Q49 of 60 — CCA Practice Exam

Code Generation with Claude Code

Your team has configured a background subagent to run automated integration test suites concurrently while developers continue coding. After two weeks, engineers report that roughly 30% of background subagent runs are silently abandoned mid-suite — the subagent stops executing but produces no error output. Examining the logs, you find that the abandoned runs all hit tool calls requiring permissions that were not pre-approved at launch time. What is the most effective architectural fix?

A. Add a system prompt instruction to the subagent telling it to request only the minimum necessary permissions and avoid tool calls it is uncertain about.

B. Audit the subagent's full tool usage during a foreground run, then enumerate all required tools in the subagent's tools frontmatter field so they can be pre-approved at background launch time.

C. Switch the integration test subagent to run in the foreground so that unexpected permission prompts can be answered interactively, accepting the loss of concurrency.

D. Set CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=1 to eliminate background execution and require developers to manually trigger each test suite run.