memory-management

You can persist agent memory across multiple tasks by applying a READ-UPDATE protocol that handles pre-task retrieval and post-task updates. This ensures context, identity, and learned insights remain coherent between sessions.

The READ-UPDATE protocol manages context by reading identity, experiences, learnings, and relationships before a task starts, then updating these files with new insights after the task completes to maintain memory persistence.

Maintaining collaboration history across multiple agents requires a memory persistence system that stores shared experiences, learnings, and relationships, allowing all agents to read and update a common context.

Yes, you can enable automatic memory activation for new tasks by setting up auto-activation hooks. These hooks trigger pre-task reads to load identity and context automatically before processing begins.

Yes, initializing agent memory requires reading specific markdown templates like identity.md, experiences.md, learnings.md, and relationships.md to establish the baseline context and validate the memory structure.

Agent context breaks down without memory persistence because there is no mechanism to carry experiences, learnings, and relationships forward, causing the agent to lose its identity and previously acquired knowledge across tasks.