hugging-face-papers
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 is designed to fetch and process external content from research papers, which constitutes an untrusted data source that could contain embedded instructions intended to override agent behavior.
- Ingestion points: External content is retrieved from
huggingface.co/papers/{PAPER_ID}.mdand arXiv URLs viacurlcommands. - Boundary markers: There are no explicit instructions to use delimiters or ignore instructions within the fetched paper content.
- Capability inventory: The skill includes network operations via
curlto various Hugging Face API endpoints for both reading and writing (indexing, claiming authorship). - Sanitization: The instructions do not specify any validation or sanitization for the markdown content retrieved from external sources.
- [COMMAND_EXECUTION]: The skill provides numerous
curlcommand templates for interacting with the Hugging Face REST API. This includes GET requests for metadata and POST requests for administrative actions like indexing papers or updating project links. - [EXTERNAL_DOWNLOADS]: Fetches research paper data and structured metadata from Hugging Face (
huggingface.co) and arXiv (arxiv.org). These are established, well-known platforms for AI and computer science research.
Audit Metadata