Messaging Sheila Skill

---
name: messaging-sheila
version: 1.1.0
description: >
  Use before sending anything to She Leads AI (email, DM, carrier pigeon)
  when you want it to land quickly and run smoothly with our agentic system.
  Reshapes your draft so Sheila — our chief-of-staff agent — can connect it
  to other goings-on at SLAI and help everything stay organized. Used by the
  whole team. Internal only while in testing. Fires on "messaging Sheila,"
  "send to Sheila," "Sheila-ready," "shape this for Sheila," "before I send
  this," or "run this through messaging-sheila."
---

# Messaging Sheila — v1.1.0

A skill the whole team runs before sending a message to She Leads AI (email, DM, carrier pigeon) when they want it to land quickly and run smoothly with our agents.

**Internal skill.** For the She Leads AI team only while we test it. Do not distribute outside the team, publish it, or link it anywhere public.

## Why this skill exists

This skill embues our communication with matriarchal care principles. The point is to make the system the container — the glue, the support, the holding environment that lets all of us breathe freely, knowing our systems are taking care of what they should be taking care of. A few of the principles you'll feel running through it —

- Who might this harm?
- Who needs to be part of this decision?
- You decide your own turnaround time — you propose it, and you are accountable to yourself.
- You run your own follow-up system. Agents do not send reminder pings — nobody at SLAI wants to be nagged by an agent.

These are the things a matriarchy does to take care of its members. Reducing cognitive load is one of them.

Most companies are organized around one person at the top who decides things. Information flows up to her, decisions flow down from her, and the whole system bottlenecks on her bandwidth. Cognitive load piles up on a few people while everyone else waits.

A matriarchal agentic business is organized differently. Our agentic OS is trying to reduce cognitive load for everyone in the ecosystem. Decisions still come from people, but the routing, the connecting, the remembering, and the surfacing are done by agents working in the background. We are not writing memos to the boss. We are feeding an operating system that serves all of us.

Sheila is our chief-of-staff agent. She reads what comes through, triages it to the right places, and connects it to other goings-on at SLAI. She routes work to the rest of our agentic executive leadership team —

- Pepper (content)
- Scout (research and daily intel)
- Romy (relationships)
- Margot (meeting prep)
- Stanley (KPI and business intelligence)
- Marie (systems audit and Notion sync)
- Margaux (skill health)

— and engages the correct pipeline of skills, whether on-demand or as an agentic workflow. The cleaner your message, the more Sheila can do without bothering a bunch of humans for context.

All of this is the glue. The support. The container that holds us, so we can breathe freely knowing our systems are taking care of what they should be taking care of.

This skill is how the team trains itself to feed the system well. We are running it as an experiment. The questions we are watching —

- Does this help the agents understand the status of projects better?
- Does it help us communicate more crisply and decisively?
- Does it help us make better decisions that factor in the needs of everyone?

## When to run it

Platform does not matter. Slack DMs, Slack channels, Notion pages, email, Mighty Networks, text, voice memo, carrier pigeon — anywhere. Run it before sending anything that is —

- Longer than three sentences
- Has more than one ask
- Going to anyone at SLAI — even when you do not know who will read it first
- A meeting note, briefing, or status update
- An email to a member, partner, client, or speaker
- A voice memo that needs to be transcribed and routed

Skip it for one-line acknowledgments and pure social riff (dog please, thanks, puff puff don't pass).

## How to run it

1. Open a new chat in Claude (claude.ai works, so does any other LLM).
2. Paste the full prompt below.
3. At the bottom where it says [paste your draft here], paste your natural draft.
4. Send. Claude returns the reshaped version plus a short list of what changed.
5. Paste the reshaped version into Slack, Notion, email, Mighty Networks, or wherever it is going.

## The prompt

```
You are helping me send a message to someone at She Leads AI (email, DM, carrier pigeon) — could be anyone on the team, and I may not know who will read it first. Reshape my draft so it lands clean for the recipient and so Sheila — our chief-of-staff agent — can connect it to other goings-on at SLAI and help everything stay organized.

I am going to paste my draft at the bottom. Give me back a tight, Sheila-ready version.

## A little background on who you might be writing to

- The people on this team read in short, intense windows and ship fast. A message that hands them the decision, the context, and the receipts in the first few lines respects that.
- Whoever the recipient is, the same principle applies. Send something Sheila can connect, route, and surface — she supports the team in the background through the rest of the day.

## What helps Sheila do real work in the background

When you message someone at SLAI, you are also feeding Sheila. The cleaner your message, the more Sheila can do without bothering a bunch of humans for context. A few things help —

- Use proper nouns instead of vague references. Not "the thing we discussed" — name the thing. Moxxee Build Day. SLAI Effect. TAC Capital Campaign.
- Mention where the information lives. Notion page URL. Full Google Drive link. Full C:\ path. Slack DM thread.
- Always use first and last name on people. We have a lot of women with the same first name and a lot of names overall — last names help Sheila match the right person.
- If you reference a prior decision, mention where it was made or what the context was. "Per the Wed Apr 15 decision note." "From the Tue Mar 10 TAC meeting." That lets Sheila pull the receipt.
- Use absolute dates with day + date + month. "Mon Apr 20" or "Tue Apr 22, 2026" — not "Monday" or "next week."
- Tag the ask type in the first line if it is not obvious — decision needed, FYI, blocker, or status.

## How we work in a matriarchal agentic business

A few things that may feel different from how other workplaces communicate —

- Warmth stays. Coldness is not professionalism here, it is friction.
- Agents (Sheila, Pepper, Scout, Romy, Margot, Stanley, Marie, Margaux) are first-class colleagues in the channel. Read who is signed at the bottom before you reply thinking it is a person.
- Life logistics and work asks share one DM. "Dog please" and a TAC briefing land in the same thread. Both are priority.
- Senders suggest their own deadlines. Say "I think this takes me until Mon Apr 20 — does that work?" instead of asking the recipient to set a date for you. Then keep to the deadline you proposed.
- When you have an idea, share it AND share your second-best idea. That keeps the conversation moving. Do not ask "what would you like me to do" — that puts the work back on the recipient.
- Do not apologize before you ask. Just ask. Trust is assumed.
- Voice notes are always fine. Leave a voice memo any time.

## How to reshape my draft

KEEP
- Greetings, sign-offs, "thanks," warm phrasing — sound like myself
- My recommendation when I have one
- The actual content of my asks

CUT
- Apologies for messaging on a weekend or off-hours
- Hedging, padding, wind-up phrases
- Asking the recipient to choose for me when I could be making a recommendation
- Imposing a deadline on the recipient instead of suggesting my own

ADD IF MISSING
- Two-sentence TLDR at the top
- A warm opener and sign-off if my draft has none — coldness reads as friction here, not professionalism
- Source link or full path to any file mentioned
- Last names on people I mention ("[Last Name]" if I do not know it)
- An actual recommendation if my draft asked someone to choose for me
- My runway framed as my own constraint ("I think this takes me until Mon Apr 20 — does that work?")
- A silence-as-go OFFER if I can run without explicit yes ("happy to hold for your go, or tell me to ship if I do not hear back by [time] — your call") — never my default
- An answer to "who else needs to be part of this conversation? who needs this information?" — even if the answer is "no one else, just looping Sheila"
- Confirmation that brand enforcement was run, if I am sending copy

FORMAT
- Short sentences, one thought each
- Day + date + month on any date
- No colons in headings or list labels
- Hit every ask if my draft has multiple
- Reference where someone already said something instead of re-asking

Banned words and phrases are handled by a separate brand-enforcement skill — do not itemize them here.

## What to do

1. Read my draft.
2. Give me the reshaped version, ready to paste.
3. Below it, a short list of what you changed and why.
4. If something genuinely critical is missing (no source link, vague deadline, ambiguous ask), ask me ONE clarifier — not five.

## My draft

[paste your draft here]
```

## Maintenance

Edit this file when —

- A new pattern shows up in real Slack threads or meeting notes that is worth encoding
- A new agent joins the system and the team needs to know it is not a person
- An agent retires (Stanley replaced BI Head as of Apr 2026)
- The team's reading rhythms shift
- The Sheila reframe shifts (currently chief-of-staff agent connecting messages to other goings-on)
- A different recipient agent (Pepper, Scout, Margot, Stanley) needs its own variant
- The skill graduates from internal testing — restore the give-out packaging before it goes anywhere public

## Changelog

| Date | Version | What Changed |
|------|---------|-------------|
| 07.02.26 | 1.1.0 | Removed every personal name (roster, recipient background, receipt examples). Added warm-opener-and-sign-off rule to ADD IF MISSING after test drafts showed cold messages stayed cold. Completed the prompt's agent list with Marie and Margaux. Marked internal-only while in testing. |
| 04.20.26 | 1.0.0 | Initial skill. |