[MODEL] Undesirable verbose comments due to system instruction

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

Other unexpected behavior

What You Asked Claude to Do

Just asked to fix css to scope it to a selector

What Claude Actually Did

Claude added some 10 lines of comment also on top of it which are very "that claude code session" specific and will not be understood an independent context.
<img width="2942" height="708" alt="Image" src="https://github.com/user-attachments/assets/dabf6115-5733-42f7-83b5-ea79101f0fdd" />
This has also resulted in claude refusing to change the code because the comment referenced a point in the plan and the new claude session treats it as unchangeable.

Expected Behavior

Should not add unnecessary comments BY DEFAULT - As a user I should get the most efficient system and the model should add meaningful comments only.
This can also confuse the model in a new claude code session because the comment it added will not be high quality and wrong hallucinated comment.

Files Affected

Modified a css file and added unnecessary comments.

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

_No response_

Claude Model

Sonnet

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.1.220 (Claude Code)

Platform

Anthropic API

Additional Context

_No response_