Agent Loops provides a robust framework for building autonomous agents within Claude Code, enabling them to reason, plan, and execute tasks independently. By implementing production-ready patterns like ReAct (Reason-Act), Plan-and-Execute, and self-correction loops, this skill allows developers to create sophisticated AI workflows that can recover from errors, manage their own memory context, and validate their own outputs for higher reliability in automated environments.
주요 기능
01ReAct (Reasoning + Acting) pattern implementation
02Plan-and-Execute multi-step task management
03Self-correction loops for automated output validation
04Sliding window memory management for context preservation
05Configurable safeguards to prevent infinite agent loops
0669 GitHub stars
