ljg-paper-river

Pass

Audited by Gen Agent Trust Hub on Apr 11, 2026

Risk Level: SAFEPROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill processes untrusted research papers from external sources, creating an indirect prompt injection surface.
  • Ingestion points: External research papers fetched via academic URLs, web searches, or read from PDF files.
  • Boundary markers: The instructions do not define boundary markers or explicit warnings to the agent to ignore embedded instructions within the ingested paper text.
  • Capability inventory: The agent uses network tools for fetching content and search, and has write access to the local file system for saving notes.
  • Sanitization: There is no evidence of sanitization or filtering of the ingested external content.
  • [DATA_EXFILTRATION]: The skill fetches research paper data from well-known academic sources and the web to perform analysis. These operations involve network access to external domains as part of the primary research functionality.
  • [COMMAND_EXECUTION]: The skill instructions specify the use of system utilities like the date command to generate unique identifiers and timestamps for document organization and file naming.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 11, 2026, 01:50 AM