Q38 of 60 — CCA Practice Exam

Code Generation with Claude Code

Your team has built a Claude Code–based assistant that helps developers analyze build artifacts. To keep the implementation simple, you used client-side direct execution: Claude is given a code execution tool, and when it invokes that tool with generated code, your application runs it locally in the same process where your internal analysis functions are defined. Three weeks after deployment, a security review flags the setup as a critical risk and cites two specific vectors. Which architectural limitation is the security review most likely identifying?

A. Because intermediate tool results are not loaded into Claude's context window, the model lacks visibility into failures and may silently retry operations, producing duplicate side effects on the local filesystem.

B. Client-side direct execution runs Claude-generated code outside of a sandbox, meaning arbitrary code executes with the same privileges as the host application, and tool invocations can be vectors for code injection.

C. The allowed_callers field is not supported in client-side execution mode, so any tool in the registry can be invoked by Claude without restriction, bypassing intended access controls.

D. Running code execution locally forces synchronous tool calls, which prevents Claude from writing Python that aggregates results across multiple tools and increases context window pressure from repeated sampling.