Social Media Fact-Check Skill

A comprehensive methodology for verifying claims in social media posts using AI assistance. This skill helps identify claims, gather evidence, analyze source independence, detect rhetorical fallacies, and provide clear verification reports.

🎯 What This Does

This skill provides a structured framework for fact-checking social media posts (50-500 words) by:

• Identifying the primary claim in a post
• Gathering supporting and disputing evidence from multiple sources
• Analyzing source independence and circular citations (CRITICAL)
• Detecting rhetorical fallacies and manipulation tactics
• Evaluating ideological alignment of sources
• Reporting findings in a clear, actionable format

💡 Why This Matters for AI Development

This project demonstrates several key competencies in AI development and prompt engineering:

Structured Prompt Engineering: Shows how to design multi-step AI workflows with clear checkpoints, validation steps, and error handling. The methodology breaks down a complex task (fact-checking) into discrete, manageable steps that AI can execute consistently.

Multi-Step AI Workflows: Demonstrates orchestrating AI through a complex process—claim identification → triage → evidence gathering → source analysis → fallacy detection → reporting—where each step builds on previous outputs.

Documentation Best Practices: Provides multiple user paths (Claude Skills, copy-paste prompts, manual methodology) showing how to make AI tools accessible to different user segments with varying technical capabilities and subscription levels.

AI Safety & Limitations: Explicitly addresses AI limitations, potential failures, and the need for human oversight—critical for production AI applications.

Real-World Application: Tackles a genuine problem (misinformation) with practical constraints (paywalls, temporal limitations, unfalsifiable claims) rather than theoretical/toy examples.

🚀 How to Use

Option 1: Claude Skills (Recommended - Requires Claude Pro)

Note: Claude Skills require a Claude Pro subscription ($20/month) or Team plan.

1. Download the SKILL.md file from this repository
2. Go to your Claude.ai account
3. Navigate to Settings → Skills → Upload Skill
4. Upload the SKILL.md file
5. Start a conversation and share a social media post or claim to verify

This provides the smoothest experience - Claude will automatically apply the methodology when you share claims.

Option 2: Copy-Paste Prompt (Free - Works with Any AI)

For users without Claude Pro, use the ready-made prompt version:

1. Download PROMPT.txt from this repository
2. Copy the entire contents
3. Paste it at the start of a conversation with any AI assistant that has web search:
   • Free Claude.ai users
   • ChatGPT (with web browsing enabled)
   • Perplexity AI
   • Other AI assistants
4. Then share the post or claim you want fact-checked

See detailed instructions in Free User Guide.

Option 3: Manual Fact-Checking

Use the methodology as a checklist for manual verification:

1. Read through SKILL.md to understand the process
2. Follow each step manually using your own research
3. Great for teaching media literacy or when AI isn't available

Option 4: Other AI Assistants

For ChatGPT Custom GPTs or other integrations:

• Use SKILL.md as a system prompt
• Adapt the format to your platform's requirements
• Ensure web search capability is enabled

Example Usage

You: Can you fact-check this post?

"BREAKING: New study shows coffee consumption reduces cancer risk by 80%. 
Researchers at major university confirm what coffee lovers have known all along!"

AI: [Follows the fact-check methodology to verify the claim]

See the examples folder for detailed fact-check demonstrations.

✨ Key Features

1. Claim Identification

Parses posts to identify the PRIMARY assertion, ensuring focus on what matters most.

2. Triage Assessment

Quickly categorizes claims as:

• Uncontroversial: Widely accepted facts
• Disputed: Claims needing verification
• Misdirected controversy: Sound main claims with contentious sub-claims

3. Evidence Gathering

Uses web search to find:

• Supporting sources
• Disputing sources
• Original/primary sources
• Publication dates and authorship

4. Source Independence Analysis ⚠️ CRITICAL

The most important component:

• Identifies circular citations
• Distinguishes primary vs. secondary sources
• Creates citation chains
• Flags when "multiple sources" are really one source cited repeatedly

5. Ideological Alignment Check

Notes if sources cluster along ideological/political lines, helping identify potential bias.

6. Rhetorical Fallacy Detection 🚨

Aggressively flags:

• Headline-text mismatches
• Appeals to authority
• Cherry-picking
• False equivalences
• Correlation/causation errors
• Emotional manipulation
• And more...

7. Clear Reporting

Delivers findings in a structured markdown report with confidence levels and caveats.

📋 Output Format

Each fact-check produces a report with:

• Main Claim Identified
• Triage (controversial or not)
• Evidence Summary (supporting and disputing)
• Source Independence Analysis (critical for validity)
• Ideological Alignment (potential bias indicators)
• Rhetorical Issues (manipulation tactics)
• Bottom Line (clear verdict with confidence level)

⚠️ Important Limitations

This methodology has inherent limitations:

• Cannot access paywalled content
• Temporal limitations (only finds currently indexed sources)
• Cannot definitively prove/disprove all claims
• Some claims may be unfalsifiable or opinion-based
• AI can make mistakes - users should verify important claims themselves

🤝 Contributing

See CONTRIBUTING.md for guidelines on improving this skill.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🎓 Use Cases

Personal & Social:

• Verifying viral claims before sharing
• Understanding if news stories have independent confirmation
• Evaluating the credibility of health/science claims

Professional Applications:

• Content Moderation: Platforms evaluating user-generated content for misinformation
• Journalism: News organizations fact-checking claims in real-time
• Research: Academic researchers analyzing information spread patterns
• AI Safety: Organizations building safety layers for AI-generated content

Educational:

• Teaching critical thinking and media literacy
• Training students to evaluate sources and identify fallacies
• Demonstrating practical AI prompt engineering techniques

Technical/Development:

• Reference implementation for structured AI workflows
• Example of documentation for multi-audience AI tools
• Template for building verification systems with LLMs

🙏 Acknowledgments

This skill emphasizes source independence analysis because many fact-checking efforts fail to distinguish between multiple independent sources and circular citations of a single source.

📞 Feedback & Support

Found an issue or have suggestions? Please open an issue on GitHub.

---

Disclaimer: This is a methodology for AI-assisted fact-checking. Always verify important claims through multiple independent sources and use critical thinking. AI can make mistakes.