subtask

Subtask is a specialized command-line interface designed to orchestrate multiple AI subagents simultaneously. It functions by creating isolated git worktrees for each task, allowing workers to research, plan, implement, and test features or bug fixes in parallel without conflicting with one another. The skill empowers the lead agent to delegate complex, multi-step projects by breaking them down into granular tasks, managing their lifecycle, and reviewing results before merging them into the base branch. This is an essential tool for development teams and individual contributors who rely on Claude Code or similar AI agents to accelerate software delivery and maintain high-quality codebases.

• Parallel task execution using isolated git worktrees ensures that multiple concurrent AI subagents can operate safely in the same repository.

• Centralized task management and tracking, including detailed logs, conversation histories, and status updates for each individual workflow.

• Flexible workflow stages allow for structured project phases such as planning, implementation, review, and ready-to-merge states.

• Automated review and squash-merge capabilities enable clean commit history and efficient integration of finished tasks.

• Transparent visibility into subagent internal thinking and debug tracing for complex logic troubleshooting.

• Use this skill when managing large, complex, or multi-faceted coding projects that require delegation to subagents.

• Always clarify requirements with the user before initiating new tasks; ensure the lead agent owns the complexity and maintains the overarching vision.

• Tasks are managed via specific CLI commands like draft, send, stage, and merge, which should be used to facilitate constant communication between the human user and the worker agents.

• Ensure all work is reviewed within the TUI or via workspace inspection before requesting final user sign-off for merging.

• The tool is designed to work in conjunction with Claude Code, leveraging git best practices to ensure repository stability even during heavy parallel development.