Q35 of 60 — CCA Practice Exam

Structured Data Extraction

Your structured data extraction pipeline uses a Claude agent to process 1,200 regulatory filings per day. A PreToolUse hook is registered to validate tool arguments before any write operation commits extracted records to the database — rejecting malformed or out-of-range values before they propagate downstream. During an incident review, you discover that on sessions where Claude hits the max_turns limit, the hook silently does not fire, and invalid records are written to the database without validation. Which architectural change most effectively closes this gap?

A. Add a stronger system prompt instruction telling Claude to always validate extracted values before calling write tools, so the hook is not the only enforcement point and prompt-level checks catch malformed records before they reach the database.

B. Increase max_turns to a significantly higher limit so that Claude is far less likely to hit the cap during normal extraction sessions, reducing how often the hook-bypass window can occur in production.

C. Add a post-session validation gate outside the agent session boundary that re-validates all records written during the session against the same schema before marking them committed.

D. Switch the PreToolUse hook matcher from an exact tool name to a wildcard pattern so it fires for a broader set of tool types, making it harder for any single tool call to bypass the validation check.