[BUG] Undisclosed background command execution burning user credits without consent — Claude Sonnet 4.6

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Claude Code (Sonnet 4.5) repeatedly executed Bash commands in the background
without disclosing this to me or obtaining my consent. This burned compute
credits silently while I had no visibility into what was running or for how long.

During a git commit task with a slow pre-commit hook, the model silently retried the failing command 6+ times in the background. The user only discovered this through unexpected task notifications.

What Should Happen?

When a command fails, the model should stop, report what failed and why, and wait for user instruction before taking any further action. Commands should never be retried silently in the background without disclosure.

Error Messages/Logs

Error Logs:
bddjl7qzz — FAILED exit 1 (ESLint SIGTERM)
btmel19z7 — FAILED exit 1 (ESLint SIGTERM)
b3jp959pn — FAILED exit 144
b0cfp242i — Completed exit 0 (ran without user knowledge)
bfbstxflt — Completed exit 0 (ran without user knowledge)
b4ahdmnsu — FAILED exit 1

Steps to Reproduce

Steps to Reproduce:

Use Claude Sonnet 4.6 via Claude Code CLI
Begin a git commit task with a slow pre-commit hook
Model silently retries the failing command in the background 6+ times
User discovers only through unexpected task notifications

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

claude-sonnet-4-6

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_