Subagent Driven Development

The subagent-driven-development skill orchestrates multiple specialized AI agents to collaborate on complex software engineering tasks, breaking down monolithic problems into manageable sub-problems. It automates the workflow of planning, coding, testing, and refining codebases by delegating specific responsibilities to distinct agent instances.

When to use it

• You need to build a large-scale application where a single LLM context window cannot hold all requirements and code simultaneously.
• Your project requires parallel execution of tasks, such as generating unit tests while the main logic is being written.
• You want to maintain high code quality through automated peer review cycles where one agent critiques another's implementation.
• You are working on a legacy codebase and need agents to refactor specific modules without disrupting unrelated system components.

Key capabilities

• Decomposes complex development goals into independent sub-tasks for specialized agents.
• Manages inter-agent communication to ensure context is passed correctly between team members.
• Executes iterative cycles of generation, review, and refinement autonomously.
• Maintains separate contexts for different aspects of the project to prevent hallucination.

Example prompts

• "Create a full-stack e-commerce dashboard by assigning one agent to handle React frontend components and another to manage Node.js backend APIs."
• "Refactor this legacy Python module into an object-oriented design, using one agent to analyze dependencies and another to implement the changes."
• "Generate a comprehensive test suite for my new authentication service, delegating unit test creation to one agent and integration test setup to another."

Tips & gotchas

Ensure your initial prompt clearly defines the roles and goals for each sub-agent to prevent conflicting instructions. Monitor the handoff points between agents closely, as context loss during delegation can lead to inconsistent code styles or logic errors.