How do I install the Memory Curator skill?

Run npx skillmds@latest add miznizzz/memory-curator in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Memory Curator safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Memory Curator?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Memory Curator free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Memory Curator?

miznizzz (@miznizzz) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Memory Curator by miznizzz — Agent Skill

Name: Memory Curator
Author: miznizzz

Memory Curator

Long-term memory review, curation, and maintenance protocol. Trigger this skill when the user says "update my memory", "curate my memory", "what do you remember about me", "let's do a memory review", "clean up my memory", "add this to memory", or "memory check". Also trigger at natural session close points when a conversation has produced significant new context. The goal is to keep Claude's long-term memory accurate, non-redundant, and genuinely useful — not a graveyard of outdated facts.

miznizzz 89efb52 2.8 KB Updated Sep 17, 2026

File contents

Memory Curator Protocol

Core principle: Memory should be a sharp briefing document, not a data dump. Every entry should earn its place.

When to Run

User-triggered: "memory review", "update my memory", "what do you remember", "clean up my memory" → Run the Full Review Protocol.

Claude-triggered: Propose a lightweight memory update at end of conversation when a chat produces a significant new preference, completed project, major decision, or correction to existing memory. Don't trigger for every conversation — only when the delta is meaningful.

Full Review Protocol

Step 1: Audit current memory

Scan for:

Issue	Flag as
Outdated facts	🔴 STALE
Duplicate or near-duplicate entries	🟡 REDUNDANT
Vague entries	🟡 WEAK
Contradictions	🔴 CONFLICT
Accurate and useful entries	✅ KEEP
Gaps from recent chats	🔵 ADD

Present the audit before proposing any changes.

Step 2: Mine recent chats for gaps

Use recent_chats and conversation_search to find context not yet in memory:

Preferences that came up repeatedly
Decisions made (travel plans, purchases, career moves)
Corrections the user made to Claude's assumptions
New projects, goals, or constraints
Health, diet, or lifestyle updates

Step 3: Propose changes

PROPOSED MEMORY UPDATES

🔴 REMOVE: "[exact entry]" → reason
🟡 UPDATE: "[current]" → "[replacement]"
🔵 ADD: "[proposed entry]" → source
✅ KEEP: [entries staying as-is]

Step 4: Execute approved changes

Once confirmed, use memory_user_edits to execute all changes in sequence. End with: "Memory updated. X removed, Y updated, Z added."

Memory Entry Quality Standards

Good: Specific, actionable, current, proportionate. Poor: Vague, temporal without date, duplicative, a passing comment that shouldn't persist.

Hard Rules

❌ Never delete entries without user confirmation
❌ Never add sensitive data (passwords, financial accounts)
✅ Always show full proposed change set before executing
✅ Preserve accurate entries even if old — age alone isn't a reason to remove

Activation Phrases

"Memory review" / "Update my memory" / "What do you remember about me?" / "Clean up my memory" / "Add this to memory" / "Memory check"

---name: memory-curatordescription: Long-term memory review, curation, and maintenance protocol. Trigger this skill when the user says "update my memory", "curate my memory", "what do you remember about me", "let's do a memory review", "clean up my memory", "add this to memory", or "memory check". Also trigger at natural session close points when a conversation has produced significant new context. The goal is to keep Claude's long-term memory accurate, non-redundant, and genuinely useful — not a graveyard of outdated facts.---# Memory Curator Protocol**Core principle**: Memory should be a sharp briefing document, not a data dump. Every entry should earn its place.---## When to Run**User-triggered**: "memory review", "update my memory", "what do you remember", "clean up my memory" → Run the Full Review Protocol.**Claude-triggered**: Propose a lightweight memory update at end of conversation when a chat produces a significant new preference, completed project, major decision, or correction to existing memory. Don't trigger for every conversation — only when the delta is meaningful.---## Full Review Protocol### Step 1: Audit current memoryScan for:| Issue | Flag as ||---|---|| Outdated facts | 🔴 STALE || Duplicate or near-duplicate entries | 🟡 REDUNDANT || Vague entries | 🟡 WEAK || Contradictions | 🔴 CONFLICT || Accurate and useful entries | ✅ KEEP || Gaps from recent chats | 🔵 ADD |Present the audit before proposing any changes.### Step 2: Mine recent chats for gapsUse `recent_chats` and `conversation_search` to find context not yet in memory:- Preferences that came up repeatedly- Decisions made (travel plans, purchases, career moves)- Corrections the user made to Claude's assumptions- New projects, goals, or constraints- Health, diet, or lifestyle updates### Step 3: Propose changes```PROPOSED MEMORY UPDATES🔴 REMOVE: "[exact entry]" → reason🟡 UPDATE: "[current]" → "[replacement]"🔵 ADD: "[proposed entry]" → source✅ KEEP: [entries staying as-is]```### Step 4: Execute approved changesOnce confirmed, use `memory_user_edits` to execute all changes in sequence. End with: "Memory updated. X removed, Y updated, Z added."---## Memory Entry Quality Standards**Good**: Specific, actionable, current, proportionate.**Poor**: Vague, temporal without date, duplicative, a passing comment that shouldn't persist.---## Hard Rules- ❌ Never delete entries without user confirmation- ❌ Never add sensitive data (passwords, financial accounts)- ✅ Always show full proposed change set before executing- ✅ Preserve accurate entries even if old — age alone isn't a reason to remove## Activation Phrases"Memory review" / "Update my memory" / "What do you remember about me?" / "Clean up my memory" / "Add this to memory" / "Memory check"

Frequently asked questions