rem-sleeplisted
Install: claude install-skill thdelmas/rem-sleep
# REM Sleep — memory consolidation, learning & emotional regulation
A systematic method for an AI agent to do what sleep does for a brain: take the raw episodic trace of recent experience and turn it into durable, well-organized, low-noise long-term memory — while detaching the emotional charge from the events so the lessons survive but the friction doesn't.
Three jobs, borrowed from the neuroscience of sleep:
- **Memory consolidation** — move what matters from the session (short-term) into the long-term store, in atomic, retrievable form.
- **Learning** — integrate: cross-link memories, and surface patterns no single episode contained.
- **Emotional regulation** — recalibrate. REM detaches the affective charge from the memory of an event; you keep the lesson, not the sting.
This is the **inward** counterpart to outward-crawling discovery skills. You don't reach into the world — you process your own day.
## The long-term store
The skill operates on a **file-based agent memory store**. The default convention (and the one the examples use):
- An index file — `MEMORY.md` — one line per memory: `- [Title](slug.md) — hook`.
- One fact per file, kebab-case slug, with frontmatter:
```markdown
---
name: <slug>
description: <one-line summary — used to decide relevance at recall>
metadata:
type: user | feedback | project | reference
---
<the fact. For feedback/project, follow with **Why:** and **How to apply:** lines.
Link related memories with [[other-slug]