ultimate-browsing

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The code intentionally includes utilities to locate, extract, decrypt, and inject local browser cookies (including OS keyring access and DPAPI decryption) and documents sending those cookies into a stealth browser session — functionality that enables credential theft and session impersonation if misused.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). SKILL.md routes user-provided URLs into python3 -m engine "<URL>", and at runtime the engine performs page content extraction/validation (Phase 1 curl_cffi probes and Phase 3 Playwright page.content()), so outsider-authored free text can be ingested via the target page they choose.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill documents runtime installation and invocation of two external tools whose upstream repositories/binaries are fetched and executed at runtime (CloakBrowser: https://github.com/CloakHQ/CloakBrowser and agent-browser: https://github.com/vercel-labs/agent-browser), and also references a raw install guide URL fetched at runtime (https://raw.githubusercontent.com/Panniantong/agent-reach/main/docs/install.md), so these external URLs represent runtime-fetched code/binaries the skill relies on.