cookieguard-browser-use 0.1.0__tar.gz

cookieguard_browser_use-0.1.0/.gitignore

@@ -0,0 +1,11 @@
+# The repository root ignores `_*`; Python packages need __init__.py, _private.py, and _node/.
+!_*
+.venv/
+dist/
+build/
+__pycache__/
+*.egg-info/
+.pytest_cache/
+# Generated by scripts/vendor_engine.mjs from the CookieGuard npm packages.
+src/cookieguard_browser_use/_node/node_modules/
+src/cookieguard_browser_use/_node/manifest.json

cookieguard_browser_use-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 CookieGuard AI
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

cookieguard_browser_use-0.1.0/PKG-INFO

@@ -0,0 +1,104 @@
+Metadata-Version: 2.5
+Name: cookieguard-browser-use
+Version: 0.1.0
+Summary: CookieGuard authentication-cookie policy checks for Browser Use agents
+Project-URL: Homepage, https://github.com/aiscihub/cookieguard-ai/tree/main/integrations/browser-use
+Project-URL: Repository, https://github.com/aiscihub/cookieguard-ai
+Project-URL: Issues, https://github.com/aiscihub/cookieguard-ai/issues
+Author: aiscihub
+License-Expression: MIT
+License-File: LICENSE
+Keywords: ai-agents,authentication,browser-use,cookies,security,testing
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Security
+Classifier: Topic :: Software Development :: Testing
+Requires-Python: >=3.11
+Requires-Dist: browser-use<0.14,>=0.13.14
+Requires-Dist: cdp-use<2,>=1.4.5
+Requires-Dist: pydantic<3,>=2.10
+Provides-Extra: test
+Requires-Dist: pytest>=8.3; extra == 'test'
+Description-Content-Type: text/markdown
+
+# CookieGuard for Browser Use
+
+**A successful agent login is not a secure session cookie.** CookieGuard for
+Browser Use watches the browser your [Browser Use](https://github.com/browser-use/browser-use)
+agent drives, checks the authentication cookies it receives, and returns
+**PASS / FAIL / INCOMPLETE** with rule IDs, explanations, and fixes. The decision comes
+from the local [CookieGuard](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/README.md) policy engine, never from the model.
+
+**Overview** · [Install](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/install.md) · [Policy](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/policy.md) · [How it works](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/how-it-works.md) · [Results](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/results.md) · [Limits & privacy](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/limits-and-privacy.md) · [Demo & tests](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/demo-and-tests.md)
+
+## Your policy decides what fails
+
+<picture>
+  <source media="(max-width: 600px) and (prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/docs/policy-dark-mobile.svg">
+  <source media="(max-width: 600px)" srcset="https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/docs/policy-light-mobile.svg">
+  <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/docs/policy-dark.svg">
+  <img alt="CookieGuardAudit options decide which findings fail. High-severity rules CG001 Missing HttpOnly, CG002 Missing Secure, CG006 invalid prefix, and CG007 regression fail by default. With fail_on set to medium, CG003 SameSite and CG005 lifetime fail too. CG004 Domain sharing fails only with require_host_only. CG008 and CG009 are review only. The result is FAIL (exit 1), INCOMPLETE (exit 2, which overrides FAIL), or PASS (exit 0)." src="https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/docs/policy-light.svg" width="100%">
+</picture>
+
+`fail_on` sets the line: a confirmed finding at or above it fails the run; anything below
+is reported. Only cookies you name in `auth_cookies` can be confirmed.
+All options and a stricter example: **[Policy](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/policy.md)**.
+
+## See it run
+
+![A Browser Use agent logs in to two local apps. Both logins succeed. CookieGuard fails the first with CG001 Missing HttpOnly (exit 1) and passes the second (exit 0).](https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/docs/cookieguard-browser-use-demo.gif)
+
+Claude (`claude-opus-5-5`) drives a real Browser Use agent through the same login twice.
+The insecure app's cookie lacks `HttpOnly`, so CookieGuard returns **FAIL — CG001** (exit 1).
+The secure app gets **PASS** (exit 0). Recorded from a real run: **[Demo & tests](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/demo-and-tests.md)**.
+
+## Add it to your agent
+
+```python
+from browser_use import Agent, BrowserProfile, BrowserSession, Tools
+from cookieguard_browser_use import CookieGuardAudit
+
+audit = CookieGuardAudit(
+    allowed_origins=['http://127.0.0.1:3000'],  # exact origins you are authorized to test
+    auth_cookies=['session_id'],                # the real session cookie name(s)
+    fail_on='high',
+)
+session = BrowserSession(browser_profile=BrowserProfile(
+    keep_alive=True,                            # required: Agent.run() otherwise closes the browser
+    enable_default_extensions=False,            # avoid the bundled cookie-consent extension
+    allowed_domains=audit.browser_allowed_domains(),
+))
+tools = Tools()
+audit.register_tools(tools)                     # optional agent-visible check
+
+await session.start()
+await audit.start(session)                      # before the login
+agent = Agent(task=..., llm=..., browser_session=session, tools=tools,
+              sensitive_data={'http://127.0.0.1': {'login_password': '...'}})
+await agent.run()
+result = await audit.finish()                   # decided by the CookieGuard engine
+await session.kill()
+
+print(result.summary)
+result.assert_policy()                          # raises CookiePolicyError on FAIL or INCOMPLETE
+```
+
+`assert_policy()` raises even if the agent never called the tool, or reported
+success. `CookieGuardAudit` is also an async context manager; `close()` stops
+collection without a result. One instance audits one workflow.
+
+Needs Python 3.11+, Node.js 24+, and Chrome or Chromium: **[Install](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/install.md)**.
+
+Install [version 0.1.0 from PyPI](https://pypi.org/project/cookieguard-browser-use/):
+
+```bash
+pip install cookieguard-browser-use
+```
+
+Status: `0.1.0`, tested with browser-use 0.13.14 on Linux (Python 3.11–3.13) and macOS (Python 3.12).
+See [tested environments](https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/docs/limits-and-privacy.md#tested-environments-and-dependencies).
+This is a community integration, not an official Browser Use package.

cookieguard_browser_use-0.1.0/README.md

@@ -0,0 +1,77 @@
+# CookieGuard for Browser Use
+
+**A successful agent login is not a secure session cookie.** CookieGuard for
+Browser Use watches the browser your [Browser Use](https://github.com/browser-use/browser-use)
+agent drives, checks the authentication cookies it receives, and returns
+**PASS / FAIL / INCOMPLETE** with rule IDs, explanations, and fixes. The decision comes
+from the local [CookieGuard](../../README.md) policy engine, never from the model.
+
+**Overview** · [Install](docs/install.md) · [Policy](docs/policy.md) · [How it works](docs/how-it-works.md) · [Results](docs/results.md) · [Limits & privacy](docs/limits-and-privacy.md) · [Demo & tests](docs/demo-and-tests.md)
+
+## Your policy decides what fails
+
+<picture>
+  <source media="(max-width: 600px) and (prefers-color-scheme: dark)" srcset="docs/policy-dark-mobile.svg">
+  <source media="(max-width: 600px)" srcset="docs/policy-light-mobile.svg">
+  <source media="(prefers-color-scheme: dark)" srcset="docs/policy-dark.svg">
+  <img alt="CookieGuardAudit options decide which findings fail. High-severity rules CG001 Missing HttpOnly, CG002 Missing Secure, CG006 invalid prefix, and CG007 regression fail by default. With fail_on set to medium, CG003 SameSite and CG005 lifetime fail too. CG004 Domain sharing fails only with require_host_only. CG008 and CG009 are review only. The result is FAIL (exit 1), INCOMPLETE (exit 2, which overrides FAIL), or PASS (exit 0)." src="docs/policy-light.svg" width="100%">
+</picture>
+
+`fail_on` sets the line: a confirmed finding at or above it fails the run; anything below
+is reported. Only cookies you name in `auth_cookies` can be confirmed.
+All options and a stricter example: **[Policy](docs/policy.md)**.
+
+## See it run
+
+![A Browser Use agent logs in to two local apps. Both logins succeed. CookieGuard fails the first with CG001 Missing HttpOnly (exit 1) and passes the second (exit 0).](docs/cookieguard-browser-use-demo.gif)
+
+Claude (`claude-opus-5-5`) drives a real Browser Use agent through the same login twice.
+The insecure app's cookie lacks `HttpOnly`, so CookieGuard returns **FAIL — CG001** (exit 1).
+The secure app gets **PASS** (exit 0). Recorded from a real run: **[Demo & tests](docs/demo-and-tests.md)**.
+
+## Add it to your agent
+
+```python
+from browser_use import Agent, BrowserProfile, BrowserSession, Tools
+from cookieguard_browser_use import CookieGuardAudit
+
+audit = CookieGuardAudit(
+    allowed_origins=['http://127.0.0.1:3000'],  # exact origins you are authorized to test
+    auth_cookies=['session_id'],                # the real session cookie name(s)
+    fail_on='high',
+)
+session = BrowserSession(browser_profile=BrowserProfile(
+    keep_alive=True,                            # required: Agent.run() otherwise closes the browser
+    enable_default_extensions=False,            # avoid the bundled cookie-consent extension
+    allowed_domains=audit.browser_allowed_domains(),
+))
+tools = Tools()
+audit.register_tools(tools)                     # optional agent-visible check
+
+await session.start()
+await audit.start(session)                      # before the login
+agent = Agent(task=..., llm=..., browser_session=session, tools=tools,
+              sensitive_data={'http://127.0.0.1': {'login_password': '...'}})
+await agent.run()
+result = await audit.finish()                   # decided by the CookieGuard engine
+await session.kill()
+
+print(result.summary)
+result.assert_policy()                          # raises CookiePolicyError on FAIL or INCOMPLETE
+```
+
+`assert_policy()` raises even if the agent never called the tool, or reported
+success. `CookieGuardAudit` is also an async context manager; `close()` stops
+collection without a result. One instance audits one workflow.
+
+Needs Python 3.11+, Node.js 24+, and Chrome or Chromium: **[Install](docs/install.md)**.
+
+Install [version 0.1.0 from PyPI](https://pypi.org/project/cookieguard-browser-use/):
+
+```bash
+pip install cookieguard-browser-use
+```
+
+Status: `0.1.0`, tested with browser-use 0.13.14 on Linux (Python 3.11–3.13) and macOS (Python 3.12).
+See [tested environments](docs/limits-and-privacy.md#tested-environments-and-dependencies).
+This is a community integration, not an official Browser Use package.

cookieguard_browser_use-0.1.0/examples/agent_login.py

@@ -0,0 +1,102 @@
+"""CookieGuard + a Browser Use agent logging in to a local demo app.
+
+    python examples/agent_login.py --mode insecure                       # scripted, no API key
+    ANTHROPIC_API_KEY=... python examples/agent_login.py --provider anthropic --mode insecure
+    BROWSER_USE_API_KEY=... python examples/agent_login.py --provider browser-use --mode secure
+    OPENAI_API_KEY=... python examples/agent_login.py --provider openai --model <model-id> --mode secure
+
+Exit code: 0 PASS, 1 confirmed policy FAIL, 2 INCOMPLETE. With --mode insecure the
+login succeeds and the exit code is 1: the agent reaching the account page does
+not make the session cookie secure.
+
+Model providers receive the task, page content, and CookieGuard's sanitized tool
+output. Credentials stay in Browser Use's sensitive_data and are typed as
+placeholders; cookie values never reach the model. The app is loopback-only.
+"""
+
+from __future__ import annotations
+
+import argparse
+import asyncio
+import os
+import sys
+
+# Browser Use telemetry would send the task, visited URLs, actions, and final answer.
+os.environ.setdefault('ANONYMIZED_TELEMETRY', 'false')
+os.environ.setdefault('BROWSER_USE_CLOUD_SYNC', 'false')
+
+from browser_use import Agent, BrowserProfile, BrowserSession, Tools  # noqa: E402
+
+from cookieguard_browser_use import CookieGuardAudit, CookiePolicyError  # noqa: E402
+from cookieguard_browser_use.demo.app import DEMO_PASSWORD, DEMO_USERNAME, DemoApp  # noqa: E402
+from cookieguard_browser_use.demo.scripted import TASK, ScriptedLoginModel  # noqa: E402
+
+KEYS = {'anthropic': 'ANTHROPIC_API_KEY', 'openai': 'OPENAI_API_KEY', 'browser-use': 'BROWSER_USE_API_KEY'}
+
+
+def make_llm(provider: str, model: str | None, origin: str):
+	if provider == 'scripted':
+		return ScriptedLoginModel(origin)  # Deterministic stand-in; not an AI model.
+	if not os.environ.get(KEYS[provider]):
+		raise SystemExit(f'{KEYS[provider]} is not set; use --provider scripted to run without a model.')
+	if provider == 'anthropic':
+		from browser_use import ChatAnthropic
+		# Adaptive thinking makes browser-use 0.13 send tool_choice "auto"; current
+		# Claude models reject the forced tool_choice it would otherwise send.
+		# Server-side fallback: if a safety classifier declines a step (observed
+		# intermittently on this login task), the API reruns it on the fallback model.
+		return ChatAnthropic(model=model or 'claude-opus-5-5', thinking={'type': 'adaptive'},
+			fallbacks=[{'model': 'claude-opus-4-8'}])
+	if provider == 'browser-use':
+		from browser_use import ChatBrowserUse
+		return ChatBrowserUse(**({'model': model} if model else {}))
+	if not model:
+		raise SystemExit('--model is required for --provider openai.')
+	from browser_use import ChatOpenAI
+	return ChatOpenAI(model=model)
+
+
+async def main(args: argparse.Namespace) -> int:
+	with DemoApp(args.mode) as app:
+		# 1. Configure the audit: exact authorized origins and the known session cookie.
+		audit = CookieGuardAudit(allowed_origins=[app.origin], auth_cookies=['session_id'], fail_on='high')
+		# keep_alive=True: Agent.run() would otherwise close the browser before audit.finish().
+		session = BrowserSession(browser_profile=BrowserProfile(
+			headless=not args.headed, keep_alive=True, enable_default_extensions=False,
+			allowed_domains=audit.browser_allowed_domains()))
+		tools = Tools()
+		audit.register_tools(tools)  # Optional: lets the agent request a sanitized provisional check.
+		try:
+			# 2. Start the browser, then evidence collection, before the login.
+			await session.start()
+			await audit.start(session)
+			# 3-4. The agent performs the authorized login and workflow.
+			agent = Agent(
+				task=TASK.format(origin=app.origin), llm=make_llm(args.provider, args.model, app.origin),
+				browser_session=session, tools=tools, use_vision=False,
+				sensitive_data={audit.browser_allowed_domains()[0]: {'login_user': DEMO_USERNAME, 'login_password': DEMO_PASSWORD}})
+			history = await agent.run(max_steps=12)
+			# 5. Evaluate with the CookieGuard engine, independent of what the agent concluded.
+			result = await audit.finish()
+		finally:
+			await audit.close()
+			await session.kill()
+		print(f'\nAgent final answer: {history.final_result()}')
+		print(f'Server-side logins: {app.logins}\n')
+		print(result.summary)
+		result.write_json(f'cookieguard-{args.mode}.json')
+	# 6. Enforce. In pytest, just call result.assert_policy().
+	try:
+		result.assert_policy()
+	except CookiePolicyError as error:
+		print(error)
+	return result.exit_code
+
+
+if __name__ == '__main__':
+	parser = argparse.ArgumentParser(description='Browser Use agent login audited by CookieGuard.')
+	parser.add_argument('--mode', choices=['insecure', 'secure'], default='insecure')
+	parser.add_argument('--provider', choices=['scripted', *KEYS], default='scripted')
+	parser.add_argument('--model', help='model id for the selected provider')
+	parser.add_argument('--headed', action='store_true')
+	sys.exit(asyncio.run(main(parser.parse_args())))

cookieguard_browser_use-0.1.0/hatch_build.py

@@ -0,0 +1,22 @@
+"""Refuse to build a wheel without a complete, verified vendored engine."""
+
+import hashlib
+import json
+from pathlib import Path
+
+from hatchling.builders.hooks.plugin.interface import BuildHookInterface
+
+
+class CustomBuildHook(BuildHookInterface):
+	def initialize(self, version, build_data):
+		engine = Path(self.root) / 'src' / 'cookieguard_browser_use' / '_node'
+		manifest_path = engine / 'manifest.json'
+		if not manifest_path.is_file():
+			raise RuntimeError('Vendored engine missing. Run: node scripts/vendor_engine.mjs --registry')
+		manifest = json.loads(manifest_path.read_text(encoding='utf-8'))
+		expected = dict(manifest['bridge'])
+		for package in manifest['packages']:
+			expected.update(package['files'])
+		for relative, digest in expected.items():
+			if hashlib.sha256((engine / relative).read_bytes()).hexdigest() != digest:
+				raise RuntimeError('Vendored engine does not match its manifest. Re-run scripts/vendor_engine.mjs.')

cookieguard_browser_use-0.1.0/pyproject.toml

@@ -0,0 +1,85 @@
+[build-system]
+requires = ["hatchling>=1.27", "hatch-fancy-pypi-readme==25.1.0"]
+build-backend = "hatchling.build"
+
+[project]
+name = "cookieguard-browser-use"
+version = "0.1.0"
+description = "CookieGuard authentication-cookie policy checks for Browser Use agents"
+dynamic = ["readme"]
+license = "MIT"
+license-files = ["LICENSE"]
+requires-python = ">=3.11"
+authors = [{ name = "aiscihub" }]
+keywords = ["browser-use", "cookies", "security", "authentication", "ai-agents", "testing"]
+classifiers = [
+  "Development Status :: 4 - Beta",
+  "Intended Audience :: Developers",
+  "Programming Language :: Python :: 3",
+  "Programming Language :: Python :: 3.11",
+  "Programming Language :: Python :: 3.12",
+  "Programming Language :: Python :: 3.13",
+  "Topic :: Security",
+  "Topic :: Software Development :: Testing",
+]
+# Tested with browser-use 0.13.14 / cdp-use 1.4.5. Node.js 24+ is required at
+# runtime for the vendored CookieGuard engine; it is not a Python dependency.
+dependencies = [
+  "browser-use>=0.13.14,<0.14",
+  "cdp-use>=1.4.5,<2",
+  "pydantic>=2.10,<3",
+]
+
+[project.optional-dependencies]
+test = ["pytest>=8.3"]
+
+[project.urls]
+Homepage = "https://github.com/aiscihub/cookieguard-ai/tree/main/integrations/browser-use"
+Repository = "https://github.com/aiscihub/cookieguard-ai"
+Issues = "https://github.com/aiscihub/cookieguard-ai/issues"
+
+[tool.hatch.metadata.hooks.fancy-pypi-readme]
+content-type = "text/markdown"
+fragments = [{ path = "README.md" }]
+
+# Only the package description gets absolute links. Keep GitHub's README relative.
+# Update these release-tag pins together with the package version.
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '''(?P<prefix>(?:src|srcset)="|!\[[^\]]*\]\()(?P<path>docs/[^"()\s]+\.(?:svg|gif))'''
+replacement = '\g<prefix>https://raw.githubusercontent.com/aiscihub/cookieguard-ai/browser-use-v0.1.0/integrations/browser-use/\g<path>'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '''(?P<prefix>\]\()(?P<path>docs/[^)\s]+\.md(?:#[^)\s]*)?)'''
+replacement = '\g<prefix>https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/integrations/browser-use/\g<path>'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = '''(?P<prefix>\]\()\.\./\.\./(?P<path>README\.md(?:#[^)\s]*)?)'''
+replacement = '\g<prefix>https://github.com/aiscihub/cookieguard-ai/blob/browser-use-v0.1.0/\g<path>'
+
+[[tool.hatch.metadata.hooks.fancy-pypi-readme.substitutions]]
+pattern = ', not yet on PyPI\.'
+replacement = '.'
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/cookieguard_browser_use"]
+# Generated by scripts/vendor_engine.mjs and ignored by git, so list explicitly.
+artifacts = [
+  "src/cookieguard_browser_use/_node/node_modules/**",
+  "src/cookieguard_browser_use/_node/manifest.json",
+]
+
+[tool.hatch.build.targets.wheel.hooks.custom]
+path = "hatch_build.py"
+
+[tool.hatch.build.targets.sdist]
+include = ["src", "tests", "examples", "scripts", "hatch_build.py", "README.md", "LICENSE", "pyproject.toml"]
+artifacts = [
+  "src/cookieguard_browser_use/_node/node_modules/**",
+  "src/cookieguard_browser_use/_node/manifest.json",
+]
+
+[tool.pytest.ini_options]
+testpaths = ["tests"]
+markers = [
+  "browser: launches a real local Chromium through Browser Use",
+]

cookieguard_browser_use-0.1.0/scripts/browser_smoke.py

@@ -0,0 +1,66 @@
+"""Bounded, step-by-step smoke check of the browser stack, for CI diagnostics.
+
+    python scripts/browser_smoke.py
+
+Each step has its own timeout and prints a marker, so a hang or crash names
+the step. No credentials or cookies are involved: it only opens the login page.
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+import sys
+import time
+
+for key, value in (('ANONYMIZED_TELEMETRY', 'false'), ('BROWSER_USE_CLOUD_SYNC', 'false')):
+	os.environ.setdefault(key, value)
+
+
+def mark(step: str, started: float) -> None:
+	print(f'[smoke] {step} ({time.monotonic() - started:.1f}s)', flush=True)
+
+
+async def bounded(step: str, awaitable, seconds: float):
+	started = time.monotonic()
+	print(f'[smoke] {step} ...', flush=True)
+	result = await asyncio.wait_for(awaitable, seconds)
+	mark(f'{step}: ok', started)
+	return result
+
+
+async def main() -> int:
+	from browser_use import BrowserProfile, BrowserSession, Tools
+	from browser_use.browser.watchdogs.local_browser_watchdog import LocalBrowserWatchdog
+
+	from cookieguard_browser_use import CookieGuardAudit
+	from cookieguard_browser_use._engine import node_major, resolve_node
+	from cookieguard_browser_use.demo.app import DemoApp
+
+	print(f'[smoke] python {sys.version.split()[0]} on {sys.platform}', flush=True)
+	print(f'[smoke] browser chosen by Browser Use: {LocalBrowserWatchdog._find_installed_browser_path()}', flush=True)
+	print(f'[smoke] node {resolve_node()} major {await node_major(resolve_node())}', flush=True)
+	with DemoApp('secure') as app:
+		session = BrowserSession(browser_profile=BrowserProfile(
+			headless=True, keep_alive=True, enable_default_extensions=False, user_data_dir=None))
+		try:
+			await bounded('BrowserSession.start', session.start(), 60)
+			version = await bounded('Browser.getVersion', session.cdp_client.send.Browser.getVersion(), 10)
+			print(f'[smoke] browser product: {version.get("product")}', flush=True)
+			audit = CookieGuardAudit(allowed_origins=[app.origin], auth_cookies=['session_id'])
+			tools = Tools()
+			audit.register_tools(tools)
+			await bounded('audit.start (engine + CDP observer)', audit.start(session), 30)
+			result = await bounded('navigate to demo login', tools.registry.execute_action(
+				'navigate', {'url': f'{app.origin}/login'}, browser_session=session), 30)
+			print(f'[smoke] navigate error: {result.error}', flush=True)
+			audit_result = await bounded('audit.finish', audit.finish(), 30)
+			print(f'[smoke] audit status without a login: {audit_result.status} (INCOMPLETE expected)', flush=True)
+		finally:
+			await bounded('BrowserSession.kill', session.kill(), 30)
+	print('[smoke] done', flush=True)
+	return 0
+
+
+if __name__ == '__main__':
+	sys.exit(asyncio.run(main()))