A deep web research skill that queries 11 Chinese and international search sources simultaneously with evidence synthesis, contradiction detection, and local Agentic RAG fallback.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agentic-rag-cn
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agentic-rag-cn using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Deep Research?
The deep-research skill is an advanced agentic RAG search implementation designed to expand your AI capabilities. Built as one of the highly specialized Openclaw Skills, this tool queries 11 distinct international and Chinese platforms (including WeChat, Zhihu, Bilibili, Baidu, Bing, and GitHub) simultaneously. It processes findings through a local Agentic RAG-CN pipeline, allowing your AI assistant to perform exhaustive investigative research, resolve information contradictions, and synthesize highly structured answers with complete source tracing.
By integrating this tool, developers can bypass shallow search limitations. When the local RAG engine is online, it serves as the ultimate investigative engine, especially for regional or technical queries on platforms like V2EX or Sogou. When offline, it gracefully falls back to basic web searches, ensuring uninterrupted operation.
Deep Research Use Cases
- Conducting deep technical research across platforms like GitHub, V2EX, and search engines.
- Querying regional Chinese platforms (WeChat, Zhihu, Bilibili, Baidu) for localized insights.
- Resolving complex information conflicts by detecting contradictions across multiple sources.
- Generating structured evidence tables and pipeline traces for academic or market research.
- Optimizing search-heavy workflows using Openclaw Skills with automated fallback configurations.
How Deep Research Works
- Health Verification: The AI agent checks if the local Agentic RAG-CN service is active via its health endpoint.
- Platform Routing: If healthy, the agent routes the search query to the deep-research endpoint. If down, it falls back to the standard web search tool.
- Multi-Source Fetching: The API queries 11 Chinese and international sources simultaneously.
- Execution Pipeline: The engine runs through a 16-step trace pipeline to process, analyze, and synthesize findings.
- Contradiction Detection & Synthesis: The API analyzes search results for conflicting claims and rates confidence.
- Structured Output: Returns the structured response, complete with evidence tables, trace logs, and direct source links.
Deep Research Setup
To integrate this skill into your ecosystem of Openclaw Skills, verify that your local Agentic RAG-CN service is running and accessible.
Check the local service health status:
curl -s http://localhost:18888/health
If the service returns a healthy status, you can issue queries directly via the /api/analyze POST endpoint using curl:
curl -s -X POST http://localhost:18888/api/analyze -H "Content-Type: application/json" -d '{"question": "YOUR_QUERY_HERE"}'
Configure your agent environment variables to point to this endpoint, ensuring fallback routes to the native web_search utility are declared in case of service interruptions.
Deep Research Data Schema & Taxonomy
The deep-research skill structure organizes and delivers data from the Agentic RAG API payload. The payload metadata taxonomy is detailed below:
| Property | Type | Description |
|---|---|---|
answer |
String | Synthesized markdown answer featuring structured evidence tables |
sources |
Array | List of URLs and resources retrieved during the search operation |
confidence |
Float | Confidence score ranging from 0 to 1 based on search consensus |
contradictions |
Array | Detected discrepancies or opposing facts found across the 11 sources |
trace |
Array | Detailed execution log detailing the 16 stages of the RAG pipeline |
Deep Research Advanced Features
- Simultaneous 11-Source Querying: Queries specialized Chinese networks (WeChat, Zhihu, Bilibili, V2EX) alongside global engines (Bing, Yandex, GitHub).
- Graceful Fallback Automation: Automatically defaults to standard
web_searchif the local Agentic RAG-CN API endpoint becomes unreachable. - Automated Contradiction Resolution: Identifies conflicting claims across indexers and flags contradictions explicitly for the user.
- 16-Step Execution Tracing: Provides absolute transparency into the search synthesis process for advanced debugging in Openclaw Skills setups.
SKILL.md
Loading
Related Openclaw Skills
Personal Productivity Consultant
An AI-powered personal productivity and energy management consultant designed to diagnose efficiency bottlenecks and build sustainable habits.
agent-browser
A powerful browser automation CLI tool that allows AI agents and developers to interact with web pages, extract structured data, and manage sessions using a simple ref-based system.
maverick-gog
An internal setup hook that non-interactively provisions the gog CLI with Google OAuth credentials from an external orchestrator over RPC.
iOS FlyVerify Integration Skill
An automated engineering skill that smoothly integrates MobTech FlyVerify into iOS projects with minimal structural modifications and full privacy compliance.
Prototype to PRD Generator
An AI-powered document generation skill that ingests prototype descriptions and screenshots to generate scope-locked, professionally formatted Product Requirement Documents (PRDs) with built-in consistency checks.
Relay GitCode CLI
An advanced command-line interface skill for automated control of GitCode API v5 platforms, managing code pipelines, repositories, and migration workflows.
Featured*
