How does this skill prevent merge conflicts during parallel execution?

It utilizes Git Worktree isolation to create separate physical directories for each issue, ensuring agents work on independent branches without overlapping or contaminating the main workspace.

Does this skill support quality checks?

Yes, it includes a ReviewAgent (Medaman) that scores implementation quality; only code meeting the specified threshold can proceed to the PR and deployment stages.

What types of agents can I run with this skill?

You can execute the full Miyabi agent suite, including the Coordinator (task decomposition), CodeGen (implementation), Review (quality assurance), and Issue (analysis) agents.

Is the Miyabi framework required to use this skill?

Yes, this skill is designed to interface with the Miyabi autonomous development framework, which is built in Rust and managed via Cargo.

Miyabi Autonomous Agent Execution

Name: Miyabi Autonomous Agent Execution
Author: ShunsukeHayashi

byShunsukeHayashi

•

デプロイメントとDevOps

Executes Miyabi coding agents with Git Worktree isolation to enable safe, parallel autonomous development workflows.

概要

This skill facilitates the execution of Miyabi's suite of specialized coding agents—including Coordinator, CodeGen, and Reviewer—by leveraging Git Worktrees for task isolation. It allows developers to process multiple GitHub issues simultaneously without merge conflicts or local workspace contamination. By orchestrating a structured lifecycle from issue analysis to pull request creation, it ensures that autonomous code generation is always followed by quality-gated reviews and standardized commit practices, making it ideal for scaling AI-driven development projects.

主な機能

Automated context generation for LLM-driven development tasks
7 GitHub stars
Orchestration of specialized agents including Coordinator, CodeGen, and Review
Full lifecycle management from issue analysis to PR creation and cleanup
Git Worktree isolation for conflict-free parallel task processing
Quality-gated review cycles with automated scoring and pass/fail logic

ユースケース

Processing multiple complex GitHub issues in parallel using isolated worktrees
Automating the end-to-end development cycle from implementation to code review
Scaling autonomous development workflows without cluttering the main working directory

概要

主な機能

Automated context generation for LLM-driven development tasks
7 GitHub stars
Orchestration of specialized agents including Coordinator, CodeGen, and Review
Full lifecycle management from issue analysis to PR creation and cleanup
Git Worktree isolation for conflict-free parallel task processing
Quality-gated review cycles with automated scoring and pass/fail logic

ユースケース

Processing multiple complex GitHub issues in parallel using isolated worktrees
Automating the end-to-end development cycle from implementation to code review
Scaling autonomous development workflows without cluttering the main working directory