Bug: Agent memory files saved in root directory instead of project folder

[AI-J-IN]

Summary

After cloning the Raven repository and running Raven directly at the root directory, the agent's memory files (such as conversation history and skill data) are saved to the root directory instead of the Raven project folder. This causes multiple Raven instances to conflict with each other, resulting in data mixing across instances.

Steps to reproduce

1. git clone https://github.com/EverMind-AI/Raven.git
2. cd Raven
3. Run Raven directly at the root directory using default configuration
4. Use the agent to complete several tasks (generating conversation history and memory files)
5. Check the root directory (.): memory files (e.g., global_mem.txt, agent conversations) are saved to the root directory
6. Clone Raven to another location and repeat the process - memory files from both instances mix in the system root directory

Expected behavior

Memory files should be saved within the Raven project directory (e.g., Raven/memory/ or Raven/data/). Each Raven instance should maintain its own isolated memory storage so that multiple instances can run independently without data conflicts.

Actual behavior

Memory files (conversation history, agent skill data, etc.) are saved to the root directory (the directory where Raven is executed). When multiple Raven instances run from different locations, their memory data gets mixed together, causing data pollution.

Environment

OS: Windows / Linux
Python: 3.10+
Raven: latest main branch

Logs or screenshots