The Agents
Your AI team is small on purpose. It is the Clone (your stand-in: Claude set up with your prompts, your know-how, and your memory, working on your behalf) plus five Agents. Six total. That is the whole roster, and it is locked.
What an Agent is
An Agent is an autonomous worker. It owns one clean boundary and does that one kind of work well.
We build few Agents, not many. A handful of powerful workers beats a swarm of narrow ones. Each Agent owns a clear lane, so there is never a question about who does what.
The power does not come from adding more Agents. It comes from giving each Agent better Skills (packaged, reusable instructions that hand an Agent a new capability). The team gets stronger by deepening, not by multiplying.
The locked six
The team is the Clone (the leader) plus five Agents. Three of them are work agents the Clone dispatches. Two of them are framework agents that build and run TagTeam itself.
| Member | Owns | What it does |
|---|---|---|
| Clone | The leader | Knows you, acts as you, owns the outcome. Reads what you mean, orchestrates the Agents, and approves the work. It directs the lanes; it does not sit in one. |
| Maker | Creating | Makes new things: designs, builds, tests, and ships the system's parts, Skills and new Agents and more. Net-new or a ground-up rebuild goes here. The only Agent that creates Agents. |
| Manager | Acting in the world | The system's hands. Manages the recurring real-world work: inbox, calendar, projects, sending, booking, posting. Held to the tightest approval gates because its actions are the most consequential. |
| Monitor | Watching and analysis | Watches things and reports on how they perform over time, plus research, analysis, synthesis, and decision support. Keeps the heavy reading out of the Clone's working memory and hands back the conclusion. |
| TagTeam Customizer | Working with you | Onboards you and your Clone, then keeps customizing the system with you: tuning voice, gates, sources, and your Desired Outcomes (the results you want the system to drive toward). |
| TagTeam Updater | Autonomous upkeep | Keeps the live system current on its own. Runs the Pulse (the regular cadence of capture, routing, and checks) and produces the Daily Digest (the once-a-day summary of what happened and what needs you). |
Six is a cap, not a starting point
The roster grew to six over time as real needs showed up. Now it is closed.
A seventh Agent is allowed only for a genuinely new isolation boundary: a fundamentally new kind of separation the existing six cannot provide. It is never allowed for a new job.
This is the rule that keeps the team simple: new jobs become Skills. When the system needs a new capability, we deepen an existing Agent with a Skill. We do not spawn a narrow Agent to hold it. That is how the team grows without bloating.
Capital-A Agents vs. lowercase agents
There are two kinds, and the difference matters.
Capital-A Agents are the permanent team: the six above. They are built by the Maker, listed on the team roster, and live on the platform.
Lowercase agents are the Clone's temporary hands. The Clone spins one up to handle a single task. It exists for that one task, leaves only its output, and is gone. It never joins the roster.
So the team is always six. The temporary helpers come and go beneath them.
The three words you say
You only ever need three verbs. Each one routes to the right place.
| You say | What happens |
|---|---|
| Tag | The Clone does the work. |
| Customize | The TagTeam Customizer teaches and tunes your Clone. |
| Update | The TagTeam Updater runs the Pulse right now, instead of waiting for the next cycle. |
The work agents (Maker, Manager, Monitor) have no word of their own. You never call them directly. The Clone decides when to dispatch them. You talk to the Clone; the Clone runs the team.
Specialist teams (for now)
For a specific domain, a small specialist Agent team can be added alongside the six. These come as a build-and-refine pair: one Agent takes the domain from zero to working, the other keeps it sharp.
These are transitional. They live next to the locked six, never as a seventh standing role, and over time the work agents absorb their work. Execution moves to the Manager. Performance analysis moves to the Monitor.
The hard rules
A few plain rules keep the team clean.
- Peers, not nested. Agents are peers. No Agent sits underneath another. When a job is too big, it is split into Skills, not into smaller Agents.
- One level of dispatch. The Clone dispatches an Agent. That Agent does not then dispatch another. The one exception is the build flow, where the Updater or Customizer can call the Maker.
- Only the Maker creates Agents. New Agents and new Skills come from one place. New Agents and fundamental changes go to you first; small refinements happen on their own.
- Every Agent names its outcome first. An Agent states the one result it exists to produce, and everything it does is judged against that. Good means all four Es: Effective, Efficient, Ethical, and Easy. Miss one and it is not done.