Memory Onboarding

Bounded institutional-memory build-out for an ecosystem with a canon/corpus. Each run, the steward agent reads the ecosystem's canon and saves the durable, load-bearing facts (thesis, tenets, refusals, stance, voice, toolkit, lineage) as institutional memories, plus short dossiers about each specialist — using the real memory tools so the write classifier runs and the facts are recallable and prompt-injected. The schedule retires after a fixed number of passes; thereafter the conversation->me...

Details

Bounded institutional-memory build-out for an ecosystem with a canon/corpus. Each run, the steward agent reads the ecosystem's canon and saves the durable, load-bearing facts (thesis, tenets, refusals, stance, voice, toolkit, lineage) as institutional memories, plus short dossiers about each specialist — using the real memory tools so the write classifier runs and the facts are recallable and prompt-injected. The schedule retires after a fixed number of passes; thereafter the conversation->memory extractor and the agents' own saves maintain memory. Designed so a canon exercises the live memory system instead of sitting as inert seed rows.