Alisson-P
UserAn agent skill that disagrees with you, on purpose, with evidence. Risk first, pre mortem always, verdict changes only on new evidence.
Categories
Indexed Skills (2)
ysat
Challenges a decision before it is made: reads the real context (code, infrastructure, email, Teams, shared files, calendar) and returns risks, blind spots and failure modes, with severity, a pre mortem, an explicit verdict, mitigations and the cheapest test that settles the question. Risk focused by design: it never opens with praise and never changes a verdict without new evidence. Use when the user says "challenge this decision", "play devil's advocate", "what could go wrong", "what are the risks of this", "discorda dessa decisao", "por que isso pode dar errado", "estou pensando em fazer X". Do NOT use for a balanced pros and cons write up, for a decision already executed, for line by line code review, or to evaluate people.
ysat-jev
Challenges a decision with a TYPED, reproducible verdict: the model answers atomic yes/no, choice and rubric questions about the decision, and the verdict is composed in code from visible coefficients instead of written in prose. Same state in, same verdict out, so pressure alone cannot move it. Returns a risk index, coverage, confidence and a JSON audit trail. Use when the user asks to "run the typed verdict", "challenge this decision with an audit trail", "reproducible risk verdict", "modo jev", "veredito tipado", "decisao tipada e auditavel". Do NOT use for a plain conversational counterpoint, use the ysat skill instead; nor for balanced pros and cons, a decision already executed, or evaluating people.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.