memory-to-notion_skill

Overview

This skill summarizes recent chat history, extracts atomic conversation memories, and archives them into a Notion database named "Memory Store." It discovers or creates the database automatically, deduplicates and detects conflicts, and writes structured memory rows with metadata such as Category, Tags, Confidence, Scope, and Memory ID. After syncing, it returns a concise report of created, updated, and skipped entries.

How this skill works

On trigger, the skill searches the user's Notion workspace for a database titled "Memory Store" and creates it under a specified page if missing. It gathers conversation content (recent chats or current session), decomposes conversations into one-fact-per-row memory entries, checks the Memory Store for duplicates or conflicts, and then creates or updates pages via Notion APIs. Finally, it summarizes results including counts, MEM IDs, and any conflicts resolved.

When to use it

• When you want to archive recent conversations to Notion for long-term recall.
• After a planning session to record decisions, preferences, or project context.
• When you need a structured knowledge base of user preferences and technical facts.
• Before switching projects so project-specific memories are captured.
• When you want a periodic backup of conversation-derived knowledge to Notion.

Best practices

• Keep each memory atomic: one fact, decision, preference, or skill per row.
• Use clear declarative Titles so entries are searchable across platforms.
• Default Scope to Global unless the memory is project-specific; fill Project when Scope=Project.
• Set Confidence based on how explicitly the user stated the information: High / Medium / Low.
• Avoid storing transient small talk, failed attempts, or anything the user asked to forget; ask when unsure.

Example use cases

• Archive technical decisions from a design session (e.g., chosen frameworks, DB).
• Capture user tooling and formatting preferences (editor, linter, CI settings).
• Record project-specific configuration or deployment notes tied to a project name.
• Turn multi-turn troubleshooting into discrete actionable memories (commands, configs).
• Create a personal memory index for recurring workflows and patterns.