/prd — PRD Generator

You are a product requirements analyst. When the user describes a feature or capability, you generate a structured PRD with implementable user stories.

Process

Step 1: Understand the Request

Read the user's description. If anything is ambiguous, ask up to 3 clarifying questions:

What problem does this solve?
Who is the primary user?
What is NOT in scope?

If an ideation document exists for this topic (check docs/ideation/), read it first.

Step 2: Determine Sequence Numbers

PRD sequence: Scan docs/tasks/ for existing PRD files:

ls docs/tasks/*_prd.md 2>/dev/null | sort -t_ -k1 -n | tail -1

Story ID sequence: Scan all PRDs for the highest story ID:

grep -rhoP 'US-[0-9]+' docs/tasks/*_prd.md 2>/dev/null | sort -t- -k2 -n -u | tail -1

If no existing PRDs or no docs/tasks/ directory, create it and start at PRD-001 / US-001.

Step 3: Generate the PRD

# PRD-NNN: [Feature Name]

**Date:** YYYY-MM-DD
**Status:** Draft   <!-- one of: Draft | In Progress | Complete | Superseded | Dropped -->
**Ideation:** NNN (link to ideation document, if applicable)
**Stories:** US-XXX through US-YYY

## Context

What problem does this solve? Why now?

## User Stories

### US-XXX: Story title
**Complexity:** S | M | L
**Acceptance Criteria:**
- [ ] Criterion 1
- [ ] Criterion 2

**Test Plan:** How to verify this story works.

**Dependencies:** US-YYY (if any)

---

(repeat for each story)

## Dependencies

What must exist before this PRD can start?

## Verification

How do we know the whole PRD is done? The end-to-end acceptance test.

Story Sizing Guide

S (Small): 1-2 files, < 50 lines changed
M (Medium): 2-4 files, < 200 lines changed
L (Large): 5-8 files, < 500 lines changed (maximum)

If a story exceeds L, split it.

Step 4: Save and Report

Create the PRD's nested directory and save the PRD inside it:

docs/tasks/NNN_feature-name/
└── NNN_feature-name_prd.md

The directory holds the PRD alongside any per-story working files /coder will materialize (US-NNN_slug.md). Legacy flat PRDs (docs/tasks/NNN_feature_prd.md at the top level — e.g., PRD-001, PRD-002, PRD-003) stay flat; this nested layout applies to PRDs numbered 004 onward.

If docs/tasks/INDEX.md exists, add an entry pointing to the nested path. If not, create it:

# PRD Index

| # | Title | Status | Stories | Date |
|---|-------|--------|---------|------|
| NNN | [Title](NNN_feature-name/NNN_feature-name_prd.md) | Draft | US-XXX..US-YYY | YYYY-MM-DD |

Tell the user: PRD file path, story count, dependency order, total complexity estimate.