Function buildSkillScenarios

plugins/plugin-eval/src/core/benchmark.js:108–142

(target)

}

function buildSkillScenarios(target) {
  return [
    {
      id: "happy-path",
      title: "Happy path implementation",
      purpose: "Run a representative task that should clearly justify using this skill inside the workspace.",
      userInput: `Use the local Codex skill "${target.name}" if it helps. Complete a representative task for this skill in the current workspace and leave the result on disk.`,
      successChecklist: [
        "The task is completed in the workspace, not only described.",
        "The final answer explains what changed.",
        "The run stays aligned with the skill's specialty.",
      ],
    },
    {
      id: "follow-up",
      title: "Focused refinement",
      purpose: "Measure whether the skill can improve an implementation cleanly instead of restarting from scratch.",
      userInput: `Use the local Codex skill "${target.name}" if it helps. Refine or extend the current workspace with one focused follow-up improvement and finish the change end to end.`,
      successChecklist: [
        "The result makes a concrete change on disk.",
        "The follow-up remains scoped and coherent.",
      ],
    },
    {
      id: "boundary-case",
      title: "Boundary handling",
      purpose: "Check whether the skill avoids overreaching when the task is only a partial fit.",
      userInput: `This task is only a partial match for the local Codex skill "${target.name}". Handle the appropriate slice in the workspace and narrow or refuse the rest honestly.`,
      successChecklist: [
        "The run sets good boundaries instead of pretending the skill fits everything.",
        "Any edits stay aligned with the justified scope.",
      ],
    },
  ];
}

function buildPluginScenarios(target) {
  return [