pm-brainstorm

Input and mode

First determine whether the user wantsa brief discussionor a formal creative solution library. When only discussing, do not write files or create directories; if relevant and valid materials already exist, they can be reused,no need to ask again. Only create after formally generating the solution library and confirming the target location docs/01-需求调研/. Ask only for missing key information. The user can stop or skip a certain thinking method at any time; do not describe skipped exploration as completed.

Cross-Agent interaction rules

When the process requires interaction with the user:

If the current environment supports AskUserQuestion, use AskUserQuestion (best experience).
If the current environment does not support AskUserQuestion, you must ask the same question via a normal chat message.
Ask only one question at a time.
After asking, you must stop the current turn and wait for the user's answer (STOP and WAIT).
When only discussing, do not write files; formal artifacts are written only after the user requests it and confirms the path.
Relevant and valid existing docs can be reused, with no need to ask again; irrelevant materials cannot replace evidence.

Execution flow

The following areformal solution library modedetailed steps. A lightweight discussion only covers the parts the user cares about and directly gives a verifiable direction in chat; the user may stop or skip method steps. When skipping, mark the unexplored parts and do not falsely claim the process has been completed.

Exception: in step 3.3, when the user chooses G "skip the entire challenge phase," skip that phase and go directly to the next step.

Step 1: Read prerequisite data

If a requirements research report exists：

Use the Read tool to read docs/01-需求调研/需求调研报告.md as reference.

If it cannot be determined from the content whether the existing report is relevant to the current product, ask the user; relevant and valid content should be reused directly.

Use AskUserQuestion to ask:

An existing requirements research report was detected. Is it applicable to the current product?

A) Applicable — continue based on the existing report, but still confirm whether the key information is accurate B) Not applicable / uncertain — enter quick mode and collect basic information again

Cross-Agent fallback: if the current environment does not support AskUserQuestion, ask the same question via a normal chat message, then STOP and WAIT for user response.

User confirms applicable → extract existing information; ask only about missing or outdated key content, with no need to repeatedly confirm item by item.
User confirms not applicable / uncertain → enter quick mode and use AskUserQuestion to collect basic information.

If there is no requirements research report：

Enter quick mode and use AskUserQuestion to collect basic information:

Ask one by one only when this information is truly missing and determines the direction; use information already provided directly.

📝 Quick mode - please provide basic information:

Question 1: What is the product name? → STOP and WAIT for user response

Question 2: Who are the target users? → STOP and WAIT for user response

Question 3: What is the core pain point? → STOP and WAIT for user response

Cross-Agent fallback: If the current environment does not support AskUserQuestion, ask the same question using a regular chat message, then STOP and WAIT for user response.

Step 2: Determine the divergence direction

Use AskUserQuestion to ask:

A formal solution library requires determining the divergence direction; a lightweight discussion can provide preliminary options based on existing information and note the assumptions.

🎯 Which direction of the product do I want to explore?

A) Product feature innovation - Explore core features and differentiated characteristics B) User growth plan - How to acquire, activate, and retain users C) Business model design - How to monetize and build a sustainable business model D) User experience optimization - How to improve usability and satisfaction E) Inspiration spark stimulation - I don't know what to do and need AI to help me find inspiration(New in v2.1) F) Other direction (please enter manually)

Cross-Agent fallback: If the current environment does not support AskUserQuestion, ask the same question using a regular chat message, then STOP and WAIT for user response.

After the user selects, record it in the variable BRAINSTORM_DIRECTION

If "Inspiration spark stimulation" is selected (new in v2.1):

Applicable scenarios: The user does not know what product to make and needs AI to actively stimulate inspiration

Execution flow：

Step 2.1: Inspiration dimension scan & Subagent parallel search (core optimization)

AI prompts the user:

🔍 Inspiration Spark Stimulation Engine started (Subagent parallel mode)

I will search for inspiration sources from 4 dimensions simultaneously, please wait...

Industry pain points → Subagent 1

Technology trends → Subagent 2

Life scenarios → Subagent 3

Cross-domain inspiration → Subagent 4

Dispatch 4 subagents in parallel to search：

Use the Agent tool to dispatch simultaneously:

Subagent 1 - 行业痛点:
  type: "general-purpose"
  prompt: |
    搜索行业痛点和用户未被满足的需求。

    搜索要求：
    1. 搜索工具优先级：首选 AnySearch（先定位 anysearch_cli.py：检查 ~/.claude/skills/anysearch/scripts/ ~/.opencode/skills/anysearch/scripts/ ~/.openclaw/skills/anysearch/scripts/ ~/.cursor/skills/anysearch/scripts/ ~/.anysearch/scripts/ 或 which anysearch_cli.py，找到后 Bash 调用 python3 <路径> search "查询词" --max_results 5），失败降级到 Exa MCP（mcp__exa__web_search_exa），最后才用 WebSearch。使用 WebSearch 时标注降级模式
    2. 搜索：site:36kr.com OR site:huxiu.com 用户痛点 2026；site:zhihu.com 用户抱怨 不满
    3. 搜索：ProductHunt 用户差评 痛点
    4. 提取 5-8 个具体痛点，每个包含：痛点描述、受影响用户、严重程度

    输出 JSON 格式：
    {"dimension":"行业痛点","findings":[{"pain_point":"描述","users":"受影响用户","severity":"高/中/低"}]}

Subagent 2 - 技术趋势:
  type: "general-purpose"
  prompt: |
    搜索2026年最新技术趋势和AI应用场景。

    搜索要求：
    1. 搜索工具优先级：首选 AnySearch（先定位 anysearch_cli.py：检查 ~/.claude/skills/anysearch/scripts/ ~/.opencode/skills/anysearch/scripts/ ~/.openclaw/skills/anysearch/scripts/ ~/.cursor/skills/anysearch/scripts/ ~/.anysearch/scripts/ 或 which anysearch_cli.py，找到后 Bash 调用 python3 <路径> search "查询词" --max_results 5），失败降级到 Exa MCP（mcp__exa__web_search_exa），最后才用 WebSearch。使用 WebSearch 时标注降级模式
    2. 搜索：AI新技术 应用场景 2026 site:mittrchina.com OR site:ifanr.com
    3. 搜索：2026 技术趋势 创业机会
    4. 提取 5-8 个技术驱动的产品机会，每个包含：技术名称、应用场景、可行性

    输出 JSON 格式：
    {"dimension":"技术趋势","findings":[{"tech":"技术","scenario":"场景","feasibility":"高/中/低"}]}

Subagent 3 - 生活场景:
  type: "general-purpose"
  prompt: |
    搜索2026年效率工具和生活方式变化带来的产品机会。

    搜索要求：
    1. 搜索工具优先级：首选 AnySearch（先定位 anysearch_cli.py：检查 ~/.claude/skills/anysearch/scripts/ ~/.opencode/skills/anysearch/scripts/ ~/.openclaw/skills/anysearch/scripts/ ~/.cursor/skills/anysearch/scripts/ ~/.anysearch/scripts/ 或 which anysearch_cli.py，找到后 Bash 调用 python3 <路径> search "查询词" --max_results 5），失败降级到 Exa MCP（mcp__exa__web_search_exa），最后才用 WebSearch。使用 WebSearch 时标注降级模式
    2. 搜索：ProductHunt 2026 最佳产品 效率工具
    3. 搜索：2026 生活方式变化 新需求
    4. 提取 5-8 个生活场景相关的产品灵感，每个包含：场景描述、用户需求、已有方案

    输出 JSON 格式：
    {"dimension":"生活场景","findings":[{"scenario":"描述","need":"用户需求","existing":"已有方案"}]}

Subagent 4 - 跨界灵感:
  type: "general-purpose"
  prompt: |
    搜索其他行业的创新商业模式和成功案例。

    搜索要求：
    1. 搜索工具优先级：首选 AnySearch（先定位 anysearch_cli.py：检查 ~/.claude/skills/anysearch/scripts/ ~/.opencode/skills/anysearch/scripts/ ~/.openclaw/skills/anysearch/scripts/ ~/.cursor/skills/anysearch/scripts/ ~/.anysearch/scripts/ 或 which anysearch_cli.py，找到后 Bash 调用 python3 <路径> search "查询词" --max_results 5），失败降级到 Exa MCP（mcp__exa__web_search_exa），最后才用 WebSearch。使用 WebSearch 时标注降级模式
    2. 搜索：创新商业模式 成功案例 2026
    3. 搜索：跨界创新 行业颠覆 案例
    4. 提取 5-8 个可借鉴的跨界灵感，每个包含：来源行业、创新点、可迁移性

    输出 JSON 格式：
    {"dimension":"跨界灵感","findings":[{"industry":"来源行业","innovation":"创新点","transferable":"可迁移性"}]}

Wait for all 4 subagents to complete, and collect the JSON results.

Step 2.2: Generate 10 inspiration sparks

AI integrates the structured data returned by the 4 subagents and generates 10 inspiration sparks, each containing:

Product concept (one sentence)
User pain point
Technical feasibility (high/medium/low)
Market potential (high/medium/low)

Output example:

#	Product concept	User pain point	Technical feasibility	Market potential
1	AI meeting minutes auto-generator	Low meeting efficiency, time-consuming to write minutes	High (speech recognition is mature)	High (rigid demand for enterprises)
2	Intelligent code review assistant	Code quality is difficult to control	High (LLM understands code)	High (developer demand)
3	Personal knowledge graph construction tool	Information is fragmented and difficult to systematize	Medium (requires knowledge graph technology)	Medium (demand from knowledge workers)
4	...	...	...	...

Step 2.3: User selects one for in-depth exploration

Use AskUserQuestion to ask:

💡 10 inspiration sparks have been generated, please select the one you are interested in:

A) Inspiration #1: {dynamically populated based on results} B) Inspiration #2: {dynamically populated} C) Inspiration #3: {dynamically populated} D) View more inspirations (scrolling display) E) Custom direction (please enter manually)

After the user makes a selection, proceed toStep 3: Progressive In-Depth Explorationto conduct detailed divergence on that inspiration.

Optimization Notes(Inspiration Spark section):

Before optimization: The main agent executes 4 searches serially, and the search results occupy approximately 8,000 tokens of context
After optimization: 4 subagents search in parallel, and the main agent only processes structured JSON, saving approximately 85% of tokens
Speed improvement: From ~4 minutes (serial search) → ~1.5 minutes (parallel subagents)

Step 2.5: Select Brainstorming Mode (new in v2.0)

Use AskUserQuestion to ask:

Formal solution libraries can choose a mode; lightweight discussions do not need to force a methodology choice.

🧠 Select brainstorming mode:

A) Conventional divergent thinking - Based on industry experience and analogical reasoning B) First principles decomposition - Derive from basic facts, break through conventional assumptions C) SCAMPER innovation method - Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse

Cross-Agent fallback: If the current environment does not support AskUserQuestion, ask the same question via a regular chat message, then STOP and WAIT for user response.

After the user makes a selection, record it to the variable BRAINSTORM_MODE

Step 3: Progressive In-Depth Exploration

Adopt different exploration paths based on the selected mode：

If "First Principles Decomposition" is selected (new in v2.0):

First Principles Core：

Decompose to basic facts
Re-derive from basic facts
Break through conventional assumptions and analogical thinking

Step 3.1: Decompose Basic Facts

Use AskUserQuestion to ask:

🔍 First Principles Decomposition - Step 1

Please list the basic facts (self-evident truths) related to "{product direction}":

Example (delivery speed):

User expectation: The faster, the better

Physical limits: Speed of light, traffic speed

Cost rule: The faster the speed, the higher the cost

Please enter the basic facts one by one:

Guide the user to enter 3-5 basic facts:

{Basic fact 1}
{Basic fact 2}
{Basic fact 3}

Step 3.2: Identify Conventional Assumptions

Use AskUserQuestion to ask:

🎭 First Principles Decomposition - Step 2

What are the conventional assumptions in the industry? (There may be cognitive biases)

Example (delivery speed):

Assumption 1: 30-minute delivery is the limit

Assumption 2: You must have your own delivery team

Assumption 3: Users are unwilling to pay a high price

Please list the industry's conventional assumptions:

Guide the user to identify 3-5 conventional assumptions:

{Conventional assumption 1}
{Conventional assumption 2}
{Conventional assumption 3}

Step 3.3: Challenge Assumptions, Re-derive

For each assumption, use AskUserQuestion to ask:

💡 First Principles Decomposition - Step 3

Challenge the assumption: "{conventional assumption}"

A) Is this assumption true? Are there counterexamples? B) What happens if this assumption is broken? C) Are there other ways to achieve the same goal? D) What is the root of this assumption? E) Other thoughts (please enter manually) F) Skip the current challenge and continue to the next assumption G) Skip the entire challenge phase and go directly to the next step

After the user selects:

Select F → Do not perform the challenge analysis for the current assumption; continue to the next assumption
Select G → Do not perform the challenge analysis for the current assumption; skip the entire challenge phase and go directly to the next step
Select A-E → Perform the AI analysis below

AI analysis:

Find the flaws in the assumption
Propose breakthrough solutions
Re-derive based on fundamental facts

Step 3.4: Generate innovative solutions

AI derives from first principles and generates breakthrough solutions:

Example output (delivery speed)：

Conventional solutions：

Build your own delivery team → high cost, difficult to scale
Third-party logistics → uncontrollable speed
Forward warehouse model → high cost, small coverage area

First-principles solutions：

Crowdsourced delivery: break the assumption that "you must have your own team"
- Use idle labor
- Low cost, highly flexible
Predictive stocking: break the assumption that "preparation starts only after the user places an order"
- Stock in advance based on AI predictions
- It is already in transit when the user places an order
Community micro-warehouse: break the assumption of "centralized warehousing"
- Distributed micro-warehouses
- 5-minute delivery becomes possible

If "conventional divergent thinking" is selected:

Key principles：

Ask only one question at a time
Provide 3-5 options for each question
The last option is always "Other (manual input)"
Based on the user's choice, go deeper step by step
Use AskUserQuestion to ask one by one
After asking, you must STOP and WAIT for user response
Cross-Agent fallback: if the current environment does not support AskUserQuestion, ask the same question in a regular chat message, then STOP and WAIT

If "product feature innovation" is selected:

Question 1：

What problem should the core feature solve?

A) Improve efficiency - let users complete tasks faster B) Lower the barrier - make it easier for new users to get started C) Enhance experience - make the process more enjoyable and fun D) Create connections - enable interaction among users E) Other (please enter manually)

Question 2：

How can differentiation be achieved?

A) Technological innovation - use new technology or unique algorithms B) Model innovation - innovative business model or operating approach C) Experience innovation - better user interface or interaction D) Service innovation - better customer service or after-sales support E) Other (please enter manually)

Question 3：

Which features are must-have (MVP)?

A) Core features + basic experience B) Core features + social attributes C) Core features + data analysis D) Core features + membership system E) Other (please enter manually)

If "user growth plan" is selected:

Question 1：

What is the main growth channel?

A) Content marketing - Attract traffic through content B) Social spread - Viral growth through user sharing C) Paid promotion - Ad placement, KOL collaborations D) Offline promotion - Ground promotion, events, exhibitions E) Other (please enter manually)

Question 2：

How to activate users?

A) Beginner guidance - A clear onboarding process B) Incentive mechanism - Check-in rewards, new user red packets C) Social driven - Friend invitations, team collaboration D) Content recommendation - Personalized content push E) Other (please enter manually)

Question 3：

How to improve retention?

A) Membership system - Levels, points, privileges B) Content updates - Continuously provide fresh content C) Community operations - User interaction, UGC incentives D) Regular activities - Limited-time events, holiday marketing E) Other (please enter manually)

If "Business Model Design" is selected:

Question 1：

What is the main revenue source?

A) Direct payment - Users purchase products or services B) Subscription - Monthly/yearly fees (membership, SaaS) C) Advertising revenue - Ad displays, feed ads D) Transaction commission - Platform commission, service fees E) Other (please enter manually)

Question 2：

What is the pricing strategy?

A) Freemium - Basic free, advanced features paid B) Tiered pricing - Different prices for different features C) Pay-as-you-go - Charge by usage or duration D) One-time payment - Buyout model E) Other (please enter manually)

Question 3：

How to lower the payment threshold?

A) Free trial - Try first, pay later B) Tiered discounts - Long-term subscription discounts C) Referral rewards - Invite friends to get discounts D) Bundling - Package with other products E) Other (please enter manually)

If "User Experience Optimization" is selected:

Question 1：

What is the biggest current experience problem?

A) Complex operation - Too many steps, process too long B) Visual confusion - Unclear interface, information overload C) Insufficient feedback - Do not know the result of the operation D) Performance issues - Slow loading, lag E) Other (please enter manually)

Question 2：

How to simplify the operation process?

A) Smart recommendations - Reduce user decisions B) One-click operation - Reduce steps C) Voice interaction - Free hands D) Auto-fill - Remember user habits E) Other (please enter manually)

Question 3：

How to enhance user satisfaction?

A) Personalization - Adjust according to user preferences B) Instant feedback - Respond immediately after operation C) Surprise design - Unexpected easter eggs or rewards D) Emotional connection - Warm copy, holiday greetings E) Other (please enter manually)

Step 4: Summarize creative solutions

AI summarizes and generates a creative solution library based on user choices.

Step 5: Generate the creative solution library document according to user requirements

When only discussing, skip writing and summarize directly in the chat. For formal deliverables, first confirm the save location and overwrite risk, then create the directory, and then use the Write tool to create docs/01-需求调研/创意方案库.md：

# 创意方案库

## 一、基础信息

- **产品名称**: {PRODUCT_NAME}
- **发散方向**: {BRAINSTORM_DIRECTION}
- **生成时间**: {当前时间}

---

## 二、核心创意方案

### 方向 1: {根据用户选择生成标题}

**核心思路**:
{根据用户回答生成的详细描述}

**具体方案**:
1. {方案细节1}
2. {方案细节2}
3. {方案细节3}

**预期效果**:
{预期的业务效果}

---

### 方向 2: {第二个方向}

**核心思路**:
{描述}

**具体方案**:
1. {细节}
2. {细节}

**预期效果**:
{效果}

---

## 三、创意评分

| 方案 | 可行性 | 创新性 | 用户价值 | 商业价值 | 总分 |
|------|--------|--------|----------|----------|------|
| 方案1 | ⭐⭐⭐⭐ | ⭐⭐⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ | 16/20 |
| 方案2 | ⭐⭐⭐⭐⭐ | ⭐⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐⭐ | 16/20 |

---

## 四、下一步建议

建议执行：

1. **/pm-clarify** - 需求细化，明确具体需求细节（推荐）
2. **/pm-market** - 市场分析，验证方案可行性
3. **/pm-priority** - 优先级排序，决定先做哪个

---

**项目状态**: 创意发散完成
**生成时间**: {时间戳}
**生成工具**: super-pm

Step 6: Output completion prompt

Use AskUserQuestion to provide the next-step options:

✅ Brainstorming complete!

📄 If a creative solution library is actually generated:docs/01-需求调研/创意方案库.md; otherwise, only report the directions in the conversation.

🎯 Suggested next steps:

A) Run /pm-clarify - Requirement refinement, clarify details (recommended) B) Run /pm-market - Market analysis, validate feasibility C) Run /pm-priority - Priority ranking D) View the idea solution library

Cross-Agent fallback: If the current environment does not support AskUserQuestion, ask the same question via a regular chat message, then STOP and WAIT for user response.

Fallback mechanism

Scenario 1: No requirement research report

Provide a quick mode option, allowing manual input of basic information.

Scenario 2: User selects "Other"

Allow the user to input manually, and the AI understands and continues exploring.

Notes

One question at a time: Strictly follow to avoid information overload
3-5 options: Provide clear options for each question
Last item "Other": Allow the user to input freely
Progressive deepening: Continue deeper exploration based on the user's choices
Markdown storage: The generated solution library is human-readable and editable

Output quality comparison

✅ Good example：

- 有数据引用：「根据 Q4 数据，留存率从 35% 降至 28%」
- 有验证来源：「数据来源：Google Analytics, 2025-12-01」
- 有明确建议：「建议将新手引导步骤从 5 步减少至 3 步」

❌ Bad example：

- 模糊结论：「数据表明留存率有所下降」
- 无来源：「根据经验，这个功能很重要」
- 没有行动建议：「留存是个问题」

Common pitfalls / Red Flags — STOP

Stop immediately and backtrack if any of the following occur:

Pitfall	Correct approach
Using "should", "probably", "seems" to draw conclusions	Must be based on actual data and verification
Claiming completion without running checks	Verify first, then state
User requests lightweight or skip	Respect the scope, mark unexplored parts, do not falsely claim completion
The thought "this time it should be fine"	Re-verify every time

Output quality check / Verification Checklist

Prerequisite dependencies satisfied (input documents/data collected)
All core steps executed
Output document generated to docs/ directory
Every judgment supported by data/evidence
2-3 follow-up skills recommended

⚠️ If any item fails → complete it before marking as done.

About this skill

Input and mode

Cross-Agent interaction rules

Execution flow

Step 1: Read prerequisite data

Step 2: Determine the divergence direction

If "Inspiration spark stimulation" is selected (new in v2.1):

Step 2.5: Select Brainstorming Mode (new in v2.0)

Step 3: Progressive In-Depth Exploration

If "First Principles Decomposition" is selected (new in v2.0):

If "conventional divergent thinking" is selected:

If "product feature innovation" is selected:

If "user growth plan" is selected:

If "Business Model Design" is selected:

If "User Experience Optimization" is selected:

Step 4: Summarize creative solutions

Step 5: Generate the creative solution library document according to user requirements

Step 6: Output completion prompt

Fallback mechanism

Scenario 1: No requirement research report

Scenario 2: User selects "Other"

Notes

Output quality comparison

Common pitfalls / Red Flags — STOP

Output quality check / Verification Checklist