The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches research metadata from the official arXiv API at export.arxiv.org. It also interacts with Hugging Face Hub APIs to manage repository content. Both sources are well-known, established services appropriate for the skill's documented functionality.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources and persists it into repository context, creating a surface for indirect prompt injection.
Ingestion points: The get_arxiv_info function in scripts/paper_manager.py fetches paper titles, author lists, and abstracts from the arXiv API. The skill also accepts these values via command-line arguments (--title, --abstract, --citation).
Boundary markers: Content added to repository README files is enclosed within structural HTML comments:  and .
Capability inventory: The skill possesses network access via the requests library and write access to Hugging Face repositories via the huggingface_hub library. It also performs local filesystem writes when generating paper templates.
Sanitization: The skill implements a _sanitize_text method that removes control characters, collapses whitespace, and neutralizes markdown code-fence breakouts and YAML document delimiters to prevent accidental or malicious formatting disruption.

huggingface-paper-publisher