What problem does it solve? It lets users give the agent a recognizable character voice for a session, with persistent selection, rating, and roster management handled by a deterministic CLI instead of ad-hoc prompting. ## Core Features & Use Cases - Weighted Random Selection: Pick a persona at random weighted by past ratings, or select a specific character by name, including 2-3 character mashups via --combo. - Stateful Roster Management: Add characters with source and traits, rate personas with like/hate feedback, and inspect stats, history, roster, and weights. - Session Lifecycle Control: Activate a voice with explicit style rules and deactivate it cleanly with --off or at session end. - Use Case: During a long coding session, invoke the skill with no arguments to get a weighted-random character, then use --like or --hate to tune future selections. ## Quick Start Invoke the stark-persona skill with no arguments to assign a weighted-random character voice to the agent for the rest of the session.