# Memory Curator

> Use when deciding whether task discoveries, decisions, preferences, architecture, failures, or lessons deserve durable repository memory.

- Skill: `ihabkhaled/memory-curator` (Agent Skill)
- Install (CLI): `npx skillmds@latest add ihabkhaled/memory-curator`
- Raw SKILL.md: https://api.skillmd.com/api/skills/ihabkhaled/memory-curator/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: ihabkhaled (https://skillmd.com/u/ihabkhaled)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/ihabkhaled/memory-curator

---


# Memory Curator

## Core principle

Durable memory must be stable, evidenced, reusable, expensive to rediscover, and save future tokens.

## Procedure

1. Classify the candidate as temporary state or durable knowledge.
2. Verify it against an authoritative source.
3. Test whether it is stable and repeatedly useful.
4. Choose preferences, architecture, decisions, recurring problems, lessons, or failure patterns.
5. Record source, scope, replacement condition, and concise value.
6. Reject raw reasoning, one-off errors, and speculation.

## Required output

Promotion decision, destination, evidence, scope, and replacement condition.

## Limits

Promote only when future rediscovery cost exceeds future reading cost.

## Common mistakes

Do not use memory to override current user intent or preserve session noise.

## Stop condition

Stop when the candidate is promoted concisely or explicitly left temporary.