A team creates a custom agent that should triage issues, inspect source files, and call a local MCP server for release-note metadata. A reviewer notices that the frontmatter defines an MCP server but also sets an empty tool list.

The correct answer is 1=Yes, 2=No, 3=Yes. 1. The empty tools list prevents the agent from using tools. This statement is true. The top-level tools: [] declaration explicitly tells the agent not to use any tools directly. While an MCP server can expose tools, the agent itself must be configured to use them. 2. The MCP server block alone makes its tools callable. This statement is false. Although the mcp-servers block defines a local MCP server with tools: [""], this only means the MCP server exposes all its tools. The agent's top-level tools: [] overrides this, preventing the agent from calling any tools, including those exposed by the MCP server. To enable tools from the MCP server, the agent's top-level tools list would need to include them (e.g., tools: ["release-notes/"]). 3. Adding selected tools can expose only required capabilities. This statement is true. By specifying individual tools or tool groups (e.g., tools: ["my-tool", "release-notes/get-metadata"]) in the agent's top-level tools list, you can precisely control which capabilities the agent has access to, promoting security and reducing unnecessary permissions.