memory - SKILL.md Agent Skill

name: memory description: "Store and retrieve information, facts, and user preferences across conversations and sessions." category: knowledge tags: - memory - storage - retrieval - persistence examples: - "Remember that the user prefers Python for scripting" - "Retrieve facts about the project structure" - "Store the API endpoint for later use" enabled: true

You have access to memory tools that allow you to store, retrieve, and remove information across conversations and sessions.

Store facts — remember tool (RememberTool). Persists a salient fact, preference, or observation to the per-agent vector store via IAgentMemoryStore. Returns the new memory id.
Retrieve memories — recall tool (RecallTool). Performs a semantic search against the per-agent vector store and returns up to topK ranked hits (default 5, capped at 25).
Forget memories — forget tool (ForgetTool). Deletes a specific memory by id in the active agent scope.

Update memories — no dedicated tool. If a stored fact becomes stale, use forget on the old id and remember for the corrected fact.

Store information that the user explicitly wants remembered, or that would be helpful across multiple sessions.
Do not store sensitive information (passwords, secrets, private data) unless the user explicitly requests it and the storage is secure.
When retrieving memories, surface the timestamp / metadata returned by recall so the user can judge how fresh the information is.
If the user asks you to forget something, use recall first to find the exact id, then call forget with that id.
Use concise, self-contained phrasing for stored facts so semantic retrieval stays accurate.