fix: integrate six focused workflow repairs with canonical registry identity

skills/agent-self-evaluation/scripts/evaluate.py

@@ -19,6 +19,7 @@ For production use, pair with an LLM judge for semantic understanding.
 """
 
 import argparse
+import errno
 import re
 import sys
 from dataclasses import dataclass, field
@@ -368,7 +369,9 @@ def _read_file_or_text(path: Optional[str], *, required: bool = False) -> Option
     try:
         with open(path) as f:
             return f.read()
-    except FileNotFoundError:
+    except OSError as error:
+        if error.errno not in (errno.ENOENT, errno.ENAMETOOLONG):
+            raise
         if required:
             print(f"Error: output file '{path}' not found", file=sys.stderr)
             sys.exit(1)
@@ -382,7 +385,10 @@ def _read_input(args: argparse.Namespace) -> tuple[Optional[str], str]:
         print("Paste agent output (Ctrl+D to finish):")
         return task, sys.stdin.read()
     if args.output:
-        return _read_file_or_text(args.task), _read_file_or_text(args.output, required=True) or ""
+        task = _read_file_or_text(args.task)
+        if task is None and not sys.stdin.isatty():
+            task = sys.stdin.read()
+        return task, _read_file_or_text(args.output, required=True) or ""
     return _read_file_or_text(args.task), sys.stdin.read()
 
 

skills/skill-comply/scripts/runner.py

@@ -94,8 +94,12 @@ def _safe_sandbox_dir(scenario_id: str) -> Path:
     """Sanitize scenario ID and ensure path stays within sandbox base."""
     safe_id = re.sub(r"[^a-zA-Z0-9\-_]", "_", scenario_id)
     path = SANDBOX_BASE / safe_id
+    resolved_path = path.resolve()
+    resolved_base = SANDBOX_BASE.resolve()
+    if not safe_id or resolved_path == resolved_base:
+        raise ValueError("Scenario sandbox must be a subdirectory of the sandbox base")
     # Validate path stays within sandbox base (raises ValueError on traversal)
-    path.resolve().relative_to(SANDBOX_BASE.resolve())
+    resolved_path.relative_to(resolved_base)
     return path
 
 

skills/skill-comply/tests/test_runner.py

@@ -9,7 +9,13 @@ from pathlib import Path
 from unittest.mock import patch
 
 import pytest
-from scripts.runner import _parse_stream_json, _setup_sandbox, run_scenario
+from scripts.runner import (
+    SANDBOX_BASE,
+    _parse_stream_json,
+    _safe_sandbox_dir,
+    _setup_sandbox,
+    run_scenario,
+)
 
 
 @dataclass(frozen=True)
@@ -21,6 +27,37 @@ class _FakeScenario:
     setup_commands: tuple[str, ...] = ()
 
 
+class TestSafeSandboxDirectory:
+    def test_empty_id_cannot_delete_the_shared_base(self) -> None:
+        with (
+            patch("scripts.runner.Path.exists", return_value=True),
+            patch("scripts.runner.shutil.rmtree") as remove,
+            patch("scripts.runner.Path.mkdir") as mkdir,
+            patch("scripts.runner.subprocess.run") as execute,
+        ):
+            execute.return_value = subprocess.CompletedProcess([], 0, stdout="", stderr="")
+            with pytest.raises(ValueError, match="sandbox"):
+                run_scenario(_FakeScenario(id=""))
+            remove.assert_not_called()
+            mkdir.assert_not_called()
+            execute.assert_not_called()
+
+    def test_resolved_base_is_not_a_scenario_directory(self) -> None:
+        with patch("scripts.runner.Path.resolve", return_value=SANDBOX_BASE):
+            with pytest.raises(ValueError, match="sandbox"):
+                _safe_sandbox_dir("alias")
+
+    def test_valid_id_keeps_existing_sanitization(self) -> None:
+        assert _safe_sandbox_dir("scenario/one") == SANDBOX_BASE / "scenario_one"
+
+    def test_resolved_escape_is_still_rejected(self, tmp_path, monkeypatch) -> None:
+        base = tmp_path / "sandbox"
+        monkeypatch.setattr("scripts.runner.SANDBOX_BASE", base)
+        with patch.object(Path, "resolve", side_effect=[tmp_path / "outside", base]):
+            with pytest.raises(ValueError):
+                _safe_sandbox_dir("escape")
+
+
 class TestSetupSandboxSkipsShellBuiltins:
     """Setup commands containing shell builtins (cd/pushd/popd) must be skipped.