knowledge-graph

To give AI agents persistent memory, this skill organizes a three-layer system with entity knowledge, daily notes, and persistent MEMORY.md, ensuring durable fact extraction and reducing repeated context gathering across sessions.

Cron-driven fact extraction works by scheduling automated workflows that pull durable facts from daily notes and append them to facts.jsonl with auto-incremented IDs, keeping living summaries fresh.

The best way to structure entity lookup is fetching lightweight context from summary.md first, then loading facts.jsonl only when deeper detail is required, balancing performance and context retrieval.

Yes, you can use this memory system across multi-agent workspaces. It applies the three-layer memory architecture to single or multi-agent setups for shared durable fact extraction and context retrieval.

To set up weekly synthesis for living summaries, initialize entities with summary.md and facts.jsonl, then enable the weekly-synthesis cron as described in the skill documentation to refresh summaries.

AI agents forget context because they lack durable storage. Without persistent memory layers like entity knowledge and facts.jsonl, agents must repeatedly gather context instead of retrieving atomic facts.