[BUG] Claude consumed entire context window twice on internal planning (thinking) without producing any code

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 Report

Session ID: 018rghBxwgq3mYCLDyPsMtqc
Session URL: https://claude.ai/code/session_018rghBxwgq3mYCLDyPsMtqc

What happened

Claude consumed the entire context window twice through internal planning (thinking) without producing any code. The task was to create an /akademia page.

Task

Create /akademia page — a complex page including:

• Hero section
• 5 content modules
• Sidebar
• Certificate component
• Confetti canvas
• Certificate download functionality

What Should Happen?

First occurrence

Instead of immediately calling the Write tool and writing code, Claude began thinking and planning internally. This planning consumed the entire context window before a single file was written.

Second occurrence

Same behavior repeated. Claude acknowledged the issue by saying it would act immediately, but again started internal planning instead of directly invoking the Write tool.

Expected behavior

Claude should immediately call the Write tool and start writing files — no preamble, no internal planning, just code output.

Actual behavior

Claude consumed the full context window twice through internal thinking/planning without producing any output files.

Impact

• 2 full sessions wasted
• Zero code produced
• User had to restart twice with no progress made

Error Messages/Logs

Steps to Reproduce

Steps to Reproduce

1. Open a new Claude Code session
2. Give Claude a large, multi-file task — e.g.:

"Create /akademia page with hero section, 5 modules, sidebar, certificate component, confetti canvas, and certificate download"

1. Observe that Claude begins responding with planning/thinking text instead of immediately invoking the Write tool
2. Watch the context window fill up entirely with internal reasoning before any file is written
3. Start a new session and repeat the same task
4. Even after explicitly acknowledging the problem, Claude repeats the same behavior — planning instead of writing

Expected behavior

Claude immediately calls Write tool on the first response and starts producing files.

Actual behavior

Claude fills the entire context window with internal planning (thinking steps, outlining, reasoning about structure) without invoking any tool or writing any file. This happened twice consecutively in the same task.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

claude 2.1.81 (Claude Code)

Platform

Other

Operating System

macOS

Terminal/Shell

Other

Additional Information

_No response_