{
  "skill_name": "weekly-review",
  "evals": [
    {
      "id": 1,
      "name": "basic-invocation",
      "prompt": "Apply the weekly-review skill: synthesize my commits from the past 7 days.",
      "expected_output": "A concise weekly synthesis of authored commits grouped into highlights by bugfix, tech debt, and net-new work.",
      "assertions": [
        "The output groups work into bugfix / tech debt / net-new categories",
        "The output covers only the requested time window",
        "The output reads as highlights, not a raw commit list"
      ]
    },
    {
      "id": 2,
      "name": "ambiguous-input",
      "prompt": "Do the weekly review.",
      "expected_output": "The skill determines the current week's commits from git authorship and reports if the window or repo is unclear.",
      "assertions": [
        "The output states the time window used",
        "The output does not fabricate commits",
        "The output handles an empty week gracefully"
      ]
    },
    {
      "id": 3,
      "name": "complex-scenario",
      "prompt": "Weekly review across 3 repos where I made 40 commits, half of them merge or chore commits.",
      "expected_output": "A cross-repo synthesis that filters mechanical commits (merges, chores) and highlights substantive work.",
      "assertions": [
        "Mechanical commits are filtered or de-emphasized",
        "The output aggregates across repos without losing per-repo context",
        "The synthesis is short enough for a status update"
      ]
    }
  ]
}