Epic Sub: Session Amnesia & The Context Priming Engine

Problem

Currently, autonomous agents suffer from "Session Amnesia". When a new session boots, the agent's context window is wiped. While the semantic vector database (ChromaDB) holds all knowledge, the agent does not intrinsically know what to query upon waking up, making it dependent on manual user prompting to reload the "Golden Path" strategic context.

Proposed Solution

Build the Context Priming Engine using the Native Edge Graph. The true purpose of the native SQLite graph within the Memory Core is to act as an "Active State Topology".

1. The Frontier Node: Maintain a specific graph pointer for the [Current Frontier] (the active Epic or task).
2. Weighted Edges: Allow the offline Dream Mode (REM) to parse past sessions, create high-priority strategic concept nodes (e.g., [Concept: Golden Path]), and explicitly link them to the Frontier with heavy numerical weights.
3. The Pre-load Injection: When an agent session initializes, the Memory Core natively traverses 2 degrees outwards from the [Current Frontier] node, grabs the highest-weighted semantic pointers, and pushes the associated ChromaDB chunks directly into the agent's startup context.

Definition of Done

• The memory-core MCP Server exposes a native getContextFrontier() (or similar) initialization hook.
• Strategic context (priorities, the "Golden Path") is pre-loaded into new agent sessions without requiring a manual user semantic prompt.
• The Agent wakes up aware of its long-term strategic agenda.