What is the primary focus of the Autonomous Agents skill?

The skill focuses on creating reliable, production-ready AI systems that can decompose goals and execute tools independently while minimizing the risk of compounding errors.

Is this skill meant for experimental or production use?

It is specifically designed for production use, emphasizing that 'autonomy is earned' and providing the guardrails necessary to move from impressive demos to consistent, real-world performance.

How does this skill prevent runaway AI costs or infinite loops?

It includes specific guidelines for setting hard cost limits, reducing step counts, and implementing robust API clients to ensure agents remain within defined constraints.

Which agent patterns does this skill support?

It provides implementation guidance for the ReAct pattern (alternating reasoning and action), the Plan-Execute pattern (separating planning from execution), and Reflection patterns for self-evaluation.

Reliable Autonomous Agents

Name: Reliable Autonomous Agents
Author: claudiodearaujo

byclaudiodearaujo

•

데이터 과학 및 ML

Architects reliable, self-correcting AI agent systems using proven patterns like ReAct and Plan-Execute to minimize error compounding in production.

This skill provides a comprehensive framework for building production-grade autonomous AI agents that can independently decompose complex goals, execute tool-based workflows, and self-correct without constant human oversight. It moves beyond brittle demos by focusing on the 'math of autonomy'—ensuring that compounding failure rates are mitigated through rigorous guardrails, reflection patterns, and cost-limiting strategies. Whether you are implementing ReAct loops or Plan-Execute architectures, this skill provides the architectural discipline needed to turn experimental agents into reliable software components.

주요 기능

01Implementation of ReAct (Reasoning + Action) and Plan-Execute patterns

02Predefined solutions for common agent failure modes and cost control

03Production-grade reliability guardrails and error-handling strategies

04Structured logging and context usage tracking for observability

05Goal decomposition and iterative reflection for self-correction

061 GitHub stars

사용 사례

01Building long-running task automation systems that require multi-step reasoning

02Orchestrating complex tool-calling sequences with built-in cost and step limits

03Developing self-healing workflows that validate agent outputs against ground truth

What are Skills?·How to Install

Install with 🐟 Skill.Fish

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

For use in Claude.ai and ChatGPT

Download Skill