{
  "context": "Evaluates how well the runner uses CPython's interpreter primitives to execute files and modules, capture their IO, and evaluate inline code while respecting normal argv and path semantics. Focus is on choosing the standard runtime APIs rather than reimplementing interpreter behavior.",
  "type": "weighted_checklist",
  "checklist": [
    {
      "name": "runpy path",
      "description": "Script files are executed through runpy.run_path (or equivalent from runpy) with a __main__ run name so top-level code runs under interpreter semantics.",
      "max_score": 25
    },
    {
      "name": "runpy module",
      "description": "Modules are executed via runpy.run_module with alter_sys enabled to honor package __main__ entrypoints instead of manual imports or subprocess calls.",
      "max_score": 25
    },
    {
      "name": "argv plumbing",
      "description": "sys.argv is temporarily set to include the invoked target and provided args for both file and module execution, then restored afterward to avoid leaking state.",
      "max_score": 15
    },
    {
      "name": "stdout/stderr capture",
      "description": "Output is captured using contextlib.redirect_stdout/redirect_stderr with io.StringIO (or equivalent) rather than suppressing prints or discarding interpreter streams.",
      "max_score": 15
    },
    {
      "name": "workdir on sys.path",
      "description": "A supplied working directory is prepended to sys.path (with restoration and cache invalidation) so relative imports inside executed code resolve correctly.",
      "max_score": 10
    },
    {
      "name": "compile/eval",
      "description": "Inline expressions are compiled with compile(..., mode='eval') and evaluated with eval (falling back to exec for statements) using caller-provided globals/locals rather than subprocesses or literal-only helpers.",
      "max_score": 10
    }
  ]
}