The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize large volumes of external data from sources like ArXiv, PubMed, and WebSearch. This creates a surface for indirect prompt injection where malicious instructions could be hidden inside an academic paper's abstract or text.
Ingestion points: Untrusted data enters via mcp__arxiv-mcp-server__read_paper, mcp__pubmed-mcp-server__pubmed_search_articles, mcp__zotero__zotero_get_item_fulltext, and WebFetch.
Boundary markers: The instructions do not provide explicit delimiters or warnings for the agent to ignore directives found within the research material.
Capability inventory: The skill has access to Bash, Write, Edit, and WebFetch, which could be abused if an injection is successful.
Sanitization: No sanitization or filtering of the external literature content is defined in the workflow.
[EXTERNAL_DOWNLOADS]: The file references/MCP_SETUP.md provides configuration instructions for installing three external MCP servers (arxiv-mcp-server, pubmedmcp, and zotero-mcp) from individual community repositories using the uvx package runner.
[COMMAND_EXECUTION]: The skill is configured with Bash in the allowed-tools section of its frontmatter, granting it the capability to execute shell commands in the user's environment to support research and writing tasks.

medical-imaging-review