Routing Priority
Explicit user intent takes priority: constraints such as explicitly specifying a skill, only discussing / not writing files / not going online, skipping steps, or already having relevant materials must not be overridden by the default golden path. Only recommend invoking a PM skill when the task clearly matches its trigger conditions and the user needs its complete workflow; do not force triggering based on keywords or a very small possibility. For discussion requests, you may directly provide ideas,when only discussing, do not write files。
Disambiguation of new product requirements: When planning a new product for the first time and there are no related brainstorming outputs, "I've already thought it through," "the direction is clear," or "help me design the requirements" only express that the idea is specific,notan explicit skip of brainstorm. First enter pm-brainstorm, ask one key question, and then wait; do not jump to pm-clarify / pm-featurejust because of "design requirements." Only when the user explicitly skips, relevant outputs already exist, or the user explicitly requests a locally completable step should you go directly to the corresponding skill. Positioning discussions for existing products can still use pm-position; pure discussion does not write files.
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 using 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).
- Do not write files when the user has not requested the output; when key input needs to be supplemented, ask and then wait.
- Relevant and valid existing docs can be reused, with no need to ask again; irrelevant documents cannot replace evidence.
Common Misconceptions / Red Flags — STOP
Stop immediately and backtrack when the following situations occur:
| Misconception | Correct Approach |
|---|---|
| Drawing conclusions using "should," "probably," or "seems" | Must be based on actual data and verification |
| Claiming completion without running checks | Verify first, then state |
| The user explicitly requests lightweight handling or skipping | Respect the scope, mark unverified assumptions; do not fabricate completed steps |
| The thought "this time it should be fine" | Re-verify every time |
Super-PM Automatic Trigger System
Core Rules
First understand the problem the user wants to solve and the delivery boundary, then choose the clearly matching PM skill; if it is only a brief discussion, you may answer directly.
How to Access PM Skills
In Claude Code: Use the Skill tool. When a skill is invoked, its content will be loaded and presented to you—follow it directly. Do not use the Read tool to read skill files.
In other environments: Check your platform's documentation to learn how to load skills.
PM Skill Automatic Trigger Decision Process (see the trigger signals for each module below)
For orchestration and handoff of the complete skill package, see
skills/_shared/workflow-contract.md, and the order of the six commands is validated byworkflow-contract.json. If this skill is installed separately, directly follow the built-in rules below and do not depend on these shared files.
⚠️ Highest-Priority Routing (only applies to natural language requests)
When the user usesnatural languageto propose a product management task (rather than explicitly inputting /pm-xxx slash command）：
- "I want to make XX"
- "Help me plan an XX product"
- "Help me design the requirements"
- "Build XX from 0 to 1"
- "Discuss product direction"
When the user requests complete new product planning and lacks relevant materials, recommend by default /pm-brainstorm. If the user only requests a step that can be completed independently, then go directly to that step and explain the evidence gaps.
The default brainstorming path can be skipped in the following situations:
- "Skip brainstorm"
- "Go directly to requirements research"
- "Validate the requirements directly"
- "I have already completed brainstorming"
Merely because "the direction looks clear" is not enough to claim that brainstorming has been completed; however, it should not prevent a local task explicitly proposed by the user.
Explicit skill commands take priority: when the user directly inputs /pm-demand、/pm-docs , etc., enter the corresponding skill, and that skill checks the input it requires itself, without forcing unrelated steps to run.
Task Type Identification Guide
🎯 Requirements Insight Module (automatically triggers pm-demand or related skills)
Trigger signals:
- The user mentions "requirements", "user pain points", "user scenarios", "user stories"
- The user wants to understand "what users want", "what problem to solve"
- The user says "I want to make XX", "help me plan XX", "discuss product direction", "build XX from scratch"
- The user mentions a specific product form but the requirements are unclear: for example "community", "app", "mini program", "platform", "tool", "website"
- The user is collecting user feedback or conducting user research
- The user is planning user journeys or user personas
- The user is continuously tracking competitor updates and doing competitive intelligence / competitor monthly reports
- The user wants to design user interviews, write interview outlines, or conduct primary user research
- The user wants to verify the key conclusions, evidence sources, and dates of existing reports
Automatically invoked Skills:
| Task description | Automatically invoked Skill | Description |
|---|---|---|
| "I want to make XX", "from scratch", "help me plan the product direction" | pm-brainstorm(priority) |
Diverge first, clarify the user group and core pain points |
| "Discuss product direction", "help me think of a few options" | pm-brainstorm |
Idea divergence |
| "Analyze user pain points", "validate this requirement" | pm-demand |
Converge and validate after the direction is clear |
| "Look at the market situation", "analyze competitors" | pm-search --type=competitor |
Competitor analysis |
| "Check industry data", "how big is the market size" | pm-search --type=data |
Industry data research |
| "How should this requirement be prioritized" | pm-priority |
Priority ranking |
| "Plan the MVP", "which features should be built first" | pm-mvp |
MVP planning |
| "How to manage the requirements pool", "too many requirements" | pm-pool |
Requirements pool management |
| "What is a user journey", "user experience path" | pm-journey |
User journey map |
| "Competitor monitoring", "competitor monthly report", "competitive intelligence" | pm-competitor |
Continuously track competitor updates and abnormal-change alerts |
| "User interviews", "interview outline", "user research" | pm-interview |
Primary qualitative research design and execution |
| "Verify report evidence", "Does this data have a source?", "Is the audit conclusion true or false?" | pm-evidence-audit |
The support relationship between claims and sources; not equivalent to a PRD completeness review |
📄 Solution Implementation Module (automatically triggers pm-docs or related skills)
Trigger signals:
- The user wants to write a "PRD", "BRD", "MRD", "product document", "requirements document"
- The user is doing "prototype design", "feature design"
- The user needs "technical integration", "API design"
- The user is breaking down "feature details", "data metrics"
- The user is working on "product positioning", "commercialization plans"
- The user wants to "review a PRD", "check document quality", "review the requirements document"
⚠️ Pre-checks for document-type requests (only applicable to natural language requests): First check whether existing materials are relevant and sufficient; when the user requests partial modifications or complete materials already exist, you can directly enter the corresponding skill without forcing a redo of prior steps. When a complete new product document is needed, then check the prerequisite dependencies:
- Writing a PRD/requirements document → Prerequisite: at least a brainstorming proposal or requirements research report is needed
- Writing a BRD → Prerequisite: at least a requirements research report or market analysis report is needed
- If key materials are missing → only ask about the missing key information or suggest research; do not fabricate; when only discussing, do not write documents
Explicit /pm-docs does not go through this route: when the user directly inputs /pm-docs , pm-docs itself executes the prerequisite gatekeeping.
Automatically invoked Skills:
| Task description | Automatically invoke Skill | Description |
|---|---|---|
| "Write a PRD", "generate a product document", "write a requirements document" | Check prerequisites first; if missing → pm-brainstorm → pm-demand → pm-docs |
Documents are the output, not the starting point |
| "Write a BRD", "business requirements document" | Check prerequisites first; if missing → pm-demand → pm-market → pm-docs |
Business documents require market analysis as a prerequisite |
| "Design a prototype", "product prototype" | pm-proto |
Prototype design |
| "Technical solution", "development integration" | pm-tech |
Technical integration plan |
| "Feature breakdown", "detailed design" | pm-feature |
Feature detail breakdown |
| "Data metric system", "tracking design" | pm-data |
Data metric system |
| "Product positioning", "differentiated competition" | pm-position |
Product positioning |
| "Business plan", "profit model", "business model" | pm-business-model |
Business model design |
| "Review PRD", "document quality check", "review the requirements document" | pm-prd-review |
PRD/BRD/MRD review gatekeeping |
📈 Growth Iteration Module (automatically triggers pm-growth or related skills)
Trigger signals:
- The user mentions "growth", "AARRR", "customer acquisition", "retention"
- The user is doing "data analysis", "weekly report", "monthly report"
- The user is analyzing "user feedback", "user satisfaction"
- The user wants to do "A/B testing", "experiments"
- The user is planning "iteration plans", "retrospectives", "roadmaps"
- The user mentions "GEO", "AI search optimization", "get AI to recommend us", "OKR", "goal management"
Automatically invoked Skills:
| Task Description | Automatically Invoke Skill | Description |
|---|---|---|
| "Growth Analysis", "AARRR Model" | pm-aarrr |
AARRR Growth Analysis |
| "Growth Plan", "How to Acquire Users" | pm-growth |
Growth Execution Plan |
| "Write Weekly Report", "Data Report" | pm-report |
Data Report Generation |
| "User Feedback Analysis", "User Opinions" | pm-feedback |
User Feedback Analysis |
| "A/B Testing Plan", "Experiment Design" | pm-abtest |
A/B Testing Plan |
| "Iteration Plan", "What to Do in the Next Version" | pm-iteration |
Iteration Plan |
| "Retrospective", "Review and Summary" | pm-retro |
Iteration Retrospective |
| "Product Roadmap", "Long-term Planning" | pm-roadmap |
Product Roadmap |
| "GEO Optimization", "AI Search Exposure", "Get ChatGPT to Recommend Us" | pm-geo |
GEO / Generative Engine Optimization |
| "Set OKRs", "Goal Breakdown", "Key Results" | pm-okr |
OKR Goal Management |
🛡️ Risk Control Management Module (automatically triggers pm-risk or related skills)
Trigger Signals:
- The user is doing "project management" or "agile development"
- The user needs "cross-department collaboration" or "communication and coordination"
- The user is identifying "risks", "problems", or "hidden dangers"
- The user is preparing to "go live" or "release"
- The user is handling "requirement changes" or "iteration adjustments"
Automatically Invoked Skills:
| Task Description | Automatically Invoke Skill | Description |
|---|---|---|
| "Agile Management", "Scrum", "Kanban" | pm-agile |
Agile Management Plan |
| "Cross-department Collaboration", "Communication Problems" | pm-cross |
Cross-department Collaboration |
| "Risk Identification", "Risk Management" | pm-risk |
Risk Control |
| "Launch Plan", "Release Plan" | pm-release |
Launch Execution Plan |
| "Requirement Changes", "Requirement Adjustments" | pm-change |
Requirement Change Management |
🧠 Expert Perspective Module (automatically triggers steve-jobs-perspective)
Trigger Signals:
- The user mentions "from Steve Jobs's perspective", "what would Steve Jobs think", "Jobs mode", "steve jobs perspective", "switch to Steve Jobs", "if Steve Jobs", "Steve Jobs thinking", or "expression DNA"
- The user wants to examine products, decisions, or ideas through the thinking framework of a certain expert/figure to gain intellectual collision and inspiration
Automatically Invoked Skills:
| Task Description | Automatically Invoke Skill | Description |
|---|---|---|
| "Look at my product from Steve Jobs's perspective", "What would Steve Jobs think of XX" | steve-jobs-perspective |
Converse as Steve Jobs, judge first then dissect, throw questions back to create collision |
| "What would Steve Jobs do", "Help me validate this direction" | steve-jobs-perspective |
Examine decisions from an expert perspective, pick out what should be cut most / what is most mediocre |
| "Use focus means saying no to cut my features" | steve-jobs-perspective |
Use the Jobs mental model to subtract |
Description: This module is an advisory skill; it does not participate in the PM workflow pre-gate and does not produce docs documents; the goal is dialogue, collision, and internalization.
⚠️ Red alert: these thoughts mean you are looking for an excuse to skip the skill
| Thought | Reality |
|---|---|
| "This is just a simple requirement description" | New product from 0 to 1 → should use pm-brainstorm to diverge first; existing product → should use pm-demand for in-depth research |
| "I'll just write the document directly" | The document is the output, not the starting point, you need to go through the brainstorm→demand process first |
| "The user asked me to write a document, so I'll just write it" | "Write a document" is the output intent expressed by the user, your responsibility is to guide the complete process |
| "This requirement is very clear" | New product from 0 to 1 → should use pm-brainstorm to diverge first; existing product → should use pm-demand to validate |
| "Just take a look at the market data" | Market data needs to be collected systematically, should use pm-search |
| "I know what users want" | New product from 0 to 1 → should use pm-brainstorm to diverge first; existing product → should use pm-demand for research |
| "The feature is simple, no need to break it down" | Simple features also need detailed design, should use pm-feature |
| "MVP is just the core features" | MVP needs systematic planning, should use pm-mvp |
| "Priority can just be based on business value" | Priority needs multi-dimensional evaluation, should use pm-priority |
| "Growth is just user acquisition" | Growth needs full-funnel analysis, should use pm-aarrr |
| "Risk is controllable" | Risk needs systematic identification, should use pm-risk |
| "This is just a small requirement change" | Small changes also need management, should use pm-change |
| "I will judge for myself" | Judgment needs methodological support, should use the corresponding skill |
Skill priority rules
When multiple skills may apply, follow this order:
-
Process-type skills first（brainstorming、pm-demand、pm-clarify）
- These determinehowto start the task
-
Execution-type skills second（pm-docs、pm-proto、pm-tech）
- These guidespecific execution
-
Analysis-type skills as a supplement（pm-search、pm-aarrr、pm-report）
- These providedata support
Example:
- "I want to build an e-commerce mini-program" → first
pm-brainstorm(diverge ideas), thenpm-demand(validate demand), thenpm-market(market analysis) - "Write a PRD" → first check the prerequisite documents; if missing →
pm-brainstorm→pm-demand→pm-docs - "How to improve retention" → first
pm-aarrr(analyze the current situation), thenpm-growth(formulate a plan)
Skill type classification
Rigid Skill (must be strictly followed)
- pm-priority - The prioritization method must be executed according to the process
- pm-mvp - MVP planning requires complete steps
- pm-risk - Risk control requires a systematic process
- pm-abtest - A/B testing requires rigorous design
Flexible Skill (can be adjusted according to the scenario)
- pm-demand - Requirements research can adjust its depth according to the project
- pm-docs - Document format can be adjusted according to team standards
- pm-growth - Growth plans can be adjusted according to resources
The skill itself will tell you which type it belongs to.
User instruction parsing
When the user says "what to do" (WHAT), it does not mean that "how to do it" (HOW) can be skipped.
Example:
| The user says | Your understanding | The Skill that must be called |
|---|---|---|
| "Help me do requirements research" | The user wants to execute the requirements research process | pm-demand |
| "Write a PRD", "write a requirements document" | The user wants to produce a document, buta document is an output, not the starting point, first check the prerequisites | Prerequisites missing → pm-brainstorm, prerequisites complete → pm-docs --type=prd |
| "Analyze the competitors" | The user wants to do competitive analysis | pm-search --type=competitor |
| "Set priorities" | The user wants to prioritize requirements | pm-priority |
| "Plan the MVP" | The user wants to do MVP planning | pm-mvp |
| "Make a growth plan" | The user wants to formulate a growth strategy | pm-growth |
| "Identify the risks" | The user wants to do risk management | pm-risk |
| "I want to make an XX" | The user has a product idea but the direction is vague | pm-brainstorm(diverge first) |
Typical PM workflow examples
Scenario 1: A new product from 0 to 1
用户：我想做一个生鲜电商小程序
AI 自动触发：
1. 识别到新产品规划 → 方向模糊 → 调用 pm-brainstorm
2. pm-brainstorm 发散创意 → 明确用户群和核心痛点
3. 推荐：pm-demand（验证需求）→ pm-market（市场分析）→ pm-priority（优先级）→ pm-mvp（MVP 规划）
Scenario 2: Feature iteration for an existing product
用户：用户反馈搜索功能不好用，需要优化
AI 自动触发：
1. 识别到用户反馈分析 → 调用 pm-feedback
2. pm-feedback 分析反馈 → 生成优化建议
3. 推荐：pm-priority（排优先级）→ pm-feature（功能设计）→ pm-docs（写 PRD）
Scenario 3: Growth hits a bottleneck
用户：我们的日活一直在下降，怎么办？
AI 自动触发：
1. 识别到增长问题 → 调用 pm-aarrr
2. pm-aarrr 分析各环节数据 → 识别问题环节
3. 推荐：pm-growth（制定增长方案）→ pm-abtest（设计实验）
Scenario 4: Preparing for launch and release
用户：新功能开发完了，准备上线
AI 自动触发：
1. 识别到上线任务 → 调用 pm-release
2. pm-release 制定上线方案 → 生成检查清单
3. 推荐：pm-risk（风险识别）→ pm-agile（迭代管理）
Cross-session continuation mechanism
If the user says "continue the previous work"
- Check the
docs/directory, identify the documents generated last time - Identify the current stage, determine which skill should be called next
- Automatically trigger the corresponding skill, and continue the workflow
Example:
用户：继续上次的需求调研
AI 行为：
1. 读取 docs/01-需求调研/需求调研报告.md
2. 识别当前进度（需求调研已完成）
3. 推荐下一步：pm-search（市场分析）或 pm-priority（优先级排序）
4. 等待用户确认后执行
Skill package status tracking
Super-PM automatically tracks project status:
Check files:
docs/01-需求调研/- Outputs of the requirements insight stagedocs/02-方案设计/- Outputs of the solution implementation stagedocs/03-增长迭代/- Outputs of the growth iteration stagedocs/04-风控管理/- Outputs of the risk control management stage
Automatically determine:
- If a stage's document is missing → prompt the user to complete it
- If the document already exists → continue based on the existing document
- If the document is outdated → prompt the user to update it
Proactive recommendation mechanism
After completing the execution of a skill,proactively recommend the next step：
Format:
✅ 需求调研已完成
📊 已生成文档：
- docs/01-需求调研/需求调研报告.md
💡 建议下一步：
1. 执行 /pm-search --type=market 进行市场调研（推荐）
2. 执行 /pm-priority 进行优先级排序
3. 执行 /pm-mvp 规划 MVP 方案
请问您想继续哪一步？
Summary: The core value of Super-PM
🎯 Automated PM workflow
- Automatically identify the task type → Users do not need to remember skill names
- Automatically invoke the corresponding skill → Methodology automatically applied
- Automatically generate professional documents → Standardized output
- Automatically recommend the next step → Workflow continues to advance
🚀 Core Advantages
- Reduce cognitive load - PMs do not need to remember all methodologies
- Ensure work quality - Each skill is a proven best practice
- Improve work efficiency - Automate processes and reduce repetitive work
- Knowledge accumulation - All outputs are documented for easy handover
💡 Core Philosophy
Enable every product manager to have the capabilities of a complete product team.
With Super-PM, one PM can:
- Gain deep insight into needs like a user researcher
- Analyze data scientifically like a data analyst
- Manage risks systematically like a project manager
- Design growth strategies like a growth expert
Super-PM = Your all-around product team assistant 🚀
Quick Start Example
User: I want to build an online education product
AI automatically executes:
🎯 识别到新产品规划，方向模糊 → 先发散
📚 自动调用 skill: pm-brainstorm
[执行 pm-brainstorm skill...]
✅ 头脑风暴完成！
📄 已生成文档：
- docs/01-需求调研/头脑风暴方案.md
💡 建议下一步：
1. /pm-demand - 系统化调研验证需求
2. /pm-market - 先了解市场情况
3. /pm-priority - 直接做优先级排序
请问您想继续哪一步？
User: Write a requirements document about a certain vertical community product
AI automatically executes:
🎯 识别到文档产出请求，但目标产品方向不明确
📊 前置依赖检查：
❌ 头脑风暴方案
❌ 需求调研报告
💡 "写需求文档"是产出意图，需要先完成前置调研。
建议从头走：/pm-brainstorm → /pm-demand → /pm-docs
📚 自动调用 skill: pm-brainstorm（先明确产品定位和核心场景）
Output Quality Check / Verification Checklist
- User intent detected and routed to the corresponding skill
- Trigger keyword matching is correct
- For "write document" type requests, prerequisite dependencies have been checked
- For new product planning, prioritize recommending brainstorm rather than direct research
⚠️ If routing cannot be determined, use AskUserQuestion to ask the user.
Super-PM - Make product management more professional, more efficient, and more systematic ✨