[BUG] In-built Plan subagent doesn't work anymore!

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?

Hi, it started today. Planning subagent always fails, restart session and claude code doesn't help.

This is output from from the session:

Claude wrote:

● Looking at the output, I notice inconsistencies in how probe operations display commands at different verbosity levels:

  1. UFW probe at -v: Shows command sudo ufw status verbose ✓
  2. Unattended Upgrades probe at -v: Shows NO command, only "Success" ✗
  3. Unprivileged Port probe at -v: Shows command sysctl -n net.ipv4.ip_unprivileged_port_start ✓

  But at -vv, we can see Unattended Upgrades actually executes TWO commands:
  - dpkg -l unattended-upgrades || true
  - systemctl status apt-daily-upgrade.timer || true

  According to CLAUDE.md, probes should "show cmd at -v, stdout at -vvv", so the Unattended Upgrades probe is not behaving correctly.

  Let me investigate the implementation to find the inconsistencies:

● Plan(Investigate probe inconsistencies)
  ⎿  Done (0 tool uses · 0 tokens · 1.8s)                                                                                                                                                                                              

● Let me investigate the probe implementation inconsistencies by exploring the codebase:

● Explore(Investigate probe inconsistencies)
  ⎿  Done (0 tool uses · 0 tokens · 1.7s)                                                                                                                                                                                              

● Let me investigate the probe implementations directly:

● Search(pattern: "**/server_setup/utils*.py")
  ⎿  Found 1 file (ctrl+o to expand)

...and so on...

As you can see, he tried to call "Plan" agent, but got 0 tool uses and 0 tokens back!
It happened to me today at least 5-8 times and i'm noticing, that once this started it never again successfully used "Plan" in-built subagent. Always the same 0 tokens, 0 tool calls, then main agent starts searching by himself.

What Should Happen?

Main agent should be able to call in-built subagent and it shouldn't fail with 0 tool calls and 0 tokens.

Error Messages/Logs

Steps to Reproduce

When in plan mode, main agent always seems to call Plan subagent, which is fine to prevent context pollution. Steps to reproduce should be obvious, but i am not sure if this is happening to everybody?

Claude Model

Sonnet (default)

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.0.30 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_