The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process large volumes of untrusted data from external academic databases, including abstracts and full-text articles. This creates a surface for indirect prompt injection, where malicious instructions could be embedded in scholarly content to influence the agent's future behavior. The skill instructions do not specify boundary markers or sanitization procedures for the retrieved content.
Ingestion points: Data is retrieved from PubMed, PMC, bioRxiv, medRxiv, arXiv, OpenAlex, Crossref, Semantic Scholar, CORE, and Unpaywall.
Boundary markers: Absent; the skill instructions direct the agent to return raw JSON or XML responses to the user without explicit delimitation.
Capability inventory: The skill uses platform-specific fetch tools or curl to perform network operations. It does not instruct the execution of retrieved data.
Sanitization: None provided for the fetched paper content.
[EXTERNAL_DOWNLOADS]: The skill references and fetches data from several well-known academic and scientific services, including NCBI (eutils.ncbi.nlm.nih.gov), arXiv (export.arxiv.org), Crossref (api.crossref.org), and Semantic Scholar (api.semanticscholar.org). These interactions are essential to the skill's primary research function.
[COMMAND_EXECUTION]: The skill instructs the agent to use curl as a fallback method for performing HTTP requests if native fetch tools like WebFetch or read_url_content are unavailable. This is a standard practice for network-enabled skills in CLI-based agent environments.

paper-lookup