Q52 of 60 — CCA Practice Exam

Structured Data Extraction

Your structured data extraction pipeline uses Claude to process 500 financial documents per day. Each extraction call invokes a get_document_fields tool that returns the full raw API response from the document management system — including internal database IDs, audit log arrays, and formatting metadata — alongside the semantic fields Claude actually needs. Over time, you notice Claude increasingly makes incorrect field-selection decisions and occasionally calls the tool a second time to "re-check" values it just retrieved. Logs confirm the tool is returning correct data. What is the most likely root cause and the most effective fix?

A. The tool description is too brief and doesn't explain which fields Claude should use for its next reasoning step; expanding the description to cover each returned field's semantics will resolve the re-checking behavior.

B. The tool response contains bloated, low-signal data that dilutes Claude's attention; refactoring the tool to return only the semantic fields and stable identifiers Claude needs to reason about its next step will eliminate the confusion.

C. Claude is losing track of already-retrieved values because the context window is filling up; increasing max_tokens will give Claude more room to retain previous tool results and avoid redundant calls.

D. The single get_document_fields tool is overloaded; splitting it into separate tools for each document field type will reduce selection ambiguity and prevent Claude from re-fetching data.