persistent-memory

To save durable facts and preferences, you can ask the agent to remember specific details like communication styles or tool quirks. It captures and curates these into a local memory store, distinguishing stable facts from one-off context to maintain workflow consistency across chats.

Managing long-lived project details requires durable memory that persists across sessions. This skill applies to chats where you provide environment configurations, using identity versus project scoping to store stable information and ensure your real preferences are retained.

Yes, you can separate user-level identity preferences from project-specific tooling memory. The skill guides choosing the appropriate memory scope, ensuring durable facts are categorized correctly as either personal identity or project environment details for accurate retrieval.

You can update or remove outdated information from your local memory store by asking the agent to correct stable behavior. The skill uses add and remove operations to keep memory entries compact, editable, and accurate over time as your preferences change.

You should not use durable memory for one-off context that should not persist. The skill distinguishes stable, project, and environment facts from temporary chat details, ensuring only long-lived, durable user facts are saved to maintain workflow consistency without clutter.