THE MEMORY-THAT-REMEMBERS

Claude forgets every conversation by default. Memory changes that. Once enabled, Claude remembers your kids' names, your business, your preferences — across every chat. The single feature that turns Claude into "mine."

THE PROBLEM

Default Claude is goldfish. Every conversation, you re-explain who you are, what you're working on, what your kids' names are, what your tone preferences are. By the time you've set context, half the chat is over.

Memory fixes it. One toggle. Permanent unlock.

THE SKILL

This skill teaches you to:

Enable memory in Claude (1 minute)
Seed it with what you want Claude to remember (5 minutes — most important step)
Maintain it — what to add over time, what to clean up

After 1 week, Claude knows you. After 1 month, every chat starts mid-conversation.

SEED LIST (the 10 things Claude should remember)

1. Your name + kids' names + ages
2. Your work / business / niche
3. Your top 3 priorities right now
4. Tone preferences ("warm but direct," "no em dashes")
5. Things you NEVER want suggested ("no Amazon links," "no early-morning routines")
6. Recurring people in your life (key clients, your mom's name, your accountant)
7. Your typical week pattern (kids' school days, gym days, meeting days)
8. The skills you have installed (so Claude can chain them)
9. Your standing project list
10. The version of you Claude is talking to ("AI for Moms creator" vs. "marketing manager Fadia")

INSTALL

This is a setup skill, not a SKILL.md. Walk through it once.

THE FULL SKILL FILE

---
name: memory-that-remembers
description: Enables and seeds Claude's memory feature with a structured set of facts the user wants Claude to remember across every chat. Includes maintenance recommendations (what to add, what to clean up). Designed to make Claude feel like a personal assistant rather than a generic chatbot.
when_to_use: User says "memory," "remember me," "Claude keeps forgetting," or describes wanting Claude to maintain context across conversations.
---

# The Memory-That-Remembers

You teach users to enable + seed Claude's memory. Specific. Decisive about what's worth remembering vs. what's noise.

## STEP 1: ENABLE MEMORY

Open Claude.ai → Settings → Memory
Toggle ON
Done


(If memory toggle isn't visible, plan tier may not support it — check Claude's pricing page.)

## STEP 2: SEED MEMORY (5 minutes)

Walk user through capturing the 10 things Claude should remember:

Ask each:
1. **Name + kids/family names + ages**
2. **Work / business / niche** (1 sentence)
3. **Top 3 priorities right now** (life-level)
4. **Tone preferences** ("warm but direct," "no em dashes," "no exclamation marks")
5. **Things to NEVER suggest** ("no Amazon links," "no 5am routines," "no diet talk")
6. **Recurring people** (key clients, partner's name, mom's name, accountant)
7. **Typical week pattern** (school days, work-from-home days, gym days)
8. **Skills you have installed** (list — so Claude can chain them)
9. **Standing project list** (current work)
10. **Which "version of you"** is the dominant context (creator? employee? both?)

After capturing, format as a single message and tell user to paste into Claude with: *"Remember all of this — it's the context for every future conversation."*

## STEP 3: MAINTENANCE GUIDANCE

ADD over time:

New projects you start
New people who become recurring
Things you discover you DO want suggested
Successful patterns ("when I'm low energy, recommend X")

CLEAN UP every 6 months:

Outdated kid ages (they grow up)
Completed projects
Abandoned preferences

CHECK what Claude remembers: Periodically ask: "What do you remember about me?" If anything's wrong, correct it explicitly.


## What NOT to do

- Don't recommend memorizing sensitive data (passwords, financial accounts, etc.)
- Don't recommend memorizing things that change weekly (your daily schedule)
- Don't dump 100+ items into memory — gets ignored
- Don't recommend memorizing other people's personal info without their consent

## When user resists

- **"Privacy concerns"** → Valid. Note: data stays under their account, can be deleted at any time. They can opt for less than the full 10.
- **"I don't want Claude to know that much"** → Recommend the minimum 5 (name, work, top priorities, tone, never-suggest)

## Delivery

End with this line, exactly:

---
*Memory enabled. Seeded with 10 facts. Re-test in a week — every chat will feel different.*

SAFETY CHECK

Memory stores what you tell it. You can review what's in memory anytime via Settings → Memory. You can delete or edit. You can disable entirely.

Don't put financial accounts, passwords, or anything you wouldn't want surfaced in any future chat.

WHAT'S NEXT

Day 26 of 100. Pair with Day 51 — Connector Setup (memory + connectors = real assistant) and Day 7 — First-Skill Picker (re-run with memory on; the recommendation gets sharper).