read-arxiv-paper

Pass

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Downloads PDF files and images from arxiv.org, which is a well-known and trusted academic repository. These downloads are required for the skill's primary function of paper analysis.
  • [COMMAND_EXECUTION]: Executes shell commands such as mkdir and curl to manage files within the user-specified Obsidian vault. These operations are restricted to the directory provided in the environment configuration.
  • [PROMPT_INJECTION]: The skill ingests and processes full text from external ArXiv papers, presenting a surface for indirect prompt injection where malicious instructions in a paper could influence the agent's output.
  • Ingestion points: Full paper content extracted from PDF and HTML sources at arxiv.org.
  • Boundary markers: No delimiters are used to separate paper content from agent instructions, and there are no warnings to ignore embedded commands.
  • Capability inventory: File system writing and directory management, plus network access via curl.
  • Sanitization: The skill does not perform any sanitization of the text extracted from papers before summarizing.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 4, 2026, 09:06 AM