The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches metadata, PDF documents, and TeX source archives from official arXiv.org domains, which are well-known services for academic research.
[COMMAND_EXECUTION]: The skill executes a local Python script (scripts/bootstrap_arxiv_paper.py) to automate workspace creation, download assets, and generate report skeletons based on research paper metadata.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted research paper content from external sources, which creates a potential surface for indirect prompt injection attacks where paper text could attempt to influence the agent's behavior.
Ingestion points: Research paper URLs, PDFs, and TeX source files downloaded from arXiv as referenced in SKILL.md and the bootstrap script.
Boundary markers: The skill uses a structured markdown template (references/report-outline.md) to define the analysis scope, although it lacks explicit instruction-ignore delimiters for the ingested paper text.
Capability inventory: The skill possesses file-system write capabilities (creating reports, metadata, and folders) and network access (to arXiv.org) via the Python bootstrap script.
Sanitization: The bootstrap script performs sanitization of paper titles for use in folder and file names and includes a safety check in the maybe_extract_source function to prevent path traversal (ZipSlip) during archive extraction.

paper-interpreter