The Agent Skills Directory

[SAFE]: The skill is designed for literature research and utilizes specialized MCP tools for querying arXiv and Semantic Scholar. No malicious patterns, obfuscation, or unauthorized access attempts were detected.
[PROMPT_INJECTION]: The skill ingests untrusted content from research papers and PDFs, which represents an indirect prompt injection surface. The risk is minimal as the skill does not possess capabilities for data exfiltration, local file modification, or arbitrary code execution.
Ingestion points: External data is fetched from research repositories via alphaxiv.get_paper_content and alphaxiv.answer_pdf_queries in SKILL.md.
Boundary markers: Not explicitly implemented within the provided instructions.
Capability inventory: Access is limited to research search and retrieval tools; no access to shell or file-writing tools.
Sanitization: No content sanitization is described for the retrieved paper text.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to retrieve research papers from well-known academic services.
Evidence: Fetches paper content and PDF data from arxiv.org and semanticscholar.org.

Literature Research