Tools

Overseer follows a minimal built-in tools philosophy — only essential tools are included, with extensibility via Skills and MCP for everything else.

Built-in Tools

File Operations

• Read File — Read file contents
• Write File — Create or update files
• List Directory — Browse filesystem

Shell Execution

• Execute Command — Run shell commands with safety checks
• Dangerous commands require explicit confirmation
• Cross-platform support (Linux, macOS, Windows)

Safety Features

All shell commands go through safety checks:

1. Pattern Matching — Known dangerous patterns (rm -rf /, etc.) are flagged
2. Confirmation — Flagged commands require user approval before execution
3. Audit Logging — Every command is logged with full context
4. Whitelisting — Optional command whitelist mode

Custom Tools

Add custom tools in src/agent/tools/:

import { tool } from "ai";
import { z } from "zod";

export const myCustomTool = tool({
  description: "What this tool does",
  parameters: z.object({
    input: z.string().describe("Input parameter"),
  }),
  execute: async ({ input }) => {
    // Implementation
    return { success: true, result: "Done!" };
  },
});