fact checker

Accuracy problems cost you relationships, while style problems only cost you a reader. This skill serves as the final gate for any professional communication, ensuring that every number, name, and citation is verified before it reaches a client or executive. It solves the two biggest risks in AI-assisted writing: confident fabrication by the model and stale facts that have changed since your last update.

The skill meticulously extracts every checkable claim, including numbers, job titles, organization names, quotes, and citations. It even identifies "implied claims" like superlatives and duration statements that writers often overlook. Each claim is categorized by its source and subjected to rigorous verification through web search or internal cross-referencing.

What it does

• Extract claims by identifying every number, name, organization, and quote in your draft.
• Audit superlatives to verify claims like "first," "only," or "fastest" instead of treating them as adjectives.
• Classify sources to separate user-provided data from model-generated text, applying harsher scrutiny to the latter.
• Verify citations to ensure links exist and actually support the claims being made.
• Propose rewrites for any claim found to be incorrect, stale, or unverifiable.

How it works

1. Trigger the audit by providing a draft and using a command like "fact-check this" or "check my claims."
2. Review the extraction table where the agent lists every quoted claim alongside its type and current status.
3. Address the red flags by reviewing the specific replacement lines for every "wrong" or "unverifiable" status.
4. Finalize the draft based on the agent's safety verdict and its identification of the single riskiest claim.

Frameworks & tools

This skill works within any Markdown-compatible environment and uses web search tools for real-time verification of public data.

Why this beats prompting it yourself

Generic prompts often miss "implied claims" and superlatives, which are the most common sources of professional embarrassment. This skill enforces a strict classification system that prevents the model from upgrading "plausible" claims to "verified" status without an independent source.

Use cases

• Executive Memos to verify internal metrics and funding figures before a board meeting.
• Sales Outreach to ensure prospect titles and company details are accurate and current.
• Public Relations to audit quotes and superlatives in press releases or launch docs.
• Thought Leadership to verify citations and dates in LinkedIn posts or op-eds.

Known limitations

Cannot verify internal metrics or private data without user-supplied context. Relies on the availability of public sources for external verification.