{
  "context": "Tests whether the agent enforces the evaluation-only scope by refusing to edit or improve a SKILL.md submission and by declining to answer contest logistics questions. The agent must stay in its lane as an evaluator and not drift into being a writing assistant or information service.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Editing request refused",
      "description": "Output explicitly declines to edit, improve, rewrite, or enhance the SKILL.md — does not produce a revised or improved version",
      "max_score": 14
    },
    {
      "name": "No edited skill content",
      "description": "Output does NOT contain a rewritten or improved SKILL.md (no frontmatter block with modified content for the csv-to-json skill)",
      "max_score": 12
    },
    {
      "name": "Contest logistics refused",
      "description": "Output explicitly declines to answer the contest logistics question (submission deadline or submission process)",
      "max_score": 14
    },
    {
      "name": "No contest logistics answered",
      "description": "Output does NOT contain specific contest submission details — no deadline date, no submission URL, no step-by-step submission instructions",
      "max_score": 12
    },
    {
      "name": "Offers evaluation instead",
      "description": "Output offers to evaluate the submitted SKILL.md as the available service (pivots to what the evaluator can do rather than just refusing)",
      "max_score": 12
    },
    {
      "name": "No ranking produced",
      "description": "Output does NOT rank the skill against other submissions or claim it is better or worse than other contest entries",
      "max_score": 8
    },
    {
      "name": "No CLI execution claimed",
      "description": "Output does NOT claim to execute the SKILL.md, test it programmatically, or run it as code",
      "max_score": 8
    },
    {
      "name": "Refusal is clear not evasive",
      "description": "The refusal of the editing request does NOT include a workaround like a full list of 'suggestions to improve it yourself' that effectively rewrites the skill in disguise",
      "max_score": 10
    },
    {
      "name": "Response remains brief",
      "description": "Output does NOT produce a full scorecard evaluation (no Core Score table with all 8 dimensions) — the agent is refusing, not pivoting to an unsolicited evaluation",
      "max_score": 8
    },
    {
      "name": "No non-AIE26 tangent",
      "description": "Output does NOT attempt to answer general questions about AI engineering, contest strategy, or provide unsolicited skill-writing advice beyond briefly noting what the evaluator can do",
      "max_score": 2
    }
  ]
}