The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documents the use of npx -y to fetch and run MCP server packages directly from the NPM registry, specifically citing the @modelcontextprotocol/server-filesystem package in SKILL.md and server-types.md.
[REMOTE_CODE_EXECUTION]: By recommending the use of npx -y, the skill facilitates the execution of remote code at runtime. This is presented as a standard method for deploying MCP servers.
[COMMAND_EXECUTION]: The documentation provides multiple patterns for executing local processes. This includes the stdio server type for running commands (e.g., python -m my_mcp_server or custom scripts) and the headersHelper feature which executes local shell scripts (e.g., get-headers.sh) to dynamically generate authentication headers.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface through the processing of external tool outputs. Ingestion points: Data enters the agent's context through outputs from configured MCP tools, such as task search results or database queries mentioned in tool-usage.md and stdio-server.json. Boundary markers: The examples do not implement boundary markers or instructions to ignore embedded commands within the data returned by these tools. Capability inventory: The skill enables significant capabilities including subprocess execution via stdio and network operations across multiple protocols (sse, http, ws). Sanitization: No specific sanitization or validation of tool-provided data is discussed or demonstrated in the provided examples.

MCP Integration