elite-longterm-memory

Elite Longterm Memory is a persistent memory architecture designed for AI agents that need reliable long-term context retention across sessions. The skill combines multiple memory strategies into a layered system that helps agents avoid losing important decisions, repeating mistakes, or forgetting user preferences. It is built around a multi-tier architecture that separates active working memory, semantic vector storage, structured historical records, and curated long-term archives. The system also incorporates a Write-Ahead Log (WAL) style protocol to improve resilience and preserve active state even during compaction or memory updates.

The skill provides five integrated memory layers including SESSION-STATE.md for active task tracking, LanceDB vector storage for semantic search and automatic recall, Git-Notes based knowledge graphs for structured and branch-aware memory management, and curated human-readable archives through MEMORY.md and daily logs. It supports automatic context recall, manual memory storage, persistent decision tracking, and optional cloud backup through the SuperMemory API. The architecture is designed to support workflows where AI agents continuously accumulate operational knowledge over time while maintaining searchable and structured memory records.

This skill is intended for developers building AI assistants, autonomous agents, or long-running automation systems using platforms such as Clawdbot, Moltbot, Claude, or GPT-based agents. Common use cases include persistent coding assistants, project management agents, research assistants, and conversational systems that require continuity across interactions. It is especially useful for environments where maintaining historical context, user preferences, and technical decisions is critical to performance and reliability.