Baseline-RAG is a statistical fact-checking skill that extracts claims and validates them against web sources to provide a confidence-scored verification report.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install baseline-rag
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 baseline-rag 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 Baseline-RAG Fact-Checking?
Baseline-RAG is a specialized tool designed for the Openclaw Skills ecosystem to bring objective verification to AI-driven interactions. It automates the process of identifying verifiable claims—including dates, statistics, and causal statements—and subjects them to rigorous analysis against real-time web data.
This skill is essential for developers building agents that require high data integrity. By calculating a heuristic confidence score and establishing baseline uncertainty bounds (CI-Level 1), it allows agents to move beyond simple text generation into the realm of evidence-based reasoning. It provides a structured way to mitigate hallucinations and present users with a transparent view of the factual reliability of any given statement.
Baseline-RAG Fact-Checking Use Cases
- Verifying statistical claims or numerical data within complex user prompts.
- Fact-checking historical or causal statements during automated research workflows.
- Providing source-backed confidence levels for technical or medical reports.
- Mitigating AI hallucinations by automatically triggering search-based verification on definitive claims.
How Baseline-RAG Fact-Checking Works
- The skill extracts specific, verifiable claims from user input, focusing on numbers, statistics, and causal attributions.
- It performs a targeted web search to identify both supporting and rejecting sources for the extracted claims.
- A confidence score is calculated using the ratio of matching sources to total sources found, adjusted for source quality.
- The system calculates a baseline Confidence Interval (CI) to account for potential heuristic variance.
- It generates a structured report that includes the findings, the confidence metric, and a list of cited sources.
Baseline-RAG Fact-Checking Setup
To integrate this skill into your agent environment, follow these steps:
- Install the skill via your agent's package manager:
openclaw install baseline-rag
- Configure your web search API credentials in your environment variables to enable the retrieval of external sources.
- Define the trigger keywords in your agent's system prompt to ensure the skill activates on terms like "verify" or "check claim".
Baseline-RAG Fact-Checking Data Schema & Taxonomy
The skill organizes its output using a structured taxonomy to ensure clarity in automated decision-making:
| Field | Description | Format |
|---|---|---|
| Finding | The specific claim being analyzed | String |
| Score | Statistical confidence percentage (0-100%) | Integer |
| CI (Baseline) | Heuristic uncertainty range (±15%) | Range |
| Sources | List of URLs and source quality indicators | Array |
| Recommendation | Suggested action: ACCEPT, VERIFY, or REJECT | Enum |
Baseline-RAG Fact-Checking Advanced Features
- CI-Level 1 statistical scoring for more nuanced confidence reporting than binary true/false outputs.
- Automated extraction of causal statements to distinguish between correlation and causation.
- Integrated upsell hooks for routing high-stakes claims to CI-95 cross-validation services.
- Source quality assessment to weight authoritative domains more heavily than unverified content.
SKILL.md
Loading
Related Openclaw Skills
TGA Data Analyzer
An automated tool to download, extract, and analyze TGA (Hortor Games) dashboard reports directly from panel URLs.
Velora Chat
A browser automation skill for interacting with Velora AI companions, supporting automated chat and image generation workflows.
OpenClaw Role Configurator
A conversational assistant designed to streamline the initialization and management of AI agent roles within the OpenClaw ecosystem.
Cross-Validate
A high-precision fact-checking skill that elevates claim verification to CI-Level 2 with tighter confidence intervals.
China Construction Commercial Mediation Expert
A comprehensive expert system for navigating Chinese construction mediation using professional legal frameworks and interest-based negotiation strategies.
NanobananaPro Prompt Master
An elite prompt engineering system designed to generate 100% compatible, professional-grade image and video prompts for the NanobananaPro platform.
Featured*
