Fact Verification via Gemini API

research

fact-checker is a Claude Code skill that automatically extracts factual claims — numbers, dates, proper nouns, statistics, causal statements, and citations — from Markdown or plain-text files and verifies each one using Gemini API grounding search. Claims are classified into five categories and assigned one of four verdicts: Verified, Needs Review, Discrepancy, or Unverifiable, with source URLs attached to every finding. Run via `fact_checker.py --input article.md` with optional `--category` and `--depth` flags (quick or thorough); a `--test` mode works without an API key. The output is a timestamped Markdown report containing a summary table and per-claim breakdowns with confidence scores. Ideal for validating blog posts, news articles, and press releases before publication.