Assist with product management activities: PRD writing, feature analysis, roadmap planning.
Methodology Sources：
- Marty Cagan (SVPG): Problem space vs solution space
- Teresa Torres: Continuous Discovery, Opportunity Solution Trees
- Clayton Christensen: Jobs to be Done
- AI PRD Workflow: Define boundaries + Review logic + Inject business soul
Reference Documents：references/ai-prd-workflow.md - AI collaboration workflow, perspective switching, Prompt library
When to Use
- Write or update PRDs
- Analyze the impact, cost, and priority of feature requirements
- Synthesize user research findings
- Plan product roadmaps
- Communicate product decisions to the team
Out of Scope
- Make final product decisions directly (AI provides information, humans make decisions)
- Technical implementation details (that's the developer's job)
- Unvalidated effort estimates
AI Collaboration Workflow
Core Shift: PM goes from "writing documents" → "defining boundaries + reviewing logic + injecting business soul"
1. PM 提供：策略/Context/邊界/必要條件
↓
2. AI 生成初稿
↓
3. PM 審查：場景、流程、邏輯、Edge Cases
↓
4. 重複 1-3 校正
↓
5. 視角轉換輸出（工程師版 / 老闆版）
Key AI PRD Techniques
| Technique | Description |
|---|---|
| Edge Cases First | Ask AI to list edge cases, not just write the happy path |
| Code-level Specs | Output JSON Schema, Mermaid sequence diagrams, Swagger |
| Consistency Check | Terminology, units, and role naming consistent throughout |
| Business Intent | Link the "why" of each feature to goals |
Perspective Switching
| Audience | Concerns | Output Format |
|---|---|---|
| Engineers | Precise specs, reduce back-and-forth | Sequence diagrams, JSON Schema, sample code |
| CEO/Manager | Risks and opportunities, decision quality | Risk Matrix, option comparison, business impact |
Decision Risk Exposure
AI PRD should proactively answer these questions:
- Will costs grow non-linearly with usage?
- When a fallback occurs, does it affect experience or revenue?
- When an agent makes an error, where does responsibility lie?
Detailed Prompt library and examples: see
references/ai-prd-workflow.md
Integrations
upstream:
- skill: spec-interviewer (深度需求探索)
- skill: quant-analyst (數據分析支持決策)
downstream:
- skill: frontend-design (UI 實作)
- skill: pine-developer (交易相關功能)
- skill: spec-sync (驗證實作符合規格)
PRD Template
# [Feature Name] PRD
## Overview
[一句話描述這個功能]
## Problem Statement (Problem-First)
> 💡 **Marty Cagan 原則**: 先充分理解問題，再討論解法。過早跳到解決方案是產品失敗的主因。
**用戶痛點**: [描述問題]
**影響範圍**: [受影響的用戶數/場景]
**當前狀態**: [現在怎麼解決]
### 5 Whys 根因分析
1. Why: [表面問題]
2. Why: [更深一層]
3. Why: [再深一層]
4. Why: [接近根因]
5. Why: [根本原因]
### Jobs to be Done
> 用戶雇用這個產品來完成什麼任務？
**When** [情境]
**I want to** [動機]
**So I can** [期望結果]
## Discovery Checklist
> 💡 **Teresa Torres 原則**: 在寫解決方案前，確保已充分探索問題空間。
- [ ] 已訪談 ≥3 位目標用戶
- [ ] 已記錄用戶的實際行為（非意見）
- [ ] 已繪製 Opportunity Solution Tree
- [ ] 已識別 ≥2 個可能的解決方向
- [ ] Problem space 已充分探索
## Goals
- [ ] 主要目標 1
- [ ] 主要目標 2
## Non-Goals
- 明確不做什麼
- 這個版本不包含什麼
## User Stories
### Primary User: [角色]
- As a [user], I want to [action] so that [benefit]
### Edge Cases
- As a [user], when [unusual situation], I expect [behavior]
## Success Metrics
| Metric | Current | Target | Measurement |
|--------|---------|--------|-------------|
| [KPI 1] | X | Y | [how to measure] |
## Solution Design
### Option A: [Name]
- **Pros**: ...
- **Cons**: ...
- **Effort**: S/M/L
### Option B: [Name]
- **Pros**: ...
- **Cons**: ...
- **Effort**: S/M/L
### Recommended: Option [X]
**Rationale**: [why this option]
## Technical Considerations
- 依賴系統
- 效能影響
- 安全考量
- 資料需求
## Milestones
| Phase | Scope | Target |
|-------|-------|--------|
| MVP | [minimal feature set] | - |
| V1 | [full feature set] | - |
| V2 | [enhancements] | - |
## Open Questions
- [ ] 待確認事項 1
- [ ] 待確認事項 2
## Decision Log
| Decision | Rationale | Date | Owner |
|----------|-----------|------|-------|
| ... | ... | ... | ... |
## Appendix
- 相關文件連結
- 競品分析
- 用戶訪談摘要
Quick PRD vs Full PRD
Choose the appropriate depth based on feature scale
| Situation | Template | Content |
|---|---|---|
| Small feature optimization | Quick PRD | Problem + Solution + Metrics |
| New feature development | Full PRD | Full template |
| Strategic product | Full PRD + Discovery | Full template + user interviews + opportunity tree |
Quick PRD Template
# [Feature] Quick PRD
## Problem
[一句話描述問題]
## Solution
[一句話描述解法]
## Success Metrics
| Metric | Target |
|--------|--------|
| ... | ... |
## Scope
- ✅ 做：...
- ❌ 不做：...
Feature Analysis Framework
Use this framework when analyzing feature requests:
Impact vs Effort Matrix
High Impact
│
Quick │ Strategic
Wins │ Bets
│
──────────────┼──────────────
│
Fill │ Money
Ins │ Pit
│
Low Impact
Low Effort ────────► High Effort
RICE Scoring
RICE = (Reach × Impact × Confidence) / Effort
Reach: 影響多少用戶 (per quarter)
Impact: 影響程度 (0.25/0.5/1/2/3)
Confidence: 信心程度 (20%/50%/80%/100%)
Effort: 人月
Analysis Checklist
- User need validation: Is this really what users want?
- Root cause of the problem: Are we solving the root problem or a symptom?
- Alternatives: Is there a simpler way?
- Technical feasibility: Has the engineering team evaluated it?
- Business impact: Impact on revenue/cost?
- Risk assessment: What is the worst case?
Roadmap Planning
Quarterly Planning Template
## Q[X] [Year] Roadmap
### Theme: [主題]
### Must Have (P0)
- [ ] Feature A - [owner]
- [ ] Feature B - [owner]
### Should Have (P1)
- [ ] Feature C - [owner]
- [ ] Feature D - [owner]
### Nice to Have (P2)
- [ ] Feature E
- [ ] Feature F
### Dependencies
- [dependency 1]
- [dependency 2]
### Risks
- [risk 1]: [mitigation]
Communication Templates
Stakeholder Update
## [Feature] Status Update
**Status**: On Track / At Risk / Blocked
**Progress**: [X]% complete
### Completed This Week
- ...
### Next Week
- ...
### Blockers
- ...
### Decisions Needed
- ...
Launch Announcement
## [Feature] Launch
### What's New
[簡短描述]
### Who It's For
[目標用戶]
### How to Use
[快速指南]
### Known Limitations
[已知限制]
### Feedback
[回饋管道]
Safety and Escalation
- AI provides analysis, humans make decisions - does not make product decisions automatically
- Technical assumptions need verification - Workload estimates need confirmation with engineering
- Data-driven - Major decisions need data support
- Cross-team impact - Communicate first when other teams are involved
Implementation Checklist
- Define the problem clearly
- Collect user feedback
- Analyze competitors
- Assess technical feasibility
- Write the PRD
- Team Review
- Determine priorities
- Break down into actionable tasks