›_ Claude Code Issues
Claude Code Issues / #24359

VSCode extension hangs/loops on complex multi-step tasks

Resolved 💬 3 comments Opened Feb 9, 2026 by cr-bh Closed Feb 12, 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?

Environment

  • Claude Code version: 2.1.11
  • VSCode version: [请填写你的版本]
  • OS: macOS (Darwin 24.2.0)
  • Model: claude-opus-4-5-20251101 (also tested with other models)

Description

The VSCode extension hangs or enters a loop when processing complex multi-step tasks, while simple tasks work normally.

Symptoms

  • Model repeatedly outputs the same text
  • Gets stuck at "Spelunking..." status
  • Appears to be in a "thinking loop" - generating reasoning but not executing tool calls
  • Eventually times out or requires manual interruption

Reproduction Steps

Tasks that work normally:

  1. Simple greeting: hello → responds correctly
  2. Simple command: run ls → executes correctly
  3. Reading small files: read config.py → works fine
  4. Reading large files: read docs/TECHNICAL_DOCUMENTATION.md → works (shows summary)
  5. Simple questions: what does this project do? → responds correctly

Tasks that hang/loop:

  1. Code modification based on feedback:

Based on my usage feedback, please modify the code: [detailed feedback]

  1. Generating comprehensive documentation:

Generate detailed technical documentation based on the current project

Expected Behavior

The extension should process complex tasks by breaking them into steps and executing tool calls sequentially.

Actual Behavior

The extension enters a loop state where it:

  • Repeatedly outputs partial responses
  • Shows "Spelunking..." but doesn't progress
  • Fails to execute tool calls
  • Requires manual cancellation (Escape/Ctrl+C)

Diagnostic Findings

  • No custom hooks configured
  • No MCP servers configured
  • Issue persists after:
  • Opening new window
  • Disabling think mode
  • Starting fresh conversation
  • Simple tasks in the same session work fine, confirming API/network is not the issue

Workaround

Breaking complex tasks into smaller manual steps works, but this defeats the purpose of autonomous task handling.

Screenshots

<img width="612" height="829" alt="Image" src="https://github.com/user-attachments/assets/81bc6078-b47c-4c84-9dcf-43b5ddc69836" />

<img width="786" height="472" alt="Image" src="https://github.com/user-attachments/assets/570cfa39-87fd-47d3-927e-a93fc8bfc1c9" />

<img width="891" height="411" alt="Image" src="https://github.com/user-attachments/assets/a0c53ef2-0781-4632-8a99-f4b6ae56dd39" />

What Should Happen?

Environment

  • Claude Code version: 2.1.11
  • VSCode version: [请填写你的版本]
  • OS: macOS (Darwin 24.2.0)
  • Model: claude-opus-4-5-20251101 (also tested with other models)

Description

The VSCode extension hangs or enters a loop when processing complex multi-step tasks, while simple tasks work normally.

Symptoms

  • Model repeatedly outputs the same text
  • Gets stuck at "Spelunking..." status
  • Appears to be in a "thinking loop" - generating reasoning but not executing tool calls
  • Eventually times out or requires manual interruption

Reproduction Steps

Tasks that work normally:

  1. Simple greeting: hello → responds correctly
  2. Simple command: run ls → executes correctly
  3. Reading small files: read config.py → works fine
  4. Reading large files: read docs/TECHNICAL_DOCUMENTATION.md → works (shows summary)
  5. Simple questions: what does this project do? → responds correctly

Tasks that hang/loop:

  1. Code modification based on feedback:

Based on my usage feedback, please modify the code: [detailed feedback]

  1. Generating comprehensive documentation:

Generate detailed technical documentation based on the current project

Expected Behavior

The extension should process complex tasks by breaking them into steps and executing tool calls sequentially.

Actual Behavior

The extension enters a loop state where it:

  • Repeatedly outputs partial responses
  • Shows "Spelunking..." but doesn't progress
  • Fails to execute tool calls
  • Requires manual cancellation (Escape/Ctrl+C)

Diagnostic Findings

  • No custom hooks configured
  • No MCP servers configured
  • Issue persists after:
  • Opening new window
  • Disabling think mode
  • Starting fresh conversation
  • Simple tasks in the same session work fine, confirming API/network is not the issue

Workaround

Breaking complex tasks into smaller manual steps works, but this defeats the purpose of autonomous task handling.

Screenshots

<img width="612" height="829" alt="Image" src="https://github.com/user-attachments/assets/81bc6078-b47c-4c84-9dcf-43b5ddc69836" />

<img width="786" height="472" alt="Image" src="https://github.com/user-attachments/assets/570cfa39-87fd-47d3-927e-a93fc8bfc1c9" />

<img width="891" height="411" alt="Image" src="https://github.com/user-attachments/assets/a0c53ef2-0781-4632-8a99-f4b6ae56dd39" />

Error Messages/Logs

Steps to Reproduce

Environment

  • Claude Code version: 2.1.11
  • VSCode version: [请填写你的版本]
  • OS: macOS (Darwin 24.2.0)
  • Model: claude-opus-4-5-20251101 (also tested with other models)

Description

The VSCode extension hangs or enters a loop when processing complex multi-step tasks, while simple tasks work normally.

Symptoms

  • Model repeatedly outputs the same text
  • Gets stuck at "Spelunking..." status
  • Appears to be in a "thinking loop" - generating reasoning but not executing tool calls
  • Eventually times out or requires manual interruption

Reproduction Steps

Tasks that work normally:

  1. Simple greeting: hello → responds correctly
  2. Simple command: run ls → executes correctly
  3. Reading small files: read config.py → works fine
  4. Reading large files: read docs/TECHNICAL_DOCUMENTATION.md → works (shows summary)
  5. Simple questions: what does this project do? → responds correctly

Tasks that hang/loop:

  1. Code modification based on feedback:

Based on my usage feedback, please modify the code: [detailed feedback]

  1. Generating comprehensive documentation:

Generate detailed technical documentation based on the current project

Expected Behavior

The extension should process complex tasks by breaking them into steps and executing tool calls sequentially.

Actual Behavior

The extension enters a loop state where it:

  • Repeatedly outputs partial responses
  • Shows "Spelunking..." but doesn't progress
  • Fails to execute tool calls
  • Requires manual cancellation (Escape/Ctrl+C)

Diagnostic Findings

  • No custom hooks configured
  • No MCP servers configured
  • Issue persists after:
  • Opening new window
  • Disabling think mode
  • Starting fresh conversation
  • Simple tasks in the same session work fine, confirming API/network is not the issue

Workaround

Breaking complex tasks into smaller manual steps works, but this defeats the purpose of autonomous task handling.

Screenshots

<img width="612" height="829" alt="Image" src="https://github.com/user-attachments/assets/81bc6078-b47c-4c84-9dcf-43b5ddc69836" />

<img width="786" height="472" alt="Image" src="https://github.com/user-attachments/assets/570cfa39-87fd-47d3-927e-a93fc8bfc1c9" />

<img width="891" height="411" alt="Image" src="https://github.com/user-attachments/assets/a0c53ef2-0781-4632-8a99-f4b6ae56dd39" />

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.11

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗