The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill interacts with reputable academic and scientific web services, including NCBI (PubMed/PMC), arXiv, Crossref, and Semantic Scholar. These network operations are necessary for the skill's functionality and target trusted domain names.\n- [COMMAND_EXECUTION]: The instructions suggest the use of platform-native tools or curl to perform standard REST API requests. This is a conventional and appropriate method for retrieving data from web-based services.\n- [SAFE]: Credential management is handled securely by instructing the agent to look for API keys in environment variables or local .env files, avoiding the risk of hardcoded secrets.\n- [PROMPT_INJECTION]: The skill identifies an indirect prompt injection surface as it processes external research data such as titles and abstracts.\n
Ingestion points: Metadata and text returned from scholarly APIs described in the references/ directory.\n
Boundary markers: None specified for the returned API content.\n
Capability inventory: Network requests via curl and WebFetch to reputable domains.\n
Sanitization: Not explicitly implemented in the instructions.\n Despite the lack of sanitization, the risk is negligible due to the authoritative and academic nature of the source databases.

paper-lookup