Preamble (run first)
echo "📊 PM-Feedback V2 - 用户反馈分析工具"
echo "支持并发分析:反馈分类 | 情感分析 | 优先级评估 | 问题归类"
echo ""
Write-time boundaries
When only discussing, do not write files; startup checks do not create directories. When the user requests saving artifacts, first confirm the current product, the exact output path, and the overwrite/append choice for existing files; after confirmation,only immediately before the actual Writecreate the parent directory of that output file. For complete installation, see skills/_shared/artifact-policy.md, installing this skill alone also follows this paragraph.
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 regular 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 generate documents or write to docs before the user answers.
- Existing docs files cannot substitute for the user's answer in this round.
Execution flow
Step 1: Collect feedback data (main agent - user interaction)
Use AskUserQuestion to ask:
📊 User Feedback Analysis - Data Source
Please provide user feedback data:
A) Read from file (input file path) B) Paste feedback content directly C) Crawl from app stores/social media (requires WebSearch) D) Use sample data for demonstration E) Other (please enter manually)
After the user chooses, obtain the feedback data.
Subagent concurrent analysis architecture
Architecture diagram
Step 2: Concurrent analysis (Subagent parallel execution)
Use the Agent tool to dispatch 4 subagents concurrently:
在一条消息中并发调用 4 个 Agent tool:
**Subagent 1: Feedback Categorization**
- type: "general-purpose"
- prompt: "分析用户反馈,进行分类(功能需求/Bug报告/体验问题/价格反馈/其他),输出到 docs/03-增长迭代/feedback-categories.md"
**Subagent 2: Sentiment Analysis**
- type: "general-purpose"
- prompt: "分析用户反馈情感倾向(正面/中性/负面),识别关键情绪点,输出到 docs/03-增长迭代/sentiment-analysis.md"
**Subagent 3: Priority Assessment**
- type: "general-purpose"
- prompt: "评估用户反馈优先级(P0-P3),基于影响面/紧急程度/实现成本,输出到 docs/03-增长迭代/priority-assessment.md"
**Subagent 4: Problem Clustering**
- type: "general-purpose"
- prompt: "对用户反馈问题进行聚类分析,识别核心问题群,输出到 docs/03-增长迭代/problem-clusters.md"
**并发执行,等待所有 subagent 完成**
Step 3: Main Agent integrates the analysis
Read all subagent analysis results:
read docs/03-增长迭代/feedback-categories.md
read docs/03-增长迭代/sentiment-analysis.md
read docs/03-增长迭代/priority-assessment.md
read docs/03-增长迭代/problem-clusters.md
Integrate into a comprehensive report:
Use Write to generate:docs/03-增长迭代/用户反馈分析报告.md
Comprehensive report structure
# 用户反馈分析报告
## 一、反馈概览
**数据来源**: [来源]
**反馈数量**: [总数]
**时间范围**: [时间段]
---
## 二、反馈分类统计
### 2.1 分类分布
[来自 feedback-categories.md]
| 类型 | 数量 | 占比 |
|------|------|------|
| 功能需求 | XX | XX% |
| Bug报告 | XX | XX% |
| 体验问题 | XX | XX% |
| 价格反馈 | XX | XX% |
| 其他 | XX | XX% |
### 2.2 高频反馈 TOP 10
1. [反馈内容] - XX 次
2. [反馈内容] - XX 次
...
---
## 三、情感分析
### 3.1 情感分布
[来自 sentiment-analysis.md]
| 情感 | 数量 | 占比 |
|------|------|------|
| 正面 | XX | XX% |
| 中性 | XX | XX% |
| 负面 | XX | XX% |
### 3.2 关键情绪点
**正面情绪**:
- [用户喜欢的地方]
**负面情绪**:
- [用户不满的地方]
---
## 四、优先级评估
### 4.1 优先级分布
[来自 priority-assessment.md]
| 优先级 | 数量 | 说明 |
|--------|------|------|
| P0 (紧急) | XX | 影响核心功能/大量用户 |
| P1 (高) | XX | 重要但不紧急 |
| P2 (中) | XX | 需要关注 |
| P3 (低) | XX | 可延后处理 |
### 4.2 P0 问题清单
1. [问题描述] - 影响:XX 用户
2. [问题描述] - 影响:XX 用户
...
---
## 五、问题聚类分析
### 5.1 核心问题群
[来自 problem-clusters.md]
**问题群 1: [问题类别]**
- 关联反馈:XX 条
- 典型描述:[用户原话]
- 根本原因:[分析]
**问题群 2: [问题类别]**
- 关联反馈:XX 条
- 典型描述:[用户原话]
- 根本原因:[分析]
---
## 六、改进建议
### 6.1 短期行动(1-2 周)
**紧急修复(P0)**:
1. [改进建议]
2. [改进建议]
**快速优化(P1)**:
1. [改进建议]
2. [改进建议]
### 6.2 中期规划(1-3 月)
**功能迭代**:
1. [功能需求] - P1 优先级
2. [功能需求] - P2 优先级
**体验优化**:
1. [优化点]
2. [优化点]
### 6.3 长期规划(3-6 月)
**战略改进**:
1. [战略建议]
2. [战略建议]
---
## 七、下一步建议
建议执行:
1. **pm-priority** - 对改进建议进行优先级排序
2. **pm-iteration** - 制定迭代计划
3. **pm-docs** - 更新产品文档
---
**分析时间**: 2026-XX-XX
**数据来源**: 用户反馈
**分析方法**: 多维度并发分析
Performance comparison
V1 vs V2 performance
| Metric | V1 (sequential analysis) | V2 (concurrent analysis) | Improvement |
|---|---|---|---|
| Analysis time | ~8 minutes | ~2 minutes | 4x |
| Main Agent context | ~40,000 tokens | ~10,000 tokens | 75% saved |
| Analysis dimensions | 1-4 sequential | 4 concurrent | - |
| Report quality | Single perspective | Multi-dimensional integration | ✅ |
Usage examples
Example 1: Read feedback from file
用户: 分析这批用户反馈,文件路径:docs/03-增长迭代/user-feedback.txt
AI: 🎯 读取反馈数据...
[Read file...]
🚀 并发分析中...
⏳ Subagent 1: Feedback Categorization - 完成 ✅
⏳ Subagent 2: Sentiment Analysis - 完成 ✅
⏳ Subagent 3: Priority Assessment - 完成 ✅
⏳ Subagent 4: Problem Clustering - 完成 ✅
🔧 整合分析结果...
✅ 生成报告: docs/03-增长迭代/用户反馈分析报告.md
📊 关键发现:
- 负面反馈占比 35%,主要集中在"搜索功能"
- P0 问题 3 个,需紧急修复
- 正面反馈集中在"界面设计"
💡 建议下一步:
1. /pm-priority - 对改进建议排序
2. /pm-iteration - 制定迭代计划
Subagent Prompt template
Feedback Categorization Subagent Prompt
你是用户反馈分类专家。执行以下任务:
**输入数据**: [反馈数据]
**分类维度**:
1. 功能需求 - 用户提出的新功能希望
2. Bug报告 - 用户遇到的错误/异常
3. 体验问题 - 用户使用过程中的不便
4. 价格反馈 - 对定价的意见
5. 其他 - 无法归类的反馈
**分析要求**:
- 对每条反馈进行分类
- 统计各类型数量和占比
- 提取高频反馈 TOP 10
**输出要求**:
生成结构化报告到:docs/03-增长迭代/feedback-categories.md
完成后立即返回结果。
Sentiment Analysis Subagent Prompt
你是用户情感分析专家。执行以下任务:
**输入数据**: [反馈数据]
**分析维度**:
1. 正面情感 - 用户满意/赞扬
2. 中性情感 - 客观陈述
3. 负面情感 - 用户不满/抱怨
**分析要求**:
- 对每条反馈进行情感判断
- 识别关键情绪点(正面/负面)
- 提取典型情感表达
**输出要求**:
生成结构化报告到:docs/03-增长迭代/sentiment-analysis.md
完成后立即返回结果。
Priority Assessment Subagent Prompt
你是优先级评估专家。执行以下任务:
**输入数据**: [反馈数据]
**评估维度**:
- P0 (紧急) - 影响核心功能/大量用户
- P1 (高) - 重要但不紧急
- P2 (中) - 需要关注
- P3 (低) - 可延后处理
**评估标准**:
1. 影响面 - 影响多少用户
2. 紧急程度 - 是否需要立即处理
3. 实现成本 - 修复难度和时间
**输出要求**:
生成结构化报告到:docs/03-增长迭代/priority-assessment.md
重点列出 P0 问题清单。
完成后立即返回结果。
Problem Clustering Subagent Prompt
你是问题聚类分析专家。执行以下任务:
**输入数据**: [反馈数据]
**聚类方法**:
1. 识别反馈中的问题关键词
2. 按问题类型进行聚类
3. 分析根本原因
**输出要求**:
生成结构化报告到:docs/03-增长迭代/problem-clusters.md
包含:
- 核心问题群(3-5 个)
- 每个问题群的关联反馈数量
- 典型用户描述
- 根本原因分析
完成后立即返回结果。
Notes
Feedback data quality
High-quality feedback:
- Specifically describes the problem
- Provides reproduction steps
- Includes environment information
Low-quality feedback:
- Vague description ("hard to use")
- Emotional expression (no specific content)
- Duplicate submissions
Analysis bias control
- Sample representativeness - Ensure diverse feedback sources
- Time window - Avoid looking only at short-term data
- Sentiment bias - Negative feedback typically outnumbers positive
- Verification mechanism - Key conclusions need cross-validation
Next-step recommendations
After completing the user feedback analysis, it is recommended to:
- pm-priority - Prioritize improvement suggestions
- pm-iteration - Develop an iteration plan
- pm-docs - Update product documentation
Super-PM - Make user feedback analysis deeper and more comprehensive 📊
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 occurs:
| Pitfall | Correct approach |
|---|---|
| Drawing conclusions using "should", "probably", "seems" | Must be based on actual data and verification |
| Claiming something is done without running checks | Verify first, then state |
| Skipping key steps due to time pressure | No exceptions; the tighter the time, the stricter you must be |
| The thought that "this time it should be fine" | Re-verify every time |
Output quality check / Verification Checklist
- Prerequisite dependencies are satisfied (input documents/data have been collected)
- All core steps have been executed
- The output document has been generated to the
docs/directory - Every judgment is supported by data/evidence
- 2-3 follow-up skills have been recommended
⚠️ If any item does not pass → complete it before marking as done.