The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from external sources (PubMed/nih.gov) through tools like pubmed_ncbi_nlm_nih_gov__get_suggestions. It also grants the agent significant capabilities, including unbrowse.scrape, unbrowse.run, and unbrowse.browse.act. There are no instructions or delimiters provided to ensure the agent sanitizes or ignores potential instructions embedded within the external data it retrieves.
Ingestion points: Data returned from nih.gov via the PubMed suggestion tool.
Boundary markers: Absent; no instructions are provided to the agent to distinguish between data and potential commands.
Capability inventory: The skill provides tools for scraping, arbitrary task execution on the target domain, and recorded browser interactions (unbrowse.scrape, unbrowse.run, unbrowse.browse.act).
Sanitization: Not present; the agent is instructed to report results from the tools without specific filtering.
[EXTERNAL_DOWNLOADS]: The skill operates by connecting to a remote MCP (Model Context Protocol) server at https://unbrowse.ai/mcp/nih.gov. This centralizes tool execution on the vendor's infrastructure.

nih