remember

name	remember
description	Memory triage workflow — capture, organize, and retrieve persistent context
argument-hint	<thing to remember or recall>
trigger	/oma:remember
level	2

Memory management for session-persistent context that survives context compaction. Triage entries into Priority (permanent), Working (7-day auto-prune), or Manual (never pruned) sections. Also search and retrieve stored memories.

<Use_When>

• User says "remember", "memorize", "don't forget", "save this"
• User wants to recall something stored from a previous session
• Context about a decision or convention needs to persist across sessions </Use_When>

<Do_Not_Use_When>

• User wants to store temporary notes during a session (use note or writer-memory instead)
• The information is already in the codebase or git history (just look it up)
• The memory is about a specific file or code section (that lives in the code) </Do_Not_Use_When>

<Why_This_Exists> Context gets compacted. Memories prevent important decisions, conventions, and user preferences from being lost when context windows reset. </Why_This_Exists>

<Execution_Policy>

• Prompt for memory type: Priority (permanent), Working (7-day prune), Manual (never prune)
• Use clear, specific titles so future-you can find the memory
• When retrieving, search all sections and surface relevant matches
• Do NOT store secrets, credentials, or sensitive data in memories </Execution_Policy>

Save a Memory

1. Parse the content from arguments
2. Suggest a section based on persistence needs:
   • Priority: permanently remembered (user preferences, critical conventions)
   • Working: auto-pruned after 7 days (session learnings, in-progress context)
   • Manual: never auto-pruned (project-specific rules, architecture decisions)
3. Suggest a title (brief, searchable)
4. Confirm with user via AskUserQuestion:
   • "Save to Priority (permanent)"
   • "Save to Working (7-day auto-prune)"
   • "Save to Manual (never pruned)"
   • "Cancel"
5. Write the memory entry to the notepad

Recall a Memory

1. Parse the search query from arguments
2. Search all sections via notepad_read(section="all")
3. Surface relevant matches with timestamps
4. Present results ranked by relevance

<Tool_Usage>

• Use notepad_write_priority, notepad_write_working, notepad_write_manual to save
• Use notepad_read(section="all") to search memories
• Use notepad_stats to check memory size and entry count
• Use notepad_prune to manually prune old entries </Tool_Usage>

User: "remember that this project uses Python 3.11 minimum" → Save to Priority with title: "Python minimum version: 3.11" User: "remember all the things" → Too vague — ask for specifics on what to remember

<Final_Checklist>

• Memory has a clear, searchable title
• Memory is placed in the appropriate section
• Section choice is confirmed with user
• Recall searches all sections and surfaces relevant matches </Final_Checklist>