Red-team-before-reliance pattern

the epistemic practice of deliberately attempting to break one's own mental model before depending on it, which surfaces weaknesses that confirmation-biased testing misses because adversarial testing operates against rather than with cognitive defaults