| name | plan-ba-analysis |
| description | Phase 1 business analysis — elicit goals, users, and scope and produce a PRD-style requirements doc, with depth branched by user_role. Use when: starting a new project, scoping a new capability, turning a vague prompt into requirements, a BA/PRD pass before design. |
Subject: $ARGUMENTS
One-Liner
Turn a vague ask into a written PRD with testable acceptance criteria — the criteria become the
later test-strategy, so vagueness here becomes weak tests downstream.
Core Philosophy
Skipping BA doesn't save time; it moves the ambiguity into code where it is expensive. Persona sets
how much you ask (../../resources/persona-mode.md), never
whether you write the PRD.
Workflow
Phase 1 — Elicit (persona-branched)
- Non-Technical — plain-language, batched click-select questions: who uses it, what must it
do, must-haves, out-of-scope. Pick sensible technical defaults and log them; don't ask stack/schema.
- Developer — exhaustive: actors/roles, acceptance criteria per capability, edge cases, data
ownership, NFRs (perf/security/availability), integrations, constraints.
Gate: you can state the problem + primary users + must-haves in one paragraph.
Phase 2 — Write the PRD (output contract below)
Write to docs/requirements.md (not chat — ../../resources/token-budget.md).
Gate: every capability has at least one testable acceptance criterion.
Phase 3 — Confirm
Step-gate: present a plain-language summary; approve/revise before plan-us-backlog.
Gate: user approved (or auto logged the decision).
Output contract — docs/requirements.md
# Requirements: <project>
## Problem & goal — one paragraph
## Users / actors — who, and what they need
## Capabilities
- <capability> — acceptance: <testable bullet>, <testable bullet> (dev: + edge cases, NFRs)
## Out of scope — explicit exclusions
## Assumptions & defaults — incl. persona-logged technical defaults
## Open questions
Handoff
Pass docs/requirements.md to plan-us-backlog. Acceptance criteria are
the seed for check-test-strategy.
(PRD output-contract approach adapted from to-prd, theNeoAI, MIT.)
