blog-factcheck

Verify statistics and claims in blog posts against cited source URLs.

1|Updated Dec 24, 2021
One-click install
npx skills add https://github.com/arthur-fontaine/arthurfontaine.fr --skill blog-factcheck-arthur-fontaine
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: blog-factcheck
Source: https://github.com/arthur-fontaine/arthurfontaine.fr/tree/main/.agents/skills/blog-factcheck
Command: npx skills add https://github.com/arthur-fontaine/arthurfontaine.fr --skill blog-factcheck-arthur-fontaine

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Blog posts often contain statistics and claims that are misquoted, outdated, or unsupported by their cited sources, and manually checking every link is slow and error-prone. This Skill automates claim extraction and source verification so you can publish with confidence. ## Core Features & Use Cases - Claim Extraction: Identifies all load-bearing claims including statistics, product, policy, ranking, comparative, and legal statements, with or without citations. - Source Verification: Validates and fetches cited URLs safely, classifies sources into quality tiers (T1-T5), detects echo clusters, and scores each claim from VERIFIED (1.0) to NOT FOUND (0.0). - Verification Report: Produces a structured report with per-claim scores, flags uncited claims as UNVERIFIED with suggested search queries, and lists recommended actions. - Use Case: Before publishing a marketing blog post citing "73% of marketers use AI (HubSpot, 2025)", run this Skill to confirm the statistic actually appears on the cited page and flag any claims lacking sources. ## Quick Start Ask the AI to fact check your blog post file, for example by saying: fact check the claims and sources in my draft post at content/blog/q3-trends.md.

Frequently Asked Questions about blog-factcheck

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I fact check statistics in a blog post?▼

Run the blog-factcheck skill on your post file and it will extract every load-bearing claim, fetch each cited URL, and score whether the claimed data appears on the source page. Uncited claims are flagged as UNVERIFIED with suggested search queries.

How are source verification scores calculated?▼

Each claim receives a confidence score: 1.0 for an exact match on the cited page, 0.7-0.9 for paraphrased matches, 0.3-0.6 when the source covers the topic but not the specific statistic, and 0.0 when the data is not found or the source is rejected.

Can it verify claims that have no source URL?▼

Claims without a URL cannot be verified and are marked UNVERIFIED. The skill suggests a search query to find a source, and if the claim names a specific organization, it suggests checking that organization's domain.

Does it work with paywalled or JavaScript-rendered pages?▼

No. WebFetch cannot access paywalled content, which scores as WEAK (0.5) with a paywall note, and JavaScript-rendered pages may return minimal content. PDF sources may not extract reliably and are flagged for manual verification.

What happens when a cited source is low quality?▼

Sources are classified into tiers, and Tier 4 (generic SEO blogs, affiliate roundups) and Tier 5 (content mills, AI spam) sources are rejected even if the wording matches. Multiple pages repeating one upstream study are treated as a single echo cluster.