# Tools, skills and recipes: choose the right building block

Learn how callable capabilities, reusable workflow guidance and implementation recipes work together in an AI build.

Canonical: https://aroai.com/resources/mcp-tools-skills-recipes/

## Start with the task

A tool answers what an AI can call. A skill explains how to approach a repeatable job. A recipe shows how to connect the pieces in a particular implementation.

## Use tools for specific capabilities

Each tool has an input schema, output behavior and access boundary. Inspect the schema before sending a request, and review the returned data and source limitations.

## Use a skill to keep process consistent

Skills can guide when to call tools, how to sequence a job and what needs human review. Host delivery varies; a prompt, resource and native Agent Skill are separate paths.

## Use a recipe to get to a first result

A recipe explains prerequisites, sample inputs, exact tool dependencies, expected outcomes, errors and review. Keep client and server versions visible when a workflow depends on specific features.

## Example: New Zealand company research

The NZ company due diligence skill uses nzbn_search and nzbn_get_entity. The recipe shows how to identify a matching result, inspect source fields and report gaps. The tools provide the register operations; the skill supplies repeatable steps.