A specialized intelligence tool for analyzing RedNote (Xiaohongshu) community signals, sentiment, and reputation through structured multimodal evidence extraction.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install rednote-research
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 rednote-research 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 RedNote Research?
RedNote Research is a sophisticated intelligence skill designed to navigate the nuanced social landscape of Xiaohongshu. It functions as a signal-discovery layer, allowing agents to move beyond surface-level vibes to extract structured, actionable data from one of the most influential community platforms. By integrating this capability with Openclaw Skills, developers can automate the synthesis of local recommendations, brand reputation, and fast-moving policy updates.
The skill operates in two distinct modes: a default public-web mode for broad discovery without authentication, and a login-enhanced browser review mode for deeper access to comments, recent posts, and creator analytics. It excels at multimodal analysis, processing not just text but also OCR fragments from screenshots, subtitles from videos, and audio transcripts to provide a holistic view of community discussion.
RedNote Research Use Cases
- Monitoring brand reputation and community sentiment (口碑) for products or services.
- Researching local recommendations for restaurants, shops, or travel destinations based on real user feedback.
- Tracking and synthesizing fast-moving platform gossip, drama, or news updates.
- Analyzing policy changes and community implementation details for education or registration tasks.
- Summarizing creator account behavior and recent posting trends for influencer marketing analysis.
- Recovering evidence from weak clues like partial screenshots, OCR fragments, or video subtitles.
How RedNote Research Works
- Target Clarification: Identifies canonical names, aliases, hashtags, and the specific time/geographic scope of the research.
- Query Expansion: Utilizes a mixed query family (overview, latest, trending, verification) to build a comprehensive search set.
- Signal Discovery: Executes searches across public-web indexed RedNote pages, official statements, and reputable media reports.
- Access Escalation: If public results are thin, the system offers an optional login-enhanced path for authenticated browser review.
- Multimodal Extraction: Normalizes findings from text, images (OCR), and videos into a structured claim log.
- Credibility Scoring: Applies a 0-5 rubric to weight evidence based on source type and corroboration.
- Report Delivery: Generates a concise decision-oriented report including snapshots, evidence lists, and suggested next steps.
RedNote Research Setup
To integrate this capability into your workflow, ensure your environment is configured to execute the included Python scripts for query expansion and claim normalization.
# Initialize the query builder for a specific research target
python scripts/query_builder.py --target "brand_name" --category "reputation"
# Normalize and summarize gathered evidence items
python scripts/claim_log_tools.py --input raw_findings.json --output report.md
Ensure you have configured the appropriate access modes in references/access-modes.md to toggle between public and authenticated sessions.
RedNote Research Data Schema & Taxonomy
The skill organizes research data using a structured claim-first pattern. Findings are normalized into the following taxonomy:
| Field | Description |
|---|---|
| Claim Type | Category (complaint, praise, official claim, rumor, etc.) |
| Theme | The topic focus (pricing, quality, service, risk) |
| Evidence Snippet | Direct quote or OCR/ASR output from the source |
| Source Class | The reliability tier of the source (official, first-hand, repost) |
| Modality | The media type (text-page, screenshot, video, audio) |
| Credibility Score | Numeric rating from 0-5 based on the verification rubric |
RedNote Research Advanced Features
- Weak Clue Recovery: Uses
scripts/recovery_query_builder.pyto pivot from thin snippets or OCR fragments to original source media. - Multimodal Capture: Supports deep analysis of screenshots, GIFs, and video subtitles to identify evidence invisible to standard text scrapers.
- Dual-Mode Workflow: Seamlessly switches between no-login public searches and explicit user-authorized login-enhanced review.
- Claim Log Normalization: Automated tools to cluster repetitive discussion patterns and resolve conflicting anecdotes.
- Sentiment Scoring Rubric: Built-in logic for weighing independent, recent, and specific evidence more heavily than viral but vague posts within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
Lobster Swarm: One-Person Company Deployment Skill
An advanced AI agent orchestration skill designed to transform single users into a high-output one-person company through multi-agent task delegation and Feishu integration.
Feishu Operations (feishu-ops)
A comprehensive automation skill for managing Feishu/Lark documents and real-time messaging workflows.
YouTube to Feishu Audio Upload
A seamless bridge for extracting audio from YouTube videos and automatically archiving them in Feishu cloud storage for your team.
Patent Writer
A professional AI assistant for drafting high-quality Chinese patent disclosures with integrated academic research and automated document generation.
OpenViking Context Database
A high-efficiency context database that implements layered loading and semantic search to drastically reduce AI agent token consumption.
Moltbook Poster
A powerful automation skill for publishing text and link posts directly to the Moltbook AI Agent community.
Featured*
