paper-fetch

SUSPICIOUS. The core OA-resolver behavior is coherent for a paper-fetching skill, but the default Sci-Hub fallback, mirror scraping, and optional stealth CloakBrowser bypass make the actual footprint broader and riskier than a normal scholarly fetcher. No confirmed malware or credential theft is shown, but external-host trust and browser-automation behavior create meaningful security risk.

The module is best characterized as a stealth/headless browser-based ‘challenge-aware’ remote fetcher that performs an in-page JavaScript evaluation to obtain base64-encoded content and streams the decoded bytes to stdout. It includes meaningful SSRF-style protections (blocks localhost/private/cloud-metadata IP ranges), but the key behavior is delegated to the unseen _FETCH_JS, which is a high-impact unknown. No explicit credential theft or direct system compromise is visible in the Python fragment, yet the challenge-bypass design and raw byte output make it plausibly risky for unauthorized content retrieval depending on how _FETCH_JS fetches/redirects and how callers use the output.