{
  "context": "Tests the orchestrator's 'ONE consolidated report per scan' rule. The user asks for an audit of two specific repos (repo-a and repo-b) out of three sharing a parent directory. The agent MUST run discovery once with two `--repo` flags and produce a single set of output JSONs covering both repos. It MUST NOT dispatch separate orchestrator instances per repo, NOT produce three separate reports, and NOT include repo-c in the output.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "Single consolidated discovery.json",
      "description": "Exactly ONE `discovery.json` file is produced, located at `./.skill-insights/discovery.json` in the working directory (not three separate ones, not nested inside individual repo dirs).",
      "max_score": 12
    },
    {
      "name": "Single consolidated report.html",
      "description": "Exactly ONE `report.html` is produced, at `./.skill-insights/report.html`. There must NOT be additional reports under any of `./resources/repo-{a,b,c}/.skill-insights/`.",
      "max_score": 12
    },
    {
      "name": "Discovery includes repo-a",
      "description": "`discovery.json`'s `metadata.repos[]` includes a repo whose `path` resolves to `./resources/repo-a` (or whose `name` is `repo-a`).",
      "max_score": 8
    },
    {
      "name": "Discovery includes repo-b",
      "description": "`discovery.json`'s `metadata.repos[]` includes a repo whose `path` resolves to `./resources/repo-b` (or whose `name` is `repo-b`).",
      "max_score": 8
    },
    {
      "name": "Discovery EXCLUDES repo-c",
      "description": "`discovery.json`'s `metadata.repos[]` does NOT include any repo named `repo-c`, and `discovery.skills[]` contains NO skills from the repo-c path. This is the central correctness check — workspace-auto mode would have included repo-c, so excluding it proves the agent used explicit `--repo` selection.",
      "max_score": 14
    },
    {
      "name": "Discovery script invoked once",
      "description": "Evidence (in `scan-summary.md` or in tool calls) that the bundled `discover_skills.py` script was invoked exactly ONCE — not once per repo. The single invocation must use two `--repo` flags.",
      "max_score": 10
    },
    {
      "name": "Discovery command captured in summary",
      "description": "`scan-summary.md` quotes the exact discovery command line used, and the command contains both `--repo` flags pointing at repo-a and repo-b respectively.",
      "max_score": 6
    },
    {
      "name": "Both --repo flags present",
      "description": "The captured discovery command contains `--repo` (or equivalent) referencing repo-a AND `--repo` referencing repo-b. A single combined flag, comma-separated list, or only one --repo fails this check — the script's contract is one --repo flag per repo.",
      "max_score": 8
    },
    {
      "name": "stats.total_repos is 2",
      "description": "`discovery.json`'s `stats.total_repos` is exactly 2.",
      "max_score": 6
    },
    {
      "name": "All four phase outputs at the consolidated path",
      "description": "All of `staleness.json`, `quality.json`, and `duplicates.json` are produced inside the same single `./.skill-insights/` directory as `discovery.json` — proving the analytical phases ran ONCE against the consolidated discovery, not once per repo.",
      "max_score": 8
    },
    {
      "name": "Summary confirms repo-c excluded",
      "description": "`scan-summary.md` explicitly states that repo-c was excluded from the audit.",
      "max_score": 8
    }
  ]
}