Delegates complex tasks to specialized AI assistants with isolated contexts and granular tool permissions within the Claude Code CLI.
The Sub-Agents skill empowers developers to build and orchestrate a team of specialized AI assistants directly within Claude Code. By creating agents with distinct system prompts and specific tool access, users can isolate complex tasks like debugging, code reviews, or data analysis into separate context windows. This prevents main conversation bloat and significantly improves the accuracy of autonomous workflows. Sub-agents can be scoped to individual projects or shared globally, allowing for a highly customizable and reusable AI-driven development environment.
주요 기능
010 GitHub stars
02Standardized YAML-based agent configuration
03Isolated context windows for specialized task execution
04Project and user-level configuration scopes
05Automatic task delegation based on agent descriptions
06Granular tool permission control per agent
사용 사례
01Automating multi-step code reviews with a dedicated expert reviewer agent
02Isolating root-cause analysis for complex bugs to keep the main chat history clean
03Creating a read-only documentation assistant to fetch and analyze external APIs