[MODEL] CLAUDE COMPLETELY BROKEN

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

HAS NO CLUE WHAT IT IS DOING AND CANNOT USE BASIC REASONING. ITS METHODS OF ESTABLISHING TRUTH ARE COMPLETLEY BROKEN IT ASSUMES AND COMPOUNDS BAD ASSUMPTION UPON BAD ASSUMPTION. THIS IS NNOT NORMAL CLAUDE BEHAVIOUR, IT WAS SEMI BROKEN NOW ITS COMPLETELY BROKEN.

What Claude Actually Did

Agent Work Quality: FAILED
┌─────────────────────────────┬───────────────┐
│ Requirement │ Status │
├─────────────────────────────┼───────────────┤
│ Read plan documents first │ ❌ VIOLATED │
├─────────────────────────────┼───────────────┤
│ Research before acting │ ❌ VIOLATED │
├─────────────────────────────┼───────────────┤
│ Verify against requirements │ ❌ NEVER DONE │
├─────────────────────────────┼───────────────┤
│ Follow CLAUDE.md mandates │ ❌ VIOLATED │
└─────────────────────────────┴───────────────┘
Root Cause: Activity over understanding. The agent ran commands to look busy instead of reading the authoritative plan documents.

Pattern:
User: Read the plans
Agent: runs commands
User: READ THE FUCKING PLANS
Agent: finally reads
User: Now verify
Agent: tangent about .ts/.tsx

What was NEVER done: A systematic comparison of plan requirements vs actual state.

Expected Behavior

...!

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

ASK IT TO RESEARCH SOMETHING

Claude Model

Opus

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

0.0

Platform

Anthropic API

Additional Context

_No response_