moai-cc-agents

Sub-agents are specialized Claude Code instances designed for focused tasks like code review or debugging. Create them by defining agent personas, setting proactive triggers, and configuring tool permissions in standardized agent files stored under .claude/agents. This enables autonomous task execution with structured handoffs.

Orchestrate agents by designing distinct personas for each task domain, establishing trigger patterns that activate agents autonomously, and implementing least-privilege tool access controls. Use standardized agent-file structures to coordinate handoffs between agents and maintain governance over execution modes and permissions.

Proactive triggers activate agents automatically based on defined conditions. Configure trigger patterns in your agent configuration to specify when agents execute for code quality checks, debugging, or architecture tasks. Combine triggers with autonomy rules and collaboration guidelines to balance automated execution with controlled handoffs.

Yes. Apply least-privilege tool access by explicitly defining which tools each agent can use in its configuration. This security model ensures agents operate only within their intended scope, with structured handoffs to other agents when tasks require elevated permissions or cross-domain expertise.

Agent design requires specifying functional and technical requirements including agent templates, trigger patterns, autonomy levels, collaboration rules, and tool-permission controls. Follow governance checklists to document execution modes, handoff protocols, and security boundaries for each agent in your .claude/agents directory.

This Skill targets users building specialized agents and assumes familiarity with Claude Code. While not prerequisite, understanding your automation goals—such as code review, debugging, or domain-specific work—helps frame agent personas and triggers effectively from the start.