What problem does it solve? Single-model feedback tends toward sycophancy, agreeing with how you framed an idea instead of challenging it. This Skill replaces one agreeable reviewer with a council of adversarial personas run as separate agents with clean context, so business ideas, build plans, launches, and pricing calls get genuinely stress-tested before you commit money or effort. ## Core Features & Use Cases - Adversarial persona council: Fans out five personas (contrarian, expansionist, first-principles, deep-researcher, buyer) as parallel agents that never see each other's output, defeating groupthink and sycophancy. - Independent judge synthesis: A separate judge agent that authored none of the analysis returns one verdict — GREEN, RESHAPE, or KILL — with scores, the biggest risk, and the biggest upside. - Mandatory 48-hour test: Every verdict includes a single falsifiable next action with a numeric threshold, so decisions convert into a runnable experiment before any code or spend. - Use Case: Before launching a $19/month subscription tool for freelancers, run the council; the buyer persona reveals churn risk, the researcher cites competitor churn data, and the judge returns RESHAPE — sell it as a $500 one-off audit to agencies instead. ## Quick Start Ask the agent to roast your business idea or plan, for example: "Roast this plan to launch a paid newsletter for indie hackers before I spend anything on it."
