What problem does it solve? When an AI agent lacks a needed capability, it is easy to pick the wrong extension mechanism or create duplicates of tools that already exist. This Skill guides you to the right approach—Forged Tool, MCP server, run_script chaining, or a skill—and walks through creating and verifying each one correctly. ## Core Features & Use Cases - Forged Tool Creation: Build a new capability as a folder containing a tool.json manifest and a backing script, with a shell command template using {{slots}} and a JSON Schema for arguments. - MCP Server Integration: Connect external MCP servers with mcp_add, activate them with reload_mcp, and browse their resources via list_mcp_resources and read_mcp_resource. - Tool Chaining: Run sequences of existing tools in a single turn with run_script, injecting step outputs into later arguments via {{N}} placeholders. - Use Case: You need a weather lookup capability that does not exist. After confirming with tool_search that nothing similar is registered, you create a Forged Tool with a Python script, verify it through tool_call, and it becomes callable immediately in the same mission. ## Quick Start Create a new tool that runs a Python script to fetch weather data for a given city, then verify it works by calling it with a real city name.
