Reason Council · cdoriga bundle Run any claim or AI output through a structured epistemic protocol to evaluate hallucination and confabulation risk. Architecturally inspired by the llm-council, adapted for evaluating claims rather than decisions. This is structured skepticism scaffolding, not a hallucination detector: the auditing model is the same model that produced the claim, so verdicts are calibration signals, not determinations. A simpler two-prompt alternative is provided as the default starting point; reserve the full protocol for high-stakes claims. MANDATORY TRIGGERS: 'reason-council this', 'is this hallucination', 'is this grounded', 'fact-check this', 'epistemic audit', 'verify this claim', 'sanity-check this', 'is this real or AI'. STRONG TRIGGERS: 'am I making this up', 'is this actually true', 'how confident should I be', 'does this hold up', 'is the AI bullshitting me', 'is this defensible'. DO NOT use for decisions between options (use the llm-council instead).
