Can I customize the completion logic of the generated agent?

Yes. The scaffolder supports types like iterationBudget, keywordSignal, and stepMachine. More advanced types like composite or custom can be manually edited in the agent.json after scaffolding.

How do I run the scaffolding command?

The skill triggers when you ask to 'create agent' or 'scaffold agent', executing a Deno script that asks for the agent name, description, and completion type.

What files does the scaffolder create?

It generates a .agent/{name}/ directory containing agent.json, steps_registry.json, a schemas folder for validation, and a prompts folder with templates for different workflow steps.

What is the intentSchemaRef used for?

It is a JSON pointer used by the Runner to identify the next action from the agent's structured output, ensuring state transitions happen correctly.

What is a Climpt Agent?

A Climpt Agent is a structured AI workflow defined by an agent.json file and a set of prompts that perform specific tasks using the Climpt framework.

Climpt Agent Scaffolder

Name: Climpt Agent Scaffolder
Author: tettuan

bytettuan

0•

数据科学与机器学习

Automates the creation of standardized Climpt agent directories and configuration files to accelerate AI agent development.

The Climpt Agent Scaffolder is a specialized tool for developers working with the Climpt framework to build custom AI agents. It eliminates manual setup by generating a complete, standardized directory structure including essential files like agent.json, steps_registry.json, and structured prompt templates for various workflow phases (initialization, continuation, and verification). By supporting multiple completion types—such as iteration budgets, keyword signals, and complex step machines—this skill ensures that your agents are built on a robust, production-ready foundation with correctly mapped schemas and intents from the start.

主要功能

01Automatic generation of .agent/ directory structures

02Pre-configured boilerplate for agent.json and steps_registry.json

03Built-in guidance for complex intentSchemaRef and JSON schema pointers

04Support for 4+ completion logic types including Step Machines and Budgets

05Standardized prompt templates for different agent lifecycle stages

060 GitHub stars

使用场景

01Creating complex multi-step agents with specific termination and verification logic

02Rapidly bootstrapping a new AI-powered code reviewer or issue tracker

03Standardizing agent structures across a team using the Climpt framework

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add tettuan/climpt agent-scaffolder

For use in Claude.ai and ChatGPT

Download Skill