The document analysis agent has a single analyze_documnet tool that takes a document and a free-text instruction parameter. During evaluation, requests like "extract the key financial metrics" often return narrative summaries, while "summarize the methodology" sometimes returns raw data tables. The synthesis agent reports that 35% of analysis results require re-requests with clarified instructions. What's the most effective way to improve reliability?

• A. Have the coordinator pre-classify each analysis request before passing instructions to the document analysis agent
• B. Keep the single tool but add an analyze_type enum parameter requiring explicit selection between extraction, summarization, and verification modes
• C. Enhance the tool description with detailed examples showing how different instruction phrasings should map to different output formats
• D. Split the generic tool into purpose-specific tools - extract_data_points, summarize_content, verify_claim_against_source - each with defined input/output contracts

You are integrating Claude Code into your Continuous Integration/Continuous Deployment (CI/CD) pipeline. The system runs automated code reviews, generates test cases, and provides feedback on pull requests. You need to design prompts that provide actionable feedback and minimize false positives.
A developer uses Claude Code to refactor a function during a development session. Before committing, the developer asks the same Claude session to review the code for issues. Later, a separate automated CI review catches several bugs that the same-session review missed.
What best explains this discrepancy?

• A. Claude retains the implementation context and prior decisions in the session, making it less likely to challenge assumptions underlying its own changes.
• B. The CI review must have used a more specific prompt, while the developer's review request was too general.
• C. The session's context window necessarily became full, leaving insufficient capacity for meaningful review.
• D. The CI environment can access the full repository, while a local Claude Code session can access only the current file.

You are integrating Claude Code into your Continuous Integration/Continuous Deployment (CI/CD) pipeline. The system runs automated code reviews, generates test cases, and provides feedback on pull requests. You need to design prompts that provide actionable feedback and minimize false positives.
In addition to your CI pipeline, your organization has enabled Claude's managed Code Review through the Claude GitHub App on this repository, and reviews run automatically on every pull request. Reviews average 18 findings per pull request. Developer feedback reveals three categories of unwanted noise: (1) style and formatting issues already enforced by your linter in CI, (2) findings on automatically generated template code under src/gen/, and (3) rendering- helper patterns that are intentional project conventions but get flagged because they resemble common anti-patterns. Only approximately four findings per pull request are genuine logic bugs.
What is the most effective way to reduce this noise while preserving the detection of genuine issues?

• A. Add custom review instructions to a GitHub Actions workflow file, using the action's prompt parameter to suppress duplicate lint findings, ignore generated template code, and apply stricter evidence requirements to rendering-related issues.
• B. Add detailed explanations to the project's CLAUDE.md describing which patterns are intentional, that linting is handled separately by CI, and that the src/gen/ directory contains automatically generated template code.
• C. Create a REVIEW.md file at the repository root containing skip rules for CI-enforced checks and generated files, together with a verification requirement that rendering-related findings cite a specific line demonstrating incorrect behavior.

You are building a structured data extraction system using Claude. The system extracts information from unstructured documents, validates the output using JSON schemas, and maintains high accuracy. It must handle edge cases gracefully and integrate with downstream systems.
Your extraction pipeline occasionally receives responses that cannot be parsed as valid JSON, causing downstream processing failures. The current implementation prompts Claude to return JSON in the response text and then parses it.
What is the most reliable approach to ensure Claude returns valid, schema-compliant structured data?

• A. Define a tool with a JSON schema specifying the expected structure, using tool use to constrain Claude's output to schema-compliant JSON.
• B. Add explicit formatting instructions to the prompt with JSON examples, emphasizing that Claude must return only valid JSON with no surrounding text.
• C. Use regular expressions to locate and extract JSON from the response text, handling cases where Claude includes explanatory text around the JSON block.
• D. Implement a retry loop that catches JSON parsing errors and re-prompts Claude with the error details, asking it to correct the malformed output.

You are building a structured data extraction system using Claude. The system extracts information from unstructured documents, validates the output using JSON schemas, and maintains high accuracy. It must handle edge cases gracefully and integrate with downstream systems.
After deployment, you find that 12% of extractions contain semantic errors that pass JSON Schema validation--for example, a duration such as "30 minutes" is incorrectly placed in an ingredient-quantity field. Human reviewers have the capacity to check only 20% of extractions.
Which approach most effectively allocates reviewer attention?

• A. Have the model output field-level confidence scores, and then calibrate review thresholds using a labeled validation set.
• B. Review all extractions from documents with formatting anomalies, such as unusual layouts or mixed content types.
• C. Randomly sample 20% of extractions for review, using corrections to track accuracy and identify error patterns.
• D. Prioritize the review of all extractions where required fields are empty or explicitly marked as not found.