Playwright Stealth Verify measures whether an automated browser presents a coherent, internally consistent fingerprint without modifying or evading detection.
The fastest way to install a skill directly from the registry.
npx clawhub@latest install playwright-stealth-verify
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
Copy this prompt to OpenClaw to install it automatically.
Help me install playwright-stealth-verify using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Get the raw skill files in a ZIP archive.
Playwright Stealth Verify is an Openclaw Skills browser automation utility for testing Playwright, Puppeteer, Selenium, and CDP-driven browsers with the liarjs library. It evaluates an existing Page object or attaches to a browser through Chrome DevTools Protocol to identify inconsistencies across JavaScript, browser APIs, workers, GPU signals, codecs, and network metadata.
The skill measures a browser rather than changing it: it ships no evasions, stealth profiles, or fingerprint modifications. Use the results to validate launch configurations, evaluate stealth-plugin effects, investigate headless behavior, and add fingerprint-quality assertions to automated test suites.
Node.js 22 or newer is required. Install liarjs as a development dependency so its version is recorded in the lockfile:
npm install --save-dev liarjs
Use it with an existing Playwright or Puppeteer Page:
import { checkPage } from "liarjs";
const result = await checkPage(page);
expect(result.score).toBeGreaterThanOrEqual(85);
const critical = result.checks.filter((c) => c.status === "bad");
expect(critical).toHaveLength(0);
For a browser already listening on port 9222:
npx [email protected] --cdp http://127.0.0.1:9222
Example with a locally launched Chromium process:
./chrome --remote-debugging-port=9222 &
npx [email protected] --cdp http://127.0.0.1:9222
Confirm ownership of the CDP endpoint before attaching. When testing a launch configuration rather than a specific running browser, prefer the default CLI behavior, which launches a throwaway profile in a temporary directory and deletes it afterward.
Use offline mode for the 32 JavaScript-layer checks without outbound requests:
npx [email protected] --offline
Without offline mode, the browser requests the liarjs network endpoint. Use endpoint to route that request to an approved deployment of the Worker, and use page only with a page the user owns. Do not navigate to third-party sites as part of a scan.
| Field | Description |
|---|---|
score |
Aggregate fingerprint coherence score used for threshold assertions. |
label |
Human-readable interpretation of the scan result. |
checks[] |
Individual checks with identifiers, statuses, and diagnostic details. |
client |
Raw client-side fingerprint collected from the browser. |
server |
Raw edge or network-layer view when online measurement is enabled. |
meta.schema |
Payload schema version. |
| Check ID | Detects | Maximum deduction |
|---|---|---|
webdriver |
navigator.webdriver remaining enabled |
40 |
native-integrity |
Injected APIs that no longer expose native-code identity | 35 |
headless-ua |
HeadlessChrome in the user agent |
30 |
worker-consistency |
Main-thread overrides missing from Web Workers | 20 |
headless-viewport |
outerHeight === innerHeight and absent browser UI |
10 |
gpu-triad |
Conflicting WebGL and WebGPU GPU names | 22 |
chrome-object |
Chrome user agent without window.chrome |
12 |
codecs |
Chrome-claiming Chromium unable to play H.264 | 6 |
The complete catalog contains 40 checks and is documented in the browser-fingerprint-audit skill reference. Online scans add network observations such as IP, ASN, HTTP version, TLS version, ClientHello shape, and headers; offline scans restrict measurement to the JavaScript layer.
Loading
A multi-source A-share decision-support skill for screening Shanghai and Shenzhen main-board stocks, analyzing market events, optimizing portfolios, and enforcing trading discipline.

A read-only Openclaw Skills integration for inventorying Huawei Cloud CCE clusters by name, ID, status, version, and flavor.

A local geospatial analysis skill that maps five-level burn severity with dNBR and tracks vegetation recovery from post-fire NDVI time series.

Amazon Listing Evidence Audit evaluates listing readiness and turns unsupported claims, unclear content, keyword issues, and missing visual proof into a prioritized action plan.

DataQuant Connector gives AI agents a secure REST API workflow for retrieving multi-market financial data, screening securities, and supporting quantitative research.

Christa is an adaptive body-shaping coach that assesses readiness, answers evidence-based nutrition questions, and generates a personalized HTML fitness report.








































