\r \r

Content Research Skill\r

Installation

npx clawhub@latest install content-research-mcbai

\r Search the web for trending articles, news, and content sources on any topic. This skill powers the MCB AI content research pipeline — finding, filtering, scoring, and organizing source material for content creation.\r \r

Search Strategy: Brave + Tavily Dual-Engine\r

\r This skill uses TWO search providers in parallel for maximum coverage:\r \r

• Brave Search — via web_search tool (built-in OpenClaw tool)\r
• Tavily — via direct API call using TAVILY_API_KEY from ~/.openclaw/.env\r \r

Tavily API Call\r

\r

POST https://api.tavily.com/search\r
Headers: Content-Type: application/json\r
Body:\r
{\r
  "api_key": "\x3CTAVILY_API_KEY>",\r
  "query": "\x3Cquery>",\r
  "search_depth": "advanced",\r
  "include_answer": false,\r
  "include_raw_content": false,\r
  "max_results": 10,\r
  "topic": "news"   // use "general" for non-news searches\r
}\r
```\r
\r
Run Tavily via `exec` with PowerShell:\r
```powershell\r
$body = @{\r
  api_key = $env:TAVILY_API_KEY\r
  query = "\x3Cquery>"\r
  search_depth = "advanced"\r
  include_answer = $false\r
  include_raw_content = $false\r
  max_results = 10\r
  topic = "news"\r
} | ConvertTo-Json\r
\r
Invoke-RestMethod -Uri "https://api.tavily.com/search" -Method Post -ContentType "application/json" -Body $body\r
```\r
\r
### Fallback Logic\r
\r
- Run Brave (`web_search`) and Tavily in parallel\r
- If Brave fails → use Tavily results only\r
- If Tavily fails → use Brave results only\r
- If both succeed → merge and deduplicate by URL\r
\r
## When to Use\r
\r
- User wants to research a topic before writing content\r
- User needs to find recent articles, news, or data about a subject\r
- User wants to discover trending content sources for LinkedIn/social media\r
- User needs to curate sources for a toplist, POV, case study, or how-to post\r
\r
## Core Workflow\r
\r
### Step 1: Understand the Research Request\r
\r
Extract from the user's message:\r
1. **Topic** — the subject to research (required)\r
2. **Source filter** — where to search (default: all sources)\r
   - `all` — All web sources\r
   - `news` — News publications only\r
   - `linkedin` — LinkedIn posts/articles (append `site:linkedin.com`)\r
   - `youtube` — YouTube videos (append `site:youtube.com`)\r
   - `blogs` — Blog posts and articles (append `blog OR article OR guide`)\r
3. **Freshness** — how recent (default: past month for web, past week for news)\r
4. **Count** — how many results to return (default: 10-15)\r
\r
If the user doesn't specify these, use sensible defaults and mention what you chose.\r
\r
### Step 2: Execute Dual Search (Brave + Tavily)\r
\r
Run BOTH providers. Each provider runs TWO queries when possible.\r
\r
#### Brave Search (web_search tool)\r
\r
**Query 1 — Web:**\r
```\r
Query: {topic} {source_filter_query}\r
count: 10\r
freshness: month\r
```\r
\r
**Query 2 — News:**\r
```\r
Query: {topic} news\r
count: 10\r
freshness: week\r
```\r
\r
#### Tavily Search (exec PowerShell)\r
\r
**Query 1 — General:**\r
```powershell\r
$env:TAVILY_API_KEY = (Get-Content "$env:USERPROFILE\.openclaw\.env" | Select-String "TAVILY_API_KEY" | ForEach-Object { $_ -replace "TAVILY_API_KEY=", "" })\r
\r
$body = @{\r
  api_key = $env:TAVILY_API_KEY.Trim()\r
  query = "{topic}"\r
  search_depth = "advanced"\r
  include_answer = $false\r
  include_raw_content = $false\r
  max_results = 10\r
  topic = "general"\r
} | ConvertTo-Json\r
\r
Invoke-RestMethod -Uri "https://api.tavily.com/search" -Method Post -ContentType "application/json" -Body $body\r
```\r
\r
**Query 2 — News:**\r
```powershell\r
# Same as above but topic = "news"\r
```\r
\r
### Step 3: Merge and Deduplicate\r
\r
1. Collect all results from Brave (web + news) and Tavily (general + news)\r
2. **Deduplicate** by URL — keep one copy per URL, prefer Tavily version (richer content)\r
3. **Sort** by relevance:\r
   - News articles first (most time-sensitive)\r
   - Then by freshness (most recent first)\r
4. **Limit** to requested count (default 15)\r
5. **Label source engine** in metadata: `[Brave]` or `[Tavily]`\r
\r
### Step 4: Process and Organize Results\r
\r
For each result, extract and structure:\r
\r
```\r
Article:\r
  - Title: [article title]\r
  - Source: [publication/website name]\r
  - URL: [full URL]\r
  - Date: [relative date, e.g. "2 hours ago", "3 days ago"]\r
  - Summary: [description/snippet from search]\r
  - Type: [News / Blog / Report / Video / LinkedIn]\r
  - Tag: [auto-detected tag, see Tag Rules below]\r
  - Engine: [Brave / Tavily / Both]\r
```\r
\r
#### Source Name Extraction\r
Clean the hostname to a readable name:\r
- Remove `www.` prefix\r
- Remove `.com`, `.org`, `.net`, `.io`, `.co` suffixes\r
- Map known domains: techcrunch → TechCrunch, crunchbase → Crunchbase, forbes → Forbes, bloomberg → Bloomberg, reuters → Reuters, etc.\r
\r
#### Auto-Tag Rules\r
Scan title + summary and apply the FIRST matching tag:\r
\r
| Tag | Pattern Keywords |\r
|-----|-----------------|\r
| Funding | fund, raise, round, series A-C, seed, valuation, invest, VC, venture |\r
| AI | ai, artificial intelligence, machine learning, LLM, GPT, Claude, OpenAI |\r
| SaaS | saas, software as a service, subscription, ARR, MRR |\r
| Tools | tool, platform, app, software, stack, framework |\r
| Trends | trend, report, survey, data, statistic, forecast, prediction |\r
| Startup | startup, founder, launch, accelerator, incubator, YC |\r
| Growth | growth, marketing, GTM, acquisition, retention, conversion |\r
\r
### Step 5: Present Results\r
\r
Present the organized results in a clear, scannable format:\r
\r
```\r
## Research Results: "{topic}"\r
Found {N} articles from {sources_count} sources\r
Sources: Brave ({brave_count}) + Tavily ({tavily_count}) → merged {total} unique\r
\r
### 📰 News\r
1. **{title}** — {source} ({date}) [{engine}]\r
   {summary}\r
   🏷️ {tag} | 🔗 {url}\r
\r
### 📝 Articles & Blogs\r
2. **{title}** — {source} ({date}) [{engine}]\r
   {summary}\r
   🏷️ {tag} | 🔗 {url}\r
\r
...\r
```\r
\r
Then ask the user which articles they want to use for content creation. If the user wants to proceed to writing, hand off to the `content-writer` skill with the selected articles.\r
\r
## Output Format\r
\r
Always provide results as a numbered list with:\r
- Clear title\r
- Source name and date\r
- Engine label [Brave] or [Tavily]\r
- Brief summary (1-2 lines)\r
- Auto-detected tag\r
- Source URL\r
\r
## Tips for Better Research\r
\r
- For funding/startup topics: search for specific company names + "funding" or "series"\r
- For trend pieces: include year/quarter in the search (e.g., "AI trends Q1 2026")\r
- For competitive analysis: search for specific company + "vs" or "alternative"\r
- For LinkedIn content: recent news performs best (past 1-2 weeks)\r
- Combine multiple source types for richer content\r
- Tavily `search_depth: "advanced"` digs deeper — use for complex topics\r
- If one engine returns fewer results than expected, note it in the summary\r
\r
## Integration with Content Writer\r
\r
After research, the user typically selects articles and moves to writing. Pass the selected articles to the content-writer skill in this format:\r
\r
```json\r
{\r
  "articles": [\r
    {\r
      "title": "Article title",\r
      "source": "Publication name",\r
      "url": "https://...",\r
      "date": "2 days ago",\r
      "summary": "Brief description",\r
      "tag": "AI",\r
      "engine": "Tavily"\r
    }\r
  ]\r
}\r
```\r
\r
See `references/source-filters.md` for detailed source filter configurations.\r
\r
\r