qiaomu-ai-prd

Turn a one-line product idea into an implementable PRD (Product Requirements Document) that both human developers and AI coding assistants can execute.

Use Cases

• You have a single product idea (e.g., “build an English vocabulary learning website” or “develop an iOS teleprompter”) and need a full, structured PRD.
• You want the PRD to include real product judgment, technical decisions, and acceptance criteria—not just a template fill-in.
• You plan to hand the PRD to an AI coding assistant or a development team for implementation.

Preview

The following is a sample PRD output (for "WordPulse", a vocabulary learning website) extracted from the README:

# WordPulse PRD

## AI Speed-Read Card
Product in one sentence: A personal word bank learning website with built-in review pacing.
Core loop: Import word bank → Practice → Mark mastery → Auto review.
Hard constraint: P0 must complete one learning and review cycle.
Recommended default: Save locally first, no account required.
Creative space: Practice animations, score feedback, review completion page can be more memorable.
Overdelivery opportunity: Generate a shareable card showing “Today you mastered 18 words”.

## Chapter 1: Product Overview
WordPulse is a web-based English vocabulary learning tool that lets self-learners study, practice, and review using their own word banks, without switching between fixed courses and scattered notes.

## Chapter 2: Overall Layout & Navigation
+--------------------------------------------------+
| Top learning status bar (100% x 64px)             |
| Today’s review: 18 words | Streak: 6 days | Start review |
+----------------------+---------------------------+
| Word bank & filter (28%) | Practice workspace (72%) |
| CET-6 core words         | abandon                 |
| Error book: 12           | [Know] [Unsure] [Don’t know] |
+----------------------+---------------------------+

## Chapter 10: Performance Metrics
| Metric | Target | Measurement method | Degradation threshold |
|---|---:|---|---:|
| First interactive time | <= 1200ms | Lighthouse mobile 4G | > 2200ms |
| Answer feedback delay | <= 80ms | Click to state change | > 180ms |

Core Capabilities

• Generate product positioning, competitive differentiation, three user personas, and feasibility boundaries.
• Provide top-level layout, core modules, real states (default, active, empty, error), normal paths, and failure paths.
• Identify differentiating features and explain why competitors usually cannot achieve them.
• Produce data models with // comments, technical architecture, and dependency choice rationale.
• Mark overdelivery opportunities, interaction details, export system, P0–P3 development priorities, and numeric performance metrics.
• Generate developer handoff notes written directly to the implementing AI, replaceable technology principles, and acceptance scripts.
• Every major module includes an ASCII diagram with realistic content.
• Must include AI Speed-Read Card, Hard Constraints / Recommended Defaults / Creative Space, and Overdelivery Opportunities.
• Data model fields have // comments; top-level objects include "version".
• Performance metrics must be numeric, not just "fast" or "smooth".

How It Works

1. You provide a one-line product idea (e.g., “Use qiaomu-ai-prd to write a PRD for an English vocabulary learning website.”)
2. You can optionally append one or more modes (see table below) to adjust the PRD’s focus.
3. The Skill generates a complete PRD document based on your request and chosen modes.
4. The PRD includes a final chapter written for the implementer, containing honest known unknowns and at least 3 acceptance scripts.

What You Need to Provide

• A one-line product idea (core requirement).
• (Optional) One or more modes from the list below.

What You Will Get

A complete PRD document containing:

• AI Speed-Read Card (product in one sentence, core loop, hard constraints / recommended defaults / creative space, overdelivery opportunities)
• Product overview
• Overall layout and navigation (ASCII diagram)
• Core modules and their states
• Data model (commented JSON)
• Technical architecture and dependency choices
• Interaction details, export system, priorities, and performance metrics
• Developer handoff notes, replaceable technology principles, and acceptance scripts

Optional Modes

Prerequisites

• Node.js installed and node --version works.
• The current agent supports a local skills directory (typically ~/.agents/skills).
• The working directory is writable if you want to save or publish PRD files.

Notes

• If the input is too short and the model does not use the skill, the output may look like a template lacking product judgment. Make sure to say “use qiaomu-ai-prd” or provide additional context about the core user and platform.
• The PRD should not contain placeholders. If it does, run scripts/lint_prd.py <file> and fix.
• If technology package size information seems unreliable, write unknown or verify official package information online before updating.
• P0 priority should be based on the smallest set that completes the core loop, not sorted by implementation difficulty.
• Low-risk implementation details should be set as Recommended Defaults or Creative Space, not Hard Constraints, to avoid over-constraining the AI.