The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for the agent to display to the user if a web search tool is not detected. These instructions recommend installing official Model Context Protocol (MCP) servers from trusted organizations and well-known services (e.g., Anthropic, Exa, Tavily, SerpAPI). The installation commands use npx for packages like @anthropic/brave-search-mcp and @exalabs/exa-mcp-server, which are recognized as safe and reputable sources.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from external websites to curate its resource map, which creates an inherent attack surface for indirect prompt injection where a malicious website could attempt to influence the agent's behavior.
Ingestion points: SKILL.md (via web_search, fetch, or curl tools to retrieve external documentation, papers, and blog posts).
Boundary markers: Absent; the skill does not explicitly instruct the agent to use delimiters or specific ignore-tags for the content retrieved from search results.
Capability inventory: The skill utilizes web_search, fetch, and curl capabilities across its search process.
Sanitization: The skill does not define specific sanitization or filtering logic for the ingested content, though it does emphasize 'rejecting low-signal results' and listicles.

finding-resources