What are Autonomous Agent Patterns?

They are architectural blueprints for creating AI systems that can independently execute tasks using tools while maintaining safety and operational reliability.

What is the 'Think-Decide-Act-Observe' loop?

It is a fundamental cycle where the agent reasons about a task (Think), chooses a tool (Decide), executes the action (Act), and processes the output (Observe) to decide the next step.

Does this skill provide sandboxing features?

It includes design patterns for sandboxed execution, featuring path validation and command allow-listing to ensure AI-generated code runs in a restricted environment.

Can I use multiple LLMs with these patterns?

Yes, the skill provides a multi-model architecture pattern that allows you to use different models for planning, reasoning, and specialized code generation.

How does the permission system protect my files?

The patterns include a multi-level approval manager that can be configured to ask for user permission before executing high-risk commands like file deletion or sudo operations.

Autonomous Agent Design Patterns

Name: Autonomous Agent Design Patterns
Author: claudiodearaujo

byclaudiodearaujo

•

生産性とワークフロー

Provides architectural blueprints and implementation patterns for building robust, safe, and autonomous AI coding agents.

This skill offers a comprehensive framework for developing autonomous agents, covering everything from core execution loops and multi-model orchestration to sophisticated tool design and safety protocols. It is particularly useful for developers building AI-driven CLI tools, coding assistants, or automated workflows that require reliable tool-calling capabilities with integrated human-in-the-loop permission systems and secure execution sandboxes inspired by industry leaders like Cline and OpenAI Codex.

主な機能

01Granular permission levels (Auto, Ask Once, Ask Each)

02JSON Schema-based tool definition and integration patterns

03Standardized Think-Decide-Act-Observe execution loops

041 GitHub stars

05Multi-model orchestration for optimized planning and reasoning

06Secure sandboxing and command validation logic

ユースケース

01Developing custom autonomous coding assistants

02Designing human-in-the-loop workflows for sensitive file operations

03Implementing safe tool-calling for LLM-based CLI utilities

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add claudiodearaujo/sistema-de-narra-o-de-livro autonomous-agent-patterns

For use in Claude.ai and ChatGPT

Download Skill