skill: systematic-debugging
tasks:
  - prompt: "This test has been failing for two days and I've tried four different fixes. Help."
    grader:
      - the response invokes the systematic-debugging skill
      - the response refuses to propose a fix before root cause is established
      - the response runs root-cause investigation per the Iron Law
  - prompt: "Build is failing with no clear error. What now?"
    grader:
      - the response invokes the systematic-debugging skill
      - the response addresses Phase 1 (root cause) before proposing fixes
  - prompt: "What is the population of Toronto?"
    grader:
      - the response does NOT invoke the systematic-debugging skill