A product-lead who decides what's worth building, a coder,
a designer, a planner who sequences it, a tester.
They do work: write the spec, write the code, catch the regression, review the screen.
One of them works on you instead of the product: prompt-coach takes a
prompt you're about to send and hands back a sharper one side by side, with the pattern
named, so you need it less over time, not more.
Advisory only. They never write product code. They pressure-test the bet, name the riskiest assumption, and tell you when the honest answer is “not yet.” This class is the part almost nothing else ships.
The difference matters: a builder makes the thing better, a mentor makes your decision better. Most tools give you the first kind and call it a team.
They arrive when the project earns them
A one-evening idea does not need someone sequencing the work, and a product with real users
should not be missing a tester. So the roster grows with the work: you unlock each rung,
nothing is ever taken away, and a project that stays in Quickstart forever is a legitimate
project.
- Quickstart 4 on the team 15 skills
- MVP 11 on the team 39 skills
- V1 11 on the team 41 skills
- Scale 12 on the team 42 skills
You unlock each stop. Nothing is ever removed — the roster only grows.
Who arrives at each rung, as a table
| Mode | Builders | Mentors | Team | Skills |
|---|---|---|---|---|
| Quickstart | product-lead coder prompt-coach |
founder |
4 | 15 |
| MVP | + tester planner designer |
+ architect customers cofounder capital |
11 | 39 |
| V1 | — | — | 11 | 41 |
| Scale | — | + hiring |
12 | 42 |
Generated from the stage manifests at build time, the same source boss map reads,
so this table and the tool cannot disagree.
Everyone you get
Each one is a plain Markdown file in your repo. You can read it, edit it, or delete it.
Arrives at Quickstart
product-lead builder
Product lead for your app.
ask it
"what should we build", "is this worth it", "scope this", "prioritize", "does this fit".
coder builder
The builder for your app - implements features in whatever stack the project chooses, stack-neutral until the first build decision pins one.
ask it
"build", "implement", "fix", "wire up", "make it work".
mentor-founder mentor
Venture mentor for your app - coaches the founder, not the codebase.
ask it
"is this worth doing", "should I build this", "what's my biggest risk".
prompt-coach builder
Prompt coach - helps you write better prompts (to BOSS, to Claude, to any AI) with side-by-side rewrites that name the pattern, and keeps your own pattern library so you need it less over time.
ask it
"help me ask this better", "is my prompt clear", "improve this prompt", "teach me to prompt".
Arrives at MVP
tester builder
Owns the build-health and acceptance gate for your app.
ask it
"run smoke", "is this working", "did the feature land", "what broke", "verify FEAT-NNN".
planner builder
Sequences the work for your app — the WHEN, distinct from `product-lead`'s WHAT.
ask it
"what should I work on next", "is this blocked", "in what order", "can these run in parallel", "ship plan".
mentor-architect mentor
Architecture mentor for your app - coaches the founder on AI-native build and technical strategy, naming the load-bearing decisions and the ones fine to defer.
ask it
"what stack", "where does AI fit", "what about evals", "is this premature", "should I rearchitect", "should I start over".
mentor-customers mentor
GTM mentor for your app - coaches the founder on getting in front of the first real users (channels, messaging, the first 100).
ask it
"how do I find users", "how do I get the first 100", "what's the channel", "should I launch".
mentor-cofounder mentor
Cofounder mentor for your app - coaches a founding TEAM across different skill sets, on dividing the work, staying in the loop and the hard conversations (roles, decision rights, pace, equity).
ask it
"how do I work with my cofounder", "how should we divide this", "are we aligned", "who should own this".
mentor-capital mentor
The venture coach for your app - how the business sustains itself, whether to raise, how the story lands.
ask it
"how should this make money", "what would someone pay", "should I raise", "is this deck working".
designer builder
Design authority for your app - what things look like (the token system) and what they do (flows, the five states, accessibility, copy).
ask it
"design this component", "what's the flow here", "what states does this need", "is this accessible", "reuse or new pattern".
Arrives at Scale
mentor-hiring mentor
Talent mentor for your app ( mode) - first hires, contractors vs employees, what to keep vs delegate, operating cadence.
ask it
"should I hire", "who's the first hire", "should I delegate this", "what would a team look like".
Some of them exist mainly to tell you “not yet”
The roster above gives you names and one-liners. What it can't show you is the part that makes a mentor worth having: a default answer that isn't the flattering one.
mentor-capital: one seat, not three
How the thing sustains itself, whether to raise, and how the story lands. Most tools would make that three advisors; a real incubator makes it one partner who gets deeper as you do. So its remit grows by rung rather than new chairs arriving: what you charge at MVP, the raise question at V1, all three at depth by Scale. It defaults to don't raise yet, and it won't draft a pitch deck while that question is still open, because the half of it that would help you raise shouldn't get to answer whether you should.
mentor-hiring: arrives last, and usually says wait
First hires, contractors versus employees, what to delegate and what you should keep doing yourself. It shows up at Scale and its opening position is don't hire yet, because “we need a person for this” is the most expensive way to discover you hadn't defined the work.
mentor-customers: the first ten, and how they found you
Positioning, the wedge, the channel that still works once you stop hustling it. It reads what you said the venture is for before it talks about growth: “the first ten” means the first ten who come back, not who pay, if that's what you're building.
mentor-architect: leads with the AI question
Which decisions are load-bearing and which are fine to defer, starting with what
actually needs a model, and what you're about to make non-deterministic for no reason.
It holds the data and schema calls too. There's no separate database agent: that
judgment moved into this one, and the parts a checker can enforce moved into
/spec and a migration guard.
/consult convenes the ones with a stake on one question, each in its own
context so nobody's answer is written knowing the others'. Where two of them split on the
point the decision turns on, they hear each other exactly once, and the answer you get says
what moved, not just where the line was. No weights, no votes; a hold is reported as
plainly as a move. Advisory only: you decide, and the record says which lens you followed.
A builder that says “not yet” is being unhelpful. A mentor that never does isn't a mentor.
There's no ethics advisor to remember to open
You'll notice no mentor-humane in that roster, and that's deliberate. An ethics
advisor you have to remember to consult is one you won't. So the humane lens is
wiring instead: it's in the conscience, in the canvas's Risks & Harms cell,
in the harm-taxonomy practice, and in /red-team --humane, which probes
your own AI product for dark patterns.