forge-mcp-tool-design

What problem does it solve?

This Skill solves unreliable MCP tool calling by giving you a disciplined, LLM-friendly way to design tool names, descriptions, JSON Schemas, responses, and errors so models select the right tool and produce parseable outputs.

Core Features & Use Cases

• LLM-first tool contracts: Enforces verb_object naming, third-person descriptions, and sibling-tool guidance so the model can infer when to call a tool.
• Strict JSON Schema hygiene: Requires per-parameter descriptions, uses enums for closed sets, sets additionalProperties: false for object types, and declares ISO 8601 formats explicitly.
• Deterministic responses & recoverable errors: Standardizes successful stable response shapes and structured error objects (code/message/retryable) so agents can retry or choose an alternate path.
• Safety for destructive actions: Requires dry_run support for mutation/deletion patterns to reduce high-risk behavior.

Quick Start

Describe the tool you want to add to your MCP server (its verb-object name, inputs, outputs, and failure modes), then write its tool schema so every parameter has a description, object schemas include additionalProperties: false, and errors return { code, message, retryable } in a consistent structured format.