read-arxiv-paper
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external LaTeX files from arXiv, which creates a surface for indirect prompt injection if those files contain instructions for the AI. Ingestion point: paper source files mentioned in SKILL.md. Boundary markers: The instructions lack delimiters or warnings to ignore instructions embedded within the paper source. Capability inventory: The skill performs network fetches, extracts archives, and writes local files. Sanitization: No sanitization of the LaTeX content is specified before processing.
- [EXTERNAL_DOWNLOADS]: Fetches paper source files from arXiv, which is a well-known repository for scientific preprints.
- [COMMAND_EXECUTION]: Performs file system tasks including directory creation, extracting tarball archives, and reading/writing to local paths.
Audit Metadata