The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's scripts (doi_to_bibtex.py, extract_metadata.py, search_pubmed.py) communicate with reputable academic metadata providers including CrossRef, PubMed, arXiv, and DataCite. These interactions are standard for citation management and utilize established, well-known services.
[PROMPT_INJECTION]: The skill exhibits an attack surface for Indirect Prompt Injection (Category 8) through the ingestion of external data.
Ingestion points: The agent retrieves research metadata, including paper titles and abstracts, from external academic APIs using scripts/search_google_scholar.py, scripts/search_pubmed.py, and scripts/extract_metadata.py.
Boundary markers: The skill does not implement delimiters or isolation techniques to distinguish untrusted external metadata from the agent's internal instructions.
Capability inventory: The agent is equipped with powerful tools such as Bash and file system access (Edit, Read, Write), which could be misused if the agent inadvertently follows instructions embedded in fetched research papers.
Sanitization: There is no evidence of sanitization, escaping, or strict validation of the text content retrieved from external sources before it is processed or presented to the agent.

citation-management