arxiv-paper-writer

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts/compile_paper.py script executes system commands to compile LaTeX documents using tools like pdflatex, bibtex, or latexmk. Additionally, scripts/bootstrap_ieee_review_paper.py uses subprocess calls to run other internal scripts within the skill's directory to manage paper plans and issues.
  • [EXTERNAL_DOWNLOADS]: The scripts/arxiv_registry.py script makes HTTP requests to the arXiv API (export.arxiv.org) and arxiv.org to fetch paper metadata and BibTeX citations. These are established academic platforms, and the script handles the data retrieval using standard libraries without suspicious overhead.
  • [PROMPT_INJECTION]: The skill ingests untrusted data from the arXiv API (titles, abstracts, and author names) to generate review content and citations. Evidence chain: (1) Ingestion point: Atom XML metadata fetched in scripts/arxiv_registry.py. (2) Boundary markers: The skill includes instructions to verify every citation and follow a gated workflow with human approval. (3) Capability inventory: Subprocess calls for document compilation and project management. (4) Sanitization: The escape_bibtex_text function in scripts/arxiv_registry.py sanitizes special characters for BibTeX entries.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 02:44 AM
Security Audit — agent-trust-hub — arxiv-paper-writer