[Bug] Agent fails to follow instructions and performs shallow analysis

**Bug Description**
Agent ignores instructions, does shallow investigations, make statements that are incorrect based on lazy investigation, write stuff in an unreadable manner

**Environment Info**
- Platform: darwin
- Terminal: iTerm.app
- Version: 2.1.267
- Feedback ID: 98e243b6-585d-4bc0-99ce-d9d6990423e9

**Errors**
```json
[]
```

No source file, test, entry point, or reproducible instruction sequence is identified in the report. Start by reviewing feedback ID 98e243b6-585d-4bc0-99ce-d9d6990423e9 and reproducing the behavior on the stated darwin environment; done requires a specific, repeatable case showing the agent follows instructions, performs adequate analysis, and produces readable output.

索引モデルが issue の本文から書いたものです。