What problem does it solve?
AI agents often lose important context across conversations, leading to inconsistent behavior and forgotten decisions. This skill provides a structured memory architecture that separates explicit facts from inferred patterns and organizes memories into Working, Daily, Projects, and Long-term tiers to preserve context over time.
Core Features & Use Cases
- Memory Types: Explicit facts ([FACT]) and deduced patterns ([DEDUCED]); automatically tag and promote signals based on frequency and relevance.
- Memory Tiers: In-session Working context, daily logs at memory/daily/YYYY-MM-DD.md, project files under memory/projects/, and a long-term MEMORY.md for recurring patterns.
- Session Summaries & Promotion: Append concise summaries after long conversations and promote stable patterns to long-term memory.
- Writing Deduced Memories: Record deduced conclusions with evidence to improve future decisions.
- Memory Quality & Maintenance: Use a checklist to ensure usefulness and guard against bloat; perform periodic compaction and audits.
- Self-Review & Production Notes: Regular automated reviews and production notes to improve memory handling.
Quick Start
Activate the memory-architecture workflow and begin capturing session context and deduced memories across conversations.
