{
  "skill_name": "control-cli",
  "evals": [
    {
      "id": 1,
      "name": "basic-invocation",
      "prompt": "Apply the control-cli skill to the following task: implement a user authentication feature with JWT tokens in a Node.js Express app.",
      "expected_output": "Output that follows the control-cli skill's guidance, including its key sections and recommendations.",
      "assertions": [
        "The output references or follows control-cli guidance",
        "The output provides actionable guidance",
        "The output is specific, not generic"
      ]
    },
    {
      "id": 2,
      "name": "ambiguous-input",
      "prompt": "The codebase is messy and needs cleanup. Apply the control-cli skill.",
      "expected_output": "The skill handles vague input by either asking clarifying questions or providing bounded guidance.",
      "assertions": [
        "The output does not hallucinate specific technologies not mentioned",
        "The output is relevant to the skill domain",
        "The output is actionable rather than purely descriptive"
      ]
    },
    {
      "id": 3,
      "name": "complex-scenario",
      "prompt": "A large enterprise team needs to adopt the practices from the control-cli skill across 5 microservices with different tech stacks. How should they apply it?",
      "expected_output": "The skill scales its guidance to a multi-team, multi-service context without losing specificity.",
      "assertions": [
        "The output addresses complexity without being vague",
        "The output provides concrete next steps",
        "The output acknowledges tradeoffs"
      ]
    }
  ]
}