Andy Xie

---
name: gemini-deep-research
description: Trigger Gemini Deep Research via browser and save results to Notion. Use when the user asks to "deep research" a topic, says "gemini deep research", or wants an in-depth research report. Execute ALL steps in the main session (browser tool requires main session access).
---

# Gemini Deep Research → Notion

## Execution Mode

**Run ALL steps in the MAIN SESSION. Do NOT spawn a subagent.**

The browser tool (OpenClaw managed profile) is only available in the main session.
Subagents cannot access the browser, so all browser automation must happen here.

Reply first: "🔬 Deep Research starting for: [topic]. This takes ~25 min. I'll update you when done."

Then execute all phases below sequentially.

---

## Instructions

Complete ALL steps below in the main session.

### Phase 1: Trigger Deep Research

1. `browser action=open profile=openclaw targetUrl="https://gemini.google.com/app"`
2. Snapshot, find the text input, type the research query. **Always prepend "请用中文回答。" to the query** so the research output is in Chinese.
3. Click **"工具" (Tools)** button (has `page_info` icon) → click **"Deep Research"** in the menu
4. Click **Send** to submit the query
5. Wait for research plan to appear (~10s), then click **"Start research"** / **"开始研究"** button
   - If snapshot-click doesn't work, use JS: `(() => { var btn = Array.from(document.querySelectorAll('button')).find(b => /Start research|开始研究/.test(b.textContent.trim())); if (btn) { btn.click(); return 'clicked'; } return 'not found'; })()`
6. Verify research started: button should be disabled, status shows "Researching X websites..." or "正在研究..."
7. Save the conversation URL from the browser

### Phase 2: Wait for Completion

1. Run `exec("sleep 1200")` (20 minutes) + `process(poll, timeout=1200000)`
2. After waking, check status via JS: `(() => { var el = document.querySelectorAll('message-content')[1]; return el ? el.innerText.substring(0, 200) : 'NOT_FOUND'; })()`
3. Look for completion signals: "I've completed your research" or "已完成"
4. If still running, sleep another 600s and check again (max 2 retries)
5. If failed/stuck after retries, announce the failure and exit

### Phase 3: Extract Report

1. Count message-content elements: `document.querySelectorAll('message-content').length`
2. The research report is in the LAST `message-content` element (usually index 2)
3. Get total length: `document.querySelectorAll('message-content')[2]?.innerText?.length`
4. Extract in 8000-char chunks using substring: `document.querySelectorAll('message-content')[N]?.innerText?.substring(START, END)`
5. Concatenate all chunks into the full report text
6. Save to a temp file: write full report to `/tmp/deep_research_<timestamp>.md`

### Phase 4: Export to Notion

**Parent page ID:** `31a4cfb5-c92b-809f-9d8a-dd451718a017` (Deep Research Database)

1. Read the Notion API key: `cat ~/.config/notion/api_key`
2. Parse the report into Notion blocks:
   - Lines starting with `#` → heading_2/heading_3 blocks
   - Bullet points → bulleted_list_item blocks
   - Regular text → paragraph blocks
   - Add a callout at top: "🔬 Generated by Gemini Deep Research on YYYY-MM-DD"
   - Split rich_text at 2000 chars
3. Create the page via Notion API:
   ```bash
   curl -s -X POST "https://api.notion.com/v1/pages" \
     -H "Authorization: Bearer $NOTION_KEY" \
     -H "Notion-Version: 2025-09-03" \
     -H "Content-Type: application/json" \
     -d '{"parent":{"page_id":"31a4cfb5-c92b-809f-9d8a-dd451718a017"},"icon":{"type":"emoji","emoji":"🔬"},"properties":{"title":{"title":[{"text":{"content":"TOPIC"}}]}},"children":[BLOCKS]}'
   ```
4. If >100 blocks, append remaining via PATCH to `/v1/blocks/{page_id}/children`
5. Rate limit: wait 0.5s between batch requests

### Phase 5: Announce

Report back with:
- Research topic
- Brief summary (2-3 key findings)
- Notion page URL: `https://www.notion.so/<page_id_without_dashes>`

## Notes

- Always use `profile="openclaw"` for browser
- Deep Research is under **"工具" (Tools) menu**, NOT the model selector
- If Gemini needs login, announce failure — user must log in manually
- The full pipeline should complete in ~25-30 min total

FILE:README.md
# Gemini Deep Research Skill for OpenClaw

An OpenClaw skill that triggers **Google Gemini Deep Research** via browser automation, extracts the full research report, and exports it to **Notion**.

## Features

- 🔬 Triggers Gemini Deep Research via OpenClaw's managed browser
- ⏳ Automatically waits for research completion (~20 min)
- 📄 Extracts the full report text
- 📝 Exports to Notion with proper formatting (headings, lists, paragraphs)
- 🤖 Runs as a subagent — non-blocking for main session

## Requirements

- **OpenClaw** with browser support (`profile="openclaw"`)
- **Gemini** account logged in via the OpenClaw browser profile
- **Notion API key** stored at `~/.config/notion/api_key`
- A **Notion parent page** to export reports to

## Installation

```bash
# Via ClawHub
clawhub install gemini-deep-research

# Or manually
git clone https://github.com/PalmPalm7/openclaw-gemini-deep-research.git ~/.openclaw/skills/gemini-deep-research
```

## Configuration

Before using, update the Notion parent page ID in `SKILL.md` to your own:

```
Parent page ID: 31a4cfb5-c92b-809f-9d8a-dd451718a017
```

Replace with your Notion page ID where research reports should be created.

## Usage

Just tell your OpenClaw agent:

> "Use Gemini deep research to research [topic]"

The skill will:
1. Open Gemini in the browser
2. Submit the query with Deep Research mode
3. Wait for completion
4. Extract the full report
5. Create a formatted Notion page
6. Report back with key findings and Notion link

## License

MIT

---
name: xiaohongshu-deep-research
description: Deep research on Xiaohongshu (小红书) topics. Use when user wants to research a topic, analyze trends, gather insights from top posts, or produce a summary report. Crawls posts via search, extracts high-engagement content, generates analysis with post links. Requires xiaohongshu-mcp service running.
---

# xiaohongshu-deep-research

小红书话题研究，自动爬取 + 分析 + 输出报告。

## 前置条件

- xiaohongshu-mcp 服务运行中 (`http://localhost:18060/mcp`)
- 已登录小红书账号

## 存储位置

```
~/xiaohongshu-research/{keyword}_{YYYYMMDD_HHmm}/
├── raw/posts.json        # 原始数据
├── analysis/summary.md   # 完整报告
└── metadata.json         # 研究配置
```

## 研究流程

1. **关键词扩展** — 主词 + 3-4 个相关词
2. **批量搜索** — `POST /api/v1/feeds/search`
3. **去重排序** — 按点赞数取 Top 20-50
4. **生成报告** — 含链接、可转发摘要

## 帖子链接格式

```
https://www.xiaohongshu.com/explore/{note_id}
```

## 报告结构

```markdown
# {主题} 小红书研究

---

## 📱 速览（可直接转发）

{2-3段自然语言总结，像朋友分享一样写}

值得看的几篇：
• {标题} — {链接}
• {标题} — {链接}

---

## 数据概览

采集 {N} 篇帖子，共 {likes} 赞 / {collects} 收藏

## Top 10

1. **{标题}** @{作者}
   {likes}👍 {collects}📁 | [查看]({link})

## 发现

{自然语言描述趋势和洞察}

## 数据

原始数据: `{path}`
```

## 写作风格

- 速览部分：像给朋友发消息，口语化，有观点
- 正文部分：简洁，重点突出，避免套话
- 不要用"本报告"、"经过分析"这类官腔

## 限制

⚠️ API 限制：帖子详情和评论无法获取，报告基于搜索结果元数据生成。

## 详细工作流

见 [references/workflow.md](references/workflow.md)

## Credits

基于 [xpzouying/xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp)

FILE:references/workflow.md
# 详细工作流

## 初始化

```bash
mkdir -p ~/xiaohongshu-research/{keyword}_{YYYYMMDD_HHmm}/{raw,analysis}
```

## 关键词扩展

```
输入: "露营"
→ 露营、露营装备推荐、露营好物、新手露营攻略
```

## 搜索 API

```bash
curl -s -X POST "http://localhost:18060/api/v1/feeds/search" \
  -H "Content-Type: application/json" \
  -d '{"keyword": "露营"}'
```

返回字段：
- `id` — 帖子 ID（用于生成链接）
- `noteCard.displayTitle` — 标题
- `noteCard.user.nickname` — 作者
- `noteCard.interactInfo.*` — 点赞/收藏/评论数

## 帖子链接

```
https://www.xiaohongshu.com/explore/{id}
```

## 数据处理

```bash
# 合并多个搜索结果，去重，按点赞排序
jq -s '
  [.[].data.feeds[] | select(.modelType == "note")] | 
  unique_by(.id) |
  sort_by(-(.noteCard.interactInfo.likedCount | tonumber))
' search_*.json > posts.json
```

## 报告模板

```markdown
# {主题} 小红书研究

> {日期} · 关键词: {keywords} · {N} 篇帖子

---

## 📱 速览

{用 2-3 段话总结发现，像给朋友发消息一样写。有观点，有态度，不要官腔。}

**值得看的几篇：**
• [{标题1}](https://www.xiaohongshu.com/explore/{id1})
• [{标题2}](https://www.xiaohongshu.com/explore/{id2})
• [{标题3}](https://www.xiaohongshu.com/explore/{id3})

---

## 数据

{N} 篇帖子 · {total_likes} 赞 · {total_collects} 收藏

## Top 10

1. **{标题}** @{作者}
   {likes}👍 {collects}📁 · [看看](https://www.xiaohongshu.com/explore/{id})

2. ...

## 发现

{用自然语言写 3-5 个洞察，不要用 bullet points 堆砌}

## 建议

{如果用户要去/要买/要做什么，给出具体建议}

---

数据: `~/xiaohongshu-research/{keyword}_{date}/raw/posts.json`
```

## 写作指南

### 速览部分（可直接转发）

写法示例：

❌ 不要这样写：
> 经过对20篇小红书帖子的分析，我们发现波士顿美食话题具有较高的用户关注度。头部内容主要集中在龙虾、中餐等品类。

✅ 这样写：
> 刷了一圈小红书，波士顿吃的还挺多的。龙虾当然是必吃，但意外发现羊杂汤特别火——可能华人都馋这口。有个博主吃了100家店做了个系列，挺靠谱的可以参考。

### 正文部分

- 直接说结论，不要铺垫
- 数字要有意义（"1489 赞"比"高互动"具体）
- 链接要能点

## 限制

- 帖子正文内容无法获取
- 评论无法获取
- 数据来自搜索结果元数据

---
name: xiaohongshu-mcp
description: Operate Xiaohongshu (小红书/RED) via local MCP service. Use when user wants to search notes, publish content (image/video), interact with posts (like/comment/favorite), or check account status on Xiaohongshu. Requires MCP service running at localhost:18060.
---

# xiaohongshu-mcp

小红书 MCP 服务操作。

## 快速开始

1. **检查服务状态**
   ```bash
   curl -s -X POST http://localhost:18060/mcp \
     -H "Content-Type: application/json" \
     -d '{"jsonrpc":"2.0","method":"initialize","params":{},"id":1}'
   ```

2. **服务未运行时启动**
   ```bash
   cd /Users/handi7/Documents/agentic-coding-projects/projects/xiaohongshu-mcp
   nohup ./bin/xiaohongshu-mcp-darwin-arm64 > mcp.log 2>&1 &
   ```

## 可用工具

| 工具 | 说明 |
|------|------|
| `check_login_status` | 检查登录状态 |
| `get_login_qrcode` | 获取登录二维码 |
| `list_feeds` | 首页推荐 |
| `search_feeds` | 搜索内容 |
| `get_feed_detail` | 笔记详情 |
| `user_profile` | 用户主页 |
| `publish_content` | 发布图文 |
| `publish_with_video` | 发布视频 |
| `post_comment_to_feed` | 发表评论 |
| `reply_comment_in_feed` | 回复评论 |
| `like_feed` | 点赞 |
| `favorite_feed` | 收藏 |
| `delete_cookies` | 重置登录 |

## 内容限制

- 标题 ≤20 字
- 正文 ≤1000 字
- 每日 ≤50 篇
- 图文流量优于视频

## References

- **首次部署**: [references/deploy.md](references/deploy.md) — 下载、登录、启动服务
- **使用指南**: [references/usage.md](references/usage.md) — 工具详解、工作流、运营知识

## Credits

Based on [xpzouying/xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp) by [@xpzouying](https://github.com/xpzouying)

FILE:references/deploy.md
# 首次部署 xiaohongshu-mcp

> Based on [xpzouying/xiaohongshu-mcp](https://github.com/xpzouying/xiaohongshu-mcp)

## 环境要求

- macOS / Linux / Windows
- 首次运行自动下载 Chromium (~150MB)

## 部署步骤

### 1. 下载预编译二进制

从 [GitHub Releases](https://github.com/xpzouying/xiaohongshu-mcp/releases) 下载对应平台版本：

```bash
# macOS ARM64
gh release download --repo xpzouying/xiaohongshu-mcp \
  --pattern "xiaohongshu-mcp-darwin-arm64.tar.gz" --dir /tmp
mkdir -p ~/xiaohongshu-mcp/bin
tar -xzf /tmp/xiaohongshu-mcp-darwin-arm64.tar.gz -C ~/xiaohongshu-mcp/bin/

# macOS Intel: xiaohongshu-mcp-darwin-amd64.tar.gz
# Linux x64: xiaohongshu-mcp-linux-amd64.tar.gz
# Windows x64: xiaohongshu-mcp-windows-amd64.zip
```

解压后包含两个可执行文件：
- `xiaohongshu-login-*` — 登录工具
- `xiaohongshu-mcp-*` — MCP 服务

### 2. 登录小红书

```bash
cd ~/xiaohongshu-mcp
./bin/xiaohongshu-login-darwin-arm64
```

- 首次运行自动下载 Chromium 到 `~/.cache/rod/browser/`
- 浏览器弹出后扫码或手机号登录
- 登录成功后 `cookies.json` 保存在当前目录
- 后续无需重复登录（除非 cookies 过期）

### 3. 启动 MCP 服务 (nohup 后台)

```bash
cd ~/xiaohongshu-mcp
nohup ./bin/xiaohongshu-mcp-darwin-arm64 > mcp.log 2>&1 &
echo $! > mcp.pid
```

服务监听: `http://localhost:18060/mcp`

验证：
```bash
curl -s -X POST http://localhost:18060/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"initialize","params":{},"id":1}'
```

### 4. 配置 MCP 客户端

**Claude Code:**
```bash
claude mcp add --transport http xiaohongshu-mcp http://localhost:18060/mcp
```

**其他 MCP 客户端配置:**
```json
{
  "mcpServers": {
    "xiaohongshu-mcp": {
      "url": "http://localhost:18060/mcp",
      "transport": "http"
    }
  }
}
```

## 本机实际部署路径

项目目录: `/Users/handi7/Documents/agentic-coding-projects/projects/xiaohongshu-mcp`

启动服务:
```bash
cd /Users/handi7/Documents/agentic-coding-projects/projects/xiaohongshu-mcp
nohup ./bin/xiaohongshu-mcp-darwin-arm64 > mcp.log 2>&1 &
```

## 日常运维

| 操作 | 命令 |
|------|------|
| 启动服务 | `nohup ./bin/xiaohongshu-mcp-darwin-arm64 > mcp.log 2>&1 &` |
| 查看日志 | `tail -f mcp.log` |
| 停止服务 | `kill $(lsof -ti:18060)` 或 `kill $(cat mcp.pid)` |
| 重新登录 | `./bin/xiaohongshu-login-darwin-arm64` |
| 查看 PID | `lsof -ti:18060` |

## 注意事项

1. **同账号单端登录**：小红书不允许同一账号在多个 web 端登录，保持 MCP 独占访问
2. **Cookies 过期**：需重新运行登录工具
3. **实名认证**：新号可能触发实名认证提醒，建议先实名
4. **发帖限制**：每天 50 篇上限

FILE:references/usage.md
# xiaohongshu-mcp 使用指南

## 前置条件

- MCP 服务运行在 `http://localhost:18060/mcp`
- 已完成小红书登录（cookies.json 存在）

## 可用工具 (13个)

### 账号管理

| 工具 | 说明 |
|------|------|
| `check_login_status` | 检查登录状态 |
| `get_login_qrcode` | 获取登录二维码（Base64 图片 + 超时时间） |
| `delete_cookies` | 删除 cookies 重置登录状态 |

### 内容浏览

| 工具 | 说明 |
|------|------|
| `list_feeds` | 获取首页推荐 Feeds |
| `search_feeds` | 搜索小红书内容 |
| `get_feed_detail` | 获取笔记详情（内容、图片、互动数据、评论） |
| `user_profile` | 获取用户主页（基本信息、关注/粉丝、笔记列表） |

### 内容发布

| 工具 | 说明 |
|------|------|
| `publish_content` | 发布图文内容 |
| `publish_with_video` | 发布视频内容（仅支持本地文件） |

### 互动操作

| 工具 | 说明 |
|------|------|
| `post_comment_to_feed` | 发表评论 |
| `reply_comment_in_feed` | 回复评论 |
| `like_feed` | 点赞/取消点赞 |
| `favorite_feed` | 收藏/取消收藏 |

## 常用工作流

### 1. 搜索并获取笔记详情

```
1. search_feeds(keyword="xxx") → 获取搜索结果列表
2. 从结果中提取 note_id 和 xsec_token
3. get_feed_detail(note_id, xsec_token) → 获取完整详情
```

### 2. 发布图文笔记

```
publish_content(
  title="标题（≤20字）",
  content="正文（≤1000字）",
  images=["/path/to/img1.jpg", "/path/to/img2.jpg"]
)
```

图片支持：
- 本地绝对路径（推荐）
- HTTP/HTTPS URL

### 3. 发布视频笔记

```
publish_with_video(
  title="标题（≤20字）",
  content="正文（≤1000字）",
  video="/path/to/video.mp4"
)
```

仅支持本地视频文件，建议 <1GB。

### 4. 批量互动

```
1. list_feeds() 或 search_feeds() 获取笔记列表
2. 遍历结果调用 like_feed / favorite_feed / post_comment_to_feed
```

## 重要参数说明

### xsec_token

- 多数操作需要 `xsec_token` 参数
- 从 `list_feeds` 或 `search_feeds` 返回结果中获取
- 与 `note_id` 配对使用

### note_id

- 笔记唯一标识
- 从搜索/推荐结果中获取

## 运营知识

- **标题**：≤20字（硬限制）
- **正文**：≤1000字（硬限制）
- **每日发帖**：≤50篇
- **图文优于视频**：推荐角度图文流量更好
- **Tags**：添加合适标签增加曝光
- **违禁词**：使用前检查内容是否含违禁词
- **禁止行为**：引流、纯搬运会被打击

## 错误排查

| 问题 | 解决 |
|------|------|
| 登录失效 | 运行 `xiaohongshu-login-*` 重新登录 |
| 无法发布 | 检查标题/正文长度限制 |
| 找不到笔记 | 确认 note_id 和 xsec_token 配对正确 |
| 服务不响应 | `tail -f mcp.log` 查看日志 |