| name | vision-writing |
| description | Use when running /business:vision, or whenever capturing a product's problem, target users, and success criteria at the highest level, before any PRD or feature work. Philosophy adapted from BMAD-METHOD's Analyst-stage discovery. |
Vision Writing
A vision statement is the top of the artifact chain: /business:prd, /business:persona,
/business:architecture, and eventually every feature spec should trace back to it. Get this
right and everything downstream has a north star; get it vague and every later document
inherits the vagueness.
Grounding in an existing codebase (brownfield adoption)
If this product already has source code (the calling command detected project manifests or a
src//lib/ tree), skim the README and top-level structure before starting the discovery
loop below, and summarize what you found back to the user in 2-3 sentences. A vision written
for an existing product should acknowledge what's already built, not read as if starting from
zero — the discovery questions in step 1-5 still apply, but they're now "what should this
become next," not "what should this be." This grounding is conversational context only: it
does not change the Vision.md template, and it never produces a file of its own.
Process
- Establish the problem before the solution. Ask what's broken or missing today, for
whom, and why it matters enough to build something. Resist jumping to "what we'll build"
before the problem is concrete.
- Name the target users/market specifically. "Everyone" is not a target market. Push for
a specific enough description that a persona could later be drawn from it.
- State goals as outcomes, not features. "Reduce investigation time for SOC analysts"
is a goal; "add a dashboard" is a feature — the latter belongs to the PRD or later, not
here.
- Make success measurable. Each success metric should be something you could actually
check later — a number, a rate, a survey result — not a feeling.
- Write explicit non-goals. What this product deliberately will not try to be, so scope
discipline exists from the very first document onward.
Self-review checklist
Red flags
| Red flag | Why it matters |
|---|
| A goal reads like a feature ("add X") | Feature-shaped goals foreclose solution space too early — reframe as the outcome X would produce |
| Target users described as "everyone" or "all businesses" | Too broad to build a coherent product or persona from |
| A success metric can't be measured with real data | It won't be checkable later, so it won't actually gate anything |
| Vision written for an existing product with no acknowledgment of what's already built | Reads as disconnected from reality; the user will discount the whole document |