hugging-face-paper-publisher

Pass

Audited by Gen Agent Trust Hub on Jun 21, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses a Python script (scripts/paper_manager.py) to perform operations. The script uses the official huggingface_hub library and standard Python packages to interact with the Hugging Face Hub and arXiv APIs.
  • [EXTERNAL_DOWNLOADS]: The skill fetches paper metadata from the well-known arXiv API (export.arxiv.org) and downloads/uploads README files to the Hugging Face Hub (huggingface.co). These are recognized and established services for the skill's intended purpose.
  • [CREDENTIALS_UNSAFE]: The skill requires an HF_TOKEN for write operations on the Hugging Face Hub. It correctly instructs users to set this via environment variables or a .env file and uses standard token retrieval methods (get_token()) rather than hardcoding credentials.
  • [PROMPT_INJECTION]: No patterns associated with prompt injection or agent instruction override were detected in the instructions or metadata.
  • [DATA_EXFILTRATION]: Network operations are restricted to communication with the Hugging Face Hub and arXiv. No unauthorized data exfiltration patterns were identified.
  • [DATA_EXPOSURE]: The skill implements a _sanitize_text method and an _escape_yaml_value helper to neutralize potential malicious markdown or YAML content received from external APIs before incorporating it into repository READMEs or local templates.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 21, 2026, 12:19 PM
Security Audit — agent-trust-hub — hugging-face-paper-publisher