The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/generate_pdf.py executes system commands to handle document conversion.
Evidence: Calls subprocess.run with a list of arguments to invoke pandoc and xelatex. This method is safe and avoids shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The skill retrieves metadata from external academic resources.
Evidence: scripts/verify_citations.py connects to doi.org and api.crossref.org using the requests library. These are well-known, authoritative services for bibliographic data.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted content from external literature databases, which presents a surface for indirect prompt injection.
Ingestion points: Bibliographic data and abstracts are ingested from JSON search results and CrossRef API responses.
Boundary markers: The instructions do not mandate the use of delimiters to separate external text from agent instructions during synthesis.
Capability inventory: The agent can perform network requests and run document conversion utilities.
Sanitization: The skill does not explicitly sanitize the text of paper titles or abstracts retrieved from external sources before they are added to the review document.

literature-review