What problem does it solve? Strategic plans, ROI estimates, and research conclusions often rest on weak evidence, hidden biases, and unquantified uncertainty. This Skill enforces statistical and inferential hygiene so agents catch confirmation bias, survivorship bias, base-rate neglect, and causal confusion before drafting conclusions. ## Core Features & Use Cases - Evidence-Level Tagging: Every load-bearing claim gets tagged from [untested] up to [multi-method], with a hard rule that no commitment rests on weak evidence alone. - 8-Check Reasoning Framework: Named-bias checks, hypothesis framing with falsifiers and pre-registered predictions, numerical literacy (mean vs median, Simpson's paradox), causal reasoning, and uncertainty quantification with P50/P90 and confidence intervals. - Research Addendum: For agent-mediated research, mandates competing-hypotheses tournaments, adversarial dives, and ACH (Analysis of Competing Hypotheses) adjudication to counter subagent confirmation bias. - Use Case: Before finalizing a market-sizing memo that cites internal search-console data as demand signal, the Skill forces you to tag it [existing-rank-only], name the confirmation-bias risk, add a falsifier, and attach a visible inferential-check line. ## Quick Start Apply the inferential-discipline checks to this strategic plan and tag every demand and ROI claim with its evidence level before I approve it.
