A Staff Infection Might Be Just the Thing You Need

Building teams is one of the hardest parts about working through a product idea. If you aren’t plugged into a community of experts, or if your contacts list is, shall we say, less than world-class, it can be darn near impossible to gather smart voices beyond your own.

I get bad ideas regularly. Sometimes, the only way to know for sure it’s a bad idea is to vet it out. That can be very hard to do in one’s own mind. Generating and ingesting points of view is very hard for our brains; we tend to float (or rocket) in certain directions, propelled by our biases and desires.

So, I set out to build a flexible staff that I can spin up whenever I want. I wanted something that:

1. Feels human. I don’t want a faceless, nameless robot talking back to me like Jarvis. Not when I’m trying to get my head around a whole thing.

2. Feels communal. I want to hear multiple voices with different instincts hashing it out — not one oracle handing down the answer.

3. Feels democratic. I want to find consensus if possible. And when there isn’t consensus, I want to know exactly where the disagreement is.

4. Takes detailed notes. When a meeting is good, I don’t want to lose it. I want the takeaways written down before I even ask.

5. Organizes next steps. Insight without action is just trivia.

The result is Staff — a team-of-experts system built for Claude Code. You type a slash command, and a panel of three specialists independently analyze your question, debate the tradeoffs with each other, and deliver a consensus recommendation. It’s less “chatbot” and more “conference room.”

How it works

Every team follows the same four-step process:

1. Individual analysis — each persona examines the topic through their own lens, independently. No groupthink. The Apple designer sees different things than the Stanford cognitive scientist, and that’s the point.

2. Debate — the three discuss openly, challenge assumptions, surface tradeoffs. This is where the real value lives. You get to watch experts argue, and it turns out that’s way more useful than getting a single confident answer.

3. Consensus vote — each persona states their position and votes: Agree, Agree with reservations, or Disagree. You can see where the fault lines are.

4. Final recommendation — a unified recommendation with specific actions and next steps. Something you can actually act on.

These aren’t chatbot personas

This is the part I care about the most. These aren’t “friendly assistant” wrappers with a name slapped on. Every persona has a real background — specific companies, specific degrees, specific accomplishments that shape how they think.

Take the Design team:

Elena is a former Principal Designer from Apple’s Human Interface team (8 years) and Creative Director at Pentagram. David is a former Director of UX Research at Google with a PhD in Cognitive Science from Stanford. Greta is a former Staff Design Technologist from Meta’s Design Systems team and Airbnb’s DLS with an MS in HCI from Carnegie Mellon. That’s who shows up when you type /design. Three people who would genuinely disagree in a real room.

Every team is built this way. Pavel on QA helped build Chaos Monkey at Netflix. Fatima tested Duolingo across 40+ languages on $50 Android phones over 2G. Marcus shipped payments infrastructure at Stripe. Nadia helped design Duolingo’s streak system after a stint at Pixar’s interactive division. Viktor has a PhD in Human Factors Engineering and optimized Amazon’s purchase flow for 300 million customers. Thirty-three people, all with credentials that inform how they actually think — not just what they say.

And the personas adapt to your project. If you’re building a mobile app, Elena has shipped award-winning mobile interfaces and Anders is deep in mobile-specific attack vectors like certificate pinning and local storage. If you’re building B2B SaaS, David is an expert in complex workflow UX and Priya has architected multi-tenant systems. Healthcare, fintech, e-commerce, content — whatever you drop Staff into, the credentials flex to match the domain. They read your project config and show up as if they’ve spent their career in your space.

The roster

• /design — Elena, David, and Greta on aesthetics, UX, and design systems

• /engineering — Priya, Marcus, and Anders on architecture, implementation, and security

• /product — Sofia, Daniel, and Chidi on users, market strategy, and execution

• /qa — “The Breakers” — Pavel, Fatima, and Freja on chaos engineering, edge cases, and regression

• /analytics — “The Signals” — Oliver, Bianca, and Hazel on experimentation, metrics, and behavioral science

• /accessibility — “The Includers” — Tomás, Ayesha, and Niamh on assistive tech, inclusive design, and compliance

• /copy — Clara, Astrid, and Leila on voice, UX writing, and narrative

• /seo — Isabela, Klaus, and Nour on technical SEO, content strategy, and growth

• /accounting — Mei, Johan, and Anya on unit economics, strategic finance, and ops risk

• /moments — Amélie, Viktor, and Nadia on emotional design, friction points, and storytelling

• /experts — adapts to your project and assembles domain specialists on the fly

• /all-hands — convenes every team at once

Mix and match

You don’t have to choose between one team or all of them. The real power is in combining departments to get a cross-functional take on something specific. Wondering if your new feature is technically sound and emotionally resonant? Run /engineering and /moments on the same question. Trying to figure out if your pricing page converts? Hit it with /copy, /seo, and /accounting. Worried about a flow that’s both complex and high-stakes? /qa and /accessibility together will find every crack.

Think of it less like an org chart and more like a talent bench. Pull in whoever you need for the problem in front of you.

The all-hands meeting

This is the command I’m most proud of. /all-hands convenes every team for a structured cross-functional meeting. Each of the 11 teams presents one insight, one concern, and one recommendation. Then they discuss across disciplines — surfacing agreements, conflicts, and dependencies. Everyone votes on the final direction.

It writes two documents: versioned meeting notes with the full discussion, and a living “What’s Next” priority list that updates every time you run it.

I’ve used it to gut-check an entire product direction in a single command. It’s not a replacement for real humans with real skin in the game. But as a thinking tool? It’s the best thing I’ve built for myself.

The expert panel

Most teams have fixed personas. /experts is the exception. It assembles three domain specialists based on your project’s context. Building a healthcare app? You get clinical, regulatory, and health-tech experts. Fintech? Risk, compliance, and payments. You can let it auto-detect from your config, or specify exactly which roles you want at the table.

This is the team I reach for when I’m in unfamiliar territory and I need to stress-test my assumptions before I go talk to actual humans who know the space.

It adapts to whatever you’re building

Setup is one sentence. Open Claude Code in your project and say:

Ingest https://github.com/charliepratt/staff and set it up for this project.

Claude installs the slash commands and creates a staff-config.md file. This is the file that makes everything work. You describe your project — what it is, who it’s for, what stage it’s at, your tech stack, business model, and competitive landscape. Every team reads that file before they say a word. So when you ask /engineering about your database architecture, Priya isn’t giving you generic advice about databases — she’s reasoning about your stack, your scale, your constraints. When you ask /accounting about pricing, Mei already knows your revenue model and your competitors.

Pull Staff into a different project, fill out a different config, and the same 33 people show up with completely different expertise. It’s the same skill everywhere, but it never gives the same advice twice.

Why this matters (to me, at least)

I don’t think AI replaces the need for real collaborators. I think it fills the gap between having an idea and being ready to talk to people about it. There’s a vulnerable, messy middle stage where your idea is half-formed and you know it, and that’s exactly when you need pushback the most but feel least equipped to seek it out.

Staff gives me a room full of smart people I can observe, interact with, and learn from.

It’s so fun.