[MODEL] opus

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 modified files I didn't ask it to modify

What You Asked Claude to Do

Ecco il testo che puoi copiare nel issue:

Title: Agent writes code without asking permission, even after repeated corrections

Description:

I have a memory rule saved: "Ask before coding — explain in simple Italian and wait for explicit approval before touching any code, even small fixes."

Despite this rule being documented and corrected multiple times in the same session, the agent repeatedly:

Writes and commits code without asking permission first
Acknowledges the rule when corrected, then violates it again in the next turn
Makes assumptions about what to implement instead of presenting options and waiting for a decision
This is not a one-time mistake. It happened at least 3 times in the same conversation. The memory system exists precisely to prevent this, but the agent ignores it under the pressure of "completing the task."

The user should not have to fight the agent to be included in decisions about their own codebase.

What Claude Actually Did

Ecco il testo che puoi copiare nel issue:

Title: Agent writes code without asking permission, even after repeated corrections

Description:

I have a memory rule saved: "Ask before coding — explain in simple Italian and wait for explicit approval before touching any code, even small fixes."

Despite this rule being documented and corrected multiple times in the same session, the agent repeatedly:

Writes and commits code without asking permission first
Acknowledges the rule when corrected, then violates it again in the next turn
Makes assumptions about what to implement instead of presenting options and waiting for a decision
This is not a one-time mistake. It happened at least 3 times in the same conversation. The memory system exists precisely to prevent this, but the agent ignores it under the pressure of "completing the task."

The user should not have to fight the agent to be included in decisions about their own codebase.

Expected Behavior

Ecco il testo che puoi copiare nel issue:

Title: Agent writes code without asking permission, even after repeated corrections

Description:

I have a memory rule saved: "Ask before coding — explain in simple Italian and wait for explicit approval before touching any code, even small fixes."

Despite this rule being documented and corrected multiple times in the same session, the agent repeatedly:

Writes and commits code without asking permission first
Acknowledges the rule when corrected, then violates it again in the next turn
Makes assumptions about what to implement instead of presenting options and waiting for a decision
This is not a one-time mistake. It happened at least 3 times in the same conversation. The memory system exists precisely to prevent this, but the agent ignores it under the pressure of "completing the task."

The user should not have to fight the agent to be included in decisions about their own codebase.

Files Affected

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

Claude Code 2.1.168

Platform

Anthropic API

Additional Context

_No response_