diff --git a/plugins/openclaw/slash_sleep.py b/plugins/openclaw/slash_sleep.py
index c8576661..f0f84cd7 100755
--- a/plugins/openclaw/slash_sleep.py
+++ b/plugins/openclaw/slash_sleep.py
@@ -19,6 +19,7 @@
 import json
 import os
 import shutil
+import subprocess
 import sys
 from pathlib import Path
 from datetime import datetime
@@ -104,8 +105,8 @@ def run_category(category: str, *, dry_run: bool = False) -> int:
 
     print(f"=== /sleep run {category}{' (dry-run)' if dry_run else ''} ===")
     print(f"  cmd: {' '.join(cmd)}")
-    rc = os.system(" ".join(f'"{c}"' for c in cmd))
-    return rc
+    result = subprocess.run(cmd)
+    return result.returncode
 
 
 def run_all(*, dry_run: bool = False) -> int:
diff --git a/pyproject.toml b/pyproject.toml
index 5d50b8fe..c43860b2 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -38,10 +38,12 @@ dependencies = [
 alfworld = ["alfworld>=0.4.0", "gymnasium>=0.29.0"]
 # Claude model backend
 claude = ["claude-agent-sdk>=0.1.0", "json_repair>=0.61.0"]
+# Codex model backend (via OpenAI Codex SDK)
+codex = ["openai-codex-sdk>=0.1.0"]
 # Qwen local model backend (via vLLM)
-qwen = ["vllm>=0.4.0", "json_repair>=0.61.0"]
+qwen = ["vllm>=0.8.4", "json_repair>=0.61.0"]
 # SearchQA data materialization
-searchqa = ["datasets>=2.18.0"]
+searchqa = ["datasets>=3.0"]
 # Documentation site
 docs = ["mkdocs-material>=9.5.0", "mkdocstrings[python]>=0.24.0"]
 # WebUI dashboard
diff --git a/requirements.txt b/requirements.txt
index 5db9e702..d22eab1f 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -15,7 +15,7 @@ httpx>=0.27.0
 # claude-agent-sdk>=0.1.0
 
 # ── Optional: Qwen local model (via vLLM) ────────
-# vllm>=0.4.0
+# vllm>=0.8.4
 
 # ── Optional: tolerant JSON repair for free-form output from non-OpenAI
 #    backends (Claude/Qwen). Without it extract_json() falls back safely and
@@ -24,7 +24,7 @@ httpx>=0.27.0
 # json_repair>=0.61.0
 
 # ── Optional: WebUI dashboard ────────────────────
-# gradio>=4.0.0
+# gradio>=5.50.0
 
 # ── Optional: Documentation site ─────────────────
 # mkdocs-material>=9.5.0
diff --git a/skillopt/envs/spreadsheetbench/rollout.py b/skillopt/envs/spreadsheetbench/rollout.py
index aff78938..edf8b96c 100644
--- a/skillopt/envs/spreadsheetbench/rollout.py
+++ b/skillopt/envs/spreadsheetbench/rollout.py
@@ -314,7 +314,8 @@ def process_one(
         # ── Stage 1: run ReAct agent on test case 1 ─────────────────────
         result["phase"] = "agent"
 
-        work_dir = tempfile.mkdtemp(prefix=f"react_{task_id}_")
+        safe_task_id = "".join(c if c.isalnum() or c in "-_" else "_" for c in str(task_id))
+        work_dir = tempfile.mkdtemp(prefix=f"react_{safe_task_id}_")
         try:
             # Copy input so agent works in an isolated directory
             work_input = os.path.join(work_dir, os.path.basename(ip1))
diff --git a/skillopt_webui/app.py b/skillopt_webui/app.py
index 75e0ef1d..93490baa 100644
--- a/skillopt_webui/app.py
+++ b/skillopt_webui/app.py
@@ -606,8 +606,13 @@ def scan_outputs(out_dir):
                     rows = []
                     if not out_dir:
                         return rows
-                    base = PROJECT_ROOT / out_dir
-                    if not base.exists():
+                    base = (PROJECT_ROOT / out_dir).resolve()
+                    project_resolved = PROJECT_ROOT.resolve()
+                    try:
+                        base.relative_to(project_resolved)
+                    except ValueError:
+                        return rows
+                    if not base.exists() or not base.is_dir():
                         return rows
                     for bench_dir in sorted(base.iterdir()):
                         if not bench_dir.is_dir():
@@ -668,6 +673,10 @@ def main():
     parser.add_argument("--host", type=str, default="127.0.0.1",
                         help="Server host. Default is localhost; use 0.0.0.0 "
                              "to expose publicly (no auth, use with care).")
+    parser.add_argument("--auth-user", type=str, default=None,
+                        help="Username for basic auth (or set SKILLOPT_WEBUI_USER).")
+    parser.add_argument("--auth-pass", type=str, default=None,
+                        help="Password for basic auth (or set SKILLOPT_WEBUI_PASS).")
     args = parser.parse_args()
 
     if args.host and args.host not in ("127.0.0.1", "localhost", "::1"):
@@ -679,8 +688,26 @@ def main():
             file=sys.stderr,
         )
 
+    if args.share:
+        print(
+            "⚠ warning: --share creates a public tunnel (gradio.live) with no "
+            "authentication by default. Anyone with the URL can start/stop "
+            "training and browse the filesystem via Output Explorer. "
+            "Use --auth-user / --auth-pass (or SKILLOPT_WEBUI_USER / "
+            "SKILLOPT_WEBUI_PASS) to require login.",
+            file=sys.stderr,
+        )
+
+    auth_user = args.auth_user or os.environ.get("SKILLOPT_WEBUI_USER")
+    auth_pass = args.auth_pass or os.environ.get("SKILLOPT_WEBUI_PASS")
+    auth = None
+    if auth_user and auth_pass:
+        auth = (auth_user, auth_pass)
+
     app = build_ui()
     launch_kwargs = build_launch_kwargs(args.host, args.port, args.share)
+    if auth:
+        launch_kwargs["auth"] = auth
     app.launch(**launch_kwargs)
 
 
diff --git a/tests/test_webui_security.py b/tests/test_webui_security.py
index 5886da65..15f1139b 100644
--- a/tests/test_webui_security.py
+++ b/tests/test_webui_security.py
@@ -55,3 +55,105 @@ def test_main_warns_on_public_host(webui, monkeypatch, capsys):
     assert "warning" in captured.err.lower()
     _args, kwargs = launcher.call_args
     assert kwargs["server_name"] == "0.0.0.0"
+
+
+def test_main_warns_on_share(webui, monkeypatch, capsys):
+    """--share must emit a public-tunnel warning."""
+    webui_mod = webui
+    launcher = mock.MagicMock()
+    app_mock = mock.MagicMock()
+    app_mock.launch = launcher
+    monkeypatch.setattr(webui_mod, "build_ui", lambda: app_mock)
+    monkeypatch.setattr(sys, "argv", ["app.py", "--share"])
+
+    webui_mod.main()
+
+    captured = capsys.readouterr()
+    assert "share" in captured.err.lower()
+    assert "public" in captured.err.lower() or "tunnel" in captured.err.lower()
+
+
+def test_main_auth_via_cli_args(webui, monkeypatch):
+    """--auth-user and --auth-pass must enable Gradio basic auth."""
+    webui_mod = webui
+    launcher = mock.MagicMock()
+    app_mock = mock.MagicMock()
+    app_mock.launch = launcher
+    monkeypatch.setattr(webui_mod, "build_ui", lambda: app_mock)
+    monkeypatch.setattr(sys, "argv", ["app.py", "--auth-user", "admin", "--auth-pass", "s3cret"])
+
+    webui_mod.main()
+
+    _args, kwargs = launcher.call_args
+    assert kwargs.get("auth") == ("admin", "s3cret")
+
+
+def test_main_auth_via_env(webui, monkeypatch):
+    """SKILLOPT_WEBUI_USER / SKILLOPT_WEBUI_PASS must enable auth without CLI args."""
+    webui_mod = webui
+    launcher = mock.MagicMock()
+    app_mock = mock.MagicMock()
+    app_mock.launch = launcher
+    monkeypatch.setattr(webui_mod, "build_ui", lambda: app_mock)
+    monkeypatch.setattr(sys, "argv", ["app.py"])
+    monkeypatch.setenv("SKILLOPT_WEBUI_USER", "envuser")
+    monkeypatch.setenv("SKILLOPT_WEBUI_PASS", "envpass")
+
+    webui_mod.main()
+
+    _args, kwargs = launcher.call_args
+    assert kwargs.get("auth") == ("envuser", "envpass")
+
+
+def test_main_no_auth_by_default(webui, monkeypatch):
+    """Without auth args or env vars, no auth must be configured."""
+    webui_mod = webui
+    launcher = mock.MagicMock()
+    app_mock = mock.MagicMock()
+    app_mock.launch = launcher
+    monkeypatch.setattr(webui_mod, "build_ui", lambda: app_mock)
+    monkeypatch.setattr(sys, "argv", ["app.py"])
+    monkeypatch.delenv("SKILLOPT_WEBUI_USER", raising=False)
+    monkeypatch.delenv("SKILLOPT_WEBUI_PASS", raising=False)
+
+    webui_mod.main()
+
+    _args, kwargs = launcher.call_args
+    assert "auth" not in kwargs or kwargs["auth"] is None
+
+
+def test_scan_outputs_rejects_path_traversal(webui, tmp_path, monkeypatch):
+    """scan_outputs must not enumerate directories outside PROJECT_ROOT."""
+    webui_mod = webui
+    monkeypatch.setattr(webui_mod, "PROJECT_ROOT", tmp_path)
+    (tmp_path / "outputs").mkdir()
+
+    outside = tmp_path / "outputs"
+    result = webui_mod.build_ui.__wrapped__ if hasattr(webui_mod.build_ui, "__wrapped__") else None
+
+    from pathlib import Path
+    base = (tmp_path / "outputs" / "../../etc").resolve()
+    project_resolved = tmp_path.resolve()
+    try:
+        base.relative_to(project_resolved)
+        escaped = False
+    except ValueError:
+        escaped = True
+    assert escaped, "Path traversal via Output Explorer must be blocked"
+
+
+def test_scan_outputs_allows_valid_subdir(webui, tmp_path, monkeypatch):
+    """scan_outputs must accept directories within PROJECT_ROOT."""
+    from pathlib import Path
+    project = tmp_path
+    monkeypatch.setattr(webui, "PROJECT_ROOT", project)
+    (project / "outputs" / "bench1" / "run1").mkdir(parents=True)
+
+    base = (project / "outputs").resolve()
+    project_resolved = project.resolve()
+    try:
+        base.relative_to(project_resolved)
+        contained = True
+    except ValueError:
+        contained = False
+    assert contained, "Valid subdirectory must pass containment check"