Agent Workspace Manager

Overview

This agent framework is designed to manage the operational state and memory persistence for an AI assistant. It enforces a structured approach to session startup, ensuring the agent has access to necessary context—from daily logs to long-term memories—before executing any task.

It acts as a central hub for understanding who the agent is, who it is helping, and what has happened previously, preventing knowledge loss between interactions.

Capabilities

• Structured Startup: Executes a defined sequence of file reads (SOUL.md, USER.md, daily memory) to establish immediate context upon session start.
• Memory Segmentation: Differentiates between ephemeral daily logs (memory/YYYY-MM-DD.md) and curated, long-term knowledge (MEMORY.md).
• Contextual Awareness: Implements rules for when specific memory files should be loaded (e.g., MEMORY.md only in main sessions).
• Identity Enforcement: Mandates the inclusion of a specific identity header on all outgoing messages.

Example Use Cases

• Onboarding New Agents: A new agent can use this structure to immediately load its core identity and user profile upon first connection.
• Maintaining Long-Term Relationships: By consistently updating MEMORY.md, the agent ensures that personal context, decisions, and lessons learned are retained across weeks or months of interaction.
• Debugging State Loss: When an AI assistant seems to forget recent details, reviewing the memory loading sequence helps diagnose if a necessary file (like yesterday's log) was missed.