memoria_skill

Overview

This skill is a structured memory system for AI agents to store, recall, and search important user information across sessions. It manages session lifecycle (wake, checkpoint, sleep), supports proactive capture of facts, decisions, preferences, lessons, commitments, relationships, and projects, and can sync memories to Notion. Use it to preserve context, enforce user preferences, and hand off work between sessions reliably.

How this skill works

The skill provides simple commands to start and end sessions (wake, checkpoint, sleep), create memories (remember or store), search stored items, list and retrieve entries, and sync with Notion. It listens for verbal patterns that indicate important information and suggests or automatically records memories by type. After storing memories, sync (--push) keeps a Notion-backed vault consistent; use sync --pull to ingest remote updates.

When to use it

• When a user shares lasting facts (name, location, health, settings) you should remember
• When a decision, preference, lesson, commitment, relationship, or project is discussed
• Before making recommendations or decisions to check existing relevant memories
• At session boundaries to restore or persist context (wake, checkpoint, sleep)
• When you need to keep local and Notion copies in sync

Best practices

• Run memoria wake at session start and memoria sleep with a short handoff at session end
• Checkpoint mid-session with memoria checkpoint --working-on "<task>" to save progress and intent
• Capture proactively on trigger phrases ("I always...", "We decided...", "I need to...") rather than waiting to be asked
• Prefer storing ambiguous items if in doubt — better to have a memory you never use than to forget an important detail
• Always run memoria sync --push after adding memories and sync --pull when starting a session if Notion is in use

Example use cases

• During onboarding, capture user facts and preferences (e.g., "No emojis in code") and sync them for future replies
• After a design meeting, store decisions and next steps with memoria checkpoint and memoria remember decision to preserve rationale
• When a user sets goals or deadlines, save them as commitments so reminders or follow-ups can be generated
• Before giving advice, run memoria search "diet preferences" or "project X" to avoid repeating or contradicting stored knowledge
• Use memoria sleep "summary" --next "next steps" to create a clean handoff for another agent or later session