How does the agent handle security and dangerous commands?

The generated harness includes a security.py module where you can define allowed and blocked commands, with three configurable security levels (High, Medium, and Low) to control the agent's permissions.

What is a long-running agent harness?

A long-running agent harness is a structured environment that allows Claude to operate autonomously over extended periods, managing its own state, executing commands, and working through a task list without constant human intervention.

How does it track progress during autonomous runs?

It uses a feature_list.json file to define specific tasks and a claude-progress.txt file to log activity, ensuring the agent can resume exactly where it left off if the session is interrupted.

Does this skill support both Python and TypeScript?

Yes, the generator provides templates and initialization logic for both Python (using uv/pytest) and TypeScript (using npm/jest/Playwright) environments.

Nonstop Agent Generator

Name: Nonstop Agent Generator
Author: seolcoding

byseolcoding

•

Productivity & Workflow

Generates long-running autonomous coding agents based on Anthropic best practices for persistent, 24/7 project development.

The Nonstop Agent Generator is a specialized capability for Claude Code designed to scaffold and manage persistent, autonomous agent harnesses. It automates the complex process of setting up agents that can work independently over long periods, featuring interactive requirement gathering, deep analysis of existing codebases (including automated test execution), and the generation of structured Python or TypeScript environments. By implementing the '2-Agent' pattern and using standardized progress tracking through feature lists, this skill ensures that AI agents remain focused, secure, and capable of incremental progress without losing state.

Key Features

01Scaffolding for robust Python and TypeScript agent harnesses

02Interactive multi-step requirement gathering via AskUserQuestion

03Standardized progress tracking with feature_list.json and git persistence

04Automated project analysis including dependency checks and test execution

058 GitHub stars

06Dual-agent pattern implementation (Initializer and Coding agents)

Use Cases

01Creating specialized research or automation agents using the Claude Agent SDK

02Bootstrapping a new autonomous agent to build a project from scratch

03Transforming an existing codebase into an automated environment for 24/7 bug fixing

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add seolcoding/nonstop-agent skills

For use in Claude.ai and ChatGPT

Download Skill