skaai Protocol 04 — The Operator's Command Center

Become the operator.

You are the connective tissue of the org. Right now you run on adrenaline and memory. In two days you build an AI Chief of Staff on your real data — one that drafts your recaps, chases your action items, assembles your QBR deck, and runs your operating cadence while you do the work you were actually hired for.

Apply for Protocol 04 See what you build

24 seats · 1 cohort / month · application-only

The operator's problem

You sit at the intersection of strategy and admin, and admin keeps winning. The topic flips every thirty minutes. The notifications pile up during your back-to-backs. You carry a quiet, constant dread that something got dropped. And you already sense the truth most operators won't say out loud: you're using maybe a tenth of what these tools can actually do — treating Claude like a faster search box instead of an operator you can delegate to.

Stop prompting.Start delegating.Keep the system.

Why this works

Build, not learn. Trust is the product.

Build, not learn

Leave with the system, not the notes.

Every module ends with a working artifact on your real data — a skill, an agent, a connector, a dashboard. By Sunday evening you own a command center, not a notebook.

Trust is the product

Every session live. Every asset delivered. Zero upsell.

Two weekend days, fully live, one instructor and two floor coaches so nobody falls behind. Module 0 is a governance clinic so operators under strict data policies build the identical system safely. Honored money-back guarantee. Transparent pricing in USD and INR. No back-end funnel, ever.

Premium restraint as proof

The build is the credential.

Your capstone is a jet-black, liquid-glass live command center your principal can actually open. Restraint is the signal: precise, fast, quiet. The kind of thing that makes a CEO ask who built it.

The turn

From running on memory to running a system.

47 notifications accumulated during your back-to-backs.A scheduled brief that triages before you're at your desk.
The nagging feeling you've dropped something.A recap skill that opens every action item itself.
A working week's hours lost to manual reporting.A near-final QBR deck generated in minutes.
“It's all in my head” — and only your head.A Knowledge MCP your whole team can query.

If this is your week

The lines operators actually use. What we build against each.

47 Slack notifications accumulated during my back-to-backs.

The Monday-Morning Brain — a scheduled brief + triage agent.

That nagging feeling I've dropped something.

The Meeting-Recap & Follow-Up skill that opens the tasks itself.

I lose roughly a working week's worth of hours every week to manual reporting.

The QBR / Board Deck Generator.

Three contradictory directives within a 45-minute span. Nobody remembers what was decided.

The Operating Cadence Engine: Status Roll-Up + Decision Log.

It's all in my head.

A connected Knowledge MCP your whole team can query.

Most people use Claude like a smarter Google.

The whole weekend — from prompting to delegating.

Indispensable to the system, in the worst way.

The capstone orchestrator that runs without being asked.

What you ship

My AI Chief of Staff

One orchestrator. One command center. Built on your real data, kept forever.

Command center · one button, six agentsfires on a schedule

Run morning briefone button

Meeting-Recap
queued
Status-Roll-Up
queued
Deck-Builder
queued
Stakeholder-Brief
queued
Digest
queued
Knowledge
queued

Run morning brief

one button · 6 agents

Meeting-Recap
transcript → clean summary
Status-Roll-Up
team updates → one report
Deck-Builder
KPIs → near-final deck
Stakeholder-Brief
one-screen pre-meeting brief
Digest
the day, in five lines
Knowledge
answers, with citations

The operator brain

A CLAUDE.md standing-context file holding your quarter priorities, your principal's profile, your top stakeholders, and your decision style. Every agent and the dashboard load it automatically.

Specialized subagents

Meeting-Recap, Status-Roll-Up, Deck-Builder, and Stakeholder-Brief — each invoked on demand: transcript to clean summary, team updates to one exec report, KPIs to a near-final deck, a one-screen pre-meeting brief.

Scoped MCP connectors

Read-scoped or synthetic connectors to Gmail, Calendar, Notion, and Slack — built safely for strict orgs.

Audit-trail hooks

Every action writes to a governance log. This is the trust answer, built in, not bolted on.

Scheduled n8n triggers

The digest and roll-up run on a cadence, without being asked.

A live liquid-glass command center

Jet-black, unifying KPIs, open action items, the decision log, and team status — auto-refreshed from the data layer. The thing your principal opens.

The Saturday demo

Press one button. Six things happen. Each operator ships a 90-second live demo to the room — one trigger fires the digest, the recap, the roll-up, and a dashboard refresh in parallel. You leave with the repo, the dashboard, and a runbook so it never rots.

The ROI

The seat is a rounding error against a Chief of Staff's annual comp.
It costs less than the reporting hours you lose to manual assembly in a single busy week.
It buys owning the system outright — instead of renting a generic one that could replace you.

The two days

Two weekend days, fully live.

Day 1 — Stop the bleeding. Clone your judgment.

Attacks the daily survival pains — context-switching, dropped balls, the notes grind, tribal knowledge — and lays the four foundations: model optimization, context orchestration, skills, MCPs. You leave Day 1 with your operator brain, a permissioned knowledge assistant over your real docs, and three to four reusable skills running on your real meetings and inbox.

Day 2 — Industrialize. Ship the command center.

Single skills become orchestrated systems: scheduled agents, n8n automations, auto-built decks, a live dashboard, async video and voice updates. Everything composes into the flagship — your AI Chief of Staff orchestrator behind a jet-black command center you demo live to the room.

From 10% to Operator: Model Optimization + Governance Sandbox

Open with the brutal mirror: a CoS prompting Claude like Google, side by side with the same task delegated as a multi-step job. Teach model optimization for operators, not engineers — when to reach for Opus vs Sonnet vs Haiku and why; why a heavy multi-server MCP setup can cost a large token budget upfront while a skill stays near-zero until invoked; why “many skills, few MCPs” is the pattern; the difference between skills (instructions), MCPs (tools), and subagents (delegated agents). Then the 30-minute Governance & Safe Sandbox clinic: scoped/read-only connectors, synthetic-copy fallback, and an audit-trail-via-hooks preview. You leave with your operator brain and you pick your data tier: live scoped, permissioned, or synthetic.

Pain: Using AI at ~10%; the “our data policy won't allow this” objection.

CLAUDE.md Operator Brain + model & governance cheat sheet

Never Drop a Ball: Meeting-Recap & Follow-Up Skill

The instructor ingests a real transcript and builds a deterministic, reusable Meeting-Recap skill live — clean summary, every decision, every action item with an owner and due date, recap drafted in the right tone. Teach skill anatomy (SKILL.md, triggers, resources). Then wire the n8n flow that auto-posts the recap to Slack and opens tasks in Asana/Linear/Jira. You build it against your own transcripts and run it on a real meeting from last week.

Pain: Dropped action items; the notes grind.

Meeting Recap skill + n8n flow (auto-post to Slack, open tasks)

Your Single Source of Truth: the Knowledge MCP

MCP taught properly for operators: what a server is, connecting Notion/Drive/Slack via the registry, scoping permissions, building a citation-backed “ask-me-anything” assistant. Make the whole team able to query it so you stop being the lookup bottleneck. You connect a real or synthetic workspace and interrogate your messiest “where is this / who owns this” questions until the answers are trustworthy.

Pain: Tribal knowledge; human FAQ; key-person risk.

Connected Knowledge MCP / RAG assistant the team can query

SOP-to-Agent Factory + Stakeholder Intelligence

Turn one messy in-the-head process — “how we run the weekly ops review,” “how we onboard a hire” — into a documented SOP and an executable skill/subagent, live. Then build a Stakeholder Intelligence agent and lightweight CRM that produces a one-screen pre-meeting brief. You convert one real recurring process and seed your CRM with your five most important stakeholders.

Pain: “It's all in my head”; influence without authority.

SOP-skill library + Stakeholder Intelligence agent & CRM

The Monday-Morning Brain: Scheduled Digest + Triage

A scheduled agent that crawls calendar + last week's notes + email + Slack into a one-page brief: week ahead, time-sensitive items, 1:1 talking points, and a “things you committed to” tracker. Layer the triage agent: classify by urgency, draft replies in the principal's saved voice, propose which of the “47 meeting requests” to decline/delegate/batch, protect focus blocks. You schedule your own brief against your real calendar and inbox.

Pain: Context-switching; inbox/calendar triage; dropped commitments.

Scheduled Brief agent + Inbox/Calendar Triage agent (saved voice)

Kill the QBR Scramble: Deck Generator

A repeatable workflow that pulls KPIs from canonical sources (Sheets/BigQuery/CRM), drafts the commentary, flags anomalies and variances, outputs a near-final deck. Teach the “freeze data at T-minus-10-days” discipline. Live, turn a real KPI export into a near-final deck in minutes — the single most convincing proof for this cohort. You regenerate a version of your last QBR/board deck and diff it against the one you built by hand.

Pain: Roughly a working week's hours lost to manual reporting; stale decks.

QBR / Board Deck Generator workflow

Operating Cadence Engine: Status Roll-Up + Decision Log

Two connected builds: a Weekly Status Roll-Up — n8n + Claude summarizer that pulls team updates, normalizes them, flags at-risk items, assembles one exec-ready report; and a persistent Decision-Log agent + database that captures “what we actually decided” with owners so decisions stop getting re-made. Plus a templated cadence layer that auto-generates recurring agendas and chases off-track check-ins. You run one live roll-up cycle and back-fill three real recent decisions.

Pain: Coordination chaos; re-litigated decisions.

Weekly Status Roll-Up n8n flow + persistent Decision-Log agent & DB

Make It Watched, Not Skimmed: Async Video/Voice Update

Turn the week's written update into a polished async deliverable: script via Claude, then a HeyGen avatar video and/or an ElevenLabs voice briefing, automated end-to-end with n8n. Teach one-script-many-formats so the same narrative serves exec, board, and team. Short module, high wow — and it makes your invisible labor visible. You produce a real ~60-second update from your own material.

Pain: Invisible labor; low-engagement updates.

Leadership Update to Video/Voice workflow (Claude to HeyGen/ElevenLabs)

CAPSTONE: “My AI Chief of Staff” Orchestrator + Live Dashboard

Compose everything into one orchestrator on Anthropic's official Chief of Staff Agent architecture: the operator brain as standing memory; the weekend's skills registered as callable skills; specialized subagents (Recap, Roll-Up, Deck, Stakeholder-Brief) invoked via the Task tool; scoped MCP connectors; hooks writing a full audit trail; output styles for executive vs detailed; a meta-router so one button fires six agents in parallel. Then the front end: a jet-black liquid-glass live command-center dashboard unifying KPIs, action items, the decision log, and team status. The dashboard is the platform; the orchestrator is the app. Close with a maintenance runbook and the ROI reframe. You ship a 90-second live demo to the room.

Pain: Being the bottleneck; no single pane of glass; looking impressive to the principal.

Composed orchestrator + jet-black command-center dashboard + runbook

Kept forever

You leave with a running system, not the notes.

artifacts, owned and kept forever

Operator Brain — a personal CLAUDE.md every agent and dashboard loads automatically.
Meeting Recap & Follow-Up — a skill + n8n flow that posts recaps and opens tasks after every meeting.
Knowledge MCP — a connected source over your real docs that your whole team can query.
SOP-skill library — a starter set that clones your judgment, one skill per recurring process.
Stakeholder Intelligence — an agent + lightweight CRM that briefs you before every meeting.
Daily/Weekly Brief + Triage — scheduled agents with your principal's saved voice profile.
QBR / Board Deck Generator — turns canonical data into a near-final deck in minutes.
Operating Cadence Engine — Weekly Status Roll-Up + a persistent Decision-Log agent and database.
Leadership Update to Video/Voice — a workflow (Claude to HeyGen/ElevenLabs, automated via n8n).
The capstone — a composed AI Chief of Staff orchestrator + a jet-black live command center, on your real data, with a runbook so it never rots.
Cheat sheets — a model-choice + data-governance one-pager and a USD + INR ROI one-pager to justify it internally.

Honest fit

Who this is for. Who it isn't.

For

The solo Chief of Staff — often “the only person in the company with that job title” — in a startup founder's office, scale-up, or enterprise.
BizOps / RevOps / ops leads who own the operating cadence, the reporting, and the cross-functional programs.
Department and team leaders who run their function's meetings, status roll-ups, and SOPs.
Operators who are AI-curious but shallow, drowning in context-switching, and tired of being the org's human FAQ.

Not for

Anyone wanting a passive course to watch later. This is live, hands-on, on your real workload.
Anyone looking for a guaranteed promotion or title. We don't sell outcomes we can't honor.
Pure individual contributors with no recurring meetings, reporting, or coordination load — the use-cases won't land.

Assumed background — No coding background required. You will write context files, build skills, and wire connectors in plain language and a guided terminal. The floor coaches exist so “I'm not technical” stops being true by Saturday lunch.

Come with real work

Prerequisites & what to bring.

Set up before Day 1

Paid Claude account with Claude Code and Claude Cowork enabled.
ChatGPT Plus (for Codex / data-wrangling).
A free n8n cloud account.
Free ElevenLabs and HeyGen trial accounts.

Bring real exports

2–3 real meeting transcripts (Otter/Fellow/Granola/Zoom).
One recent board / QBR / leadership-update deck.
A messy KPI spreadsheet or CRM export.
Read access to one real Notion/Drive/Slack workspace.

Governance, answered live

Day 1 opens with a 30-minute Governance & Safe Sandbox clinic. Operators under strict data policies build the identical system on a permissioned, scoped, or synthetic copy. Nobody is blocked. The security objection dies on day one, with scoped MCP access shown live.

A laptop (Mac or Windows) and a real, in-progress workload — not a sandbox.

Straight answers

Objections, answered.

It's a rounding error against a Chief of Staff's annual comp, and less than the reporting hours you lose in a single busy week. You keep every artifact forever.

No coding background assumed. You write context files and wire connectors in plain language, with two floor coaches on the floor. By Saturday lunch, “I'm not technical” stops being true.

Those rent you a generic agent that could replace you. We teach you to build a bespoke one that proves you — your stack, your voice, owned and portable.

Then you're using a fraction of the capability. This is the gap between prompting and delegating multi-step work — and you leave with the system, not a habit.

Module 0 is a governance clinic. Scoped, read-only, or fully synthetic — you build the identical system safely, and the audit trail is built in.

You build on your own transcripts, KPIs, and workspace during the workshop — not demos. If it doesn't run on your reality by Sunday, that's on us.

The overload is the reason to come. The system you build is what gives the time back.

You leave with a maintenance runbook and the repo. The capstone is built to survive contact with reality.

Every session is live, every asset is delivered, and there is no back-end upsell. The build is the credential.

From the room

What operators build.

Voices from the founding cohort.

“Walk into the week already knowing what slipped and what to do about it — a Monday brief that lands before you're at your desk. That's what this room is built to hand you.”

The Monday-Morning BrainChiefs of staff · founding cohortThe Monday-Morning Brain

“Regenerate the quarterly board deck in an afternoon instead of a week — the deck generator you build and keep, on your own numbers.”

QBR / Board Deck GeneratorOps & chiefs of staff · founding cohortQBR / Board Deck Generator

“Build a command center good enough that your principal asks who made it — and the answer is you. The outcome you're here for.”

My AI Chief of StaffFounder's-office operators · founding cohortMy AI Chief of Staff

Application-only · one cohort a month · ~24 seats