CCA-F: Claude Certified Architect — Foundations

By Anthropic · Question Mode

QuestionQ78

Context Management & Reliability

You are creating developer-productivity tools with the Claude Agent SDK. The agent assists engineers in exploring unfamiliar codebases, understanding legacy systems, generating boilerplate code, and automating repetitive work. It uses the built-in tools (Read, Write, Bash, Grep, Glob) and integrates with Model Context Protocol (MCP) servers.

Your codebase-exploration tool saves session IDs so engineers can continue investigations across work sessions. Yesterday, an engineer spent an hour analyzing a legacy authentication module and building context about its architecture and dependencies. They want to continue today. The session ID is valid, but version control indicates that a teammate's merge modified 3 of the 12 files the agent previously read overnight. Which approach best balances efficiency and accuracy?

Explanation

Resumed Claude sessions preserve their conversation history, so retaining the earlier analysis avoids duplicating work. Known changes to three previously read files can invalidate only the context or conclusions dependent on those files; explicitly identifying them lets the agent re-read and reassess the affected areas while preserving the valid analysis of the other nine files. This balances current-code accuracy with efficient reuse of session context.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!