hugging-face-paper-publisher

Pass

Audited by Gen Agent Trust Hub on Sep 10, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes research paper metadata (titles, abstracts, and authors) fetched from the external arXiv API and user-provided inputs, which are then interpolated into markdown templates and Hugging Face repository README files.
  • Ingestion points: Metadata is retrieved from the arXiv API via PaperManager.get_arxiv_info and through CLI arguments like --title and --abstract in scripts/paper_manager.py.
  • Boundary markers: The skill uses HTML comment markers <!-- paper-manager:start --> and <!-- paper-manager:end --> to delimit generated content when updating repository READMEs.
  • Capability inventory: The skill performs file writes to the local filesystem, makes network requests to external APIs, and uses the huggingface_hub library to upload modified files (README updates) to user repositories.
  • Sanitization: The script implements _sanitize_text to neutralize markdown code fence breakouts and YAML document delimiters, and _escape_yaml_value to wrap strings in quotes and escape internal characters, reducing the risk of injection attacks from untrusted paper metadata.
  • [COMMAND_EXECUTION]: The skill's primary functionality is delivered through a CLI script (scripts/paper_manager.py) intended to be executed by the user via uv run or standard Python, involving various operations such as indexing, linking, and creating research articles.
  • [EXTERNAL_DOWNLOADS]: The skill fetches paper metadata from the well-known arXiv service (export.arxiv.org) and interacts with Hugging Face Hub (huggingface.co) to retrieve and update repository contents.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 10, 2026, 12:52 AM
Security Audit — agent-trust-hub — hugging-face-paper-publisher