Anthropic Claude Certified Architect – Foundations

1. A company is building a long-running data pipeline agent that processes thousands of records over several hours. The agent crashes midway through the process. The team wants to ensure that when the agent restarts, it does not reprocess already-completed records. Which pattern should the architect implement to achieve this?

2. A team is building an agentic coding assistant using Claude Code hooks. They want to inject contextual feedback after every tool execution — for example, appending linting results after a file write operation. Which hook should they configure, and what can it do?

3. A company's security team is conducting a review of the development team's MCP server configuration practices. They find that the .mcp.json file in the main repository contains the following entry: { "database-server": { "command": "node", "args": ["server.js"], "env": { "DB_PASSWORD": "prod-secret-abc123" } } }. What is the primary security risk and the recommended remediation?

4. A developer is building an MCP server that provides both a search_documents tool (which performs a full-text search with ranking) and a documents resource (which returns a specific document by URI). A colleague suggests consolidating both into a single tool. Why should the architect keep them as separate MCP primitives?

5. A team maintains a large set of shared coding guidelines in a file called shared_guidelines.md that is 800 lines long. They want their root CLAUDE.md to remain concise but still apply all these guidelines. What is the recommended Claude Code approach?