Tool results in English are understood by Claude and the response is generated in thedetected language
B. Translate the system prompt into each language and use the matching one per request 
C. Use a translation MCP tool to convert everything to the customer's language before Claude processes it
D. Create separate agent configurations per language 

A developer is implementing streaming for their Claude integration. They want to display a 'typing'indicator while Claude is generating thinking blocks, and switch to displaying text when the responsetext starts. Which streaming events signal this transition?

A. A 'phase_change' event indicates the switch from thinking to text generation 
B. The content_block_stop event for the thinking block followed by content_block_start for the text block signals the transition
C. The message_delta event includes a 'current_phase' field 
D. Thinking and text generation happen in separate streaming connections 

content_block_start  (type: "thinking")  ? show typing indicator
    delta events...                       ? still thinking
content_block_stop                        ? thinking done ?
content_block_start  (type: "text")      ? switch to text display ?
    delta events...                       ? stream text to UI
content_block_stop                        ? response complete

A team builds an MCP tool that manages Kubernetes clusters. The tool includes operations likescale_deployment, delete_pod, and drain_node. These are high-impact operations that should requireconfirmation. How should the tool design handle dangerous operations?

A. Add 'DANGEROUS:' prefix to tool descriptions and rely on Claude to warn the user 
B. Separate dangerous tools into a different MCP server that requires admin credentials 
C. Require the user to type 'CONFIRM' before each dangerous operation 
D. Implement a two-phase execution: the tool first returns a preview of what will happen (dry run), and requires a second call with a confirmation token to execute

Phase 1 — Preview:
User/Claude calls scale_deployment(replicas=0)
Tool returns: {
  "preview": "This will scale deployment 'api-server' to 0 replicas, 
               causing downtime for all users",
  "confirmation_token": "abc123xyz",
  "expires_in": 60
}

Phase 2 — Execute:
Claude shows preview to user, user confirms
Tool called again with confirmation_token: "abc123xyz"
? Operation executes

A senior developer is optimizing their token costs. They discover that 60% of their API cost comesfrom input tokens (most content is repeated context). Only 15% comes from output tokens, and 25%from thinking tokens. What optimization provides the biggest cost reduction?

A. Reduce output length with max_tokens to save on the 15% output cost 
B. Reduce thinking budget to save on the 25% thinking cost 
C. Switch to a smaller model that has lower per-token costs across all categories 
D. Implement prompt caching to save on the 60% input cost — cache reads cost only 10% of base input price

Without caching:  60% of bill = input tokens
With caching:     60% × 10% = 6% of bill for same content

Savings = ~54% reduction in total API cost

A developer is configuring Claude Code for a project that uses both JavaScript and Rust. TheJavaScript code follows Prettier formatting while Rust follows rustfmt conventions. How shouldCLAUDE.md handle this dual-language setup?

A. Include both formatting conventions in the root CLAUDE.md with clear section headers 
B. Create separate CLAUDE.md files in the js/ and rust/ directories 
C. Use .claude/rules/ with glob-based rules: one file with **/*.js,**/*.ts pattern for Prettier rules, and another with **/*.rs pattern for rustfmt rules
D. Rely on the language-specific formatters and don't include formatting rules in CLAUDE.md 

.claude/
  rules/
    javascript.md   # applies to **/*.js, **/*.ts
    rust.md         # applies to **/*.rs

A platform engineer is building a system where multiple Claude agents share information through acentralized message bus. Agent A publishes analysis results, Agents B and C consume results relevantto their domains. The messages include metadata, status, and data payloads. What is the keyconsideration for the message format?

A. Use the same structured format that each consuming agent can parse without additional tool calls, with clear metadata for routing and versioning
B. Use plain text messages for maximum compatibility 
C. Use different formats per agent pair for maximum efficiency 
D. Let each agent define its preferred input format 

{
  "metadata": {
    "source_agent": "agent_a",
    "message_id": "uuid-123",
    "timestamp": "2026-04-30T10:00:00Z",
    "version": "1.2",
    "domain": "financial_analysis"
  },
  "status": "completed",
  "data": {
    "result": "...",
    "confidence": 0.95
  }
}

A developer is using Claude's structured output to generate JSON configuration files. They need theoutput to include comments explaining each configuration option. JSON doesn't support comments.What is the best structured output approach?

A. Use json_object mode which allows comments in the output 
B. Use YAML mode instead of JSON for comment support 
C. Generate the JSON and comments in separate API calls 
D. Generate a schema with parallel fields: each config option has a 'value' field and a 'comment' field, e.g., {port: {value: 8080, comment: 'Server listening port'}}

{
  "port": {
    "value": 8080,
    "comment": "Server listening port"
  },
  "debug": {
    "value": false,
    "comment": "Enable verbose logging for development"
  }
}

An AI engineer is implementing an evaluation framework for their agent system. They want to measurewhether the agent asks for clarification when a customer's request is ambiguous rather than makingassumptions. What evaluation metric captures this behavior?

A. Response accuracy — measuring whether final answers are correct 
B. Clarification rate — measuring the percentage of ambiguous inputs where the agent asks for clarification before acting, compared against a ground truth set of inputs that require clarification
C. Response time — faster responses indicate the agent made assumptions 
D. Token efficiency — lower tokens mean more decisive responses 

Clarification Rate = (Ambiguous inputs where agent asked for clarification)
                     ?????????????????????????????????????????????????????
                        (Total ambiguous inputs in ground truth set)
                        
                        Target: as close to 100% as possible

A developer wants to add a keyboard shortcut in Claude Code that switches between regular mode andplan mode during interactive sessions. What is the default keyboard shortcut for toggling plan mode?

A. Shift+Tab to toggle between plan mode and regular mode 
B. Tab to cycle between modes 
C. Ctrl+P to toggle plan mode 
D. Escape to enter plan mode 

A team is building an MCP server that integrates with Salesforce. Their tool catalog includes:search_contacts, get_contact_by_id, update_contact, create_opportunity, get_opportunity_pipeline,update_opportunity_stage, generate_report, and send_email. This is 8 tools. During testing, Claudeoccasionally picks the wrong tool (e.g., search_contacts when get_contact_by_id is more appropriate).What improves tool selection accuracy?

A. Add detailed descriptions that clearly differentiate each tool's purpose with usage guidance: e.g.,'search_contacts: Use when you need to FIND contacts by name, email, or company. NOT forretrieving a known contact (use get_contact_by_id instead)'
B. Reduce to 4 tools by combining related operations 
C. Add a tool_router tool that Claude calls first to determine which tool to use 
D. Rename tools with numbered prefixes: tool1_search, tool2_get, etc. 

A technical lead is migrating their agent from a 200K token context model (Claude Sonnet 4.5) to a 1Mtoken context model (Claude Sonnet 4.6). They expect to handle longer documents. What otherconsideration should they account for with the larger context?

A. The 1M context model has the same pricing per token, so costs only increase proportionally to usage 
B. Prompt caching minimum thresholds differ between models — Sonnet 4.5 has 1024 token minimum while Sonnet 4.6 has 2048 token minimum, affecting caching strategy
C. The architecture doesn't need any changes — just update the model parameter 
D. The response quality is identical regardless of context length 

A developer has a prompt engineering challenge: they need Claude to fill in a form that has 20 fields,but typically only 5-8 fields are relevant for any given request. The remaining fields should be null.Using structured output with all 20 fields as required causes Claude to hallucinate values for irrelevantfields. What schema design fixes this?

A. Make all 20 fields optional (nullable) with clear descriptions indicating when each field applies 
B. Use a dynamic schema generated per-request based on the input context 
C. Keep all fields required but add a 'confidence' field alongside each to flag uncertain values 
D. Create 4 different schemas for different form types, each with only the relevant required fields 

A solutions engineer is troubleshooting an agent that uses the Agent SDK. The agent should use the'search' tool before answering questions, but it frequently answers from training knowledge withoutsearching. The tool is properly defined and the system prompt says to 'use the search tool to findanswers.' What is the most effective fix?

A. Set tool_choice: {'type': 'tool', 'name': 'search'} for the first turn to force the initial search, then switch to 'auto' for subsequent turns
B. Rewrite the system prompt to be more emphatic: 'ALWAYS use the search tool' 
C. Set tool_choice: 'any' to force tool use on every turn 
D. Add a PreToolUse hook that blocks non-search responses 

A team maintains a Claude Code project with a growing CLAUDE.md file that's now 800 lines. ClaudeCode seems to be ignoring rules defined in the latter half of the file. What is the recommendedapproach for managing large CLAUDE.md files?

A. Delete rules that seem to be ignored — they're probably unnecessary 
B. Increase Claude's context window to ensure the full file is processed 
C. Convert the CLAUDE.md to a more compact YAML format 
D. Split into a concise root CLAUDE.md with the most critical rules, use @import for detailed sections, and use .claude/rules/ files for context-specific rules with glob patterns

your-project/
??? CLAUDE.md              # Critical rules only, ~50-100 lines
??? .claude/
    ??? rules/
        ??? testing.md     # Auto-loaded everywhere
        ??? security.md    # Auto-loaded everywhere
        ??? frontend/
            ??? react.md   # paths: src/**/*.tsx — loads only for React files

A developer builds an MCP tool that generates images using Stable Diffusion. The tool takes a promptand returns a base64-encoded image. Claude then describes the image to the user. However, Claude'sdescriptions don't match the generated images because Claude can't actually see the tool output as animage. How should this be architectured?

A. Have Claude generate the image description from the original prompt without seeing the actual image
B. Have the tool return a text description of what it generated alongside the image 
C. Store the image and return a URL that Claude mentions to the user 
D. Include the base64 image in a subsequent message with image content type so Claude can actually see and describe it

Your team is designing an agent evaluation framework. They need to test whether the agent handlestool failures gracefully across 50 different error scenarios. Running each scenario manually takes 10minutes. What is the most efficient approach?

A. Create a comprehensive test prompt that covers all 50 scenarios in one conversation 
B. Test 10 representative scenarios and extrapolate for the remaining 40 
C. Hire a QA team to manually test each scenario 
D. Use the Message Batches API to run all 50 scenarios as a batch — each scenario is a separate request with a mocked tool failure, and results are analyzed programmatically

A developer's agent needs to process CSV files uploaded by users. The files vary from 100 rows to 1million rows. For small files, direct context inclusion works well. For large files, the context can't holdall data. What scalable architecture handles both cases?

A. Always process the full CSV regardless of size, relying on compaction if needed 
B. Always chunk the CSV into 100-row segments and process sequentially 
C. For small files (<5K rows): include directly in context. For large files: load into a database and provide SQL query tools for the agent to explore the data on demand
D. Convert all CSVs to summary statistics before processing 

A development team is implementing context awareness (citations) for a RAG application. They wantto display inline citations like 'According to [Source A], the market grew by 15%.' The API returnscitation objects with character-level indices. How should the client render these?

A. Use the citation indices to insert reference markers into the text, linking each marker back to the specific passage in the source document
B. Append all citations as footnotes at the end of the response 
C. All of the above are valid rendering approaches depending on UX requirements
D. Create a sidebar with source highlights that appear when hovering over cited text 

A team is evaluating their Claude Code setup and wants to understand the permission model. They adda new tool via MCP that can execute shell commands. When an agent tries to use this tool, it gets apermission error. Where are MCP tool permissions managed?

A. In .claude/settings.json under the 'allowedTools' and 'blockedTools' arrays, with glob pattern matching for MCP tool names
B. In the CLAUDE.md file under an 'allowed_tools' section 
C. In the MCP server configuration file 
D. Permissions are managed per-conversation and reset each time 

A developer builds an MCP server that provides a search tool. The search tool returns results withURLs. After deployment, they notice Claude sometimes fabricates URLs that look similar to realresults but don't exist. What server-side measures prevent this?

A. Add URL validation in the system prompt instructing Claude to only use URLs from tool results 
B. Use context awareness (citations) to track which URLs come from tool results vs. model generation
C. Implement client-side URL validation that checks all URLs in Claude's response 
D. Return URLs as structured data in tool results (not embedded in text) and instruct Claude to reference results by ID rather than regenerating URLs

A product team is designing a fraud detection agent for a bank. The agent analyzes transactions in realtime and flags suspicious activity. False positives block legitimate transactions, causing customerfrustration. False negatives allow fraud through. Currently the agent flags 30% of transactions, ofwhich 90% are false positives. What analysis framework should be applied?

A. Lower the sensitivity threshold to reduce false positives 
B. Use a tiered confidence system: high-confidence fraud ? block immediately, medium-confidence? add friction (2FA), low-confidence ? allow with monitoring. Each tier has different precision/recalltradeoffs
C. Remove automated blocking entirely and rely on post-transaction review 
D. Train a separate classifier to pre-filter the agent's results 

A developer notices that their Claude Code session becomes sluggish after extended use. Responsestake longer and quality decreases. They suspect context window saturation. What is the recommendedmaintenance approach for long coding sessions?

A. Restart Claude Code to get a fresh context window
B. Delete all CLAUDE.md files temporarily to reduce context consumption 
C. Use /compact to manually trigger compaction, then verify Claude's understanding of the current task by asking it to summarize what it knows 
D. Switch to a model with a larger context window 

A DevOps engineer is building an agent that must generate responses in strict compliance with acompany's brand voice guide (a 15-page document). They've included the full brand guide in thesystem prompt. Responses generally match the brand voice, but occasionally Claude generates offbrand content for edge-case topics not covered in the guide. How can they improve consistency?

A. Reduce the brand guide to key bullet points to avoid Claude getting confused by verbose instructions
B. Add few-shot examples showing correct brand voice application for various scenarios, including edge cases similar to the problematic ones
C. Use a PostToolUse classifier that checks every response against the brand guide and regenerates if off-brand
D. Fine-tune a custom model on brand-compliant responses 

A developer is building a multi-step data pipeline with Claude. Step 1 extracts data from PDFs, Step 2normalizes the data, Step 3 performs analysis, and Step 4 generates a report. Each step's output feedsinto the next. After Step 2, the normalized data is 40K tokens. They worry Step 3 will lose accuracyover long context. What prompting pattern optimizes this pipeline?

A. Run all 4 steps in a single conversation to maintain context
B. Combine Steps 1-2 and 3-4 into two larger steps to reduce the number of calls 
C. Run all steps in parallel with separate conversations 
D. Use prompt chaining: run each step as a separate API call, passing only the necessary output from the previous step as input to the next 

An engineering manager is building an enterprise agent platform that serves multiple departments.Each department has its own MCP servers, tools, and access controls. The IT department should accessinfrastructure tools, HR should access employee tools, and Finance should access accounting tools.How should tool access be scoped?

A. Give all departments access to all tools and rely on prompt instructions for scope 
B. Create separate Claude API keys per department with different tool permissions 
C. Use per-agent tool configurations that expose only department-appropriate tools, with MCP server permissions enforcing access control at the tool execution layer
D. Use a single MCP server that checks the caller's department before returning results 

A developer wants to use Claude for generating regex patterns from natural language descriptions.They use the Messages API with a simple prompt. The generated regex often has edge cases that fail.What approach improves regex generation quality?

A. Use a larger model (Opus instead of Sonnet) for better pattern generation 
B. Ask Claude to generate multiple regex variants and pick the best one 
C. Use extended thinking to let Claude reason through edge cases 
D. Provide test strings (both matching and non-matching examples) in the prompt, then use a tool that tests the generated regex against all examples and returns failures for iteration