Q19 of 60 — CCA Practice Exam

Developer Productivity with Claude

Your team has deployed a Claude-based developer productivity assistant that uses a memory tool to persist project context, coding conventions, and architecture decisions across sessions. After six months of active use, developers report that the assistant is becoming noticeably slower to start tasks — logs show that the initial memory read call now takes 4–6 seconds and returns over 200KB of content on every invocation. The assistant reads the full memory directory on every session start. What is the most effective architectural change to address this?

A. Increase the token budget allocated to the memory read call so Claude can process larger file contents without hitting truncation errors on any single invocation.

B. Enforce a maximum return size on the memory read command and implement pagination, so Claude retrieves a bounded portion of memory per call and fetches additional pages only as needed.

C. Add a system prompt instruction telling Claude to skip reading memory files that are unlikely to be relevant to the current task, relying on file names to judge relevance before reading.

D. Archive old memory files into a compressed backup and require developers to manually delete obsolete entries whenever the assistant's startup latency becomes unacceptable.