Subagents prompt for permission on every tool call starting v2.1.56

Resolved 💬 5 comments Opened Feb 25, 2026 by csharpner Closed Apr 19, 2026

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?

Bug description:

Starting with v2.1.56, subagents (spawned via the Task tool) prompt the user for permission on every single tool call — including Read, Glob, Grep, and basic Bash commands. This makes subagents completely unusable.

Impact:

  • Two simultaneous Claude Code sessions both broke at the same time when the auto-update to 2.1.56 installed at ~01:40 local time
  • Subagents previously inherited the parent session's approved permissions; they no longer do

To reproduce:

  1. Run Claude Code v2.1.56
  2. Trigger any Task tool call that spawns a subagent
  3. Observe: every tool call inside the subagent triggers a user permission prompt

Expected behavior:

Subagents should inherit the parent session's permission mode. If the parent session has auto-approved a tool (or is in auto-approve mode), subagents should not re-prompt.

Actual behavior:

Every subagent tool call — Read, Edit, Write, Glob, Grep, Bash — prompts for permission regardless of parent session approvals.

Workaround:

2 options:

  1. Stop spawning Task subagents entirely and implement everything directly in the main session.
  2. Manually added permissions to ~/.claude/settings.json

I did not attempt a rollback.

Environment:

  • OS: Linux (Ubuntu)
  • Claude Code version: 2.1.56
  • Previous working version: Unsure: 4 updates were published over night. Whatever was latest before those updates.
  • Four versions released in ~24 hours: 2.1.52 → 2.1.53 → 2.1.55 → 2.1.56

What Should Happen?

Subagents should inherit the parent session's permission mode. If the parent session has auto-approved a tool (or is in auto-approve mode), subagents should not re-prompt for every tool call.

Error Messages/Logs

Steps to Reproduce

  1. Run Claude Code v2.1.56
  2. Trigger any Task tool call that spawns a subagent
  3. Observe: every tool call inside the subagent (Read, Edit, Write, Glob, Grep, Bash) triggers a user permission prompt, even if the parent session has auto-approved those tools

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

2.1.55

Claude Code Version

2.1.56

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_

View original on GitHub ↗

This issue has 5 comments on GitHub. Read the full discussion on GitHub ↗