The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of MCP servers from various remote sources, including npm packages, GitHub repositories, and direct URLs.
It utilizes npx add-mcp, a tool provided by Neon (a trusted organization), to streamline multi-agent configuration.
It provides instructions for downloading and running official servers from trusted organizations such as GitHub, Cloudflare, MongoDB, and Vercel.
It guides the user to the official MCP Registry at registry.modelcontextprotocol.io to discover third-party servers.
[COMMAND_EXECUTION]: The skill uses command-line interfaces (claude mcp, npx, docker) to install, configure, and execute external server code.
SKILL.md and references/multi-agent.md detail the use of npx add-mcp and npx -y for installing servers as stdio transports.
references/search.md describes the use of docker run as a deployment option for isolated server environments.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection through the ingestion of untrusted external data.
Ingestion points: references/search.md details querying the MCP Registry API (registry.modelcontextprotocol.io/v0.1/servers) and performing web searches for server details.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are present when processing registry search results.
Capability inventory: The skill has the capability to write to configuration files (e.g., ~/.claude.json, .vscode/mcp.json) and execute installation commands (npx, claude mcp).
Sanitization: There is no evidence of sanitization for fields returned by the registry API (e.g., server.name, server.description) before they are presented to the agent or used in installation commands.
[CREDENTIALS_UNSAFE]: The skill manages sensitive API keys and tokens required for various MCP servers.
It correctly advises using environment variables (${VAR}) in configuration files rather than hardcoding secrets.
It provides guidance on editing config files like ~/.claude.json to securely add credentials that are difficult to pass via CLI flags.

mcp-management