[MODEL] claude-opus-4-6

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude made incorrect assumptions about my project

What You Asked Claude to Do

I asked: Read all README.md files and the @documentation (if available) to gain context!
You need to remove the screen capture feature from the UI and adress all down
stream changes that might come along. Use /codebase-tracer to trace the
screencaputer feature to correctly erase it. /effort max

What Claude Actually Did

1. Claude did read all files and checked
2. Used the correct skill /codebase-tracer (it specifically enforces claude to check ALLd dependencies from a little change and iterate inductive and deductive reasoning.
3. Claude asses correctly that there are two pipellines with a similar name.

Now it chose both and started to simply delete all backend files that had connection to the simple UI feature that I wanted to be removed. The update is right. Claude is now more literal.

Expected Behavior

1. Claude did read all files and checked
2. Used the correct skill /codebase-tracer (it specifically enforces claude to check ALLd dependencies from a little change and iterate inductive and deductive reasoning.
3. Claude asses correctly that there are two pipellines with a similar name.
4. Finde downstream dependencies that might be used for buffering or cache storage and delete them as well.

Files Affected

I am sorry its company code that I can present.

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

No, only happened once

Steps to Reproduce

_No response_

Claude Model

Opus

Relevant Conversation

Impact

High - Significant unwanted changes

Claude Code Version

2.1.112

Platform

Anthropic API

Additional Context

Maybe long conversation. Maybe MCP (only sequential thinking and github). Context was 45% in this sesssion