browsing

SUSPICIOUS. The skill’s capabilities mostly fit its browsing purpose, but its footprint is high-risk: stealth/WAF bypass, third-party content routing, cookie extraction/injection, and autonomous interaction with real logged-in sessions. Official tool provenance reduces malware confidence, yet the combination is disproportionate for a generic browsing router and poses significant account/privacy risk.

This fragment describes an automated browsing/fetch skill whose core behavior is explicitly oriented toward detecting WAF/anti-bot challenges and attempting circumvention via TLS fingerprint impersonation and browser/network reconnaissance. It additionally documents an endpoint-discovery loop that identifies internal JSON/GraphQL APIs from captured browser traffic and then repeatedly calls them for pagination/collection—an abuse pattern consistent with scraping and access-control bypass. No direct payload-style malware (credential theft, persistence, or explicit exfiltration) is shown here, but the documented operational capability and runtime auto-installation of third-party tooling (including patch/stealth-like Playwright adaptations) make the overall security risk high and warrant deeper review of the actual engine implementation and dependencies.

No direct evidence of traditional malware (RCE/backdoor/exfiltration) is present in the provided fragment. However, the content explicitly enables WAF/anti-bot evasion by impersonating browser TLS/HTTP fingerprints and—optionally—obtaining and injecting clearance cookies into curl_cffi to access protected endpoints. This capability is high-risk for unauthorized automation. Additionally, the runtime pip auto-install approach increases supply-chain exposure if versions/integrity are not pinned and verified.

No overt backdoor or explicit malicious networking is present in the shown code, but the module is a high-sensitivity browser data harvester. It captures full rendered HTML and—most importantly—raw cookie values from a persistent Chromium profile and outputs them in plaintext JSON to stdout. If used against authenticated targets or with retained session cookies, it can facilitate session/cookie theft or covert scraping via the caller’s pipeline.

This fragment documents a high-risk capability: extracting sensitive Google/YouTube authentication/session cookies from local browser storage (including decrypting Chromium cookies using macOS Keychain) and injecting them into an automated browser session via CDP to reuse an authenticated state. Even without the implementation code, the described cookie selection, decryption/key access, token export, and CDP injection workflow are strongly consistent with session hijacking/credential reuse and warrant urgent review and restrictive use controls.

This code is primarily an automation/scraping-style collector, not a self-propagating malware module. However, it performs high-risk privacy/security operations: it navigates to an arbitrary URL without allowlisting and it captures and exports sensitive browser state (full HTML and cookies, including name/value/domain) from a persistent profile provided by the caller. If the profile contains authenticated/session cookies, the stdout envelope can effectively enable session hijacking or sensitive data exfiltration by anyone who can access the script’s output or logs. The use of a stealth plugin and persistent profile path increases the likelihood of intentional evasion and session reuse.