atomic-agent-designer

Designs atomic agents via a structured two-flow workflow into an Agent Card.

Updated Apr 1, 2026
One-click install
npx skills add https://github.com/moky-me/nice --skill atomic-agent-designer
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: atomic-agent-designer
Source: https://github.com/moky-me/nice/tree/main/atomic-agent-designer
Command: npx skills add https://github.com/moky-me/nice --skill atomic-agent-designer

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Этот скилл помогает пользователю спроектировать атомарных агентов через структурированный процесс: от идеи до детальной спецификации (Agent Card), обеспечивая последовательность действий и управляемость.

Core Features & Use Cases

  • Пятифазная методология: MAP, GROUND, CONTRACT, VALIDATE, AGENT CARD — с двойным потоком.
  • Контракт-first: вход/выход определяются до описания процесса, с явными пределами.
  • Living Card: Kolb-журнал обновляется после реального использования, превращая карту в эволюционный артефакт. Пример: проектирование агента для автоматизации повторяющихся задач в процессе разработки.

Quick Start

Задайте тему или цель, и скилл проведёт через пять фаз, чтобы выдать Agent Card.

Frequently Asked Questions about atomic-agent-designer

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I design atomic agents through structured dialogue to get an actionable specification?

Designing atomic agents through structured dialogue uses a five-phase workflow to produce an actionable Agent Card. It guides you from idea to specification using MAP, GROUND, CONTRACT, VALIDATE, and AGENT CARD phases.

What is the contract-first approach for defining agent inputs and boundaries?

The contract-first approach defines inputs and outputs before describing the agent's process, establishing explicit boundaries. This ensures predictable behavior by enforcing constraints and one DIKW transition per agent upfront.

How do I map DIKW transformations and Kolb learning cycles when building agents?

Mapping DIKW transformations and Kolb learning cycles is supported through task-mode and process-mode workflows. These dual flows integrate real-world prompts to align learning cycles with data, information, knowledge, and wisdom transitions.

Does the atomic agent design methodology require validating examples before generating an Agent Card?

The atomic agent design methodology enforces three-example validation to ensure reliability. Validating examples before generating an Agent Card guarantees the agent handles expected inputs correctly within its defined boundaries.

What is a living Agent Card and how does it evolve after real-world use?

A living Agent Card is an evolutionary artifact that updates after real-world use via a Kolb journal. It tracks experiential learning to continuously refine the agent's specification based on actual performance and feedback.

Can I compose multiple atomic agents together if each handles one DIKW transition?

Composing multiple atomic agents is supported because each agent handles exactly one DIKW transition. This composable structure allows you to chain agents together to manage complex workflows while maintaining strict individual boundaries.