Help us improve
Share bugs, ideas, or general feedback.
From wechat-insight
Extracts and analyzes local WeChat Mac 4.x chat history — exports encrypted messages, generates daily reports, customer insights, contact labels, and interactive HTML reports with heuristic emotion/MBTI analysis.
npx claudepluginhub caigee-cmd/wechat-insight --plugin wechat-insightHow this skill is triggered — by the user, by Claude, or both
Slash command
/wechat-insight:analyzing-wechat-chatsThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
> **前置依赖**:本 skill 不是自包含的,运行需要 `./wechat-insight` CLI(仓库:https://github.com/caigee-cmd/wechat-insight)。
Summarizes WeChat group chat highlights into a structured digest using the local wx-cli binary. Maintains per-group history and per-user profiles across runs with privacy guardrails.
Automates WeChat Desktop on macOS/Windows via screenshots, clicks, typing for sending messages, reading chats, searching contacts, forwarding, broadcasting.
Ingests recent messages from WhatsApp chats (group or direct) into Obsidian vault as daily markdown files. Auto-creates Decision Log stubs for keywords like incident, outage. Invoke via /ingest-whatsapp <chat> [--days N].
Share bugs, ideas, or general feedback.
前置依赖:本 skill 不是自包含的,运行需要
./wechat-insightCLI(仓库:https://github.com/caigee-cmd/wechat-insight)。触发本 skill 时按以下顺序判断当前 CLI 位置:
- 当前目录有
wechat-insight可执行文件 → 直接用./wechat-insight ...~/.local/share/wechat-insight/wechat-insight存在 →cd ~/.local/share/wechat-insight后用./wechat-insight ...- 都没有 → 提示用户、并提议执行一行安装(macOS only):
脚本会 clone 仓库到curl -sL https://raw.githubusercontent.com/caigee-cmd/wechat-insight/main/install.sh | bash~/.local/share/wechat-insight、创建 venv、安装 Python 依赖。装好后再cd进去触发本 skill。./wechat-insightlauncher 会自动使用.venv,不需要手动source activate。
这是一个 本地微信分析工作台 v1。
当前已经可用的能力:
启发式分析能力(基于聊天文本的统计规则推测,不是医学诊断、心理测评或模型级结论,结果仅供参考):
规划中的下一阶段:
./wechat-insightdoctorsetupexportfeaturesdailylabelscustomer| 命令 | 作用 | 状态 |
|---|---|---|
./wechat-insight doctor | 检查配置状态 | ✅ 可用 |
./wechat-insight setup | 首次提取密钥并生成配置 | ✅ 可用 |
./wechat-insight list | 列出群聊和联系人 | ✅ 可用 |
./wechat-insight export | 导出聊天记录 JSONL | ✅ 可用 |
./wechat-insight features | 生成 feature 层 | ✅ 可用 |
./wechat-insight daily | 生成日报 | ✅ 可用 |
./wechat-insight digest | 一键导出并生成自动化日报 | ✅ 可用 |
./wechat-insight labels | 生成联系人标签模板 | ✅ 可用 |
./wechat-insight customer | 生成客户 / 商业分析 | ✅ 可用 |
./wechat-insight report-data | 汇总展示层统一 JSON 载荷 | ✅ 可用 |
./wechat-insight html | 生成本地可打开的单文件 HTML 报告(内置交互式 React 工作台) | ✅ 可用 |
./wechat-insight emotion | 情绪分析(启发式) | ✅ 可用 |
./wechat-insight mbti | MBTI 推测(启发式) | ✅ 可用 |
./wechat-insight speech | 口癖统计(启发式) | ✅ 可用 |
./wechat-insight social | 社交图谱(启发式) | ✅ 可用 |
每次优先执行:
./wechat-insight doctor
判断逻辑:
./wechat-insight setup
脚本会自动:
wxid 和数据库路径~/.config/wechat-insight.json~/.config/wechat-keys.json注意:
./wechat-insight list
常见命令:
./wechat-insight export
./wechat-insight export --days 7
./wechat-insight export --start 2026-04-01 --end 2026-04-25
./wechat-insight export --contacts "老婆"
./wechat-insight export --chats "AI编辑器技术讨论"
推荐流程:
./wechat-insight export --days 7
./wechat-insight daily
也可以直接指定输入:
./wechat-insight daily --input ~/.wechat-insight/data/messages_*.jsonl
日报当前已包含:
本 skill 不负责定时和推送;OpenClaw 负责调度、推送和失败重试。
推荐让自动化宿主每天执行:
./wechat-insight doctor
./wechat-insight digest --today --stdout
执行规则:
doctor 非 0:提示用户先人工执行 ./wechat-insight setupdigest 返回 0:从 stdout 读取 Markdown,或读取 DIGEST_REPORT_PATH=... 指向的文件digest 仍会返回 0,并生成“暂无可分析消息”的日报setup,因为它需要用户登录微信和 Frida 注入推荐流程:
./wechat-insight export --days 30
./wechat-insight labels --apply-suggestions
./wechat-insight customer --labels ~/.config/wechat-insight-contacts_labels.json
customer 当前已包含:
customer / vendor / unknown./wechat-insight features
生成内容包括:
./wechat-insight labels
./wechat-insight labels --apply-suggestions
./wechat-insight labels --limit 20
标签模板会:
role / notessuggested_rolesuggested_role_reasonreview_priority_score--apply-suggestions 的规则:
unknown 的 role当前必须明确说明:
report-datahtml优先走:
./wechat-insight doctor./wechat-insight setupdailycustomer优先走:
./wechat-insight export --days 7
./wechat-insight daily
优先走:
./wechat-insight export --days 30
./wechat-insight labels --apply-suggestions
./wechat-insight customer --labels ~/.config/wechat-insight-contacts_labels.json
当前应该回答:
./wechat-insight mbtiemotion / speech / social / html 一起看,而不是单独解读~/.config/wechat-insight.json~/.config/wechat-keys.json~/.config/wechat-insight-contacts_labels.json目录:
~/.wechat-insight/data/
主要文件:
messages_YYYYMMDD_YYYYMMDD.jsonlexport_meta.json目录:
~/.wechat-insight/features/
主要文件:
messages_enriched_*.jsonldaily_features.jsonlchat_features.jsonlcontact_features.jsonl目录:
~/.wechat-insight/reports/
主要文件:
daily_*.mdcustomer_report.mdreport_payload_*.jsondashboard_*.html当前导出层已经有这些关键字段:
is_selfdirectionreal_sender_id当前 feature 层已经有这些关键衍生字段:
message_iddate / hour / weekdaychat_typecontent_cleanis_questionis_action_itemis_scheduleis_business_signalis_quote_signalis_support_signalis_negative_signaltopic_tagsemotion_label这意味着:
示例:
{
"contacts": {
"客户A": {
"role": "customer",
"suggested_role": "customer",
"suggested_role_reason": ["quote_signal", "business_signal"],
"notes": "已成交客户",
"review_priority_score": 62,
"total_messages": 12,
"inbound_messages": 8,
"outbound_messages": 4,
"last_message_at": "2026-04-25 20:55:21",
"business_signal_count": 2,
"quote_signal_count": 1,
"support_signal_count": 0,
"negative_signal_count": 0
}
}
}
建议角色:
customervendorfamilyfriendadspamunknown~/.config/当前这个 skill 的真实定位是:
微信本地分析工作台 v1
已经成熟可用的主线:
doctorsetuplistexportfeaturesdailylabelscustomerreport-datahtmlemotionmbtispeechsocial已落地的交付层:
report-data:输出统一展示载荷html:生成本地可打开的单文件 HTML 报告(内置交互式 React 工作台)规划中待落地的交付层: