What problem does it solve? Designers and AI agents tend to elaborate the first idea that comes to mind, producing polished variations of a single bet instead of genuinely different directions. This Skill enforces a diverge-then-converge loop grounded in NN/g parallel-design evidence, so alternatives are compared before any one direction is built. ## Core Features & Use Cases - Forced structural divergence: Generates exactly three concepts, each carrying an opposing product priority, with a swap test that catches cosmetic-only differences. - Concept sheet documentation: Produces a deliverable per concept covering the bet, behavior sketch, honest best case, named risk, and the goal it best serves. - Criteria-first convergence: Writes scoring criteria before ranking, then outputs one recommendation via either select-the-winner or synthesize-across-candidates, with losers harvested as layers. - Use Case: When designing a conversational shopping feature, use it to produce three trust-model concepts (people-based, reasoning-based, effortlessness-based), score them against pre-written criteria, and hand one recommendation to the decision-maker. ## Quick Start Ask the agent to explore three divergent design concepts for your feature and recommend one with explicit convergence criteria.
