arxiv-paper-writer

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill demonstrates safe behavior by utilizing local validation scripts and templates to manage the research paper workflow.
  • [EXTERNAL_DOWNLOADS]: The skill fetches metadata and BibTeX citations from official arXiv repositories (export.arxiv.org). These downloads are from a well-known service and are used for the skill's primary research purpose.
  • [COMMAND_EXECUTION]: The skill invokes local LaTeX compilation utilities (pdflatex, bibtex, latexmk) to generate PDFs. These executions are restricted to the project directory and use static command patterns.
  • [PROMPT_INJECTION]: Indirect prompt injection surface exists via ingestion of external paper metadata. 1. Ingestion points: scripts/arxiv_registry.py (arXiv API response). 2. Boundary markers: Absent. 3. Capability inventory: subprocess.run (LaTeX tools) and file writing. 4. Sanitization: Absent. The risk is considered minimal and inherent to the task of research assistance when using specialized academic sources.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 07:42 AM