A Claude application that worked well in testing is now occasionally returning outputs that mention information not present in the input. The development team initially assumed the model was hallucinating, so they asked you to troubleshoot.
What would you do first?

• A. Add a system prompt instruction telling the model not to invent information, on the grounds that prompt-level instructions are the fastest fix for hallucination concerns.
• B. Apply a retrieval-augmented generation pattern to ground the responses in source content before any further investigation of the production traces.
• C. Replace the current model with a larger one to reduce the chance of hallucination, on the grounds that larger models tend to hallucinate less in typical applications.
• D. Examine production traces to identify whether the issue is hallucination by the model, context loss, prompt injection, or another failure mode before recommending a fix.

A teammate has submitted a pull request that adds a Claude-powered feature to your service. The code works, but the prompt and model selection are hard-coded inline, error handling is missing, and there are no tests for the integration.
What would you request during code review?

• A. Request changes that move prompt and model configuration to a configurable location and add tests, treating the missing error handling as a follow-up release item.
• B. Approve the pull request as-is, on the grounds that the feature works in the happy path and the missing pieces can be added in follow-up commits.
• C. Approve the pull request and add the missing pieces yourself in a follow-up commit so the teammate can move on to other work immediately.
• D. Request changes that move prompt and model configuration to a configurable location, add error handling for Claude API failures, and add tests for the integration.

Your team uses several plugins across multiple Claude applications, and a recent plugin update introduced a regression. The team had not been tracking plugin versions, so the team cannot easily identify which version was previously working. How would you address this?

• A. Upgrade every plugin to the latest version on a regular cadence to keep version drift small, on the grounds that drift contributes to regression risk.
• B. Treat plugins as untrackable third-party code and rely on plugin authors to communicate breaking changes when they happen, with no internal version tracking.
• C. Add explicit plugin version tracking to the project's configuration so the team can identify, pin, and upgrade plugin versions deliberately.
• D. Stop using all plugins until the team can rebuild equivalent functionality directly into the application code, treating plugin avoidance as a way to remove version-related risk.

Your Claude application runs long agentic workflows where the agent makes many tool calls, and the conversation history grows quickly. After about 20 tool calls, you notice the agent's responses become less focused and sometimes ignore earlier task constraints.
How would you address this?

• A. Remove tool calling from the workflow entirely so the agent operates as a single text-generation step with no tool outputs accumulating in the context window.
• B. Apply context engineering techniques such as tool output pruning or compaction to keep the active task state visible while reducing the volume of older content.
• C. Restart the agent every five tool calls to prevent any drift, with the agent losing all task state at each restart point during the workflow.
• D. Increase the model's context window so the agent can hold every tool output at full detail across the entire workflow no matter how many tool calls it accumulates.

You are designing an agent that handles a multi-step research task. You want the agent to break the task into smaller pieces, hand each piece to a focused subagent, and consolidate the results.
The agent pattern you would apply is...

• A. A memory pattern that stores the entire research history in advance, before any subtask begins execution.
• B. A context-window pruning pattern that drops each subtask's content after the agent moves on.
• C. A single tool-use loop that includes every tool the agent might need across all subtasks.
• D. An orchestrator and subagent pattern with specialized subagents assigned to each subtask.