What happens if a task fails during orchestration?

The orchestrator includes an automated review and consolidation loop that can enter a 'fix_required' state to address issues before completing the workflow.

Can I run multiple tasks in parallel?

Yes. The orchestrator uses a file manifest system; tasks with non-overlapping 'writes' are executed in parallel to optimize performance.

What is a Kiro spec in this context?

A Kiro spec is a structured configuration file that defines the tasks, requirements, and dependencies for a feature, serving as the blueprint for the orchestrator.

How does the skill distinguish between frontend and backend tasks?

The orchestrator analyzes task semantics: 'code' tasks like APIs and database logic are assigned to Codex, while 'ui' tasks like React components and CSS are assigned to Gemini.

Does this skill require manual input between steps?

No. Aside from an initial mandatory confirmation, the skill is designed to complete the entire orchestration loop automatically until all tasks are finished.

Multi-Agent Orchestrator

Name: Multi-Agent Orchestrator
Author: PeterFile

byPeterFile

•

생산성 및 워크플로

Orchestrates complex multi-agent workflows to automate software development tasks from Kiro specifications.

소개

The Multi-Agent Orchestrator is a sophisticated coordination layer for Claude Code that automates software development by managing specialized AI agents. By parsing Kiro specifications, it intelligently assigns backend tasks to Codex and UI tasks to Gemini, handling the entire implementation lifecycle including parallel execution, automated code reviews, and state synchronization. It is designed for developers who want to move from manual task execution to high-level project steering, providing a fully automated loop that maintains project state and pulse updates throughout the build process.

주요 기능

Specialized agent assignment for backend (Codex) and frontend (Gemini) tasks
Intelligent parallel execution based on file write/read dependency analysis
Real-time project tracking via AGENT_STATE.json and PROJECT_PULSE.md
Fully automated multi-agent workflow loop from task dispatch to review
25 GitHub stars
Automated code review and consolidation cycles with fix-loop capabilities

사용 사례

Implementing large-scale features that span both frontend components and backend logic
Automating the end-to-end development and review cycle for high-velocity software teams
Managing complex refactoring projects requiring coordinated changes across multiple files

소개

주요 기능

Specialized agent assignment for backend (Codex) and frontend (Gemini) tasks
Intelligent parallel execution based on file write/read dependency analysis
Real-time project tracking via AGENT_STATE.json and PROJECT_PULSE.md
Fully automated multi-agent workflow loop from task dispatch to review
25 GitHub stars
Automated code review and consolidation cycles with fix-loop capabilities

사용 사례

Implementing large-scale features that span both frontend components and backend logic
Automating the end-to-end development and review cycle for high-velocity software teams
Managing complex refactoring projects requiring coordinated changes across multiple files