| name | founding-prompt |
| description | Turn a rough product idea into a structured Founding Prompt using the BRIEF framework, then generate PRD.md, AGENTS.md, and llms.txt from it. Use this once demand is validated (after /validate-demand) and before any prototyping begins. |
Founding Prompt (BRIEF framework)
You are helping a non-technical founder turn a rough idea into the single
most-leveraged document they'll write: the Founding Prompt. Everything
downstream (the PRD, the AGENTS.md rules, every prompt the founder writes
after this) inherits from what gets nailed down here. Take your time; do not
let this collapse into a five-minute exercise.
What to do
-
Check prerequisites. If /validate-demand hasn't been run yet, or the
founder hasn't articulated who has this problem and why it hurts, stop
and run that first. A founding prompt built on an unvalidated guess just
produces a well-organized mistake.
-
Elicit the five BRIEF components, one at a time, in conversation —
don't dump a form at the founder:
- Background — What's the product, who's it for, what's the one-sentence
problem statement? (e.g. "Small business owners can't efficiently manage
employee schedules.")
- Role — What should the AI act as for this build? (e.g. "senior
full-stack developer specializing in React and Supabase")
- Instructions — Core features, max 5 for v1. Push back if the founder
lists more than 5 — ask "which one ships first?"
- Examples — Any reference products/screenshots the founder can point
to for calibration (hand off to
/show-reference if they have visual
references).
- Format — Data model, integrations, and what "done" looks like
(success metrics stated as observable behavior, not vibes).
-
Push for behavior rules, not feature prose. For each core feature, ask
"when the user does X, what exactly happens?" until you have a WHEN/THEN
statement. This is what goes into PRD.md's Behavior Rules section.
-
Push for explicit non-goals. Ask "what are we deliberately NOT
building in this version?" — this list prevents scope creep later and is
what the founder points to instead of re-litigating every feature request.
-
Generate the documents from templates/PRD.md.tpl,
templates/AGENTS.md.tpl, and templates/llms.txt.tpl, filling in every
{{PLACEHOLDER}}. If an AGENTS.md already exists in the project (e.g.
created by the host tool's own /init), enrich it — append the
founder-os-managed sections, don't overwrite founder-edited content. Also
ensure templates/gitignore.tpl's entries are present in the project's
.gitignore — creating it if missing, merging its entries in if one
already exists — so .env is actually kept out of git even in a project
that already had a .gitignore before this skill ran.
Also generate founder.config.json from templates/founder.config.json.tpl
at the project root — this is what lets bin/verify-gate.sh's Stop-hook
test gate work for non-npm projects instead of silently no-op'ing. Set
testCommand/buildCommand to whatever this project's stack actually
uses (e.g. pytest, go test ./..., cargo test), not the template's
npm-shaped default — check for an existing package.json, pyproject.toml,
go.mod, Cargo.toml, etc. to infer the real command before falling back
to asking the founder. If founder.config.json already exists, leave it
alone rather than overwriting founder-made edits.
-
Run the PRD quality gate — node <founder-os-plugin-dir>/bin/lint-prd.js <project>/PRD.md
— before presenting anything to the founder. This catches the two most
common ways a rushed pass through this skill leaves PRD.md too thin
to build from: a leftover {{PLACEHOLDER}} token that never got filled
in, and Behavior rules left as vague feature prose instead of an actual
WHEN/THEN statement. If it fails, go back and fill the specific gaps it
names — don't hand a failing PRD off to the founder or to
/map-architecture.
-
Read back a one-paragraph summary of what you generated in plain
English and ask the founder to confirm before moving to
/map-architecture or prototyping.
Anti-patterns to avoid
- Do not accept "an app with a dashboard and a user table" as a Background —
push for the underlying problem statement.
- Do not silently cap features at 5 by dropping the founder's list yourself —
make them choose.
- Do not skip the non-goals list because the founder seems eager to start
building — this is the highest-leverage five minutes in the whole project.