fact-checkerlisted

# Fact Checker One wrong number in a published piece costs more than a hundred right ones earn. Readers who catch a single error discount everything else, corrections travel slower than the original mistake, and in professional contexts a bad stat can mean legal or reputational damage. This skill exists because the failure mode is silent: wrong claims read exactly like right ones. The only defense is checking each one on purpose. A second reason this skill exists: text drafted from memory (yours or the user's) inherits stale facts. A claim that was true in 2023 can be false today. Treat "I remember this being true" as a hypothesis, never as a verification. ## Workflow ### 1. Extract claims Read the draft and pull out every checkable factual assertion. A claim is checkable if a source could prove it right or wrong. Number each one and keep the exact wording from the draft, because precision matters: "the largest producer" and "one of the largest producers" are different claims with different truth values. Not claims (skip these): opinions, predictions, the user's personal experiences, hedged generalities ("many people prefer..."), and genuinely common knowledge (water boils at 100°C at sea level). When in doubt whether something is common knowledge, treat it as a claim. ### 2. Triage by risk Verify in this order, because these categories fail most often and cost most when wrong: 1. **Numbers**: statistics, prices, percentages, dates, counts 2. **Quotes**: attributed s