A red-team agent is designed to fail your article — its objective is to find every error, inconsistency, misleading claim, and unverified assertion in a draft before it's published. Unlike a standard fact-checker (which attempts to verify claims), a red-team agent attempts to falsify them, using a different set of sources and reasoning strategies than the original fact-checking process.
Red Team Agent Design
Objective: "Your goal is to identify every factual claim in this article that might be wrong, misleading, or missing important context. You are not trying to agree with the article — you are trying to find errors." Methodology: For each specific claim, the agent searches for counter-evidence (sources that contradict the claim), alternative interpretations, and missing context that would change the meaning of the claim. Output: A list of potential problems with severity ratings (critical, moderate, minor), supporting evidence for each concern, and recommended editorial actions. Human review: A human editor reviews all critical and moderate concerns before publication. Minor concerns are addressed at their discretion.