quaternius-assets

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill performs network requests to quaternius.com and potentially user-provided URLs to retrieve asset metadata using urllib.request. While the primary purpose is legitimate asset discovery, the inspect tool can be directed to arbitrary URLs if the agent is manipulated.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests and parses HTML content from external web pages, returning this data to the agent context.
  • Ingestion points: Metadata extraction in scripts/_quaternius.py via the search and inspect functions.
  • Boundary markers: None implemented; results are returned as raw dictionaries without delimiters or instruction-bypass warnings.
  • Capability inventory: Network operations via urllib.request.urlopen.
  • Sanitization: Employs html.unescape and regular expressions to strip tags from the fetched HTML.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 01:31 AM