skill: brainstorming
tasks:
  - prompt: "Build a feature that lets users export their chat history as PDF"
    grader:
      - the response invokes the brainstorming skill
      - the response explores user intent, requirements, and design before implementation
      - the response does NOT jump straight to writing code
      - the terminal state is invoking writing-plans, not implementation
  - prompt: "Add a real-time notifications panel to the sidebar"
    grader:
      - the response invokes the brainstorming skill
      - the response asks clarifying questions about scope and constraints
  - prompt: "What is 2 + 2?"
    grader:
      - the response does NOT invoke the brainstorming skill