novel-assistant

This module is best characterized as a local Chrome automation/CDP bootstrapper that intentionally clones sensitive Chrome authentication/session database files into a dedicated debug profile and then launches Chrome with CDP enabled (including --remote-allow-origins=*). While the file itself shows no outbound exfiltration or overt malware behavior, the combination of (a) sensitive credential material handling and (b) expanded access to the debugging surface makes it high-risk if misused, executed in an untrusted context, or if other processes/users can access the debug directory or CDP port. Treat as security-sensitive tooling rather than a benign helper.

The provided module appears to be a legitimate CDP-based web scraping/markdown-generation tool, but it contains a notable security red flag: it performs dynamic JavaScript execution via evalJSON/buildDetailJS (with explicit eval-style logic) using parameters derived from scraped/remote page data. While there is no clear evidence of overt malware (exfiltration/backdoors/mining) in this fragment, the eval-driven injection surface and the additional local artifact builder execution increase supply-chain risk. Further review of evalJSON/buildDetailJS parameter handling and escaping is recommended to assess exploitability.

SUSPICIOUS: the skill is coherent as a browser CDP helper, but its footprint is high-risk because it intentionally reuses authenticated sessions and extracts cookies/tokens through a third-party browser-control CLI. Install trust is acceptable, but credential/session access is broader than a typical scraping helper and warrants caution.

This module is a high-abuse automation helper. Its most significant risk is the combination of host-level execSync with dynamically constructed command strings and a direct CDP 'eval' pathway that executes caller-supplied JavaScript in a controlled browser context. While the snippet contains no explicit malware/exfiltration code by itself, it provides powerful primitives that can be used maliciously if any upstream inputs (port/js/args) are attacker-controlled or insufficiently constrained. Treat as security-sensitive and require strict input validation and safe invocation patterns (e.g., avoiding string-based execSync, using argument arrays, and allowlisting commands/js).