关于 review-analyst-agent
review-analyst-agent 是用于分析产品评论的技能,通过 4 个专门 Agent 汇总情感、投诉、好评和功能需求,并形成改进建议。它适合需要从多来源评论中确定问题优先级的产品团队。
主要功能
- 01从多个来源收集评论
- 02分析情感与趋势
- 03按频率和严重性归类投诉
- 04识别好评与功能需求
- 05生成优先级矩阵和行动计划
适用场景
- 01分析自家产品的客户反馈
- 02比较产品与竞品的评论
- 03从应用商店和 Reddit 提取功能需求
- 04根据评论确定改进优先级
安装
复制下面的命令,在终端中运行。
npx skills add https:/ / github.com/ michaelboeding/ skills --skill review-analyst-agent安装后在 Claude Code 中输入/review-analyst-agent确认已加载。
包含的文件
除 SKILL.md 外,安装时一起下载的参考资料和脚本。
agents/improvement-recommender.mdissue-identifier.mdreview-scraper.mdsentiment-analyzer.md详细介绍
评测分析 Agent
分析产品评测,找出问题并确定改进优先级。
此技能使用 4 个专用 Agent,从不同角度分析评测,然后综合得出可执行的建议。
产出内容
| 输出 | 描述 |
|---|---|
| 情感概览 | 整体情感分布(正面/中性/负面) |
| 主要投诉 | 按频率和严重程度排序的问题列表 |
| 主要好评 | 客户喜爱的方面(应予以保留/强调) |
| 功能请求 | 客户希望拥有但目前不存在的功能 |
| 优先级矩阵 | 关键/重要/锦上添花的改进项 |
| 行动计划 | 具体建议及预期影响 |
前提条件
- 可访问网络以抓取评测
- 无需 API 密钥
工作流
第 1 步:确定产品和来源(必需)
⚠️ 不要跳过此步骤。使用交互式提问——每次只问一个问题。
提问流程
⚠️ **对以下每个问题都使用 AskUserQuestion 工具。**不要只在回复中列出问题——请使用该工具创建包含所示选项的交互式提示。
Q1:产品
"我将为你的产品分析评测!首先——这是什么产品?
(产品名称或 URL)"
等待回复。
Q2:来源
"应该去哪里查找评测?
- Amazon
- App Store / Google Play
- G2 / Capterra
- 以上所有来源
- 或指定来源"
等待回复。
Q3:背景
"这是你的产品还是竞争对手的产品?
(有助于确定建议的侧重点)"
等待回复。
Q4:问题
"是否有你希望我验证或调查的已知问题?
- 有——请描述
- 没有——找出所有问题"
等待回复。
快速参考
| 问题 | 决定事项 |
|---|---|
| 产品 | 分析对象 |
| 来源 | 从哪里抓取评测 |
| 背景 | 建议的侧重点 |
| 问题 | 分析的重点领域 |
第 2 步:收集评测
使用浏览器工具从以下来源抓取评测:
| 来源类型 | 平台 |
|---|---|
| 电商 | Amazon、Walmart、Target、Best Buy |
| 软件 | G2、Capterra、TrustRadius、Product Hunt |
| 应用 | App Store、Google Play Store |
| 综合 | Trustpilot、BBB、Yelp |
| 社交媒体 | Reddit、Twitter/X、YouTube 评论 |
| 论坛 | 特定产品的社区 |
每条评测收集以下内容:
- 评分(如有)
- 日期
- 评测文本
- 有用票数(如有)
第 3 步:并行运行专用分析 Agent
部署 4 个 Agent,分别从不同角度进行分析:
Agent 1:评测抓取器
重点:从多个来源查找并收集评测
Tasks:
- Navigate to review platforms
- Extract review text and ratings
- Collect metadata (date, helpful votes)
- Handle pagination
- De-duplicate reviews
Agent 2:情感分析器
重点:分析情感及情绪模式
Analyze:
- Overall sentiment (positive/neutral/negative)
- Emotional intensity
- Frustration indicators
- Satisfaction indicators
- Sentiment trends over time
Agent 3:问题识别器
重点:对投诉分类并发现规律
Identify:
- Common complaint themes
- Frequency of each issue
- Severity indicators
- Specific quotes as evidence
- Root cause patterns
Agent 4:改进建议器
重点:确定优先级并提出修复建议
Recommend:
- Priority ranking of issues
- Specific improvement suggestions
- Expected impact of each fix
- Quick wins vs long-term investments
- Competitive gaps to address
第 4 步:综合生成分析报告
将所有 Agent 的输出整合为结构化报告:
{
"product": {
"name": "Product Name",
"sources_analyzed": ["Amazon (342 reviews)", "Reddit (89 posts)", "G2 (56 reviews)"],
"total_reviews": 487,
"date_range": "Jan 2025 - Jan 2026",
"analysis_date": "2026-01-04"
},
"sentiment": {
"overall_score": 3.8,
"breakdown": {
"positive": 62,
"neutral": 18,
"negative": 20
},
"trend": "Improving (up from 3.5 six months ago)",
"net_promoter_estimate": 32
},
"top_complaints": [
{
"rank": 1,
"issue": "Battery drains too fast",
"frequency": 47,
"percentage": "23% of negative reviews",
"severity": "High",
"sample_quotes": [
"Battery only lasts 2 hours, not the 8 advertised",
"Have to charge it 3x per day",
"Battery life is a dealbreaker"
],
"root_cause": "Hardware limitation or software optimization needed",
"recommendation": "Improve battery capacity or optimize power consumption",
"expected_impact": "Could improve rating by 0.3-0.5 stars"
},
{
"rank": 2,
"issue": "App crashes frequently",
"frequency": 32,
"percentage": "16% of negative reviews",
"severity": "High",
"sample_quotes": [
"App crashes every time I try to sync",
"Lost all my data after app crashed"
],
"root_cause": "Sync functionality stability",
"recommendation": "Stability audit of mobile app, fix crash on sync",
"expected_impact": "Could reduce 1-star reviews by 15%"
}
],
"top_praise": [
{
"feature": "Build quality",
"frequency": 89,
"percentage": "45% of positive reviews",
"sample_quotes": [
"Feels premium in hand",
"Solid construction, very durable"
],
"recommendation": "Emphasize in marketing, protect in future versions"
}
],
"feature_requests": [
{
"request": "Water resistance",
"frequency": 23,
"sample_quotes": [
"Wish I could use it in the rain",
"Would pay extra for waterproof version"
],
"recommendation": "Consider for v2 or premium tier"
}
],
"competitor_mentions": [
{
"competitor": "Competitor X",
"context": "Switching from",
"frequency": 15,
"sentiment": "Mixed - some prefer us, some prefer them"
}
],
"priority_matrix": {
"critical": [
{"issue": "Battery life", "reason": "Top complaint, high severity"},
{"issue": "App crashes", "reason": "Causes data loss, drives 1-star reviews"}
],
"important": [
{"issue": "Water resistance", "reason": "Frequent request, competitive gap"}
],
"nice_to_have": [
{"issue": "Color options", "reason": "Low frequency, low impact"}
]
},
"action_plan": [
{
"priority": 1,
"action": "Fix app crash on sync",
"effort": "Medium",
"impact": "High",
"expected_outcome": "Reduce 1-star reviews by 15%"
},
{
"priority": 2,
"action": "Improve battery life or set realistic expectations",
"effort": "High",
"impact": "High",
"expected_outcome": "Improve rating by 0.3-0.5 stars"
},
{
"priority": 3,
"action": "Add water resistance to roadmap for v2",
"effort": "High",
"impact": "Medium",
"expected_outcome": "Address top feature request"
}
]
}
第 5 步:提供可执行的洞察
交付消息:
"✅ 评测分析已完成!
产品: [Name] 已分析评测: 来自 [Sources] 的 [Count] 条 整体情感: [Score](正面评价占 [Positive]%)
排名前三的问题(按频率):
- 🔴 [Issue 1] - 占投诉的 [X]%
- 🔴 [Issue 2] - 占投诉的 [X]%
- 🟡 [Issue 3] - 占投诉的 [X]%
客户喜爱的方面: ✅ [Praised feature 1] ✅ [Praised feature 2]
优先行动: → 优先修复 [Top Issue]——预计可将评分提高 [X]
需要我:
- 深入分析某个问题?
- 对比竞争对手的评测?
- 跟踪一段时间内的变化?
- 制定改进路线图?"
与其他 Agent 的集成
review-analyst-agent
↓ "Battery is top complaint"
product-engineer-agent
↓ "Design better battery solution"
patent-lawyer-agent
↓ "Check if solution is patentable"
copywriter-agent
↓ "Update marketing to address concern"
| Agent | 如何使用评价数据 |
|---|---|
product-engineer-agent | 了解需要修复或改进的内容 |
competitive-intel-agent | 与竞争对手的评价进行比较 |
market-researcher-agent | 验证市场需求 |
copywriter-agent | 在营销内容中回应用户顾虑 |
pitch-deck-agent | 展示以客户为中心的改进 |
media-utils | 根据分析生成 PDF 报告 |
生成 PDF 报告
完成分析后,询问是否要生成 PDF:
“要我为这份评价分析生成一份 PDF 报告吗?”
python3 ${CLAUDE_PLUGIN_ROOT}/skills/media-utils/scripts/report_to_pdf.py \
--input review_analysis.md \
--output review_analysis.pdf \
--title "Customer Review Analysis" \
--style business
Agents
| Agent | 文件 | 侧重点 |
|---|---|---|
| Review Scraper | review-scraper.md | 查找并收集评价 |
| Sentiment Analyzer | sentiment-analyzer.md | 分析情感倾向模式 |
| Issue Identifier | issue-identifier.md | 对投诉进行分类 |
| Improvement Recommender | improvement-recommender.md | 排定优先级并提出建议 |
示例提示词
你的产品:
“分析 Amazon 上我们蓝牙耳机的评价”
竞争对手:
“大家在抱怨 Notion 的哪些方面?”
比较:
“比较我们产品与竞争对手 X 的评价”
聚焦功能:
“从 App Store 和 Reddit 中查找针对我们移动应用的功能请求”
优先级:
“根据客户反馈,我们应该先修复什么?”
趋势:
“过去 6 个月的情感倾向发生了怎样的变化?”