Claude Certified Architect – Foundations Practice Question 95: Bash Sandbox Auto Allow
Claude Certified Architect, Foundations · Test 10 · question 5 of 10
Developers on a team are interrupted by a permission prompt for nearly every build and test command Claude Code runs in Bash. Security does not want developers to bypass permission checks entirely, and wants commands that run on a laptop to be unable to write outside the project or reach arbitrary network hosts. Which approach reduces the prompts while keeping those limits?
Show answer and explanation
Answer:
B is correct. The Bash sandbox is an operating-system-enforced boundary around the shell commands Claude runs; by default commands can write to the working directory, a temp directory and added directories, and new network domains need approval. In auto-allow mode Claude Code approves a command with no prompt when it runs inside the sandbox, and explicit deny rules, rm/rmdir on critical paths and content-scoped ask rules still apply. Because the OS enforces the limits, approving automatically stays safe.
A is wrong because skipping all permission checks removes the boundary the security team wants. C is wrong because a blanket allow rule approves commands with no containment, so nothing limits where they write or connect. D is wrong on two counts: regular permissions mode keeps the regular prompts even for sandboxed commands, and the sandbox covers shell commands only; Claude's file tools, MCP servers and hooks run outside it.
Reference: Configure the sandboxed Bash tool