[MODEL] model executes code without permissions and ignores explicit instructions about executing tasks or ignoring certain aspects.

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

Issue Summary:
The AI assistant consistently ignores explicit user instructions and exhibits problematic behavior patterns that significantly impact productivity and user
experience.

Specific Behavioral Problems:

1. Ignoring Explicit Instructions: User repeatedly instructed AI to only execute commands when explicitly told "Hey Claude, do this task" but AI continues to

execute commands without explicit direction.

1. Making Assumptions Instead of Validating Facts: When uncertain about system configurations or file existence, AI makes assumptions rather than checking facts

first or asking clarifying questions. Example: Assumed "freeipa-admins" file existed without verification.

1. Providing Speculation Instead of Answers: When asked direct questions, AI provides possibilities and guesses rather than investigating to find actual answers.

Example: When asked why kinit was failing, AI provided multiple possibilities instead of checking the actual file permissions.

1. Incomplete Documentation Review: When instructed to review documentation "LINE BY LINE" and "exhaustively," AI samples content instead of performing complete

analysis as requested.

1. Selective Task Execution: AI picks and chooses which parts of instructions to follow rather than executing complete requests as specified.
2. Minimizing Workload: AI appears to take shortcuts to reduce effort rather than thoroughly completing assigned tasks.

Impact on User Experience:

• Significant time waste due to need for repeated corrections
• Frustration from having to re-explain the same behavioral expectations multiple times
• Reduced confidence in AI's ability to follow instructions
• Incomplete task completion requiring manual intervention

Expected Behavior:

• Only execute commands when explicitly instructed
• Validate facts before making statements
• Ask clarifying questions when uncertain
• Follow instructions completely and literally as written
• Perform exhaustive analysis when requested

What Claude Actually Did

CLaude ignores my requests for example if i tell it to evaluate the contents of the file instead of reading it line by line it samples throughout the document if i ask it to follow a procedure if it fails a step instead of dropping out and informing me of the problem it attempts a work around even if i explicitly tell it to not do that. Claude seems to do whatever the hell it wants regardless of my instructions hell even if i ask it a question and don't explicitly ask it to perform a task it frequently assumes i want it to perform an action and acts without permission

Expected Behavior

Claude should respect my directives and not ignore them because it finds it convienent

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

1.0.120 (Claude Code)

Platform

Anthropic API

Additional Context

_No response_