Define Requirements and a Specification
Use this page to pick the requirements skill for the work in front of you: a
product brief or a PRFAQ, then a PRD, for a product; or froid-spec for the
contract implementation reads.
Which Path Applies
Section titled “Which Path Applies”Choose a Planning Path decides the size. Most
work goes straight to froid-spec from whatever defined the intent: a forged
idea, a PRFAQ summary, a brainstorm intent, an issue. A PRD becomes necessary
when several people must agree on what the product is or several epics must
stay aligned; Plan Inside an Organization
covers that setting. This page covers what each document is for and which
skill writes it. None of them replaces the spec: every epic still ends up as a
SPEC.md that Build reads, and when there is a PRD, that is where the spec’s
answers come from.
The brief, PRD, and UX skills steer the same way. Each opens with a brain
dump: tell it everything and point it at any files you have. Then you choose a
working mode. Fast path batches the remaining gaps into a question or two
and drafts the whole document with [ASSUMPTION] tags for you to correct.
Coaching path pulls the thinking out of you section by section and pushes
back where an answer is thin. Sessions can be paused and resumed.
Product Brief: Capture Conviction
Section titled “Product Brief: Capture Conviction”Run froid-product-brief to create, update, or validate a brief: a one- to
two-page account of the product concept, right-sized to its purpose. A passion
project does not get investor-grade rigor; a pitch input does. The coach reads
the stakes early and calibrates how hard it pushes.
Use it when your concept is relatively clear and you want it written down
before a PRD. It produces brief.md plus addendum.md, which holds the depth
that belongs later rather than in the brief: rejected alternatives, options
considered, technical constraints, sizing data. froid-prd reads both. For
serious market sizing or competitor teardowns it hands you to
Deep Recon.
PRFAQ: Working Backwards
Section titled “PRFAQ: Working Backwards”Run froid-prfaq for Amazon’s Working Backwards method as a challenge. You
write the press release announcing the finished product before anything is
built, then answer the hardest questions customers and stakeholders would
ask. Solution-first and technology-first openings get redirected to the
customer’s problem. Vague answers get challenged. When you are stuck it offers
concrete reframings rather than repeating the question.
A run moves through five stages: customer, problem, stakes, and concept; the press release; the customer FAQ; the internal FAQ on feasibility and trade-offs; and a verdict on the concept’s strength. Claims about the market and competitors are checked against current research, not assumed. If after a few exchanges you cannot name a customer or a problem, it sends you back to brainstorming or Forge Idea instead of forcing it.
Use it when you want the concept stress-tested before committing resources.
If you cannot write a compelling press release, the product is not ready. It
produces the PRFAQ document plus a short summary a PRD or spec can read, and
it accepts -H for an unattended first draft when you supply the customer,
problem, stakes, and concept up front.
Brief and PRFAQ both feed a PRD, and the PRFAQ summary can feed froid-spec
directly when no PRD is needed. Choose by how much challenge you want: the
brief is collaborative discovery, the PRFAQ is the harder path. Neither is
required; froid-prd starts from a brain dump on its own.
PRD: Agree on What and Why
Section titled “PRD: Agree on What and Why”Run froid-prd to create, update, or validate a Product Requirements Document.
- Create runs discovery, then drafts. On the Coaching path you also pick an entry point: Vision + Features for capability-first products and internal tools, or Journey-led for consumer and multi-stakeholder products, where user journeys are told with a named protagonist.
- Update reconciles the PRD with a change signal, surfacing conflicts with earlier decisions before applying anything.
- Validate critiques without changing and produces a findings report.
The PRD describes capabilities, not implementation: features grouped, with
functional requirements under stable IDs and non-functional requirements in
their own section. Technical choices go to addendum.md. Length scales with
stakes, from about two pages for a hobby project to as long as the
requirements need for a launch.
It answers “what should we build and why.” It does not say how; that is Design UX and Architecture. It does not divide work into stories; that is Break Work into Stories and Track It. If you open it with a one-pager in mind or an idea to vet, it points you at the brief or the PRFAQ instead.
Spec: The Contract Implementation Reads
Section titled “Spec: The Contract Implementation Reads”Run froid-spec to turn an intent into a short contract that Build reads.
SPEC.md has five fields: Why, Capabilities (each with an intent and a
success condition), Constraints, Non-goals, and Success signal. Tables,
diagrams, glossaries, and documents other skills already wrote sit beside it;
the spec points at them rather than copying them.
The spec writes the contract; it does not help you figure out what you want.
Rich input is extracted with no questions. Sparse input gets a choice: a
best-effort draft where every gap becomes an open question, or a guided walk
through the five fields. Input too thin to use (“an app for hikers”) is sent
to froid-prd. Input too large is the other failure: a few tens of thousands
of tokens is the practical ceiling, so condense a pile of material first.
froid-spec is the only writer of SPEC.md. Do not hand-edit it; run the
skill again with the change and it updates the spec in place, keeping
capability IDs stable. The PRD, UX, and architecture skills can run in any
order and feed the same spec. After every run it reports assumptions it made
and open questions it could not answer, for you to resolve.
On request, Story Breakdown turns one spec into an epic. It walks the
capabilities and constraints with you, proposes a story per independently
reviewable slice, and asks for each one whether you want a checkpoint before or
after implementation. The result is the ordered stories.yaml beside
SPEC.md. That is the whole planning set for a spec-backed epic; see
Choose a Planning Path
for how the epic then runs.
What Comes Next
Section titled “What Comes Next”With a PRD in hand for multi-epic work, decide whether the work needs shared design decisions: Design UX and Architecture. With a spec in hand for one epic, ask for Story Breakdown and go to Break Work into Stories and Track It.