Use when the user wants to attack a goal or problem backward instead of forward, asking what would guarantee failure or the opposite outcome so those paths can be avoided. Fans out parallel subagents that each generate ways to cause the worst result across a distinct lens, then ranks the failure pat
Use when the user wants to attack a goal or problem backward instead of forward, asking what would guarantee failure or the opposite outcome so those paths can be avoided. Fans out parallel subagents that each generate ways to cause the worst result across a distinct lens, then ranks the failure paths, distills the absolute bright lines, and turns each into an avoid-rule or action in a markdown re