pinchtab-stealth-score

Run the stealth-score sweep, which navigates the browser through 15 public bot-detection sites like sannysoft, creepjs, and browserscan. An agent captures per-site metrics such as webdriver flags, trust scores, and fingerprint hashes into a structured JSON report.

Run the skill with the 'both' argument to execute sequential Docker-based sweeps for chrome and cloak. The summarizer then prints a divergent-metrics table showing only the rows where the two browsers disagree on real values.

The sweep covers 15 sites including sannysoft, rebrowser, iphey, whoer, browserscan, pixelscan, incolumitas, amiunique, browserleaks, creepjs, coveryourtracks, and fingerprint-demo. The list is defined in an editable index file, so sites can be added or removed without code changes.

Yes, Docker must be running because each browser benchmark executes inside an isolated PinchTab container bound to host port 9867. The up.sh script builds the image from current source and fails if the container does not become healthy.

A non-zero detection score on cloak that matches chrome's baseline reflects a generic Chromium signal, not a CloakBrowser regression. The results are advisory rather than a CI gate, so compare divergent metrics against the chrome baseline before concluding a problem.

Each browser sweep takes roughly 12-15 minutes once images are cached, so a both run takes about 25-30 minutes. First-time cloak image builds add approximately 10 minutes.