Social Intel Hub

社媒情报中心 - 多平台爬虫 + 数据分析 + AI洞察 + 词云 + 趋势追踪

Security Scan

The skill's stated purpose (multi‑platform social crawler + analysis) largely matches its code, but metadata and instructions reference hardcoded user paths and rely on an external 'mcporter' tool — these mismatches and implicit trust in mcporter are concerning and should be reviewed before installing.

Purpose & Capability

The name/description (multi‑platform crawler, analysis, wordclouds) aligns with what the code does: it runs searches via a local 'mcporter' CLI, parses platform results, generates analytics and wordclouds. Requiring python3 and wordcloud/matplotlib/openpyxl is proportional. However the SKILL.md metadata lists specific user directories (/Users/kk/openclaw-media, /Users/kk/openclaw-crawl4ai) and a core script path under /Users/kk/.openclaw which are user-specific and not justified by the generic description.

Instruction Scope

Runtime instructions and the included script invoke the local 'mcporter' binary to call remote TikHub services, read/write cache and output under /tmp, and reference a hardcoded workspace path (/Users/kk/.openclaw/...). The instructions thus depend on an external binary and a particular local environment; they also give the skill freedom to call arbitrary mcporter methods (server.method), which broadens its runtime scope beyond simple 'search and analyze'.

Install Mechanism

There is no automated install spec besides a manual pip suggestion (pip3 install wordcloud matplotlib openpyxl). No downloaded archives or remote installers are used. This is low risk, but the presence of an included Python script means the skill will execute code locally — review is recommended.

Credentials

No environment variables or credentials are requested (good), but the metadata declares required directories under a specific user's home and requires the 'mcporter' binary. Requiring access to specific paths in another user's home is disproportionate for a general-purpose skill and suggests the package was packaged from a developer machine without sanitization. The skill also implicitly trusts the 'mcporter' CLI to access external services — that is a privileged dependency and should be validated.

Persistence & Privilege

The skill does not request always:true, does not alter other skills, and only writes cache/output to /tmp and its own output dir. It does perform local caching and file exports (CSV/Excel/images) which is expected for this functionality.

What to consider before installing

This skill appears to implement the described crawler and analysis, but there are red flags you should consider before installing: - Hardcoded developer paths: SKILL.md and the script reference /Users/kk/... and expect directories that likely don't exist on your machine. Ask the author why those directories are required or sanitize them before use. - Trusted external CLI: The skill uses a local 'mcporter' binary to call remote TikHub services (mcporter call server.method). You must trust this binary and its configuration — inspect mcporter, its config, and what endpoints it will call. If mcporter is malicious or misconfigured it could exfiltrate data or perform unexpected network actions. - Local execution risk: The package includes and runs a Python script that will execute subprocess calls and write cache/output files under /tmp. Review social_intel.py fully (network calls, file reads) and run in an isolated environment (container/VM) if you have any doubt. - Mitigations: run the script in a sandbox, validate or replace hardcoded paths, confirm mcporter binary origin and behavior, and verify that no sensitive local files or credentials are referenced by the script or mcporter before giving it network access. If you want, I can: (1) point out exact lines in social_intel.py that call mcporter and write files, (2) suggest a safe wrapper to run it in a temp environment, or (3) produce a sanitized SKILL.md that removes user‑specific paths.

Social Intel Hub v2 — 社媒情报中心

一句话需求，结构化数据 + 可视化 + AI洞察，全套出来。

触发方式

"爬小红书服装数据并分析" "分析微博AI相关内容" "生成奶茶的词云图" "对比咖啡和茶饮的差异"

核心脚本

/Users/kk/.openclaw/workspace/skills/social-intel/social_intel.py

支持平台

平台	参数	主要数据
小红书	xhs	笔记/点赞/收藏/评论数
微博	weibo / wb	帖子/转发/评论数
B站	bilibili / bili	视频/点赞/浏览/弹幕
抖音	douyin / dy	视频/点赞/浏览/评论
快手	kuaishou / ks	短视频/点赞/浏览
TikTok	tiktok	视频
YouTube	youtube / yt	视频/字幕

意图索引

用户意图	触发条件	功能
搜索 + 分析	"爬/分析/查看+平台+关键词"	默认模式
多平台并行	"所有平台+关键词"	--all-platforms
竞品对比	"对比/比较+关键词1+关键词2"	--compare
趋势追踪	"趋势/近期+关键词"	--trend
词云图	"词云/可视化+关键词"	--wordcloud
评论采集	"评论+关键词"	--comments
数据导出	"导出/csv/excel+关键词"	--export

使用示例

基础搜索 + 分析报告

python3 social_intel.py -k "咖啡" -p xhs -m 30 --analyze

输出：高频标签 TOP15 + 高赞内容 + 日期趋势 + AI洞察prompt

多平台并行搜索

python3 social_intel.py -k "AI" --all-platforms -m 20

同时爬取微博/B站/小红书/抖音/快手，一次看全网反应

竞品对比

python3 social_intel.py -k "对比" -p xhs --compare "咖啡" "奶茶" "气泡水"

输出对比表：哪个话题更热、标签差异、用户偏好

趋势追踪（近7天）

python3 social_intel.py -k "新品" -p xhs --trend

每天一条快照，自动判断上升/下降/平稳趋势

生成词云图

python3 social_intel.py -k "穿搭" -p xhs -m 30 --wordcloud -o /tmp/chuanda.png

采集指定笔记评论

python3 social_intel.py -k "咖啡" -p xhs --comments 69aae46f000000001a0298f1

导出数据

# CSV
python3 social_intel.py -k "运动鞋" -p xhs -m 50 --export csv -o /tmp/shoes.csv

# Excel（多平台对比）
python3 social_intel.py -k "咖啡" --all-platforms -m 30 --export excel -o /tmp/coffee_multi.xlsx

参数说明

参数	说明
-k / --keyword	搜索关键词（必填）
-p / --platform	平台，默认 xhs；all 表示全部平台
-m / --max	最大笔记数，默认 20
--pages	搜索页数，默认 1（每页约 20 条）
--analyze	输出完整分析报告（默认开启）
--all-platforms	并行搜索所有平台
--compare KW1 KW2...	竞品对比模式
--trend	近7天趋势分析
--wordcloud	生成词云图 PNG
--comments NOTE_ID	采集指定笔记评论
--export csv|excel|both	导出格式
-o / --output	输出文件路径
--no-cache	禁用缓存，强制重新请求

分析报告内容

• 📈 总体指标：总点赞/收藏/浏览/评论 + 均值
• 🏷️ 高频话题：TOP20 标签 + 可视化条形图
• 🔥 TOP3：点赞最高 + 评论最多 + 浏览最多
• 📅 趋势：按日期聚合 + 热度条形图
• 💡 AI洞察：自动生成可复制的 LLM 分析 prompt

缓存机制

• 自动缓存：每个搜索请求缓存 24 小时
• 缓存路径：/tmp/social_intel_cache/
• 重复查询走缓存，不消耗 TikHub 余额
• 用 --no-cache 强制刷新

依赖

包	用途	安装
wordcloud	词云生成	pip3 install wordcloud
matplotlib	词云可视化	pip3 install matplotlib
openpyxl	Excel 导出	pip3 install openpyxl

底层架构

用户需求 → social_intel.py（统一入口）
    ├── TikHub MCP（主力，零登录，7平台并行）
    ├── 24h 缓存（节省余额）
    ├── 数据分析（词频/趋势/竞品对比）
    ├── 词云生成（wordcloud）
    ├── 导出（CSV/Excel）
    └── AI洞察 prompt（喂给 MiniMax LLM）

后续可扩展方向

• MediaCrawler 评论情感分析（需登录）
• MiniMax LLM 直接生成洞察文字
• Cron 定时趋势监控 + 飞书推送
• 热度预测模型