refactor(claw-eval): use the standard <task>/task.json layout

CHANGELOG.md

@@ -7,6 +7,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/)
 and this project adheres to [Semantic Versioning](https://semver.org/).
 
 ## [Unreleased]
+### Changed
+- The `claw-eval` port now uses the same `test-cases/<suite>/<task-identifier>/task.json` layout as the native corpora, instead of flat `<task-identifier>.json` files. Case discovery in `clawbench-batch` and the TUI is a plain `*/task.json` search again, and the `validate-task` workflow covers the suite without special-casing.
 
 ## [0.10.0] - 2026-08-30
 ### Added

src/clawbench/runner/batch.py

@@ -81,7 +81,7 @@ def discover_models(patterns: list[str] | None, all_models: bool) -> list[str]:
 
 
 def _case_id(d: Path) -> int | None:
-    """Extract the numeric task ID from V1/V2/flat Claw-Eval case names."""
+    """Extract the numeric task ID from V1/V2/Claw-Eval case names."""
     match = re.match(r"^(?:v\d+-|ce-)?[A-Za-z]?(\d+)", d.stem)
     if not match:
         return None
@@ -104,16 +104,8 @@ def _resolve_cases_dir(cases_dir: str | Path) -> Path:
     return path
 
 
-def _flat_case_files(base: Path) -> list[Path]:
-    return [
-        p
-        for p in base.glob("*.json")
-        if p.is_file() and p.name not in {"eligibility-report.json"}
-    ]
-
-
 def _all_cases_in(base: Path) -> list[Path]:
-    return [p.parent for p in base.glob("*/task.json")] + _flat_case_files(base)
+    return [p.parent for p in base.glob("*/task.json")]
 
 
 def discover_cases(
@@ -139,8 +131,6 @@ def discover_cases(
             for d in expanded:
                 if d.is_dir() and (d / "task.json").exists():
                     dirs.append(d)
-                elif d.is_file() and d.suffix == ".json":
-                    dirs.append(d)
     elif case_range:
         dirs = sorted(_all_cases_in(base), key=_case_sort_key)
     else:

src/clawbench/tui.py

@@ -383,13 +383,8 @@ def _case_sort_key(case: str) -> tuple[int, int, str]:
 
 def load_cases(cases_dir_name: str = "test-cases") -> list[str]:
     cases_dir = ASSET_ROOT / cases_dir_name
-    flat_cases = [
-        p.stem
-        for p in cases_dir.glob("*.json")
-        if p.name not in {"eligibility-report.json"}
-    ]
     cases = sorted(
-        [*(p.parent.name for p in cases_dir.glob("*/task.json")), *flat_cases],
+        (p.parent.name for p in cases_dir.glob("*/task.json")),
         key=_case_sort_key,
     )
     if not cases:

tests/test_host_tasks.py

@@ -18,13 +18,7 @@ def _suite_base(suite: str) -> Path:
 
 def _task_files_for_suite(suite: str) -> list[Path]:
     base = _suite_base(suite)
-    task_files = [path for path in base.glob("*/task.json") if path.is_file()]
-    task_files.extend(
-        path
-        for path in base.glob("*.json")
-        if path.is_file() and path.name != "eligibility-report.json"
-    )
-    return sorted(task_files)
+    return sorted(path for path in base.glob("*/task.json") if path.is_file())
 
 
 @pytest.mark.parametrize("suite", sorted(batch.CASE_SUITES))
@@ -37,10 +31,8 @@ def test_builtin_case_suites_are_discoverable(suite: str) -> None:
 
     assert cases, f"{suite} should contain at least one case"
     for case in cases:
-        if case.is_dir():
-            assert (case / "task.json").is_file()
-        else:
-            assert case.is_file()
+        assert case.is_dir(), f"{case} should be a task directory"
+        assert (case / "task.json").is_file()
 
 
 @pytest.mark.parametrize("suite", sorted(batch.CASE_SUITES))

tests/test_tui_helpers.py

@@ -129,7 +129,7 @@ def test_tui_dataset_and_case_helpers_use_known_suites() -> None:
     )
 
 
-def test_tui_load_cases_reads_flat_and_directory_suites() -> None:
+def test_tui_load_cases_reads_every_builtin_suite() -> None:
     v1_cases = tui.load_cases("test-cases/v1")
     claw_eval_cases = tui.load_cases("test-cases/claw-eval")