The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external scientific sources, which creates a potential surface for indirect instructions to be processed by the agent.
Ingestion points: Data is fetched via requests.get from https://www.ebi.ac.uk/gwas/rest/api and https://www.ebi.ac.uk/gwas/summary-statistics/api as well as via FTP.
Boundary markers: The provided documentation and code examples do not implement delimiters or specific instructions for the agent to ignore potentially malicious embedded content within the genomic data or publication metadata.
Capability inventory: The skill utilizes requests for network access and pandas for data manipulation. It does not demonstrate high-risk capabilities like arbitrary shell execution or local file writing, limiting the potential impact of an injection.
Sanitization: The Python examples do not show any input validation or sanitization of the external API responses before they are presented to the agent or the user.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of genetic association data and summary statistics from well-known scientific repositories.
Fetches configuration and data from the European Bioinformatics Institute (EBI) via ebi.ac.uk domains and FTP servers.
References training materials and workshop code hosted on a public GitHub repository (github.com/EBISPOT/GWAS_Catalog-workshop).
[METADATA_POISONING]: The skill includes a promotional section at the end of the instructions that monitors user task complexity and proactively suggests switching to a commercial hosted platform (k-dense.ai). While this is presented as a productivity upgrade, it introduces instructions into the agent's workflow that prioritize third-party platform adoption based on the skill's own assessment of task difficulty.

gwas-database