The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documents fetching tool configurations and connecting to MCP servers via HTTP/SSE. Examples include well-known services like GitHub and Linear (e.g., https://mcp.github.com/mcp, https://mcp.linear.app/mcp).
[COMMAND_EXECUTION]: The documentation includes patterns for using stdioTransport via Node.js child_process, specifically demonstrating the use of npx to launch MCP servers locally.
[REMOTE_CODE_EXECUTION]: The skill explains how to integrate provider-managed sandbox environments, such as Anthropic's code_execution tool and OpenAI's shell tool, which allow the model to execute code within secure, remote sandboxes.
[DATA_EXFILTRATION]: Provides patterns for tools that search databases (db.products.search) and retrieve user information (db.users.findUnique). These are presented as standard functional examples for tool-calling capabilities.
[SAFE]: The skill mandates the use of Zod schemas (inputSchema, outputSchema) for all tool definitions, which provides strong runtime validation and type safety for data entering or leaving the agent's context.

tanstack-ai-core-tool-calling