Cross-Validated Search for Openclaw

A source-backed verification skill for OpenClaw that performs live web searches, deep page reading, and evidence-aware claim checking.

wd041216-bit
v16.0.0
Mar 24, 2026
0
829
0

Install & Download

1. ClawHub CLI

The fastest way to install a skill directly from the registry.

npx clawhub@latest install cross-validated-search

2. Manual Installation

Copy the skill folder to one of these locations

Global
~/.openclaw/skills/
Workspace
<project>/skills/

Priority: Workspace > Local > Bundled

3. Prompt Installation

Copy this prompt to OpenClaw to install it automatically.

Help me install cross-validated-search 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 Cross-Validated Search?

Cross-Validated Search is a high-performance Openclaw Skills extension designed to provide AI agents with a rigorous verification workflow. It bridges the gap between raw web data and factual reliability by offering tools that not only search the web but also ingest full page content to classify claims based on supporting or conflicting evidence.

By integrating this skill, developers can move beyond simple LLM hallucinations. The tool enforces a structured approach to information gathering, ensuring that every claim made by the agent is backed by specific, citation-ready source URLs. This makes it an essential component for research-heavy AI applications and automated fact-checking pipelines.

Cross-Validated Search Use Cases

  • Verifying factual claims using real-time search results to prevent AI hallucinations.
  • Generating comprehensive evidence reports with explicit support and conflict scores for legal or academic research.
  • Deep-diving into specific web pages when search snippets provide insufficient context for complex queries.
  • Automating the monitoring of recent news or technical releases using time-limited search parameters.

How Cross-Validated Search Works

  1. Execute a live search using the search-web command to gather relevant snippets and URLs from the web.
  2. Utilize browse-page to extract full-text content from the most promising sources for deeper context.
  3. Run verify-claim to process the gathered text and assign heuristic support or conflict classifications to a specific statement.
  4. Generate an evidence-report which synthesizes the findings into a citation-ready summary including next steps for the agent.
  5. Review the explicit verdict and citation URLs to formulate a final, evidence-backed response.

Cross-Validated Search Setup

To install the Cross-Validated Search skill, run the following command in your terminal:

pip install cross-validated-search

For an enhanced experience, it is recommended to configure a self-hosted SearXNG instance by setting the environment variable:

export CROSS_VALIDATED_SEARCH_SEARXNG_URL='your_searxng_url'

Cross-Validated Search Data Schema & Taxonomy

The skill organizes its output into structured formats, often available in JSON, to facilitate multi-agent interoperability within Openclaw Skills.

Feature Data Handled Output Key
Search Results Snippets, URLs, Metadata results
Verification Support/Conflict Scores verdict
Deep Analysis Page-aware content flags page_aware
Evidence Citation-ready source links citations

Cross-Validated Search Advanced Features

  • Multi-provider support defaulting to DuckDuckGo (ddgs) with a recommended upgrade path to self-hosted SearXNG.
  • Deep verification mode (--deep) for page-aware analysis that goes beyond top-level search snippets.
  • Support and Conflict scoring system to highlight discrepancies in web data rather than just aggregating top results.
  • JSON output flags for easy integration into automated DevOps or data analytics pipelines.

SKILL.md


Loading

Related Openclaw Skills

METADATA

Github Stars: 0
forks: 0

Featured*