[BUG] claude code CLI

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] Opus 4.6 hallucinating fake tool calls and ignoring user input in Claude Code CLI

Environment

• Platform: macOS
• Model: Claude Opus 4.6
• Claude Code: Latest version (freshly reinstalled)

Bug Description

Opus 4.6 in Claude Code CLI does not respond to user input normally. Instead, it:

1. Generates fake function calls in XML format that are never executed
2. Produces meta-analysis of the conversation context instead of answering
3. Then says "I don't have file system access" — even though Claude Code CLI does

Sonnet works perfectly with the same setup. The issue is Opus-specific.

Reproduction

1. Open Claude Code CLI → set model to Opus 4.6
2. Type any simple message (e.g., "Hello, can you hear me?")
3. Opus responds with hallucinated context analysis or fake tool calls instead of a normal reply

Example

User: "Can you hear me?"

Opus 4.6:

The total context of the conversation is 1191 tokens... Let me break down the key elements... I don't have enough context to provide a meaningful response.

Sonnet (same input): Responds normally.

What Should Happen?

<img width="866" height="671" alt="Image" src="https://github.com/user-attachments/assets/aeca8c17-fc8d-491a-bf54-f02886f498b3" />

Error Messages/Logs

Steps to Reproduce

[BUG] Opus 4.6 hallucinating fake tool calls and ignoring user input in Claude Code CLI

Environment

• Platform: macOS
• Model: Claude Opus 4.6
• Claude Code: Latest version (freshly reinstalled)

Bug Description

Opus 4.6 in Claude Code CLI does not respond to user input normally. Instead, it:

1. Generates fake function calls in XML format that are never executed
2. Produces meta-analysis of the conversation context instead of answering
3. Then says "I don't have file system access" — even though Claude Code CLI does

Sonnet works perfectly with the same setup. The issue is Opus-specific.

Reproduction

1. Open Claude Code CLI → set model to Opus 4.6
2. Type any simple message (e.g., "Hello, can you hear me?")
3. Opus responds with hallucinated context analysis or fake tool calls instead of a normal reply

Example

User: "Can you hear me?"

Opus 4.6:

The total context of the conversation is 1191 tokens... Let me break down the key elements... I don't have enough context to provide a meaningful response.

Sonnet (same input): Responds normally.

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

4.6

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_